MEC - 109 Research Methods in Economics PDF

MEC-109
Research Methods
Indira Gandhi
in Economics
National Open University
School of Social Sciences
Block
1
RESEARCH METHODOLOGY: ISSUES AND
PERSPECTIVES
UNIT 1
Research Methodology: Conceptual Foundation 7
UNIT 2
Approaches to Scientific Knowledge: Positivism and Post
Positivism 21
UNIT 3
Models of Scientific Explanation 43
UNIT 4
Debates on Models of Explanation in Economics 57
UNIT 5
Foundations of Qualitative Research: Interpretativism and
Critical Theory Paradigm 72
Expert Committee
Prof. D.N. Reddy Prof. Romar Korea
Rtd. Professor of Economics Professor of Economics
University of Hyderabad, Hyderabad University of Mumbai
Mumbai
Prof. Harishankar Asthana
Professor of Psychology Dr. Manish Gupta
Banaras Hindu University, Varanasi Sr. Economist
National Institute of Public
Prof. Chandan Mukherjee
Finance and Policy
Professor and Dean
New Delhi
School of Development Studies
Ambedkar University, New Delhi Prof. Anjila Gupta
Professor of Economics
Prof. V.R. Panchmukhi
IGNOU, New Delhi
Rtd. Professor of Economics
Bombay University and Former Prof. Narayan Prasad (Convenor)
Chairman ICSSR, New Delhi Professor of Economics
IGNOU, New Delhi
Prof. Achal Kumar Gaur
Professor of Economics Prof. K. Barik
Faculty of Social Sciences Professor of Economics
Banaras Hindu University, Varanasi IGNOU, New Delhi
Prof. P.K. Chaubey Dr. B.S. Prakash
Professor, Indian Institute of Public Associate Professor in Economics
Administration, New Delhi IGNOU, New Delhi
Shri S.S. Suryanarayana Shri Saugato Sen
Rtd. Joint Advisor Associate Professor in Economics
Planning Commission, New Delhi IGNOU, New Delhi
Course Coordinator: Prof. Narayan Prasad
Block Preparation Team
Units Resource Person IGNOU Faculty
(Format, Language and Conent editing)
1 Prof. Narayan Prasad
Professor of Economics Prof. Narayan Prasad
Block Editor
IGNOU, New Delhi Professor of Economics
(Content)
IGNOU, New Delhi
2 Prof. S.G. Kulkarni
Prof. Prem Vashistha
Professor of Philosophy Prof. Narayan Prasad
NCAER, New Delhi
University of Hyderabad, Hyderabad Professor of Economics
IGNOU, New Delhi
3,4 Prof. D. Narshimha Reddy
Rtd. Professor of Economics Prof. Narayan Prasad
University of Hyderabad, Hyderabad Professor of Economics
IGNOU, New Delhi
5 Dr. Nausheen Nizami
Assistant Professor in Economics Prof. Narayan Prasad
Gargi College, University of Delhi Professor of Economics
New Delhi IGNOU, New Delhi
Print Production
Mr. Manjit Singh
Section Officer (Pub.)
SOSS, IGNOU, New Delhi
October, 2015
 Indira Gandhi National Open University, 2015
ISBN-978-81-266-
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the Indira Gandhi National Open University.
Further information on Indira Gandhi National Open University courses may be obtained from the
University's office at Maidan Garhi. New Delhi-110 068.
Printed and published on behalf of the Indira Gandhi National Open University, New Delhi by the
Director, School of Social Sciences.
Laser Typeset by : Tessa Media & Computers, C-206, A.F.E.-II, Okhla, New Delhi
Printed at :
MEC-109 RESEARCH METHODS IN
ECONOMICS
After completion of Master’s degree in Economics, many of you may intend to

start your career as professional economist. As professional economists, you
would be required to carry out the task of analyzing many specific economic
situations and indicate their impact on economic policy framework. Some of
you may also pursue research degree programmes. In order to perform these
tasks, you need to be quipped with the various constituents of Research Methods
and the different techniques applied in data collection/analysis. The present course
aims to cater to this need. The theoretical perspectives that guides research, tools
and techniques of data collection and methods of data analysis together constitute
the research methodology. Quantitative Research and Qualitative Research are
associated with different paradigms. Accordingly approaches to social enquiry
varies. Mixed methods by combining of at least one quantitative method and
one qualitative method in measurement, collection or analysis of data is
increasingly being used to strengthen the validity of research findings. The present
course deals with all these aspects. The course comprises of 6 Blocks.
Block 1: This block deals with the theoretical perspectives/foundations guiding

both quantitative and qualitative research. The essence of three paradigms i.e.
positivism and post positivism. Interpretitivism and critical theory associated
with different research strategies and methodologies have been discussed. This
block covers the entire breadth of main trends of development in the philosophy
of science and debates in the methodology of economics. The block has been
divided into 5 units. Unit 1 deals with the conceptual foundation related to
Research Methodology and its constituents, approaches to social enquiry, research
strategy, research process, an elementary idea of hypothesis and measurement
scales of variables. Unit 2 is devoted to scientific methodology covering
Positivists’ verification approach, Pauper’s critical rationalism, and Thomas
Kuhan’s paradigm’s shift approach. The fundamental differences between
Popper’s and Positivists’ views at the one hand, and differences between Popper’s
view and Kuhan’s views on the other, have also been explained. Model of
scientific explanation broadly within the framework of positivist theory of
scientific method constitutes the core of Unit 3. Basic rules of formal reasoning,
models of explanation, role of logical reasoning in the formulation of research
problem, and the problems involved in systematic explanation of phenomenon
have been discussed in this unit. Unit 4 presents the view of different economists
about the models of explanation in economics in the positivists mainstream
framework and a brief discussion on the alternative methods of explanation.
Unit 5 throws light on essentials of interpretativism and critical theory paradigm
guiding the frameworks of qualitative research. Based upon different ontological
and epistemological positions, these paradigms analyse the nature of reality
differently.
Block 2 on research design and measurement issues sets the tone and context to
provide balanced treatment to quantitative and qualitative research. The block
comprises of three units. Research design and mixed methods, the characteristics
of quantitative methods and qualitative methods, sampling design, the various
issues relating to measurement of variables have been covered in this block.
Research Methodology: Block 3: In order to validate the economic theory, we need their empirical
Issues and Perspectives
verification. Further, in order to examine disparity in income, the inequality
measures are important. Similarly in order to describe the development status of
an area in terms of several dimensions, we need to develop skill to construct
composite index numbers. For all these purposes, good exposure to methods of
regression models, in-equality measures and composite index is required.
Block 3 meets this requirement.
Block 4: In order to analyse the quantitative and qualitative data, more analytical
techniques like Multi Variate Analysis: Factor Analysis, Canonical Correlation
Analysis, Cluster Analysis, Correspondence Analysis, and Structural Equation
Models are being increasingly used under either mono method or mix methods.
Block 4 explains all these techniques.
Block 5: A large range of data analysis techniques are used in carrying out
qualitative research. However, three important methods namely participatory
methods, content analysis, and action research – relevant for conducting the
qualitative research studies in the area of Economics have been covered in this
Block.
Block 6: The availability of data is crucial for undertaking research. For this,
one is expected to be familiar with the different databases, and sources, the
concepts used in data compilation and the agencies involved in data collection.
Block 6 focuses on these aspects of database of Indian Economy.
4
Research Methodology:
UNIT 1 RESEARCH METHODOLOGY: Conceptual Foundation
CONCEPTUAL FOUNDATION
Structure
1.0 Objectives
1.1 Introduction
1.2 Research Methodology and its Constituents
1.3 Theoretical Perspectives
1.4 Approaches to Social Enquiry
1.5 Research Strategies
1.6 Research Process
1.7 Hypothesis: Its Types and Sources
1.8 The Nature, Sources and Types of Data
1.9 Measurement Scales of Variables
1.10 Let Us Sum Up
1.11 Key Words
1.12 Some Useful Books
1.13 Answer or Hints to Check Your Progress Exercises
1.0 OBJECTIVES
After going through this unit, you will be able to:
• Explain the research methodology and its constituents;
• State the various philosophical perspectives that guide research in social sciences;
• Describe various research approaches and strategies that are applied to answer
the research questions;
• Describe the various steps involved in the research process; and
• Discuss the various types of research designs appropriate for different types
of research.
1.1 INTRODUCTION
Research plays a significant role in human progress. It inculcates scientific and
inductive thinking and promotes the development of logical thinking and
organization. The role of research in several fields of applied economics has
increased in modern times. As an aid to economic policy, it has gained importance
for policy makers in understanding the complex nature of the functioning of the
economy. It also occupies special significance in analysing problems of business
and industry. Social scientists study the relationships among many variables
seeking answers to complex issues through research. In short, research aids the
process of knowledge formation and serves as an important source of providing
policy suggestions to different business, government and social organization.
The knowledge of the critical perspectives or philosophy of science, techniques
5
Research Methodology: of data collection and tools or methods of analyzing data are essential for
undertaking research in a systematic manner. In this unit, we shall focus our
attention on important constituents of research methodology i.e., research
perspectives, research approaches and strategies and data types and its sources,
hypothesis formulation and measurement scale of variables. At the initial stage
of research, several questions may arise in your mind – what do we mean by
research? How is the term ‘research methodology’ distinct from ‘research
techniques’ or ‘research methods’? Which type of research approach can be
applied to a particular situation or context? What are the steps involved in the
research process and how to design a research project? We shall take up these
issues in this unit. Let us begin by explaining the term research methodology
and its constituents.
1.2 RESEARCH METHODOLOGY AND ITS

CONSTITUENTS
Research in common parlance refers to a search for knowledge. It can be defined
as a scientific and systematic enquiry either to discover new facts or to verify old
facts, their sequences, interrelationships, causal explanation and the adherence
to natural laws governing them. It thus aims to discover the truth by applying
scientific methods.
Research Methodology is a wider term. It consists of three important elements:
i) theoretical perspectives or orientation to guide research and logic of enquiry,
ii) tools and techniques of data collection, and
iii) methods of data analysis.
Research Methods, comprises of research techniques and tools. Research

techniques refer to the practical aspects of collecting data and the way the
information/data obtained/collected is organized and analysed. Tools are the
instruments that are used for data collection and its analysis. It includes
questionnaire/schedules, dairies, check lists, maps, photos, drawings etc. Census
and survey methods are mainly used to collect quantitative data. In qualitative
research, data is generated/complied by way of participant observation, semi
structured interviews, life histories, experiments, pilot studies, scenarios etc. Data
analysis involves a set of statistical techniques used in establishing relationships
between the different variables and in evaluating the accuracy of the results.
Thus, methodology, methods and tools/techniques are three distinct elements of

the research process. Any one of these three elements by itself may not be adequate
in many situations. For instance, no data can be systematically collected without
adequate knowledge of techniques of data collection. Similarly, data can not be
explained without comprehending the philosophy or perspective behind the
characteristics underlying the variables to which the data relates. A sound
knowledge of statistical techniques is also necessary to analyse the data efficiently.
1.3 THEORETICAL PERSPECTIVES

Theoretical perspectives relate to theories of knowledge which lies within the
domain of philosophy of social science. The key concept associated with the
perspectives is the paradigm. Let us start to discuss with the concept of paradigm.
6
Paradigm: A paradigm is a comprehensive belief system, world view or Research Methodology:
Conceptual Foundation
framework that guides research and practice. It consists of
• At the basic or fundamental level, a philosophy of science that makes a
number of assumptions about fundamental issues relating to nature and
characteristics of truth or reality (ontology) and the theory of knowledge
dealing with how can we know the things that exist (epistemology).
• World view, conceptual and theoretical framework that guides research and
practices used in the field.
• General methodological prescriptions including tools to be used for
information/data collection and data analysis to conduct the work within
the paradigm.
The exact number of world views and the names associated with a particular
paradigm vary from author to another but four paradigms in the context of research
approaches in social sciences, are important:
• Positivism and post positivism
• Critical theory
• Interpretivism
• Pragmatism.
The peculiar features of these paradigms are:
• They differ on the question of reality,
• They offer different reasons or purposes for doing research,
• They use different types of tools for data collection and sources of
information also vary,
• They resort different ways of deriving meaning from the collected data,
• They vary in the relationship between research and practice.
Some authors have classified these paradigms into three categories by re-naming
them as:
Realism Constructivism Pragmatism
• Realism begins by assuming that there is a real world that is external to the
experience of any particular person and the goal of research is to understand
that world.
• Constructivism begins by assuming that everyone has unique experience

and beliefs and it posits that no reality exists outside of those perceptions.
• Pragmatism considers realism and constructivism as two alternate ways to

understand the world. However, the questions about the nature of reality are
less important than questions about what is meant to act and experience the
consequence of those actions.
7
Research Methodology: The knowledge of all these perspectives enable a researcher to make a meaningful
choice about
1) the research problem;
2) the research questions to investigate this problem;
3) the research strategies to answer these questions;
4) the approaches to social enquiry that accompany these strategies;
5) the concept and theory that direct the investigation;
6) the sources, forms and types of data;
7) the methods for collecting and analyzing the data to answer these questions.
The characteristics of positivism and post positivism as theoretical and

methodological perspectives to scientific knowledge have been provided in the
next unit i.e. Unit 2 of this course. Similarly the characteristics of interpretivisms
and critical theory perspective to conduct social research will be taken up in
Unit 5 of this course.
1.4 APPROACHES TO SOCIAL ENQUIRY

Broadly two types of approaches are used in conducting research in social
sciences: quantitative and qualitative. The studies conducted within the
perspective (framework) of positivism/post-positivism/realism generally resort
the quantitative approach and are termed as ‘quantitative research’. Quantitative
research integrates purposes and procedures that are deductive, objective and
generalized. Emphasis is laid on the construction of general theories which are
applied universally. Well controlled procedures with large number of cases are
followed in conducting the studies.
On other hand, the studies conducted within the perspective of critical theory
and interpretivism paradigms are termed as qualitative research. By using
induction as a research strategy, qualitative research creates the theory and
discovery through flexible, emergent research designs. It tries to evolve meaning
and interpretation based on closer contacts between researchers and the people
they study. Thus qualitative research consists of purposes and procedures that
integrate inductive, subjective and contextual approaches. Based on the above
outlines on the types of research – we may say that there are two basic approaches
to research viz., quantitative approach and qualitative approach.
Mixed methods research by combining quantitative methods and qualitative

methods are being used in social sciences. Hence, mixed method design by
integrating quantitative and qualitative approach has also emerged as an approach
to social enquiry.
Quantitative approach can be further sub-classified into: inferential approach;

experimental approach; and simulation approach.
In Inferential approach, database is established through survey method and

inference is drawn about characteristics or relationship of variables. In
experimental approach, greater control is exercised over research environment.
Some variables are manipulated to observe their effect on other variables.
8
Simulation approach refers to the operation of a numerical model that represents Research Methodology:
the structure of a dynamic process. It involves the construction of an artificial
environment in which relevant information/data can be generated. Given the
values of initial conditions, parameters and exogenous variables, a simulation is
run to represent the behaviour of the process over time.
Qualitative approach deals with the subjective assessment of attitudes, opinions

and behaviour of respondents in the field. Results are generated either in non-
quantitative form or in a form which are subjected to relatively less rigorous
quantitative treatment. Various techniques like group discussions, projective
techniques, in-depth interviews etc., are used. The typical characteristics of both
these approaches may be summarized in the following Table:
Table 1.1: Typical Characteristics of Qualitative and Quantitative Approach

Characteristics Qualitative Approach Quantitative Approach
1) Typical data Participant observation, semi- Laboratory observations,
collection structured interviews, group questionnaire, schedule or
methods discussion report cards etc. structured interviews.
2) Formulation of Open/loosely specified questions Closed questions (hypothesis)
questions and and possible answers. Questions and answer categories to be
answers and answers are exchanged in prepared in advance.
two way communication between
researcher and respondent.
3) Selection of Information maximization guides Representativeness as
respondents the selection of respondent. proportion of population N.
Every respondent may be unique Randam sample selection,
(key person). sample size according to
assumptions about distribution
in population N.
4) Timing of Parallel with data collection After data collection
analysis
5) Application of Descriptive methods of analysis Standard statistical methods
standard are used. Mixed methods are also are frequently used.
methods of used.
analysis
6) The role of Existing theories are typically A-priori deducted theories are
theories in the used only as point of departure opeationalised and tested on
analysis for the analysis. Theories are data. The process of analysis
further developed by forming is basically deductive.
new concepts and relations. The
contents of the new concepts are
studied and illustrated. Practical
application of theory is
illustrated by cases.
1.5 RESEARCH STRATEGIES

Four basic strategies can be adopted in social research depending upon the
researcher’s belief/reliance on perspective/paradigm about the nature of reality.
9
Issues and Perspectives Paradigm Research Strategy
Positivism Induction
Post positivism/Realism Deduction
Critical realism Retroduction
Interpretativism Abduction
Each strategy has different starting points (Lewis-Beck, 2004):

1) The inductive strategy begins with collection of data from which
generalisation is made and can be used as an elementary explanation;
2) The deductive strategy starts out with a theory that provides a possible answer.
The theory is tested in the context of a research problem by collection of
relevant data;
3) The retroductive strategy starts out with a hypothetical model of a mechanism
that could explain the occurance of a phenomenon under investigation;
4) The abductive strategy starts with laying the concepts and meanings that
are contained in social quarters account of activities related to a research
problem.
Check Your Progress 1

1) Distinguish between Research Methodology and Research Methods.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) How does the knowledge of research perspectives help a researcher to
undertake research studies in social sciences?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) How is retroductive research strategy different from abductive strategy?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
10
4) Explain the term ‘ontology’ and ‘epistomology’. Research Methodology:
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
1.6 RESEARCH PROCESS

Research process refers to different steps involved in a desired sequence in
carrying out research. However, this does not mean that these steps are always in
a given sequence. The various steps involved in research can be illustrated in a
flow chart as shown in the figure below:
Defining Research Problem
• Objectives
• Research Questions
Review of Literature
Formulation of Hypothesis
Research Design
Collection of Data: Source, tools, and methods
Analysis of Data
Report Writing
Fig. 1.1: Flowchart of Research Process
The above activities or steps overlap continuously rather than following the
prescribed sequence strictly. The steps are not mutually exclusive. The order
illustrated is meant only to provide a procedural guideline for research. The steps
are briefly elaborated below:
a) Defining the research problem: Selecting and properly defining the

research problem is the first foremost step. The problem to be investigated
must be defined categorically. It is important to identify the general area of
interest or a particular aspect of a subject matter desired to be studied. Initially,
the problem may be stated in a borad way and later it can be narrowed down
in operational terms. Essentially two steps are involved in formulating the
11
Research Methodology: research problem: (i) Understanding the problem thoroughly; and (ii)
rephrasing it into meaningful terms from operational/analytical point of
view.
It is better to select the subject that is familiar with easy access to research
material and data sources. Apart from the topic, following points need to be
stated clearly in the research problem:
1) rationale behind the research problem;
2) the aims and objectives as per the requirements of the research questions.
The statement of the objectives determines the data to be collected,
hypothesis to be tested, techniques for data collection and analysis to
be adopted, and the relations intended to be explored;
3) the research questions in the light of the objectives and the theoretical
arguments/foundation on which it rests;
4) developing the ideas through discussions; and
5) re-phrasing the research problems identified in (i) above into a working

proposition.
The different steps to be followed while defining the research problem,

therefore are:
• statement of the problem first in a general way to be later sharpened

with the help of literature review,
• understanding the nature of the problem, and
• surveying the available literature.
In addition to the above, the following points should also be observed while
defining the research problem:
• Technical terms and words or phrases used in the research problem

should be explicitly defined.
• Basic assumptions or postulates relating to the research problem need

to be clearly stated.
• A clear and unambiguous statement of the investigation should be

provided.
• The time-period required and the scope of the study must be duly
stated.
• The sources of data and its limitations must be explicitly mentioned.

b) Review of Literature: The review of literature is meant to gain insight on
the topic and gain knowledge on the availability of data and other materials
on the theme of proposed area of research. The literature reviewed may be
classified into two types viz. (i) literature relating to the concepts and theory
12
and (ii) empirical literature consisting of findings in quantitative terms by Research Methodology:
studies conducted in the area. This will help in framing research questions
to be investigated. Academic journals, conference proceedings, government
reports, books etc. are the main sources of literature. With the spread of IT,
one can access a large volume of literature through internet.
c) Formulation of Hypothesis: Specification of working hypothesis (or

hypotheses) is the next step of research process. A hypothesis is a tentative
statement made which needs to be tested for its logical and empirical
confirmation. Hypothesis can be formulated as a proposition or set of
propositions providing most probable explanation for occurrence of some
event or specified phenomenon. Hypotheses when empirically tested may
either be accepted or rejected. A hypothesis must, therefore, be capable of
being tested. A hypothesis stated in terms of a relationship between the
dependent and independent variables are suitable for econometric treatment.
The manner in which hypothesis is formulated is important as it provides
the required focus for research. It also helps in identifying the method of
analysis to be used.
Prior thinking about the subject, examination of the available data and
material related to the study, discussion with colleagues and experts help
the researcher in formulation of hypothesis. Exploratory or descriptive
research can be carried out even without hypothesis.
d) Research Design: Research design is the logical conceptual structure within

which research is conducted. It is the blueprint for the collection,
measurement and analysis of data. Detailed discussion on Research Design
has been provided in Unit 6 of Block 2.
e) Collection of Data: Collection of data is an essential part of the research

process. Data can be primary or secondary. Data collected by the researcher,
say by a survey, is primary. The data already collected by some agency and
available in some published form is secondary. There are two main techniques
of data collection – (i) census and survey (ii) observation method. Primary
data can also be collected by means of experiments (e.g., yield under certain
conditions, observation at many time points of a certain phenomena, etc.).
Intensive fieldwork methods include observation, interview, case study, etc.
A survey is usually conducted by the canvassing of a questionnaire. Issues
of data collection and sampling design have been discussed in detail in
Unit 7 of Block 2.
f) Analysis of Data: Analysis involves steps like categorization, coding,

tabulation, etc. The principle for classification or categorization of data has
to be based on the problem under study or the hypothesis formulated. The
category must be exhaustive and sufficient for classifying all responses.
They must be distinct, separate and mutually exclusive. Coding involves
grouping of responses falling under a particular category. Tabulation is a
means of organizing the responses to facilitate comparisons bringing up the
inherent relations between two or more variables. It is an orderly arrangement
of data in columns and rows. Analysis and inference is usually aided by the
application of different statistical and econometric techniques. Some of the
major techniques usually employed in research have been reflected in the
summary chart 1.2 13
Issues and Perspectives Summary Chart 1.2
ANALYSIS OF DATA
Processing of Data Analysis of Data

(Preparing data for analysis)
Editing Descriptive and Inferential analysis/

Classification
Using Causal Analyses Statistical analysis
Percentages
Coding Tabulation
Estimation of Testing of
parameter values hypotheses
Bivariate analysis Multi – variate analysis

Uni-
(Analysis of two (simultaneous analysis of
dimensional
variables or more than two variables
analysis Parametric test
attributes in a two – / attributes in a multiway Non-
way classification) classification parametric
tests or
Distribution
free tests
Point Interval
estimate estimate
(Calculation of several measures
mostly concerning one variable) Simple regression and
simple correlation (in
i) Measures of Central
respect of variables) Multiple regression*
Tendency;
and multiple
ii) Measures of dispersion; correlation/partial
iii) Measures of skewness; Association of correlation in respect
iv) One-way ANOVA, Index attributes of variables
numbers, Time series ( t h r o u g h
analysis; and coefficient of Multiple discriminant
v) Others (including simple association and analysis (in respect of
correlation and regression in coefficient of attributes)
simple classification of paired contingency)
data)
Multi-ANOVA (in
Two-way ANOVA respect of variables)
Canonical analysis
(in respect of both
variables and
attributes)
Other types of
analyses (such as
factor analysis,
cluster analysis)
Most of these analytical techniques have been covered in Units 9,10,11 and 12
of Block 3 and Units 13,14,15,16 and 17 of Block 4 of this course.
14
Report Writing: Originality and clarity are the two vital components of research Research Methodology:
report. It is the ultimate test of one’s analytical ability and communication skills.
It is an exercise involving the organization of ideas. The research report needs to
be presented in such a manner that the readers can grasp the context, methodology
and findings easily. The report comprise of two parts: the preliminary pages and
the main text. In the preliminary pages, the report should indicate the title of the
research study, name of the researcher (and his team members) and the name of
the institution and/or the month/year of preparation of the report. This should be
followed by a ‘preface’ in which the main context of preparing the report along
with key findings must be presented. Towards the end of the ‘preface’, the
important sources/persons can suitably be acknowledged.
The main text begins with an introductory chapter followed by the major aspects
of the study organized into different chapters. The introductory chapter should
contain a clear statement of the objectives of the study, rationale behind the
study, a brief summary of the literature review, hypotheses tested (if any) and the
definitions of the major concepts employed in the study. The methodology adopted
in conducting the study must also be fully explained along with an explicit mention
of the limitations of the study. The subsequent parts of the main text, should
present the major aspects of the study arranged in a logical sequence splited into
appropriate sections and subsections. The inter-connection between different
sections should be properly maintained so that the report reads smoothly.
The implications of the results of the study should be stated towards the end of
the report. The implications may comprise of:
i) the inferences drawn from the study;
ii) the conditions which may limit the extent of generalizations of the inferences
drawn; and
iii) the questions that remain unanswered along with new areas for further
research identified. The conclusion drawn from the study should be clearly
related to the objectives/hypotheses stated in the introductory section.
The report may also include an ‘executive summary’ outlining the context and
methodology, and major findings of the study. The ‘executive summary’ is placed
right at the beginning (i.e. before the introductory chapter) so as to provide a
concise picture of the entire report.
1.7 HYPOTHESIS: ITS TYPES AND SOURCES

Hypothesis are potential explanations that can account for our observations of
the external world. They usually describe cause and effect relationships between
a proposed mechanism or process (the cause) and our observations (the effect).
In quantitative research, there are two methods of hypothesis generation: deductive
method and inductive method.
If hypothesis is generated from a theory, it is called deductive approach. On the
other hand, if it is generated from observation, it is termed as inductive approach.
Deductive approach: Theory → Hypothesis→Observation→Confirmation.
Inductive approach: Observation→Pattern→Tentative Hypothesis→Theory. 15
Research Methodology: In qualitative research, a hypothesis might be framed in terms of social setting
having certain features, which through observation, can be confirmed or falsified.
In survey or experimental research or otherwise hypothesis testing establishes
the statistical significance of a finding.
1.8 THE NATURE, SOURCES AND TYPES OF DATA

For undertaking any meaningful research in terms of situational assessment,
testing of models, development of theory, evaluation of economic policy, data is
essential. The availability of data therefore, determines the scope of analysis. In
any research, the researcher is expected to state the sources of the data used in
the analysis, their definitions, and methods of collection.
The data may be of three types; Time series, Cross-section and Pooled.
1) Time Series Data: It is a set of observations on the values that a variable

takes at different times. Such data may be collected at regular time intervals
such as daily (i.e. prices, whether reports etc.), weekly (like money supply
figures), monthly (i.e. consumer price index etc.) quarterly (i.e. GDP),
annually (i.e. government budget etc.).
2) Cross Section Data: Cross-section data are data on one or more variables
collected at the same point of time. For example the data on the census of
population collected by the Registrar General of India.
3) Pooled Data: In pooled data, the elements of both time series and cross
section are clubbed. For example, over a period of time say from 2000 to
2013, we have data on saving, investment and GDP across Indian states.
Panel, Longitudinal or Micro Panel Data: This is a special type of pooled data
in which the same cross-sectional (say a family or firm) is surveyed overtime.
The Sources of Data: The data used in empirical analysis may be collected by a
governmental agency (e.g. CSO, NSSO, RBI, Labour Bureau etc.), an
international agency (e.g. International Monetary Fund (IMF) or a private
organization. Such data is called secondary data because these are collected from
secondary sources. The details about the kind of secondary data compiled by the
different data collecting agencies have been presented in Block 6 of this course.
The data collected by the investigator or researcher through field work is termed
primary data. Such data is collected by using different tools like questionnaire,
schedule, interview etc. under quantitative approach and participant observation,
open ended interview, group discussion, key information etc. ( under qualitative
approach). The methods for collecting the primary data have been discussed in
Unit 7 of Block 2.
1.9 MEASUREMENT SCALES OF VARIABLES

Measurement is the process of observing and recording the observation that are
collected. A variable can be measured at four levels: ratio scale, interval scale,
ordinal scale and nominal scale. The suitability of analytic technique depends on
the measurement scale. A particular econometric/statistical technique that may
16 be suitable for ratio scale variables may not be suitable for nominal scale variables.
It is therefore desirable to know the distinctions among the four types of Research Methodology:
measurement scales. These are discussed below:
1) Nominal Scale: Under the nominal scale, the data is recorded into categories,
without any order or structure. In other words, if against any question/
statement, response is recorded simply yes/no, then the scale will be nominal.
It has no order and there is no distance between yes and no.
The statistical techniques that can be used with nominal scales data are:
Mode, cross tabulation – with chi-square. Other highly sophisticated
modeling techniques available for nominal scale data are – logistic Linear
Regression Model, Principal component analysis, factor analysis etc.
2) Ordinal Scale: In terms of power of measurement, ordinal scale comes
next to nominal scale. Recording Ranks against various options/choices is
the simplest ordinal scale. If you are asked by a researcher to rank 5 fruits
from most favorable to least favorable, he is essentially asking you to create
an ordinal scale of preference.
Median, Mode, rank order correlation, non-parametric correlation and
modeling techniques are used with ordinal data.
3) Interval Scale: In a situation when we not only talk about differences in
order but also differences in the degree of order, it is referred to as interval
scale. For example, if we are asked to rate our satisfaction with a piece of
software on a 7 point scale, from dissatisfied to satisfied, we are using an
interval scale.
Mean and stardard deviation, correlation, regression, analysis of variance,
factor analysis techniques can be used with interval scale data.
4) Ratio Scale: A ratio scale is the top level of measurement and satisfies the
following properties:
i) Measurement of each observation of a variable in numerals (quantitative
terms) and hence possible to work out the ratio of two observations.
For a variable X taking two values X1 and X2 the ratio will be X1 / X2.
ii) Measurement of distance between two observation X1 and X2 i.e. (X2 – X1).
ii) Indication of the natural ordering (ascending or descending) of the
elements of a variable. Therefore, comparison such as X2 > X1 or X2 > X1
are meaningful.
The statistical techniques used in interval scale can easily be used in ratio scale
also. You will find elaborate discussion on measurement scales and scaling
techniques in Unit 8 of Block 2.

1) In what way review of literature helps a researcher?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
17
Research Methodology: 2) What is meant by a hypothesis?
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) Distinguish between time series data and cross section data.
......................................................................................................................
......................................................................................................................
......................................................................................................................
4) State the various measurement scales of data.
......................................................................................................................
......................................................................................................................
......................................................................................................................
1.10 LET US SUM UP

Research plays a significant role in human progress. It inculcates scientific temper
and logical thinking. Research refers to a scientific and systematic enquiry aiming
either to discover new facts or to verify the old facts. The theoretical perspectives,
tools and techniques of data collection and methods of data analysis together
constitute the Research Methodology. Broadly there are two basic approaches of
research; quantitative and qualitative. Quantitative approach generates the data
in quantitative terms. This can, therefore, be subjected to quantitative analysis.
Subjective assessment of attitudes, opinions and behaviour are tackled by the
qualitative approach to research. Research process involves seven steps –
identification of research problem, review of literature, objectives, formulating
the hypothesis, finalizing research design, collection of data, analyzing the data
and report writing. Data can be of three types: time series, cross section and
panel data. A variable can be measured at four levels: ratio scale, interval scale,
ordinal scale and nominal scale. The suitability of analytic technique depends on
the measurement scale. Hypothesis is a tentative explanation of any event or
phenomenon in the external world. There are two methods of hypothesis
generation: deductive method and inductive method. Thus, this unit provides an
overview of the conceptual foundation of Research Methodology and enables
the learner to comprehend the various processes involved in carrying out the
research study.
1.11 KEY WORDS

Coding : A system of symbols, letter or words used in
transmitting messages.
Epistemology : Epistemology refers to the theory of knowledge of
how human beings come to have knowledge of the
world around them – of how we know what we know.
Broadly there are two theories: Rationalism and
Empiricism.
18
Rationalism : Rationalism is based on the idea that reliable Research Methodology:
knowledge is derived from the use of “pure” reason.
Empiricism : Empiricism envisages that the knowledge of the world
can be obtained only through direct sense-experience.
Experimental : Research in which independent variables are
Testing Research manipulated.
Ontology : It is a branch of philosophy that is concerned with the
nature of reality. It deals with the theories about what
makes up reality.
Induction : Induction is a process for moving from particular
statements to general statements. This logic is used
in social sciences to produce theory from data.
Deduction : Deduction is a process used to derive particular
statements from general statements. A hypothesis is
deduced from a theory and is tested by empirical data.
Abduction : Abduction refers to the process of moving from the
way social actors describe their way of life to
technical, social scientific description of that social
life. It has two stages: (a) describing these activities
and meaning and (b) deriving categories and concepts
that can form the basis of an understanding or an
explanation of the problem at hand.
Retroduction : Retroduction refers to the process of going back from,
below, or behind observed patterns or regularities to
discover what produces them. This logic of enquiry
focuses to locate the structures and mechanisms that
have produced the regularity. These structures and
mechanisms envisage the tendencies or powers of
things to act in a particular way.
Research Hypothesis : Research hypothesis is a predictive statement that
relates an independent variable to a dependent
variable. Such a hypothesised relationship is usually
meant to be tested by research method. Predictive
statements, which cannot be objectively verified, or
the relationships that are assumed which cannot be
tested do not qualify as research hypothesis.
Tabulation : Setting out the data/information in tabular form.
1.12 SOME USEFUL BOOKS

1) Kothari C.R. (1985): Research Methodology – Methods and Techniques,
Wiley Eastern Publication Chapter 1-3 pp 1 to 67.
2) LeWis-Beck Michael S; Bryman Aan, Tim Futing, Liao (Ed) (2004): The
Sage Encyclopedia of Social Sciences Research Method, Volumes 1,2 and
3, Sage Publications, New Delhi.
19
Research Methodology: 3) David L. Morgan (2014): Integrating Qualitative and Quantitative Methods,
Sage Publications, New Delhi
1.13 ANSWERS OR HINTS TO CHECK YOUR

PROGRESS EXERCISES
1) See section 1.2
2) See section 1.3
3) See section 1.5
4) See section 1.3

1) Review of literature enables the researcher to know about the availability of
relevant material and data. It also helps in framing the research questions.
2) See section 1.7.
3) A set of values of a variable at different times is time series data whereas a
set of values of one or more variables collected at a point of time is cross
section data.
4) Various measurement scales of a variable are: ratio scale, interval scale,
ordinal scale and nominal scale.
20
UNIT 2 APPROACHES TO SCIENTIFIC Conceptual Foundation
KNOWLEDGE: POSITIVISM AND

POST POSITIVISM
Structure
2.0 Objectives
2.1 Introduction
2.2 Positivist Philosophy of Science
2.2.1 Central Tenets of Positivist Philosophy
2.3 Attack on Positivist Philosophy of Science
2.4 Karl Popper’s Philosophy of Science
2.5 Criticism against Karl Popper’s Philosophy of Science
2.6 Thomas Kuhn’s Philosophy of Science
2.7 Popper Versus Kuhn
2.8 Let Us Sum Up
2.9 Keywords
2.0 OBJECTIVES
• Describe the aim of science as a cognitive enterprise;
• Explain the basic ideas of positivism;
• Appreciate the arguments or criticisms leveled against positivist philosophy
of science;
• Explain the central tenets of Popper’s philosophy of science;
• State the fundamental differences between Popper’s and positivists’ views
about theory of scientific method;
• Assess Karl Popper’s philosophy of science;
• Describe the essentials of Kuhn’s philosophy; and
• Discuss the basic differences between the Popperian and the Kuhnian models
of science.
2.1 INTRODUCTION
Science has not only shaped our mode of living in the world but also our ways of
thinking about the world. It is for this reason it has acquired a central place in the
intellectual life of our times. Because of its central place in the modern culture,
three disciplines have emerged which has made science itself their object of
inquiry. These three disciplines are: History of Science, Sociology of Science
and Philosophy of Science. Whereas science studies the world natural or social 21
Research Methodology: or psychological, these three disciplines study science itself. History of Science
and Sociology of Science are essentially twentieth century disciplines. No doubt,
Philosophy of Science has a great past built by the contributions of individual
philosophies in different centuries. However, as a widespread and coordinated
discipline, it is also an essentially twentieth century phenomenon.
History of Science is a study of the development of scientific ideas and institutions

in a chronological order. Sociology of Science is the study of the relation between
social factors and forces on the one hand and the growth of scientific ideas and
institutions, on the other. In short, it is an inquiry into the relation between science
and society. History of Science deals with science as a historical phenomenon
and sociology of science is concerned with science as a social institution.
Philosophy of Science is a study of Science as a cognitive enterprise, that is, as a
knowledge-seeking activity. The enquiry we call Philosophy of Science addresses
questions like “What is the aim of Science?”, “What is the method of Science?”
In what sense, if any, is Science objective, rational and progressive?”
It is obvious that answers to questions such as these are of importance in figuring

out how if at all, and in what sense, if any, social sciences including Economics,
can claim to be scientific.
2.2 POSITIVIST PHILOSOPHY OF SCIENCE

Positivism (which is also called “logical positivism”) is a movement in Philosophy
of the first half of the 20th century. Positivists debunked the whole of traditional
Philosophy by attacking Metaphysics which was the most important branch of
Philosophy. Metaphysics was so central to Philosophy that Philosophy was
virtually identified with Metaphysics. Metaphysics sought to answer most general
and basic questions like “What is Ultimate Reality?” “Does God exist? And is
so what is God’s relation with the world?”, “Is there a soul?”, “What is the
relation between the mind and the body?”, “Is man free or are human actions
determined?” etc., Positivists maintained that Metaphysics was a spurious
discipline because metaphysical statements (such as “Matter is Ultimate reality”
or “Ultimate Reality is Spiritual” etc.,) are meaningless since they are not
verifiable in experience. “A statement” they claimed “is meaningful if and only
if it is verifiable.” Apart from being anti-metaphysical, they were empiricists
i.e., according to them, experience is the source of knowledge. They called
themselves “Neo-Empiricists” to distinguish themselves from the Traditional
Empiricists of 17th and 18th centuries like Locke, Berkeley and Hume.
2.2.1 Central Tenets of Positivist Philosophy

In science we start with particular observations regarding what is the case. Using
the method of induction, we arrive at definitions which are statements about the
essential nature of things. We, then, using the method of deduction, on the basis
of definitions, arrive at demonstrations which show why things must be what
they are. Thus, the aim of science is two fold: Definition and Demonstration.
The method of science, broadly, is of two types: Induction and Deduction. All
science must proceed from “What is” given to us in particular observations to
“what must be” as shown by the demonstrations. The path of science is an arch
whose two end points are observation and demonstration.
22
In the whole span of three centuries – from the beginning of the 17th century till Approaches to Scientific
Knowledge: Positivism and
the end of the 19th century – two views stand out prominently as answer to the Post Positivism
question regarding the aim and method of science. The first view is called
Induction. The second view is called Hypothesisism, according to which the
method of science is the method of Hypothesis. Francis Bacon is the Father of
Inductivism and Rene Dercartes is the Father of Hypothesisism. The two views
provide two distinct models of scientific practice, which we may call Baconian
and Cartesian models of Scientific practice.
Positivists worked out a well-knit philosophy of science. Here are some of the
central tenets of the Positivist Philosophy of Science:
1) Science is qualitatively distinct from, superior to and ideal for all other
areas of human endeavor (Scientism).
2) The distinction, superiority and idealhood that Science enjoys is traceable
to its possession of method (Methodologism).
3) There is only one method common to all sciences, irrespective of their subject
matter (Methodological monism).
4) That method which is common to all sciences, natural or human, is the
method of Induction (Inductivism).
5) The hallmark of science consists in the fact that its statements are
systematically verifiable.
6) Scientific observations are or can be shown to be “pure” in the sense that
they are theory-free.
7) The theories are winnowed from facts or observations.
8) The relation between observation and theory is unilateral in the sense theories
are dependent on observations whereas observations are theory-independent.
9) To a given set of observation based statements, there corresponds uniquely
only one theory (just as from a given set of premises in an argument, only
one conclusion follows).
10) Our factual judgments are value-neutral and our value judgments have no
factual content (Fact-Value Dichotomy thesis); hence, science being the
foremost instance of factual inquiry, does not have value commitments.
11) That all scientific explanation must have the following pattern:
L1…………………………………….Ln
I2……………………………………….In
Therefore, E.
Where L1……………………………………..Ln is a set of laws, I2………………………………In is a

set of statements describing initial conditions and E is the statement
describing phenomenon to be explained. That is to say, to explain a
phenomenon scientifically is to deduce its description from a set of laws
(which are called “Covering Laws”) via a set of statements describing initial
conditions. In sum, all explanation worthy to be called ‘scientific’ must
contain laws and involve deduction (Hence, this is called Deductive-
Nomologism where ‘nomological’ means ‘concerning laws’). 23
Research Methodology: 12) The aim of science is either economical description of phenomena or precise
prediction of facts and not providing an account of observations in terms of
unobservables. Hence, scientific theories are not putative descriptions of
the unobservable world. The aim of science has nothing to do with alleged
reality of such a world (Anti-Realism).
13) Unlike other areas of activity, science is progressive in the sense that scientific
change is always change for better, whereas other areas exhibit just change:
the progress of science consists in the accumulation of observations on the
one hand, and, cumulative growth of theories,on the other hand. The latter
means that any new theory includes the old theory(plus something). Thus
the growth of science essentially exhibits continuity.
14) Science is objective in the sense that its theories are based on ‘pure’
observations or facts which as theory free. Interpretations may be subjective
but observations/facts are objective because they are free from interpretation/
theory.
15) Science is rational because the principle of Induction which is central to the
method of science is rationally defensible, in spite of Hume’s skepticism
regarding its rational defensibility.
Positivists tried to justify the principle of Induction by invoking the concept of

pure observation. According to them, theories are arrived at on the basis of the
Principle of Induction. If we can show that theories are very closely related to
pure observations, the Principle of Induction stands rationally justified. They
tried to work out a whole project to demonstrate rational justification of the
Principle of Induction on these lines.
2.3 ATTACK ON THE POSITIVIST PHILOSOPHY

OF SCIENCE
It must be noted that the concept of pure observation is necessary for positivist
philosophy of science for showing that science is objective and that science is
rational. Hence, the centrality of this concept to the positivist philosophy of
science. Therefore, the collapse of this concept to the positivist philosophy of
science though other theses of positivist philosophy of science listed above also
were demolished by its opponents. Let us briefly look at the arguments which
demolished the positivist thesis of pure observations i.e., thesis at Sl.No.6 in the
list above.
Firstly, observations presuppose some principle of selection. We need relevant
observations. In science it is the problem that decides what is a relevant
observation and thus provides the principle of relevance. Hence, there can not
be observations without a – prior problem.
As Popper says “Before we can collect data, our interest in data of a certain kind
must be aroused; the problem always comes first”. It may be objected that we
become aware of the problems because of observations and hence observations
come first and therefore positivists are right. But this objection does not hold.
Two persons might make same observations but one may come out with a problem
and the other may not. Therefore, mere observations would not generate problems
24 in science. Usually problems are generated when there is a clash between what
we observe and what we expect. Of the two persons making the same observations, Approaches to Scientific
one comes out with a problem because he sees a conflict between what he observes Post Positivism
and what he expects whereas the other observer may have not expectation which
conflicts with what he observes. The former believes in a theory which produces
certain expectations which conflict with his observations and hence he comes
out with a problem. In other words, a – prior belief in a theory is necessary for a
problem to be generated and a – prior awareness of the problem is necessary for
making relevant observations. Thus, theory precedes observations.
Secondly, in science observations are taken into account only if they are described
in a language that is currently used in a particular science. An observation,
however, genuine, is no observation unless it is expressed in a recognized idiom.
It is the theory which provides the idiom or language to be used to describe facts
or observations. It is relevant to quote the words of Pierre Duhem, a distinguished
physicist and philosopher:
“Enter a laboratory. The table crowded with an assortment of apparatus: an electric

cell, silk covered copper wire, small cups of mercury, spools, a mirror mounted
on an iron bar; the experimenter is inserting into small openings the metal ends
of ebony-headed pins: the iron bar oscillates and the mirror attached to it throws
a luminous band upon a celluloid scale: the forward backward motion of this
spot enables the physicist to observe the minute oscillations of the iron bar. But
ask him what he is doing. Will he answer “I am studying the oscillations of an
iron bar which carries a mirror?” No, he will say that he is measuring the electrical
resistance of the spools. If you are astonished, if you ask him what his words
mean, what relation they have with the phenomenon he has been observing and
which you have noted at the same time as he, he will answer that your question
requires a long explanation and that you should take a course in electricity”.
Thirdly, most of the observations in science are made with the help of instruments.
These instruments are constructed or designed in accordance with the
specifications provided by some theories. These theories, one may say, form the
software of these instruments. Belief in the reliability of these instruments implies
the acceptance of these theories which has gone into the making of these
instruments. This observations presuppose prior acceptance of theories.
Fourthly, observations in science need to be legitimized i.e., ratified by a theory.

An example makes the point clear. We all know that Galileo used some telescopic
observations to support the Helio-centric theory against the geo-centric theory
of Ptolemy and his followers. His opponents did not consider the telescopic
observations adequate. Why did they not? No doubt, they had belief in the
reliability of telescope; they had no problem in using telescope for terrestrial
purposes i.e., making observations of earthly objects. They opposed the extension
of telescopic observations to the celestial sphere i.e., regarding heavenly bodies.
Their argument was that the normal factors like background and neighbourhood
which help our normal perceptions are absent in the sky. Further, it is impossible
to directly verify whether telescopic observations of heavenly bodies are accurate.
They rightly demanded from Galileo a theory of light which would justify the
extension of the telescope from terrestrial to celestial sphere. Galileo had no
such theory. But he rightly believed that such a theory could be provided in
future so that telescopic observations would get justification. Thus, while the
opponents of Galileo insisted that the telescopic observations be justified by an
25
Research Methodology: optical theory at the same time as their acceptance, Galileo maintained that the
justification could be provided subsequent to their acceptance. It may be noted
that both sides accepted that the telescopic observations needed justification in
term of a theory of light.
All this does not mean that observations are theory-dependent whereas theories
are observation-independent. Observations and theory are interdependent though
it is not easy to clarify what the nature of this interdependence is. However,
positivists were wrong in claiming that observations are theory –independent.
To say that observations are theory dependent is to say that observation is not a
passive reception but an active participation of our cognitive faculties equipped
with prior knowledge which we call theory. After all, observations are not ‘given’
but ‘made’.

1) What is the aim of science?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) What do you mean by the term ‘inductivism’?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) What are the features of hypothesis?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
4) Match the following:
A B
i) The relation between observation i) Value judgments
and theory
ii) Factual inquiry under positivist ii) Precise prediction of facts
approach does not have
iii) The aim of Science is iii) Objective
iv) Science is iv) Unilateral
26
5) Give two counter arguments against the positivist thesis of pure observation. Approaches to Scientific
...................................................................................................................... Post Positivism
......................................................................................................................
......................................................................................................................
2.4 KARL POPPER’S PHILOSOPHY OF SCIENCE

Karl Popper was the first to react against the positivist philosophy of science. In
fact he started attacking it quite early. But his attack on positivists and his
alternative to the positivist philosophy of science came to be widely known at
the beginning of the second half of the 20th century. Popper’s theory of science,
particularly his theory of scientific method has won a lot of admirers among
scientists and philosophers. As we know, positivists tried to work out a
sophisticated version of Inductivism. Popper worked out a sophisticated
version of Hypothesisism. We shall briefly consider his views on the nature of
science.
According to Popper, the central task of philosophy is not to solve Hume’s

problem or problem of Induction as thought by Positivists. This is because (1)
the problem of Induction cannot be solved, and (2) it need not be solved because
the method of science is not the method of Induction. The central task of
philosophy of science, Popper maintains, is to solve what he calls the problem of
demarcation or Kant’s problem i.e., the problem of identifying the line of
demarcation between science and non-science. Popper maintains that what
distinguishes science from the rest of our knowledge is the systematic
falsifiability of scientific theories. This falsifiability is the line of demarcation
between science and non-science. Falsifiability is the criterion of scientificity. A
statement is scientific if and only it is falsifiable.
In accordance with what he considers to be the hallmark of scientific theories,

Popper puts forward what he considers to an adequate model of scientific method.
He characterizes his model of scientific method as Hypothetico-Deductive
model. According to him, the method of science is not method of Induction but
the method of Hypothetico-Deduction. What are the fundamental differences
between these methodological models”? Firstly, the inductivist model maintains
that our observations are theory-independent and therefore are indubitable.
That is to say, since observations are theory-independent, they have
probability Value 1. It also says that our theories are only winnowed from
observations and therefore our scientific theories have the initial probability
value 1 in principle. Of course, inductivists admitted that in actual practice, the
theories may contain something more than what observation based statements
say, with the result that our actual theories may not have been winnowed from
observations.
Hence, the need for verification arises. Popper rejects the inductivist view that
our observations are theory-free and hence rejects the idea that our observation
statements have probability equal to 1. More importantly, he maintains that
theories are not winnowed from observations or facts, but are free creations of
human mind. Our scientific ideas, in other words, are not extracted from our
observations; they are pure inventions. Since our theories are our own
27
Research Methodology: constructions, not the functions of anything like pure observations, which
according to Popper are anyway myths, the initial probability of our scientific
theories is zero.
From this it follows that whereas according to the inductivists what scientific
tests do is to mrely find out whether our scientific theories are true, according to
Popper scientific tests cannot establish the truth of scientific theories even when
the tests give positive results. If a test gives a positive result, the inductivists
claim that the scientific theory is established as true, whereas according to Popper
all that we calim is that our theory has not yet been falsified. In Popper’s scheme
no amount of positive result of scientific testing can prove our theories. Whereas
the inductivists speak of confirmation of our theories in the face of positive
results of the tests, Popper only speaks of corroboration. In other words, in the
inductivist scheme we can speak of scientific theories as established truths,
whereas in the Popperian scheme, a scientific theory, however, well supported
by evidence remains permanently tentative. We can bring out the fundamental
difference between verificationism (inductivism) and falsificationism
(Hypothetico-Deductivism) by drawing on the analogy between two systems of
criminal law. According to one system, the judge has to start with the assumption
that the accused is innocent and consequently unless one finds evidence against
him, he should be declared innocent. According to the other, the judge has to
start with the assumption that the accused is a culprit and consequently, unless
evidence goes in his favour, he should be declared to be a culprit. Obviously the
latter system of criminal law is harsher than the former. The inductivist scheme
is analogous to the former kind of criminal law, where as the Hypothetico-
Deductive scheme is akin to the latter one.
The steps of Scientific Procedure

In the Popperian scheme we begin with a problem, suggest a hypothesis as a
tentative solution, try to falsify our solution by deducing the test implications of
our solution, try to show that the implications are not borne out and consider our
solution to be corroborated if repeated attempts to falsify it fail. Thus, problem,
tentative solution, falsification and corroboration constitute the steps of scientific
procedure. Popper’s theory of scientific method is called Hypothetico-
Deductivism because, according to him, the essence of scientific practice consists
of deducing the test implications of our hypothesis and attempting to falsify the
latter by showing that the former do not obtain, whereas according to Inductivism
the essence of scientific practice consists of searching for instances supporting
the generalization arrived on the basis of some observations and with the principle
of induction.
Popper claims that the Hypothetico-Deductive model of scientific method is

superior to inductivist model for the following reasons:
Firstly, it does justice to the critical spirit of science by maintaining that the aim
of scientific testing is to falsify our theories and by maintaining that our scientific
theories are, however corroborated, going to permanently remain tentative. In
other words, the Hypothetico-Deductivist view presents scientific theories as
permanently vulnerable with the sword of possible falsification always hanging
on their head. The inductivist view of scientific method makes science a safe
and defensive activity by portraying scientific testing as a search for confirming
instances and by characterizing scientific theories as established truths. According
28
to Popper, the special status accorded to science is due to the fact that science Approaches to Scientific
embodies an attitude which is essentially open-minded and anti-dogmatic. Post Positivism
Hypothetico-Deductivism is an adequate model of scientific practice because it
gives central place to such an attitude.
Secondly, Popper thinks that if science had followed the inductivist path; it would
not have made the progress it has. Suppose a scientist has arrived at a
generalization. If he follows the inductivist message, he will go in search of
instances which establish it as a truth. If he finds an instance which conflicts
with his generalization, what he does is to qualify his generalization saying that
the generalization is true except in the cases where it has to be held unsupported.
Such qualifications impose heavy restrictions on the scope of the generalization.
This results in scientific theories becoming extremely narrow in their range of
applicability. But if a scientist follows the Hypothetico-Dedectivist view, he will
throw away his theory once he comes across a negative instance instead of pruning
it and fitting it with the known positive facts. Instead of being satisfied with a
theory, tailored to suit the supporting observations, he will look for an alternative
which will encompass not only the observations which supported the old theory
but also the observations which went against the old theory and more importantly
which will yield fresh test implications. The theoretical progress science has
made can be explained only by the fact that science seeks to come out with
bolder and bolder explanations rather than taking recourse to the defensive method
of reducing the scope of the theories to make them consistent with facts. Hence,
Popper claims that the Hypothetico-Deductive model gives an adequate account
of scientific progress. According to him, if one accepts the inductivists account
of science one fails to give any explanation of scientific progress.
Thirdly, the Hypothetico-Deductive view according to Popper avoids the

predicament encountered by inductivist theory in the face of Hume’s challenge.
As we have seen, Hume conclusively showed that the principal of induction
could not be justified on logical grounds. If Hume is right, than science is based
upon an irrational faith. According to Hypothetico-Deductivist view, science
does not use the principle of induction at all. Hence, even though Hume is right,
it does not matter since science follows the Hypothetico-Deductivism are so
radically different that the latter in no way face any threat akin to the one faced
by the former. In this connection, he draws our attention to the logical asymmetry
between verification, the central component of the inductivist scheme, and
falsification, the central component of the Hypothetico-Deductivist scheme. They
are logically asymmetrical in the sense that one negative instance is sufficient
for conclusively falsifying a theory, whereas no amount of positive instances are
sufficient for conclusively falsifying a theory, whereas no amount of positive
instances are sufficient to conclusively verify a theory. It may be recalled that
Hume was able to come out with the problem of induction precisely because a
generalization (all theories according to Inductivism are generalizations) cannot
be conclusively verified.
How does Popper characterize scientific progress? According to him, one finds
in the history of science invariable transitions from theories to better theories.
What does the work ‘better’ stand for? It may be recalled that, according to
Popper, no scientific theory, however, corroborated can be said to be ‘true’. Hence,
Popper drops the very concept of truth and replaces it by the concept of
Verisimilitude (truth-likeness or truth-nearness) in his characterization of the
29
Research Methodology: goal of science. In other words, through science cannot attain truth, i.e., though
our theories can never be said to be true, science can set for itself the goal of
achieving higher and higher degrees of Verisimilitude, i.e., successive scientific
theories can progressively approximate to truth. So, in science we go from theory
to better theory and the criterion for betterness is Verisimilitude. But what is the
criterion for Verisimilitude? The totality of the testable implications of a
hypothesis constitutes what he calls ‘the empirical content’ of the hypothesis.
The totality of the testable implications which are borne out constitute the truth
content of the hypothesis and the totality of the testable implications which are
not borne out is called the falsity content of the hypothesis. The criterion of the
Verisimilitude of a theory is nothing but truth content minus the falsity content
of a theory. In the actual history of science we always find, according to Popper,
theories being replaced by better theories, that is, theories with higher degree of
Verisimilitude. In other words, of the two successive theories, at any time in the
History of Science, we find the successor theory possessing greater Verisimilitude
and is therefore better than its predecessor. In fact, according to him, a theory is
rejected as false only if we have an alternative which is better than the one
at hand in the sense that it has more testable implications and a greater number
of its testable implications are already borne out. The growth of science is
convergent in the sense that the successful part of the old theory is retained in the
successor theory with the result the old theory becomes a limiting case of the
new one. The growth of science thus shows continuity. In other words, it is the
convergence of the old theory into the new one that provides continuity in the
growth of science. It must also be noted in this connection that unlike the
Inductivists or Positivists, Popper is a Realist in the sense, according to him,
scientific theories are about an unobservable world. This means that the real
world of the unobservable thought can never be captured entirely by our theories,
evidence that through the gap between Truth and our theories can never be
completely filled, it can be progressively reduced. Consequently, the real world
of unobservable will be more and more like what our theories say though not
completely so.
How does Popper establish the objectivity of scientific knowledge? Inductivists

sought to establish the objectivity of science by showing that scientific theories
are based upon pure observations. The so-called ‘pure’ observations were
supposed to be absolutely theory-free. They are only ‘given’ and hence, free
from subjective interpretations. Popper, as we have seen, rightly rejects the idea
of pure observations. Consequently, he cannot accept the inductivist account of
the objectivity of science. What engenders scientific objectivity, according Popper,
is not the possibility of pure observation, but the possibility of inter-subjective
testing. In short, science is objective because it is public and it is public because
its claims are inter-subjectively testable.
To the question, “Which comes first, observation or theory?” the Inductivist

answers ‘observation’. Popper answers ‘earlier observation or earlier theory’.
To him the question is as illegitimate as the question, “which comes first, egg or
hen?” which can be answered only by saying ‘earlier egg or earlier hen?’
It will be convenient if we list the main theses of Popper’s philosophy of science

arranged in a manner with our list of the theses of the positivist philosophy of
science:
30
1) Science is qualitatively distinct from, superior to and ideal for all other Approaches to Scientific
areas of human endeavour (scientism). Post Positivism
2) The distinction, superiority and idealhood that science enjoys is traceable
to its possession of a method (Methodologism).
3) There is only one method common to all science irrespective of their subject
matter (Methodological Monism).
4) That method which is common to all sciences, natural and human, is the
method of Hypothetico-Deduction (Hypothetico-Deductivism).
5) The hallmark of science (i.e., the distinguishing mark of science) consists
in the fact that its statements are systematically falsifiable (falsifiability).
6) Scientific observations are not and cannot be shown to be pure; that is, they
are theory-dependent.
7) Theories are not winnowed from observations or facts; they are pure
inventions of human mind i.e., only conjectures and not generalizations
based on ‘pure observations’.
8) The relation between observation and theory is one of interdependence.
9) To a given set of observation-statements there might correspond more than
one theory.
10) Our factual judgements may have value commitments and our value
judgements may have cognitive content (hence fact-value dichotomy is
unacceptable); science is not value neutral but the value commitments can
be critically discussed and therefore they are not subjective.
11) All scientific explanations must have deductive-nomological pattern and
thus the thesis of Deductive-Nomologism is acceptable.
12) The aim of science is to provide an account of observable world in terms of
unobservable entities and to provide accounts of those unobservable entities
in terms of further unobservable entities. Unobservable entities are,
therefore, real and our theories are putative descriptions of such real entities
(‘Realism’).
13) Unlike other areas of human activity, there is progress in science which
consists in going from one theory to a better theory. Here, ‘better’ means
‘more true’. ‘More to true’ means ‘the world of unobservables’. In short,
science is progressive in the sense our successive theories in any domain of
science exhibit greater and greater verisimilitude or truth-nearness i.e. the
match between our theories and reality. Unlike positivists, Popper rejects
the idea that progress of science is characterized by cumulative growth of
theories. According to him, a new theory is entirely new and not an old
theory plus an epsilon as Positivists thought. Thus, in Popper’s scheme, the
growth of science is essentially discontinuous. Of course, Popper makes
some room for continuity also when he says that old theory (at least true
part of it) is a limiting case of the new theory.
14) Science is not objective in the sense scientific theories are based on pure
observations as positivists thought because there are no pure observations.
Science is objective in the sense its theories are inter-subjectively testable.
31
Research Methodology: 15) Lastly, science is not rational in the sense the principle of Induction can be
rationally justified as Positivists thought. The principle of Induction cannot
be rationally justified; nor is it used by science. Science is rational in the
sense it embodies critical thinking. Apart from insisting that our theories be
falsifiable, science has institutional mechanisms for practicing and promoting
critical thinking. What is rationality other than critical thinking? Positivists
and Popper differ from each. The theses (1), (2), (3) and (11) are common
to both Positivists and Popperians. Popper rejects most of other theses of
Positivists, especially their central thesis which concerns the idea of pure
observation. Finally, he agrees with the Positivists that science is uniquely
progressive, objective and rational; but his notions of progressiveness,
objectivity and rationality of science are entirely different from those of
Positivists.
2.5 CRITICISM AGAINST KARL POPPER’S

PHILOSOPHY OF SCIENCE
A serious lacuna in Popper’s position concerns his idea of scientific progress.
First of all, according to Popper, the growth of science is essentially discontinuous

in the sense that a new theory which displaces an old theory is not the old theory
plus an epsilon because it is entirely new. Yet, he seeks to make room for continuity
in the growth of science by insisting that the old theory is a limiting case of the
new theory. In this connection he cites an example of Newtonian mechanics and
Relativistic mechanics. The former is the limiting case of the latter in the sense
that in a certain domain both give the same results. Thus the former is contained
in the latter. Hence there is some continuity in the growth of science. But Popper
overlooks the fact that such examples of an old theory being a limiting case of
the new one are rare. For example, it is absurd to say that Phlogiston theory is a
limiting case of oxygen theory or that Ptolemy’s theory is limiting case of
Copernican theory.
Secondly, Popper says that successive theories in any domain exhibit increasing
verisimilitude i.e., truth nearness. That is, reality constituted by unobservable
entities is more like what a new theory says than what its immediate predecessor
says. This means that following Popper we have to say that the ultimate
constituents of matter are more like fields as the present physical theory says
than like particles (atoms) as claimed by Newtonian theory. This is unintelligible.
What does it mean to say that the ultimate constituents of matter are more like
fields than particles called atoms? Either they are like fields or like particles.
Thirdly, when Popper says a new theory is better than the old one (in the sense
it is more true), he assumes that the two theories can be compared. This means
that they have something common which makes them comparable. But this has
been ably questioned by Thomas Kuhn who sought to show that when one
fundamental theory replaces another, the two theories are so radically different
as to make any talk of comparison between them highly questionable. We shall
now turn to his views.
32
Check Your Progress 2 Approaches to Scientific
1) On what basis, according to Popper, a line can be demarked between science Post Positivism
and rest of knowledge:

......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Identify the main differences between Inductive Model and Hypothetico-
Deduction Model.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) What are the Central components of Hypothetico-Deductive Model?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
4) In what sense, science is rationale?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2.6 THOMAS KUHN’S PHILOSOPHY OF SCIENCE

Thomas Kuhn’s work The Structure of Scientific Revolutions is a milestone
in the history of the 20th century of philosophy of science. A brief exposition of
his basic ideas are given below.
According to Kuhn, in the life of every major science there are two stages (1)
Pre-paradigmatic stage, and (2) paradigmatic stage. In the pre-paradigmatic
stage one finds more than one mode of practicing that science. That is, there was
a time in the history of Astronomy when different schools of Astronomy practised
Astronomy differently. So is the case with Physics, Chemistry and Biology. In
that stage their situation was similar to that which obtains today in areas like art,
philosophy and even medicine wherein divergent modes of practicing these
disciplines co-exist. Today, we speak of schools of Art (e.g., painting), schools
of Philosophy and systems/schools of medicine. But today we do not speak of
Schools of Astronomy or Physics or Chemistry or Biology.
33
Research Methodology: This is, according to Kuhn, in areas like art, philosophy and medicine that did
not, and cannot make a transition from pre-paradigmatic stage to paradigmatic
stage, which marks the disappearance of plurality, that is, disappearance of
schools. In other words, the transition means replacement of plurality by monolith
i.e., a uniform mode of practice.
Such a transition is made possible, Kuhn claims, by acquisition of a paradigm.

When a science makes such a transition, we may say, it has bcome ‘mature’ or
‘science’ in the proper sense of the term. Astronomy was the first to make such a
transition followed by Physics, Chemistry and Biology in that order. Social
Sciences are still, according to him, in the pre-paradigmatic stage, though
Economics is showing signs of such a transition. This is evident from the fact
that in Social Sciences there is no consensus on fundamentals as we can see
prevalence of distinct schools in every Social Science.
So, the transition to maturity is effected by acquiring a paradigm by a science.

The question is “What is a paradigm?”.
We all know that Ptolemy’s Almagest Newton’s Principia and Darwin’s Origin
of the Species are path-breaking works in the areas of Astronomy, Physics and
Biology respectively. According to Kuhn, these works provided paradigms for
these disciplines. They did so by specifying the exact manner in which these
disciplines ought to proceed. They laid the ground rules regarding what problems
these disciplines must tackle and how to tackle them. Hence, paradigms are
Universally recognized achievements that for a time provide model problems
and solutions to community of practitioners. Hence, in the first place, a paradigm
specifies what that ultimate constituents of that sphere of reality, which a particular
science is inquiring into, are.
Secondly, it identifies the model problems. Thirdly, it specifies the possible

range of solutions. Fourthly, it provides the necessary strategies and techniques
for solving the problems. Lastly, it provides examples which show how to solve
certain problems. In other words, a paradigm is a disciplinary matrix of a
professional group. Once a science possesses a paradigm, it develops what Kuhn
calls, a ‘normal science tradition’. Normal science is the day-to-day research
activity purporting to force of nature into conceptual boxes provided by the
paradigm. The practitioners of normal science, that is, a scientist who engages in
day-to-day research, internalizes the paradigm by professional education. This
explains the prevalence of textbook culture in science education.
Of course, scientific practice is not exhausted in terms of day-to-day research or

‘normal science’. When a paradigm fails to promote fruitful, interesting and
smooth normal science, it is considered to be in a crisis. The deepening of the
crisis leads to the replacement of the existing paradigm by a new one. This process
of replacement is called ‘scientific revolution’. Therefore, scientific revolutions
are “the tradition-shattering complements to the tradition bound activity of normal
science.” Thus, once a science enters the paradigmatic stage, it is characterized
by (1) normal science, and (2) revolutions. In sheer temporal terms, normal science
occupies much larger span than revolutionary science. That is to say, science is
revolutionary once a while and mostly it is non-revolutionary or normal. Also
the scientific activity engaged in by most of the practitioners can be characterized
aptly in terms of noral science. On account of this temporal and numerical
34
magnitude we can say that much of the scientific activity as we ordinarily Approaches to Scientific
encounter is normal though this normal course is occasionally interrupted by Post Positivism
revolutions which change the form, content and direction of the process of the
scientific activity, which is basically normal by which we mean a non-
revolutionary committed and tradition bound activity. Normal science demands
a through going convergent thinking and hence is preceded by an education that
involves ‘a dogmatic initiation in a pre-established tradition that the student is
not equipped to evaluate’. “Normal science is an activity that purports not to
question the existing paradigm but to (1) “Increase the precision of the existing
theory by attempting to adjust the existing theory or existing observation in order
to bring the two into closer and closer agreement”, and (2) “to extend the existing
theory to areas that it is expected to cover but in which it has never before been
tried”. In other words, normal science consists of solving puzzles that are
encountered in forcing nature into the conceptual boxes supplied by the reigning
paradigm.
It is in this way Kuhn attempts to account for the smooth, defined and directional
character of day-to-day scientific research in terms of the features of what he
calls “Normal Science”. Normal Science has no room for any radical thinking. It
is limited to the enterprise of solving certain puzzles in accordance with the
rules specified by the paradigm. These rules are never questioned but only
accepted and followed. The aim of scientific education is to ensure that the
paradigm is internalized by a student. In other words, the professional training in
science consists in accepting the paradigm as given and equipping oneself to
promote the cause of paradigm by giving a greater precision and further
elaboration. The day-to-day scientific research does not aim at anything
fundamentally new but only at the application of what has already been given,
namely the theoretical ideas and the practical guidelines for solving certain
puzzles. It is in this sense that normal science is a highly tradition-bound
activity.
As pointed out earlier, normal science purports to force nature into the conceptual
boxes provided by the reigning paradigm by solving puzzles in accordance with
the guidelines provided by the paradigm whose validity is accepted without
question. During this process of puzzle solving, certain hurdles may be
encountered. We than speak of “anomalies”. That is, an anomaly arises when a
puzzle remains puzzle defying every attempt to resolve it within the framework
of the paradigm. But, appearance of one or two anomalies is not sufficient to
overthrow a paradigm. The ushering in of the ear of a new paradigm has to be
preceded by the appearance of not one or two minor anomalies, but many major
ones. In order to declare a paradigm to be crisis-ridden, what is needed is an
accumulation of major anomalies. But there is no clear cut and objective criterion
to decide which anomalies are major and how many anomalies must accumulate
to declare a paradigm to be crisis-ridden. In other words, there is no criterion
which decides whether a perceived anomaly is only a puzzle or the symptom of
a deep crisis. The issue will be decided by the community of the practitioners of
the discipline through the judgment of its peers. Once the scientific community
declares the existing paradigm to be crisis-ridden, the search for the alternative
begins. Of course the crisis-ridden paradigm will not be given up unless a new
theory is accepted in its place. It is only during this transitional period of search
for the new paradigm that the scientific debates become radical.
35
Research Methodology: During the process of search for an alternative, the scientific community has to
make a choice between competing theories. In this choice, the evaluation
procedures of normal science are of no help, “for these depend in part upon a
particular paradigm and that paradigm is at issue.” The issue concerning the
paradigm choice can not be settled by logic and experiment alone. What ultimately
matters is the consensus of the relevant scientific community. Kuhn points out,
“that question of value can be answered only in terms of criteria that lie outside
of normal science altogether, and it is that recourse to external criteria that most
obviously makes paradigm debates revolutionary.” Thus a paradigm choice cannot
be explicated in the neutral language of mathematical equations and experimental
procedures, but in terms of specific perceptions which a scientific community as
a social entity entertains about what it considers to be the basic values of its
professional enterprise. In other words, the ultimate explanation of a theory choice
is not methodological but sociological. Hence in Kuhn’s scheme, the idea of
scientific community as a social entity is axiomatic. That is to say, according to
him, “If the term “paradigm” is to be successfully explicated, scientific
communities must first be recognized as having an independent existence”. This
means that one must explain scientific practice in terms of paradigms and
paradigmatic changes and the latter are to be explicated in terms of a particular
scientific community which shares the paradigms and brings about paradigmatic
changes. Thus, the concept of scientific community is basic to the concept of
paradigm. The concept of scientific community can be explicated in sociological
terms. Hence, the ultimate terms of explication of scientific activity are
sociological.
What is the relation between the old paradigm, which is overthrown and the new
paradigm which succeeds it? Kuhn’s answer to this question is extremely radical.
According to him, in no obvious sense one can say that the new paradigm is
better or truer that the old one. Kuhn maintains that the two successful paradigms
cut the world differently. They speak different languages. In fact when a paradigm
changes, to put it metaphorically, the world changes. With his characteristic
lucidity he says, “the transition from a paradigm in crisis to new one from which
a new tradition of normal science can emerge is far from accumulative process,
one that is achieved by an articulation or extension of the old paradigm. Rather,
it is a reconstruction of the field from new fundamentals, a reconstruction that
changes some of the field’s most elementary theoretical generalizations as well
as many of its,… methods and applications.” This apart Kuhn contends that the
two paradigms talk different languages. Even if the same terms are used in two
paradigms, the terms have different meanings. What can be said in the language
of one paradigm can not be translated into the other language. Based on these
reasons, Kuhn claims that the relation between two successive paradigms is
incommensurable. No wonder Kuhn compares paradigm shift to gestalt shift.
With this, the idea of scientific progress as a continuous process and the idea of
truth as the absolute standard stand totally repudiated. Kuhn advances what might
appear to be an undiluted relativism according to which truth is intro-paradigmatic
and not inter-paradigmatic. That is to say what is true is relative to a paradigm
and there is no truth lying outside all paradigms.
36
Approaches to Scientific
2.7 POPPER VERSUS KUHN Knowledge: Positivism and
Post Positivism
Some of the radical implications of Kuhn’s position can be brought about by
juxtaposing his views with those of Popper.
Firstly, the hallmark of science according to Popper is critical thinking. In fact

science exemplifies critical thinking at its best. Since critical thinking considers
nothing to be settled and lying beyond all doubt, fundamental disagreements and
divergent thinking must and in fact do characterize science. As we have seen,
according to Kuhn, what constitutes the essence of scientific practice is normal
science and we have also seen why normal science is a highly tradition-bound
activity, an activity made possible by a consensus among the practitioners who
share a paradigm. Thus if Popper sees the essence of science in divergent thinking
and fundamental disagreements, Kuhn sees the essence of science in convergent
thinking and consensus. In other words, the hallmark of science according to
Kuhn is tradition-bound thinking. In fact, according to Kuhn, what distinguishes
science from other areas of creative thinking is that whereas in science one finds
institutional mechanisms of enforcing consensus, the other areas suffer from
perpetual disagreements even on fundamentals.
Secondly, if Popper considers the individual to be the focus of scientific activity,

Kuhn bestows that status upon the scientific community. Both positivists and
Popper looked upon science as the sum total of the work of individual scientists
working in accordance with a method though the Positivists and Popper
fundamentally differed on the characterization of that method. As opposed to
this individualistic account of scientific enterprise, Kuhn propounds a
collectivistic view of scientific activity. In Kuhn’s scheme, it is the scientific
community, which constitutes the pillar of stability and locomotive of change.
This is borne out by the fact that according to Kuhn the scientific community has
institutional mechanism, like peer review, by which it can settle all the issues
such as whether an anomaly is a symptom of crisis, how many anomalies suffice
to warrant the search for an alternative paradigm, what factors are to be considered
in choosing a new theory for the status of the paradigm etc.
Thirdly, Popper and Kuhn differ fundamentally in their attitude towards the
transition from one theory to another theory in science. According to Popper, we
can explain every case of change of theoretical framework in terms of certain
norms which science always adopts and follows meticulously. In fact, scientific
rationality consists in following these norms. But Kuhn contends that an adequate
explanation of change of theoretical framework must be in terms of the value
judgments made by a community while making the choice. According to Kuhn,
recourse to the so-called methodological norms explains nothing. From the point
of view of Popper, Kuhn is an irrationalist because he sets aside methodological
norms and seeks to explain theory change exclusively in terms of non-rational or
sociological factors like value commitments of a professional group. Whatever
be the merit of Popper’s attack on Kuhn as an irrationalist, we can say that Kuhn’s
construct of scientific practice is sociological. That is to say, according to him,
scientific activity can not be understood by trying to find out the absolute standards
which have guided the scientific activity in all ages. It can only be understood in
terms of the specific judgments which a community makes at a particular juncture
regarding what it considers to be its value commitments as a professional group.
37
Research Methodology: The above juxtaposition of Popper and Kuhn brings out the radical implications
of Kuhn’s views regarding the nature of scientific practice. However, in one
respect Kuhn is very close to Popper. Both, like positivists, content that there is
something unique to science though they differ in their explanation of what that
uniqueness consists of. Positivists maintain that the hallmark of science is the
systematic verifiability of its claims. According to Popper, the uniqueness of
science consists in the systematic falsifiability of theories. Accounting to Kuhn,
it is consensus which marks out science from other areas of human endeavour.
That is to say, Kuhn, like Positivists and Popper, does not question whether
science is really unique. That is to say, instead of raising critical questions about
the status that science has acquired in the contemporary culture, Kuhn only seeks
to provide an alternative account of how it has acquired that status. In that sense
Kuhn’s position is quite conservative.
In this unit, we had a brief look at the 20th century thinking on the nature of
science. It is very difficult to decide which view is the correct one, though
Positivist view has been shown to be highly inadequate. The question is “How
should we practise social sciences so as to make them scientific”?
Some social scientists take the Positivist recommendation: “collect data, extract
a generalization, verify the generalization and formulate a law”. Those social
scientists who are inspired by the Popperian view, take seriously the Popperian
advice “Formulate a problem, provide a tentative solution, try to falsify it, if the
solution survives treat it as a corroborated theory but not as a confirmed one”.
Still others go by Kuhn’s view of science and think that the task of social sciences
today is to arrive at paradigms in different social scientific disciplines. This will
enable social sciences to overcome ideological commitments which generate
differences even at a fundamental level. According to them, the consensus so
generated will bring social sciences close to natural sciences.

1) Which two stages come in the life of every major science?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) What do you understand by the term ‘Paradigm’?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
38
3) What is scientific revolution? Approaches to Scientific
...................................................................................................................... Post Positivism
......................................................................................................................
......................................................................................................................
......................................................................................................................
4) How did Popper differ with Kuhn about essence of science?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
5) Match the following:
A B
1) Individual as the locus of scientific study i) Positivists
2) Explanation of theory in terms of value judgment ii) Popper
3) Theory changes in terms of certain norms iii) Kuhn
4) Verification as the hall mark of science iv) Popper
2.8 LET US SUM UP

By virtue of shaping our mode of living and our ways of thinking about the
world, science has acquired a central place in intellectual world. Three disciplines
namely history of science, sociology of science and philosophy of science have
made science itself their object of inquiry. Philosophy of science deals with
broadly two questions: what is the aim of the science and what are the methods
of science? Different scholars have answered these questions differently.
Regarding aims and methodology two views have been dominant. ‘inductivism’
by Francis Bacon and ‘Hypothesism’ by Rene Decartes. Positivism has been a
movement in the philosophy of the first half of the 20th Century. Inductivitism,
systematical verifiable statements, pure scientific observations, unilateral
relationship between observation and theory, value neutral theory, precise
prediction of facts as aim of science, progressive/objective/rational arguments
as essentials of science are some of the important central tenets of positivism.
The concept of pure observation in positivist philosophy has been attacked on
several grounds.
Karl Popper and Thomas Kuhn are the two most important philosophers of mid
20th century. They offered alternatives to positivism and shaped the contemporary
debate on theory of scientific knowledge. Popper rejected the central thesis of
positivism i.e., idea of ‘pure observation’. He propounded the Hypothetico-
Deductive model as an adequate and superior model of scientific method.
According to him, the scientific statements are systematically falsifiable. The
scientific observations are theory-dependent. Theories are pure inventions of
human mind. The theory and observation are interdependent. Science is not value
39
Research Methodology: neutral and therefore values are not subjective. Science is progressive. Science
is objective in the sense that its theories are inter-subjectivity testable. Science,
according to Popper, is rational in the sense that it embodies critical thinking.
Owing to commonality of views of Popper with those of positivists regarding
scientism, methodologism, methodological monism and Deductive Nomologism,
Popper is sometimes viewed as semi-positivist.
Karl Popper’s ideas of scientific progress, verisimilitude and comparison of old

theory with new theory have been criticized.
Thomas Kuhn sees the life of major sciences in two stages – Pre-paradigmatic
stage and paradigmatic stage. In the pre-paradigmatic stage, more than one mode
are practised. In the process of transition a discipline acquires the stage of
paradigm. When a paradigm fails to promote fruitful, interesting and smooth
normal science, It is considered in crisis. Deepening of crisis leads to replacement
of the existing paradigm, which is termed as scientific revolution.
Thus, on the question how we should practise social sciences so as to make them
‘scientific, three views have so far been propounded. Positivist recommendation
is to ‘collect data, extract generalization, verify the generalization and formulate
a law’. Popper’s view is “formulate a problem, provide a tentative solution, try
to falsify it, if solution survives, treat it as a corroborated theory but not as a
confirmed one’. Which one out these three is correct – is difficult to decide?
Consensus on the three models will bring social science close to natural sciences.
2.9 KEY WORDS

Inductivism : A scientific method of knowledge focusing that
scientific knowledge is generated by arriving
at generalizations on the basis of pure
observations using the principle of induction.
Hypothesisism : Hypothesisism is a method of knowledge which
envisages that knowledge is created or
generated by putting forth the hypothesis having
test implications. Hypothesis is a statement
describing the unobservable entities which lie
behind observations.
Realists : Those who support Hypothesis are realist in
connection with the status of theoretical/
unobservable entities.
Anti-Realists Inductivists are also called as Anti-Realists.
Paradigm : Paradigm refers to the scientific practice
comprising of a particular a world view, concept,
assumptions, theories, models used by the
researcher in carrying out the research studies.
The concept was introduced by Thomas Kuhn.
Deductive : Derivation of a conclusion by reasoning or by
inference in which the conclusion follows
necessarily from the premises.
40
Methodological Dualism : Referring to positivist belief in separation of Approaches to Scientific
the knower from the known or of the subject Post Positivism
from object.
Methodological Monism : In contrast to the comparatmentalisation of
dualism, monism view the world as a “seamless
web”. In terms of Gouldner’s argument, the
separation between the knower and the known
must be overcome.
Pseudo Science : Pseudo science refers to any body of knowledge
or practice, which purports to be scientific or
supported by science but which is judged to fall
outside the domain of science.

Mark Blang (1992): The Methodology of Economics, Cambridge University Press,
London.
Uma Devi. S. (1994): Economic Theory and Methods of Reasoning, Har-Anand

Publication, New Delhi.
Williams, M., & May, T. (1996): Introduction to the Philosophy of Social

Research, London: UCL Press.
Kuhn, Thomas, (1970); The Structure of Scientific Revolutions. The University

of Chicago Press: Chicago.
Popper, Karl (1972); Conjectures and Refutations: The Growth of Scientific

Knowledge, Rutledge and Kegan Paul: London.

PROGRESS EXERCISES
1) The aim of science is two fold: Definition and Demonstrations.
2) See Sub-section 2.2.1
4) A B
i) iv)
ii) i)
iii) ii)
iv) iii)
5) See Section 2.3
41
Research Methodology: Check Your Progress 2
1) On the basis of falsifiability of scientific theories.
2) Inductivist model’s basic proposition is that observations are pure i.e. theory
– independent and theories are derived from pure observations. In
Hypothetico-deductive model, theories are perceived as our mental (own)
constructions and are not the functions of pure observations.
3) Hypothetico-Deductive model’s central components are: (i) theories are our
own constructions reflecting unobservable entities, (ii) science is progressive
in terms of increasing very similitude, (iii) science is rational in terms of
critical rationalism and objective in terms of inter-subjectivity.
4) Science embodies critical thinking.

1) (i) Pre-paradigmatic stage, (ii) Paradigmatic stage
2) The constituents of reality, model problems, range of solutions, necessary
strategies and techniques for solving the problems together constitute the
paradigm.
3) The process of replacement of existing paradigm is known as ‘scientific
revolution’.
4) According to Popper, the essence of science lies in divergent thinking and
fundamental disagreements whereas Kuhn sees essence of science in
convergent thinking and consensus.
5) (1) ii, (2) iii, (3) iv, (4) i.
42
Approaches to Scientific
UNIT 3 MODELS OF SCIENTIFIC Knowledge: Positivism and
Post Positivism
EXPLANATION
Structure
3.0 Objectives
3.1 Introduction
3.2 Unified View of Rules of Positivism
3.3 Search for the Criterion of Cognitive Significance
3.4 Rules of Logic or Rules of Correct Reasoning
3.4.1 ‘Categorical Syllogism’
3.4.2 Hypothetical Syllogism
3.4.3 ‘Logical Truth’ and ‘Material Truth’
3.4.4 Deductive Fallacies
3.5 Models of Scientific Explanation
3.5.1 Hypothetico-Deductive Model
3.5.2 Covering-Law Models
3.5.2.1 Deductive-Nomological (D-N) Model
3.5.2.2 Inductive-Probablistic (I-P) Model
3.5.2.3 Critical Appraisal of Covering-Law (D-N,I-P) Models of Explanation
3.6 Explanation in Non-Physical Sciences
3.7 Let Us Sum Up
3.8 Key Words
3.10 Answers or Hints to Check Your Progress Exercises
3.0 OBJECTIVES
• Explain the basic rules of formal reasoning;
• State the role of logical reasoning in the formulation of a research problem;
• Describe the basic rules of structural relationship between assumptions,
theory and conclusions; and
• Appreciate the problems involved in systematic explanation of phenomena.
3.1 INTRODUCTION
Explanation has three meanings: One interpretation of explanation is to remove
a perplexity or solve mystery i.e., ‘puzzle solving’. The second meaning is to
change the unknown into the known. The third interpretation of explanation is
to give causes of phenomenon to be explained. The first two meanings are too
relativistic and subjective. It is the third one emphasizing causal explanation
43
Research Methodology: subserving regularity, and law like explanations that fits the meaning of scientific
explanation. Here, we are concerned with one mainstream approach to science
viz,. Positivism. We may begin this unit by recognizing the complexity of
explaining subject matter of philosophy of science. The first two units make it
clear, that there is no general agreement on Positivism as the only approach to
pursue knowledge. It is at least over fifty years, since when various limitations
and misconception of the rules and propositions of Positivism have been brought
to light. But yet, paradoxically, the mainstream notion of science and scientific
explanation is dominated by the tenets of Positivism. This is partly because of
the rigorous and prescriptive nature of positivist propositions which remain simple
and attractive, and partly because of the open ended and speculative nature of
the alternative approaches. The enthusiasm with which many mainstream
economists still prefer to refer to Economics as a Positive Science is yet another
indication that alternative approaches in the Philosophy of Science have not yet
percolated to Economics. Because of these reasons this unit on ‘Models of
Scientific Explanation’, and the next one on ‘Debates on the Models of Scientific
Explanation in Economics’ are essentially confined to Positivist Models of
Explanation.
Let us begin with a broad unified view of Positivism in terms of certain rules. It
will be followed by certain rules of deduction or formal logical reasoning that
underlie Positivist Models of explanation. Afterwards, the various Models of
Explanation and their limitations, will be discussed.
3.2 UNIFIED VIEW OF RULES OF POSITIVISM

Our emphasis is on the Models of Scientific Explanation under Positivism.
Positivism is procrustean in nature. Its propositions, often contested, evolved
over a long period. It passed through several stages, from radical empiricism to
logical positivism to logical empiricism. And the propositions of Positivism at
various stages varied. Nevertheless, we shall confine here to a unified view of
Positivism.
Kilakowski provides such a unified view by simplifying Positivism into a set of

four fules:
K1 The Rule of Phenomenalism

The Rule emphasizes phenomena as the basis of knowledge. Sensory experience
is the basis. The basis of knowledge is only the record of that which is actually
manifested in experience. It is phenomenon not noumina. It is existence, not
essence. Simply, it is facts and objective facts only that form the basis of
knowledge.
K2 The Rule of Nouminalism

Nouminalism refers to Insight representative of facts. Any insight formulated in
general terms can not have any real reference other than individual facts. Therefore
every abstract science is abridging the recording of experience, and gives no
extra independent knowledge.
44
K3 The Rule of Value-free Statements Models of Scientific
Explanation
Knowledge is value-free. The rule refutes to call value judgments and normative
statements as knowledge. There is no room for good or bad or ethical judgments
in statement of science or knowledge.
K4 The Rule of Unity of Science

There is only one method of scientific knowledge. The Unity of science refers to
a single fundamental form from which all other laws are ultimately derived.
3.3 SEARCH FOR THE CRITERION OF

COGNITIVE SIGNIFICANCE
In the above summation of Positivism into a set of rules, K 1, Rule of
Phenomenalism, may appear to be closer to the early inductivist phase of
Positivism. The inductivist view of science emphasized observation of facts and
to proceed towards by inductive inference to the formulation of universal laws
about these facts. But Positivism underwent considerable sophistication to
accommodate deductive reasoning as a part of scientific explanation.
Yet another aspect of Positivism which is important in understanding its approach

to scientific explanation is the criterion of cognitive significance. According to
Positivism, the basic criterion for distinguishing ‘scientific knowledge’ or
meaningful statements from ‘non-scientific’ or meaningless statements is
testability. In the early phase the emphasis was on of verification – only complete
verification by observational evidence to be considered empirically meaningful.
But soon it was found to be too strict, and also faced the problem ‘induction’.
Karl Popper suggested falsification testing instead of verification. But both
verification and falsification tests are found to be too strict criteria, and
alternatively, Rudolf Camap suggested confirmation. Confirmation is a relative
concept. Instead of a single test for rejection or acceptance of a hypothesis, a
series of positive results are supposed to increase the confirmation of it. Test
instances confirmed or disconfirmed hypotheses, and hypotheses could be ranked
according to their degree of confirmation relative to the available evidence. Later,
emphasis shifted from testing individual theories to evaluation of competing
theories on the basis of their relative degree of confirmation. But facts and testing
remain basic ingredients of Positivist explanation.
3.4 RULES OF LOGIC OR RULES OF CORRECT

REASONING
Just as facts and testing, rules of logic have become important part of the whole
structure of scientific explanation under Positivism. Here we briefly review the
main stay of deductive reasoning viz. ‘syllogism’ or ‘rules of logic’ or ‘rules of
correct reasoning’.
3.4.1 ‘Categorical Syllogism’

Two categorical statements if logically formulated, taken together would lead to
a conclusion. The two statements should serve as the premises. If there is no
premises, then there will be a fallacy or enthymeme. 45
Research Methodology: For example, the following two examples stand for categorical syllogisms:
I) All As are B } → Major Premis
} Premises
C is an A } → Minor Premis
Therefore, C is B → Conclusion
II) All Politicians are corrupt } → Major Premis

} Premises
Blogg is a politician } → Minor Premis
Therefore, Blogg is corrupt → Conclusion
The following is an example of a fallacy of enthymeme because it leaves out the

premises:
Blogg is a politician.
Therefore, he is corrupt.
3.4.2 Hypothetical Syllogism

Hypothetical Syllogism is referred to as Modes Poneus, affirming the antecedent.
The major premise has if……………….then form.
The following are two examples:
If A is true, then B is true Therefore, B is true.

(Antecedent) (Consequent)
If Blogg is a politician, then he is corrupt Therefore, Blogg is corrupt

(Antecedent) (Consequent)
Hypothesis statements may have a major Hypothesis (H) and auxiliary

(A1A2………..An) hypotheses.
3.4.3 ‘Logical Truth’ and ‘Material Truth’

All statements which are logically true are not necessarily “materially true”. For
example:
All politicians have two tongues.
Blogg is a politician
Therefore, Blogg has two tongues.
The above statement is “logically true” but “materially” not true.
3.4.4 Deductive Fallacies

Deductive fallacies are those where premises do not lead necessarily to the stated
conclusions. There are three main types of deductive fallacies:
1) Logical (formal) Fallacies (Fallacy of affirming the consequent)
Correct Reasoning or ‘Modus Poneus’ ‘Logical (Formal) Fallacy
(Affirming the antecedent) (Affirming the consequent)
46
If A is true, then B is true. If A is true, B is true. Models of Scientific
Explanation
A is true. B is true.
Therefore, B is true. Therefore, A is true.
2) Verbal Fallacies (Fallacy of Composition)
The verbal fallacy involves a statement where something which is true of the
part is also made true of the whole. For example, in case of people waiting to see
a procession:
If one rises on one’s tip-toes, one can see better.
If people rise on their tip-toes, they can see better.
3) Material Fallacies (All others): Material fallacies are of several kinds.

a) Post hoc ergo propter hoc: (After this, therefore because of this).
Because event B occurs after event A, then event B is necessarily caused
by event A.
b) “Argument by Analogy”: It refers to a statement where, “A is similar to
B, therefore whatever is true of A is also true of B”. For example,
Singapore was backward in 1950s. It developed faster because of “open
economy policies”. India was backward in 1950s. Therefore, India would
have prospered by ‘open economy policies’.
c) “Appeal to Authority”: It refers to a statement where the truth is sought
to be asserted by referring to an authority. “A is true because (so & so)
say it so”. For example, “India is shining because Milton Friedman has
said so”.

1) What do you mean by the term ‘Scientific explanation’?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) What is the difference between Phenomenalism and Nominalism.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) State the rules of logic. Explain ‘categorical syllogism’.
......................................................................................................................
......................................................................................................................
......................................................................................................................
47
Research Methodology: 4) What is deductive fallacy?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3.5 MODELS OF SCIENTIFIC EXPLANATION

“Men do not think they know a thing till they have grasped the ‘why’ of it”
- Aristotle
In the light of understanding of the basic building blocks of knowledge viz.,
facts, testing and formal logic, we shall now turn to the Positivist Models of
Scientific Explanation. The basic building blocks of scientific explanation are
theories. The status, structure and functions of theories and theoretical terms
become essential part of scientific explanation. The aim of scientific research is
not only to discover and describe events and phenomena in the world but also,
and more importantly, to explain and understand why these phenomena occur as
they do.
The early Positivists like Aguste Comte and Ernest Mach did not believe in any
role to explanation in science. Mach believed that theories are mere heuristic
devises and talked of theories being “eliminative fiction” useful only in organizing
data. But Positivism has traveled a long distance since, and by mid-1950s was
referring to explanation as the Chief objective of Science. “To explain the
phenomena in the world of our experience, to answer the question “Why?” rather
than only the question “What?” is one of the foremost objectives of all rational
enquiry…”. Explanation removes puzzlement and provides understanding.
However, there exist considerable differences of opinion as to the function and
the essential characteristics of scientific explanation.
What follows is the description of some of the basic models of scientific

explanation within the positivist approach. First, we shall discuss the hypothetico-
Deductive Model (H-D Model), and then turn to Covering Law Models
encompassing the Deductive-Nomological (D-N) and Inductive-Probablistic
(I-P) Models.
3.5.1 Hypothetico-Deductive Model

At a time when the Positivists were not ready to concede any role to theoretical
explanation, Carnap and Hempel, in their writings, came out with a model –
which later came to be known as Hypothetico-Deductive (H-D) model of
explanation. The H-D model not only describes the structure of theories, but
provides answers to the questions of the status and functions of theories, as well.
H-D model explicitly addresses the problems of a theory’s structure. The
propositions in a deductive system are seen as being arranged in an order of
levels. Higher-level hypotheses will refer to theoretical entities, all which need
not be tested. Lower-level hypotheses describe observable phenomena and are
tested against reality. H-D model addresses the problem of status of theoretical
48
terms which may not be testable directly, to gain meaningfulness indirectly by Models of Scientific
Explanation
the successful confirmation of theory in which they are embedded.
In H-D model, theories as a whole are tested by comparing their deduced

consequences (prediction) with data. Every theoretical term need not be given
empirical counterparts via correspondence rules. H-D model also turns the old
realist-instrumentalist controversy into a moot debate. Realists claim that all
theoretical terms must refer to real entities and theories, which do not, are false.
Instrumentalists insist that theories are only instruments for predictions. Only
relevant question for them is whether theories are adequate for prediction. H-D
model seems to accommodate both these concerns and defuse the controversy.
The H-D model rejects the earlier notion that theories have no role but are only
instruments. H-D model emphasizes the following positive functions of theories:
1) They allow generality in the specification of scientific laws.
2) They possess ‘a certain formal simplicity’ which allows the use of ‘powerful
and elegant mathematical machinery’.
3) They can serve the practical function of allowing the scientist to discover
interdependencies among observable.
4) They are convenient and fruitful heuristic (intellectual) devices, often serving
an explanatory function of their own.
Thus, the H-D model without minimizing the importance of observable

phenomena for scientific explanation, allowed far more substantial role for
theories and theoretical terms than did their predecessors.
3.5.2 Covering-Law Models

Carl G. Hempel and Paul Oppenheim developed Deductive-Nomological
(D-N) Model of scientific explanation and much later, Hempel extended it to
include Inductive-Probablistic (I-P) Model. These two models together are
referred to as Covering-Law Models of explanation.
3.5.2.1 Deductive-Nomological (D-N) Model

Hempel and Oppenheim in their paper entitled “Studies in the Logic of
Explanation” advanced an account of scientific explanation, which later came to
be known as deductive-nomological (D-N) model. They divided an explanation
into two major constituents – the explanandum and the explanans. The
Explanandum means the sentence describing the phenomenon to be explained
(not that phenomenon itself). The Explanans mean the class of those sentences,
which are adduced to account for the phenomenon. The explanans fall into two
sub-classes, viz., certain antecedent conditions C1 , C2 …….Ck and certain general
laws L1, L2…….L3. If a proposed explanation is to be sound, conditions of
adequacy must be satisfied. The following four conditions of logical and empirical
adequacy must be satisfied:
i) Logical Conditions of Adequacy

R1) The explanandum must be a logical consequence of the explanans; in
other words, the explanandum must be logically deduced from the
49
Research Methodology: information contained in the explanans, for otherwise, the explanans
would not constitute adequate grounds for the explanandum.
R2) The explanandum must contain general laws, and these must actually
be required for the derivation of the explanandum. Though not a
necessary condition, the explanans must contain at least one statement
which is not a law.
R3) The explanans must have empirical content, i.e., it must be capable,
at least in principle, of test by experiment or observation. The point
deserves special mention because certain arguments which have been
offered as explanations in the natural and in the social sciences violate
this requirement.
ii) Empirical Condition of Adequacy

R4) The sentences constituting the explanans must be true.
The characteristics of explanation discussed may be summarized into

the following schema:
{C1, C2……..CK Statements of Antecedent}
Logical { conditions } Explanans

Deduction { }
{L1, L2………Lr General Laws }
{E Description of the empirical } Explanandum

phenomenon to be explained }
Hempel and Oppenheim argued that the same formal analysis, including the
four necessary conditions, applies to scientific predictions as well as to
explanation. They went on to argue that an explanation is not fully adequate
unless its explanans, if taken account of in time, could have served as a basis for
predicting the phenomenon under consideration. Consequently, whatever is said
about the logical characteristics of explanation or prediction will be applicable
to either, even if one of them is mentioned. This resulted what is known as
Covering-Law Model extended to the D-N Model of Explanation.
Extension of characteristics of explanation to prediction could formally be

presented in the following schema:
{C1, C2……..CK Statements of Conditions}
Logical { } Predecens
Deduction { }
{L1, L2………Lr General Laws }

{P Description of the empirical } Predecendum
Prediction }
50
The symmetry visualized between explanation and prediction has often drawn Models of Scientific
Explanation
criticism.
3.5.2.2 Inductive-Probabilistic (I-P) Model

In 1960s Hempel developed Inductive-Probabilistic Model. In the I-P model the
explanans consists of sentences describing the requisite initial conditions along
with statistical laws and is expected to confer upon the explanandum highly
logical, or inductive probability. The characteristics are similar to D-N Model.
The symmetry holds and hence another Covering-Law Explanation.

1) Identify the basic blocks of knowledge.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Explain how Hypothetico-Deductive Model defuses the conflict between
the realists and instrumentalists.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) State two characteristics of Deductive-Nomoligical (D-N) Model.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
3.5.2.3 Critical Appraisal of Covering-Law (D-N, I-P) Models of

Explanation
The Covering-Law Model of explanation offered by the D-N and I-P models are
subject to a number of limitations. For convenience we shall confine here to the
limitations of D-N model, some of which are also applicable to I-P model.
First is about the necessity of laws and theories in the pursuit of scientific
explanation. Laws are distinct from ordinary statements. Ordinary statements or
statement of mere facts refer to particular statements and express “contingent
truths:, like “John went home and had dinner”. Laws refer to universal statements
and express “necessary truths”. For example, “all objects are attracted to one
another with a force inversely proportional to the square of the distances between
them” is a universal and law like statement: However, there are many apparently
51
Research Methodology: universal statements which are not really universal statements. These may be the
ones that involve mere “accidental generalizations” like, “all coins in my pocket
are nickels”.
This distinction between laws and ordinary statements does not make laws as
necessary requirement for explanation. Particularly in biology, astronomy and
social sciences explanations need not be in the form of laws. Explanations may
be in the form of questions and answers. The often cited example is: Why do
Hopi Indians continue to perform rain making ceremonies even though their
ceremonies do not normally produce desired result? The explanation is to reinforce
group identity, but there is no law involved. Similarly: Why do all vertebrates’
hearts beat regularly? The explanation is to circulate blood throughout the body
which helps to maintain uniform temperature. This explanation again doesn’t
require any law like statement.
Second, explanation requires not only the factors one cites to be casually relevant
but also that they be causes. D-N model is often satisfied by citing relevant
causes. Causes explain effects, but effects do not explain their causes, and effects
of common cause do not explain one another. Based on David Hume’s famous
guillotine of time, two contiguous events do not become cause and effect. Inferring
event B as the effect of event A because it followed the later, amounts to the
“Billiard Ball Model of Causation”, where the ball reaching the pouch is seen,
‘as if’ the player knew all the laws of time and motion of the billiard ball.
Third, much of the criticism of Covering Law Model of explanation relates to

the symmetry thesis, that explanation and prediction are inherently linked, and
that only difference is the sequence. But critics point out that explanation need
not predict. For example, Darwin’s Theory of Evolution provides a great deal of
explanation on the origin and evolution of species, but it does not predict anything.
Similarly, prediction need not explain anything. For instance, prediction of
weather does not explain anything.
Fourth, the D-N Model ignores the divergence between the norms and practices
and thereby eliminates much of science from the ambit of explanation. Many
functional or technological explanations get eliminated because they are not
through any laws or regularities. Purposive behaviour needs explanation of
motivation and in motivated behaviour future affects the present. Determining
motives to be classified in antecedent conditions and not in causal relation.
Hempel and Oppenheim also recognize the problems in applying their model of
explanation in social sciences. Human behaviour often is characterized by
uniqueness and irrepeatability. It depends upon situation and previous history of
individuals.
Notwithstanding all these limitations, “many philosophers would still maintain

that the D-N Model is an important starting point for studying scientific
explanation and there is something right and important about it.” (Hausman,
1994, p.9). Particularly in the mainstream economics, scientific explanation is
strongly associated with Positivist models of explanation and thus D-N Model
becomes an important reference point for the debates on explanation in economics.
52
Models of Scientific
3.6 EXPLANATION IN NON-PHYSICAL SCIENCES Explanation
There is a view that explanation in biology, psychology, and the social sciences
has the same structure as in the physical sciences. But the causal type of
explanation is essentially inadequate in fields other than physics and chemistry.
And it is more so in the case of social sciences which involve explanation of
purposive behaviour.
Hempel and Oppenheim point out to some of these limitation of applying models
of scientific explanation to social sciences. First, events involving the activities
of humans singly or in groups have a peculiar uniqueness and irrepeatability
which make them inaccessible to causal explanation. Causal explanation
presupposes uniformities and repeatability which may not hold for phenomena
in social sciences.
Second, the establishment of scientific generalizations and explanatory principles

for human behaviour is impossible because the reactions of an individual in a
given situation depend not only upon that situation, but upon the previous history
of the individual.
Third, the explanation of any phenomenon involving purposive behaviour calls

for reference to motivations and thus for teleological rather than causal analysis.
Many of the explanations which are offered for human actions involve reference
to goals and motives. Motivational explanations often tend to have less cognitive
significance.
Attempts to extend scientific models of explanation to non-physical sciences

without closer understanding of the context often lead to inappropriate
explanations. One possibility is that there are contexts or areas within social
sciences like economics where there is room for application of scientific or to be
specific, positivist models of explanations. But at the same time there are vast
aspects of economics where such application may not be easily amenable.

1) State the difference between laws and ordinary statement.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Describe two limitations of application of scientific models to social sciences.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
53
Issues and Perspectives 3.7 LET US SUM UP
Positivism as an approach to the pursuit of knowledge is procrustean in nature.
It has passed through several stages from radical empiricism to logical positivism
to logical empiricism. Its propositions, often contested, evolved over a long period.
The main characteristics (unified view) of positivism lie in a set of its four rules-
phenomenalism, nouminalism, the rule of value free statements and the rule of
unity of science.
Testability under the positivism has been basic criteria for distinguishing scientific
knowledge from non-scientific or meaningless statement. In early phase, emphasis
was on verification. Later on Karl Popper suggested falsification testing in place
of verification. Subsequently, emphasis shifted from falsification to confirmation.
Rules of logic have become important part of the whole structure of scientific
explanation under positivism. There are two types of deductive reasoning-
categorical syllogism and hypothetical syllogism.
Theories are the basic building blocks of scientific explanation. Hence the status,
structure and functions of theories and theoretical terms are essential parts of
scientific explanation.
The basic modes of scientific explanation within the positivist approach are:
Hypothetico-Deductive Model (HD), Deductive-Nomological (DN) Model, and
Inductive Probabilistic (IP) models. The DN and IP models are subject to several
limitations. The application of scientific models to social sciences has several
limitations; peculiar uniqueness of human activities, difficulty in scientific
generalizations and explanatory principles for human behaviour, etc. Hence
applications of scientific models to non-physical sciences without closer
understanding of the context often lead to inappropriate explanation.
3.8 KEY WORDS

Deduction : Deduction is a process used to derive particular
statements from general statements.
Empiricism : ‘Empiricism’ refers to distinctive approach that

envisages that all knowledge comes from the senses.
Falsification : The concept developed by Karl Popper (1959) is a

philosophy of Science also known as critical
rationalism. It uses the logic of deduction to provide
the foundation for the Hypothetico-Deductive
Method. He rejected the idea that observation
provides the formulation for scientific theories.
Theories are invented to account for observation, not
derived from them. Observations are used to try to
reject false theories. Theories that survive this critical
process are provisionally accepted but never proven
to be true.
54
Induction : Induction is a logical process of moving from Models of Scientific
Explanation
particular statements to general statements. This is
used in the social sciences to produce theory from
data.
Logical Positivism : An approach to philosophy of science that envisages

that all meaningful statements are either empirical
or logical in character. This means that they are either
open to test by observational evidence or are matches
of definition and therefore, tautological. According
to this approach, no fundamental difference exists
between natural and social sciences.
Modelling : Refers to a process of creating a simplified

representation of a System or phenomena with
hypotheses to describe the system or explain the
phenomenon.
Theory : Refers to set of general laws that are related in

networks and are generated by inductive or deductive
form of logic. In other words, theory can be described
as a set of propositions of different levels of
generality.
Verification : Refers to the use of empirical data, observation, test

or experiment to confirm the truth or rational
justification of a hypothesis.

Brody, Baruch A. (1970); Readings in the Philosophy of Science, Prentice Hall
Inc., Englewoodcliffs, New Jersey.
Bryant, C.G.A (1985); Positivism in Social Theory and Research, London,
Macmillan.
Caldwell, Bruce J (1984); Beyond Positivism: Economic Methodology in the
Twentieth Century, George Allen and Unwin, Boston.
Hausman, D.M (1994); The Philosophy of Economics: An Anthology, Second
Ed., Cambridge University Press, Cambridge.
Popper, Karl (1959); The Logic of Scientific Discovery, Basic Books, New York.
Stewart, F. (1979); Reasoning and Method in Economics, McGraw-Hill Book
Co., London.

PROGRESS EXERCISES
1) Scientific explanation refers to give causes of phenomenon to be explained.
55
Research Methodology: 2) Phenomenon refers to a situation in which facts and objective facts based
on experience forms the basis of knowledge. It is existence, not essence.
On the other hand, nouminalism is insight representative of facts. It envisages
that any abstract concepts used in scientific explanation must also be derived
from experience.
3) Broadly there are two rules of logic. Categorical syllogism, and hypothetical
syllogism.Logical formulation of two categorical statements serving as
premises is categorical syllogism.
4) See Sub-section 3.4.4.

1) Facts, testing and formal logic.
3) See Sub-section 3.5.2.1
1) Laws refer to universal statements expressing necessary truths whereas
ordinary statement refers to particular statements of mere facts and express
‘contingent truths’.
2) See Section 3.6
56
Models of Scientific
UNIT 4 DEBATES ON MODELS OF Explanation
EXPLANATION IN ECONOMICS
Structure
4.0 Objectives
4.1 Introduction
4.2 Classical Political Economy and Ricardo’s Method
4.2.1 Ricardo’s Method
4.2.2 N.W.Senior
4.2.3 J.S.Mill
4.2.4 J.E.Cairnes
4.2.5 J.N.Keynes
4.3 Robbins, Positivism and Apriorism in Economics
4.4 Hutchison and Logical Empiricism in Economics
4.5 Milton Friedman and Instrumentalism in Economics
4.5.1 Goal of Science
4.5.2 Relationship Between Significance of Theory and ‘Realism’ of Assumptions
4.5.3 Critical Appraisal of Friedman’s Instrumentalism
4.6 Paul Samuelson and Operationalism
4.6.1 Operationalism
4.6.2 Descriptivism
4.7 Theory – Assumptions Debate in Economics: A Long View
4.8 Amartya Sen on Heterogeneity of Explanation in Economics
4.8.1 Heterogeneity of Substance and Methods of Economics
4.8.2 Testing and Verification
4.8.3 Value Judgements and Welfare Economics
4.8.4 Formalisation and Mathematics
4.9 Let Us Sum Up
4.10 Key Words
4.0 OBJECTIVES
The unit aims at tracing the broad contours of the debate relating to bringing
explanation in Economics closer to the Models of Positive Science. Patterning
explanation in Economics on the lines of Positivist Models of Explanation has
been a major methodological obsession in Economics. But the subject matter of
Economics is not as easily amenable as physical science for application of rules
of Positivism. After going through this unit, you will be able to:
• state the efforts made by mainstream economists to adopt the model of
explanation of Positivism;
57
Research Methodology: • explain the limitations, difficulties and differences in adopting ‘scientific’
models of explanation; and
• appraise the present state of explanation in Economics.
4.1 INTRODUCTION
The urge to model Economics or Political Economy as a science on the lines of
physical sciences dates back to the period of Classical Economics. David Ricardo
claimed that the principles of Political Economy were as accurate as the laws of
gravity. Though Ricardo did not write much on the methods or models of
explanation, his theories of political economy reflected abstract-deductive
thinking which was claimed later as constituting basic methodology of Political
Economy. It is ‘Ricardo’s habit of thinking’ equated with hypothetico-deductive
model of explanation that set the tone for the classical writers on methodology
to carry forward the view of Political Economy as a pure science. We shall begin
by briefly examining the views on the nature of Political Economy and its Method
of Explanation in the Ricardian tradition as enunciated in the writings of
N.W.Senior, J.S.Mill, J.E.Cairnes and J.N.Keynes. Then, we shall turn to Lionel
Robbins who asserted that Economics was a pure positive science and emphasized
apriorism as its method of explanation. This will be followed by T. Hutchison’s
insistence on empiricism if Economics were to be called science at all. The next
two sections will discuss the foundations for present day mainstream practice of
Economics in the form of ‘instrumentalism’ and ‘operationalism’ in the writings
of Milton Friedman and Paul Samuelson respectively. Throughout, the focus is
on the limitations of these approaches and the ad-hoc means by which problems
are sought to be overcome. Though the unit is confined to the explanation in
Economics in the Positivist mainstream, towards the end, a brief discussion of
the alternative methods of explanation will be presented with a view to expose
the student to the potential of plurality of methods of explanation in Economics.
4.2 CLASSICAL POLITICAL ECONOMY AND

RICARDO’S METHOD
4.2.1 Ricardo’s Method
The right method of economic enquiry as such was not discussed by Adam Smith.
Smith’s approach to methodology was, as it was analysed by later economists,
somewhat enigmatic. There are instances which make some to point out that he
was the first to raise political economy to the state of deductive science. But in
the same breath there are many instances in his writings to as the founder of
inductive historical method. Yet his overall approach often tends to be moral and
ethical, distancing him from the attributes of any positivist explanation. For these
reasons, on issues of methodology of Classical political economy, one invariably
turns to David Ricardo.
Ricardo did not write on methodology of economics but his method explaining
the principles of political economy is clearly discerned as positive, abstract and
deductive, and it has come to be called ‘hypothetico-deductive model of
explanation’. It vigorously denies that facts can speak for themselves and
contends that it is theory that can make sense. Ricardo died in 1823, leaving
behind the claim that political economy was a science. His ‘Principles of Political
58
Economy’, which was a set of laws based on deductive reasoning, measures up Debates on Models of
Explanation in Economics
to that claim. Questions were raised soon whether a deductive body of laws
without empirical content could claim the status of science. The task of defending
Ricardo’s method and the claim of political economy as a science was left to his
followers of classical political economy. The defence came from the writings of
N.W.Senior, J.S.Mill, J.E.Cairnes and J.N.Keynes.
4.2.2 N.W. Senior

It was N.W.Senior in his 1927 lecture, which was later elaborated as outline of
the Science of Political Economy, (1936), that first made a statement on the
distinction between a pure and strictly positive science, and an impure and
inherently normative art of economics. He went on to claim scientific status for
the Ricardian system. Science of Political Economy, according to him, rests
essentially on “a very few general propositions, which are the result of observation,
or consciousness, and which almost every man, as soon as he hears them, admits,
as being familiar to his thoughts”. From this conclusions are drawn which hold
true only in the absence of “particular disturbing causes”.
Senior reduced these “very few general propositions” of political economy into
the following four:
1) that every person desires to maximize wealth with as little sacrifice as
possible;
2) that population tends to increase faster than the means of subsistence;
3) that labour working with machines is capable of producing a positive net
product; and
4) that agriculture is subject to diminishing returns.
4.2.3 J.S. Mill

It is senior, who for the first time introduces the concept of ‘economic man’ or
‘maximizing man’, as could be seen from the first proposition above. J.S.Mill,
though staunch advocate of inductive method and a stout defender of
methodological monism in science, was nonetheless sympathetic to Ricardo’s
deductive method of political economy. Mill, in his essay on the Definition of
Political Economy and on the Method of Investigation Proper to It (1936), treats
political economy as an “inexact science: and a ‘separate science:, and thereby
defends deductive method in economics. Elaborating on Senior’s notion of
‘economic man’, Mill observes “political economy proceeds….under the
supposition that man is a being who is determined, by the necessity of his nature,
to prefer a greater proportion of wealth to a smaller in all cases…” Mill does
consider “economic man” as fictional man to facilitate the propositions of political
economy. The pages on ‘economic man’ in Mill’s essay are followed immediately
by the characterization of political economy as “essentially an abstract science;
that employs “the method of a priori”. The “disturbing causes” and “the tendency
laws” in political economy necessitate ceteris paribus clauses.
4.2.4 J.E. Cairnes

J.E.Cairnes in his Character and Logical Method of Political Economy (1875),
goes on to argue that political economy is a hypothetical, deductive science, and
its conclusions “will correspond with facts only in the absence of disturbing
59
Research Methodology: causes, which is, in other words, to say that they represent not positive but
hypothetical truths”. By late 19th century there were growing doubts on the method
of political economy. J.N.Keynes tried to argue that what is referred to as deductive
reasoning in political economy was indeed preceded by inductive inferences
based on observation of human behaviour.
4.2.5 J. N. Keynes
J.N. Keynes, in The Scope and Method of Political Economy (1890), argued
that deductive reasoning was indeed preceded by concrete observations. Deductive
reasoning comes at a later stage drawing upon specific individual experiences.
J.N.Keynes is known for comprehensive reconciliation of the classical
methodological approaches.
The following observation stands as a testimony to his comprehensive

understanding of the methodological approaches underlying a complex subject
like economics. He observes “…economic science deals with phenomena that
are more complex and less uniform than those with which the natural sciences
are concerned; and its conclusions, except in their most abstract form, lack both
the certainty and the universality that pertain to physical laws. There is a
corresponding difficulty in regard to the proper method of economic study; and
the problem of defining preconditions and limits of the validity of economic
reasonings become one of exceptional complexity. It is, moreover, impossible to
establish the sight of any one method to hold the field to the exclusion of others.
Different methods are appropriate, according to the material available, the
stage of investigation reached, and the object in view; and hence arises the
special task of assigning to each its legitimate place and relative importance”.
One could read this even to this day, as a manifesto of methodological pluralism,
that is increasingly thought of as appropriate for economics. This aspect is further
discussed, when we turn to Sen on economic methodology.

1) Explain the characteristics of Ricardo’s method of explanation.
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) State the general prepositions advanced by N.W.Senior in support of scientific
status of the Ricardion system.
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) What is methodological pluralism?
......................................................................................................................
......................................................................................................................
......................................................................................................................
60 ......................................................................................................................
Debates on Models of
4.3 ROBBINS, POSITIVISM AND APRIORISM IN Explanation in Economics
ECONOMICS
Lional Robbins in his classic tract, An Essay on the Nature and Significance of
Economic Science (1935 ed.) presented the dominant methodological view point
of the early twentieth century economic thought, which stressed subjectivism,
methodological individualism, and the self-evident nature of the basic
postulates of economic theory. His objective was to counter the criticism from
the historical and institutional schools that basic economic propositions were
not amenable for Quantification, therefore not testable and hence not eligible to
be called a science. Robbins countered by suggesting that basic propositions of
Economics were deductions from a series of postulates. Chief of these postulates
include the assumption involving the experience of the way in which scarcity of
goods shows itself in the world of reality and how individuals order their
preferences. These are universally acknowledged facts of experience. Robbins
offers a coherent account of the status of terms used in economic theory. The
fundamental assumptions are combined with subsidiary hypotheses which allow
us to apply the theory to actual situations.
For Robbins rationality as a postulate in economics doesn’t merely mean

maximizing behaviour, but implies consistency of choice. Rationality as
consistency of choice is stated in the sense that if A is preferred to B and B to C,
A will be preferred to C. Robbins asserts that the assumption of rational conduct,
and with it the assumption of perfect foresight in Economics, are ‘expository
devices’. They are not realistic assumptions. According to Robbins, empirical
studies may be of use for the short-term prediction of possible trends, but they
do not provide the grounds for discovering ‘empirical laws’. The proper uses of
realistic (empirical) studies are three: to check on the applicability of theoretical
constructions to particular concrete situations, to suggest auxiliary postulates to
be used with the fundamental generalizations, and to bring to light areas where
pure theory can be reformulated or extended.
Robbin’s insistence on Economics being based on true postulates born out of

experience, which need not be tested, brings him closer to apriorism of the
Austrian School. Economic laws are derived from postulates based on a priori
truths of experience and therefore there is no need for testing them. Robbins’
claim that Economics is a pure theory and still a Positive science may be somewhat
puzzling. His claim of Positivism for Economics appears to be based more on
his insistence that Economic laws are value- free. He argued that there was no
room for ethical considerations in Economics, and in that sense too it is ‘pure
theory’.
Robbins’ claim of ‘pure theory’ or apriorism status to Economics provoked

devastating criticism, especially from T. Hutchison. The point of criticism is the
‘emptiness of the propositions of pure theory’. If Economics is to be called as a
positive science, Hutchison insisted that it should be testable. Otherwise it would
end up as a ‘pseudo-science’. Second, the assumption of ‘perfect expectations’
is baseless and therefore the rationality postulate is unrealistic. Third, there is
need for more extensive use of empirical techniques in Economics instead of
merely claiming all postulates are prior truths. Finally, Hutchison calls the use of
‘psychological method’ (or introspection) as ground for asserting the fundamental
postulates. 61
Issues and Perspectives 4.4 HUTCHISON AND LOGICAL EMPIRICISM IN
ECONOMICS
Terence Hutchison’s The Significance and Basic Postulates of Economic Theory
(1938) marks a clear turning point in the Twentieth Century Economic
methodology and laid the foundation for a clear logical positivist turn to the
practice of Economics. Hutchison, in his book, sets out on search for and making
clear the foundations of modern economic theory. He states that economics is a
science and as such it must appeal to facts, otherwise one would be engaging in
‘pseudo-science’. What sets apart the empirical propositions of science from
those of other intellectual endeavours is their testability.
According to Hutchison, science contains statements which are either conceivably

falsifiable by empirical observation or are not. Those which are not falsifiable
are tautologies and are thus devoid of empirical content. The primary reason
why the propositions of pure theory have no empirical content is that they are
posed in the form of deductive inferences. In Economics, widespread use of
ceteris paribus clause robs empirical content. Irresponsible use of ceteris paribus
clause make microeconomic theory effectively unfalsifiable.
Hutchison examines the formal structure of economic theory which consists of a

series of deductions from basic postulates. These deductions are analytical in
nature. These analytical statements are logical statements without any empirical
content. Therefore, pure economic theory is empty and thus there is need for
more empirical content. Hutchison insisted on the need for testability, and
falsification tests even for the basic assumptions. The insistence of testing not
only the theory, but the assumptions as well, for validity sparked-off a wider
debate in economics later. But the immediate response was to describe Hutchison
as an ‘ultra-empiricist’.
Hutchison’s attempt to establish the analyticity of fundamental generalizations

of economic theory and to wean economists away from pure theory did not
succeed very much. But his proposal to make economics increasingly testable or
falsifiable form; to increase empirical investigations of various aspects of
economics, and the need for economists to abandon their psychological method
received universal acceptance by economists.
One of the interesting aspects of Hutchison’s contribution relates to the issue of

testing assumptions as well as the theory or hypothesis. This provoked a debate
between Hutchison and Fritz Machlup. Machlup believes that only deduced or
‘lower level’ hypotheses require ‘verification’, and he outlines how such testing
could be carried out. While Hutchison does not require that every statement in a
theory be tested, he does insist that each be ‘testable’ and one should be able to
conceive how a test could be carried out. Further, Hutchison prefers that the
behavioural postulates of economics reflect the actual observed and statistically
recorded behaviour of economic agents. Machlup requires no such
correspondence. The crucial behavioural postulate is the assumption of rational,
maximizing behaviour. And Machlup is more reasonable in arguing that this
assumption need not be tested directly, which was insisted by Hutchison, but
indirect testing of the same could be done. Much before the debate on testing of
assumptions sparked-off by Hutchison’s contribution, Milton Friedman’s
62
independent contribution suggested that testing of assumptions as such was not Debates on Models of
an important issue. However, such a proposition did lead to much controversy as
we shall see in the next section.
4.5 MILTON FRIEDMAN AND

INSTRUMENTALISM IN ECONOMICS
Milton Friedman’s “The Methodology of Positive Economics”, the lead article
in his book Essays in Positive Economics (1953), is among the best-known pieces
of methodological writings in economics. For over two decades the
methodological prescriptions advanced in his essay became widely accepted
among many working economists, in spite of wide controversy it generated.
Rightly, as observed by Caldwell, it was a remarkable masterpiece of marketing
certain notions of explanation as the basis of the methodology of positive
economics. In what follows, we shall discuss in brief Friedman’s contribution
including some of the major points of criticism. Friedman sets out by stating that
the purpose of his essay is to counter criticism of two pillars of neo-classical,
economics viz.,1) the maximization behaviour and 2) The model of perfect
competition. He goes on to establish the methodological foundations of neo-
classical economics by posting his own version of positivism which later came
to be branded as “instrumentalism”.
4.5.1 Goal of Science

According to Milton Friedman, the ultimate goal of positive science “is the
development of a ‘theory’ or ‘hypothesis’ that yields valid and meaningful
…..predictions about the phenomenon”. The criteria for acceptability of theory
or hypothesis are three:
1) that theory / hypothesis be logically consistent and contain categories

which have meaningful empirical counterparts;
2) that there are ‘substantive hypotheses’ which are testable; and
3) that “the only relevant test of the validity of a hypothesis is comparison of

its predictions with experience”.
Further, when an infinite number of hypotheses generally consistent with an

observed set of facts generally exist, other criteria-simplicity and fruitfulness-be
invoked to chose among competing hypotheses.
4.5.2 Relationship Between Significance of Theory and

‘Realism’ of Assumptions
Friedman is critical of those who instead of testing a theory for predictability, try
to evaluate theory by the ‘realism of their assumptions? For Friedman, disparaging
a theory for having ‘unrealistic assumptions’ is ridiculous. He tries to show that
most significant theories are actually characterized by descriptively inaccurate
assumptions. He observes: “Truly important and significant hypotheses will be
found to have “assumptions” that are widely inaccurate, descriptive
representations of reality, and, in general, the more significant the theory, the
more unrealistic the assumptions”. (emphasis added).
63
Research Methodology: Realism of assumptions does not matter except when certain hypotheses are
derived from assumptions, and when there is no test available, realist description
is in order. With these two exceptions, realism of assumptions do not matter for
the following reasons:
1) Theory is supposed to lead to simplification of complex reality by abstraction.

There is inverse relationship between realism and abstraction. The more
realism one insists upon, the more abstract and the complex the explanation.
2) The uninformativeness of knowledge of discrepancies between assumptions

and facts, makes one to treat realism of assumptions as of no importance.
He cites the example of Galileo’s Law of Falling Bodies:
“If a body falls toward the earth in vacuum, its instantaneous acceleration is
constant, s=½gt2. The realism of vacuum need not be tested. Since the
prediction is true, then the assumption be treated as if true. Turning to
economic theory, Friedman cites the “Billiard Ball Player” example and
where once the ball reaches the pouch one could assume as if the player
knew all the laws of time and motion and extends it to profit maximization:
‘under behave as if they were seeking rationally to maximize their expected
returns……and have full knowledge of the data needed to succeed in this
attempt”.
3) The presence of ‘undesigned’ classes of implications can be avoided by as

if assumption.
Milton Friedman concludes his essay by asserting unity of method in all positive
sciences. For him there are no methodological differences between natural and
social sciences. He regards economics as “objective” as physical sciences. His
main conclusions are:
1) that realism of assumptions is “largely” irrelevant to validation of theories,

which ought to be judged ‘almost’ solely in terms of their instrumental value
in generating accurate predictions.
2) standard economic theory has an excellent predictive record as judged by

countless application to specific problems.
3) the dynamics of competition over time accounts for this splendid track record.
4.5.3 Critical Appraisal of Friedman’s Instrumentalism

Milton Friedman’s methodological approach is criticized widely. The major
criticism relates to the proposition that realism of assumptions does not matter.
Friedman’s assertion that not only the realism of assumptions is necessary but
“in general, the more significant the theory, the more unrealistic the assumptions”
is called by Samuelson as a flamboyant exaggeration, as the extreme version of
“F-Twist” (meaning Friedman-Twist). It is one thing to say that proving realism
of all assumptions is difficult, but altogether different to say that further the
assumptions move away from realism, closer is the theory to predictions. Even
Fritz Machlup, L. Qsoland and Nagel, who were generally in defence of
Friedman’s methodology, would insist on the realism of at least second order
assumptions. It is pointed out that direct evidence about assumptions is not
64
necessarily more difficult than testing a theory. Test assumptions may yield Debates on Models of
important insights into the theory. Caldwill points out that philosophically,
Friedman’s insistence on prediction not explanation may result in correlation
not causation, i.e., it may result in “measurement without theory”. Blaug points
out that accurate predictions are not the only relevant test of the validity of a
theory and it would be impossible to distinguish between genuine and spurious
correlations. Further, Friedman’s instrumental attitude to theories, ignores ‘truth
value’ of theories as many philosophers of science would insist upon.

1) Identify the important features of methodological view point of L. Rabins.
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) What is the important contribution of Hutchison in the area of logical
empiricism?
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) Which criteria have been laid down by Milton Friedman for acceptability of
theory or hypothesis?
......................................................................................................................
......................................................................................................................
......................................................................................................................
4) Identify the basis of the major criticism against Friedman’s instrumentalism
proposition.
......................................................................................................................
......................................................................................................................
......................................................................................................................
4.6 PAUL SAMUELSON AND OPERATIONALISM

Paul Samuelson’s methodological contribution, though not comparable to his
substantial work in economic theory, does offer certain insights into the turn of
economics towards logical empiricism. Samuelson is against a priorism. He is
more akin to logical empiricism of Hutchison’s kind. He does show the influence
of Popper’s variety of ‘rational reconstruction’. His contribution to economic
methodology could be seen in terms of two theses, which he advances. One has
come to be described as ‘operationalism’ and the other as ‘descriptivism’. We
shall elaborate these two theses.
65
Research Methodology: 4.6.1 Operationalism
According to him, economists should seek to discover ‘operationally meaningful
theorems’. Theories are strategically simplified description of observable and
refutable empirical regularities. He begins with the criticism of Friedman for the
latter’s F-Twist on realism of assumptions and emphasizes the need for
methodological clarity. He proposes the thesis of “logical equivalence” as the
basis for methodological consistency. According to “logical equivalence:, theories
are merely equivalent restatements of assumptions and conclusions, i.e., A=B=C,
where A is defined as ‘assumptions’, B is defined as theory and C as consequences
or predictions.
Theory (B) consists of “a set of axioms, postulates or hypotheses that stipulate

something about observable reality… “ The set is either confirmable or refutable
in principle by observation. A theory has a set of consequences (C) which are
logically implied by theory, and a set assumptions (A) which logically implies
the theory. The degree of “realism”, “factual conrrectness”, “empirical validity”
or “truth” of any one of A,B,C is shared by the other two. Referring to ‘F-Twist’,
he observes first, it is a contradiction to maintain all consequences (C) are valid
and the theory (B) and the assumptions (A) are not valid. Second, it is absurd to
maintain, in case where only some of the consequences (C) are valid, that the
theory (B) and assumptions (A) are important though invalid. The part of the
theory set and assumptions set corresponding to the invalid part of the
consequence set should be eliminated.
4.6.2 Descriptivism
Samuelson has certain exalted view of explanation, which he considers different
from the usual notion in science. He observed, “scientists never” explain” any
behaviour, by theory or any other hook. Every description that is superseded by
a “deeper explanation” turns out …to have been replaced by still another
description ….” An explanation, as used legitimately in science, is a better kind
of description and not something that goes ultimately beyond description. It is
this emphasis and elaboration of ‘description’ that earns Samuelson’s approach
the title of ‘descriptivism’. Why only description? First, a theory is just description
of observable experience, a convenient mnemonic representation of empirical
reality. Second, knowledge consists essentially of observational reports. A theory
expressible in observational language is superior to those which are not.
Explanations are ultimate. Apriorisms must be avoided; hence theories should
be expressed in observational language. All known theories in science are
expressible in terms of observational statements, i.e., basic statements.
Samuelson is criticized for not practicing what he preaches methodologically.

Machlup, citing Samuelson’s famous work on ‘factor price equalisation’, accuses
him for not following his own norm of deriving “operationally meaningful
theorem” based on unrealistic assumptions. Another famous criticism is that of
Stanley Wong’s against the methodological foundations of Samulson’s revealed
preference theory.
66
Debates on Models of
4.7 THEORY – ASSUMPTIONS DEBATE IN Explanation in Economics
ECONOMICS: A LONG VIEW

The spectacular advance of modern science is usually attributed very largely to
the development of deductively related statements known as theories. Theories
in science brought improvements in order and clarity, broadened the scope of
generalizations and scored extraordinary predictive success. However, economic
theory does not command as much respect as theories in physical sciences. Ever
since the classical economics, there has been criticism of unrealistic assumptions
of economic theory. There have been reasoned replies from time to time. Jack
Melitz (1965), in one of his classic papers on the theory – assumptions controversy,
provides an account of the response to the criticism since late Twentieth century.
Here is a brief summary, which helps us to understand the problems of explanation
that persist in economics.
During 1880-1920 advocates of economic theory adopted a moderate and

conciliatory stance. They agreed that economic theory made false assumptions,
and admitted that the value of the theory depended greatly on the degree of
accord between the assumptions and the facts. Yet they insisted, first, that the
assumptions did correspond broadly with events. Second sacrifice of some
accuracy for simplicity was justified in view of complexity of reality. Further,
they emphasized importance of combining the use of simplifying assumptions
with protective measures:
1) The need to pursue more inductive studies in all areas of economics.
2) The determination of reasonable proximity between major assumptions and
facts before application of economic theory.
3) The alteration of assumptions to suit the particular case involved.
During 1920s and 1930s with notable advance of purely logical branch of
economics, the advocates of economic theory lost appeal. The practitioners of
economic theory with the support of Robbins and the Austrians turned to a
priorism. There was a stance that economic theory is its own reward.
In 1948-1953, as we have seen earlier, Milton Friedman tried to supply a logical

foundation for the developing attitude that the realism of assumptions is not a
genuine, or only a secondary concern. False assumptions in economics do not
result in a handicap. Assumptions must simply work, and yield reliable results.
The criticism that assumptions in economics do not correspond with facts is
besides the point. In 1955 Firtz Machlup joined forces with Friedman, claiming
to bring with him the support of experts in the philosophy of science and the
whole tradition of political economy. The result is that economics continues to
live with a certain methodological inadequacy. The other reasons for this disquiet
in explanation in economics are summed up in the next section.
4.8 AMARTYA SEN ON HETEROGENEITY OF

EXPLANATION IN ECONOMICS
By now it is clear that explanation in economics is nowhere near being in a
satisfactory state. In a broad based critique of contemporary economic
67
Research Methodology: methodology, Amartya Sen (1989) draws attention to the heterogeneity of the
subject matter of economics. Any attempt to think of mono-method for all the
diverse concerns of economics is bound to cause the kind of disquiet that we
experience today. He cogently argues for heterogeneity of methodological
approaches in economics.
4.8.1 Heterogeneity of Substance and Methods of Economics

According to Sen, economics as a subject is concerned with many different types
of problems. The diversity of the discipline of economics should be kept in view
to achieve an adequate grip on the methodological issues in the subject. The
subject of economics includes three diverse, though interrelated, exercises:
1) Predicting the future and causally explaining the past events.
2) Choosing appropriate descriptions of states and events in the past and the
present, and
3) Providing normative evaluations of states, institutions, and policies.
These exercises are interrelated but each requires a different methodological
approach. For instance, the method of scientific explanation that insists on
prediction is concerned only with the first set of exercises. The ‘methodology of
economics’ has to admit enough diversity to be able to deal with other classes of
problems as well.
4.8.2 Testing and Verification

Testing and Verification are important for many types of economic analyses since
they are concerned with causal relationships and with making predictions. But
these are not suitable for all economic theories. For instance evaluative exercises
are not open to testing. Normative evaluation is a different discipline from that
of making predictions on the basis of causal hypotheses. Similarly, some
descriptive propositions do not have predictive content, and “testing” would be
the wrong operation to seek. As far as causal theories are concerned, the need for
testing them with empirical information is fairly accepted in principle by
economists. Conceptual and analytical issues need to be explored very
substantially to understand what types of relationships might be involved.
Analyses at this stage are not meant for testing and verification. Of course, one
should not end at this stage and move up to the next where testing is possible.
4.8.3 Value Judgments and Welfare Economics

Sen draws attention to the fact that following Robbins value judgments are kept
out of economics. “The decision to keep economics ‘value-free’ would, of course,
militate against the subject of welfare economics as such”. Welfare economics
still is accepted as important and in this domain to keep economics value-free
may not be a value that would be appreciated. In evaluative exercises in welfare
economics descriptive methods becomes indispensable.
4.8.4 Formalisation and Mathematics

Mathematics helps to an extent in formal reasoning and helped in systematization
of many economic propositions. However, there are severe limitations of formal
language of mathematics. Not all economic propositions can be reduced into
mathematics. Lack of balance often has resulted in certain degree of over-
68
formalization of economics. While recognizing the positive contribution of Debates on Models of
mathematics in lending rigour to certain economic propositions, the negative
contribution is through over concentration on mathematics to the exclusion of
other disciplines. Such excesses of formalization need to be corrected.

1) What do you mean by the term ‘logical equivalence’?
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Do you agree that different methodological approaches are needed in
Economics?
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) To what extent mathematics should be allowed to use in systematization
of economic proposition?
......................................................................................................................
......................................................................................................................
......................................................................................................................
4.9 LET US SUM UP

Keeping in view that the subject matter of economics is complex and
heterogeneous, any single model of explanation is bound to face serious
limitations. Yet there has been an ambition in the profession from the times of
classical political economy, which has turned towards a kind of arrogance that
economics is a science like any physical sciences and should conform to the rule
of one methodological approach, viz., logical positivism. This has been subjected
to severe criticism on its limitations.
The mainstream economics is not able to come out of this obsession “It is about
the one methodological rule which has dominated economics since early 1960s.
The rule at issue is the methodological requirement that economic models or
theories, if they are going to be given serious consideration by practicing
economists, must be shown to be testable, where successful test of theory is
defined as a falsification of that theory. A testable theory is a falsifiable theory”.
The exploratory lesson on explanation that we had should help a student of

economics to move out of this obsession. It is not that abandon testing but we
must abandon the habit of considering evaluation, description, and even advocacy
as not being part of pursuing knowledge in economics. Heterogeneity of
economics needs diversity of approaches if the disquiet of disaffection that
pervades economics is to be overcome.
69
Issues and Perspectives 4.10 KEY WORDS
Determinism : A term used to describe an argument or
methodology that simply reduces casuality to a
single set of factors acting more or less directly
to produce outcomes.
Hypothesis : Hypothesis refers to a tentative statement that can

be tested by applying the methods of particular
science.
Empiricism : Empiricism is generally regarded as being at the

heart of the modern scientific method that our
theories should be based on our observations of
the world rather than on intuition or faith; that is,
empirical research and a posteriori inductive
reasoning rather than purely deductive logic.
Ideology : A systematic body of concepts about human life

or culture; a manner or the content of thinking
characteristic of an individual or group or culture;
also refers to the integrated assertion and theories
that constitute a socio-political programme.
Methodological Dualism : Referring to positivist belief in separation of the

knower from the known or of the subject from
object.
Methodological Monism : In contrast to the compartmentalization of

dualism, monism views the world as a “seamless
web”. In terms of Gouldner’s argument, the
separation between the knower and the known
must be overcome.
Nomothetic : Relating to discovery of general laws or relating

to the discovery of universal law.
Pseudo Science : Pseudo science refers to any body of knowledge

or practice, which purports to be scientific or
supported by science but which is judged to fall
outside the domain of science.

Blaug, Mark (1980); Methodology of Economics, Cambridge Univeristy Press,
Cambridge.
Caldwell, Bruce (1982); Beyond Positivism: Economic Methodology in the

Twentieth Centrury, George Allen and Unwin, London.
Friedman, Milton (1953); Essays in Positive Economics, University of Chicago

Press, Chicago.
70
Hausman, Daniel M. (1994); The Philosophy of Economics: An Anthology, Debates on Models of
Cambridge University Press, Cambridge, 2nd Edition.
Melitz, Jack (1965); “Friedman and Machlup on the Significance of Testing

Economic Assumptions”, Journal of Political Economy, pp.37-60.
Sen, Amartya (1989); “Economic Methodology: Heterogeneity and Relevance”,

Social Research, Vol.56, No.2 (Summer 1989), pp. 299-329.

PROGRESS EXERCISES
1) Positive, abstract and deductive briefly called hyupothetico-deductive

model of explanation.
3) Methodological pluralism refers to comprehensive reconciliation of different

methods and acceptance of the idea that different methods are appropriate
depending upon the object in view, the stage of investigation reached and
the material available.
1) Stress on pure theory bases on a priori truth of experience and without any
need for testing, no room for ethical consideration, rationality as maximizing
behaviour and as consistency of choice.
2) The insistence of testing both theory as well as assumptions.
4) The proposition that realism of assumption does not matter.
1) Theories are merely equivalent restatements of assumptions and conclusions.
2) See Section 4.8
71
Issues and Perspectives UNIT 5 FOUNDATIONS OF QUALITATIVE
RESEARCH: INTERPRETATIVISIM
AND CRITICAL THEORY PARADIGM
Structure
5.0 Objectives
5.1 Introduction
5.2 Interpretive Paradigm
5.3 Critical Theory Paradigm
5.4 Applications in Research: Illustrative Cases
5.5 Let Us Sum Up
5.6 Exercises
5.7 Key Words
5.0 OBJECTIVES
This unit introduces the two widely used and well-established paradigms namely,
interpretive and critical theory. Unlike post-positivism, these paradigms are the
pillars of qualitative research. After going through this unit, you will be able to:
• describe the essence of interpretive and critical theory paradigm;
• discuss their theoretical framework;
• elucidate the nature of reality and methodology underlying these two
paradigms; and
• state the examples and applications of interpretive and critical theory
paradigm in economics.
5.1 INTRODUCTION
As has been discussed earlier in Unit 1, you have come to know that a paradigm
essentially refers to a comprehensive belief system or theoretical framework/
world view that guides research and practice in a particular field. At its basic
level, it has a philosophy of science that makes a number of fundamental
assumptions about the nature of truth and what it means to know. Unlike
positivism, that is based on empiricism and deductive reasoning for verification
of truth, interpretive and critical theory paradigms are based on subjectivism
and substantialism for the search of truth. The underlying ontological and
epistemological assumptions of interpretitivism and critical theory paradigms
have been discussed in the subsequent sections. It is important to note that these
two paradigms form the backbone of philosophical science of qualitative research.
72
Before embarking on meaningful research using qualitative approach, it is Foundations of Qualitative
Research: Interpretativism
important to understand the meaning, nature of reality, methodology and the and Critical Theory
essence of these two paradigms. Let us begin with interpretative paradigm. Paradigm
5.2 INTERPRETIVE PARADIGM

Interpretive research assumes “that our knowledge of reality is gained only through
social constructions such as language, consciousness, shared meanings,
documents, tools, and other artifacts” (Klein & Myers, p. 69). As the term
indicates, interpretive paradigm looks for understanding of a particular context.
Interpretivists believe that it is important to understand the context in which
research is conducted for proper interpretation of the data. The interest of
interpretivists lies not in the generation of a new theory, but to judge or evaluate
and refine interpretive theories. Researchers using an interpretive approach aim
to uncover meaning towards a better understanding of the issues involved. The
underlying ontological assumption of interpretive paradigm is subjectivism as
here reality is viewed as socially constructed and interpreted. Epistemological
assumption is that knowledge of reality is obtained from the accounts that social
actors provide. Neuman (1997) affirms that “social reality is based on people’s
definition of it” (p. 69). From the previous assertions, it is apparent that interpretive
researchers do not recognize the existence of an objective world. On the contrary,
they see the world strongly bounded by particular time and specific context.
Therefore, the epistemological question, “What is the nature of the relationship
between the knower or would-be knower and what can be known” (Guba &
Lincoln, 1994, p. 108) must be answered in a consistent way with the ontological
view. The interpretive researcher’s epistemological assumption is that “findings
are literally created as the investigation proceeds” (Guba & Lincoln, p. 111).
Moreover, they explicitly recognise that “understanding social reality requires
understanding how practices and meanings are formed and informed by the
language and tacit norms shared by humans working towards some shared goal”
(Orlikowski & Baroudi, 1991, p. 14).
Interpretive research focuses on identifying, documenting, and ‘knowing’ –

through interpretation of :
– world views,
– values,
– meanings,
– beliefs,
– thoughts and the general characteristics of life events, situations, ceremonies
and specific phenomena under investigation,
– the goal being to document and interpret as fully as possible the totality of
whatever is being studied in particular contexts from the people’s viewpoint
or frame of reference.
Interpretivists assert that all research is influenced and shaped by pre-existing

theories and world-views of the researchers. The terms, procedures and data of
research have meaning because a group of scholars has agreed on that meaning.
73
Research Methodology: Research is thus a socially constructed activity. Three other approaches support
the philosophy of interpretive paradigm. They are Verstehen, hermeneutics and
phenomenology. Verstehen stresses on the understanding of the particulars of a
situation, hermeneutics emphasizes on the importance of language and context
in understanding and phenomenology on people’s perception of the world.
Methodology
Foundationalism is an approach that asserts research can begin with self-evident
truths which can serve as the starting point for our understanding of the world.
Interpretivists are anti-foundationalists as they believe that the standards that
guide research are products of a particular group or culture. They use a broad
range of qualitative methods and gather thoughtful reflections of experienced
practitioners. Interpretative approaches rely heavily on naturalistic methods
(interviewing and observation) and their methods generally ensure an adequate
dialogue between the researchers and the people with whom they interact to be
able to construct a meaningful reality. These meanings emerge from the research
process. Certain methodologies used in interpretive research are:
Naturalistic inquiry, that uses first-hand observation to understand human action

and studies real-life situations as they unfold. It is non-manipulative and non-
controlling, hence has lack of predetermined constraints on outcomes;
Phenomenologic methodologies, rely on descriptions of conscious experiences
to develop understanding of the meaning of human action;
Constructivism, makes use of perception or self-experience in making and
structuring knowledge;
Ethnographic inquiry, rely on self-experiencing the culture of participants in the
field;
Symbolic interactionism, understands human action as per the meanings derived
by the human beings for particular objects and people.
Some other methods for data-collection are surveys, interviews, field observation,
witness accounts, focus group discussion. Interpretive research adds to the
understanding of different contexts and situations. Interpretivists argue that
research results should be applied to higher, conceptual level. From the data, a
researcher tries to understand multiple perspectives on the same topic. For
analyzing the meaning of data, interpretive researchers may use and conduct
research by methods from another paradigm (eg. Post-positivism) also and doing
so would lead to reinterpreting the meaning of results by another perspective.
With a philosophical alignment with interpretive naturalistic orientations,

interpretive description acknowledges the constructed and contextual nature of
human experience that at the same time allows for shared realities (Thorne, Reimer
Kirkham, & MacDonald-Emes, 1997). Key axioms of naturalistic inquiry, such
as those delineated by Lincoln and Guba (1985), provide philosophical
underpinnings for research design, including:
1) There are multiple constructed realities that can be studied only holistically.
Thus, reality is complex, contextual, constructed, and ultimately subjective.
74
2) The inquirer and the “object” of inquiry interact to influence one another; Foundations of Qualitative
indeed, the knower and known are inseparable. and Critical Theory
Paradigm
3) No a priori theory could possibly encompass the multiple realities that are
likely to be encountered; rather, theory must emerge or be grounded in the
data.
Research Strategy for Interpretivism

Broadly there are four distinct research strategies that work with different
ontological assumptions. Abduction strategy best fulfills the needs of interpretive
researcher as it starts with laying the concepts and meanings that are contained
in social quarter’s accounts of activities related to a research problem.
As interpretivists construct reality on the accounts of social actors present in a

specific situation, it is important to take into account the different perspectives
of each participant. One can understand the complexity of the issue through the
following example, excerpts taken from Andrade, 2009.
Let’s imagine a scenario at the beach in which a huge wave is approaching the
shore. There is an excited surfer on top of the big wave and two scared children
in a small inflatable boat right below the colossal wave. On the shore, a girl is
admiring her boyfriend’s dexterity and the petrified children’s mother is watching
the looming mass of water approaching the boat. On the adjacent cliff there is a
relaxed monk meditating on the infiniteness of the universe, while enjoying the
sea breeze and the sound of the sea. If we want to conduct research on what that
wave means for beach-goers, our results will depend on who the respondent is.
Interviewing one of the participants would give insights from that participant’s
perspective only, which may be insufficient, or even misleading, because their
personal and intimate experiences with the wave are quite different from that of
the others. If the interpretive researcher wants to create an integral and persuasive
piece of research around this phenomenon, each participant’s different
perspectives should be included.
To conduct interpretive research on a certain setting, intense and long-term

participant observation is required, followed by deliberate and long-term reflection
on what was observed. Interpretive researchers start out with the assumption
that access to reality (given or socially constructed) is only through social
constructions such as language, consciousness and shared meanings. Interpretive
studies generally attempt to understand phenomena through the meanings that
people assign to them and interpretive methods of research in IS are “aimed at
producing an understanding of the context of the information system, and the
process whereby the information system influences and is influenced by the
context. Interpretive research does not predefine dependent and independent
variables, but focuses on the full complexity of human sense making as the
situation emerges.
The life-cycle of theory-generation under interpretive paradigm has been depicted

in the following flow-chart:
75
Issues and Perspectives Review relevant theory (literature)
Define research questions
Determine suitable research methods and site(s)
Experimental, observational, action research or case studies
Analysis using qualitative or quantitative methods
Theory suggestion, confirmation, constrains or extension
Publish Conclusions
Criticism
Researchers who wish to use interpretive framework for their case-studies are
cautioned not to lose theoretical sensitivity. Strauss and Corbin (1990, p. 41)
describe theoretical sensitivity as the “awareness of the subtleties of meaning of
data” and elaborate that “one can come to a research situation with varying degrees
of sensitivity depending upon previous reading and experience with or relevant
to that area.” Ultimately, the researcher has to evaluate the relevance of their
preliminary theoretical framework vis-à-vis the actual findings (Urquhart, 2001,
2007). The criteria to be kept in mind for using interpretive case studies aiming
for theory-building can be seen in the appendix to the unit.
Differences of Interpretive paradigm from post-positivism can be broadly
summarized in the table below:
Differences between Post-positivism and Interpretivism on Five major Issues
Post-positivism Interpretivism
Nature of reality External to human mind Socially constructed
Purpose of research Find Universals Reflect understanding
Acceptable methods Scientific method Subjective and Objective
and data research methods are acceptable
76
Foundations of Qualitative
Post-positivism Interpretivism Research: Interpretativism
and Critical Theory
Meaning of data Falsification Understanding is contextual. Paradigm
Use to test theory Universals are deemphasized
Relationship of Separate activities. Integrated activities.
research to practice Research guides practice Both guide and become the
other.
(Source: Foundations of qualitative research by Jerry W. Willis (2007) p.95)

1) Explain how interpretive framework is a departure from post-positivism?
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Outline the methodology of interpretive research using a suitable example
from your locality.
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) How do the ontological and epistemological assumptions underlying
interpretive paradigm guide the researcher to the methodology of research,
data-collection and analysis?
......................................................................................................................
......................................................................................................................
......................................................................................................................
5.3 CRITICAL THEORY PARADIGM

Critical theories share some ideas of the interpretative paradigm, but what makes
it different is that critical paradigm focuses on oppression. Critical social scientists
believe it necessary to understand the lived experience of real people in context.
Persons can perceive reality outside them and represent that reality with language.
Also, reality is defined by the interaction between the knower and the known.
Critical approaches examine social conditions and uncover oppressive power
arrangements. ‘Critical theory’ is a term that can apply to a number of movements
in the social sciences. As it is based on ideology of class conflict, the critical
theorists whose objective is to find out power-relationships among the different
sections of the society (between the doers and the oppressed), it is sometimes
referred as ‘ideologically oriented inquiry, neo-Marxism, materialism, the
Frankfurt School and freireism. Critical theory was built on the foundation of
Marxism which perceived conflicts between classes in society and believed that
it could be reformed only via radical transformation of the society. The roots of
the critical theory can be traced to the Frankfurt school or the Institute for Social
77
Research Methodology: Research. It was established as a school of thought primarily by five Frankfurt
school theoreticians; Herbert Marcuse, Theodor Adorno, Max Horkheimer, Walter
Benjamin and Erich Fromm. The underlying ontological assumption of critical
theory paradigm is substantialism wherein matter constitutes the reality. However,
epistemologically people interpret reality differently in different times and places.
Critical Theory is a theoretical tradition developed most notably by Horkeimer,
Adorno, Marcuse at the Frankfort School. Their work is a critical response to
the works of Marx, Kant, Hegel and Weber.
• Historical ontology – assumes that there is a ‘reality’ that is apprehendable.
This is a reality created and shaped by social, political, cultural, economic,
ethnic and gender-based forces that have been reified or crystallized over
time into social structures that are taken to be natural or real. People,
including researchers, function under the assumption that for all practical
purposes these structures are real. Critical theorist believe this assumption
is inappropriate.
• Modified transactional or subjectivist epistemology – we cannot separate
ourselves from what we know and this inevitably influences inquiry. What
can be known is inextricably tied to the interaction between a particular
investigator and a particular object or group.
Purpose of Research
Research that aspires to be critical seeks, as its purpose of inquiry, to confront
injustices in society. Following a tradition associated with Antonio Gramsci,
critical researchers aim to understand the relationship between societal structures
(especially those economic and political) and ideological patterns of thought
that constrain the human imagination and thus limit opportunities for confronting
and changing unjust social systems. Critical theorists are committed to
understanding the complexity of such relations, however, and thus distance
themselves from what they see as reductionist Marxist approaches. Critical
theorists hold that these earlier approaches offered no ability to explain social
change. Thus, in contrast to what they believe was an overemphasis on the
determinative nature of economic and political structures, critical theorists are
interested in social change as it occurs in relation to social struggle. Critical
researchers assume that the knowledge developed in their research may serve as
a first step toward addressing such injustices. As an approach with a definite
normative dimension, the research aims for a transformative outcome, and thus
is not interested in “knowledge for knowledge’s sake.”
Critical Theory is multi-disciplinary. It finds its applications in anthropology,
economics, art criticism, education, history, psychology, political science,
sociology and theology. An interesting question is why and where to apply critical
theory paradigm in research. To begin with, research and practice are integrated
activities in the critical paradigm. It not only studies power relationships which
are critical factors in the society, the injustices and the inequities, the contradictions
and the incoherencies, but also aims at helping and empowering those who are
oppressed to free themselves from their oppression.
Critical theory paradigm should be applied where the objective of the researcher
is to analyse the power structures in a set-up or social problems arising of such
structures. For example, it can be used to understand wage inequities between
male, female and child workers in the society.
78
Nature of reality: Critical theorists like post-positivists believe that reality is Foundations of Qualitative
material and external to human mind but they interpret it differently. They and Critical Theory
question/critique the existing reality and analyze why such a reality exists and Paradigm
whether it contributes positively/negatively to human mind. Further, they also
advocate evolving ways and means to reform the existing system if it is bifurcating
the society into the have’s and the have not’s.
Methodology: Critical theorists suggest two kinds of research methodologies,

namely ideology critique and action research, for undertaking research work.
They rely on methods combining observation and interviewing with approaches
that foster conversation and reflection. They try to challenge guiding assumptions
and they begin this by asking people to reflect and question their current
experience with regard to values identified. In doing research, they not only try
to define a situation but change the situation.
As a research methodology, critical theory adopts an overtly critical approach to

inquiry. It precedes with an attitude of suspicion, calling into question not only
the data itself, but also the researcher, the research design and interpretation of
findings. From a critical theory viewpoint, the task of the social scientist has
three dimensions:
1) To understand the ideologically distorted subjective situation of some
individual or group;
2) To explore the forces that has caused that situation;
3) To show that these forces can be overcome through awareness of them on
the part of the oppressed individual or group in question.
The appropriate research strategy for critical research is Retroduction as it begins
with a hypothetical mode of a mechanism that could explain the occurrence of a
phenomenon under investigation.
• Critical theoretical approaches tend to rely on dialogic methods; methods

combining observation and interviewing with approaches that foster
conversation and reflection. This reflective dialogic allows the researcher
and the participants to question the ‘natural’ state and challenge the
mechanisms for order maintenance. This is a way to reclaim conflict and
tension.
• Rather than naming and describing, the critical theorists try to challenge
guiding assumptions.
• Critical theorists usually do this by beginning with an assumption about

what is good (e.g. autonomy, democracy) and asking people in a social group,
culture or organization to reflect on and question their current experience
with regard to the values identified (e.g. To what extent are they an
autonomous worker?)
• Critical theorists not only just try to describe a situation from a particular
vantage point or set of values (e.g. the need for greater autonomy or
democracy in a particular setting), but also try to change the situation.
79
Research Methodology: “Objective” analysis
In their embrace of a normative perspective, Critical theorists make no claims
that their analyses are “objective” in the sense usually meant by logical positivists.
In fact, critical theorists argue that the subjective/objective dualism masks the
ways in which both positions are limited by the social forces that inform all
human action and analysis. Critical qualitative research acknowledges
subjectivism in the sense that learnings and interpretations cannot be based on
logic and scientific analysis only. While it affirms that knowledge can never be
separated completely from the researcher’s own experience, it rejects the notion
that all analyses are relative. It asserts that rational analysis is fundamental to
human emancipation, and hence embraces what Morrow (1994) calls critical
realism
Data analysis and verification

Critical researchers assume that their task is to expose the hidden assumptions
that guide both research respondent statements and often, initial analyses of data.
Researchers therefore bring a level of scrutiny to their task that includes rooting
out the meanings of what is left unsaid as well as that which is stated. The research
is verified as other members of the research community offer corroboration that
has come from their own research experiences.
Sample representativeness, typicality, and generalizability

In a response similar to that of constructivism, critical researchers employing
qualitative research would note that we are not seeking to explain the “typical”
person, but to analyze that person’s possibilities and limits within a culture. In
this approach, individuals are not seen as “types” or members of aggregate groups
(although they may be both of these). Individuals instead are approached as beings
that inhabit subject positions that are possible within a culture. Because individuals
are members of society and must act within the society, they share certain
understandings and meanings; if they did not, they could be considered insane,
which in societal terms is the designation given to persons whose social realities
have no seeming connection to those around them.
Validity
The test of validity in the case of critical constructivist research is directly related
to its stated purpose of inquiry. The research is valid to the extent that the analysis
provides insight into the systems of oppression and domination that limit human
freedoms, and on a secondary level, in its usefulness in countering such systems.
Criticism
One of the charges against critical theory is its tendency toward elitism. With its
proponents’ commitment to the idea that research can bring about a better and
more equitable world, critics charge that critical theorists tend to assume that
they are not only more capable of analyzing a situation than most; they are better
equipped to offer a prescriptive plan of action. Critics charge that this often
brings theorists outside of their realms of expertise so that the insights they offer
are naive and unworkable in the contemporary setting.
Further, critics charge that critical theorists can be unwilling to listen to the
experiences of those most adversely affected by current policies and the status
80
quo, as they tend to focus their analyses on persons and institutions in positions Foundations of Qualitative
of power and authority. This, critics note, causes critical theorists to be out of and Critical Theory
touch with the very persons they purport to be most interested in helping. Paradigm
Differences between critical theory framework and post-positivism on five major

issues are as given below:
Post-positivism Critical Theory

Nature of reality External to human Material and external to the
mind human mind.
Purpose of research Find Universals Uncover local instances of
universal power relationships
and empower the oppressed.
Acceptable methods Scientific method Subjective inquiry based on
and data ideology and values; both
quantitative and qualitative
methods are acceptable.
Meaning of data Falsification Interpreted through ideology;
Use to test theory used to enlighten and
emancipate.
Relationship of Separate activities. Integrated activities.
research to practice Research guides Research guides practice.
practice
(Source: Foundations of qualitative research by Jerry W. Willis (2007))

1) Is critical framework a departure from post-positivism and interpretivism?
Give reasons in support of your answer?
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) What points should be kept in mind while defining the methodology of
critical research?
......................................................................................................................
......................................................................................................................
......................................................................................................................
3) Critical framework is sometimes known as the Frankfurt school. Why?
Elaborate on the foundations of critical theory.
......................................................................................................................
......................................................................................................................
......................................................................................................................
81
Issues and Perspectives 5.4 APPLICATIONS IN RESEARCH:
ILLUSTRATIVE CASES
Case 1: ‘A study on ‘invisible’ labour-force in India
Invisible workforce comprises of care-workers who work daily for low/negligible
wages however their work is critical to the success of the household/enterprise.
Primarily constituting of housewives, nannies, cleaners, etc., these workers are
engaged in ‘care’ work and their work is generally under-valued or over-looked.
A news daily recently reported that as per Census 2011 figures nearly 160 million
women in India aged between 15-59 years reported themselves as not working
but were primarily involved in domestic work, care work and rearing families.
In this case, the social actors are the housewives, nannies, cleaners, etc.
An interpretive researcher by using qualitative methods like phenomenology

and symbolic interactionism would try to understand the root causes of such
women not involved in economic work. Research could begin with penning
down research questions like ‘why the women have not taken up work outside
home’ or ‘Are their social pressures for the women sticking to domestic work’ or
even ‘Is the decision to work at home independent or curbed’. Next, the researcher
can shortlist a region predominantly populated with invisible workers and use
techniques of PRA/RRA to gather the requisite data. The perspectives on invisible
work may differ among the social actors in this case too. For example if the
researcher is probing the causes of invisible work he may observe that causes
may be social, cultural or at times political too. It would be important to gather
perspectives of not only unpaid care-workers but also their family members as
to why they did not try to change their current situation, village head-men to
understand whether it is due to lack of economic opportunities in the area and
also to know whether illiteracy is the cause of invisibility of women workers.
Case 2: Absenteeism among students of primary school in rural areas

Universal primary education being one of the goals of Millennium Development
Goals of UNDP, all developing countries are in the race of providing elementary
and primary education to children in rural and urban areas. While enrolment in
elementary and primary education has definitely increased in India in the last
decade, but equally troublesome statistics have emerged quoting high incidence
of absenteeism among students of primary school in certain rural areas. An
interpretive researcher using qualitative methods and tools can take different
perspectives of the social actors (read primary school children, teachers and
parents) for absenteeism in primary schools.
Case 3: A study of employer’s attitude towards distance/online-educated job

applicants.
Distance educated youth are perceived to be less efficient and trained as compared
to classroom educated youth. Instances of preferential bias have been observed
in this respect often in the job market. An in-depth study on the employer’s
mindset can be undertaken by an interpretive researcher to explore the causes of
the same. Extensive literature review is a pre-cursor to such research as that will
underpin the historical causes of preferential attitude and negative perception of
the employers. An interpretive researcher can employ actual observational
techniques or use witness accounts of the interviews/interactions of the employer
82
with distance/online educated job applicants. Research questions may range from Foundations of Qualitative
finding variability in the questions posed to such candidates from the ones who and Critical Theory
were educated in the traditional mode to quizzing grounds for negative perception Paradigm
of employers for the same. For example, a study quoted the following reasons
for dislike of distance educated job applicants by an employer: lack of rigor, lack
of face-to-face interactions, increased potential for academic dishonesty,
association with diploma mills, concerns about online students’ true commitment
evident from regularly venturing to a college or university physical location,
considered by some to be an important part of the educational experience.
Interestingly all of the above cases can also be studied in a critical framework as
only the questions posed would change and power-relationships would be judged.
Critical theory is also used subsequently after interpretive research as it is easier
to critically analyze a problem after undertaking a deeper understanding of that
problem. For example, after understanding the reasons behind the existence of
invisible workers, a critical researcher may critically analyze the role of power
relationships between the care workers and their families, social circle and the
environmental factors leading to it. Awareness campaigns and policies may be
designed to further empower them to be able to participate in the economic
workforce and to allow labour statisticians to recognize their care work within
the ambit of ‘work’.
1) Using a suitable example from economics, select a theme and write a proposal
in about 500 words using interpretative research methodology.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
2) Explain the scenario when both interpretative and critical theory framework
can be applied in the same research? Do they supplement each other? Why
or why not?
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
5.5 LET US SUM UP

The three paradigms – postpositivism, interpretivism and critical theory – are
the dominant guiding frameworks in the research literature in the social sciences.
Interpretivism and critical theory are two of the contemporary frameworks of
qualitative research besides positivism and post-positivism paradigms. Based
upon different ontological and epistemological positions, these paradigms analyze
nature of reality differently. Interpretivism is based on socially constructed reality
83
Research Methodology: whereas critical research is based on the critique of social reality. These paradigms
have emerged as a response to a perceived problem in society. Critical theory
was a response to the complexities of modern nation states that often lead to
domination and exploitation of one group by the other. Interpretivism proposed
that we abandon the search for generalizable truths and laws about human
behaviour and concentrate instead on local understanding.
5.6 EXERCISES
1) Consider the interpretivist philosophy of science. What to you see as the
most significant departure from post-positivism and why?
2) Consider a topic of research in an area of interest to you. What would be the
purpose of research on the topic if it were conducted from a post-positivist
perspective? Critical or Interpretive?
3) What does interpretivism have in common with the critical paradigm? In
your field which are more important: the commonalities or the differences?
5.7 KEY WORDS

Paradigm : A comprehensive belief system, world view or
framework that guides research and practice.
Ontology : It is concerned with the nature of reality and inquires

the characteristics of things that exist, is a major aspect
of metaphysics- a branch of philosophy.
Epistemology : It is concerned how can we know? It is also a major

aspect of metaphysics.
5.8 SOME USEFUL BOOKS/WEBLINKS

Willis, Jerry, W (2007), Foundations of Qualitative Research: Interpretive and
Critical Approaches, Sage Publications, USA.
Clark, Lynn Schofield, ‘Critical Theory and Constructivism: Theory and Methods
for the Teens and the New Media @ Home Project’ retrieved online from http:/
/www.ihrcs.ch/?p=92
‘Critical theory paradigms’ retrieved online from http://www.qualres.org/

HomeCrit-3518.html
5.9 ANSWER OR HINTS TO CHECK YOUR

PROGRESS EXERCISES
1) See criticism, Section 5.2
2) See Methodology, Section 5.2
3) See Section 5.2
84
Check Your Progress 2 Foundations of Qualitative
1) See Section 5.3 and Critical Theory
Paradigm
2) See Section 5.3
3) See Section 5.3
1) See Section 5.4
2) See Section 5.4, yes they supplement each other.
85
APPENDIX I
Criteria for Interpretive Case study aiming at Theory Building
Criterion Definition Specific case study Grounded theory1

tactic principles
Construct Establishing • Use multiple • Corroboration
validity correct sources for • Theoretical
operational evidence sufficiency
measures for • Establish chain
the concepts of evidence
being studied • Have key
informants
review draft case
study report
Internal Establishing • Do pattern • Theoretical coding
validity causal matching
relationship • Do explanation-
as building
distinguished • Address rival
from spurious explanations
relationships • Use logic models
External Establishing • Use theory in • Theoretical
validity the domain to single case generalisation
which a studies
study’s • Use replication
findings can logic in multiple
be case studies
generalized
Reliability Demonstrating • Use case study • Chain of evidence as
that a study protocol afforded by
can be • Develop case grounded theory
repeated with study database method
the same
results
(Case study methodology criteria, Yin, 2003, p.24)
1
The discovery of theory from data is known as grounded theory and it provides the opportunity
86 for the researcher to theorise from evidence existing in the data.
MEC-109
Research Methods
Indira Gandhi in Economics
Block
2
RESEARCH DESIGN AND MEASUREMENT
UNIT 6
Research Design and Mixed Methods Research 5
UNIT 7
Data Collection and Sampling Design 25
UNIT 8
Measurement and Scaling Techniques 53
Expert Committee
University of Hyderabad, Hyderabad University of Mumbai, Mumbai
Prof. Harishankar Asthana Dr. Manish Gupta

Professor of Psychology Sr. Economist
Banaras Hindu University National Institute of Public Finance and Policy
Varanasi New Delhi
Prof. Chandan Mukherjee Prof. Anjila Gupta

Professor and Dean Professor of Economics
School of Development Studies IGNOU, New Delhi
Ambedkar University, New Delhi
Prof. V.R. Panchmukhi Prof. Narayan Prasad (Convenor)

Bombay University and Former IGNOU, New Delhi
Chairman ICSSR, New Delhi
Prof. Achal Kumar Gaur Prof. K. Barik

Professor of Economics Professor of Economics
Faculty of Social Sciences IGNOU, New Delhi
Banaras Hindu University, Varanasi

Professor, Indian Institute of Associate Professor in Economics
Public Administration, New Delhi IGNOU, New Delhi

Course Coordinator and Editor: Prof. Narayan Prasad

Unit Resource Person Editor
(Format, Language and Content)
6 Prof. Narayan Prasad Prof. Narayan Prasad
IGNOU, New Delhi IGNOU, New Delhi
7 Sai S.S. Suryanarayan Prof. Narayan Prasad
Rtd. Joint Advisor Professor of Economics
8 Prof. A.K. Gaur Prof. H.S. Asthana
Professor of Economics Professor of Psychology
BHU, Varanasi BHU, Varanasi
&
Prof. Narayan Prasad
Professor of Economics, IGNOU, New Delhi
Print Production
Mr. Manjit Singh
October, 2015
© Indira Gandhi National Open University, 2015
ISBN-978-81-266-
Printed at :
BLOCK 2 RESEARCH DESIGN AND
MEASUREMENT
Research design is a logical structure of an enquiry and its formulation is
guided and determined by the research questions raised in the problem under
investigation. Apart from specifying the logical structure of data, research
design also test and eliminate alternative explanation. Broadly, the
observational design sampling design and statistical design are covered in
Research Design. The various attributes of people, objects or concepts are
being increasingly included in explanation of human behaviour in Economics.
Hence, these individual traits, attitudes need to measure for deeper analysis.
Research Design and measurement issues, therefore constitute the theme of this
block. The block comprises of 3 Units.
Throwing light on the concept of research design and various types of research,
Unit 6 covers the rationale and forms of mixed methods research with
illustration of three case studies.
Unit 7 provides comprehensive knowledge of all the elements for collecting

the quantitative data for research study. It covers the methods and tools of data
collection and the various issues related to sampling design.
Unit 8 deals with the various measurement scales, scaling techniques, and
criteria for good measurement and sources of errors in measurement.
Research Design and
Mixed Methods
UNIT 6 RESEARCH DESIGN AND MIXED Research
METHODS RESEARCH
Structure
6.0 Objectives
6.1 Introduction
6.2 Types of Research
6.2.1 Theoretical and Applied Research
6.2.2 Descriptive and Explanatory Research
6.2.3 Quantitative and Qualitative Research
6.2.4 Conceptual and Empirical Research
6.2.5 Other Types of Research
6.3 Research Design
6.4 Research Design vs. Research Methods
6.5 Research Methods
6.5.1 Quantitative Methods
6.5.2 Qualitative Methods
6.5.3 Mixed Methods
6.6 The Rationale for Mixed Methods Research
6.7 Forms of Mixed Methods Research Designs
6.8 Case Studies of Mixed Methods Research Design
6.9 Let Us Sum Up
6.0 OBJECTIVES
After going through this Unit, you will be able to:
• state the various types of research;
• explain the concept and meaning of research design;
• make a distinction between research design and research methods;
• delineate the characteristics of quantitative research and qualitative research;
• discuss the purposes of mixed methods research; and
• appreciate the application of mixed methods research in conducting
research studies in social sciences in general and economics in particular.
6.1 INTRODUCTION
Due to confusion or lack of clarity between research design and research
methods, we often use these two terms interchangeably. As a result, the
research design is evaluated on the basis of strengths and weakness of the
method rather its ability to draw relatively unambiguous conclusions. Hence 5
Research Design and
Measurement efforts have been made here to throw light on the distinction between these two
terms. Traditionally two approaches – quantitative research (QN) associated
with post positivist paradigm and qualitative research (QL) associated with
interpretative paradigm have been followed in social sciences. However, in
recent years increasing reliance on mixed methods research by combining
quantitative and qualitative methods in a single research project has been
observed. Hence in this unit, the concept of mixed method research, its types
and application have been covered. Application of mixed method research in a
subject like economics which is essentially quantitative in nature have been
illustrated by case studies. Before taking up the issue of research design and
research method, it is desirable to have an idea about various types of research.
Hence let us begin to introduce the various types of research.
6.2 TYPES OF RESEARCH

There are different ‘types of research’ depending on their nature and field of
specialization. While a classification based on a dichotomous distinction (like,
theoretical/applied, descriptive/analytical, conceptual/empirical, etc.) is
possible, it is necessary to recognize that there may be overlapping features
rendering such a classification less perfect to that extent. It is, therefore, useful
to first of all take a look at the underlying basic features so as to be able to
identify the different types of research within their respective connotations and
usage.
A variable which can be measured can take values or scores either in

quantitative or qualitative terms. For instance, yield of an agricultural output,
height/weight of individuals, etc. can be measured and expressed in
quantitative terms. On the other hand, characteristics like one’s feelings or
opinion (e.g., good/bad, male/female, agreeing/disagreeing, yes/no, etc.) are
attributes on which a choice can be expressed (and a score assigned) depending
on the possible alternatives or choices. Thus, variables like the performance of
an artist, respondent’s gender, etc. on which responses can only be categorized
or grouped are ‘qualitative variables’. From this angle, variables on which
data can be collected and expressed in quantified terms may be called as
‘quantitative variables’. Even variables which are expressed in qualitative
terms can be counted whereupon their total numbers become quantified
expressions. A popular example of developmental economics which ranks
different countries based on their status attained in a number of
areas/variables(some of which like the political freedom/choice enjoyed by the
country, though expressed in some quantitative measure, is qualitative in its
nature) is the human development index. In actual research (particularly
research in the field of social sciences), it is important to note that variables of
both the nature normally co-exist. In view of this, the analysis pertaining to
each type of variable/data needs different methods and techniques.
The type of research would also vary depending on the objectives of the study.
Research design varies with the type of research one likes to pursue. With this
background, we can now proceed to know the distinctions between different
types of research.
6.2.1 Theoretical and Applied Research

Research can either be theoretical or applied. The former i.e., theoretical
research can also be considered as ‘fundamental research’ as its outcome serves
6
Research Design and
as a foundation for all subsequent development in the field. Fundamental (or Mixed Methods
basic) research mainly concerns with formulation of theory with knowledge Research
perceived as an end in itself. It, thus, aims at obtaining knowledge of the
logical processes involved in a phenomenon. It pertains to the quest for
knowledge about a phenomenon without concern for its practical use. Such a
research may either verify the old theory or establish a new one. For example,
fundamental research in economics may consist of research to develop and
improve economic theories or evolve quantitative techniques to measure
parameters such as multiplier effect, elasticity of demand and supply, etc.
Fundamental research is essentially positive in nature.
Applied research, on the other hand, aims at finding a solution for a problem
facing society or industry. It is, thus, applied to practical situations or contexts.
While pure research discovers principles and laws, applied research discovers
ways of applying them to solve specific problems. It is useful to test the
theories developed empirically and can as a result also contribute to improving
the tools and techniques of measurement. Basic research, therefore, can be
treated as building blocks for applied research. Illustrations of applied research
in economics can be measurement of poverty, employment, rural development,
agriculture, environment, etc.
6.2.2 Descriptive and Explanatory Research

Descriptive research describes a situation, events or social systems. It aims to
describe the state of affairs as it exists. Surveys and fact-finding enquiries of
different kinds are part of descriptive research. Survey methods of all kinds
including comparative and correlational methods are used in descriptive
research studies. A survey of socio-economic conditions of rural/urban labour
is an area of descriptive research. In descriptive research studies, the
researchers have no control over variable. They can report only what has
happened or is happening. Descriptive research deals with questions like ‘how
does X vary with Y? or ‘how does malnutrition vary with age and sex?, etc.
Explanatory research, aims at establishing the cause and effect relationship.

The researcher uses the facts or information already available to analyse and
make a critical evaluation of the data/information. An example of explanatory
research is: ‘whether increase in agricultural productivity is explained by
improved rural roads?’
6.2.3 Quantitative and Qualitative Research

Quantitative Research is the systematic and scientific investigation of
quantitative properties and phenomena and their relationships. The objective of
quantitative research is to develop mathematical models, and to test
hypotheses. Based on ontological and epistomological premises, quantitative
research is characterized by the following attributes:
• A belief in a single reality.

• The pursuit of identifying universal laws.
• Separation of knower (researcher) from known (researched).
• The possibility and necessacity of value free research.
• Pursuit of universal laws (findings) beyond the limit of research/social
context. 7
Research Design and
Measurement • The tendency to work with large, representative sample.
• An emphasis on deductive research via falsifiable hypothesis.
• Formal hypothesis testing.
In contrast, Qualitative Research captures a set of purposes associated with

meaning and interpretation. Stress is laid on the socially constructed nature of
reality, the intimate relationship between the research and researched and
situational constraints that shape the enquiry. The key attributes associated with
qualitative research are:
• A belief in a constructed reality, multiple realities, co-existence realities.

• An interdependence between the knower and the known.
• Impossibility to separate the researcher from the research subject.
• Heavy role of context in research process.
• The impossibility to generalize research findings beyond the limits of the
immediate context.
• Non-separation of cause and effect.
• The explicit focus on inductive, exploratory research approaches.
• The tendency to work with small and purposely chosen sampling.
• Analyses holistic system.
6.2.4 Conceptual and Empirical Research

Conceptual research is related to abstract ideas or theory. It is generally used
by philosophers and thinkers to develop new concepts or to reinterpret the
existing theories.
Empirical research relies on experience or observation. It is data based. It is

subject to verification by observation and experiment. This type of research is
particularly useful when validation or verification of an aspect is required.
6.2.5 Other Types of Research

Research may be exploratory or formalized. Exploratory research aims at
developing the hypothesis rather than testing a pre-conceived hypothetical
contention or notion. Formalised research studies deal with a definitive
structure within which specific hypotheses are tested. Historical research
utilizes existing documents to study events of the past. Research can also be
experimental or evaluative. Experimental research aims at identifying the
causal factors by means of experiments. In evaluative research, the cost
effectiveness of a programme is examined. Such research addresses the
question of the efficiency of a programme and are useful in taking policy
decisions on issues like whether the programme is effective and/or needs
modification or re-orientation? Whether it should be continued?
Action research is another type of research. It deals with real world problems
aimed at finding out practical solutions or answers to them. It gathers feedback
which is then used to generate ideas for improvement. You will find details on
Action Research in Unit 20.
8
Research Design and
Check Your Progress 1 Mixed Methods
Research
1) Distinguish between quantitative variable and qualitative variable.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
2) State the main objectives of applied research?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
3) In what sense quantitative research is different from qualitative research?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
4) What is the distinction between explanatory and descriptive research?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
6.3 RESEARCH DESIGN

Research design is a logical structure of an enquiry. Given the research
question or theory, what type of evidence is needed to answer the question (or
to test the theory) in a convincing way – constitutes the essence of the research
design. Let us use an analogy to understand the term ‘research design’. While
constructing a building, the first decision to be arrived at is: whether we need a
high rise office building, a factory, a school or a residential apartment etc.?
Until this is decided, we cannot sketch a plan and order material or setting
critical dates for completion of the project dates. Similarly, a social researcher
needs to be clear about the research questions and then the research design will
flow from the research questions. The function of a research design is to ensure
that the evidence obtained enables us to answer the initial research questions as
unambiguously as possible. Obtaining relevant evidence entails specifying the
type of evidence we need to answer the research question, to test a theory, to
evaluate a programme or to accurately describe some phenomenon. The issues
of sampling, method of data collection (e.g. questionnaire, observation,
document analysis), design of questionnaire etc. are all subsidiary to what
constitute the evidences that need to collect to answer the research questions.
Thus the research design ‘deals with a logical problem and not a logistical
problem. Apart from specifying the logical structure of the data, it also aims to
test and eliminate alternative explanation of results.
9
Research Design and
Measurement The research design comprises of the following components:
i) the sampling design; (the type of sampling method i.e. random or non
random sampling);
ii) the statistical design (i.e. the sample size and the method to draw the
sampling to be adopted);
iii) the observational design (i.e. the instrument of collection of data);
iv) the operational design i.e. the specific details by which the procedure in
(i), (ii) & (iii) above are to be carried out.
6.4 RESEARCH DESIGN VS. RESEARCH

METHODS
Research design is different from the research methods. The research methods
are made of data collection and techniques of data analysis whereas the
research design is a logical structure of an enquiry. There is nothing intransic
about any research design that requires a particular method of data collection.
How the data are collected is irrelevant to the logic of the research design.
Confusing research designs with research methods leads to poor evaluation of
designs. For example equating cross sectional designs with questionnaires or
case studies with participant observation result in evaluation of research
designs against the strengths and weakness of the method rather than their
ability to draw relatively unambiguous conclusion or to select between rival
plausible hypothesis.
6.5 RESEARCH METHODS

Research methods as explained in Unit 1 refers to tools and techniques of data
collection and data analysis. Broadly Research Methods are put under two
categories: (i) Mono Methods, (ii) Mixed Methods. Within mono methods,
there are two approaches (i) quantitative (QN) methods (ii) qualitative (QL)
methods. Since quantitative methods are associated with positivist and post-
positivist paradigm, quantitative method is also termed as (QN) method
research. Similarly by virtue of the fact that QL methods are associated with
interpretativist paradigm and critical theory paradigm, research conducted by
the qualitative method is also termed as qualitative methods research.
Research Methods
Mono-Methods Mixed Methods of

Quantitative
Quantitative and
Methods
Qualitative Methods
Qualitative
Methods
10
Research Design and
6.5.1 Quantitative Methods Mixed Methods
Research
The researcher using this approach tests theory through empirical observations
and inclines to establish cause and effect relationship. Well predesigned
structured questionnaire and structured interviews are resorted as tool to collect
the data. The variables that can be measured are emphasized. The descriptive,
analytical and inferential statistical techniques are used to analyse the data.
Experimental research design is used to make the logical structure of the
research study. Generalization of laws and their replication is emphasized.
6.5.2 Qualitative Methods

Under qualitative approach, a researcher starts with observation as a basis for
generating theory and concentrates on meaning of observations. He studies
events as they occur in naturalistic settings. He allows interview topics to
emerge during conversation and listens to others’ interpretation and
perspectives.
Under qualitative method, open-end questionnaire, key informants, group

discussion and unstructured interviews, documents, interview transcripts etc.
are used as tools to collect data. Content analysis, Case study, Action research
participatory method, Cluster analysis, Factor analysis, Correspondence analysis,
Context analysis, are used in analyzing qualitative data. You will find details of
these techniques in Unit 13, 14, 15, 16 and 17 of Block 4 and 18, 19 and 20 of
Block 5 of this course.
6.5.3 Mixed Methods

The combination of at least one qualitative and at least one quantitative
component related to measurement scale, tools of data collection or data
analysis technique in a single research study/project is known as mixed
methods research. For example it can be in the following forms:
• Employment of more than one measurement procedure in different kind of

situations such as different ways of measuring levels of job satisfaction.
• Employment of two or more methods of data collection. For example we
may employ observation, interview and questionnaire to collect the data
about decent work.
• Employment of two or more data analysis techniques for example, content
analysis and factor analysis.
Mixed methods research has gained a tremendous popularity in social,

behavioural and related sciences in recent years. The rationale for mixed
method design research is to take the best of qualitative (QL) and quantitative
methods (QN) and combine them. However, many debates on mixed method
research design are based on methodological arguments due to the ways QN
method is linked to positivist paradigm and QL method to Interpretative and
critical theory paradigm.
As discussed above, keeping in view the diametrical oppositional attributes of

QL and QN methods, one may argue that these methods are not compatible to
each other and in such a situation how a truce between two kinds of approaches
be useful to conduct the meaningful research?
Notwithstanding the philosophical debate between QN and QL approach.

Mixed methods research designs are justified primarily by exploiting the
11
Research Design and
Measurement strength of each method and by combining their respective strengths within one
single research design. The argument of incompatibility between two methods
on philosophical ground is not valid because of the following reasons:
(Bergman, 2008)
i) The QL and QN methods represent large and heterogeneous familities of

methods which vary within their own family to such an extent that it is
difficult to identify unique set of qualities or attributes that encompasses the
characteristics of one family of methods, clearly distinct from the
characteristics of the members of the other family. Most characteristics
encompass either only a subgroup of members of the family or are also
applicable to some members of other family. Hence there is a need to re-
think the division of labour between QL and QN methods in order to better
understand the possibilities and functions of methods to better justify and
apply mixed method design.
ii) The proposition of existence of one single reality under QN method is

inconsistent with research applications. The emergent structures are based
on co-construction between the researchers’ selection and understanding of
items in a questionnaire, their choice of analytic strategy and their
interpretation of the statistical output on the one hand, and the way the
respondents interpreted the survey questions with the given social, political,
historical and economic context, on the other hand.
From a methodological perspective, it is not proper to declare one approach

more or less valid, valuable or scientific. Instead how to understand and
analyze data need to be based to a large extent on the consistency formed
between how to understand data in conjunction with the specific research
question, rationale, aim etc. only in connection with the specificities of the
research goals it makes sense to delimit the nature of reality. Thus, the
decision on whether the researcher deals with one single reality, a
constructed reality, multiple realities, multiple constructed realities or a co-
constructed reality is unrelated to whether patterns in the data are detected
via statistical analysis or otherwise.
iii) A small vs large samples: The sample size for many QN studies is often
quite small and sample size for some QL studies can indeed be large.
iv) Regarding hypothesis testing, many researchers engaged in QL research,

pursue conjectures that are embedded either in their research questions, the
kind and the way they collect data, the way they analyse the data and the
way, they protect themselves from selective reportings of their findings. In
many types of well-established statical analyses, the QN methods like
cluster analysis, factor analysis, correspondence analysis, multi-
dimensional scaling are used in exploring data structures.
In order to avoid misconceptions and mistakes while deciding the research

design issues, the following points need to be kept in mind.
a) We refrain ourselves from making specific argument for or against the QL

and QN method for a specific research project. We need to focus our efforts
more explicitly on embedding and justifying our selected methods
according to our research questions, data needs, theoretical grounding and
research design.
12
Research Design and
b) Data collection methods (i.e. unstructured narrative interview, survey Mixed Methods
research based on closed-ended questions) and data analysis methods Research
(qualitative content analysis, discourse analysis, quantitative content
analysis) should be differentiated.
c) We ought to be more aware of the actual inductive and deductive analytic

phases of our research projects.
1) Which constituent of research study does guide the research design?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) State the functions of research design?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) What is the distinction between Mono Method and Mixed Method?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
6.6 THE RATIONALE FOR MIXED METHODS

RESEARCH
Mixed methods research has gained tremendous popularity in the social,
behavioural and related sciences in recent years. This increasing popularity of
mixed methods research is reflected in terms of the claims by mixed method
researchers, increase in the publications on this topic and the inclusion of
mixed methods designs in the text books. The argument for combining
qualitative and quantitative methods is put forwarded on the following grounds.
i) The research methods associated with both quantitative and qualitative

research have their own strengths and weaknesses. Combining them
allows the researchers to offset their weakness to draw on the strengths of
both.
ii) Mixed methods designs enable the researcher to provide a comprehensive

account of the area of the enquiry in which he or she is interested if both
quantitative and qualitative methods are used. for example, causal
explanation of any event or phenomena can be captured by quantitative
methods whereas the reason explanation and pattern explanation or the
explanation involving processes can better be captured by qualitative
methods.
13
Research Design and
Measurement iii) Quantitative and qualitative methods be used to answer different
questions in a study. Further mixed methods can be utilized in order to
gain complementary views about the same phenomenon or relationship.
Research question for the two strands of the mixed study address related
aspects of the same phenomenon.
iv) Mixed methods design can be used to develop the research further.
Questions for one strand emerge from the inferences of a previous one
(sequential mixed methods), or one strand provides hypothesis to be
tested in the next one.
v) Mixed methods may also be used to assess the credibility of inferences
obtained from one approach (strand). Exploratory and explanatory/
confirmatory questions in such situation, may be useful.
vi) Mixed methods are also used to obtain divergent pictures of the same
phenomenon. These divergent findings can ideally be compared and
contrasted.
vii) Quantitative research provides an account of structures in social life but
qualitative research provides sense of process.
viii) Mixed methods research entails to generate hypothesis by using
qualitative method and testing the hypothesis resorting quantitative
method within a single project.
ix) Mixed methods research enables to present diversity of views i.e. combining
researchers’ and participants’ perspectives through quantitative and
qualitative research respectively by way of uncovering relationships between
variables through quantitative research and revealing meaning among
research participants through qualitative research.
6.7 FORM OF MIXED METHODS RESEARCH

DESIGNS
Connecting qualitative and quantitative methods through sequencing and
prioritising, there can be four forms of mixed methods research designs.
1) Preliminary qualitative inputs to core quantitative method (qual-

QUANT): With a view to use the strengths of qualitative methods to
contribute to core quantitative methods, preliminary qualitative methods are
used for collecting and interpreting the quantitative data. Preliminary
qualitative methods can also be used as an input to generate hypotheses,
develop content for questionnaires etc.
2) Supplementary qualitative method as a follow up to core quantitative
method (QUANT-qual): Supplementary qualitative method as a follow up
to core quantitative method can be used to extend what is learned from
quantitative study. It can demonstrate bases for results, examine reasons for
failed hypotheses, explore the theoretical significance of outliers and so on.
3) Preliminary quantitative inputs to core qualitative methods design
(quan-QUAL): The study undertaken by using core quantitative method
design gets inputs by way of using preliminary quantitative method for
collecting and interpreting the qualitative data. Preliminary quantitative
14
Research Design and
method can also locate major differences between sub-groups, guide Mixed Methods
purposive sampling, establish results to pursue and so on. Research
4) Supplementary quantitative method to core qualitative method design

(QUAL-quan): Supplementary qualitative method as a follow up to core
qualitative method can be used to develop measures of key concepts and
relationships across different samples.
6.8 CASE STUDIES OF MIXED METHODS

RESEARCH DESIGN
i) Studies where the quantitative component is dominant in terms of the
framework of positivism and post positivism paradigm and inductive and
deductive logic of enquiry guides the study, quantitative approach is more
dominant and the researchers specialize in quantitative work. In such
studies, qualitative data and their manipulation are shaped by the nature of
research questions raised in the quantitative part of the project. The
qualitative data are treated and transformed in effect into quantitative
variables with the purpose of fleshing out the explanations required for the
quantitative results. In such cases quantitative component follows
qualitative component mixed methods. An illustration of a Research Study
which has used Mixed Method pre dominated by the Quantitative Method
is given below in the box:
CASE 1: Mixed Method Research Design: Conducted the study within

the framework of post positivist paradigm using inductive
logic of inquiry and analyzing the qualitative data by
quantitative technique of data analysis
Study: ‘Decent Work: Concept, Measurement and Status in India’:
Ph.D thesis by Ms. Nausheen Nizami Ph.D Research Scholar,
(Economics) IGNOU, N. Delhi
The study was conducted within the framework of positivist paradigm using
inductive logic of enquiry with the objectives to (i) examine the applicability
of indicators of ‘decent work in India (ii) develop indicators and tools to
measure ‘decent work (iii) evaluate the status of decent work in IT
companies and to identify the socio-economic attributes influencing status of
decent work, and (iv) know the employer’s perspective in provision of
decent work’.
Methodology: Decent work sums up the aspirations of the people in their

professional life and is a revolutionary agenda of International Labour
Office. Decent work is any such work which ensures provision of fair and
free employment to all men and women of economically productive age-
group in conditions of fairness, equity, security and dignity. It is a
multidimensional concept that applies to both formal and informal sectors of
an economy. The study is based on primary data collected through well
designed questionnaire and telephonic interviews. Measurement of decent
work was undertaken using both quantitative and qualitative methods of data
analysis. A characteristic feature of this study was to transform the
qualitative data in a manner so as to be able to use quantitative techniques of
data analysis.
15
Research Design and
Measurement
Quantitative Techniques to analyse qualitative data
a) Composite Index construction for Decent Work
With a view to measure the extent and range of decent work
provision, decent work indices were constructed. These indices have
been made for all the indicators of decent work used in this study by
coding, normalising and aggregating the responses as the data was
qualitative in nature. Thereafter, the standard formula developed by
United Nation’s Development Programme (UNDP) for the
construction of HDI and other indices was applied.
b) Factor analysis of decent work indicators

The technique of factor-analysis was adopted to short-list the most
relevant and reliable indicators of decent work. The principal
component method was used to outline components (i.e. indicators in
this study) explaining the maximum variance in the dependent
variable.
c) Phi ф Coefficient of Correlation Analysis

Since the variables were dichotomous in nature, correlation analysis
was attempted to test whether there was statistically significant
associations between the responses on different decent work
indicators. Partial correlation analysis examined the relationship
between few socio-demographic variables and composite Decent
work Index keeping the effect of other variables constant.
d) Chi-square test for independence

Chi-square test was conducted to test the independence of the results
between the two scales of Decent work index (For example, 0 and
0-1) and results of each decent work indicator for those two scales.
Thus an efficient mix of quantitative and qualitative methods

were used to deal with the key issues related to decent work in
this doctoral thesis.
Major Results:
1) Composite decent work index reflects deficit in the status of
decent work for a vast majority (94%) of IT employees.
2) Adequacy of earnings and productive employment are directly
associated with decent work which is in consonance with the
features of search and matching theory of employment.
3) Longer working hours have a significant impact on the physical
and emotional well-being of an employee. A vast majority of IT
employees (95%) were found to be working for longer hours.
4) Work-life balance and adequate earning and productive
employment were found to be correlated. Majority of the
employees in IT sector were found to be earning adequate
earnings but their work-life balance was disrupted owing to
several reasons including longer working hours. This implies that
higher earnings of the employees in the IT industry are necessary
16
Research Design and
Mixed Methods
but not a sufficient condition for the employees. Their higher earnings Research
are not resulting into their well-being (mental) as reflected in their
deteriorating health status and imbalance between professional and
personal lives.
5) Social security mechanisms, social dialogue and fair treatment at work

places are major aspects of decent work owing to ensure sustainability
of economic well-being of the employees.
6) Decent work reflects decent work place but a decent work place may
not always translate into decent work because of the gap of the nature
of work and work amenities and environment. This implies that deficit
in decent work to the employees is correlated with deficit in decent
work places.
7) Age and social class are the prominent attributes in influencing decent
work status.
8) Despite the presence of premier educational institutions like IITs,

RECs, etc. demand-supply gap in the available manpower for IT
industry is the major concern of the employers pushing them to face the
problem of adverse selection. A direct implication of this situation is
the low employment elasticity of IT industry with respect to its growth
and increasing contribution to GDP.
ii) Studies where qualitative component had priority in terms of the frame
work of interpretative paradigm and abductive logic of enquiry. In such
situations, the researchers identify themselves primarily as qualitative
researchers. In such studies, the tools used for data/information collection is
mixed in nature. The modalities of the different types of data are
maintained and are not treated as commensurate. Integration of data takes
place as a part of interpretative process referred to as analytic or
interpretative integration. In such studies quantitative techniques can be
used to analyse the perceptions of the people which may be qualitative in
nature. An illustration of the research study using mix of qualitative and
quantitative methods in the manner discussed here in is given below:
CASE 2: Mixed Methods Research Design: Conducted the study

within the framework of Qualitative Research Approach
using Quantitative Methods for data analysis
Study: Gender specific impacts of climate variation in Agriculture:
By Mamta Mehar, Ph.D Research Scholar, IGNOU, N. Delhi
The Main Objective of the study was to understand the gender differences
in agriculture and analyse the linkages between gender and climate
variation using mixed method approach. The research questions addressed
in this study were: (i) Does there exist difference in gender perception on
climate variability? (ii) Whether male and female farmers follow similar
coping strategies to mitigate risk and shocks in agriculture? The study
focused on to understand how gender and climate variation in agriculture
are interrelated with the support of empirical evidences.
17
Research Design and
Measurement Methodology
This study was conducted within the framework of qualitative approach
anchored in interpretative paradigm using quantitative methods –
questionnaire as a tool and survey as a method to collect the data and
multivariate probit model to find out probability in deciding adaption
strategy by male and female to climate variability.
This research study used case study method within a socio-cultural

framework. Case study methodology derives much of its rationale and
methods from ethnography characterized by “an instance in action” in a
“real life context when the boundaries between the phenomenon and
context are not clear”. Keeping in view, the logic of enquiry, the research
questions were analysed using both qualitative and quantitative methods.
The qualitative analysis was done within the framework of “Interpretative
paradigm of qualitative research”. Since the research objective was to
understand human behaviour within the surrounding context of climate
change, where it was difficult to isolate the phenomena of climate
variation and perception of males and females towards it, conducting of the
present study under interpretive paradigm was appropriate. The associated
research strategy used in this study was abduction. Abduction is the logic
used to construct descriptions and explanations that are grounded in the
everyday activities as well as in the language and meanings used by social
actors. The social actors in the present proposal are men and women
farmers, and their activities to be analysed are specific to decision making
in climate change situation. The interpretative paradigm using abduction
strategy was done via literature review as well as synthesis from findings
of focus group discussion during the survey. In the wake of theoretical
understanding, it was tried to quantify the impact using appropriate
empirical analysis and using data from a baseline survey conducted by
CCAFS in 2013 of 641 farm households in twelve villages of Vaishali
districts in Bihar, India; it was attempted to quantify the gender
differentiated impact.
Results
The linkage between gender and climate variation in agriculture was more
influenced from subjective experiences of individuals or society. The
movement of understanding “is constantly from the whole to the part and
back to the whole” (Gadamer, 1976). According to Gadamer, it is a
circular relationship. Since the study aims to understand the perception of
human in context of climate variation, it cannot be explained entirely by
any scientific paradigm where hypothesis are pre-structured and to some
extent results are known. The interpretative paradigm points out that
positivism, in its first attempt to model the social after the natural sciences,
fails to see that unlike nature, social reality exists only insofar as lay
members create that reality in meaningful interaction (Fuchs S., 1992:
198). But the point of interpretive paradigm is that we must first
understand social worlds from within, before we develop scientific models
and explanations (Fuchs, S. 1992: 205). According to Willis (1995)
interpretivists are anti-foundationalists, who believe there is no single
correct route or particular method to knowledge. Walsham (1995) argues
that in the interpretive tradition there are no ‘correct’ or ‘incorrect’
theories. Instead, they should be judged according to how ‘interesting’ they
are to the researcher as well as those involved in the same areas.
18
Research Design and
Mixed Methods
The linkages between gender and climate variation in context of Research
agriculture are bi-directional. Changes in climate event effect agriculture
through reduction in yield, increase in occurrence of pests or crop disease,
changes in time of activities for example, changes in winter time could
result in change in harvesting or sowing days of respective crops. All this
affects farmers’ livelihood and work activity in different ways. For
example, women (and children) being the ones traditionally charged with
water collection roles, have to travel longer distances after drought.
Evidences suggest decrease in yield and thereby total availability of food
consumption in a household basket usually has been seen as
complemented by reduction in meals of female. This exposes them to
health issues. Male farmers are usually seen to migrate in search of
nonfarm opportunities outside their village domain, putting all burden or
responsibilities on women. Moreover mostly in rural areas women are
less likely to get education as compare to male, they are less allowed to
mobile and do more household related activities, and they have less
control over the family assets.
The other direction of gender differentiated impacts in context of climate

events in agriculture can be viewed through understanding their coping
strategies. The adaption of particular strategies are often influenced by the
priorities and options available for women and men for coping these
strategies. Constraints on women’s mobility, and behavioural restrictions
hinders their ability to make decision on various matters and thus they
have less choices of coping strategies available. Additionally inequitable
access to assets and resources such as land, knowledge, technology,
limited power in decision-making, education, health care and food have
been identified as determinant factors behind adaptation of different
coping strategies of male versus female. The results of quantitative
analysis also supports this argument.
The females are mostly illiterate (71 per cent) and have little access to
resources. Only 9 per cent of farm household have reported to provide
land entitlement to at least one female member. Additionally the gender
power dynamics is seen to favour only one gender i.e. male. The decision
making role in different household related expenditures as well as in
agriculture activities is dominated by male. Considering the perception of
male vis-à-vis female farmer about effects of climate change has shown a
similar pattern. More than 90 per cent of farmers are aware about climate
variation and feels that weather conditions adversely affect their
agriculture activities. However again, the decisions to cope with the risk
due to climate have suggested different behaviour, priorities by male and
female. Survey results suggest that almost 60 per cent of surveyed farmers
have adapted at least one coping strategy and in total adoption of more
than 30 coping strategies are reported. The results of multivariate probit
model suggest that male farmers have higher probability in deciding
adaptations through additional jobs as well as specific agriculture
strategies such as crop rotation and use of hybrid seed in farming
practices.
19
Research Design and
Measurement iii) Studies presenting diversity of views i.e. combining researcher’s and
participants’ perspectives through quantitative and qualitative research.
Researchers uncover relationship between variables through quantitative
methods and reveal meaning among research participants through
qualitative methods. Thus a comprehensive account of the area of the
enquiry is presented. An illustration of the study using this form of mixed
method research is given below:
CASE 3: Mixed Method Research –used Quantitative Methods for data

analysis and Qualitative Methods for revealing and
understanding participants’ perspective on quality of education
Study: ‘Measuring Quality of Education and Its Determinants: Indian

Context’: Ph.D thesis submitted by Ms. Charu Jain, Ph.D
Research Scholar, (Economics) IGNOU, New Delhi
The study was carried out with the objectives to (i) examine the linkages of
secondary education with various socio-economic outcomes at all India
level and state level, (ii) identify various determinants (including students’
family background characteristics, and schools’ characteristics) affecting
students’ performances and teaching efficiencies in senior secondary
schools in Delhi. To fulfill these objectives, both quantitative and
qualitative research approaches were applied.
Quantitative Method: The quantitative approach was used to test the

hypothesis generated. Two set of questionnaires one related to the teachers
and the other to students were got filled up through in-depth face to face
interviews with the respondents. The student questionnaire was translated in
Hindi language to enhance the understanding level of students in Hindi
medium government schools in Delhi. Within translation procedure, the
student questionnaire was first translated in Hindi and thereafter re-
translated in English to re-check the accuracy and meaning of questions.
The teachers’ questionnaire was provided in English language only. The
information thus retrieved from these questionnaires was used to measure
the performance of students and teacher analyzing thereby the overall
quality of education at lower secondary and senior secondary levels of
education. The information collected through quantitative approach was
statistically analyzed and tested using various statistical softwares like
SPSS and STATA?
Qualitative Method: In the qualitative approach, direct observation

method and key informer interviews were used to supplement the
information/data collected through quantitative method by way of two sets
of structured questionnaires. Qualitative research method enabled the
researcher to approach the research topic from the perspective of teachers
and students. Further, the qualitative techniques were used to examine
social processes which otherwise could not have been captured by
traditional quantitative measures.
The direct observation technique enabled researcher to learn about the

behaviour of the people under study: students and teachers in the natural
setting i.e. schools through observing. It gave the researcher the first hand
information on the quality aspects of school and classrooms and working
conditions for teachers, which were quite useful in doing the analysis.
20
Research Design and
The field notes, which were taken down during the period of fieldwork, were Mixed Methods
instrumental during analysis and interpretation of the results, as they helped in Research
recalling the incidents that took place at the time of data collection. Face-to-
face interviews with the key informants were also conducted within
qualitative approach. Key informant interviews were qualitative informal
interviews with people who knew what was going on in the community. Our
purpose of conducting key informant interviews was to collect information
from people who had firsthand knowledge about the overall functioning of
school. The key informers in the sample schools were selected using
purposive sampling technique and in this case the key informants were either
senior teacher in school or senior administrative person in schools.
Interviewing key informants enabled the researcher to look at the underlying
issues and problems with varying perspectives. These interviews were
conducted using an informal approach. Initially the questions were drafted in
a form of well designed open ended short questionnaire for conducting face
to face interviews with the respondents. One respondent or a group of 2-3
people from each school were selected purposively for gathering this
information, depending upon how well they were informed about the school.
Apart from these informal interviews, the school guards/receptionists were

also contacted informally to gather the information on schools. The
information thus collected through qualitative approach was first entered in
Excel in a pre-designed format and then coded so that it can be further
analysed statistically. Moreover, the responses to few open ended questions
were reviewed and inferences were drawn which were quite useful in
providing suggestions and recommendation for this study.
Main Results
1) Secondary education increases income level of individuals to a significant
extent. One unit increase in Gross Enrollment Ratio (GER) at secondary
level leads to increase in the personal disposable income by 0.65 units.
2) The educational background of households determine their occupational
pattern. The households with at least secondary education are either
salaried or business class, while those with elementary level of education
are more concentrated in agricultural/wage earning activities. Higher the level
of education among females, greater the proportion of female work
participation rate.
3) Secondary education bears highest impact on improving health
indicators. Secondary education has high potential for bringing
demographic transitions in terms of decline in maternal mortality rates,
infant mortality rates, fertility rates, death rates etc.
4) Secondary education attainment brings behavioral changes in individuals
in terms of higher savings and diversion in the expenditure pattern.
5) School resources, teachers and teaching quality, student’s family
background, mass media exposure and self motivational factors of
students determine their outcome to significant extent.
6) Within school characteristics, cleanliness in school, well qualified
teachers their friendly and supportive attitude with students and quality
of school infrastructure influence students performance positively. Lack
of student motivation, large class size, lack of school resources and
facilities, lack of adequate teaching material affect the teaching abilities
adversely.
21
Research Design and
Measurement Check Your Progress 3
1) Give two situations where Mixed Method research is superior to Mono

Method?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Give an example of Mixed Method where qualitative method pre-
dominates quantitative method?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) State the different forms of Mixed Methods?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
6.9 LET US SUM UP

Traditionally two distinct research approaches – Quantitative (QN) and
Qualitative (QL) – have been followed in undertaking research in social
sciences. The former is associated with positivism and post positivism
paradigm and the latter principally with interpretative paradigm and critical
theory paradigm. Depending upon the nature and field of specialization
research studies may be of various types. Important among these are –
theoretical and applied, descriptive and explanatory, conceptual and empirical,
exploratory and experimental etc.
Research design is a logical structure of enquiry specifying the types of

evidences needed to answer the research questions. Research methods are made
of tools for data collection, type of data, sampling design, sample size,
techniques for data analysis etc. There is nothing intransic about any research
design that requires a particular method of data collection. Research Methods
can broadly be of two types – Mono Methods and Mixed Methods. Within
mono methods, it can be either quantitative or qualitative. The combination of
at least one qualitative and at least one quantitative component in a single
research project is known as mixed methods research. The mixed methods have
gained popularity in carrying out research in social and behavioral sciences
because combining them allows the researchers (i) to offset their (QN and QL
methods) weaknesses, (ii) provide comprehensive account of the area of the
enquiry, (iii) to answer different types of questions. Studies undertaken under
mixed methods can be of four types – (i) preliminary quantitative methods
22
Research Design and
input design and core qualitative methods as core design (quan-QUAL), (ii) Mixed Methods
supplementary quantitative methods design as follow up to core quantitative Research
methods design (QUAL-quan), (iii) preliminary qualitative methods input
design and core quantitative methods design (qual-QUAN), (iv) supplementary
qualitative methods design as follow up to quantitative methods design
(QUAN-qual).

Bergman Manfred Max (2008): The Strawmen of Qualitative-Quantitative
Divide and Their Influence on Mixed Method Research in Advances in Mixed
Methods Research, edited by Manfred Max Bergman, Sage Publications Ltd.
London.
Creswell John W, Clark Kicki L Plano and Amanda L. Garrett (2008):
Methodological Issues in Conducting Mixed Methods Research Designs in
Advances in Mixed Methods Research, Sage Publications Ltd. London.
Tashakkori Abbas and Teddlie Charles (2008): Quality of inferences in Mixed
Methods Research in Advances in Mixed Methods Research, Sage Publications
Ltd. London.
Morgan David L (2014): Integrating Qualitative and Quantitative Methods- A
Pragmatic Approach, Sage Publications, London, N. Delhi, Singapore.
6.11 ANSWERS OF HINTS TO CHECK YOUR

PROGRESS EXERCISES
1) A variable expressed in terms of numerals like 1, 2, 3,.. etc. is called
quantitative variable. For instance, yield of an agricultural output,
height/weight of individuals, etc. can be measured and expressed in
quantitative terms. On the other hand, a variable expressed in terms of the
choices or opinion about attributes/ characteristics like one’s feelings or
opinion is called qualitative variable.
4) Descriptive research describes a situation, a phenomenon or event or a
social system. It aims to describe the state of affairs as it exists. On the
other hand, explanatory results aim to establish cause and effect
relationship.
1) Research questions guide the research design.

2) There are two major functions of the research design: (i) to determine the
type of evidences (data/information) which are needed to answer the
research questions, (ii) to eliminate the alternative explanation of the
results.
23
Research Design and
Measurement 3) If a research study is confined to use either quantitative or qualitative
method, it is said to be a mono method research. The combination of at
least one quantitative and at least one qualitative component in a single
research project is known as mixed methods research.
1) See Section 6.6

2) See Section 6.7
3) See Section 6.7
24
Data Collection and
UNIT 7 DATA COLLECTION AND Sampling Design
SAMPLING DESIGN
Structure
7.0 Objectives
7.1 Introduction
7.2 An Overview of the Unit
7.3 Method of Data Collection
7.4 Tools of Data Collection
7.5 Sampling Design
7.5.1 Population and Sample Aggregates and Inference
7.5.2 Non-Random Sampling
7.5.3 Random or Probability Sampling
7.5.4 Methods of Random Sampling
7.5.4.1 Simple Random Sampling with Replacement (SRSWR)
7.5.4.2 Simple Random Sampling without Replacement (SRSWR)
7.5.4.3 Interpenetrating Sub-Samples (I-PSS)
7.5.4.4 Systematic Sampling
7.5.4.5 Sampling with Probability Proportional to Size (PPS)
7.5.4.6 Stratified Sampling
7.5.4.7 Cluster Sampling
7.5.4.8 Multi-Stage Sampling
7.6 The Choice of an Appropriate Sampling Method
7.7 Let Us Sum Up
7.0 OBJECTIVES
• appreciate different methods of collecting data;
• acquire knowledge of different tools of data collection;
• define the key terms commonly used in quantitative analysis, like
parameter, statistic, estimator, estimate, inference, standard error,
confidence intervals, etc.;
• distinguish between random and non-random sampling procedures for data
collection;
• appreciate the advantages of random sampling in the assessment of the
“precision” of the estimates of population parameters;
• acquire knowledge of the procedure for drawing samples by different
methods;
25
Research Design and
Measurement
• develop the ability to obtain estimates of key parameters like population
and sample aggregates, proportion, mean, etc.; and of the “precision” of
such estimates under different sampling methods; and
• appreciate the feasibility/appropriateness of applying different sampling
methods in different research contexts.
7.1 INTRODUCTION
Studies of the behaviour of variables and of relationships amongst them
necessitate measurement of the variables involved. Variables can be
quantitative variables like GNP or qualitative variables like opinions of
individuals on, say, ban on smoking in public places. The former set assumes
quantitative values. The latter set does not admit of easy quantification, though
some of these can be categorised into groups that can then be assigned
quantitative values. Research strategies thus adopt two approaches, quantitative
and qualitative. We shall deal with the quantitative approach in this unit. The
various measurement scales and scaling techniques to measure the qualitative
data will be discussed in Unit 8.
The basic ingredient of quantitative research is the measurement in quantitative

terms of the variables involved or the collection of the data relevant for the
analytical and interpretative processes that constitute research. The quality of
the data utilised in research is important because the use of faulty data in such
endeavour results in misleading conclusions, however sophisticated may be the
analytical tools used for analysis. Research processes, be it testing of
hypotheses and models or providing the theoretical basis for policy or review
of policy, call for objectivity, integrity and analytical rigour in order to ensure
academic and professional acceptability and, above all, an effective tool to
tackle the problem at hand. Data used for research should, therefore, reflect, as
accurately as possible, the phenomena these seek to measure and be free from
errors, bias and subjectivity. Collection of data has thus to be made on a
scientific basis.
7.2 AN OVERVIEW OF THE UNIT

How to assemble data on a scientific basis? There are broadly three different
methods of collecting data. These are dealt with in section 7.3. The tools that
one can use for collecting data – the formats and the devices that modern
technology has provided – are enumerated in section 7.4. There are situations
where it is considered desirable to gather data from only a part of the universe,
or a sample selected from the universe of interest to the study at hand, rather
than a complete coverage of the universe, for reasons of cost, convenience,
expediency, speed and effort. Questions then arise as to the manner in which
such a sample should be chosen – the sampling design. This question is
examined in detail in section 7.5. The discussion is divided into a number of
sub-topics. Concepts relating to population aggregates like mean and variance
and similar aggregates from the sample and the use of the latter as estimates of
population aggregates have been introduced in sub-section 7.5.1. There are two
types of sampling: random and non random. Non random sampling methods
and the contexts in which these are used are described in sub-section 7.5.2. A
random sample has certain advantages over a non random sample — it
provides a basis for drawing valid conclusions from the sample about the
parent population. It enables us to state the precision of the estimates of
26
population parameters in terms of (a) the extent of their variation or (b) an Data Collection and
interval within which the value of the population parameter is likely to lie with Sampling Design
a given degree of certainty. Further, it even helps the researcher to determine
the size of the sample to be drawn if his project is subject to a sanctioned
budget and permissible limits of error in the estimate of the population
parameter. These principles are explained in sub-section 7.5.3. Eight methods
of random sampling are then detailed in sub-section 7.5.4. These details relate
to (i) operational procedures for drawing samples, and (ii) expressions for (a)
estimators of parameters and measures of their variation and b) estimators of
such variation where the population variation parameter is not known. Different
sampling procedures are also compared, as we go along, in terms of the relative
precision of the estimates, they generate. Finally, the question of choosing the
sampling method that is appropriate to a given research context is addressed in
section 7.6. A short summing up of the Unit is given in section 7.7.
7.3 METHOD OF DATA COLLECTION

There are three methods of data collection – the Census and Survey Method,
the Observation Method and the Experimental Method. The first is a carefully
planned and organised study or enquiry to collect data on the subject of the
study/enquiry. We might for instance organise a study on the prevalence of the
smoking habit among high school children – those aged 14 to 17 - in a certain
city. One approach is to collect data of the kind we wish to collect on the
subject matter of the study from all such children in all the schools in the city.
In other words, we have a complete enumeration or census of the population
or universe relevant to the enquiry, namely, the city’s high school children
(called the respondent units or informants of the Study) to collect the data we
desire. The other is to confine our attention to a suitably selected part of the
population of high school children of the city, or a sample, for gathering the
data needed. We are then conducting a sample survey. A well known example
of Census enquiry is the Census of Population conducted in the year 2011,
where data on the demographic, economic, social and cultural characteristics of
all persons residing in India were collected. Among sample surveys of note
are the household surveys conducted by the National Sample Survey
Organisation (NSSO) of the Government of India that collect data on the socio-
economic characteristics of a sample of households spread across the country.
The Observation Method records data as things occur, making use of an
appropriate and accepted method of measurement. An example is to record the
body temperature of a patient every hour or a patient’s blood pressure, pulse
rate, blood sugar levels or the lipid profile at specified intervals. Other examples
are the daily recording of a location’s maximum and minimum temperatures,
rainfall during the South West / North East monsoon every year in an area, etc.
The Experimental Method collects data through well designed and controlled
statistical experiments. Suppose for example, we wish to know the rate at
which manure is to be applied to crops to maximise yield. This calls for an
experiment, in which all variables other than manure that affect yield, like
water, quality of soil, quality of seed, use of insecticides and so on, need to be
controlled so as to evaluate the effect of different levels of manure on the yield.
Other methods of conducting the experiment to achieve the same objective
without controlling “all other factors” also exist. Two branches of statistics –
The Design and Analysis of Experiments and Analysis of Variance — deal
with these.
27
Research Design and
Measurement 7.4 TOOLS OF DATA COLLECTION
How do we collect data? We translate the data requirements of the proposed
Study into items of information to be collected from the respondent units to be
covered by the study and organise the items into a logical format. Such a
format, setting out the items of information to be collected from the respondent
units, is called the questionnaire or schedule of the study. The questionnaire
has a set of pre-specified questions and the replies to these are recorded either
by the respondents themselves or by the investigators. The questionnaire
approach assumes that the respondent is capable of understanding and
answering the questions all by himself/herself, as the investigator is not
supposed, in this approach, to influence the response in any manner by
interpreting the terms used in the questions. Respondent-bias will have to be
minimised by keeping the questions simple and direct. Often the responses are
sought in the form of “yes”, “no” or “can’t say” or the judgment of the
respondent with reference to the perceived quality of a service is graded, like,
“good”, “satisfactory” or “unsatisfactory”.
In the schedule approach on the other hand, the questions are detailed. The
exact form of the question to be asked of the respondent is not given to the
respondent and the task of asking and eliciting the information required in the
schedule is left to the investigator. Backed by his training and the instructions
given to him, the investigator uses his ingenuity in explaining the concepts and
definitions to respondents to obtain reliable information. This does not mean
that investigator-bias is more in the schedule approach than in the questionnaire
approach. Intensive training of investigators is necessary to ensure that such a
bias does not affect the responses from respondents.
Schedules and questionnaires are used for collecting data in a number of ways.
Data may be collected by personally contacting the respondents of the survey.
Interviews can also be conducted over the telephone and the responses of the
respondent recorded by the investigator. The advent of modern electronic and
telecommunications technology enables interviews being done through e-mails
or by ‘chatting’ over the internet. The mail method is one where (usually)
questionnaires are mailed to the respondents of the survey and replies received
by mail through (postage pre-paid) business-reply envelopes. The respondents
can also be asked (usually by radio or television channels or even print media)
to send their replies by SMS to a mobile telephone number or to an e-mail
address.
Collection of data can also be done through mechanical, electro-mechanical or

electronic devices. Data on arrival and departure times of workers are obtained
through a mechanical device. The time taken by a product to roll off the
assembly line and the time taken by it to pass through different work stations
are recorded by timers. A large number of instruments are used for collecting
data on weather conditions by meteorological centres across the country that
help assessing current and emerging weather conditions. Electronic Data
Transfers (EDT) can also be the means through which source agencies like
ports and customs houses, where export and import data originate, supply data
to a central agency like the Directorate General of Commercial Intelligence and
Statistics (DGCI&S) for consolidation.
The above methods enable us to collect primary data, that is, data being
collected afresh by the agency conducting the enquiry or study. . The agency
28
concerned can also make use of data on the subject already collected by Data Collection and
another agency or other agencies – secondary data. Secondary data are Sampling Design
published by several agencies, mostly Government agencies, at regular
intervals. These can be collected from the publications / compact discs or the
websites of the agencies concerned. But such data have to be examined
carefully to see whether these are suitable or not for the study at hand before
deciding to collect new data.
Errors in data constitute an important area of concern to data users. Errors can
arise due to confining data collection to a sample. (sampling errors). It can be
due to faulty measurement arising out of lack of clarity about what is to be
measured and how it is measured. Even when these are clear, errors can creep
in due to inaccurate measurement. Investigator bias also leads to errors in data.
Failure to collect data from respondent units of the population or the sample
due to omission by the investigator or due to non-response (respondents not
furnishing the required information) also results in errors. (non-sampling
errors). The total survey error made up of these two types of errors need to be
minimised to ensure quality of data.
7.5 SAMPLING DESIGN

We have looked at methods and tools of data collection, chief among which is
the sample survey. How to select a sample for the survey to be conducted?
There are a number of methods of choosing a sample from a universe. These
consist of two categories, random sampling and non-random sampling. Let us
turn these methods and see how well the results from the sample can be utilised
to draw conclusions about the parent universe.
But first let us turn to some notations, concepts and definitions.
7.5.1 Population and Sample Aggregates and Inference

Let us denote population characteristics by upper case (capital) letters in
English or Greek and sample characteristics by lower case (small) letters in
English. Let us consider a (finite) population consisting of N units Ui (i =
1,2,….N). Let Yi (i = 1,2,….N) be the value of the variable y, the characteristic
under study, for the ith unit Ui (i = 1,2,…..N). For instance, the units may be the
students of a university and y may be their weight in kilograms. Any function
of the population values Yi is called a parameter. An example is the population
mean ‘µ’ or ‘M’ given by (1/N) iYi, where i stands for summation over i = 1
to N. Let us now draw a sample of ‘n’ units ui (i = 1,2,…..n)1 from the above
population and let the value of the ith sample unit be yi (i = 1,2,…..n)2. In other
words, yi (i = 1,2,….n) are the sample observations. A function of the sample
observations is referred to as a statistic. The sample mean ‘m’ given by
(1/n) iyi , i (i =1 to n), is an example of a statistic.
Let us note the formulae for some important parameters and statistics.
Population total Y = iYi, i stands for summation over i = 1 to N (2.1)
Population mean = ‘µ’ or ‘M’ , = (1/N) iYi, i , i = 1 to N (2.2)
1
The sample units are being referred to as ui (i = 1,2,…..n) and not in terms of Ui as we do not
know which of the population units have got included in the sample. Each ui in the sample is
some population unit
2
The same reasons apply for referring the sample values or observations as yi (i = 1,2,…n) and
not in terms of the population values Yi . yi wil be some Yi. 29
Research Design and
Measurement Population variance σ2 = (1/N) iYi
2
– M2 , i , i = 1 to N (2.3)
Population SD = σ = +√[(1/N) iYi
2
– M2] , i , i = 1 to N (2.4)
Sample mean = (1/n) iyi , i , (i =1 to n) (2.5)
Sample variance s2 = (1/n) i yi 2 – m2 , i i = 1 to n (2.6)
= [ss]2 /n , where [ss]2 = (yi – m)2 = yi 2 – n m2 =
sum of squares of sample observations from their mean ‘m’ (2.7)
Sample standard deviation ‘s’ = +√[(1/n) i yi 2 – m2 ], i , i = 1 to n (2.8)
Population proportion P = (1/N) iYi = Nı / N, (where Nı is the number
of units in the population possessing a specified characteristic) (2.9)
σ2 = (1/N) i Yi2 – M2 = P – P2 = P(1 – P) = PQ,
where Q = [(N – Nı)/N] = (1 – P). (2.10)
m = p, (proportion of units in the sample with the specific characteristic) (2.11)
s2 = p(1 – p) = pq, where p is the sample proportion and p + q = 1 (2.12)
2
[ss] = npq (2.13)
The purpose of drawing a sample from a population is to arrive at some
conclusions about the parent population from the results of the sample. This
process of drawing conclusions or making inferences about the population
from the information contained in a sample chosen from the population is
called inference. Let us see how this process works and what its components
are. The sample mean ‘m’, for example, can serve as an estimate of the value
of the population mean ‘µ’. The statistic ‘m’ is called an estimator (point
estimator) of the population mean ‘µ’. The value of ‘m’ calculated from a
specific sample is called an estimate (point estimate) of the population mean
‘µ’. In general, a function of sample observations, that is, a statistic, which can
be used to estimate the unknown value of a population parameter, is an
estimator of the population parameter. The value of the estimator calculated
from a specific sample is an estimate of the population parameter.
The estimate ‘m1 ’ of the population parameter ‘µ’, computed from a sample,
will most likely be different from ‘µ’. There is thus an error in using ‘m1’ as an
estimate of ‘µ’. This error is the sampling error, assuming that all measurement
errors, biases etc., are absent, that is, there are no non-sampling errors. Let us
draw another sample from the population and compute the estimate ‘m2‘ of ‘µ’.
‘m2 ‘ may be different from ‘m1’ and also from ‘µ’. Supposing we generate in
this manner a number of estimates mi (i = 1,2,3,…….) of ‘µ’ by drawing
repeated samples from the population. All these mi (i = 1,2,3,….) would be
different from each other and from ‘µ’. What is the extent of the variability in
the mi (i = 1,2,3,….), or, the variability of the error in the estimate of ‘µ’
computed from different samples? How will these values be spread or scattered
around the value of ‘µ’ or the errors be scattered around zero? What can we say
about the estimate of the parameter obtained from the specific sample that we
have drawn from the population as a means of measuring the parameter,
without actually drawing repeated samples? How well do non-random and
random samples answer these questions? The answers to these questions are
important from the point of view of inference.
30
Let us first look at the different methods of non-random sampling and then Data Collection and
move on to random sampling. Sampling Design
7.5.2 Non-Random Sampling

There are several kinds of non-random sampling. A judgment sample is a
sample that has been selected by making use of one’s expert knowledge of the
population or the universe under consideration. It can be useful in some
circumstances. An auditor for example could decide, on the basis of his
experience, on what kind of transactions of an institution he would examine so
as to draw conclusions about the quality of financial management of an
institution. Convenience Sampling is used in exploratory research to get a
broad idea of the characteristic under investigation. An example is one that
consists of some of those coming out of a movie theatre; and these persons may
be asked to give their opinion of the movie they had just seen. Another
example is one consisting of those passers by in a shopping mall whom the
investigator is able to meet. They may be asked to give their opinion on a
certain television programme. The point here is the convenience of the
researcher in choosing the sample. Purposive Sampling is much similar to
judgement sampling and is also made use of in preliminary research. Such a
sample is one that is made up of a group of people specially picked up for a
given purpose. In Quota Sampling, subgroups or strata of the universe (and
their shares in the universe) are identified. A convenience or a judgement
sample is then selected from each stratum. No effort is made in these types of
sampling to contact members of the universe who are difficult to reach. In
Heterogeneity Sampling units are chosen to include all opinions or views.
Snowball Sampling is used when dealing with a rare characteristic. In such
cases, contacting respondent units would be difficult and costly. This method
relies on referrals from initial respondents to generate additional respondents.
This technique enables one to access social groups that are relatively invisible
and vulnerable. This method can lower search costs substantially but this
saving in cost is at the expense of the representative character of the sample.
An example of this method of sampling is to find a rare genetic trait in a person
and to start tracing his lineage to understand the origin, inheritance and
etiology of the disease.
It would be evident from the description of the methods given above that the
relationship between the sample and the parent universe not clear. The
selection of specific units for inclusion in the sample seem to be subjective and
discretionary in nature and, therefore, may well reflect the researcher’s or the
investigator’s attitudes and bias with reference to the subject of the enquiry.
A sample has to be representative of the population from which it has been

selected, if it is to be useful in arriving at conclusions about the parent
population. A representative sample is one that contains the relevant
characteristics of the population in the same proportion as in the population.
Seen from this angle, the non-random sampling methods described above do
not yield representative samples. Such samples are, therefore, not helpful in
drawing valid conclusions about the parent population and the way these
conclusions change when another sample is chosen from the population. Non-
random sampling is, however, useful in certain circumstances. For instance, it
is an inexpensive and quick way to get a preliminary idea of the variable under
study or a rough preliminary estimate of the characteristics of the universe that
helps us to design a scientific enquiry into the problem later. It is thus useful in
exploratory research.
31
Research Design and
Measurement
1) Name the three methods of the data collection
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) What is meant by a ‘parameter’? Defining the term ‘statistic’, indicate the
expressions for population/sample mean and population/sample mean and
population/sample variance.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Mention the most important purpose of studying the population on the basis
of a sample. In this context, define the terms ‘estimator’ and ‘estimate’ with
a suitable example.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) Defining the term ‘representative sample’. Indicate how is ‘random
sampling’ principally different from that of ‘non-random sampling’? What
could be the use of the latter despite its major drawback vis-à-vis the
former?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
7.5.3 Random or Probability Sampling
Random sampling methods, on the other hand, yield samples that are
representative of the parent universe. The selection process in random sampling
is free form the bias of the individuals involved in drawing the sample as the
units of the population are selected at random for inclusion in the sample.
Random sampling is a method of sampling in which each unit in the population
has a predetermined chance (probability) of being included in the sample. A
sampling design is a clear specification of all possible samples of a given type
with their corresponding probabilities. This property of random sampling helps
us to answer the questions we raised at the end of sub-section 2.6.1 above. That
is, we can make estimates of the characteristics of the parent population from
the results of a sample and also indicate the extent of error to which such
estimates are subject or the precision of the estimate. This is better than not
knowing anything at all about the magnitude of the error in our statements
regarding the parent population. Let us see how random sampling helps in this
regard.
32
Data Collection and
Sampling Design
A) Precision of Estimates – Standard Errors and Confidence Intervals
We noted earlier (the last paragraph of sub-section 7.5.1) that the sample mean
(an estimate of the population mean ‘µ’) will have different values in repeated
samples drawn from the population and none of these may be equal to ‘µ’.
Suppose that the repeated samples drawn from the population are random
samples. The sample mean computed from a random sample is a random
variable. So is the sampling error, that is, the difference between‘µ’ and the
sample mean. The values of the sample means (and the corresponding errors in
the estimate of ‘µ’) computed from the repeated random samples drawn from
the population are the values assumed by this random variable with
probabilities associated with drawing the corresponding samples. These will
trace out a frequency distribution that will approach a probability distribution
when the number of random samples drawn increases indefinitely. The
probability distribution of sample means computed from all possible random
samples from the population is called the sampling distribution of the sample
mean. The sampling distribution of the sample mean has a mean and a standard
deviation. The sample mean is said to be an unbiased estimator of the
population mean if the mean of the sampling distribution of the sample mean is
equal to the mean of the parent population, say, µ. In general, an estimator “t”
of a population parameter “θ” is an unbiased estimator of “θ” if the mean of
the sampling distribution of “t”, or the expected value of the random variable
“t”, is equal to “θ”. In other words, the mean of the estimates of the parameter
made from all possible samples drawn from the population will be equal to the
value of the parameter. Otherwise, it is said to be a biased estimate. Supposing
the mean of the sampling distribution of sample mean is Kµ or K+µ, where K
is a constant. The bias in the estimate can be easily corrected in such cases by
adopting m/K or (m – K) as the estimator of the population mean.
The variance of the sampling distribution of the sample mean is called the
sampling variance of the sample mean. The standard deviation of the
sampling distribution of sample means is called the standard error (SE) of the
sample mean. It is also called the standard error of the estimator (of the
population mean), as the sample mean is an estimator of the population mean.
The standard error of the sample mean is a measure of the variability of the
sample mean about the population mean or a measure of the precision of the
sample mean as an estimator of the population mean. The ratio of the standard
deviation of the sampling distribution of sample means and the mean of the
sampling distribution is called the coefficient of variation (CV) of the sample
mean or the relative standard error (RSE) of the sample mean. That is,
CV or RSE = C = standard deviation / mean (2.14)
CV (or RSE) is a free number or is dimension-less, while the mean and the
standard deviation are in the same units as the variable ‘y’. (These definitions
can easily be generalised to the sampling distribution of any sample statistic
and its SE and RSE.)
We have talked about the unbiasedness and precision of the estimate made
from the sample. What more can we say about the precision of the estimate and
other characteristics of the estimate? This is possible if we know the nature of
the sampling distribution of the estimate.
33
Research Design and
Measurement
The nature of the sampling distribution of, say, the sample mean, or for that
matter any statistic, depends on the nature of the population from which the
random sample is drawn. If the parent population has a normal distribution
with mean µ and variance σ2 or, in short notation, N (µ, σ2), the sampling
distribution of the sample mean, based on a random sample drawn from this, is
N (µ, σ2/n). In other words, the variability of the sample mean is much smaller
than that of the variable of the population and it also decreases as the sample
size increases. Thus, the precision of the sample mean as an estimate of the
population mean increases as the sample size increases.
As we know, the normal distribution N (µ, σ2) has the following properties:
i) Approximately 68% of all the values in a normally distributed population
lie within a distance of one standard deviation (plus and minus) from the
mean,
ii) Approximately 95% of all the values in a normally distributed population
lie within a distance of 1.96 standard deviation (plus and minus) of the
mean,
iii) Approximately 99% of all the values in a normally distributed population
lie within a distance of 2.576 standard deviation (plus and minus) of the
mean.
The statement at (iii) above, for instance, is equivalent to saying that the
population mean µ will lie between the observed values (y – 2.576 σ) and (y +
2.576 σ) in 99% of the random samples drawn from the population N(µ, σ2).
Applying this to the sampling distribution of the sample mean, which is N(µ,
σ2/n), we can say that
Pr.[(m – 2.576 σ / √n) < µ < (m+ 2.576 σ / √n )] = 0.99 (2.15)
or that the population mean µ will lie between the limits computed from the
sample, namely, (m – 2.576 σ/√n) and (m+ 2.576 σ√n) in 99% of the samples
drawn from the population. This is an interval estimate, or a confidence
interval, for the parameter with a confidence coefficient of 99% derived from
the sample.
The general rule for constructing a confidence interval of the population

mean with a confidence coefficient of 99% is: the lower limit of the
confidence interval is given by the “estimate of the population mean minus
2.576 times the standard error of the estimate” and the upper limit of the
interval by the “estimate plus 2.576 times the standard error of the estimate”.
(2.16)
B) Assessment of Precision – Unknown Population Variance

If the parent population is distributed as N (M, σ2) and σ2 is not known, we
make use of an estimate of σ2. The statistic ‘s2 ‘ given in formula 2.6 can be
one such, but this is not an unbiased estimate of σ2 as E (s2) = [(n – 1)/ n] σ2.
We, therefore, by using (2.6) and (2.7) have:
v(y)=[ns2 /(n – 1)] as an unbiased estimate of σ2. (2.17)
v(y) = [1/(n – 1)] [ss]2 or, v(y) = [(1/(n - 1)][ i yi 2 – nm2 ] (2.18)
34
As the sampling variance of the sample mean ‘m’ is σ2/n, an unbiased estimate Data Collection and
Sampling Design
v(m) of the sampling variance will be v(y)/n. Let us now consider the statistic
defined by the ratio,
t = (m – M) / [√v(y) / √n] (2.19)
The numerator is a random variable distributed as N(0, σ2/n) and the

denominator is the square root of the unbiased estimate of its variance. The
sampling distribution of the statistic ‘t’ is the Student’s t-distribution with
(n – 1) degrees of freedom. It is a symmetric distribution. A confidence interval
can now be constructed for the population mean M from the selected random
sample, say with a confidence coefficient of (1 - α )%. The values of ‘tα ‘ for
different values of α = Pr.[t > tα ] + Pr.[(- t) < (- tα)] = 2 Pr.[ t > tα ] and
different degrees of freedom have been tabulated in, for instance, Rao, C.R.
and Others (1966). The confidence interval with a confidence coefficient
(1 - α) for the population mean M would be as in 2.19 below – easily computed
from the sample observations.
[m – tα v(m) < M < m + tα v(m)] (2.20)
We note that the rule 2.16 applies here also except that we use (i) the square
root of the unbiased estimate of the sampling variance of the estimate of the
population mean in the place of the standard error of the estimate of the
population mean, and (ii) the relevant value of the ‘t’ distribution instead of
the normal distribution ……… (2.21)
We have so far dealt with parent populations that are normally distributed.
What will be the nature of the sampling distribution of the sample mean when
the parent population is not normally distributed? We examine this question in
the next sub-section C.
C) Assessment of Precision–Parent Population has a Non-Normal

Distribution
The Central Limit Theorem ensures that, even if the population distribution is
not normal,
• the sampling distribution of the sample mean will have a mean equal to the
population mean regardless of the sample size, and
• as the sample size increases, the sampling distribution of the sample mean
approaches the normal distribution.
Thus for large ‘n’ (sample size), say 30 or more, we can proceed with the steps
mentioned in sub-section A above. Further, the Student’s t-distribution also
approaches the normal distribution as ‘n’ becomes large so that we can use the
statistic ‘t’ in sub-section B as a normally distributed variable with mean 0 and
unit variance for samples of size 30 or more. We may then adopt the procedure
outlined in sub-section A.
D) Determination of Sample Size

Random sampling methods also help in determining the sample size that is
required to attain a desired level of precision. This is possible because the
standard error and the coefficient of variation C.V. of the estimate, say, sample
mean ‘m’, are functions of ‘n’, the sample size. C.V. is usually very stable over
the years and its value available from past data can be used for determining the
35
Research Design and
Measurement
sample size. We can specify the value of C.V. of the sample mean that we
desire as, say, C(m) and calculate the sample size with the help of prior
knowledge of the population C.V., namely, C. That is,
C(m) = C /√n ; so that √n = C/C(m), or, n = [C/C(m)]2 (2.22)
Or we can define the desired precision in terms of the error that we can
tolerate in our estimate of ‘M’ (permissible error) and link it with the desired
value of C(m). Then,
n = [2.576C / e]2 , where the permissible error e = (m – M)/ M. (2.23)
If the sanctioned budget is F for the survey: Let the cost function be of the
form F0 + F1 n, consisting of two components – overhead cost and cost per unit
to be surveyed. As this is fixed as F, F = F0 + F1n, and the sample size
becomes n = (F – F0 )/F1. The coefficient of variation of C(m) is not at our
choice in this situation since it gets fixed once ‘n’ is determined. We can,
however, determine the error in the estimate of ‘m’ from this sample (in terms
of the RSE of m), if the population CV, C is known. If further we suppose that
the loss in terms of money is proportional to the value of RSE of m, say, Rs. ‘l’
per 1% of RSE of ‘m’, the total cost of the survey becomes, L(n) = F0 + F1n +
l (C/ √ n). We can then determine the sample size that minimises this new cost
(which includes the cost arising out of loss). Differentiating L(n) w.r.t n and
equating to zero and simplifying,
n = [(l/2) (C/F1)]2/3 (2.24)
See also the sub-section below on stratified sampling.
7.5.4 Methods of Random Sampling

We have so far dealt with random samples drawn form a population. We did
not specify the size of the population. We had assumed that the population is
infinite in size. In practice, a population may have a size N, however, large. Let
us, therefore, consider drawing random samples of size ‘n’ from a population
of size ‘N’. We shall consider the following methods of random sampling:
a) Simple Random Sampling (With Replacement) [SRSWR],
b) Simple Random Sampling (Without Replacement) [SRSWOR]
c) Interpenetrating Sub-Samples (I-PSS),
d) Systematic Sampling (sys),
e) Sampling with Probability Proportional to Size (pps)
f) Stratified Sampling (sts),
g) Cluster Sampling (cs) and
h) Multi-Stage Sampling (mss)
We shall indicate in the following sections a description of the above methods,

the relevant operational procedure for drawing a sample and the expressions/
formulae for (a) the estimator of the population mean/total/ proportion, (b) the
sampling variance of the sample mean/total/population and (c) unbiased
estimate of the sampling variance.
36
Check Your Progress 2 Data Collection and
Sampling Design
1) Briefly describe the meaning of the terms ‘sampling distribution of a
statistic’ and also what is meant by unbiased estimator of a parameter.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) How is random sampling procedure helpful in correcting for the bias of an
estimate? Illustrate this with the help of an example.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Explain the terms ‘coefficient of variation’ and ‘relative standard error’.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
7.5.4.1 Simple Random Sampling with Replacement (SRSWR)

The method: This method of drawing samples at random ensures that (i) each
item in the population has an equal chance of being included in the sample and
(ii) each possible sample has an equal probability of getting selected. Let us
select a sample of ‘n’ units from a population of ‘N’ units by simple random
sampling with replacement (SRSWR). We select the first unit at random, note
its identity particulars for collection of data and place it back in the population.
We choose at random another unit – this could turn out to be the same unit
selected earlier or a different one, note its identity particulars and place it back.
We repeat this process ‘n’ times to get an SRSWR sample of size ‘n’. In such a
sample one or more units may occur more than once. A sample of ‘n’ distinct
units is also possible. It can be shown that the number of possible samples that
can be selected by SRSWR method is N n and that the probability of any one
sample being chosen is 1/ N n.
Operational procedure for selection of the sample by SRSWR method:

Tables of Random Numbers are used for drawing random samples. These
tables contain a series of four-digit (or five-digit or ten-digit) random numbers.
Supposing a sample of 30 units is to be selected out of a population of 3000
units. First allot one number from the set of numbers to 0001 to 3000 as the
identification number to each one of the population units. The problem of
drawing the sample of size 30 then reduces to that of selecting 30 random
numbers, one after another, from the random number tables. Turn to a page of
the Tables at random and start noting down, from the first left-most column of
(four or five or ten-digit) random numbers, the first four digits of the numbers
from the top of the column downwards. Continue this operation on to the
second column till the required sample size of 30 is selected. If any of the
random numbers that comes up is more than 3000, reject it. If some numbers
(< 3000) get repeated in the process, it means that the corresponding units of
the population would be selected more than once, this being sampling with
replacement.
37
Research Design and
Measurement
Estimators from SRSWR samples (using notations set down earlier):
msrswr = (1/n) i yi ( i , i = 1 to n) is an unbiased estimator for M (2.25)
Note: If a unit gets selected in the sample more than once, the corresponding
value of yi will also have to be repeated as many times in the summation for
calculating msrswr .
Sampling Variance of msrswr : V(msrswr) = σ2/n = [1/n] [E(yi 2 ) – M2 ] (2.26)

Standard Error of msrswr : SE(msrswr) = σ / √n (2.27)
CV or RSE of msrswr : C(msrswr) = (1/ √n)(σ / M) = C(y)/√n. (2.28)
Note that the sampling variance, SE and CV(RSE) of the sample mean in
SRSWR is much less than SE and CV of the variable y and these decrease as
the sample size increases. The precision of the sample mean in SRSWR, as an
estimator of M increases as the sample size increases. However, the extent of
decrease in the standard error will not be commensurate with the size of the
increase in the sample size. We would need an unbiased estimator of σ2, as σ2
may not be known. This is
v(y) = [1/(n – 1)][ss]2 (2.29);

Therefore, v(msrswr) = v(y)/n (2.30)
an unbiased estimate of Y, or, Y*srswr = Nm (2.31)
V(Y*srswr) = N2 (σ2 /n) (2.32)
v(Y*srswr) = N2 (1/n)[1/(n – 1)] (yi – msrswr)2 (2.33)
the sample proportion ‘psrswr’ is an unbiased estimate of P (2.34)
V(psrswr) is PQ/n (2.35); and v(psrswr) = npq/(n – 1) (2.36)
C(psrswr )-( = √[(1/n)(PQ)]/ P = [1/√n)]√[Q/P]. (2.37)
Confidence intervals for the population mean/proportion and the sample size
for a given level of precision and/or permissible error can now be derived easily.
7.5.4.2 Simple Random Sampling without Replacement (SRSWOR)

The Method: This method of sampling is the same as SRSWR but for one
difference. If a unit is selected, it is not placed back before the next one is
selected. This means that no unit gets repeated in a sample. Operationally, we
draw random numbers between 1 and N and if a random number comes up
again, it is rejected and another random number is selected. This process is
repeated till ‘n’ distinct units are selected. It can be shown that the number of
samples of size n that may be selected from a population of ‘N’ units by this
method is NCn = N ! /[ (N – n) ! n ! ] =[N(N –1) (N – 2) …(N – n + 1)] / [n(n – 1)
(n – 2)…….1]. The probability Psrswor(S) of any one of the samples being
chosen is, 1/ NCn .
Estimators from SRSWOR samples:

msrswor = (1/n) i yi , i , i =1 to n, is an unbiased estimator of M (2.38)
V(msrswor) = [(N – n)/ (N – 1)] [σ2/n]
= [(N – n)/(N – 1)][1/n][(1/N) i(Yi – M)2], i, i=1 to N (2.39)
38
V(m)srswor < V(msrswr) since (N – n) / (N – 1) is less than 1 for n > 1, (2.40) Data Collection and
Sampling Design
Both msrswor and msrswr are unbiased estimators of M but msrswor is a more
efficient estimator of M than msrswr . The factor [(N – n)/(N – 1)] in (2.40) is
called the finite population correction or finite population multiplier. The
finite population correction required for finite population need not, however,
be used when the sampling fraction (n / N) is less than 0.05.
v(msrswor) = [(N – n)/N][1/n][1/(n – 1)][ i (yi – m)2] , i , i= 1 to n

=[(N– n)/N][1/n][1/(n – 1)][ss]2, (2.41)
Unbiased estimate of population total Y, that is Y*srswor = Nmsrswor (2.42)
2
V(Y*srswor) = N V(msrswor ) (2.43)
2
unbiased estimate of V(Y*srswor), namely, v(Y*srswor) = N v(msrswor (2.44)
C(Y*srswor) = C(msrswor ) (2.45)
the sample proportion ‘p’ is an unbiased estimate of ‘P’ in SRSWOR
also. (2.46)
V(p) = [(N – n)/ (N – 1)] [PQ/n] , where P + Q =1 (2.47)
v(p) = [(N – n)/N] [pq/(n – 1)] , where p + q = 1 (2.48)
C(p) = √[(N – n)/(N – 1)] √[(1/n) √[Q/P] and (2.49)
1) A population has 80 units. The relevant variable has a population mean of

8.2 and a variance of 4.41. These SRSWR samples of size (i) 16, (ii) 25 and
(iii) 49 are drawn from the population. What is the standard error (SE) of
the sample means from the three samples? Is the extent of reduction in SE
commensurate with that of the increase in sample size?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) What are the results when the sampling method in drawing the three
samples in problem 1 above is changed to SRSWOR? What is your advice
regarding the choice between increasing the sample size and changing the
sampling method from SRSWR to SRSWOR?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Indicate whether the following statements are true (T) or false (F). If false,
what is the correct position?
1) The standard error of the sample mean decreases in direct proportion
to the sample size. (T/F)
2) SRSWOR method of sampling is more advantageous than SRSWR
for a sampling fraction of 0.02 (T/F)
3) If Y* = Nm and Variance of m is V(m), the variance of Y* is NV(m).

(T/F)
39
Research Design and
Measurement 7.5.4.3 Interpenetrating Sub-Samples (I-PSS)
Suppose a sample is selected in the form of two or more sub-samples drawn
according to the same sampling method so that each such sub-sample provides
a valid estimate of the population parameter. The sub-samples drawn in this
way are called interpenetrating sub-samples (I-PSS). This is operationally
convenient, as the different sub-samples could be allotted to different
investigators. The sub-samples need not be independently selected. There is,
however, an important advantage in selecting independent interpenetrating
sub-samples. It is then possible to easily arrive at an unbiased estimate of the
variance of the estimator even in cases where the sampling method/design is
complex and the formula for the variance of the estimator is complicated.
Let {t i}, i = 1,2,…,h be unbiased estimates of a parameter θ based on ‘h’

independent interpenetrating sub-samples. Then,
t = (1/h) i ti ,( i , i = 1 to h) is an unbiased estimate of θ (2.50)
v(t) = [1/h(h – 1)] [ i (t i – t)2 ], ( i , i = 1 to h) is an unbiased
estimate of V(t) (2.51)
If the unbiased estimator ‘t’ of the parameter θ is symmetrically distributed (for
example, normally distributed), the probability of the parameter θ lying
between the maximum and the minimum of the ‘h’ estimates of θ obtained
from the ‘h’ sub-samples is given by:
Prob.[Min of {t1, t2 ,---- t h }< θ < Max of {t1, t2,-----t h}] = [1–(1/2)( h – 1)](2.52)
This is a confidence interval for θ from the sample. The probability increases
rapidly with the number of I-P sub-samples – from 0.5 (two sub-samples) to
0.875 (four sub-samples).
7.5.4.4 Systematic Sampling

The Method: Let {Ui}, i = 1,2,………. N be the units in a population. Let ‘n’
be the size of the sample to be selected. Let ‘k’ be the integer nearest to N/n -
denoted usually as [N/n] — the reciprocal of the sampling fraction. Let us
choose a random number from 1 to k, say, ‘r’. We then choose the rth unit, that
is, Ur . Thereafter, we select every kth unit. In other words, we select the units
Ur, Ur+k , Ur+2k ,………… This method of sampling is called systematic
sampling with a random start. ‘r’ is known as the random start and ‘k’ the
sampling interval. There would thus be ‘k’ possible systematic samples, each
corresponding to one random start from 1 to k. The sample corresponding to
the random start ‘r’ will be
{Ur+jk }, j = 0,1,2, r+jk ≤ N.

The sample size of all the ‘k’ systematic samples will be ‘n’ if N = nk. All the
‘k’ systematic samples will not have a sample size ‘n’ if N ≠ nk. For example,
if we have a sample of 100 units and we wish to select systematic samples of
size 14, the sampling interval is k = [100/15] or 7 The samples with the
random starting 1 and 2 will be of size 15 while the other 5 systematic samples
(with random starts 3 to 7) will be of size 14.
In systematic sampling, units of a population could thus be selected at a

uniform interval that is measured in time, order or space. We can for instance
choose a sample of nails produced by a machine for five minutes at the interval
40
of every two hours to test whether the machine is turning out nails as per the Data Collection and
desired specifications. Or, we could arrange the income tax returns relating to Sampling Design
an area in the order of increasing gross income returned and select every
fiftieth tax return for a detailed examination of the income of assesses of the
area. Systematic samples are thus operationally easier to draw than SRSWR or
SRSWOR samples. Only one random number needs to be chosen for selecting
a systematic sample.
Estimators from Systematic Samples:

An unbiased estimator of the population mean M based on a systematic
sample is given by a slight variant of the sample mean, namely,
msys* = (k/N) i yi , i , i =1 to n*, n* is the size of the selected sample
and k the sampling interval v(2.53)
If N = nk, msys* = m the sample mean. If N ≠ nk, there is a bias in using the
sample mean as the estimator for M, and
the bias in using the sample mean as an estimator of M is likely to be small in
the case of systematic samples selected from a large population. (2.54)
The disadvantages, referred to above, in systematic sampling, namely, N not
being a multiplier of the sample size n and the sample mean not being an
unbiased estimator of the population mean can be overcome by adopting a
procedure called Circular Systematic Sampling (CSS). If ‘r’ is the random
start, and k the integer nearest to N / n, we choose the units.
{Ur+jk }, if r+jk ≤ N and {Ur+jk - N }, if r+jk > N ; j = 0,1,2,………… ( n - 1).
Taking the earlier example of selecting a systematic sample of size 14 from a

population of 100 units (N == 100, k = 7 and n = 15) all the samples can be
made to have a size of 15 by adopting the CSS. A random start of 5 will lead
to the selection of a sample of the 15 units 5,12,19,26,33,40,47,54,61,
68.75,82,89,96 and 3 (96 + 7 – 100). This procedure ensures equal probability
of selection to every unit in the population.
Besides constancy of the sample size from sample to sample, the CSS
procedure ensures that mr the sample mean is an unbiased estimate of the
population mean. (2.55)
Let nk = N. Then m* = m. There are k possible samples, each sample with a

probability of 1/k. Let the sample mean of the r-th systematic sample be mr =
(1/n) i yir, where yir is the value of the characteristic under study for the i-th
unit in the r-th systematic sample, summation is from i = 1 to n. As already
noted mi is an unbiased estimator of M or E(mr) = M. We thus have k possible
unbiased estimates of M. Denoting the sample mean in systematic sampling as
msys, the sampling variance of msys, and related results of interest are:
V(msys)= σb2 (the between-sample variance). (2.56)

V(msys) = V(y) – σw2 , where σw2 is within-sample variance. (2.57)
Equation 2.57 shows that (i) V(msys) is less than the variance of the variable
under study or the population variance, since σw2 is > 0 and (ii) V(msys) can be
reduced by increasing σw2, or by increasing the within-sample variance. (ii) would
happen if the units within each systematic sample are as heterogeneous as
41
Research Design and
Measurement
possible. Since we select a sample of ‘n’ units from the population of N units
by selecting every k-th element from the random start ‘r’, the population is
divided into ‘n’ groups and we select one unit from each of these ‘n’ groups of
population units. Units within a sample would be heterogeneous if there is
heterogeneity between the ‘n’ groups. This would imply that units within each
of the n groups would have to be as homogeneous as possible. All these
suggest that the sampling variance of the sample mean is related to the
arrangement of the units in the population. This is both an advantage and
disadvantage of systematic sampling. An arrangement that conforms to the
conditions mentioned above would lead to a smaller sampling variance or an
efficient estimate of the population mean while a ‘bad’ arrangement would lead
estimates that are not as efficient.
7.5.4.5 Sampling with Probability Proportional to Size (PPS)

The Sampling Method: We have so far considered sampling methods in
which the probability of each unit in the population getting selected in the
sample was equal. There are also methods of sampling in which the probability
of any unit in the population getting included in the sample varies from unit to
unit. One such method is sampling with probability proportional to size (pps) in
which the probability of selection of a unit is proportional to a given measure
of its size. This measure may be a characteristic related to the variable under
study. One example may be the employment size of a factory in the past year
and the variable under study may be the current year’s output. Does this
method lead to a bias in our results, as units with smaller sizes would be under
represented in the sample and those with larger sizes would be over
represented. It is true that if the sample mean ‘m’ were to be used to estimate
the population mean M, m would be a biased estimator of M. However, what is
done in this method of sampling is to weight the sample observations with
suitable weights at the estimation stage to obtain unbiased estimates of
population parameters, the weights being the probabilities of selection of the
units.
Estimates from pps sample of size 1: Let the population units be {U1,U2, ----
---UN}.Let the main variable Y and the related size variable X associated with
these units be{Y1, X1; Y2, X2; …………YN, XN}. The probability of selecting
any unit, say, Ui in the sample will be Pi = (X i / X), where iX i = X, where
i , i = 1 to N. Let us select one unit by pps method. Let the unit selected thus
have the values y1 and x1 for the variables y and x. The variables y and x are
random variables assuming values Yi and X i respectively with probabilities Pi ,
i = 1,2,…………N. The following results based on the sample of size 1 can be
derived easily:
An unbiased estimator of population total Y is Y*(1) pps = y1 /p1 (2.58)

An unbiased estimator of M is m*(1)pps = (1/N) Y*(1)pps = (1/N)(y1 /p1) (2.59)
V[Y*(1) pps] = i (Yi2 /Pi ) – Y2 (2.60)
V[m*(1)pps ] = (1/N2) V[Y*(1)pps] = (1/N2) [ 2

i (Yi /Pi ) – Y2] (2.61)
These show that the variance of the estimate will be small if the Pi are
proportional to Yi.
Estimators from pps sample of size > 1 [pps with replacement (pps-wr)]
42
A sample of n ( > 1) units with pps can be drawn with or without replacement. Data Collection and
Let us consider a pps-wr sample. Let {yi , pi} be respectively the sample Sampling Design
observation on the selected unit and the initial probability of selection at the i–
th draw, i = 1,2,----n. Each (yi / pi) , i = 1,2,----- n in the sample is an unbiased
estimate [Yi(pps-wr)*] of the population total Y and V(Yi(pps-wr)*) = i (Yr 2 /Pr) –
Y2, r , r = 1 to n. (see 2.60). Estimates from pps-wr samples are:
Y*pps-wr = (1/n) i (yi / pi) = (1/n) i Y*(i)pps-wr ; i , i = 1 to n. (2.62)

V(Y*pps-wr ) = (1/n)[ r (Yr 2 /Pr) – Y2] ; r , r = 1 to N. (2.63)
V(mpps-wr ) = (1/N2 )(1/n)[ r (Yr 2 /Pr) – Y2] ; r , r = 1 to N. (2.64)
v (Y*pps-wr )= [1/{n(n – 1)}][ r (yr2 / pr2 – nY*2]; r , r = 1 to n;
(using 2.51) (2.65)
Operational procedure for drawing a pps-wr sample: The steps are:
1) Cumulate the sizes of the units to arrive at the cumulative totals of the
unit sizes. Thus
Ti - 1 = X1 + X2 + -----+ Xi –1 ; Ti = X1 + X2 +---+ Xi - 1 + Xi = Ti - 1 + Xi ;
i = 1, 2,……..., N.
2) Then choose a random number R between 1 and TN = X1 + X2
+………..+ XN = X .
3) Choose the unit Ui if R lies between Ti - 1 and Ti, that is, if Ti - 1 < R ≤ Ti .
The probability P(Ui) of selecting the i-th unit will thus be P(Ui) = (Ti -
Ti – 1) /TN = Xi / X = Pi
4) Repeat the operation ‘n’ times for selecting a sample of size n with pps-wr.
7.5.5.6 Stratified Sampling
The Method: We might sometimes find it useful to classify the universe into a
number of groups and treat each of these groups as a separate universe for
purposes of sampling. Each of these groups is called a stratum and the process
of grouping stratification. Estimates obtained from each stratum can then be
combined to arrive at estimates for the entire universe. This method is very
useful as (i) it gives estimates not only for the whole universe but also for the
sub-universes and (ii) it affords the choice of different sampling methods for
different strata as appropriate. It is particularly useful when a survey organistion has
regional field offices. This method is called Stratified Sampling.
Let us divide the population (universe) of N units into k strata. Let Ns be the
number of units in the s-th stratum. Ysi be the value of the i-th unit in the s-th
stratum. Let the population mean of the s-th stratum be Ms. Ms = (1/Ns) Ysi,
i, i = 1,2,----,Ns (that is over the units within the s-th stratum) and the
population M is =(1/N) sNs Ms = sWs Ms , where Ws = (Ns / N) and s being
over the strata s = 1,2,……,k). Suppose that we select random samples from
each stratum and the sampling method for different strata are different. Let the
unbiased estimate of the population mean Ms of the s-th stratum be ms.
Denoting ‘st’ for stratified sampling, an unbiased estimator of M is given by
mst = s Ws ms = (1/N) s Ns ms, s , s = 1 to k. (2.66)

2
V(mst) = sWs V(ms) = (1/N2 ) s Ns2 V(ms), s , s= 1 to k (2.67)
43
Research Design and
Measurement Cov.(ms,mr) = 0 for s ⋅≠ r ; (samples from diff. strata are independently
chosen) .. (2.68)
Yst* = sYs* ; s , s = 1 to k. (2.69)
V(Yst*) = sV(Ys*), s, s = 1 to k. (2.70)
Thus estimators with smaller variance (efficient estimators) can be obtained in

stratified sampling if we form the strata in such a way as to minimise intra-
strata or within-strata variation, that is, variance within strata. This would
mean maximising between-strata or inter-strata variation, since the total
variation is made up of within-strata and between-strata variation. In other
words, units in a stratum should be homogeneous.
Stratified sampling enables us to choose the sample we wish to select by

drawing independent samples from each of the different strata in to which we
have grouped the universe. How do we allocate the total sample size ‘n’ among
the different strata? One way is to allocate the sample size to different strata in
proportion to the size of individual strata measured by the number of units in
these strata, namely, Ns, [ s Ns = N, (s= 1 to k)]. This method is especially
appropriate in situations where no information is available except the sizes of
the strata. The sample sizes for the samples from the stratum, say, the s-th
stratum, would then be ns = n(Ns/N) and s ns can easily seen to be equal to ‘n’.
There are other methods like allocation of the sample size among strata in
proportion to the stratum totals of the variable under study, that is, Ys , the
stratum total of the s-th stratum, ‘s’ = 1 to k. We shall not go into the details of
other methods here except one situation, namely, when we have a fixed budget
F sanctioned for the survey. Let the cost function F be of the form F0 + nsFs ,
( s , s = 1 to k), where F0 , ns and Fs are respectively the overhead cost, the
sample size in stratum ‘s’ and the per unit cost of surveying a unit in stratum
‘s’ (s = 1,2,-----, k). We can determine the optimum stratum-wise-sample-size
by minimising the sampling variance of the sample mean (2.67) subject to the
constraint that the cost of the survey is fixed. The stratum-wise optimum
sample size is given by
ns = [(F - F0)] [Ws√(Vs / Fs )] / [ s Ws√(VsFs)] , s = 1, k. (2.71)
The stratum sample size should, therefore, be proportional to Ws√(Vs / Fs). The
minimum variance with the ns so determined is,
Min. V(mst ) = [ Ws√(VsFs)]2 / (F - F0)] (2.72)

1) When is PPS method adopted?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) When will the sampling variance of Ypps * be small?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
44
3) Say True (T) and False (F): Data Collection and
Sampling Design
a) PPS and stratified sampling can be combined with other sampling
methods. (T/F)
b) V(mst) is reduced by ensuring that units within individual strataare
heterogeneous. (T/F)
c) The size of a stratified sample can be allocated among the strate,
the size being the number of population units in a stratum. (T/F)
4) A systematic sample of size 18 has to be selected from a population of 124.
What problems do you face in selecting the sample? Is the sample mean the
unbiased estimator of the population mean M? How do you overcome these
problems?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
7.5.4.7 Cluster Sampling

The method: Supposing we are interested in studying certain characteristics of
individuals in an area. We would naturally select a random sample of
individuals from all the individuals residing in the area and collect the required
information from the selected individuals. We might also think of selecting a
sample of households out of all the households in the area and collect the
required details from all the individuals in the selected households. The
households in the area are clusters of individuals and what we have done is to
select a sample of such clusters and to collect the information needed from all
the individuals in the selected clusters instead of selecting a random sample of
individuals from all persons in the area. What we have done is cluster
sampling. Cluster Sampling is a process of forming suitable clusters of units
and surveying all the units in a sample of clusters selected according to an
appropriate sampling method. The clusters of units are formed by grouping
neighbouring units or units that can be conveniently surveyed together.
Sampling methods like srswr, srswor, systematic sampling, pps and stratified
sampling discussed earlier can be applied to sampling of clusters by treating
clusters themselves as sampling units. The clusters can all be of equal size or
varying sizes, that is, the number of units can be the same, or vary from cluster
to cluster. Clusters can be mutually exclusive, that is, a unit belonging to one
cluster will not belong to any other cluster. They could also be overlapping.
Estimates from cluster sampling: Let us consider a population of NK units

divided into N mutually exclusive clusters of K units each – a case of clusters
of equal size. The population mean M and the cluster means are given
respectively by M = (1/N) sms, s being over clusters s = 1 to N and ms =
(1/K) i Ysi, i being from i = 1 to K within the s-th cluster. Let us draw a
sample of one cluster by srs. The cluster mean mc-srs (the subscript c-srs
denotes cluster sampling with srs) is an unbiased estimate of M. The
sampling variance of the sample cluster mean is
V(mc-srs) = (1/N) s (ms – M)2 = σ b2 = Variance between clusters; s ,

s = 1 to N; (2.73)
45
Research Design and
Measurement
Let us compare V(mc-srs) with the sampling variance of the sample mean when
K units are drawn from NK units by SRSWR method. How does the “sampling
efficiency” of cluster sampling compare with that of SRSWR?. The sampling
efficiency of cluster sampling compared to that of SRSWR, Ec/srswr , is defined
as the ratio of the reciprocals of the sampling variances of the unbiased
estimators of the population mean obtained from the two sampling methods.
The sampling variances and sampling efficiency are
V(msrswr) = (1/K) [(1/NK) s

2
iYsi – M2] = σ2/K (2.74)
σ2 = σw2 + σb2 = within-cluster variance + between-cluster variance. (2.75)
Ec/srswr > 1 if σw2 > (K – 1)σb2 (2.76)
Thus, cluster sampling is more efficient than SRSWR if the within-cluster

variance is larger than (K –1) times the between-cluster variance. Is this likely?
This is not likely as the between-cluster variance will usually be larger than the
within-cluster variance due to within-cluster homogeneity. Cluster sampling is
in general less efficient than sampling of individual units from the point of
view of sampling variance. Sampling of individuals could provide a better
cross section of the population than a sample of clusters since units in a cluster
tend to be similar.
7.5.4.8 Multi-Stage Sampling

We noted in the sub-section on cluster sampling that random sampling of units
directly is more efficient than random sampling of clusters of units. But cluster
sampling is operationally convenient. How to get over this dilemma? We may
first select a random sample of clusters of units and thereafter select a random
sample of individual units from the selected clusters. We are thus selecting a
sample of units, but from selected clusters of units. What we are attempting is a
two-stage sampling. This can thus be a compromise between the efficiency of
direct sampling of units and the relatively less efficient sampling of clusters of
units. This type of sampling would be more efficient than cluster sampling but
less efficient than direct sampling of individual units. In the sampling
procedure now proposed, the clusters of units are the first stage units (fsu) or
the primary stage units (psu). The individual units constitute the second stage
units (ssu) or the ultimate stage units (usu).
This procedure of sampling can also be generalised to multi-stage sampling.

Take for instance a rural household survey. The fsu’s in such a survey may
consist of districts, the ssu’s may be the tehsils or taluks chosen from the
districts selected in the first stage, the third stage units could be the villages
selected from the tehsils or taluks selected in the second stage and the fourth
and the ultimate stage units (usu’s) may be the households selected from the
villages selected in the third stage. Such multi-stage sampling procedures help
in utilising such information related to the variable under study as may be
available in choosing the sampling method appropriate at different stages of
sampling. In a multi-stage sampling, estimates of parameters are built up stage
by stage. For instance, in two-stage sampling, estimates of the sample
aggregates relating to the fsu’s are built up from the ssu’s using the sampling
method adopted for selecting the ssu’s. These estimates are then used with the
sample probabilities of selection of fsu’s to build up estimates of the relevant
population parameters.
46
Data Collection and
7.6 THE CHOICE OF AN APPROPRIATE Sampling Design
SAMPLING METHOD
We have considered a number of random sampling methods in the foregoing
sub-sections. A natural question that arises now is – which method is to be
adopted in a given situation? Let us consider this question, although the
answer to it lies scattered across the foregoing sub-sections. The choice of a
sampling design depends on considerations like a priori information available
about the population, the precision of the estimates that a sampling design can
give, operational convenience and cost considerations.
1) When we do not have any a priori information about the nature of the
population variable under study, SRSWR and SRSWOR would be
appropriate. Both are operationally simple. However, SRSWOR is to be
preferred, since V(msrswor) < V(msrswr). This advantage holds only when the
sampling fraction is not small, or N and n are not large.
2) Systematic sampling is operationally even simpler than SRSWR and
SRSWOR, but it should not be used for sampling from populations where
periodic or cyclic trends/variations exist, though this difficulty can be
overcome if the period of the cycle is known. V(msys) can be reduced if the
units chosen in the sample are as heterogeneous as possible. But this will
call for a rearrangement of the population units before sampling.
3) When additional information is available about the variable ‘y’ under study,
say, on a variable (size variable) ‘x’ related to ‘y’, the pps method should
be preferred. The sampling variance of Y* (or m) gets reduced when the
probability of selection of units Pi = ( Xi /N) are proportional to Yi , that is,
the size Xi is proportional to Yi or the variables x and y are linearly related
to each other and the regression line passes through the origin. In such
cases pps is more efficient than SRSWR. Further, this method can be
utilised along with other sampling methods and their relative efficiencies.
pps is operationally simple. Pps-wor combines the efficiency of SRSWOR
and the efficiency-enhancing capacity of pps. However, most of the
procedures of selection available, estimators and their variance for pps-wor
are complicated and are not commonly used in practice. This is particularly
so in large-scale sample surveys with a small sampling fraction, as in such
cases sampling without replacement does not result in much gain in
efficiency. Hence unless the sample size is small, we should prefer pps-wr.
4) Stratified sampling comes in handy when we wish to get estimates at the
level of sub-populations or regions or groups. This method also gives us the
freedom to choose different sampling methods/designs in different strata as
appropriate to the group (stratum) of the population and the opportunity to
utilise available additional information relating to the stratum. The
sampling variance of estimators can also be brought down by forming the
strata in such a way as to ensure homogeneity of units within individual
strata. In fact, the stratum sizes can be so chosen as to minimise the
variance of estimators, when there is a ceiling on the cost of the survey.
Stratified sampling with SRS, SRSWOR or pps-wr presents a set of
efficient sampling designs.
5) Sometimes, sampling of groups of individual units than direct sampling of
units might be found to be operationally convenient. Supposing it is easier
to get a complete frame of clusters of individual units than that of units or,
47
Research Design and
Measurement
only such a frame, and not that of the units, is available. (e.g. households
are clusters of individuals). In such circumstances, cluster sampling is
adopted. This is in general less efficient than direct sampling of
individual units, as clusters usually consist of homogeneous units. A
compromise between operational convenience and efficiency could be
made by adopting a two-stage sampling design, by selecting a sample of
individual units (second stage units) from sampled clusters (the first stage
units). A multi-stage design would be useful in cases where clusters have to
be selected at more than one stage of sampling.
6) Finally, we can use the technique of independent I-PSS in conjunction with

the chosen sampling design to get at (i) an unbiased estimate of V(m) for
any sampling design or estimator of V(m), however complicated, (ii) a
confidence interval for ‘M’ based only on the I-PSS estimates (when the
population distribution is symmetrical) and (iii) a tool for monitoring the
quality of work of the field staff and agencies.
7) SRSWOR, stratified sampling with SRSWOR and, when available

information permits, pps-wr and stratified sampling with pps-wr, turn out
to be a set of the more efficient and operationally easy designs to choose
from. I-PSS can also be used in these designs where possible and
necessary.

1) We wish to study the wage levels of factory labour. What type of sampling
method would you adopt for the study and why if (a) just a list of factories
is available with the Chief Inspector of Factories of different State
Governments, (b) if the list in (a) above also gives the total number of
employees in the individual factories at the end of last year and (c) the list
also indicates both the kind of product manufactured in the factory along
with the information specified in (b) above.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
7.7 LET US SUM UP

There are broadly three methods of collecting data. The array of tools used for
data collection by such methods has expanded over time with the advent of
modern technology. Confining data collection efforts to a sample from the
population of interest to the study inevitably leads to questions like the use of
random and non-random samples. Judgment sampling, convenience sampling,
purposive sampling, quota sampling and snowball sampling all belong to the
latter group. The absence of a clear relationship between a non-random sample
and the parent universe and the presence of the researcher’s bias in the
selection of the sample render such samples useless for drawing valid
conclusions about the parent population. But these methods are inexpensive
and quick ways of getting a preliminary idea of the universe for use in
designing a detailed enquiry and in exploratory research. Random samples, on
the other hand, are free from such drawbacks and have properties that help in
arriving at valid conclusions about the parent population.
48
The simplest of the sampling methods – SRSWR - ensures equal chance of Data Collection and
selection to every unit of the population and yields a sample in which one or Sampling Design
more units may occur more than once. ‘msrswr’ is an unbiased estimator of M.
Its precision as an estimator of M increases as the sample size increases.
SRSWOR yields a sample of distinct units. ‘msrswor’ is also unbiased for ‘M’.
SRSWOR is a more efficient than SRSWR as V(msrswor) < V(msrswr). But this
advantage disappears when the sampling fraction is small (< 0.05). Both
provide an unbiased estimator of V(m). An operationally convenient procedure
- interpenetrating sub-samples (I-PSS) – also provides an unbiased estimator
of V(m) for any sampling design and estimator for V(m), however complicated.
Systematic sampling is a simple and operationally convenient method used in

large-scale surveys that requires only a random start and the sampling interval
k = [N/n] for drawing the sample. A slight variant of ‘m’ is unbiased for ‘M’.
Circular systematic sampling takes care of problems that arise when N/n is not
an integer.. An unbiased estimate of V(m) is not possible but this problem can
be tackled easily. Systematic sampling is not recommended when there is a
periodic or cyclic variation in the population. This problem too can be
overcome if the period of the cycle is known.
An example of methods where the probability of selection varies from unit to

unit is pps. The “size” could be the value of a variable related to the study
variable. In pps, each yi / pi , where yi is the value of the study variable
associated with the selected unit and pi the probability of selection of the unit,
is an unbiased estimate (Y*) of the population total Y and [(1/N) Y*] an
unbiased estimator of M. As V(Y*) is small if the probabilities Pi are roughly
proportional to Yi , pps sampling is more efficient than SRS if the size
variable x is proportional to y, that is, x and y are linearly related and the
regression line passes through the origin. pps sampling can be done with
SRSWR, SRSWOR or systematic sampling. In pps-srswr, [(1/n) i(yi / pi)], ( i
i = 1 to n), is an unbiased estimator of Y. This being the mean of n independent
unbiased estimates with the same variance V(Y*), v(Y*) can be derived using
the I-PSS technique.
Stratified Sampling is used when (i) estimates are needed for subgroups of a
universe or (ii) the subgroups could be treated as sub-universes. It gives us the
freedom to choose the sampling method as appropriate to each stratum.
Estimates of parameters are available for the sub-universes (strata) and these
can then be combined over the strata to get estimates for the entire universe. SE
of estimates based on stratified sampling can be small if we form the strata in
such a way as to minimise intra-strata variance. Each stratum should thus
consist of homogeneous units, as far as possible. Stratum-wise sample sizes
can also so chosen as to minimise the variance of estimators.
Another operationally convenient sampling method, cluster sampling, is to

sample groups of units or clusters of units at random and collect data from all
the units of the selected clusters. For example, the household is a cluster of
individuals. SRSWR, SRSWOR, pps or systematic sampling can be used for
sampling clusters. Cluster sampling is, in general, less efficient than direct
sampling of units from the point of view of sampling variance. The question
here is one of striking a balance between operational convenience and cost
reduction on the one hand and efficiency of the sampling design on the other.
We could improve the efficiency of cluster sampling by selecting a random

sample of units from each of the selected clusters - introduce another stage of
49
Research Design and
Measurement
sampling. This is two-stage sampling. This would be more efficient than
cluster sampling but less efficient than direct sampling of units. Multi-stage
sampling can also be done. Such designs are commonly used in large-scale
surveys as these facilitate the utilisation of information available and the choice
of appropriate sampling designs at different stages.
Thus while non-random sampling methods are useful in exploratory research

and preliminary work on planning of enquiries, random sampling techniques
lead to valid judgments regarding the universe. Among random sampling
methods, SRSWOR, stratified sampling with SRSWOR and, when available
information permits, pps-wr and stratified sampling with pps-wr, turn out to
be a set of the more efficient and practically useful designs to choose from. I-
PSS can also be used in these designs where possible and necessary.
7.8 SOME USEFUL BOOKS & REFERENCES
Burgess, R.G.(ed) : Field Research: A Sourcebook and Field

(1982) Manual. (Contemporary Social Research 4),
George Allen and Unwin, London.
Des Raj & Chandok : Sampling Theory, Narosa Publishing House,
P(1998) New Delhi.
Levin, Richard I. & : Statistics for Mangement, Fifth Edition,
Rubin, David S. (1991) Prentice-Hall of India (Private) Limited, M-97
Connaught Circus, New Delhi – 110001.
Krishniah, P.R. & Rao, : Handbook of Statistics – Vol. 6 : Sampling,
C.R. (Eds.) (1988) North Holland, Amsterdam.
Mukhopadhyay, : Theory & Methods of Survey Sampling,
Parimal (1998) Prentice-Hall of India Pvt. Ltd., New Delhi.
Murthy, M.N. (1967) : Sampling Theory and Methods, Statistical
Publishing Society, 204, Barrackpore Trunk
Road, Kolkota-700108.
Rao, C.R., Mitra, S.K. : Formulae and Tables for Statistical Work,
and Matthai, A. (1966) Statistical Publishing Society, Kolkota –
700108.
Sampath, S. (2005) : Sampling Theory & Methods, Second Edition,
Narosa Publishing House, New Delhi, Chennai,
Mumbai, Kolkata.
Singh, Kultar (2007) : Quantitative Social Research Methods, Sage
Publications (Pvt.) Limited, New Delhi
Singleton Jr., Royce A. : Approaches to Social Research, 4th Edition,
& Straits, Bruce C. Oxford University Press, New York – Oxford.
(2005)
Viswanathan, P.K. : Business Statistics – An Applied Orientation,
(2007) Darling Kindersely (India) Pvt. Ltd. – licensees
of Pearson Education in South Asia.
50
Data Collection and
7.9 ANSWERS OR HINTS TO CHECK YOUR Sampling Design
PROGRESS EXERCISES
1) Three methods of data collection are: the census and survey method, the
observation method, and the experimental method.
2) A function of population values is a parameter; a function of sample
observations is a statistic.
1 1 1 1
M = å Yi , m = å yi , s 2 = å Yi 2 - M 2 , and s 2 = å yi2 - m 2 ,
N N N N
where M/m is the mean of the population/sample values are the
expressions of population/sample means and population/sample variance.
3) The purpose of drawing/studying a sample is to arrive at inferences about
the characteristics of the unknown population. Thus, the sample mean is
an estimator of the population mean (of the variable under study). The actual
value of the estimator when computed from a sample is the estimate.
4) A representative sample is one that possesses the characteristics of the
population in the same proportion as in the population. A conscious or
subjective selection of units introduces a bias that erodes the sample’s
capacity to be a true representative of the characteristics of the population.
It is for this reason that a random sampling procedure is preferred. Although
non-random sampling is drawn without any regard to the rigors of a random
sampling procedure, it gives a rough idea of the unknown features of the
population under focus. Implemented judiciously, it can serve the objectives
of providing preliminary inputs for planning a detailed enquiry later.
1) Sample observations in a random sample are random variables. A statistic
is a function of sample observations and is, therefore, also a random
variable. This random variable assumes different values in repeated
random samples. These values form a frequency distribution of the
statistic. This frequency distribution will, in the limit, tend to a probability
distribution that gives the probabilities of the statistic assuming different
values. This probability distribution is called the sampling distribution of
the statistic. Let a statistic ‘t’ be an estimator of a population parameter,
say, T. The estimator ‘t’ is an unbiased estimator of the parameter ‘T’, if
the mean of the sampling distribution of ‘t’ is ‘T’. Since the mean of the
sampling distribution of ‘t’ can also be denoted by E(t), we can express
this condition as E(t) = T.
2) Let ‘t’ be an estimator of ‘T’. The sampling distribution of the statistic ‘t’
can be derived only if the statistic is based on a random sample selected
from the population with a parameter ‘T’. The mean of the sampling
distribution can then be derived from the sampling distribution of ‘t’. That
is, E(t) can be derived. If E(t) is ≠ T and E(t) = KT or T+K, we can correct
the estimator ‘t’ as (t/K) or (t – K) to arrive at an unbiased estimator of T.
An example is the unbiased estimator of σ2.
3) See Sub-section 7.5.3 A
51
Research Design and
Measurement
4.41 2.1
1) The answer for (i) would be = = 0.525 . The answer for (ii) is
16 4
4.41 2.1
= = 0.42 . The percentage decrease in SE works out to [(0.525
25 5
– 0.42)x 100] / [0.525] = 20%. The percentage increase in the sample size
9
is = *100 = 56.25% . The extent of increase is larger than the extent of
16
decrease in the standard error. Try to work out (iii) and the compare the
results of (ii) & (iii).
2) Use the formula 2.39 for variance of the sample mean under SRSWOR.
Note that, for the same sample size,
N −n n −1
V ( msrswor ) − V ( msrswr ) = 1 − V ( msrswr ).
N −1 N −1
64
For n =16 and using SRSWOR sampling, SE works out to * 0.525 = 0.425 .
79
Compare it with the SE for n = 25 in problem 1. SE of the sample mean
was reduced to more or less this level by increasing the size of the SRSWR
sample from 16 to 25. Work out (ii) & (iii)
3) a) F. SE(m) decreases in inverse proportion to the square root of the
sample size.
b) F. SRSWOR is more advantageous if the sampling fraction is > 0.5.
c) F. V(Y*) = N2 V(m).
1) When information on an auxiliary variable ‘x’ related to the variable ‘y’
under study is available for each unit of the population.
2) When Pi are proportional toYi . That is, when the variable ‘x’ and ‘y’ are
linearly related to each other and the regression line passes through the origin.
3) (a) T; (b) F: they should be homogeneous; (c) T.
4) The integer ‘k’ nearest to 124/18 is 7. Choose a random start between 1
and 7 for selecting the sample. Since N = 124 is not a multiple of n = 18,
two problems arise. (i) Samples with random start 1 to 5 are of size 18 and
k 7
others of size 17. The statistic
N
∑ yi =
124
∑ yi and not the sample
mean is an unbiased estimate of M. Both the problems at (i) and (ii) are
overcome by adopting the circular systematic sampling procedure.

1) a) State-wise lists are available. Stratified sampling with each State (or part
of it depending on the number of factories) being a stratum is an obvious
choice. We need to collect details relating to workers in the factories.
Within each stratum a cluster of factories can be selected and wage data
relating to all the workers in each selected factory may be collected.
Alternatively, we could adopt a two-stage sampling – select a sample of
factories (fsu’s) and a sample of workers (ssu’s) from the selected
factories, to enhance the efficiency of sampling. SRSWOR would be
appropriate for each stage of selection.
52
Measurement and
UNIT 8 MEASUREMENT AND SCALING Scaling Techniques
TECHNIQUES
Structure
8.0 Objectives
8.1 Introduction
8.2 Concept of Measurement
8.3 Measurement Issues in Research
8.4 Scales of Measurement
8.5 Criteria for Good Measurement
8.5.1 Reliability
8.5.2 Validity
8.5.3 Practicality
8.6 Errors in Measurements
8.7 Scaling Techniques
8.7.1 Comparative Scaling Techniques
8.7.2 Non-Comparative Scaling Techniques
8.8 Let Us Sum Up
8.9 Key Words
8.0 OBJECTIVES
• state the concept of measurement and its need;
• explain the various scales of measurement;
• discuss the various scaling techniques; and
• identify the criteria for good measurement.
8.1 INTRODUCTION
Measurement is the foundation of scientific enquiry. As we know that research
begins with a ‘problem’ or topic. Thinking about a problem results in
identifying concepts that capture the phenomenon being examined. Concepts
are mental ideas representing the phenomenon. The process of
conceptualization involves defining the concepts abstractly in theoretical terms.
Operationalisation of concepts for research purpose involves moving from the
abstract to the empirical level. This underlines the need to measure the various
attributes of the people, the characteristics of objects or phenomenon. Further
issues like decent work, human wellbeing, happiness, quality of education etc.
which have several qualitative and quantitative dimensions are emerging
important issues of research in economics. Such issues are being probed
53
Research Design and through development of indicators and composite indexes which necessarily
Measurement involve measurement of such indicators. Hence, a student of economics is
expected to be well versed in the measurement scales, criteria of good
measurement, and important scaling techniques. This unit throws light on these
issues. Let us begin to discuss the concept of measurement.
8.2 CONCEPT OF MEASUREMENT

Simply speaking the process of assigning numbers to various attributes of
people, objects or concepts is known as measurement. Technically,
measurement is a process of mapping aspects of a domain to other aspects of a
range according to some rule of correspondence.Tyler(1963) defines
measurement as, “assignment of numerals according to rules”. Nunally
(1970)viewed that, “Measurement consists of rules, for assigning numbers to
objects in such a way to represent quantities of attributes”. Thus, Nunally
focuses on both rules and manner in which numbers are assigned to an object.
According to Campbell, “Measurement is defined as the assignment of

numerals to objects or events according to rules”. The fact that numerals can be
assigned under different rules leads to different kinds of scales and different
kinds of measurement.
Thus, measurement is a process to assign numbers or other symbols to

characteristics of objects according to certain rules. Assigning numbers permit
statistical analysis of the resulting data and facilitate the communication of
measurement rules and results. The rules for assigning numbers hence need to
be standardized and uniformly applicable. In the assignment process, there
must be one to one correspondence between the numbers and characteristics
being measured. In this process, variables can be divided into two basic types –
quantitative andqualitative variables.
8.3 MEASUREMENT ISSUES IN RESEARCH

A number of concerns arise in the process of measurement which need to be
addressed by the researcher. Some of the important issues include the following:
1) Whether the underlying characteristics of the concept allows ordering
(ordinal level) or categorizing (nominal level)?
2) Whether the features of the concept are discrete or continuous with fine
gradation?
The answer of these two questions will enable to determine the level of
measurement inherent to the concept. As a general rule, efforts should be
made that measurement represent the highest scale of measurement for a
concept because it allows the use of more powerful statistical techniques
for analysis.
3) Another important issue pertains to number of indicators for measurement
of a concept. Some simple concepts can be measured by one indicator
whereas abstract concepts are measured with more than one indicators.
How many indicators are appropriate to measure a concept is to be decided
by the researcher.
4) Another measurement issue concerns the source of valid and reliable
measure. Hence, considerable attention should be given to identify valid
and reliable measures for the study.
54
5) Measurement should be free from measurement errors like invalidity error Measurement and
and unreliability. Scaling Techniques
6) Measurement validity is distinct from internal validity and external validity

because these two are separate research design issues.
7) A final measurement issue concerns the proper use of available data. For
that, a research should be well aware of all available data sources and the
types of data available with the various data compilation agencies.
8.4 SCALES OF MEASUREMENT

Before discussing different levels of measurement, let us remember that there
are three postulates of measurement: (i) Equalities or identities, (ii) Rank order,
(iii) Additivity.
You were introduced the elementary concept and types of scales of
measurement in Unit 1. Even at the cost of repetition let us recall that Stanley
S.Stevens (1946) in his seminal paper, “On the theory of scales of
measurement” published in “Science” classified types of scales into four
categories: (1) Nominal, (2) Ordinal, (3) Interval, (4) Ratio.
1) Nominal Scale
A qualitative scale without order is called nominal scale. In nominal scale
numbers are used to name identity or classify persons, objects, groups,
gender, industry type. Even if we assign unique numbers to each value, the
numbers do not really mean anything. This scale neither has any specific
order nor it has any value. In case of nominal measurement, statistical
analysis is attempted in terms of counting or frequency, percentage,
proportion, mode or coefficient of contingency. Addition, subtraction,
multiplication and division are not possible under this scale.
2) Ordinal scale
A qualitative scale with order is called ordinal scale. In ordinal scale
numbers denote the rank order of the objects or the individuals. Numbers
are arranged from highest to lowest or lowest to highest. For example,
students may be ranked Ist, 2nd and 3rd in terms of their academic
achievement. The statistical operations which can be applied in ordinal
measurement are median, percentiles and rank correlation coefficient.
Ordinal scales do not provide information about the relative strength of
ranking. This scale does not convey that the distance between the different
rank values is equal. Ordinal scales are not equal interval measurement.
Further this scale does not incorporate absolute zero point.
3) Interval scale
Interval scale includes all the characteristics of the nominal and ordinal
scale of measurement. In interval scale numerically equal distance on the
scale indicate equal distances in the attributes of the object being measured.
For example, a scale represent marks of students using the attributes range
0 to 10, 10 to 20, 20 to 30, 30 to 40 and 40 to 50, and so forth. The mid
point of each range (ie. 5, 15, 25, 35 and 45 etc.) are equidistance from
each other. The data obtained from an interval scale is known as interval
data. The appropriate measures used in this scale are: arithmetic mean,
standard deviation, Karl Pearson’s coefficient of correlation and tests like t-
test and f-test. We cannot apply coefficient of variation in the interval scale.
55
Research Design and 4) Ratio Scale
Measurement
Ratio scale has all the characteristics of nominal, ordinal and interval scale.
It also possesses a conceptually meaningful zero point in which there is a
total absence of the characteristic being measured. Ratio scales are common
among physical sciences rather than among social sciences. The examples
of ratio scales are the measures of height, weight, distance and so on. All
measures of central tendencies that can be used in this scale include
geometric and harmonic means.
The properties of these four scales can be summarized in following tabular

form:
Properties Category Ranking Equal Zero Point

Scale interval
Nominal
Ordinal
Interval
Ratio
Properties Indicates Indicates Indicates Absolute

Scale Difference Direction of Amount of Zero
Difference Difference
Nominal ×
Ordinal × ×
Interval × × ×
Ratio × × × ×
You will notice in the above tables that only the ratio scale meets the criteria for
all four properties of scales of measurement. Interval and Ratio scale data are
sometimes referred to as parametric and Nominal and Ordinal data are referred to
as non-parametric.
Examples of Scales of Measurement –
Scale Example Statistics
Nominal Gender Frequency, percentage, proportion Mode,
Yes-no Coefficient of contingency, chi-square (x2)
Students roll number
Objects/Groups
Ordinal Class Rank, Median, percentile, rank, correlation,
Socio Economic status Range
Academic Achievement
Interval Student grade point, Mean, correlation, Range, Standard
Temperature, Calendar deviation, rank order variance, Karl
dates, Rating Scale pearson’s correlation, t-test, f-test etc.
Ratio Weight, Height, Salary, Mean, Median, Mode, Range, Variance,
Frequency of buying a Standard deviation, coefficient of variation,
product rank order variance, Karl pearson’s
correlation, t-test, f-test
56
Measurement and
8.5 CRITERION FOR GOOD MEASUREMENT Scaling Techniques
There are three major criteria for evaluating measurement:

1) Reliability
2) Validity
3) Practicality
8.5.1 Reliability
Reliability refers to the degree to which the measurement or scale is consistent
or dependable. If we use same construct again and again for measurement, it
would lead to same conclusion. Reliability is consistency in drawing conclusion.
Reliability and Validity
. . ..
. . ..
Reliable and Valid Valid but not Reliable Reliable but not valid
8.5.2 Validity
Validity refers to the extent an instrument or scale tests or measures what it
intends to measure. This means validity is the extent to which differences
found with a measuring instrument reflect true differences among those being
tested. This can be done by seeking other relevant evidences. There are three
types of validity: content validity; criterion validity and construct validity.
Content validity indicates the extent to which it provides coverage of the issues
under study. Criterion validity examines how well a given measure relates to
one or more external criterion based on empirical observations. And, construct
validity explains the variation observed in tests conducted on several
individuals. Construct validity is closely related with factor analysis.
8.5.3 Practicality
From practical view point, measure should be economical, convenient and
interpretable. The measure should not be lengthy and difficult. The measure
can be done by highly specialized persons.
8.6 ERRORS IN MEASUREMENTS

Measurement need to be precise and unambiguous for validity and reliability of
any research study. Hence, measurement should be either free from errors or
have minimum error. Any measurement usually involves two types of errors:
• Measurement invalidity
• Unreliability
Measurement invalidity refers to the degree to which the measure incorrectly

captures the concept.
57
Research Design and Unreliability refers to inconsistency in what the measure produces under
Measurement repeated uses. If any measure, on average give some score for a case on
variable and gives other score, when it is used again, it is said to be unreliable.
The possible sources of error are:

a) Respondent by not expressing strong negative feelings.
b) Situation wherein interviewer and respondent are not in good rapport.
c) Measurer – errors may also creep because of incorrect coding, faulty
tabulations and statistical calculations or behavior or attitude of the
interviewer.
d) Instrument errors may also arise because of the defective measuring
instrument.
1) What do you mean by measurement?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Give examples of interval scale and ratio scale.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) How will you examine the validity of a measurement?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) Identify three measure concerns that need to be addressed in the process of
measurement.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
58
Measurement and
8.7 SCALING TECHNIQUES Scaling Techniques
Scaling describes the way of generation of a continuum upon which measured

objects are located. Broadly, scaling techniques can be classified into two
categories: comparative, and non comparative scaling.
Scaling Techniques
Comparative Non-comparative
Scaling Techniques Scaling Techniques
Paired Rank Constant Continuous Itemised Category Cumulative

Comparison order Sum Rating Scale Rating Scale Scale Scale
8.7.1 Comparative Scaling Techniques

Comparative scales involve the comparison of objects directly with one
another. For example, in a study of consumer preferences for different telecom
services, a respondent may be asked to rank choice according to his
preferences. In this technique, data must be interpreted in relative terms and
have only ordinal or rank order properties. In comparative scales small
differences between stimulus objects can be detected.
Example
Rank the following factors in order of your importance in choosing mobile
services (Assign 1 to most important and 5 to least important factor. Please do
not repeat the ranks.)
Factor: Price, connectivity, call drop, Internet, Download speed.
Rank :……… ……… …… ……. ……….
There are various comparative scaling techniques. Three most commonly used
comparative scaling techniques are discussed here.
a) Paired comparison
b) Rank order
c) Constant sum
a) Paired Comparison
The paired comparison is most commonly used scaling technique. This
method is simply a binary choice. In this method respondents choose the
stimulus or items in each pair that has the greater magnitude on the choice
dimension they are instructed to use. This method is used when the study
requires to distinguish between the two objects. The data obtained in this
method is ordinal in nature.
59
Research Design and Example
Measurement
In a study of consumer preferences for the two brands of milk product i.e.
Amul and Mother Dairy, a consumer is asked to indicate which one of the
two brands he would prefer for personal use:
1) Which milk product of the following you prefer on the basis of taste.
Please tick mark (√).
Amul Mother Dairy
2) Which milk product will you prefer on the basis of packaging? Please
tick mark (√).
Amul Mother Dairy
3) Which product will you prefer on the basis of price? Please tick mark (√).
Amul Mother Dairy
This technique is useful when the researcher wants to compare two or

more than two objects. If there are more than two objects (for example
n objects) to compare, the total comparison will be;
n [n - 1]
Number of comparison = ;
2
n= number of objects.
b) Rank Order
This method is very popular among researchers and provides ordinal data.
In this method, respondents are provided various objects and asked to rank
the objects in the list. Rank order method is less time consuming. In this
method, if there are n objects, only (n-1) decisions need to be made.
Respondent can easily understand the instructions for ranking. This
technique produces ordinal data.
Example
Rank the various brands of mobile phone in order of preference. The most
preferred can be ranked 1, the next as 2 and so on. The least preferred will
have the last rank. No two brands should receive the same rank number.
60
Measurement and
Scaling Techniques
Sl. No. Brand Rank

1 Nokia
2 Motorola
3 Samsung
4 HTC
5 Karbonn
6 Spice
7 Xolo
8 Gionee
9 LG
10 Sony
c) Constant Sum
In constant sum scaling, respondents are asked to allocate a constant sum
of units to a specific set of objects on the basis of pre-defined criterion. If
an object is not important, the respondent can allocate zeropoint and if an
object is most important may allocate maximum points out of the fixed
points. The total fixed points are 100. The total may be taken as some
other value depending on the study.
Example
Allocate preferences regarding a movie based on various predefined
criteria. Respondents were asked to rate each criteria in such a way that the
total becomes 100.
Criteria Respondent Preference

Content Quality 20
Music 20
Sound 15
Story 30
Breadth of Coverage (local, regional 15
Supplements, national & global)
Total 100
The data obtained in this method is considered an ordinal scale. The advantage
of this method is that it provides fine discrimination among objects without
consuming too much time. However, allocation over large number of objects
may create confusion for the respondent. This method cannot be used as a
response strategy with illiterate people and children.
61
Research Design and 8.7.2 Non Comparative Scaling Techniques
Measurement
Non-comparative scaling techniques involve scaling of objects independent to
some specific standard. Respondent evaluate only one object at a time. For
example, in a study of consumer preferences for different telecom services, a
consumer may be asked to rate a list of factors that he/she would consider
while choosing a particular telecom company.
Example
Please rate the following factors you think important in choosing mobile
service. Rate 1 to least important and 5 to most important (rating could be on
any scale. In this example, we have used a 5 point scale).
Factor Price Connectivity Call Download Internet

Drop Speed
Rating
In this scaling technique, data is usually in interval scale. It can be continuous,

metric/numeric also. Some of the commonly used non-comparative scaling
techniques are –
i) Continuous Rating Scale

ii) Itemised Rating Scale
iii) Category Scale
iv) Cumulative Scale
i) Continuous Rating Scale

This rating scale is also known as graphic rating scale.In continuous
rating scale, respondents indicate their rating by marking at appropriate
position on a line. The line is labeled at both ends from one extreme
criterion to the other. The line may contain points 0,10,20,…..,100.
Example
How would you rate a magazine with regard to its quality.
Quality Scale Measurement

Indicators
Content Most_________________x__________________Least
Coverage Above 80 60 40 20 0
Language Most__________x_________________________Least
Presentati Most_________________x__________________Least
on Style
A very large number of ratings are possible if the respondents are literate
to understand and accurately differentiate the objects. The data generated
from continuous rating scale can be treated as numeric and interval data.
The researcher can divide the line into as many categories as desired and
assign scores based on the categories under which the ratings fall.
62
ii) Itemized Rating Scale Measurement and
Scaling Techniques
In this scale, respondents are provided with a scale having numbers/
descriptions associated with each category. The respondents are asked to
select the best fitting category with the object.The commonly used
itemised rating scale are:
a) Likert Type Scale (Summated Scale)

b) Semantic Differential Scale
c) Stapel Scale
a) Likert Type Scale (Summated Scale)

Likert (1932) proposed a simple and straight forward method for scaling
attitudes that is most frequently used today. This scale is also known as
summated rating scale. Summated scales consist of a number of
statements which express either a favourable or unfavourable attitude
towards the given object to which the respondant is asked to respond to
each of the statements in terms of several degrees of agreement or
disagreement. Let us understand this method with an example.
Against the following statements relating to psychological well-being, please

tick mark any one of the following options shown against the statement
Sl. Statements Strongly Agree Undecided Disagree Strongly

No. Agree Disagree
5 4 3 2 1
1 I always live in present
and do not worry about
future.
2 Many people think without
purpose in life but I am not
such type of persons.
3 I generally feel sense of
perfection during the
moments I live in.
4 I am very well in
managing and discharging
most of my
responsibilities.
5 I give importance to the
innovative ideas and
experiences that challenge
my thought about me and
the world.
Depending on the wording of an individual item, an extreme answer of strongly

agree or strongly disagree will indicate the most favorable response on the
underlying attitude measured by the questionnaire. This scale is relatively easy
and quick to compute. The data in this scale is of interval scale.
Some of the important advantages of this scale are: (i) it is relatively easy to
construct as it can be performed without a panel of judges, (ii) it is considered
more reliable because respondents answer each statement included in the
63
Research Design and instrument, (iii) it can easily be used in respondent centred and stimulus
Measurement centred studies, (iv) this takes less time to construct. However, there are several
limitations of this type of scale i.e. (i) with this scale, we can simply examine
whether respondents are more or less favourable to a topic but it is difficult to
tell how much more or less they are, (ii) This does not rise more than to a
ordinal scale structure, (iii) the total score of an individual respondent has little
clear meaning since a given total score can be secured by a variety of answer
patterns.
b) Semantic Differential Scale

This scale was developed by Osgood, suit &Tannenbaum (1957). Semantic
differential scale includes a seven point scale in comparison to Likert’s five
point scale. The semantic differential scale is based on the proposition that an
object can have several implied or suggestive meanings to an expressed
opinion. The semantic differential scale can be scored on either -3 to +3 or 1
to 7. This scale can also provide interesting comparison between products
organizations and so on. The results of this scale are further analysed by the
factor analysis.
Example: A researcher has developed an item seeking your perceptions

regarding magazine. Please mark√on each line that best indicates your
perception. Please be sure to mark every scale and do not omit any scale.
India Today Magazine is
+3 +2 +1 0 -1 -2 -3
Easy to read ____ : ____ : ____:____:_√__:____:____ Hard to read
Modern ____ : ____ : ____:____:_√__:____:____ Old fashion
Rational ____ : ____ : ____:____:____:____:____ Emotional
Unreliable ____ : ____ : ____:____:____:____:____ Reliable
Worthless ____ : ____ : ____:____:____:____:____ Valuable
Unorganized ____ : ____ : ____:____:____:____:____ Organized
Orthodox ____ : ____ : ____:____:____:____:____ Liberal
Vain ____ : ____ : ____:____:____:____:____ Modest
Semantic differential scale provides a very convenient and quick way of

gathering impressions on one or more than one concept. The data generated
from this scale can be considered as numeric in some cases, and can be
summed to arrive total scores adjectives. It must define a single dimension and
each pair must be bipolar opposites labeling the extremes.
Like Likert Type Scale, this scale has also several advantages: (i) it is an
efficient and easy way to secure attitude from a large sample, (ii) the total set
of responses (both direction) gives a comprehensive picture of the meaning of
an object as well as a measure of the subject of the rating, (iii) it is a
standarised technique that can be easily repeated.
However, this technique escapes many of the problems of response distortion

found with more direct methods.
64
c) Stapel Scale Measurement and
Scaling Techniques
Stapel scales are named after John stapel who developed these scales. Stapel
scale consists of a single criterion in the centre with 10 categories numbered
from -5 to +5 without a neutral point (zero). The scale is usually presented
vertically. Negative rating indicates that the respondent inaccurately describes
the object and positive rating indicates that the respondent describes the object
accurately.
Example
Rate the Departmental store by marking (√) on the following factors. +5
indicate that the factor is most accurate for you and -5 indicate that the factor is
most inaccurate for you.
+5 -5
+4 -4
+3 -3
+2 -2
+1 -1
Services Products
-1 +1
-2 +2
-3 +3
-4 +4
-5 +5
In this scale, data generated is interval data. In this method, data can be
collected through telephonic interview. The data obtained from Stapel
scale can be analysed in the same way as semantic differential scale.
iii) Category Scale

In Category scaling method, objects are grouped into a predetermined
number of categories on the basis of their perceived strength along certain
dimension. This scale is a dichotomous scale. This method is useful for
socio demographic questions. The response is this category we get,
typically ‘Yes’ or ‘No’ type. Data in this category is either nominal or
ordinal. Under this category scale, instructions and response tasks are
quick and simple, and many options can be included.
Example
A) Single Category Scale
1) Do you own a house?
Yes No
65
Research Design and 2) Do you own a car?
Measurement
Yes No
3) Do you own a Mobile phone?
Yes No
b) Multiple Category
1) Please indicate your income by marking (√) in the income
group you fall
5000 - 10000
10000 - 15000
15000 - 20000
20000 - 25000
iv) Cumulative Scale (Guttman Scale)

Like other scales, it consists of series of statements to which a respondent
expresses his agreement or disagreement. The statements are in a form of
cumulative series i.e. in a way, an individual who replies favourably to
say item no. 3 also replies favourably to item no. 2 and 1 and so on. The
individual’s score is worked out by counting the number of points
concerning the number of statements he answers favourably. Knowing
the total score, we can estimate as to how a respondent has answered
individual statements constituting cumulative scales.
Let us understand with the illustration of an example.

Here is an example of a Guttman scale - the Bogardus Social Distance
Scale:
(Least extreme)
1) Are you willing to permit immigrants to live in your country?
2) Are you willing to permit immigrants to live in your community?
3) Are you willing to permit immigrants to live in your neighbourhood?
4) Are you willing to permit immigrants to live next door to you?
5) Would you permit your child to marry to an immigrant?
(Most extreme)
Cumulative scale (scalogram analysis) like any other scaling technique has
several advantages as well as limitations. The advantages are – It assures that
only a single dimension of attitude is measured, (ii) Researcher’s subjective
judgement is not allowed to creep in the development of scale since the scale is
determined by the replies of respondents.
66
Its main limitation is that (i) In practice perfect cumulative or unidimesional Measurement and
scale are very rarely found and approximation is used, (ii) Its developmental Scaling Techniques
procedure is cumbersome in comparison to other scaling methods.

1) What is the distinction between comparative measuring scale and non
comparative measuring scale?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) What do you mean byLikert Scale? Give its two examples.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
3) What are the sources of measurement errors?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
8.8 LET US SUM UP

Operationalisation of concepts for research purpose need measurement of
various attributes of the people, the characteristics of objects or phenomenon.
Measurement is a process to assign numbers to the characteristics according to
certain rules. Measurement scales are of four types – nominal, ordinal, interval
and ratio. A number of concerns arise in the process of measurement.
Important among them are: the scale measurement i.e. the validity, reliability
and practicality, number of indicators for measurement etc. Scaling techniques
are broadly of two types – comparative, and non-comparative. Within
comparative category, paired, rank order and constant sum are important
techniques. Within non-comparative techniques, four sub-categories i.e.
continuous rating scale, itemized rating scale, category scale and cumulative
scale have been covered in this unit. All these scaling techniques have their
relative advantages and limitations and are used depending upon the goal and the
nature of characteristics of objects/phenomenon to be measured.
8.9 KEY WORDS

Measurement : The assignment of numerals to objects or events
according to rules.
Nominal : Numbers are used to name identity or classify
persons, objects, groups or gender.
67
Research Design and Ordinal Scale : Scale with order is called ordinal scale.
Measurement
Interval Scale : In this scale numerically equal distance on the
scale indicate equal distances in the attributes of
the object being measured.
Ratio Scale : A conceptually meaningful zero point in which
there is a total absence of the characteristics
being measured.
Validity : The extent to which an instrument or scale tests
or measures what the measurement or scale i.e.
consistent or dependent.
Reliability : The degree to which the measurement or scale
i.e. consistent or dependent.
Practicality : Measure should be economical convenient and
interpretable.
Errors in : Discrepancies between the obtained scores and
Measurement the corresponding scores.
Scaling : The way of generation of a continuum upon
which measured objects are located.
Comparative Scaling The comparison of objects directly with one
another.
Paired Comparison : A binary choice, respondents choose the objects
in each pair that has the greater magnitude on
the choice dimension.
Rank Order : Respondents are provided with various objects
and asked to rank the list of objects.
Constant Sum : Respondents are asked to allocate a constant
sum of units to a specific set of objects with
respect to predefined criterion.
Non Comparative : Scaling of object independently to some specify
Scaling standard.
Continuous Rating Respondents indicate their rating by marking at
Scale appropriate position on a line.
Itemised Rating Scale Respondents are asked to select the best fitting
category with the object.
Likert Scale : Respondents are asked to select the best fitting
category with the object.
Semantic Differential Seven point scale based on the proposition that
Scale an object can have several implied or
suggestive meanings to an expressed opinion.
Stapel Scale : A single criterion in the centre with 10
categories numbered from -5 to +5 without a
neutral point.
Category Scale : Objects are grouped in to a predetermined
number of categories on the basis of their
perceived strength along certain dimension.
68
Measurement and
8.10 SOME USEFUL BOOKS/PAPERS Scaling Techniques
Stevens S.S. (1946): On Theory of Scales of Measurement, Science New Series,

Vol. 103, No. 2684, pp 77-680.
Kothari C.R. (2008): Research Methodology Methods and Techniques, New

Age International Publishers. Chapter 5 page 69 to 82.
Babbie Earl, (2010): The Practice of Social Research, 12th Edition,

WodsworthCengage Learning, CA, USA.
Malhotra, N.K. and S. Dash (2009): Marketing Research, An Applied

Orientation, Pearson Education, New Delhi.
Singh, A.K.: Tests Measurements and Research Method in Behavioural

Sciences, Bharti Bhawan Publishers and Distributors, Patna.
Gregory R.J. (2006): Psychological Testing History, Principles and Applications,

Pearson Education, New Delhi, India.
Michael S. Lewis-Beck (ed) (2004): The Sage Encyclopedia of Social Science

Research Methods, Vol. 2, Page no. 161 to 165, Sage Publications, New Delhi,
India.

PROGRESS EXERCISES
1) See Section 8.2

2) See Section 8.4
3) See Section 8.5
4) See Section 8.3
1) See Sub-section 8.7.1 and 8.7.2
2) See Sub-section 8.7.2 under the subhead Itemised Rating Scale
3) See Section 8.6.
69
MEC-109
Research Methods
Block
3
QUANTITATIVE METHODS-I
UNIT 9
Two Variable Regression Models 5
UNIT 10
Multivariable Regression Models 37
UNIT 11
Measures of Inequality 59
UNIT 12
Construction of Composite Index in Social Sciences 93
Expert Committee
Varanasi New Delhi
Bombay University and Former IGNOU
Chairman ICSSR, New Delhi New Delhi
Faculty of Social Sciences IGNOU
Banaras Hindu University, Varanasi New Delhi
Former Joint Advisor Associate Professor in Economics
Unit Resource Person IGNOU Faculty
(Format, Language and Content Editing)
9 Dr. Anoop Chatterjee Prof. Narayan Prasad
Associate Professor in Economics Professor of Economics
ARSD College (University of Delhi), Delhi IGNOU, New Delhi
10 Shri B.S. Bagla Prof. Narayan Prasad
Associate Professor in Economics Professor of Economics
PGDAV College (University of Delhi), Delhi IGNOU, New Delhi
11 Prof P.K. Chaubey Prof. Narayan Prasad
IIPA, New Delhi Professor of Economics, IGNOU, New Delhi
12 Dr. Sunil K Mishra Prof. Narayan Prasad
Fellow, Institute for Human Development Professor of Economics, IGNOU, New Delhi
New Delhi
Print Production
Mr. Manjit Singh
October, 2015
ISBN-978-81-266-
Printed at :
BLOCK 3 QUANTITATIVE METHODS-I
Economic theory is basically concerned about economic laws. These laws (or
hypotheses) are qualitative statements on the economic behaviour of various
agents both at the micro and macro levels. The validity of such qualitative
statements, however, depends upon their rigorous empirical verification by
means of the available data.
The implicit argument between the economic theory and its empirical
verification is that if the stated laws holds good in the real world situation, it
should be reflected in the displayed pattern of the relevant data. The
identification and the examination of such patterns are conducted by employing
various statistical techniques. The primary statistical tool employed for such a
purpose is the Regression Modeling. Hence, a thorough knowledge of
Regression Models is absolutely essential for conducting any serious research
in Applied Economics.
An economic law or theory essentially postulates some relationship that exist
among the economic variables. Suppose the relationship between two variables
Q and P (i.e. Quantity and Price) is expressed in the form (say linear form)
ܳ ൌ ߙ ൅ ߚܲ. Such an expression amount to an exact linear relationship which
is hard to find in reality. Usually, various non-deterministic factors (called
random factors) affect such a relationship. Therefore, in order to allow for the
effect of such random factors, a random component is incorporated into the
model i.e. ܳ ൌ ߙ ൅ ߚܲ ൅ ܷ, where ܷ is a random variable. With this, the strict
mathematical relationship expressed before becomes statistical in character.
Regression technique can now be applied to estimate the two parameters ߙ and
ߚ by using the actual data. This is the crux of a Two Variable Regression
Model explained in Unit 9.
Unit 10 explains the multiple regression models by including more than two
independent random variables. The same procedure as done in the case of two
variable regression models is applied to examine whether the estimated
parameters correspond to the postulated relationship. The treatment accorded to
the Two Variable Regression Models and the Multiple Regression Models in
this course is first limited to the ordinary least square (OLS) method
subsequently, another method viz., the maximum likelihood method (MLM) is
also explained. The two methods taken together equip a researcher with the
basic tools for undertaking empirical research.
Improvement in well being of the poor has been one of the important goals of
economic policy and to a significant extent it is determined by the growth and
distribution of its income. Distribution patterns have an important bearing on
the relationship between average income and poverty levels. Extreme
inequalities are economically wasteful. Further, income inequalities also
interact with other life-chance inequalities. Hence reducing inequalities has
become priority of public policy. Unit 11 deals with the conceptual and
measurement aspects of income inequality.
The complex social and economic issues like child deprivation, food security,
human development, human wellbeing etc. are difficult to measure in terms of
single variable. They are expression of several indicators. Composite Index is
an important statistical techniques to express the single value of several
interdependent or independent variables. Hence, Unit 12 throws light on
various methods to construct Composite Index.
1 Two Variable Regression
UNIT 9 TWO VARIABLE REGRESSION Models
MODELS
Structure
9.0 Objectives
9.1 Introduction
9.2 The Issue of Linearity
9.3 The Non-deterministic Nature of Regression Model
9.4 Population Regression Function
9.5 Sample Regression Function
9.6 Estimation of Sample Regression Function
9.7 Goodness of Fit
9.8 Functional Forms of Regression Model
9.9 Classical Normal Regression Model
9.10 Hypothesis Testing
9.11 Let Us Sum Up
9.12 Exercises
9.13 Key Words
9.16 Answers or Hints to Exercises
9.0 OBJECTIVES
After reading this unit, you will be able to:
• know the issue of linearity in regression model;
• appreciate the probabilistic nature of the regression model;
• distinguish between the population regression model and the sample
regression model;
• state the assumptions of the classical regression model;
• estimate the unknown population regression parameters with the help of the
sample information;
• explain the concept of goodness of fit;
• use various functional forms in the estimation of the regression model;
• state the role of classical normal regression model; and
• conduct some tests of hypotheses regarding the unknown population
regression parameters.
5
Quantitative Methods-1 T
9.1 INTRODUCTION
The empirical research in economics is concerned with the statistical analysis
of economic relations. Often these relations are expressed in the form of
regression equations involving the dependent variable and the independent
variables. The formulation of an economic relation in the form of a regression
equation is called a regression model in econometrics. In such kind of a
regression model, the variable under study is assumed to be a function of
certain explanatory variables. The major purpose of a regression model is the
estimation of its parameters. In addition, it is also useful for testing hypotheses
and making certain forecasts. However, our concern will be mainly with the
estimation of parameters and testing of some hypotheses. At this stage, we
shall refrain from discussing the issue of forecasts from a regression model. A
regression model that contains one explanatory variable is called a two variable
regression model. In this unit, we shall focus on a two variable regression
model. It may be mentioned here that we are essentially focusing on what is
known as the two variable classical regression model.
9.2 THE ISSUE OF LINEARITY

When we talk about a regression model; at our level, what we have in our mind
is a linear regression model. It is important to have a clear idea about the
concept of linearity in the context of a regression model.
Suppose, we have a regression equation like,

Y = α + βX
Generally, we call this a linear regression equation. By linearity we often mean
a relationship, in which, the dependent variable is a linear function of the
independent variable. In this case, the graphical representation of the regression
equation is a straight line. However, there can be another interpretation of
linearity. Consider the following regression equation,
Y = α + β X + γX 2
Now, this regression equation is linear in parameter, in the sense that the
highest power of any parameter (constant) is one. But this equation is non-
linear in variable because the highest power of the independent variable is two.
In fact, conventionally speaking, it is a quadratic regression equation. In this
way, we can have any number of polynomial regression equations with the
highest power of the independent variable going up to any positive integer.
And all these equations may be linear in parameter.
But consider this regression equation,

Y =α + β 2X
This regression equation is linear in variable, but non-linear in parameter
because the power of β is two.
We should note now that from the point of view of regression analysis, we
shall consider only those models which are linear in parameter, no matter
whether they are linear in variable or not. After all, the main purpose of a
regression model is the estimation of its parameters. These parameters
have to be linear for the purpose of their straightforward estimation, say,
6
1 by least square method. For example, the parameters of the second regression Two Variable Regression
Models
equation presented above, can be easily estimated by a simple extension of the
least square method that we have studied in Unit 8. It should be mentioned here
that there are some regression equations that are non-linear in parameters. By
applying suitable transformations they can be reduced to linear in parameter
regression equations. We may consider the following regression equation
Y = αX β
This regression equation in its present form is non-linear in parameter.
However, by applying log transformation, we obtain the following equation
log Y = log α + log β X
It can be clearly seen that the equation has now been transformed into one that
is linear in the log of parameters. We can easily estimate log values of these
parameters by using the usual least square method and then obtain the
estimated values of the original parameters by applying the antilog procedure.
However, some times, the regression equations can be of non-linear in
parameter type that no transformation can render them to a linear in parameter
form. Such equations should be the basis of non-linear or intrinsically non-
regression models. There is no direct method available for the estimation of the
parameters of this kind of regression models and they are generally estimated
by following some standard iteration procedure. Any discussion on such
iteration procedure, however, is beyond the scope of the present unit.
9.3 THE NON-DETERMINISTIC NATURE OF THE

REGRESSION MODEL
We have already referred to the statistical nature of the regression equation in
the last unit. By the very nature of social science, we cannot expect the
relationships that may exist among different variables to be exact or
deterministic. There is always some random element involved in them. As a
result, for a particular value of the independent variable X, the value of the
dependent variable Y cannot be exactly determined from such a relationship.
Let us consider the example of a consumption function. Here, for a given level
of income, when we try to identify the corresponding level of consumption, in
all probability, we shall not be able to obtain a definite value of consumption;
instead, a host of values will be available. And this will be the case with all
possible levels of income. The diagram below shows this.
Consumption
Income
Fig. 9.1: Bi-Variate Population
7
Quantitative Methods-1 Thus, the dependent variable Y tends to be probabilistic or stochastic in nature. T
In fact, for each value of the independent variable X, there will be a distribution
of the values of the dependent variable Y. Accordingly; we specify the
regression model by incorporating a random or stochastic variable. As
mentioned in the previous unit, it should be clear that in our formulation the
dependent variable is stochastic but the independent or explanatory variables
are non-stochastic in nature. We should point out here that at an advanced
treatment of the regression model, even the explanatory variables are
considered to be stochastic in nature. It is the element of randomness either in
the dependent variable alone or both in the dependent and independent
variables, that make the regression model non-deterministic in nature.
9.4 POPULATION REGRESSION FUNCTION

The first step in the regression model is the conceptualization of a population
regression function. Let us assume that there is a bi-variate population of (X, Y)
of size N. Suppose, in this population, Y is the dependent variable and X is the
independent variable. Let X and Y be related to each in a linear fashion in the
following way:
Y = α + βX + U
The above equation is called the population regression function. Here, α and β
are the unknown parameters. It is clear from our discussion on linearity that
this function is linear in variable as well as in parameter.
In this population regression function, the variable U deserves some attention.

We have already mentioned that in social Science, there cannot be an exact
relationship and it is at the most statistical in nature. In our formulation, in fact,
Y is stochastic or random, but X is non-stochastic or deterministic in nature.
Consequently, a random variable like U is introduced in the population
regression function to incorporate this element of randomness of a statistical
relationship. The random variable U is also called the disturbance term. It is a
sort of catch – all variables that represent all kinds of indeterminacies of an
inexact relationship. Thus it may represent, for example,
a) Inherent randomness in human behaviour: The unpredictability of human

psyche is also reflected in the human behaviour. As a result, even after
taking into account of all possible factors, a regression equation cannot
fully explain the value of the dependent variable for the given values of all
possible explanatory variables.
b) Effect of omitted variables: Sometimes for the sake of parsimony, all the
explanatory variables are not included in a regression model. The
disturbance term can be taken as a representative variable of all such
omitted variables.
c) Effect of measurement error: Sometimes, it is not possible to measure the
values of the dependent variable accurately. Consequently, the disturbance
term is introduced in the regression model to represent the combined effect
of all such possible sources of measurement error.
d) Error in the formulation: Often, the functional form of the regression
model proves to be far from a correct depiction of the underlying
relationship between the dependent variable and the independent variables.
8
1 We incorporate the disturbance term to correct the distortions that may Two Variable Regression
Models
arise from a wrong specification of the regression model.
We should note that apart from the above-mentioned factors, there might be
other unidentifiable factors that might also be influencing the dependent
variable in an unknown manner. We introduce the disturbance term for
incorporating the effect of all such unknown random factors. It should be clear
that for some of the reasons mentioned above the disturbance term may assume
a positive value and for some other factor it may tend to be negative.
Consequently, for all practical purposes, the net or the mean effect of the
random disturbance can be taken to be zero.
Difference between the Disturbance Term and the Intercept

There is a difference between the interpretation of the disturbance term and that
of the intercept term α in our regression equation. As we have noted, that the
disturbance represents the random effects of the known and unknown variables
that have not been included in the regression model. The intercept term, on the
other hand, stands for all such variables, that are known to have some definite
non-random effects on the dependent variable. For example, in the demand
function, if we just formulate a two variable regression equation between the
quantity demanded and the price instead of a multiple regression equation, then
we are consciously not including variables like prices of other commodities and
the income of the consumer. And all these variables affect the quantity
demanded in a known fashion. So, the intercept term here represents the
average effect of all such known factors that are not explicitly included in the
model.
It should be clear now that it is the disturbance term that makes the relationship
statistical in nature and renders the entire procedure so rich in content.
Population Regression Line

The main objective of the regression analysis is to obtain the value of the
dependent variable for a given value of the independent variable. This can be
attempted by running a regression on the population regression function
Y = α + βX + U , i.e., by fitting a regression line through the population cluster
shown in Figure 9.1. Obviously, the value of Y that we shall obtain from such a
population regression line will be an average value. In other words, we shall
get the equation for the population regression line by applying the expectation
to the population regression function, Y = α + βX + U . Thus, the population
regression equation will be given by
E (Y / X ) = α + βX
It may be mentioned here that while applying the expectation operator, E (U)
can be taken as zero because, as we have discussed above, the mean effect of U
tends to be zero.
Sometimes we call the above expression, the population regression line and
loosely write it as
Y = α + βX
In this expression, however, we should be clear that Y stands for the conditional
mean of Y for a given X.
9
Quantitative Methods-1 Assumptions of the Classical Regression Model T
As we have mentioned earlier, the population regression model

Y = α + βX + U is unknown in the sense that that the numerical values of its
parameters are not known. As a result, we are not in a position to find the value
taken by Y on an average for a given value of X. This means that we have to
estimate these parameters from sample information. We already know that a
very simple and popular method of estimation is the least square procedure.
However, we should ensure that these least square estimates faithfully
represent the unknown population parameters, otherwise, the whole purpose is
defeated. This is possible only if the disturbance term U satisfies some
assumptions. These assumptions along with two other that we have already
mentioned about, are known as the assumptions of the classical regression
model. The assumptions are:
1) The disturbance term U has a zero mean for all the values of X, i.e., E (U) = 0.
2) Variance of U is constant for all the values of X, i.e., V ( X ) = σ 2 . It may be
mentioned here that this assumption is known as the assumption of
homoscedasticity in the econometric literature.
3) The disturbance terms for two different values of X are independent i.e.,
cov(U i , U j ) = 0, for i ≠ j .
4) X is non-stochastic.
5) The model is linear in parameter.
The assumption of non-stochasticity of X leads to a corollary. It is, X and U are
independent i.e., cov( X ,U ) = 0 .
9.5 SAMPLE REGRESSION FUNCTION

To put the whole discussion in the proper perspective; we have an unknown bi-
variate population. The only information that we have is some randomly
selected values of Y from this population that correspond to some fixed values
of X. Suppose, in this way we have n pairs of values of X and Y. So, we can say
that we have a random sample of size n. Our purpose is to estimate the
parameters of this unknown population from our sample information so that we
can have a reasonable estimate of an average value of Y for a given value of X
that is valid for the entire population. In other words, our objective is to
estimate the population regression function from the sample observations. We
can proceed in that direction by discussing the concept of a sample regression
function. The form of the sample regression function is quite similar to that of
the population regression function. It can be presented as
Yˆ = αˆ + βˆX + Uˆ
In the above expression, Yˆ , αˆ , βˆ and Uˆ should be interpreted as the sample

estimators for their respective unknown population counterparts. Thus, we are
hypothesizing that corresponding to the linear population regression function
Y = α + βX + U , there is a linear sample regression function Yˆ = αˆ + βˆX + Uˆ .
Our purpose is now to estimate the population regression function from the
sample regression function. It means that we have to estimate α̂ and βˆ from
the given sample and take them as the estimates for the unknown population
10
1 parameters. It should be clear that due to sampling fluctuations, we might Two Variable Regression
Models
obtain different values of α̂ and βˆ from different samples. What we have to
ensure is that on the average, they represent the population parameters. Thus,
the entire issue now boils down to the estimation of α̂ and βˆ . This is what is
known as the estimation of the sample regression function and we are going to
discuss it in the next section.
9.6 ESTIMATION OF SAMPLE REGRESSION

FUNCTION
The commonly used procedure is the least square method that we discussed in
the previous unit. To recapitulate, we have a sample of observations
represented by a cluster of points in the sample space and we have to pass a
regression line through this cluster in such a fashion that the sum of the squares
of the deviations of the observed values of Y from their estimated values from
the regression line for different values of X is minimum. We are reproducing
the two relevant diagrams from Unit 8 for the sake of convenience.
0 X
Fig. 9.2: Sample Scatter-Plot
Thus, we have a scatter of sample values as shown in the above diagram and
we have to pass a regression line through it. We can summarise the least square
procedure for obtaining such a line now. Putting mathematically the procedure
amounts to Minimize
∑u ∑ (Y − Yˆ ) = ∑ (Y − αˆ − βˆX )
2 2
2
= with respect to α̂ and βˆ .
Following the usual minimization procedure, we obtain two normal equations
given by
∑ Y = nαˆ + βˆ ∑ X
and
∑ XY = αˆ ∑ X + βˆ ∑ X 2
.
11
Quantitative Methods-1 Solving the two normal equations simultaneously, we obtain T
βˆ =
n∑ XY − ∑ X ∑ Y
=
∑ (X − X )(Y − Y )
n∑ X − (∑ X ) ∑ (X − X )
2 2 2
and
αˆ =
∑ X ∑ Y − ∑ X ∑ XY = Y − βˆX .
2
n ∑ X − (∑ X )
2 2
Let us now derive an important result for the estimated slope coefficient βˆ .
Writing lower case letters for deviations
βˆ =
∑ xy = ∑ x(Y − Y ) = ∑ xY − Y ∑ (X − X )
∑x 2
∑x 2
∑x 2
But ∑ (X − X ) = 0
Thus βˆ =
∑ xY
∑x 2
x
Now putting k = , as xx’s
s are
aregiven.
given.
∑ x2
We have
βˆ = kY∑
Thus, βˆ is a linear function of the observed values of Y. This is an important
result, which we shall use later in the issue of hypothesis testing.
The following diagram shows a regression line, with the values of α̂ and βˆ
given by the above expressions fitted into our sample scatter.
0 X
Fig. 9.3: Sample Scatter-Plot with the Sample Regression Line
12
1 It may be noted here that the sample regression line has the equation, Two Variable Regression
Models
Y = αˆ + βˆX with the values of α̂ and βˆ obtained from the above-mentioned
least square method. It can be clearly seen that the values of α̂ and βˆ can be
calculated in terms of sample observations only and no other information that
is not known is required for this purpose. As we can see, both α̂ and βˆ are
linear functions of X and Y. Consequently, they are called linear estimators.
Gauss-Markov Theorem
The least square estimators α̂ and βˆ can be taken as the estimators of the
unknown population parameters α and β because they satisfy certain desirable
properties. We can state them without proofs below.
Under the assumptions of the classical regression model as mentioned earlier,

1) Least square estimators are linear. As we have already seen from the
expressions of α̂ and βˆ , they are linear functions of the variables.
2) Least square estimators are unbiased i.e., E ( α̂ ) = α and E ( βˆ ) = β . This

means, if we consider different values of the least square estimates
obtained from a number of random samples for a given population on the
average, they will be equal to the unknown population parameters. They
will not be systematically overestimating or underestimating the population
parameters.
3) Among all the linear unbiased estimators, least square estimators have
the minimum variance. In this sense, they are termed as the efficient
estimators.
All these properties of the least square estimators lead to what is known as the
Gauss-Markov Theorem. The theorem states:
Under the assumptions of the classical linear regression model,
among all the linear unbiased estimators, the least square estimators
have the minimum variance. In other words, the least square
estimators are the best linear unbiased estimators or in short, BLUE.
Standard Error of the Regression Estimate

We have already stated that due to sampling fluctuations, we should expect to
obtain different values of the least square estimates α̂ and βˆ from sample to
sample. Consequently, if we measure the standard deviations of these values of
the two least square estimates from their respective expectation or mean, we
shall get an idea about the extent to which these estimates are affected by the
sampling fluctuations. In other words, the standard deviations of the least
square estimates can be taken as a measure of the precision of these estimates.
The standard deviations of the least square estimates are known as the
standard errors of the estimates. It should be clear that the standard errors
are the standard deviations of the sampling distributions of the least square
estimates. The standard deviations or standard errors are obtained by taking the
positive square root of the variances of α̂ and βˆ . The expressions for both the
variance and standard of the least square estimators are given below.
13
Quantitative Methods-1
var (αˆ ) =
∑X σ 2
2
T
n∑ (X − X ) 2
se(αˆ ) =
∑X σ 2
2
n∑ (X − X ) 2
()
var βˆ =
σ2
∑ (X − X )
2
()
se βˆ =
σ
∑ (X − X )
2
It should be recollected that σ is the standard deviation of the error term U in

our population regression model and we assume that it is a constant. However,
since the population regression model is unknown, it follows that σ is also
unknown. As a result, for the calculation of variance and standard error of the
least square estimates α̂ and βˆ , we need to estimate σ . It can be proved that
an unbiased estimator of σ is
∑Uˆ 2
. Here, n – 2 is what is known as the

n−2
degrees of freedom in statistics. This concept you must have studied in your
compulsory course in statistics. Now, Uˆ = Y − Yˆ is the sample regression error
term and accordingly we can calculate it. Thus, by replacing σ by its unbiased
estimator we can compute the standard errors of both α̂ and βˆ . It should be
noted here that we can write the unbiased estimator of σ as
∑ (Uˆ − Uˆ )
2
σˆ =
∑Uˆ 2 =
n−2 n−2
This is, because, Û is the mean of the sample regression errors. Now, while
calculating it, the positive errors will tend to cancel the negative errors and as a
result, it will be reduced to zero. From above expression of the estimator of σ ,
it can easily be interpreted as the standard deviation of the sample observations
about the estimated sample regression line. The sample estimator σˆ =

∑Uˆ 2
n−2
is known as the standard error of estimate or the standard error of the
regression.
9.7 GOODNESS OF FIT

In the earlier section, we have discussed the procedure for fitting the sample
regression line and its purpose. Once such a regression line is fitted, we may be
interested in examining how good has been the fit in the sense how faithfully it
can describe the unknown population regression line. This is known as the
issue of goodness of fit. In this matter, the regression error term or residual Û
14
1 plays an important role. Small quantities of residuals imply that a large Two Variable Regression
Models
proportion of variation in the dependent variable has been explained by the
regression equation and consequently, the fit is good. Similarly, large quantities
of residuals obviously point to a poor fit. At this stage, what we are interested
in is to obtain a quantitative measure of the goodness of fit that is free of any
unit for the purpose of comparability. In the previous unit, we have referred to
the coefficient of determination, which is the square of the correlation
coefficient. It can be shown that this coefficient of determination acts as a
measure of goodness of fit. We may consider it now. The variation in the
dependent variable Y about its mean can be conceptualized as
var (Y ) = ∑ (Y − Y )
2
We can decompose it into two components. The first is the variation explained
by the regression. The second is the portion that remains unexplained by the
regression. Thus we can write
(Y − Y ) = (Y − Yˆ )+ (Yˆ − Y )
Squaring and applying summation on both the sides
∑ (Y − Y ) = ∑ (Yˆ − Y ) + ∑ (Y − Yˆ ) + 2∑ (Yˆ − Y )(Y − Yˆ )

2 2 2
Now let us try to find the value of ∑ (Yˆ − Y )(Y − Yˆ ). We have the regression
equation
Y = αˆ + βˆX + Uˆ
or Y = αˆ + βˆX + Uˆ
Subtracting the second equation from the first equation
Y − Y = βˆ (X − X ) + Uˆ , because Uˆ =
∑Uˆ = 0, as Uˆ = 0.
n
∑
Writing lower case letters for the deviations of the variables from their mean,
we have the sample regression equation in the form
y = βˆx + uˆ
Now
å (Yˆ - Y )(Y - Yˆ )
= å (Yˆ - Y ) Uˆ
= å YUˆ ˆ - Y Uˆ
å
= å YUˆ ˆ , because Uˆ = 0
å
= å (aˆ + bˆ X) Uˆ
= aˆåUˆ + bˆå XUˆ ∵ å Uˆ = 0

= bˆ å XUˆ
We have
15
Quantitative Methods-1 x= X -X T
Þåx = å X - X ( )
or å xUˆ =å X - X Uˆ = å XUˆ - X åUˆ =å XUˆ
( )
Now
å xUˆ = å x ( y - bˆ x) = å xy - bˆ å x 2
We know, bˆ = (
å X - X )(Y - Y ) = å xy
å( X - X ) åx 2
Putting the value of bˆ in the above expression for å xUˆ ,
å xUˆ = å xy - å x å x = å xy - å xy = 0
xy 2
å 2
Hence,
∑ (Yˆ − Y )(Y − Yˆ ) = 0
Thus,
∑ (Y − Y ) = ∑ (Yˆ − Y ) + ∑ (Y − Yˆ )
2 2 2
∑ (Y − Y ) ∑ (Yˆ − Y )
2 2
If is defined as the total sum of squares (TSS), as the
∑ (Y − Yˆ )
2
explained sum of squares (ESS) and as the residual sum of squares
(RSS),
We have
TSS = ESS + RSS
Dividing both the sides by TSS,
TSS ESS RSS

= +
TSS TSS TSS
Defining the ratio of ESS to TSS as R 2 , we have
RSS ESS
R2 = 1- =
TSS TSS
It should be clear now that R 2 or coefficient determination can be interpreted

as the proportion of total variation in Y explained by the regression of Y on X.
Let us now consider a numerical example to fix the concepts and ideas that we
have discussed so far.
Example 9.1
The hypothetical figures for the total labour force and the number of employed
out of that for the period 1991-2000 are given below in Table 9.1. Run a
regression of the number of employed on the total labour force.
16
1 Table 9.1: Actual Employment and Labour Force Two Variable Regression
Models
Employed Labour Force
Year (million) (million)
1991 100 120
1992 125 140
1993 140 165
1994 160 185
1995 175 200
1996 195 210
1997 230 250
1998 245 255
1999 270 305
2000 295 320
Let Y be the dependent variable employed and X be the independent variable

labour force. The detailed calculation for working out the regression is shown
below:
Year Y X XY X2 y = Y −Y x = X − X x2 y2 xy
1991 100 120 12000 14400 -93.5 -95 9025 8742.25 8882.5
1992 125 140 17500 19600 -68.5 -75 5625 4692.25 5137.5
1993 140 165 23100 27225 -53.5 -50 2500 2862.25 2675
1994 160 185 29600 34225 -33.5 -30 900 1122.25 1005
1995 175 200 35000 40000 -18.5 -15 225 342.25 277.5
1996 195 210 40950 44100 1.5 -5 25 2.25 -7.5
1997 230 250 57500 62500 36.5 35 1225 1332.25 1277.5
1998 245 255 62475 65025 51.5 40 1600 2652.25 2060
1999 270 305 82350 93025 76.5 90 8100 5852.25 6885
2000 295 320 94400 102400 101.5 105 11025 10302.25 10657.5
Sum 1935 2150 454875 502500 40250 37902.5 38850
Mean 193.5 215
bˆ = å
xy 38850
= = 0.965217 and aˆ = Y - bˆ X = 193.5 - (0.965217 ´ 215) = –14.217
å x 2
40250
17
Year Yˆ = αˆ − βˆX Uˆ = Y − Yˆ Û 2
Yˆ − Y (
Yˆ − Y )
2
1991 101.8043 -1.80434 3.255643 -91.6957 8408.094

1992 121.1087 3.89132 15.14237 -72.3913 5240.503
1993 145.2391 -5.23911 27.44822 -48.2609 2329.114
1994 164.5434 -4.54344 20.64289 -28.9566 838.4821
1995 179.0217 -4.0217 16.17407 -14.4783 209.6212
1996 188.6739 6.32613 40.01992 -4.82613 23.29153
1997 227.2826 2.71745 7.384535 33.78255 1141.261
1998 232.1086 12.89137 166.1873 38.60864 1490.627
1999 280.3695 -10.3695 107.5262 86.86949 7546.307
2000 294.8477 0.15226 0.023183 101.3477 10271.36
Sum 0.00045 403.8043 37498.67
In the next stage we compute the following:
∑X ∑ (Yˆ − Y )
2
V (βˆ ) =
2
σ2 ESS
, V (αˆ ) = σ and R =
2
=2
.
∑x n∑ x ∑ (Y − Y )
2 2 2
TSS
But the population variance σ2 is unknown. So we have to use an unbiased

sample estimator σ2 for the same. Such an unbiased estimator is given by
s =
ˆ 2 å Û 2
, where n – 2 is the degrees of feedom. Here, n – 2 = 10 – 2 = 8.
n-2
Therefore, sˆ 2 = å
Û 2 403.8043
= = 50.475537.
n-2 8
sˆ 2 50.475537
Thus, V bˆ =()
åx 2
=
40250
= 0.001254
() ()
s.e. bˆ = V bˆ = 0.001254 = 0.0354118
V (aˆ ) = åX 2
sˆ 2 =
502500
´50.475537 = 63.016042
nå x 2
10 ´ 40250
s.e.(aˆ ) = V (aˆ ) = 63.016042 = 7.9382644
ESS 37498.67
R2 = = = 0.989345
TSS 37902.5
Finally, we present a summary of all the important results concerning the

above-mentioned question.
18
Models
βˆ = 0.965217 ()
V βˆ = 0.001254 and ()
s.e. βˆ = 0.0354118
αˆ = −14.0217 V (αˆ ) = 63.016042 and s.e.(αˆ ) = 7.9382644
σˆ 2 = 50.475537
R 2 = 0.989345 Degrees of Freedom (d . f .) = 8
The estimated regression line is given by
Yˆ = −14.0217 + 0.965217 X
It may be mentioned here that it is this kind of a summary of the results that is
generally used for the further analysis of the regression model. Let us now
interpret some of the results from the above summary. The regression line
estimates the average employment (Y) for a given level of labour force (X). The
slope coefficient βˆ = 0.965217 estimates the rate of change of employment
with respect to labour force. For example, if 100 more persons start looking for
jobs, about 97 of them actually get employed. The intercept αˆ = −14.0217 can
be interpreted as the average combined effect of all those variables that might
also affect employment but have been omitted for the purpose of the above-
mentioned regression. The coefficient of regression R 2 = 0.989345 indicates
that about 99 per cent of the variation in employment can be explained by a
variation in the labour force, which is indeed high and indicates a good fit to
the given sample.
1) Discuss the meaning of linearity in the regression model.

……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) How is the regression model non-deterministic in nature?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
19
Quantitative Methods-1 3) Discuss the assumptions of the classical regression model. T
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
4) State the Gauss-Markov Theorem.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
5) Explain the concept of goodness of fit.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
9.8 FUNCTIONAL FORMS OF REGRESSION

MODEL
We have already mentioned that our focus is on a linear regression model. The
issue of linearity has also been considered in Section 9.2. It is in fact linear in
parameter regression model that is relevant for us. However, linear in
parameter regression models may also have different functional forms. We
shall now briefly discuss four such regression models.
1) Linear Model
This functional form is the most common one and we have already discussed
it. Its equation is given by
Y = α + βX + U
As we know, it is both linear in parameter and linear in variable. This model
can be estimated by the ordinary least square (OLS) method. The least square
20
1
estimators α̂ and βˆ are the unbiased estimators of the unknown population Two Variable Regression
Models
parameters α and β .
Here the regression coefficient β measures the rate of change of Y per unit
change of X.
Example 9.2
In this example we shall consider the issue of fitting a Keynesian consumption
function to the Indian data. Table 9.2 presents the GDP at factor cost and final
consumption expenditure (FCE) figures for the Indian economy during the
period 1980-2001 at 1993-94 prices.
Table 9.2: GDP and Final Consumption Expenditure (Rs. Crore)
Year GDP FCE Year GDP FCE

1980 401128 390671 1991 701863 620806
1981 425073 407728 1992 737792 637307
1982 438079 415924 1993 781345 666950
1983 471742 446396 1994 838031 695825
1984 492077 461675 1995 899563 740109
1985 513990 485090 1996 970082 794093
1986 536257 504854 1997 1016595 826834
1987 556778 525558 1998 1082747 888513
1988 615098 557527 1999 1148367 952489
1989 656331 584970 2000 1198592 972932
1990 692871 610169 2001 1267945 1027254
FCE
GDP
Fig. 9.3: Scatter of Final Consumption Expenditure against GDP
21
Quantitative Methods-1 The above scatter of final consumption expenditure against GDP been obtained T
from Table 9.2. The scatter makes it clear that there seems to be a linear
relationship between the two variables. Consequently, we have run a linear
regression between the variables by using the data of Table 9.2. The results are
presented below:
FCE = 108206.4 + 0.719674 GDP
()
s.e. aˆ = 6233.203 ()
s.e. bˆ = 0.007865
t (aˆ ) = 17.35968 t (bˆ ) = 91.50314

R 2 = 0.997617 d . f . = 20
A high R 2 of more than 0.99 implies a tight fit. The estimated βˆ = 0.719674
indicates a marginal propensity to consume of about 72 percent and this can be
quite expected in the Indian economy. We shall explain the significance of
degrees of freedom and t values later in the issue of tests of hypotheses.
However, we shall continue to present these two statistics in our subsequent
examples also.
2) Log-linear Model
This model is also known as log-log, double-log or constant elasticity model.
Its original form is given by
Y = αX β e U , Where, e is the base of natural log (e is approximately equal to

2.718). In the original form, the model is non linear in nature. However, by
applying the log transformation, the model is transformed to
ln Y = ln α + β ln X + U
If we put ln Y = y, ln α = a, β = b and ln X = x , the model reduces to
y = a + bx + U
From the above transformation, it is very clear that the model now becomes
linear in parameter and linear in log of the variables; and thus, can be estimated
by the ordinary least square method. Here, a will be estimating ln α and then
by taking the antilog of a, we shall obtain the estimate for α . The regression
coefficient b = β deserves our special attention. It measures a change in log Y
d ln Y
per unit of a change in log X. In the language of calculus, β = . Now, a
d ln X
change in the log of some variable implies a proportional change in it. Thus, β
is the ratio of a proportional change in Y to a proportional change in X. in other
words, β measures the elasticity of Y with respect to X. This is the rationale
for the log linear regression model being termed as the constant elasticity
model.
Example 9.3
This is an example from Gujrati (2003). The table below presents the quarterly
data on total personal consumption expenditure (PCExp) and expenditure on
durables (ExpDur) for the U.S. economy measured in 1992 billions of dollar
for the period 1993:1-1998:3.
22
1 Table 9.3: Personal Consumption Expenditure and Expenditure on Durables in Two Variable Regression
U.S.A. Models
Year PCExp ExpDur
1993:1 4286.8 504
1993:2 4322.8 519.3
1993:3 4366.6 529.9
1993:4 4398 542.1
1994:1 4439.4 550.7
1994:2 4472.2 558.8
1994:3 4498.2 561.7
1994:4 4534.1 576.6
1995:1 4555.3 575.2
1995:2 4593.6 583.5
1995:3 4623.4 595.3
1995:4 4650 602.4
1996:1 4692.1 611
1996:2 4746.6 629.5
1996:3 4768.3 626.5
1996:4 4802.6 637.5
1997:1 4853.4 656.3
1997:2 4872.7 653.8
1997:3 4947 679.6
1997:4 4981 648.8
1998:1 5055.1 710.3
1998:2 5130.2 729.4
1998:3 5181.8 733.7

Source: Damodar Gujrati (2003), Basic Econometrics
We have plotted the log of expenditure on durables against the log of total
personal consumption expenditure in Figure 9.4 below.
23
Log Exp Dur
Log PCExp
Fig. 9.4: Scatter of Log of Exp on Durables against Log of Personal Con Exp
The scatter above is clearly indicative of a linear relationship between the log
of the two variables. As a result, we fit in a log linear model to the data on total
personal consumption expenditure and the expenditure on consumer durables.
We present some of the results from the fit below.
ln ExpD ur = - 9.697098 + 1.905633 ln PC Exp

s .e . (ln aˆ ) = 0.434127 ()
s .e. bˆ = 0.051370
t (ln aˆ ) = - 22.33702 t (bˆ ) = 37.09622

R 2 = 0.984969 d . f . = 21
We have an R 2 of about 0.99. This is indicative of a good fit. In this fit, the
slope coefficient is an estimate of the elasticity of expenditure on durables with
respect to personal consumption expenditure. Thus, βˆ = 1.905633 indicates
that the expenditure on durables is highly elastic to total personal consumption
expenditure.
3) Semi-Log Model
One of the common forms of semi-log regression models is the so-called
growth rate model. The model is expressed as
log Y = α + β t + U
In this model the dependent variable is expressed in log, whereas, the
independent variable is the time period, and it is measured in absolute term.
Since, log operator is applied only to the left hand side of the equation, the
model is called the semi-log model. In this model, the slope coefficient β is a
measure of the proportional change in the dependent variable for a unit change
in the time period. Thus it measures the growth rate of the dependent variable.
24
1 Example 9.4 Two Variable Regression
Models
In this example, we are going to verify the so-called ‘Hindu Rate of Growth’ of
about 3.5 per cent that Indian economy consistently witnessed in 1960s and
1970s.
In Table 9.4, we present India’s GDP at 1993-94 prices for the period 1960-
1980.
Table 9.4: India’s GDP during 1970-1980
Year GDP Year GDP
1960 206103 1971 299269
1961 212499 1972 298316
1962 216994 1973 311894
1963 227980 1974 315514
1964 245270 1975 343924
1965 236306 1976 348223
1966 238710 1977 374235
1967 258137 1978 394828
1968 264873 1979 374291
1969 282134 1980 401128
1970 296278
Figure 9.5 presents the scatter of the log of GDP against the time period
denoted by the letter t.
Log GDP
Time
Fig. 9.5: Scatter of India’s GDP against Time
The scatter above suggests a linear relationship between the log of GDP and
the time period. Therefore, we run a regression of log GDP on time period t.
we present the result below.
25
Quantitative Methods-1 log GDP = 12.19473 + 0.033729 t T
s.e.(αˆ ) = 0.012517 ()
s.e. βˆ = 0.000997
t (αˆ ) = 974.2658 t (βˆ ) = 33.83607
R 2 = 0.983675 d . f . = 19
The value of R 2 is quite high. The estimated slope coefficient of 0.034

indicates a rate of growth of 3.4 per cent during the period 1960-1980. This
adequately supports the phenomenon of ‘Hindu rate of Growth’ in the 60s and
the 70s.
4) Reciprocal Model
In the reciprocal model, the dependent variable is regressed on the reciprocal of
the independent variable. Its functional from is
1
Y =α +β +U
X
Although the model is non-linear in variable (because the power of X is –1), it
is linear in parameter. Hence, it can be estimated by the OLS method. The
model has an important characteristic. As the independent variable X increases
1
indefinitely, the term β approaches zero, β being a constant. Consequently,
X
the dependent variable Y approaches a limiting value or an asymptotic value
equal to the intercept α . An application of this model can be in the relationship
between per capita GNP and the child mortality. As per capita GNP increases,
the infant mortality rate is expected to fall but we cannot expect it to fall
independently if per capita GNP continues to increase indefinitely. In such a
situation, in all probability, the mortality rate will tend to a limiting value.
Gujrati (1993) presents an example of applying the reciprocal model to study
the relationship between the per capita GNP and the infant mortality rate by
using the cross-section data for 64 countries. His results are presented below.
1
Y = 81 . 79436 + 27237 . 17
X
s .e .(αˆ ) = 10 . 8321 ()
s .e. βˆ = 3759 . 999
t (αˆ ) = 7 . 5511 ()
t βˆ = 7 . 2535
R 2 = 0 . 4590
where, Y is the child mortality rate (the number of deaths of children under the
age of 5 per year per 1000 live births) and X is the per capita GNP at constant
prices. The above-mentioned results show that if per capita GNP increases
indefinitely, the infant mortality approaches a limiting rate of about 82 deaths
per 1000 children born.
Another application of the reciprocal model can be in the estimation of the

Phillips curve. This curve proposes an inverse relationship between the change
in unemployment rate and that in inflation rate. With a growth in
unemployment rate, the growth in inflation rate is expected to fall. In fact
Phillips curve postulates that if the unemployment rate increases beyond its so-
called natural rate, the growth in the inflation rate should become negative.
26 However, with an indefinite increase in the unemployment rate, the inflation
1 growth cannot be expected to fall indefinitely. It should stabilize at some Two Variable Regression
Models
negative limiting value. Thus, Phillips Curve can be an appropriate case for the
use of the reciprocal model.
9.9 CLASSICAL NORMAL REGRESSION MODEL

We have already seen that a two variable classical linear regression model can
be presented as
Y = α + βX + U
In this model, U is the population disturbance term. We can estimate the
unknown parameters of this regression model from the sample information by
using the least square or, as it is sometimes called, ordinary least square
method. The least square estimates possess some desirable properties if the
population disturbance U satisfies some five assumptions that we have
discussed in Section 9.3. These five assumptions are adequate for the
estimation purposes. But the scope of a regression model is just not restricted
to the estimation of the parameters. An important purpose of the sample
estimates is to test some hypotheses about the unknown population regression
parameters with their help. And we can do this if we make some assumption
about the distribution of U. This we do by making the assumption of normality
for U. The assumption essentially means that the population regression
disturbance term follows normal distribution with mean zero, a constant
variance equal to σ 2 and a zero covariance. In fact U has a conditional
distribution, in the sense, that, for each of the given values of the non-
stochastic independent variable X, we might have a distribution of different
values of U. This will be clear if we reproduce the diagram of Figure 9.1 with
the unknown population regression line fitted into it.
Consumption
Income
Fig. 9.6: Bi-Variate Population with the Unknown Regression Line
In Figure 9.6, we can see that for a given observed value of income, we can
have different observed values of consumption represented by the bold dots on
each of the vertical lines. The vertical distance between an observed value of
consumption and the corresponding estimated value from the regression line
for a given income level measures the value of the disturbance term. Thus there
27
Quantitative Methods-1 are many possible values of U for each of the given income levels. As a result, T
there are conditional distributions of U for different levels of income.
The significance of the normality assumption is that these conditional

distributions of U should all be independently normally distributed with the
same mean equal to zero and, the same variance equal to σ 2 . Mathematically
speaking,
E (U i ) = 0 for all is
V (U i ) = σ 2
for all is
Cov (U iU j ) = 0 for i ≠ j
In other words, the population disturbance variable should have independent

and identical normal distribution.
If we make this additional normality assumption along with the earlier-

mentioned five assumptions about the disturbance term U, then the regression
model is called the classical normal regression model.
It may be mentioned here that the normality assumption is important not only
for hypothesis testing but also for the fact that under this assumption we can
use an alternative procedure for estimating the population regression
parameters. This procedure is known as the method of maximum likelihood
estimation. Although for regression parameters the method gives the same
estimates as the least square estimates, the approach followed is quite different.
In fact this procedure has some stronger theoretical characteristics than the
least square method. However, we are not discussing this procedure here
because it is beyond our scope.
In the next section, we shall discuss the issue of hypothesis testing and see how
the normality assumption plays a crucial role there.
9.10 HYPOTHESIS TESTING

Considering the two variable regression model
Y = α + βX + U
We might be interested in examining whether the unknown parameter α or β
assumes a particular value or not. This is what is known as hypothesis testing
in statistics. We can conduct such a test from the sample estimate of α or β as
the case might be. Although we may test some hypothesis about the intercept
α , our main concern in regression model is the slope coefficient β . We have
the estimated sample regression function
Yˆ = αˆ + βˆX + Uˆ
Here, βˆ is the sample estimate of β and we shall use it for estimating the
unknown β . As mentioned earlier that βˆ might vary from sample to sample
collected from the same population. As a result, βˆ is a random variable with
some probability distribution. Now, for the purpose of hypothesis testing, we
need to know the form of this distribution of βˆ .
28
1 It is here that the assumption of normality helps us. Let us consider it. From Two Variable Regression
Models
Section 9.6, we know that βˆ = ∑ kY and thus, it is a linear function of the
observed values of Y. But, Y = α + β X + U . Consequently, we can
see, βˆ = ∑ k (α + β X + U ). Now, we already know that the ks, the Xs and the
parameters are all given. As a result, finally, βˆ becomes a linear function of the
random disturbance variable U. Thus, βˆ has the same distribution as U.
Therefore, the assumption of normality U, implies that βˆ is distributed as
normal.
()
We have already seen that E βˆ = β and se βˆ = () σ
. It follows
∑ (X − X )
2
then that
 
 σ 
β ~ N β,
ˆ

 ∑ (X − X ) 
2
 
You must have learnt from your compulsory course in quantitative methods,
that now a standard normal variable can be formed with mean equal to zero and
standard deviation equal to one and it can be used for testing of hypotheses
regarding β . But the problem here is that the population standard deviation σ
is unknown and we have to use an estimate for it. However, we know that an
unbiased estimate of σ is given by σˆ 2 =

∑Uˆ 2 , where, Û is the sample
n−2
regression error term and is therefore, computable, and n is the size of the
sample. Again you must be knowing that when we standardize βˆ by
subtracting its mean from it and divide it by its estimated standard deviation, it
no longer follows normal distribution and it in fact follows a student-t
distribution with n-2 degrees of freedom. This student-t distribution has a mean
equal to zero and a standard deviation equal to one. Thus, it is this standard
student-t distribution that is used for testing of hypothesis about β . We have a
table for such a standard student-t distribution for different degrees of freedom.
And like the standard normal table, this table is put into use for such tests of
hypotheses.
Let us see now how we can conduct some test of hypothesis regarding the
unknown population regression coefficient by considering some examples that
we have already discussed in Section 9.7. Consider Example 9.2. In this
example, we have considered the issue of the consumption function in the
Indian economy by regressing final consumption expenditure on GDP. To
begin with, we might be interested in examining whether the relationship is
significant or not. Thus, our focus is on the regression coefficient β and by the
nature of the inquiry, we have to conduct a two-tailed test. We can proceed by
setting our null hypothesis and the alternative hypothesis in the following
manner:
Null hypothesis H0 : β = 0
Alternative hypothesis H1 : β ≠ 0
29
Quantitative Methods-1 The acceptance of the null hypothesis implies the rejection of the alternative T
hypothesis and this in turn implies that on the basis of the sample information
there does not seem to be any significant relationship between GDP and the
final consumption expenditure. On the other hand, the rejection of the null
hypothesis implies the acceptance of the alternative hypothesis and this in turn
implies that on the basis of the sample information there seems to exist
significant relationship between GDP and the final consumption expenditure.
The next step is to construct our test statistic, which is conventionally denoted
by t. Thus
t= ()
bˆ - E bˆ
s.e (bˆ )
Under the null hypothesis,
bˆ - 0 bˆ
t= =
Uˆ 2 Uˆ 2
ån-2 ån-2
And this is distributed as a student-t with n-2 degrees of freedom. From the
results of the Example 9.2 we can easily compute the value of this t statistic
since, both the values of the estimated regression coefficient (0.719676) and
the standard error of the estimated regression coefficient (0.007865) are
mentioned there. In fact, even the computed value of the t statistic (91.50314)
is given. In addition, the degree of freedom is also given (20). Normally, any
test of hypothesis is conducted either at 1 per cent level of significance or at
5% level of significance. From the student-t distribution table we can find out
the critical values of the test statistic t for both 1 per cent level of significance
and at 5% level of significance at a degree of freedom of 20 for two-tailed test.
The values are 2.845 and 2.086 respectively. Thus, the computed value of t far
exceeds both the critical values. Thus, on the basis of the sample information
we cannot accept the null hypothesis. This amounts to accepting the alternative
hypothesis. Consequently, personal consumption expenditure does seem to be
dependent upon GNP in India during the sample period 1980-2001.
Once the relationship between GDP and personal consumption expenditure has
been established some further test can be conducted about the likely value of
the marginal propensity to consume out of GNP. For example we can conduct a
test regarding whether the marginal propensity to consume is 80 per cent or
not. Thus we have
Null hypothesis H 0 : β = 0.80
Alternative hypothesis H 1 : β ≠ 0.80
Thus our t statistic in this case is given by
t=
()
βˆ − E βˆ
=
0.719676 − 0.80 − 0.080324
= = 10.212841 .
()
s.e βˆ 0.007865 0.007865
Now this value of the test statistic also exceeds the critical values of 2.845 and
2.086 for a degree of freedom of 20 at 1 per cent and 5 per cent levels of
significance respectively. Thus, on the basis of the sample information the
30
1 difference between estimated value of β and its mean is so much that even in Two Variable Regression
Models
1 out 100 cases or 5 out of 100 cases we do not expect to obtain such a
difference. Thus on the basis of the sample information, we are not in a
position to accept the null hypothesis. Hence, in all probability during the
sample period of 1980-2001, India’s marginal propensity to consume out GDP
has not been as high as 80 per cent.
In the above example, we considered two-tailed tests of hypotheses. This

however, does not rule out the possibility of conducting one-tiled tests. It all
depends upon the type of inquiry that we intend conducting.
Thus, a discussion on the test of hypotheses finally marks the end of our
discussion on two variable regression model.
1) When will you use a log linear regression model?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) How do you interpret the estimated slope coefficient of a log linear

regression model?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
3) If you want to estimate India’s rate of growth of per capita income during
the period 1980-2000, what should be the functional form of your
regression model?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
31
Quantitative Methods-1 4) Explain the concept of a classical normal regression model. What is its use? T
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
5) Explain briefly the steps for testing the significance of the regression
coefficient.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
9.11 LET US SUM UP

Our focus has been exclusively on the two variable regression model. The
purpose of this unit has been to specify what is known as the classical
regression model in the literature for the population. It is essentially a linear in
parameter regression model. This regression model is stochastic in nature in the
sense that here the dependent variable is taken as a random variable as against
the independent variable being considered as non-random in nature. The
element of stochasticity is introduced by including a random error or
disturbance term in the model. Our major concern has been to estimate the
unknown parameters of the population regression model from the known
sample information and tests some hypotheses about these parameters. In this
regression model we usually make five assumptions about the disturbance term
in order to effectively estimate the parameters by implementing the ordinary
least square procedure. The estimates thus obtained satisfy some desirable
properties as enunciated in the Gauss-Markov Theorem. In this connection, it
may be mentioned here that if some of these assumptions are not fulfilled,
some problems regarding the sample estimates of these unknown parameters
arise. These problems and their solutions have not been discussed in this unit.
We have also considered different functional forms of the regression model
that might be relevant to use in different situations. For the purpose of
hypothesis testing, we have to extend the assumptions of the classical
regression model to include one more assumption about the probability
distribution of the disturbance term. The resultant regression model is then
known as the classical normal regression model. We have considered the use of
this model for the tests of hypotheses.
32
9.12 EXERCISES Models
1) The following table presents the data for broad money supply M 3 (Rs.
Crore) for India during 1980-2002. Fit in an appropriate regression model
and estimate the rate of growth of money supply in India during this
period.
Table 9.5: India’s Broad Money Supply 1980-2002
Year M3
1980 50966
1981 59793
1982 68515
1983 80577
1984 95295
1985 111096
1986 130653
1987 153207
1988 179687
1989 213856
1990 249493
1991 292403
1992 344238
1993 399048
1994 478196
1995 552953
1996 642631
1997 752028
1998 901294
1999 1056025
2000 1224092
2001 1420025
2002 1647976
2) Consider the following hypothetical data. Fit in a log linear regression

model and interpret the estimated regression coefficient. Show all the
calculations.
33
Dependent Variable (Y) Independent Variable (X)
189.8 173.3
172.1 165.4
159.1 158.2
135.6 141.7
132.0 141.6
141.8 148.0
153.9 154.4
171.5 163.5
183.0 172.0
173.2 161.5
188.5 168.6
205.5 176.5
236.0 192.4
257.8 205.1
277.5 210.1
291.1 208.8
284.5 202.1
274.0 213.4
279.9 223.6
297.6 228.2
297.7 221.3
328.9 228.8
351.4 239.0
360.4 241.7
378.9 245.2
Adapted from Maddala (2001)
3) In Example 9.4, the sample estimate of India’s rate of growth during the
period 1960-80 has been obtained to be about 3.4 per cent per year. Test
the hypothesis that during the same period the rate of growth has in fact
been 4 per cent per year.
9.13 KEY WORDS

Assumptions of the Classical Regression Model:
1) The disturbance term U has a zero mean for all the values of X, i.e., E (U) = 0.
2) Variance of U is constant for all the values of X, i.e., V ( X ) = σ 2 .
34
1 3) The disturbance terms for two different values of X are independent i.e., Two Variable Regression
Models
cov(U i , U j ) = 0, for i ≠ j .
4) X is non-stochastic.
5) The model is linear.
Classical Normal : A linear regression model, in which, in

Regression Model addition to the five assumptions of the classical
regression model, one more assumption of the
error term being normally distributed is made.
Classical Regression : The conventional regression model whose
Model parameters are estimated by the ordinary least
square procedure.
Functional Forms of : Different kinds of linear regression models.
Linear Regression
Model
Gauss-Markov : Under the assumptions of the classical
Theorem regression model, among all the linear
unbiased estimators, the least square estimators
have the minimum variance. In other words,
the least square estimators are the best linear
unbiased estimators or BLUE.
Goodness of Fit : The ratio of the explained sum of squares to
the total sum of squares. Also known as the
coefficient of determination.
Hypothesis Testing : Conducting tests regarding hypotheses made
about the unknown parameters of the
population regression model with the help of
the estimated sample regression function.
Linear Regression : A regression model that is essentially linear in
Model parameter.
Non-Deterministic : A regression model in which the dependent
Regression Model variable is random or probabilistic but the
independent variable is non-random.
Parameters : Unknown intercept and the slope coefficient of
a population regression model.
Population Regression : A linear regression model with a stochastic
Function error term for a given population.
Sample Regression : The sample regression function that is fitted
Function into the sample data for estimating the
population regression model.
Standard Errors of : Standard deviations of the estimated sample
Estimate regression intercept and slope coefficient.
Two Variable : A regression model with one explanatory
Regression Model variable.
35
Gujrati, Damodar N. (2003); Basic Econometrics, McGraw-Hill, New York,
U.S.A.
Maddala, G.S. (2002): Introduction to Econometrics, John Wiley & Sons
(Asia), Singapore.
Pidyck, Robert S. & Rubinfeld, Daniel L. (1991); Econometric Models &
Economic Forecasts, McGraw-Hill, New York, U.S.A.

PROGRESS EXERCISES
1) See Section 9.2
2) See Section 9.3
3) See Section 9.4
4) See Section 9.6
5) See Section 9.7

1) See Section 9.8
2) Measures the elasticity of the dependent variable with respect to the
independent variable.
3) See Section 9.8
4) See Section 9.9
5) See Section 9.10
9.16 ANSWERS OR HINTS TO EXERCISES

1) 15.9 per cent
2) Do Yourself
3) Do yourself
36
I Multivariable Regression
UNIT 10 MULTIVARIABLE REGRESSION Models
MODELS
Structure
10.0 Objectives
10.1 Introduction
10.2 Regression Model with Two Explanatory Variables
10.2.1 Estimation of Parameters
10.2.2 Variance and Standard Errors
10.3 Interpretation of Regression Coefficients
10.3.1 Goodness of Fit: Multiple Coefficient of Determination: R2
10.3.2 Analysis of Variance (ANOVA)
10.4 Inclusion and Exclusion of Variables
10.5 Generalisation to n-explainatory Variables
10.6 Problem of Multi-co-linearity
10.7 Problem of Hetero-scedasticity
10.8 Problem of Autocorrelation
10.9 Maximum Likelihood Estimations
10.10 Let Us Sum Up
10.11 Key Words
10.12 Answers or Hints for Check Your Progress Exercises
10.0 OBJECTIVES
This Unit aims at equipping the learners with techniques of multiple regressions
which come handy in analysing a number of situations where one is required to
study impact of a number of independent variables (factors) on some particular
dependent variable. After going through this unit, you will be able to:
• have a fairly good comprehension of technique of multiple regression
analysis;
• appreciate the need for extending the basic idea of regression as discussed
in Unit 8;
• apply those techniques for more realistic model formulation of explaination
of economic phenomena; and
• spot and avoid pit falls which may arise (in quantitative analysis).
on accounts of auto-correlation, hetero-scedasticity and multi-co-linearity on
the one hand and mis-specification of the model, (in the sense of including
irrelevant as well as excluding relevant variables from the model) on the other.
10.1 INTRODUCTION
In Unit 8, you have studied basics of the Classical Linear Regression Model.
You regressed dependent variable Y on independent variable X. In other
37
Quantitative Methods-I words, you tried to explain changes in Y in terms of the changes in X. You had M
hypothesised a linear relationship between Y and X of this type:
Y = α + βX
∧ ∧
Then, you went on to ‘estimate’ the two constants α and β . Once the
estimates were ready, you had the estimated model or ‘relationship’ with you
given by
∧ ∧ ∧
Υ = α + β x
In the present unit we are going to extend this type of analysis further to make
it more ‘realistic’ and ‘comprehensive’. We will do it in a number of steps:
first of all, we shall introduce one more explainatory variable and re-examine’
the model. Next step will be to generalise the model to n-explainatory
variables. At the next stage, we try to interpret the partial regression co-
efficients. After that, we will try to examine how large should the number ‘n’
be – that is how many explainatory variables must be included in the model
and what should be the statistical touch stone for arriving at such a decision.
Then, we examine the conditions, or ‘assumptions’, which make these
extensions and generalisations possible. We also examine the possible effects
of violation of one or more assumptions. We shall, in particular, pay attention
to problems of multi-co-linearity, Hetero-scedasticity and auto-correlation.
Then, we take a look at another class of estimates, the Maximum Likelihood
Estimators. Finally, we give a brief exposition to some uses of multiple
regression analysis in economics.
10.2 REGRESSION MODEL WITH TWO

EXPLANATORY VARIABLES
At the outset, please note one small change in notation: in section 10.1 – the
introduction to the present unit, we have used the specification of model as
Y = α + βX
However, now onwards, we will write the model in slightly different form:
Y = β 0 + β1 X 1 This form helps us in presentation in the sense that as and
when, we add more explanatory variables X2, X3, etc. we simply indicate their
coefficients with respective β1, β2, β3, etc.
So the model that we consider in the present section has two explanatory
variables X1, and X2. The non-stochastic specification will be:
E( Y ) = b 0 + b1 X 1 + b 2 X 2 (10.1)
The same model in stochastic form will have ‘error terms’ included:
Y = b 0 + b1 X 1 + b 2 X 2 + m (10.2)
= E( Y ) + m (10.3)
We can use subscripts‘t’ with Y0, X1, X2 and µ to denote ‘tth’ observation.
Thus, the above equations can be written as
38
I Yt = b 0 + b1 X1t + b 2 X 2 t Multivariable Regression
Models
Yt = b 0 + b1 X1t + b 2 X 2 t + m t and
Yt = E (Yt ) + m t respectively
In the relationships above, the component ( β 0 + β 1 X 1t + β 2 X 2t ) is systematic

or deterministic component, which is equal to mean value E(Yt ) ─ or a point
on the regression line.
The component µ t is non-systematic random component determined by factors

other than X1 and X2.
The model specified by equations 10.1 and 10.2 is a linear regression model
which is linear in parameters.
10.2.1 Estimation of Parameters: The Ordinary Least Squares

Approach
The Ordinary Least Squares Approach
We collect a sample of observations on Y1, X1 and X2 and write down sample
regression function
Yt = b0 + b1 X 1t + b2 X 2 t + et (10.4)
Note that b0, b1, b2 have replaced the corresponding population parameters β1,
β2, β3 here. The random component µ t replaced by et or the sample error term.
So, simply speaking:

b0 = the estimator for β0
We know from the Unit 8 that the principle of ordinary least squares selects
those values for unknowns (b0, b1, b2 ) such that residual sum of squares (RSS)
= ∑ et2 is minimum. We can develop this idea by rewriting equation 10.4 as
et = Yt - b0 - b1 X1t - b2 X 2 t (10.5)
Square on both sides and sum up to get RSS:
å et2 = å [Yt - b0 - b1 X 1t - b2 X 2t ]
2
(10.6)
Differentiating the (10.6) w.r.t b0, b1, b2 and equating to zero gives us the
three normal equations:
Y t = b0 + b1 X 1t + b2 X 2 t (10.7)
å Yt X 1t = b0 å X 1t + b1 å X 12t + b2 å X 1t X 2 t (10.8)
å Yt X 2t = b0 å X 2t + b1 å X 1t X 2t + b2 å X 22t (10.9)
39
Quantitative Methods-I These three equations give us the following expressions for b0, b1, and b2 M
respectively:
b0 = Y - b1 X - b2 X 2 (10.10)
b1 =
(å y x )(å x
t 1t ) - (å y x )(å x x )
2
2t t 2t 1t 2t
(10.11)
(å x )(å x ) - (å x x )
2 2 2
1t 2t 1t 2t
and
b2 =
(å y x )(å x
t 2t ) - (å y x )(å x x )
2
1t t 1t 1t 2t
(10.12)
(å x )(å x ) - (å x x )
2 2 2
1t 2t 1t 2 t
The lower case letters denote, as usual, the deviations from the respective
means. Thus y t = ( y t − Y ), x1t = ( X 1t − X ) and x 2t = ( X 2t − X 2 ). .
10.2.2 Variance and Standard Errors

The variances of OLS estimators given in 10.10, 10.11 and 10.12 above are
given in terms of means and deviations of X1 and X2 and the variance of
population error terms µ t. Thus,
æ ö
ç 2 2 ÷
ç 1 X 1 å x2 t + X 2 å x1t - 2 X 1 X 2 å x1t x2t ÷ 2
2 2
Var( b0 ) = ç + ÷s (10.13)
çn å x12t å x22t - ( å x1t x2t )2 ÷
ç ÷
è ø
Standard Error of b0 = Var( b0 ) (10.14)
å x22t
Var (b1 ) = .s 2 (10.15)
å x1t å x2 t - (å x1t x2 t )
2 2 2
SE (b1 ) = Var (b1 ) (10.16)
and
å x12t
Var (b2 ) = .s 2 (10.17)
å x12t å x22t - (å x1t x2 t ) 2
SE (b2 ) = Var (b2 )

(10.18)
When σ2 is unknown and its t unbiased OLS estimator is obtained, we find that
å ei2
sˆ 2 = (10.19)
n-3
The denominator of 10.19 shows the degrees of freedom. In a sample of size
‘n’ we exhaust 3 degrees of freedom in estimating b0, b1, and b2. Note that this
40 reasoning is quiet general, in the sense that if out of a sample of size ‘n’, we
I estimate ‘k’ parameters, remaining degrees of freedom will be n ─ k. We call Multivariable Regression
Models
σˆ = σˆ 2 , the standard error of estimate/regression.
The residual sum of squares in 10.19 is ∑ et2 = ∑(Yt − Yˆt ) 2 . This expression is
equivalent to the following.
å et2 = å yt2 - b1 å y1 x1t - b2 å yt x2 t (10.20
10.3 INTERPRETATION OF REGRESSION

COEFFICIENTS
Mathematically, b1 and b2 represent the partial slopes of regression plan with
respect to X1 and X2 respectively. In other words, b1 shows the rate of change
in Y as X1 alone undergoes a unit change, keeping all other thing, constant.
Similarly, b2 represents rate of change of Y as X2 alone changes by a unit while
other things are held constant.
Note: This interpretation is quite similar to what you have been doing in
economic theory. Recall the law of demand: quantity demanded of a
commodity varied inversely with its price, ‘ceteris paribus’ that is when other
things were kept ‘constant’. What were those other things? Those were prices
of complements and substitutes, income of the consumer, and tastes/likings/
disliking etc.
10.3.1 Goodness of Fit: Multiple Coefficient of

Determination:R2
Recall that is case of a single independent variable, r2 measured goodness of fit
of the fitted sample regression line. It gave you the percentage total variation in
dependent variable Y which has been explained by the single explanatory
variable X. Correspondingly, when we have two explanatory variables X1 and
X2, we would like to know the proportion of total variation in Y = ∑ y t2
explained by X1 and X2 jointly. This information is conveyed by R2 ─ the
multiple coefficient of determination. Thus,
ESS
R2 =
TSS
When ESS is explained sum of squares and TSS is total sum of squares. You
are familiar with the relationship between TSS, ESS and RSS from Unit 8.
You know that ESS = b1 ∑ y t x 1t + b 2 ∑ y t x 2 t
and RSS = ∑ y t2 − b1 ∑ y t x1t − b2 ∑ y t x 2t

Therefore R2 can be computed as
b1 ∑ y t x1t + b2 ∑ y t x 2t
R2 = (10.21)
∑ y t2
You can get the quantities in the above formula from your computation sheet
for the normal equations. The R2 lies between 0 and 1. Closer it is to one;
41
Quantitative Methods-I better is the fit – implying estimated regression line is capable of explaining M
greater proportions of variation in Y. The positive square root of R2 is called
coefficient of multiple correlation.
10.3.2 Analysis of Variance (ANOVA)

We know the relationship:
TSS = ESS +RSS
This is equivalent to saying:
å yt2 b1 å yt x1t + b2 å yt x2t å et2

(TSS) = (ESS) + (RSS)
A study of the components of TSS is called analysis of variance in the context

of regression. You know that every sum of squares has some degrees of
freedom (df) associated with it. In our above 2-explainatory variable case, the
degrees of freedom will be
TSS = n–1
RSS = n–3
ESS = 2
We can put this information in form of a table:
Table 10.1: ANOVA Table for 2-Explainatory Variable Regression

Source of Variation Sum of Squares (SS) df δS
Mean S.S =
df
Due to regression b1 ∑ yt xt + b2 ∑ yt x2 t 2 ESS
(ESS) 2
Due to residuals ∑ et2 n-3 ∑ et2
n−3
TSS ∑ y12 n-1
One may be interested in testing a null hypothesis H0: B1 = B2=0. In such a

case we find that
ESS / df
that is ratio of the variance explained by X1 and X2 to unexplained
RSS / df
variance follows F distribution with 2 and n-3 degrees of freedom. In general,
if regression equation estimates ‘k’ parameters including intercept, than F ratio
has (k-1) df in numerator and (n-k) df in the denominator.
Interpretation: Larger the variance explained by fitted regression line, larger

the numerator will be in relation to denominator. Thus a larger F value is
evidence against the ‘truthfulness’ of H0: B1 = B2=0. That is, F values larger
than one will indicate that hypothesis of both the variables X1 and X2 having no
effect on Y cannot sustain.
42
I We can also express F values in terms of R2. Multivariable Regression
Models
R 2 /(k − 1)
F=
(1 − R 2 ) /(n − k )

1) What is multiple regression? Explain with an example.
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
2) State the assumptions of classical linear multiple regression model.
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
3) What do you mean by linearity of regression model? Explain?
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
4) How do you interpret coefficients of multiple regression model?
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
43
Quantitative Methods-I M
10.4 INCLUSION AND EXCLUSION OF
EXPLANATORY VARIABLES
The Adjusted R 2 : R 2
It has been noted that as we add more and more explanatory variables Xs , the
explained sum of squares keeps on rising. Thus, R2 goes on increasing as we
increase Xs, the explained sum of squares keeps on rising. Thus, R2 goes on
increasing as we increase Xs. But, notice that adding each additional variable
‘eats up’ one degree of freedom and our definition of R2 makes no allowance
for loss of degrees of freedom. Hence, the thinking that you can improve the
goodness of fit by suitably increasing the number of variables may not be
justified  We know that TSS always has (n-1) degrees of freedom. Therefore,
comparing two regression models with same dependent variable but
differencing number of independent variables will not be justified. Therefore,
we must adjust our measure of goodness of fit for degrees of freedom. This
measure is called adjusted R2, denoted by R 2 . It can be derived from R2 as
follows:
2 (n - 1)
R = 1 - (1 - R 2 ) (10.22)
(n - k )
Therefore, it is recommended that, one must include new variables only if

(upon inclusion) R 2 increases and not otherwise. A general guide is provided
by ‘t’ statistic, if absolute value of the co-efficient of added variable is greater
than one, retain it (Note that ‘t’ value is calculated under the hypothesis that
population value of that co-efficient is zero).
10.5 GENERALISATION TO N-EXPLANATORY

VARIABLES
In general, our regression model may have a large number of independent
variables. Each of those variables can, on priority grounds, be expected to have
some influence over the ‘dependent’ or ‘explained’ variable. Consider a very
simple example. What can be determinants of demand for potatoes in a
vegetables market? One obvious choice will be the price of potatoes. What else
can affect the quantity demanded? Could it be availability of vegetables which
can be paired off with potatoes? In that case, prior of a large number of
vegetables which are cooked along with potatoes will become ‘relevant
explanatory variables’. You cannot ignore income of the community that
patronizes the particular market. Needless to say, the dietary preferences of
members of the households can also affect the demand and so on. In the next
Unit, we shall discuss techniques which help us restrict the analysis to a
selected ‘few variables, though theoretic considerations may find a huge
number of them to be ‘useful’ and ‘powerful’ determinants. In fact, in
economic theory, we usually append the phrase Ceteris paribus, with many a
statements. This phrase means keeping all other things constant. That means,
we may focus on impact of only a few selected variables on the dependent
variable while assuming that all other variables remain ‘unchanged’ during the
period of analysis. This assumption may not hold so, we have to juggle the
need to include more and more variables in our model with the ‘gains’ in
44
I explainatory power of the model. We have developed, in previous section Multivariable Regression
Models
(10.4) a working touchstone for inclusion of more variables in terms of
improvement in R 2 and have tried to give it ‘practical’ shape in form of
magnitude of ‘t’ values of the relevant slope parameters.
With these considerations in mind we can generalise the linear regression

model as follows:
We hypothesize that in population, the dependent variable Y depends upon n

explainatory variables, X1, X2, ………..Xn. We also assume that the
relationship is linear in parameters. Three more assumptions are made and they
have very significant bearing on the analysis. These are:
a) Absence of Multi-co-linearity;
b) Absence of Hetero-scedasticity; and
c) Absence of Autocorrelation
We will discuss the complications, which arise because of violations of these

assumptions in section 10.6, 10.7 and 10.8 respectively. So our Classical
Linear General Regression Model is :
Yt = β 0 + β 1 X 1t + ⋯⋯⋯⋯⋯ + β n X nt
or
where it is understood that Y and each X will have a large numbers of values
t=1………….N, forming (n+1) ‘tuples’ (Y11 X11, X21, ……………..Xn1).
We can simply write
Y = b1 X 1 + b 2 X 2 .......................b n X n + m (10.23)
in Matrix equation form, we get
Y = Xβ+U (10.24)
Y1  β1  U 1 
Where Y = ⋮  , β = ⋮  and U = ⋮ 
Yn   β n  U n 
 X 11 , X 21 , X 31 , ⋯⋯⋯⋯⋯ X k 1 
 
 X 12 , X 22 , X 32 , ⋯⋯⋯⋯⋯ X k 2 
also X = ⋮ 
 
⋮ 
 X 1n , X 2n , X 3n , ⋯⋯⋯⋯⋯ X kn 
 
We assume that (1) expected values of error terms are equal to zero; that is
E(ui) = 0 for all ‘i’. In matrix notation
45
Quantitative Methods-I 0 M
E(u i )   
E (u ) = ⋮  = 0 = 0
 ⋮ 
 E (u n )  
0
2) The error terms are not correlated with one another and they all have
same variance for σ 2 all sets values of the variables X. That is,
E (u i u j ) = 0; ∀i ≠ j and
E (u12 ) = σ 2 ∀i
in matrix notation:
 E (u12 ), E (u1u2 )⋯⋯⋯⋯⋯ E (u1un )  σ 2 0 0 0

   
 E (u1u2 ), E (u2 )⋯⋯⋯⋯⋯ E (u2un )   0 σ
2 2
σ 0
E UU  = 
'
 = = σ 2In
⋮  0 0 σ 2
0 
   
 E (u1un ), E (u2 un )⋯⋯⋯⋯ E (un )   0
2
0 0 σ2
3) Number X1i……………..Xk are all real numbers and are free of random
errors or disturbances.
4) The matrix X has been linearly independent columns. It implies that

number of observations exceeds number of co-efficient to be estimated. It
also implies there exists no exact linear relationships between any of the X
variables.
Note: Assumptions that E(UiUj) = 0 means that error terms are not correlated.
The implication of diagonal terms in matrix E(UU′) being all equal to σ 2 is
that all error terms have same variance, σ 2 . This is also called assumption of
homo-scedasticity. The last assumption implies absence of multi-co-linearity.
We can write the regression relation for the sample as:
e = Y – Xb
where e, Y, X and b are appropriate matrices.
Sum of squared residuals will be
φ = ∑ e12 = ∑(Yi − b1 X 1i ⋯⋯⋯ + bk X ki ) 2 (10.25)

′
= e ′e = [Y − Xb] [Y − Xb]
= Y 1Y − 2b′X ′Y + b ′X ′Xb
′ ' ′
Note: b X Y is scalar and is therefore equal to its transpose Y Xb.
by equating 1st order partials of φ w.r.t each bi, to zero, we get k normal
equations. This set of equations in matrix form is:
46
I ∂φ Multivariable Regression
= −2 X ′Y + 2 X ′Xb = 0 (10.26) Models
∂b
X ′Xb = X ′Y (10.27)
when X has rank equal to k, the normal equation 10.27 will have a unique
solution and least squares estimator b is equal to:
b = [X ′X ] [X ′Y ]
−1
(10.28)
We have assumed b to be estimator for β and thus E(b) = β , therefore we
can rewrite 10.28 as
b = [X ′X ] X ′′[Xβ + u ]
−1
= [X ′X ] X ′Xβ + [X ′X ] X ′U
−1 −1
[
= β + X 1 X X 1U ]
1
[
∴ E (b) = E (β ) + E X 1 X ]
−1
]
X 1 E ( µ ) = E ( B) + [X ′X ]. X ′E ( µ )
=β
Variance of b = σ 2 ( X ′X ) −1
Notes
1) In this course our objective is simply to introduce the concepts. Those who
plan to pursue the concepts at much more rigorous level can study our
course on Econometric Methods (MEC) – included as optional course of
M.A.(Economics) Programme.
2) The other ideas regarding coefficient of determination R2 and adjusted R2

remain the same as they were developed for two explainatory variable case.
Now we can safely turn to discussions about non-satisfaction or violation of

assumptions.
10.6 THE PROBLEM OF MULTI-CO-LINEARITY

Many a times our X variables may be found to have some other linear
relationships among themselves. This vitiates our classical regression model.
Let us illustrate it with help of our 2 explainatory variable model.
Yi = β 0 + β 1 X 1i + β 2 X 2i + U 1
Let us give specific names to variables, say, X1 is price of commodity Y and X2
is family income. We expect β1 to be negative and β2 to be positive. Now we go
one step further. Let Y be demand for milk, X1 be price of milk and let the
family wise demand for milk is being estimated for a family, which produces and
sells milk! Clearly, larger the value of X1 higher the magnitude of X2 will be.
In such situations, the estimation of co-efficient will not be possible. Recall, we
wanted variables X in our matrix equations to be linearly independent. If that
conditions is not satisfied X matrix becomes singular, that is its determinant
tends to vanish. Thus, there will be no solution to the normal equation 10.26
(or 10.27).
47
Quantitative Methods-I However, if co-linearity not perfect, we can still get OLS estimates and they M
remain best linear unbiased estimates (BLUE) – though one or more partial
regression co-efficient may turn out to be individually insignificant.
Not only this, the OLS estimates still retain property of minimum variance.
Further, it is found that multi co-linearity is essentially a sample regression
problem. The X variables may not be linearly related in population but some of
our suppositions while drawing a sample may create a situation of multiple
linear relations.
The practical consequences of multi-co-linearity. Gujarali, (D.N.) has listed

the following consequences of multiplicity of linear relationships:
1) Large variances /SEs of OLS estimates

2) Wider confidence intervals
3) Insignificant ‘t’ ratios for β parameters
4) A high R2 despite few significant t values
5) Instability of OLS estimators: The estimators and their standard errors
(SEs) become very sensitive to small changes in data.
6) Sometimes, even signs of some of the regressions may turn out to be
theoretically unacceptable like a rise in income having negative impact on
demand for milk.
7) When many regressions have insignificant coefficients, their individual
contributions to the explained sum of squares cannot be assessed properly.
The multi-co-linearity can be detected by:

1) high R2 but few significant ‘t’ ratios,
2) high pair wise correlation between explainatory variables. One can try
partial correlations, subsidiary or auxiliary regressions as well. But each
such technique increases burden of calculations.

1) When do you decide to drop a variable from the regression equation? Why?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) When do you include more variable(s) into your model? Why?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
48
I 3) “Inclusion of more variables always increases R2 the goodness of fit. So to Multivariable Regression
Models
make a regression model ‘good’ what we need to do is simply increase the
number of explanatory variables”. Do you agree/disagree with this
statement? Give reasons.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) What is multi-co-linearity? What are its consequences?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
10.7 PROBLEM OF HETERO-SCEDASTICITY

The Classical Linear Regression Model has a significant underlying assumption
in homo scedasticity, that is, all the error terms are identically distributed with
mean equal to zero and standard deviation equal to σ (or variance equal to σ2).
σ2 What happens when this second part of assumption regarding distribution of
variance does not hold? As a result, in symbolic terms E (u i ) 2 = σ i2 , that is, if the
expectation of squared errors is no longer equal to σ2  each error term has its
own σ2, or variance varies from observation to observation.
It has been observed that usually time series data does not suffer from this
problem of hetero scedasticity but in cross-section data, the problem may
assume serious dimensions. The consequences of hetero scedasticity: If the
assumption of homoscedasticity does not hold, we observe the following
impact on OLS estimators.
1) They are still linear

2) They are still unbiased
3) But they no longer have minimum variance – that is we cannot call them
BLUE: the Best Linear Unbiased Estimators. In fact, this point is relevant
both for small as well as large samples.
4) Reason for this problem hinted at in (3) above is that generally, OLS
estimators have some bias built into their formulae. We try to rectify that
making use of degrees of freedom.
For instance σˆ 2 , (the estimator for true population σ2) given by ∑ e12 df
no longer remains unbiased. And this very σˆ 2 enters into calculation of
standard errors of OLS estimates.
49
Quantitative Methods-I 5) Since, estimates of standard errors are themselves no longer reliable, we M
may end up drawing wrong conclusions using conventional reasoning
based on procedures for testing the hypothesis.
How to detect Hetero scedasticity

In applied regression analysis, plotting the residual terms can give us important
clues about whether or not one or more assumptions underlying our regression
model hold. The pattern exhibited by ei2 plotted against the concerned variable
can provide important clue. If no pattern is detected – homoscedasticity holds
or hetero scedasticity is absent. On the other hand, if errors form a pattern with
variable  expanding, increasing linear or changing is some non-linear manner
thus hetero scedasticity is certainly present.
Some tests have been designed to detect presence of Hetero scedasticity, using
various statistical techniques. Prominent ones are: Park Test, Glejser Test,
Whites General Test, Spearman’s Rank correlation Test, Goldfield - Quadnt
Test etc. But in this unit, the limitation of space does not permit us to go into
their details. We are forced to refer the learners again the course on
Econometric Method for details in this regard.
How to tackle the hetero scedasticity?

Our ability to tackle the problem will depend upon the assumptions. We can
really make about error variance. Thus, the following situations may emerge
i) When σ 12 is known
Here the CLRM
Yi = β 0 + β 1 X 1 + u i can be transformed, dividing each value by

corresponding σ1 thus,
Yi  1  X  Ui
= β 0   + β 1  i  +
σ1 σ1   σi  σi
Ui
This effectively transforms error terms to U i = which is homo-scedastic
σ1
and therefore, the OLS estimators will be free of disability caused by
hetero scedasticity. The estimates of β0 and β1 in this situation are called
Weighted Least Squares Estimators (WLSEs).
ii) When σ2 is unknown: we make some further assumptions about error

variance:
iia) Error variance proportional to the Xi s. Here, the Square Root

transformation is enough. We divide on both sides by X i . Thus, our
regression line looks like:
Yi b X U
= 0 + b1 i + i
Xi Xi Xi Xi
50
I 1 Multivariable Regression
= b0 + b1 X i + U i Models
Xi
Ui
Here U i = and this is sufficient to address the problem.
Xi
iib) Error Variance proportional to X12. Here, instead of division by X i , we

divide by Xi on both the sides and estimate
Yi 1 U
= b0 + b1 + i
Xi Xi Xi
1
= b0 + b1 + U i
Xi
Ui
The error term will be U i = and this will be free of hetero scedasticity,
Xi
facilitating use of CLS techniques.
iii) Respecification of Model: Assigning a different functional form to the

model, in place of speculating about the nature of variance may be found
expedient. We can estimate this model:
ln Yi = β 0 + β1 ln X i + U i
This loglinear model is usually adequate to address our concerns.
10.8 PROBLEM OF AUTOCORRELATION

The classical regression model also assumes that disturbance terms Uis do not
have any serial correlation. But, in many situations this assumption may not
hold. The consequences of the presence of serial or auto correlation are similar
to those of hetero scedasticity: the OLS are no longer BLUE.
Symbolically no autocorrelation means E(Ui Uj)= 0 when i ≠ j .

Autocorrelation can arise in economic data on account of many factors:
i) there may be various reasons which cause cyclical up and down swings in
economic time series. These tendencies continue till something happens
which reverse them. This is called inertia.
ii) Misspecification of model in the form of under specification can be a
cause of autocorrelation: you have fewer X variables in the model,
leaving out rather large systematic components to be clubbed with errors.
iii) Cob-web phenomenon is another factor which may create this problem in
certain types of economic time series (especially agricultural output etc.).
iv) Polishing of data – like adding monthly data to make quarterly or
quarterly to make half yearly series etc. can also be responsible for the
autocorrelation – as the averaging involved dampens the fluctuations of
the original data.
51
Quantitative Methods-I The consequences of autocorrelation are not different from those of hetero M
scedasticity listed in 10.7 above. Here too OLS estimators are biased or are not
BLUE, t & F tests are no longer reliable. Therefore, computed value of R2 is
not reliable estimate of true goodness of fit.
There are many tests for detecting autocorrelation – varying from visual
inspection of error plots, the Runs Test, Swed-Eisenhart critical runs test. But
most commonly used in Durbin-Watson d test defined as
n
∑ (e
t =2
t − et −1 ) 2
d= n
∑e
t =1
2
t
However, again, we are holding back information on practical detections and

avoidance of problem of autocorrelation for the reasons of limitation of space
here.
10.9 THE MAXIMUM LIKELIHOOD

ESTIMATIONS
Sometimes another class of estimators is used in place of OLS. This class of
estimators is called maximum likelihood estimators (MLEs). The MLEs
possess some stronger theoretical properties. But it also requires a stronger
assumption about distribution of error terms. Moreover, when errors follow
normal distribution, we find that OLS and MLE methods give identical
estimates of β parameters, both in simple and in multiple regressions.
However, MLE estimate of σ2 is biased. Hence, if one uses assumption of
normal distribution of Uis and persists with OLS, one does not miss out on any
thing that may be advantageous in MLE.
The Method: Simple Regression Illustration

In the model Y1 = β 0 + β1 X 1 + U i , the Y1 is normally and independently
distributed with means β0 + β1Xi and variance σ2. Therefore, the joint
probability density function of
Y1, Y2, Y3…………………….Yn with above mean and variance will be
(
f Y1 .................Yn β 0 + β1 X i , σ 2 ) ML - 1
But given the independence of Ys, this function can be written as product of ‘n’
individual density functions, or
(
f Y1 .................Yn b 0 + b1 X i + U i , s 2 ) ML-2
= π ëéYi b 0 + b1 X i + U i , s 2 ûù
é 2ù
ê-Y (Yi - b 0 - b1 X i ) ú
ê ú
1 ê
2
s2 ûú
where f (Yi ) = .e ë ML-3
s 2p
52
I Putting this value for each Yi, in ML-2. We get the likelihood function (LF): Multivariable Regression
Models
 1 (Yi − β 0 − β1 −U i )2 
1 − ∑ σ2

 2 
LF ( β 0 , β 1 , σ 2 ) = .e ML - 4
σ n
( 2π )
n
The method of maximum likelihood estimation is nothing but maximization of

the (LF) given in ML-4 above. We can differentiate logarithms of (LF) with
respect to β0, β1 and σ2 and equate the respective partials to zero to get the
requisite estimation relations. So as
1 (Y − β 0 − β1 X 1 − U i )
2
n
ln( LF ) = n nσ 2 − ln( 2π ) − ∑ i ML - 5
2 2 σ2
therefore
∂ ln (LF) − 1
= 2 ∑ (Y1 − β 0 − β1 X i )(−1) = 0 ML - 6
∂β 0 σ
∂ ln (LF) − 1
= 2 ∑ (Y1 − β 0 − β1 X 1 )(− X i ) = 0 ML - 7
∂β1 σ
∂ ln (LF) − n 1
4 ∑ 1
= + (Y − β 0 − β1 X i )2 = 0 ML - 8
∂σ 2
2σ 2
2σ
Simplification of ML-6 and ML-7 give us
å Yi = nb 0 + b1 å X i ML-8a
å X iYi = b 0 å X i + b1 å X i2 ML-9
Which are same as OLS normal equations.
Substituting values of ML (=OLS) estimates obtained by simultaneously

solving ML-8 and ML-9 into ML-7 gives us
1
σ2 =
n
∑ (Yi − β 0 − β1 X i ) 2
1
=
n
∑ U i2 ML - 10
1
Expectation of σ~ 2 , E (σ~ 2 ) = E (∑ U i2 )
n
n−2 2
= σ
 n 
2
σ2 − σ2
n
or σ~ 2 is biased downwards.
Multiple Regression: An example: The following regression were run on

SPSS. Edited summary of results is as follows:
53
Quantitative Methods-I India’s Imports, Exports and Foreign Investment Inflows M
All figures are in Millions of US dollars

Year var00001 var00002 var00003
1991-1992 19411.00 17865.00 133.00
1992-1993 21882.00 18537.00 559.00
1993-1994 23306.00 22238.00 4153.00
1994-1995 28654.00 26330.00 5138.00
1995-1996 36678.00 31797.00 4892.00
1996-1997 39133.00 33470.00 6133.00
1997-1998 41484.00 35006.00 5385.00
1998-1999 42389.00 33218.00 2401.00
1999-2000 49671.00 36822.00 5181.00
2000-2001 50536.00 44560.00 5862.00
2001-2002 51413.00 43827.00 6693.00
2002-2003 61412.00 52719.00 4555.00
var00001=Imports
var00002=Exports
var00003=Foreign Investment Inflow
Regression of Imports on Exports and Foreign Investment Inflow

R R2 R
2
Standard error
of Estimate
0.983 0.966 0.958 2726.228
Coefficients
B Standard Error t
Constant -1655.737 2658.578 -0.623
var00002 1.263 0.104 12.192
var00003 -0.288 0.522 -0.552
Coefficients Correlation
Correlation Var00003 Var00002
Var00003 1.0000 -0.670
Var00002 -0.670 1.0000
Covariance Var00003 0.272 -3.625E-02
Var00002 -3.625E-02 1.073E-02
54
I ANOVA Multivariable Regression
Models
Sum of Sqrs. df Mean sqrs. F
Regression1.9E+09 2 9.4E+08 127.097
Residual 6.7E+07 9 7432320
Total 2.0E+09 11
Regression of Foreign Investment Inflow on Exports and Imports

Standard error
R R2 R-2
of Estimate
0.684 0.468 0.349 1712.334
Coefficients
B Standard Error t
Constant -321.545 1702.075 -0.189
var00001 -0.114 0.206 -0.552
var00002 0.272 0.257 1.060
Coefficients Correlation
Correlation Var00002 Var00001
Var00002 1.0000 -0.982
Var00001 -0.982 1.0000
Covariance
Var00002 6.590E-02 -5.92E-02
Var00001 -5.192E-02 4.24E-02
ANOVA
Sum of Sqrs. df Mean sqrs. F

Regression 2.3E+07 2 1.2E+07 3.952
Residual 2.6E+07 9 2932089
Total 5.0E+07 11
The above regressions are based on certain ‘a prior’ expectations. We expect

that imports into India during the period 1991-92 to 2002-03 depend upon
exports from India and foreign investment inflows into the country. The idea is
that exports pay for imports and foreign investment inflow ‘facilitates’ the
country to import more. Our results confirm our theoretic expectations.
Regression on imports on exports and investment inflows shows that R2 =
0.966 and R 2 = 0.958 . This shows that model has high explanatory power. It
explains over 95 per cent of the variation. Typical computer output gives
information on coefficients, correlation and co-variances, co-linearity diagnostics,
residual analysis of variance etc. We have run another regression too − the
results of which are reported above. This is regression of foreign investment
55
Quantitative Methods-I inflows on exports and imports. Theoretic expectation was that investment M
inflows are determined by magnitudes of exports and imports. However, this
model gives rather disappointing results. R2 = 0.468 and R 2 = 0.349 only. In
other words, our model explains less than 35 per cent of the variations. It is not
desirable to persist with it. The same is reflected in low values of β1 & β2
parameters and their high standard errors.

1) What is hetero-scedasticity? What are its consequences?
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
2) What is auto-correlation? When does it arise? What are its consequences?

……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
3) What are maximum likelihood estimations? Why do we persist with least

squares estimators most of the time.
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
……………………………………………………………………………
10.10 LET US SUM UP

We began with extension of the simple linear regression model to incorporate
one more explanatory variable. Our next step was to interpret the coefficients
of partial regression. Afterwards, we tried to design statistical touch stone for
56
I inclusion of more variables into the model. This also gave us some guidelines Multivariable Regression
Models
to drop the ‘undesirable’ variable as well. We then moved on to more tedious
and mathematically more demanding extension to ‘n’ variable model. We then
attempted to analyse the effects of multi-co-linearity, hetero scedasticity and
auto-correlation respectively. We also made comments on techniques of
identification of these problems and some strategies to get rid of the problems
as well. Further, we discussed the class of estimators called maximum
likelihood estimators (MLEs). We made comparison between MLEs and Least
square estimators as well. Finally, we ran some regressions on SPSS package
between India’s imports, exports and foreign investment inflows to illustrate
some of the points, which arose in course of our discussions. However, we
must again draw the attention of learners to the fact that limitations of space
did not permit us to go in for an exhaustive discussion of the concepts touched
upon. Those who want to have detailed knowledge about the issues and
concepts involved in multiple regression models are advised to go through our
optional course MECE-001 Econometric Methods.
10.11 KEY WORDS
Multiple Regressions : Regression of one dependent variable on more

than one independent variables.
Partial regression : Coefficients of individual predictors in multiple

Coefficients regressions.
Coefficient of multiple : Ratio of variation explained or accounted for by

determination  R2 the regression model to the total variation.
Adjusted coefficient of : It is coefficient of determination adjusted for the

multiple determination loss of degrees of freedom. It helps us against
 R 2
increasing R 2 by incorporating unnecessary
explanatory variables.
Multi-co-linearity : Existence of linear relationships between

explanatory variables in addition to one between
dependent variable and them.
Hetero scedasticity : Differences between variances of the error

terms. This problem is more serious in cross-
sectional data.
Auto-co-relation : Correlations between the errors and their

previous values. This problem is encountered
very often when dealing with time series data.

PROGRESS EXERCISES
1) Read Section 10.2
57
Quantitative Methods-I 4) Read Section 10.3 M

58
I Measures of Inequality
UNIT 11 MEASURES OF INEQUALITY
Structure
11.0 Objectives
11.1 Introduction
11.2 Positive Measures
11.2.1 Relative Range
11.2.2 Relative Inter-Quartile Range
11.2.3 Relative Standard Variation
11.2.4 Standard Deviation of Logarithms
11.2.5 Champernowne Index
11.2.6 Hirschman-Herfindahl Indices
11.2.7 Kolm’s Index
11.3 Gini Index
11.3.1 Gini as a Measure of Dispersion
11.3.2 Simple Computational Device
11.4 Lorenz Curve
11.4.1 Geometrical Definition
11.4.2 Properties of Lorenz Curve
11.4.3 A Measure Based on Area
11.4.4 A Measure Based on Length
11.5 Normative Measures
11.5.1 Dalton Index
11.5.2 Atkinson Index
11.5.3 Sen Index
11.5.4 Theil Entropy Index
11.6 Let Us Sum Up
11.7 Key Words
11.8 Exercises
11.0 OBJECTIVES
• explain the various positive measures of inequality;
• discuss the computational device for construction of Gini index; and
• describe the normative measures of inequality propounded by Dalton,
Atiknson, Sen and Theil.
59
11.1 INTRODUCTION
Improvement in well-being of the poor has been one of the important goals of
economic policy and to a significant extent it is determined by the growth and
distribution of its income. Distribution patterns have an important bearing on
the relationship between average income and poverty levels. Extreme
inequalities are economically wasteful. Further, income inequalities also
interact with other life-chance inequalities. Hence reducing inequalities has
become priority of public policy.
It therefore, becomes pertinent to measure income inequalities. Various

measures have been developed over a period of time to study the level of
inequalities in different situations. Broadly these measures can be put under
two categories (i) positive measures, and (ii) normative measures. The
measures which capture the inequality of income without value judgment about
social well-being are known as positive measures. Range, quartile range,
standard deviation, Gini ratio, Lorenz curve etc. are positive measures of
inequality. On the other hand, the measures that essentially involve value
judgement about social welfare are called normative measures. The index
propounded by Dalton, Atkinson, Sen, Theil and Kakwani are normative
measures. We shall discuss all these measures one by one in this unit. Gini
Coefficient of Inequality and Lorenz Curve will receive particular attention, as
they are very popular in literature.
11.2 POSITIVE MEASURES

If all values in a distribution are not equal, which means that there is dispersion
in the distribution, hence, there exists inequality in the distribution. If a
measure is developed to capture this non-equality in values without giving
explicit consideration to its consequences with respect to social well-being or
economic significance in a particular context, the measure is known as
positive. It means that the measure is bothering about the fact whether it is
measuring inequality of lengths of iron nails or incomes of wage earners in a
village. Nevertheless, many of them are standard statistical measures and their
social implications and/or social consequences can still be studied. Some of
these measures can be arrived from normative approach as well.
Let us consider an income distribution xi i=1,2,…,N over N persons and with

mean income . Let the relative share of total income with person i be
designated as qi, which is naturally given by xi/N. The cumulative share of
total income with the persons not having more than xi income can be given as
Qi. However, when we have an income having frequency more than one, the
proportion of people with income xi can be denoted by pi and the cumulative
proportion of people with income no less than xi as Pi. In this case, obviously
the relative share of total income would be given by f i xi / Nµ where f i
denotes the frequency of occurrence of income xi and N= f i . It will not be
necessary to use two different subscripts for distinguishing the two cases for
the purpose as the context would make it clear whether subscript i stands for a
person with income xi or for the group of persons with income xi .
60
I We have implicitly assumed that the data are arranged in the increasing (non- Measures of Inequality
decreasing) order by magnitude of income so that symbolic representation is
easy. The same could be accomplished by decreasing (non-increasing) order.
We intend to cover important measures along with their variants in this Unit.
We shall also, albeit briefly, discuss their properties and weaknesses.
To recapitulate, we recount what we said above in a formal way
xi fx
qi = or i i
Nµ Nµ
i
Qi = ∑ qi ,
i −1
1 f
pi = or i
N N
i
Pi = ∑ pi ,
i −1
11.2.1 Relative Range

A measure of relative dispersion can be taken a measure of inequality. It is
defined as the relative range by
Maxi xi − Mini x i
RR1 = (RR.1)
µ
that is, the relative difference between the highest income and the lowest
income. If income is equally distributed, then RR1 = 0 and if one person
received all the income, then RR1 is maximum. If one wants to make the index
lie in the interval between 0 and 1, one can define it as
Maxi xi − Mini xi
RR2 = (RR.2)
Nµ
which means it is the gap between the maximum share and the minimum share.
That is,
RR2 = Maxi qi − Mini q i (RR.3)
Though Cowell has suggested division of range by Mini xi,, which does not
serve, in our view, any purpose. Two other normalization or standardization
procedures that make it unit-free and contain it in (0,1) interval are suggested
below:
Maxi xi − Mini xi
RR3 = (RR.4)
Maxi xi
and
Maxi xi − Mini xi
RR4 = (RR.5)
Maxi xi + Mini xi
61
Quantitative Methods-I The basic weaknesses of these range-based measures are that they are not M
based on all values and therefore they do not reflect the change in inequality if
there is any transfer of income between two non-extreme recipients.
Instead of considering extreme values at either end, which may not be even
known, some scholars have toyed with the idea of the ratio between the mean
income of the highest fractile (percentile, quintile or decile) and that of the
lowest counterpart. They term it as the extreme disparity ratio (EDR).
Naturally, this ratio is not contained in the interval (0.1). This ratio is
independent of µ as well. The measure will not reflect the transfer of income
that does not involve the extreme fractiles.
11.2.2 Relative Inter-Quartile Range

Sometimes, extremism of the relative range is sought to be moderated by
restricting the distribution between the 10th and 90th percentile or sometimes to
interquartile range. Bowley (1937) suggested relative quartile deviation as the
index of inequality:
3 1
xq − xq
B= 3 1
(B.1)
xq + xq
r
where x q represents the income level which divides the population in r and
(4-r) quartiles. B is zero for degenerate distribution where everybody has the
same income and unity if the lowest 75 per cent people have no income at all.
Though the extremes are moderated in comparison to the measure of range, it

has an obvious weakness that the measure takes into account only 50 per cent
of the distribution. Further, a transfer of income between two persons without
causing either or both of them cross x q1 or x q 3 would not change the measured
level of inequality. Thus, the index suffers from all weaknesses of the earlier
proposals except that of extremism. Its highest value reaches when the lowest
75 per cent people do not possess any income.
A variant of this measure is inter-quartile ratio, which can be defined as the

75th percentile (3rd quartile) income minus 25th percentile (1st quartile) income
divided by the median ( x q 2 ) income.
11.2.3 Relative Standard Variation

The standard deviation divided by the mean can be used as one measure of
dispersion. It is:
σ
RSD = (RSD.1)
µ
where σ and µ are standard deviation and mean of the distribution.
It can be equivalently defined as the standard deviation of relative incomes.

Using definition of σ , one can find out that it lies in the interval of 0 and
(N –1)1/2, not in (0,1). The highest value depends on the size of distribution.
62
I Since the measure uses all values, any transfer of income would be reflected in Measures of Inequality
the measure. However, it should be noted that the measure is equi-sensitive to
transfers at all levels. Whether a given amount d is transferred
between x j =Rs.400 and x k =Rs.500, or between x j =Rs.10,000 and x k =
Rs.10,100, the changed in RSD is exactly the same.
We may finally note that the square of RSD is also quite often used as another
measure of inequality, which is known as the coefficient of variance. Quite a
few scholars suggest use of variance as a measure of inequality but we have not
considered it here primarily because it is not unit-free. We think that an
inequality measure must be unit-free.
11.2.4 Standard Deviation of Logarithms

One way of attaching greater importance to transfers at lower end (as required
by Sen) is to consider some transformation of incomes. This transformation
can easily be attained by considering the logarithms that stagger the income at
lower levels.
This measure is defined in either of the following two ways:
1/ 2
1 N

SDL1 = 
N
∑
i =1
(log µ − log xi ) 
2

(SDL.1)
1/ 2
1 N

SDL2 = 
N
∑
i =1
(log µˆ − log xi ) 
2

(SDL.2)
where µ and µ̂ are the arithmetic and geometric means respectively. While
standard statistical literature prefers use of geometric mean the more common
practice in literature on income inequality is one of using arithmetic means.
Cowell (1995) prefers to define these in terms of variance and calls the square
of SDL1 as the logarithm variance (V1) and the square of SDL2 as the variance
of logarithms (V2). Name of the second is clear from the expression but that of
the first is derived from the fact that (log x-log x ) could be written as log
(x/ x ). One can see that V1 is equal to V2 plus log ( µ̂ /µ).
As these measures are in terms of ratios of incomes, any proportionate change

in incomes would leave the magnitude of inequality unchanged when measured
by these indices. But, unfortunately, a transfer from a richer person to a poorer
person may raise the magnitude of inequality, particularly if the poorer person
has income more than 2.72 times the mean of the distribution.
While the lower limit, irrespective of formula, is zero when everybody has the
same income, the upper limit depends on the size of distribution and
approaches infinity when N is large and when everybody except the richest,
receives income equal to one unit (as zero is inadmissible in logarithmic
transformation.) Further, if we face grouped data, it is convenient to use µ in
place of µ̂ and µ i in place of xi .
63
Quantitative Methods-I The variance of logarithms is however decomposable. It is a property that is M
being given emphasis of late. It can be shown that V2 is the sum of between-
group component and within group component, latter being population-
weighted sum of within-group V2 ’s.
11.2.5 Champernowne Index

Champernowne (1973) makes use of the idea of geometric mean. It is a well
known fact of an unequal distribution that its geometric mean is smaller than
the arithmetic mean. The additive inverse of the ratio of geometric mean to
arithmetic mean can duly be considered as an index of inequality. Formally, the
index could be written as:
µ̂
CII = 1 − (CII.1)
µ
where µ and µ̂ are, as stated earlier, arithmetic and geometric means of the
income distribution. It is easy to see that its value is bound between 0 and 1.
One can obviously think of another measure where geometric mean is replaced
by harmonic means.
These measures are sensitive to transfer to income and change is greater when
the transfer takes place at lower end of the distribution. They are sensitive to
transfer of income between two persons. One can try it by replacing x j and x k
by ( x j -d) and ( x k +d) respectively and finding out the direction of the change.
Or, one can use differential calculus.
The trouble with these indices is that they cannot be defined when any of the
income is zero.
11.2.6 Hirschman-Herfindahl Indices

These indices were developed in the course of studying the commodity
concentration in trade by Hirschman (1945) and in characterizing market
monopoly in industry by Herfindahl (1950). Later, they were more used in
capturing autonomy and dependence of units in a federation.
If each unit is a class in itself, pi =1/N, i=1,2…,N. Then concentration could be

captured through use of qi ’s . As the sum of qi ’s is always 1, Hirschman
devised a measure which would capture the inequality among them. He
proposed square root of the sum of squares of shares qi i=1,2…,N. That is,
1/ 2
 N 
H 1 =  N ∑ qi2  (H.1)
 i =1 
which could be generalized as

1/ a
 N 
H =  N ∑ q ia 
*
1 ,a > 1
 i =1  (H.2)
64
I Herfindahl devised a very similar measure, which has been more popular than Measures of Inequality
the original (H.1). This is just the sum of share squares:
N
H 2 = ∑ qi2 (H.3)
i =1
N
H 2* = ∑ qia , a ≥ 1. (H.4)
i =1
It is clear that, besides inequality among the shares, the value of these measures
depends on N—the fewness or largeness of the number of units. For N=2, it
has been suggested that (1/N) could be subtracted from (H.3)
N
1
H 3 = ∑ q i2 − (H.5)
i =1 N
The minimum value of H 3 is zero. But it serves no great purpose. When N=2,
for q1 =0.99 and q 2 =0.01, while H 2 =0.98, H 3 =0.48. H 2 scores definitely
better than H 3 in characterizing the scene of monopoly.
11.2.7 Kolm’s Index

Let there be N incomes such that N= nm where n is the number of different
incomes and each income has m recipients. The number of equal pairs with a
given income would be m(m-1)/2 and total number of equal pairs would be
n.m(m-1)/2. Total number of all pairs would obviously be N(N-1)/2-nm(nm-
1)/2. One can think of an `equality’ index in terms of nm(m-1)/nm(nm-1)=(m-
1)/(N-1). The inequality index could then be constructed by subtracting it
from1: 1-(m-1)/(N-1)-(N-m)/(N-1)=m(n-1)/(nm-1)=(nm-m)/(nm-1). In case,
income x1 has f i recipients, the measure is:
Σf i 2 − N N 2 − Σf i 2
K = 1− = (K.1)
N ( N − 1) N ( N − 1)
The purpose of developing this curiosum due to Kolm (1996) is just to make
one feel that there could be a variety of simple ways to approach the issue of
measurement of inequality.

1) Define relative range measures of inequality. List out relative merits.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
65
Quantitative Methods-I 2) How relative inter-quartile range is better than relative range? M
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
3) What is the relative mean deviation? If a transfer of income is between two
persons both having income lower than the mean, will it change the
magnitude of this index?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
4) Compare the two versions of standard logarithmic deviations.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
5) What is import of Champernowne Index?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
6) What is Hirfindahl index? What are its areas of application?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
66
I 7) What is the message from the Kolm’s index? Calculate the Kolm index for Measures of Inequality
a distribution, which frequency 5 for value Rs.5 lakh and frequency 5 with
value Rs.10 lakh and therefore total size 10 and the arithmetic mean 7.5.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
11.3 GINI INDEX

This coefficient of concentration, as it is usually called, owes to an Italian
statistician by the name of Corrado Gini (1912). Modern practice is to call it
just Gini. This index in its origin is positive. There are a number of ways in
which this coefficient can be expressed. There are also a number of ways in
which it can be interpreted. People have also derived it as a measure of
inequality under plausible axioms in welfare theoretic framework. The index
satisfies many axioms proposed in literature for an index of inequality.
First, we shall discuss those definitions and expressions, which can be derived
as a measure of dispersion. Besides giving its expressions for its frequency
data for grouped observations, we shall discuss its welfare theoretic
interpretations.
11.3.1 Gini as a Measure of Dispersion

Recall that mean deviation and standard deviation, which are measures of
dispersion, seek the deviation from arithmetic mean. Also recall that one of the
logarithmic measure sought deviation from the geometric mean. However, one
may ask why to seek dispersion in terms of deviations from any mean? Why
not compare all pairs and seek the differences. In order to consider positive
values of differences, we either take modal values of deviations before
averaging (mean deviation) or sum the squares of deviations and take root of
the mean of the squared differences.
Corrado Gini (1912) proposed to consider all the differences, that is all pairs of
values. By contrast, the range measure of dispersion considers only one pair of
highest value and lowest value. When xi and xj denote ith and jth incomes
respectively and i, j=1, 2,…, N, we can see that the aggregate of absolute
differences is given by
N N
∑∑ x
i =1 j =1
i− xj (G.1)
and because total number of differences is N2, the mean of absolute differences
can obviously be written as
N N
1
N2
∑ ∑x
i =1 i =1
i− xj (G.2)
67
Quantitative Methods-I where the differences with the self have also been counted and the difference of M
xi with xj is treated as separate from that of xj and xi though numerically they
are the same. This is also said to be the case with replacement. Expression
(G.2) ranges between 0 and 2µ.
In the case of without replacement, the sum is obviously to be divided by

N(N-1) as there are N deviations with the self. It is not difficult to see that the
numerical value of the sum remains the same.
In order to make it serve as a measure of inequality, (G.2) can be divided by µ

to produce what can be called coefficient of mean difference (CMD):
1 N N
CMD = ∑
N 2 µ i =1
∑x
i =1
i − xj (G.3)
CMD fulfils the idea of scale independence. However, the expression (G.3)
ranges between 0 when everybody has the same income and 2[=2N/(N-1)]
when only one person has all the income. In order to make it satisfy the interval
(0,1), we can further divide it by 2. The result is Gini coefficient of
concentration or Gini index of inequality:
N N
1
G= ∑
2 N 2 µ i =1
∑x j =1
i − xj (G.4)
or
1 N N
G= ∑
N 2 µ i =1
∑
x j ≤ xi
( xi − x j ) (G.5)
conceived as an aggregate of only positive differences, though normalized by

the number of all differences and the mean income. Kendall and Stuart define
this as ‘one half of the average value of absolute differences between all pairs
of incomes divided by the mean income’.
The index can also be defined in terms of population proportions and income
shares. If the income-share of individual i is denoted by qi, that is,
xi
qi = , (G.6)
Nµ
then the expression (G.4) can also be written as

N N
1
G=
2N 2
∑ ∑q
i =1 j =1
i −qj (G.7)
In the case of a discrete distribution, each individual constitutes (1/N)th of the

population, that is,
1
pi = . (G.8)
N
Therefore one can also write (G.7) as

68
I N N Measures of Inequality
1
G=
N
∑∑
i =1 j =1
p j qi − pi q j . (G.9)
An obvious question is: why not pi qi − p j q j in the expression (G.9)? For

understanding this, let us consider the Gini coefficient for the groups.
Let µ r and µ s denote mean incomes of rth and sth groups (say, families)
respectively and r,s=1,2…g. Then, Gini for the groups can be defined as
g Ng
1
G= ∑
2 N 2 µ r =1
∑
s =1
µr − µ s f r f s (G.10)
where fr and fs are frequencies of the groups r and s respectively. This can
obviously be written as
1 g g
µr µ s
G= ∑
2 r =1
∑
s =1
−
µ µ
pr p s (G.12)
where
fr f
pr = and p s = s (G.12)
N N
Now, let us note the share of total income with the group r:
µr fr µ
qr = = pr r (G.13)
Nµ µ
Then, the expression (G.12) can be written in either of the two ways (G.14) and
(G.15)
1 g g
qr q
G= ∑
2 r =1
∑
s =1
− s
p r ps
(G.14)
or
1 g g
G= ∑
2 r =1
∑
s =1
ps qr − pr q s (G.15)
It is easy to see that g=N and ps=pr=(1/N) when fr and fs are all equal to1.
In statistics literature we emphasize frequency aspects, in economics literature

we find it convenient, expression-wise, to treat each individual with single
income though there is no bar for xi=xj.
11.3.2 Simple Computational Device

Two years after giving his index to terms of relative mean differences, Gini
(1914) showed that the index is exactly equal to one minus twice the area under
Lorenz curve (to be discussed later). That is,
G = 1− 2A (LG.1)
where A is the area under the Lorenz curve, as shown in Fig. 11.1
69
Quantitative Methods-I (0,1) (1,1) M
B
--
---
----
-----
---------
------------
L
----------------
-------------------
------------------------
O ------------------------------------- A
(0,0) (1,0)
Fig. 11.1
However, normally, we do not specify and estimate a smooth relationship

between Q and P. Instead, we obtain the curve by plotting cumulative
proportions of people in classes and cumulative shares of their incomes, where
classes are arranged according to increasing per capita income values:
µ 1 ≤ µ 2 ≤ µ 3 ≤ .... ≤ µ r ≤ ... ≤ µ g . (LG.4)
strictly speaking, equality sign is useless in this presentation.(We have written
it in deference to Theil (1967). Plotting Pr and Qr, we obtain the Fig. 11.2. We
can see that the area below the Lorenz curve can be conceived as consisting of
several trapeziums. A trapezium could be seen as consisting of a rectangle and
a triangle. Summing the areas of all trapeziums (say g in number), we can get
the area A . Substituting it in (LG.1), we get the following expression for
computing Gini coefficient G:
g
G = 1 − ∑ ( Pr − Pr −1 )(Qr + Qr −1 ) (LG.5)
r =1
Qr
Tr
Pr
Fig. 11.2
70
I Check Your Progress 2 Measures of Inequality
1) Define Gini ratio.

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Show the difference in approach in defining Gini from other measures of
dispersion.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Write the expression for computing Gini.

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) How can you compute Gini ratio?

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
11.4 LORENZ CURVE
Lorenz curve is a powerful geometrical device to compare two situations of

distribution with regard to their level of inequality. Devised some hundred
years ago by Max O. Lorenz (1905) to measure concentration of wealth, it is
still very widely used in empirical studies on inequality. The device can be
used for comparing inequality of distribution of any measurable entity such as
71
Quantitative Methods-I income, wealth (land, capital), consumption, expenditure on an item (say, food M
or education), etc. The distribution may be over persons or households. But the
device can also be used to measure inequality of tax collection or expenditure
incurred by states or federal grants received by different states. We can
compare pre-tax and post-tax distributions in order to study the efficacy of
instrument of tax.
Lorenz (1905) studied a number of methods then in use to gauge the level of,
or change in the level of, inequality. Most of these measures used fixed-income
classes in data tabulation and made inter-temporal comparison, employed
changes in percentage of recipients of class incomes in each of the fixed-
income classes or movement of persons from one class to another and so on.
Finding them unsatisfactory, he comes to the conclusion that changes in
income and changes in population both have to be simultaneously taken into
account and in a manner that ‘fixed-ness’ of income classes gets neutralized.
In fact, this measurement relates to comparison and in most cases, we are in a

position to compare but there are situations of non-comparability. However, a
few of inequality measures that are capable of numerical representation in
terms of a scalar number, and therefore called summary measure, are found to
be based on the Lorenz curve.
It may be pointed out that the curve was independently introduced by Gini
(1914). It is therefore, quite often referred to as Lorenz-Gini curve as well. We
shall, however, stick to more common usage and call it Lorenz Curve.
11.4.1 Geometrical Definition

The Lorenz curve of concentration of incomes is the relationship between the
cumulative proportions of recipients, usually plotted on the abscissa, and the
corresponding cumulative shares of total income with the recipients, usually
plotted on the ordinate. If population proportions and income shares of class j
are denoted by pj and qj and cumulative proportions and shares upto class i, by
Pi and Qi then
i
Pi = ∑ p j , 1≥ pj ≥ 0 (GD.1)
j =1
and
i
Qi = ∑ q j , 1≥ qj ≥ 0 (GD.2)
j =1
The relationship between Pi and Qi is given by the curve
Qi = L( Pi ), 1 ≥ Pi ≥ 0, 1 ≥ Qi ≥ 0 (GD.3)
and the point on the curve by (Pi , Qi ). Naturally, the first point is (0,0) and the
last one on the curve, (1,1). It is also clear that Qi ≤ Pi i=1, 2,…, N-1 if there
are N classes of incomes. It means no point will make an angel of more than
45˚ with the abscissa at the origin. Then, one can be sure that the Lorenz curve
lies in the lower triangle of Lorenz Box of the unit square. See Fig. 11.3 in
which OLB shows the Lorenz curve (Fig. 11.3).
72
I (0,1) (1,1) Measures of Inequality
B
Qr
O A
(0,0) Pr (1,0)
Fig. 11.3
11.4.2 Properties of the Lorenz Curve

Now it is easy to see that the extreme case of perfect equality is given by the
diagonal OB which represents Pi=Qi, i=1, 2,…, N. The other extreme of perfect
inequality will be given by a curve OAB. The diagonal OB is often designate
as the egalitarian line or line of equality. The other diagonal CA is known as
the alternative diagonal and is useful to study the symmetry of the curve. The
tringle OAB with sharp kink of 90˚ at A can be said to be the line of perfect
inequality. See Fig. 11.4.
Fig. 11.4
We can finally note the following properties:

i) 1 ≥ pi ≥ 0; 1 ≥ q i ≥ 0, i = 1,2,..., N
ii) 1 ≥ Pi ≥ 0; 1 ≥ Qi ≥ 0, i = 1,2,..., N − 1
iii) P0 = Q0 = 0; PN = Q N = 1
iv) Pi ≥ Q i , i = 1, 2 ,.., N − 1
73
Quantitative Methods-I By drawing a Lorenz Curve, we can know whether a given distribution is equal M
or unequal. We do not yet know how much unequal a given distribution is.
When we draw two or more Lorenz Curves, we can compare the distributions
as regards their levels of inequality. The curve closer to the diagonal of
equality has lower level of inequality than the one away from it (Fig. 11.5). But
we do not know yet the level of inequality. And even this comparison is
possible only when the curves do not intersect (Fig. 11.6).
Fig. 11.5 Fig. 11.6
However we can devise some measures, which are based on the Lorenz curve.
In case the Lorenz curves intersect, reducing the distributions into single real
number is the only option. So we shall discuss only two such proposals.
11.4.3 A Measure Based on Area

We have noted that if Lorenz curve coincides with the diagonal of equality, the
inequality is nil and if Lorenz curve coincides with the two sides of the square,
the inequality is full.
In the case of non-intersecting Lorenz curves, it is clear that the curve closer to
the diagonal of equality will circumscribe smaller area between itself and the
diagonal of equality than the one, which is farther. Which is what it should be.
We can therefore devise a measure of inequality by dividing the area OLB
(Fig. 11.4) by the area of triangle OAB, which is the maximum possible area
between the diagonal of equality and Lorenz curve. As the area of OAB is
(1/2), the measure turns out to be twice the area between the diagonal of
equality and the Lorenz curve. In other words, Lorenz coefficient of
concentration (LCC) is:
Area OLB
LCC = = 2 Area OLB
∆OAB
Since this turns out to be exactly equal to Gini coefficient, we are not
elaborating it any further.
11.4.4 A Measure Based on Length

This is a measure proposed by Kakwani (1980). The length of the Lorenz
curve, denoted by ℓ, cannot fall below √2, which is the length of the egalitarian
line and cannot exceed 2, which is the sum of the lengths of the two arms of the
lower triangle. In order to produce a measure with the minimum value 0 and
the maximum value 1, following exercise can be suggested:
74
Minimum Actual Maximum
Length of the Curve √2 ℓ 2
Length of the Curve - √2 0 ℓ-√2 2-√2
Length of the Curve -√2 0 ℓ-√2 1

Maximum length -√2 2-√2
So this measure is clearly:
LK= (ℓ-√2)/ (2-√2)
In both the cases, one can draw actual graphs and actually measure the area and
the length and calculate the indices for level of inequality. Those who wish to
carry out a more sophisticated exercise will have to estimate smooth functions.

1) Enumerate the properties of Lorenz curve.
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
2) When will comparison between two Lorenz curve fail to compare
inequality in two distribution?
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
3) What is the relationship between Lorenz curve and Gini coefficient.
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
4) What is Kakwani’s measure of inequality, which is based on the Lorenz
curve.
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
…………………………………………………………………………….
75
11.5 NORMATIVE MEASURES
The measures that essentially involve judgement about values through
specification of social welfare function are called normative measures. The
arguments of this nature were first advanced by Dalton, pretty eight decades
ago in 1920 for constructing what are today called normative measures of
inequality.
Reacting to an observation by Pearsons (1909) that ‘the statistical problem

before the economists in determining upon a measure of inequality in the
distribution of wealth is identical with that of the biologist in determining upon
a measure of the inequality in the distribution of any physical characteristic’,
Dalton (1920) pointed out that ‘economist is interested, not in distribution as
such, but in effects of the distribution upon the distribution (and total amount)
of economic welfare which may be derived from income’. The objection to
great inequality of income, he further points out, is due to the resulting loss of
potential economic welfare that could accrue to people in the absence of it.
Yet, it has to be noted that inequality though defined in terms of economic

welfare, has to be measured in terms of income. This idea due to Dalton has
been conceded by subsequent contributions. Using the notion of social welfare
function in construction gives rise to normative measures of inequality.
It may be instructive to remember that the discussion would revolve around

three issues:
1) the relationship between income of a person and his welfare;

2) the relationship between personal income-welfare functions; and
3) the relationship between personal welfare and social welfare.
It may be noted that utility is the word mostly used for personal welfare
whereas for welfare of society the phrase social utility is rarely used.
There are two major indices in this category: Dalton’s index and Atkinson’s
index. In Atkinson’s index a new idea is introduced and that is of equally
distributed equivalent income. Actually there are two sub-approaches within
normative approach. One is Dalton’s and the other is Atkinson’s. While in
Dalton’s approach present social welfare is compared with that could be
obtained by equally distributing the total income, in Atkinson’s approach the
present level of income is compared with that of equally distributed level of
income, which generates the present level of social welfare. Sen has
generalised the Atkinson’s index. Theil’s index based on information theory
could be suggested here only to sort of complete the unit.
11.5.1 Dalton Index

For each individual, Dalton assumes, marginal economic welfare diminishes as
income increases. It means income-welfare function
U i = U i ( xi ), i = 1,2,...N (D.1)
(where Ui is welfare of person i possessing income xi)

76
I is concave, suggesting that (∂ U i /xi) >0 but (∂2 U i /xi2) < 0. Dalton further Measures of Inequality
assumes that economic welfare of different persons is additive. Thus, in his

scheme, social welfare is a simple aggregation of personal welfares. In other
words, social welfare W is given by
N
W = ∑ U i ( xi ) (D.2)
i =1
He further assumes that the relation of income to economic welfare is the same
for all members of the community. That is,
U i = U ( xi ) , i=1,2,…,N (D.3)
In that case, the relation (D.2) can be expressed as
N
W = ∑ U ( xi ) (D.4)
i =1
which makes it clear that whosoever gains in welfare, the addition to the social
welfare is the same. For any given level of social welfare, any distribution of
welfare among the members of the society is permissible. However, one must
remember that the relation of individual income to their welfares is concave.
Therefore, transfer of income from A to B will not lead to symmetric change in
welfares of those two persons involved in the transaction. The result is some
impact on W the measure of social welfare.
From Fig. 11.7, we may compare the situation when two individuals, both
possessing the same relation, have two different income levels, with that when
they have the same (mean) income. We may note that the sum of the welfare
of person 1(BB’) and the welfare of person 2(DD’) is less than the twice of
CC’ which is the level of welfare enjoyed by both the persons when they have
equal income. It is easy to see that the loss suffered by person 2, that is D’E, is
overcompensated by the gained by person 1, which is C’F.
D’
C’ E
B’
F
B C D
X1 µ X2
Fig. 11.7
77
Quantitative Methods-I This demonstrates that, under assumptions by Dalton, an equal distribution is M
preferable to an unequal one for a given amount of total income from the
viewpoint of social welfare. In fact, for a given total of income, the economic
welfare of the society will be maximum when all incomes are equal. The
inequality of any given distribution may therefore be defined as
∑U (µ )
i =1 NU ( µ )
D1 = N
= N
(D.7)
∑ U ( x ) ∑U ( x )
i =1
i
i =1
i
which is equal to unity for an equal distribution and greater than unity for an
unequal one. It may therefore be preferred to define the Dalton’s index as
NU ( µ )
D2 = N
(D.8)
∑U ( x )
i =1
i
which is obviously zero for an equal distribution. How large can it be? It will
depend on the values of U(0), U(µ) and U(Nµ) when N and µ are given, not
necessarily 1. Later writers have therefore preferred to define Dalton’s index in
the following form, which inverts the arguments of D2 subtract it from 1:
∑U ( x )
i =1
i
U
D =1− = 1− (D.9)
NU ( µ ) U (µ )
It looks as if the index is contained in the interval (0,1). However, there are
many valid concave functions where it may not hold true. For example, if we
have U ( xi ) =log xi , then D=1– {log µˆ / log µ }. Given the fact that µ̂ < µ, D
would turn out to be a negative number for µ<1. And µ could be less than 1 as
x can be measured in any unit. It would be the same case U ( xi ) =1/ xi .
However, in order to obtain numerical magnitude, it is not sufficient to define

the index. Dalton (1920) points out that though defined in terms of economic
welfare, inequality has to be measured in terms of income. Then, no unique
measure of inequality will emerge. It will verily depend on the particular
functional relationship assumed. Dalton himself considered two such functions
for the purpose of illustration. The first is related to Bernaulli’s hypothesis. It
holds that proportionate additions to income (in excess of that required for bare
subsistence-poverty line) make equal additions to personal welfare, That is,
dx i
dU i = orU i = log x i + ci (D.10)
xi
Under the assumption that every person has the same functional relationship,
the Dalton’s index can be given as
log µˆ + c
D =1− (D.11)
log µ + c
where µ̂ is the geometric mean of personal incomes. The other formulation he
discusses is given as
78
I dx i 1 Measures of Inequality
dU i = 2
orU i = c − (D.12)
x1 xi
where c is the maximum welfare one can obtain when x → ∞ . Dalton’s index
in this case would turn out to be:
c − (1 / µ~ )
D =1− (D.13)
C − (1 / µ )
where µ~ is the harmonic mean.
11.5.2 Atkinson Index

Atkinson (1970) objects to Dalton’s measure because D is not invariant with
respect to positive linear transformations of personal income-welfare functions.
This was pointed out by Dalton himself but he could not resolve it.
Atkinson seeks to redefine the index in such a way that measurement would be
invariant with respect to permitted transformations of welfare numbers.
Atkinson does it through devising what he calls 'equally distributed equivalent
income'. Both the distributions, the original and the new one, are supposed to
yield the same level of welfare.
In order to make the concept clear, we put a few artifacts along with the actual
distribution. Let us first note that for an actually distributed income vector xi,
i=1,2…,N (call it vector a), there is only one equally distributed income vector
with each element equal to µ (call it vector b) but there are a number of
equivalently distributed, vectors (call them vectors c). See Chart 1. An
equivalent income distribution is one, which has the same level of welfare as
that of currently given distribution. However, one of these equivalent
distributions (vectors c) is `equal’ as well. This is called equally distributed
equivalent income vector, shown as vector (d) in the Chart 1. As µ is the mean
level of current distribution, µ* may be used for designating the level of
equally distributed equivalent income.
CHART-I
Vector (a) Actually distributed income vector x1, x2 …., xi, …., xN.
Vector (b) Equally distributed income vector µ, µ, …., µ, …, µ.
Vector (c) Equivalently distributed income vector x1* , x 2* , …, x1* , …, x *N .
Vector (d) Equally distributed equivalent income vector µ*, µ*, …, µ*, …, µ*
If should be obvious that Wb ≥ Wa = Wc and Wc = Wd . Then, Wa = Wd . W

represents social welfare with respective distributions of income vectors. It is
clear that µ ≥ µ *. µ * is defined by the additive social welfare function having
symmetric individual utility functions such as:
N
1
U (µ ) =
N
∑U ( x )
i =1
i (A.1)
or equivalently
1 N

µ * = U −1 
N
∑U ( x )
i =1
i (A.2)
79
Quantitative Methods-I The index due to Atkinson is then defined as the additive inverse of the ratio of M
equivalent mean income to actual mean income:
µ*
A =1− (A.3)
µ
which is said to lie between zero (complete equality) and 1 (complete
inequality). We can see that A cannot be 1 unless µ* is zero, which is an
impossibility for any distribution with µ>0. If complete inequality is defined as
the situation when only one person grabs all the income, we can see that
µ m*
A = 1− (A.4)
µ
where
N
NU (µ *) = ∑ U ( xi )
i =1
and
NU (µ * m ) = ( N − 1)U (0) + U ( Nµ ).
This index is not scale independent unless some restriction is imposed on the
relationship U. If this requirement has to be met, Atkinson points out, we may
have to have the following form
 β 1−ε
α + xi , ε ≠ 1
U ( xi ) =  1−ε (A.5)
log e xi , ε =1
Note that we need є ≥0 for ensuring concavity and є >0 for ensuring strict
concavity. This is a homothetic function and is linear when є=0. We may note
that є cannot exceed 1 as in that case the varying component assumes inverse
relationship. α is usually negative so that U(xi) is not positive for xi = 0.
Otherwise, when xi = 0, Ui = α which means that welfare is positive even
when income is zero. This is generally not acceptable. On the contrary, a
negative α would be more acceptable. When є =1, α is infinitely large and
negative.
Since є can be zero, Atkinson’s requirement is not strict concavity. Sen (1973)
has a question. He asks to consider two distributions (0,10) and (5,5) along
with
U ( xi ) = α + β ( xi ) (A.6)
Then, he points out the level of social welfare would be ( 2α + 10 β ) whatever

the distribution. µ* would be 5 in both the cases. µ is of course 5. The measure
of inequality A is therefore zero. So, both the distributions are ethically equal.
This is obviously absurd. Therefore, the relation (A.5) should be defined with
the restriction є>0. We should also note that (A.4) is an iso-elastic marginal
utility function.
80
I є is the inequality-aversion parameter and has close resemblance with risk- Measures of Inequality
aversion premium. Atkinson proposed to draw on the parallel formally with the
problem of measuring risk. He finds his concept of equally distributed
equivalent income very closely resembles with risk-premium or certainty
equivalent income as used in the theory of decision-making under uncertainty.
In case we seek to introduce this restrictive personal income-welfare function

along with simple aggregation of individual welfares to constitute the social
welfare, into the inequality measure A, we will have
1− ε 1 /(1−ε )
1 N
 xi  
A = 1−  ∑ 
i =1  µ
  ,ε ≠ 1 (A.7)
 N  
The question is now narrowed down to choosing є. As є rises, more weight is

attached to transfers at the lower end of the distribution and less to that at the
top. When є rises, (A.7) assumes the function mini (xi), which only takes
account of transfers to the very lowest income group (and is therefore not
strictly concave). When є = 0,Ui is linear. As a consequence, A is always zero.
This means A has no descriptive content at all. When є→1, A turns out to be
1/ N
N
 xi  µˆ
A = 1− ∏   =1− (A.8)
i =1 µ  µ
which is the same as Champernowne index (CII.1). For values of є between 0

and 1, the expressions may not be very neat. Parameter є is often chosen to be
½ or 1/3 or 2/3.
11.5.3 Sen Index

There are people who feel rather strongly that the social valuation of the
welfare of individuals should depend crucially on the incomes of their
neighbours too. Then, why should society add simply individual welfares?
One may also question the assumption of one welfare function for all
individuals. If we do so, we should go for broad social welfare function such
as
W = W ( xi , x 2 ,..., x N ) (S.1)
which is just symmetric, quasi-concave and increasing in individual income
levels. Then, a more general normative measure of inequality can be defined
by devising the concept of ‘generalized equally distributed equivalent income’.
This is obviously the level of per capita income x* which, if shared by all,
would produce the same level of W as is generated by the present actual
distribution. That is,
x* = x W ( x*, x*,.., x*) = W ( x1 x 2 ,..., x N ) (S.2)
Under the assumption that (S.1) is quasi-concave, x* ≤ µ for every distribution
of income. The index S would then be
x*
S = 1− (S.3)
µ
81
Quantitative Methods-I which is but a generalized version of A. If utilitarian framework is employed, M
then S and A turn out to be indistinguishable.
These measures, it may be noted, clearly suggest that there exists a

redistribution equivalent of growth so far as the concern is about raising the
welfare.
11.5.4 Theil Entropy Index

Theil (1967) poses a question: Does information theory supply us with a
‘natural’ measure of income inequality among N individuals, which is based on
income shares? He answers: Yes. Here is a short introduction.
Let us start with income share of individual i:
N
xi
qi = > o such that ∑ qi = 1 (T.1)
Nµ i =1
When xi = µ, i=1, 2,…, N, that is, when distribution is equal, we have

1
qi = i = 1, 2,...N (T.2)
N
We have complete inequality when some xi=Nµ and x j = 0, j ≠ i. It implies
that qi=1 for some i and q j = 0, i ≠ j .
In information theory, one way of defining entropy of probabilities pi is

N
1
H = ∑ p i. log (T.2)
i =1 Pi .
Replacing probabilities by shares, we have

N
1
H = ∑ qi. log (T.3)
i =1 qi .
which can be taken as a measure of equality. For the situation of complete

equality, we can see that H is equal to logN and for that of complete inequality
H is zero. We can therefore define Theil index T as
N
1
T = log N ∑ qi . log
i =1 qi .
N N
1
= ∑ qi log N − ∑ qi . log
i =1 i =1 qi
N
= ∑ qi log N .qi (T.4)
i =1
This measure is motivated by the notion of entropy in information theory. But

one can see that it can be interpreted in the traditional normative framework
with
82
I 1 Measures of Inequality
U i = qi log (T.5)
qi
and
N
W= ∑ U i (q i ). (T.6)
i =1
We may note that (T.5) depends on xi as well as on µ along with N and U that it
is concave with respect to xi.
While the lower limit of T is zero, its upper limit log N increases as the number
of individuals increases. To many people, it is objectionable. However Theil
(1967) chooses to defend it. When society consists of two crore persons and
one grabs all and when society consists of two persons and one grabs all,
cannot have the same level of inequality. The former case is equivalent to the
situation in which one crore out of two crore people have nothing and the other
one crore have equal income. Maximum value for two-person society is log 2,
and that for two crore-person society is 7 log 2. Some researchers still insist
that the measure should be normalized by dividing it by log N.

1) What is social welfare function, according to Dalton?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Discuss Dalton index of inequality.

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Give the logic behind Atkinson index.

………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
83
Quantitative Methods-I 4) How is Sen index distinct from Atkinson’s index? M
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
5) Discuss Theil’s entropy index of inequality.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
11.6 LET US SUM UP

Owing to adverse impacts of economic inequality both on poverty and on
growth, reducing inequality has been a priority of public policy. Various
measures of income inequality can be put under two categories: positive
measures and normative measures. The positive measures capture the
inequality of income without value judgements. These include range quartile
range, standard deviation, Gini ratio, etc. Lorenz curve belongs to this
category. It measures inequality to the extent of comparing two distributions.
The measures, which essentially involve value judgements about social
welfare, are normative measures. These include indices propounded by Dalton,
Atkinson, Sen, and Theil.
Easy comprehension and easy computation, range of variation and amount of

information needed the desirable properties of the measures of economic
inequality. In order to judge the efficacy of an inequality index, several axiom
have been set up. However, these axioms have been relegated to the Appendix.
11.7 KEY WORDS

Co-efficient of Mean : Mean of all pair-wise differences divided by the
Difference mean of differences has been termed as
coefficient of mean difference in this text.
Dispersion : The fact that values of a variable are not all the
same is known as dispersion. The spread or
scattering of the distribution is measured by a
measure of dispersion.
84
I Extreme Disparity : The ratio of the highest value to the lowest Measures of Inequality
Ratio value is known as extreme disparity ratio.
Normative Measures of : Measures of inequality, which are articulated
Inequality through the explicit incorporation of social
welfare function or social welfare
considerations, are known as the normative
measures of inequality.
Positive Measures of : Measures of inequality, which are based in
Inequality statistical properties of distribution, are known
as the positive measures of inequality.
Relative Standard : Standard deviation of a distribution divided by
Deviation its mean is known as Relative Standard
Deviation.
Standard Logarithmic : Standard deviation of logarithms of values in a
Deviation distribution is known as Standard Logarithmic
Deviation. Though logically the deviations of
logarithmic values should be taken from the
logarithm of geometric mean but at times they
are taken from logarithm of arithmetic mean.
Therefore, there are two versions.
Social Welfare : An index of social well-being, often articulated
Function as a function of individual utilities or individual
incomes or individual consumption baskets,
with or without labour disposition, or individual
rankings of potential state of affairs.
11.8 EXCERCISES
1) Following the adapted distribution of monthly per capita expenditure (in

Rs.) in rural India in the 60th round of the NSS over January to June 2004:
Class 50- 225- 255- 300- 340- 380- 420- 470- 525- 615- 755- 950-
225 255 300 340 380 420 470 525 615 755 950 1200
Avg. Exp. 100 240 280 325 365 405 450 500 580 700 850 1100
Percentage 2.4 2.7 6.4 8.3 9.6 9.6 10.8 10.0 12.2 12.6 6.7 8.8
of persons
Calculate as many positive measures as you can.

2) Following is distribution data of operational holdings from agriculture
census 1976-77:
Holding Marginal Small Semi- Medium Large All
medium
Definition Unit 0.0-1.0 1.0-2.0 2.0-4.0 Ha 4.0-10.0 Above
Ha Ha Ha 10.0 Ha
Number ’000 44523 14728 11666 8212 2440 81569
Area ’000 Ha 17509 20905 32428 49628 42673 163343
Draw Lorenz curve and compute Gini ratio.

85
Chaubey, P. K. (2004), Inequality: Issues and Indices, Kanishka Publishers,
Distributors, Delhi.
Cowell, Frank A. (1995), Measuring Inequality, Prentice Hall/Harvester

Wheatsheaf, London;
Sen, A.K. (1997) On Economic Inequality, Oxford University Press, Oxford.

(with Annexe by James E. Foster.

PROGRESS EXERCISES
1) Range is the difference between the maximum and minimum value. With
a view to ensuring that the index of inequality based on this measure of
dispersion is unit-free and/or is confined in the interval of (0,1), various
ways of normalizations could be considered. See Sub-section 11.2.1.
2) In the relative range only extreme values are considered, which may not
be representatives of the distribution. It is like comparing the poorest
(who may be a few) with the richest (who may be one). Inter-quartile
measures take into account the middlemost distribution with 50 percent
recipients. See Sub-section 11.2.2
3) In the mean deviation, all values in a way are considered. The mean of
absolute deviations is divided by the mean of the distribution to yield
relative mean deviation. See Sub-section 11.2.3. Since the sum of
deviations on one side of the mean will not change with the transfer of
income contemplated, the magnitude of the index would not change.
4) See Sub-section 11.2.4. In one version the deviations are taken with
respect to logarithm of geometric mean and in the other with respect to
that of arithmetic mean. With some mathematical manipulation, one can
find out that former is smaller than the latter by square of the difference
between the logarithms of geometric mean and arithmetic mean.
5) It is a straight application of the fact that geometric mean of a distribution
is smaller than its arithmetic mean. Of course, when the values are greater
than 1!
6) Hirfindahl index is sum of squares of the shares with each recipient,
which of course varies with the number of recipients. Equal distribution
of shares between two recipients will yield a value of 0.5 and between
three recipients, 0.333. It is therefore more used as a measure of
concentration. See Sub-section 11.2.5.
7) The message is that one can try on one’s own to devise new methods. See
Sub-section 11.2.6. For second part, the answer is 5/9. Try it out.
1) Gini ratio is one half of the average value of absolute differences between
all pairs, including with self, of values divided by the arithmetic mean.
86
I See the definition by Kendall and Stuart in Sub-section 11.3.1. Have a Measures of Inequality
look at expression (g.4) and (G.7).
2) The basic difference lies Gini index from other measures based on
dispersion that in articulating all differences are considered while in others
either few differences are considered or deviations from arithmetic/
geometric means are considered.
4) By writing out in a table, columns for class intervals or values,

frequencies, class total values, cumulative frequencies, cumulative total
values, cumulative proportions and cumulative shares in respect of each
class. For using expression (LG.5), consecutive moving differences (or
sums) of proportions and consecutive moving sums (or differences) need
to be computed in two additional columns. MS-Excel will do well.
1) Look at the Fig. 11.4 and Sub-section 11.4.2. Try writing the properties in
language.
2) When the two Lorenz curve will intersect, it will not be possible to say on
balance which distribution is more unequal. In fact, one section in that
case will be more unequal and the other section less unequal in
distribution 1 in comparison to their counterparts in the distribution 2.
3) The value of Gini coefficient is equal to twice the areas inscribed between
line of equality and the Lorenz curve. It is equivalent to saying that G is
equal to (1-A) where A is the areas below Lorenz curve in the unit box.
4) Kakwani’s measure of inequality is normalized length of the Lorenz curve.
1) Dalton’s social welfare function is a simple aggregation of welfare (utility)

functions of the individuals constituting the society. In addition, all
individuals are supposed to have the same income utility function.
2) Sub-section 11.5.1. Write Dalton’s proposal and its modern version. Also
point out that though conceived in terms of utility, Dalton held that the
index has to be measured in terms of income only.
3) An inequality measure should not change with linear transformation of

personal utility function. Since Dalton’s index does not respect this
property, Atkinson is not happy. He therefore devises a new artifact called
‘equally distributed equivalent income’ and suggests a new utility function
called iso-elastic marginal utility function.
4) Sen index is different from the Atkinson’s index in one respect that he opts
for a social welfare function, which has as individual incomes its arguments
and is symmetric and quasi-concave. See Sub-section 11.5.3.
87
Appendix
AXIOMS OF INEQUALITY MEASURES
For any statistical measure, some of the desirable properties that are described
in standard textbooks are (i) simplicity of comprehension, (ii) ease of
computation, (iii) range of variation, and (iv) amount of information needed.
However, we discuss below only those properties, which are peculiar to the
measures of economic inequality.
We are often faced with situations where we have to compare two distributions
with the help of an index with regard to their level of inequality. The two
distributions may belong to two different countries at a point of time, to a
country at two points of time, or to two situations—say, one before tax and the
other after tax or before and after interpersonal transfers etc.
People have set up some intuitively appealing properties in order to judge the
efficacy of an inequality index. The first set of properties was given as
‘principles’ by Dalton (1920). Today, in literature, they are known as axioms.
We propose to discuss some common axioms. It may be pointed out at the
outset that these axioms almost ignore the question whether inequality is an
issue, which matters more (or less) in an affluent society or in a poor one. The
whole discussion will assume that all incomes are positive though we know for
sure in case of business failure or crop failure, incomes can well be negative or
zero.
1) Axiom of Scale Independence

If there are two distributions of equal size such (N) that each element of one
distribution is a multiple θ of the corresponding element of the other
distribution, i.e.,
x12 = θ x11 i=1,2,…,N,
then the numerical magnitudes of inequalities of both the distributions should

be the same, i.e.,
I ( x11 , x12 ,..., x 1N ) = I ( x12 , x 22 ,..., x N2 )
where the inequality measure I is shown as a function of the distribution
( x1 , x2 , ...., x N ).
Obviously, it also satisfies the idea that the level of inequality should not
change when the scale of measurement changes, say, from rupees to paise or
bushels to quintals.
It does also mean that equal proportionate additions to all incomes would not
change the level of inequality for
x i (1 + λ ) = θx i i =1,2,….,N,
The proportionate addition could even be negative. Thus, it is a question of

88 shares in the cake, not the size of the cake. It is very obvious that an inequality
I measure is defined in terms of shares si because a proportionate change in all Measures of Inequality
incomes leaves the shares unchanged.
However, this axiom goes against Dalton’s principle of proportionate additions

to income, which stated that equal proportionate additions (subtractions) should
diminish (increase) the level of inequality. Perhaps, Dalton could not see that,
equal proportionate addition is theoretically, equivalent to change in the scale
of measurement. It should so happen in the case of a measure of relative
dispersion is obvious enough.
The axiom covers the cases of proportionate taxation/subsidies. It may be

noted that such additions do not change individual (class) shares of total
income and, therefore, the Lorenz curve remains unchanged. All measures
based on the Lorenz curve shall therefore satisfy this axiom.
It should not mean that change in the size of cake is immaterial. In the social
welfare, size of cake and distribution of cake both matter. It is only in a limited
context of measurement of inequality that this property is considered desirable.
Lorenz (1997) mentions an objection raised against this axiom in terms of non-
proportionate increase in well-being of different income holders, which means
diffusion of well-being when incomes increase but concentration when
incomes decrease. Thus, this idea existed much before Dalton mentioned it.
One could easily see that this objection incorporates the idea of diminishing
marginal utility. The true province of the axiom then is the unit of
measurement.
2) Axiom of Population Size Independence

The level of inequality remains unaffected if a proportionate number of persons
is added to each income level.
This suggests that the magnitude of inequality in the distribution of the cake
should depend on the relative number of receivers with different levels of
income. If we merge two economies of identical distributions of the same size
N, then, in the consequent economy of size 2N, there shall be the same
proportion of the merged population for any given income. Such replications
will leave the inequality level unchanged. The axiom is also known as the
Principle of Population Replication.
This exactly corresponds to Dalton’s principle of proportionate additions of

persons. Since the Lorenz curve remains unchanged so long as proportions of
people in each class remains the same, the measures based on the Lorenz curve
would satisfy this axiom.
Let us have however a counter-intuitive example. Let us a have two-person

world in which one person is having no income and the other having is having
all. Let us replicate the economy. Now there is a four-person world in which
two are sharing destitution with zero income but the other two are sharing
positive income equally. Earlier there was no equality; now each 50 per cent of
population is sharing the income equally. So, some scholars do not accept it.
89
Quantitative Methods-I 3) Axiom of Equal Income Addition M
If the distribution 2 xi2 ,i=1,2,….,N is obtained by addition of equal amount d

(say, through pension) to each element of distribution 1 xi1 , i= 1,2,…N, i.e.,
xi2 = xi1 + d
then inequality level of distribution 2 should be lower than that of distribution 1

i.e.,
I ( x12 , x 22 ,..., x N2 ) < I ( x11 , x12 ,....x 1N )
naturally, subtractions (say, taxation) of equal amount from each income would
reverse the inequality sign. It can be noted that in the former situation, the
shares of the poorer persons increase and in the latter, they decrease. This
axiom exactly corresponds to Dalton’s principle of equal additions to incomes.
Now, we propose to discuss two very important axioms relating to transfer of

an income from a person to another when other things remain the same. The
former may be called Pigou-Dalton condition and the latter, Sen condition.
4) First Axiom of Income Transfer (Pigou-Dalton Condition)

If an equal transfer from a richer person to a poorer person takes place, then the
level of inequality is strictly diminished, provided that the equalizing transfer
amount is not more than the difference between two incomes involved. Any
number of such transfers taking place between any two consecutive income
units will not cause any change in the ranking of income units and therefore
such a process of transfers may be called the rank-preserving equalization.
This axiom requires an inequality measure to be sensitive to transfers at all

levels of income and, thus, at least a function of all incomes.
This axiom corresponds to Dalton’s principle of income transfer. Dalton

(1920) argued that an inequality measure must have this minimal property.
Since in this context Pigou’s contribution (1912) is found significant, Sen
(1973) designated this axiom as Pigou-Dalton condition. Following him, a
number of contributors in the field have given it the name of ‘P-D condition’.
Most of the indices, barring relative range and relative mean deviation, pass
this test. This axiom is also known as weak transfer axiom because it suggests
the direction but not the magnitude of change in the level of inequality.
5) Second Axiom of Income Transfer (Sen Condition)

If we consider two transfers, one at a time, at different points of scale, then the
transfer at lower end of scale should have greater impact than its counterpart at
higher end of the scale. According to Sen, (1973), the impact on the index
should be greater if the transfer takes place from a person with an income level
of, say Rs.1000 to someone with Rs.900 than a similar transfer from a man
with Rs.1000100 to someone with Rs.1000000.
We may see many measures do not satisfy any of the two conditions of transfer
and some satisfy only the first one. Those that satisfy the second transfer axiom
automatically satisfy the first transfer axiom.
90
I 6) Axiom of Symmetry Measures of Inequality
If distribution ( x1p , x 2p ,..., x Np ) were a permutation of distribution ( x1 , x 2 ,..., x N ) ,

then the inequality level of both the distributions would be the same.
This implies that if two persons interchange their income positions, inequality
measure does not change. Thus the axiom ensures impartiality between
individuals for non-income characteristics: The evaluator does not distinguish
between Amar, Akabar and Anthony; nor between Shiela and Peter; or between
Mr. Pygmy and Ms. Dwarfy. Further, it means that the inequality depends only
on the frequency distribution of incomes.
7) Axiom of Interval
The inequality measure should lie in the closed interval of (0,1).
The measure is supposed to assume the value of zero when all incomes are
equal, which means when all persons have equal income and the value of unity
when only one individual gets all the income (and other have zero incomes, not
negative incomes).
Most people tend to agree with the axiom. A few, notably Theil (1967) and
Cowell (1995), disagree. They hold that the situation of one person grabbing
all the income in a society of 2 persons cannot be described by the same level
of inequality as that of one person doing so in a society of 2 crore persons. It
would not be easy to assert that in the case of 2-person society the level of
inequality is unity when one person has all the income and the other has none.
In one case, 50 percent population is having positive income, in the other only
0.00000005 percent. Some people therefore qualify the axiom by saying that
when one person gets all the income the measure approaches unity in the limit
as the number increases.
When a measure has a finite maximum, it is easy to transform such an index

into the one, which has maximum value 1. Most measures, though not all, have
zero as their minimum value. But question that Cowell (1995) raises is that
there are many ways in which the measure could be transformed so that it lies
in the zero-to-one range but each transformation has different cardinal
properties.
8) Axiom of Decomposability
Suppose population can be sub-divided into several groups and an over-all
index of inequality was a function of group-wise indexes and if the population
mean can be expressed as weighted average of group means, the population
index of inequality can be regarded as decomposable. The groups might be
defined as comprising of people in different occupations, residents of different
areas, with different religious or educational backgrounds etc.
However we find a lot of overlapping in these groups. This leads sum of the
weights to differ from unity.
Not all indices are found to be decomposable. Gini coefficient, a very popular
measure is decomposable only if the constituent groups are non-overlapping.
Cowell (1995) has conducted a beautiful experiment. First, he computes four
inequality measures for two distributions of same size and same mean-each
divided into two groups of equal size in a manner that there is no overlapping:
91
Quantitative Methods-I Population A: (60,70,80), (30,30,130) M
Population B: (60,60,90), (10,60,120)
Now, it is found that the group means and population means in two
distributions are the same and group inequalities in B are higher than their
counterparts in A. But when we compute overall inequalities, one of the
measures suggests that the magnitude of inequality in B is lower than that in A.
And the measure used is Gini, which is very popular among economists. As he
says, ‘strange but true’. If the component inequality magnitudes are higher and
the weights are the same, how could overall measure be lower? It is therefore
impossible to express overall inequality (change) as some consistent function
of inequality change in the consisted groups.
These are all intuitively appealing axioms. There does remain scope for
formulating other axioms. In the literature on poverty measurement one finds a
plethora of axioms developed by a number of contributors working in that
areas. But we shall be content with these only.
92
93
I Construction of Composite
UNIT 12 CONSTRUCTION OF COMPOSITE Index in Social Sciences
INDEX IN SOCIAL SCIENCES

Structure
12.0 Objectives
12.1 Introduction
12.2 Composite Index: The Concept
12.3 Steps in Constructing Composite Index
12.4 Dealing with Missing Values and Outliers
12.5 Methods to Construct Composite Index
12.5.1 Simple Ranking Method
12.5.2 Indices Method
12.5.3 Mean Standardization Method
12.5.4 Range Equalization Method
12.6 Principal Component Analysis (PCA)
12.6.1 Conducting PCA and Analyses of Results
12.6.2 Use of Output Indicators
12.6.3 Use of Weight
12.6.4 Limitations of Principal Component Analysis
12.7 Merits and Limitations of Composite Index
12.8 Let Us Sum Up
12.9 Exercises
12.10 Some Useful Books/References
12.0 OBJECTIVES
• state the concept of composite index;
• describe the process of constructing the composite index;
• explain the various methods of composite index;
• discuss the merits and limitations of composite index; and
• learn how to interpret the results derived from composite indexes.
12.1 INTRODUCTION
In social sciences research, many a times the complex social and economic
issues like child deprivation, food security, human well-being, human
development etc. are difficult to measure in terms of single variable. The
reason being that such issues have several dimensions and indicators. For
example, it is difficult to explain the status of development of a district in terms
93
Quantitative Methods-I of a single variable because development is reflected in terms of several Co
indicators. Some of such variables/indicators are quantitative type while others
are of qualitative nature. In such situations, Composite index plays an
important role to express the single value of several inter-dependent or
independent variables. Further, the composite index makes it possible to
compare the performance among region/states or districts etc. Hence,
composite indexes are being recognized as a useful tool for policy analysis.
Composite indexes can also be resorted to make comparison among different
regions/sectors where wide range of variables are used.
In this Unit, we shall therefore discuss the concept of composite index, the
process of their construction, various methods, their uses, and limitations. The
study of the unit will also enable you to learn how to interpret the results. Let
us begin to explain the concept of composite index.
12.2 COMPOSITE INDEX: THE CONCEPT

A composite index is an expression of a single score made by combination of
different scores to measure a given variable or a group of variables. It
expresses quantity or place (position) of multi facet aspects of a concept. The
UNDP (2005) has explained that ‘a composite index expresses a quantity or a
position on a scale of qualitative multi-faceted aspects….. which is relevant for
information of the society’. An index can be a combination of independent
indicators, or the average of an accumulation of selected indicators. The index
represent specific concept or highlighting specific sector or areas like status of
dalit (dalit deprivation index), food situation (food security/insecurity index),
status of child (child deprivation/development index), quality of human
development (human development index) etc.
The index that we construct is the outcome of some unidirectional variables or

indicators. If an index is constructed by taking positive indicators, the higher
value of index implies higher development and lower values imply lower
development. For example, in case of index related to child development, if the
variables are positive directional, the final index can be termed as ‘child
development index’. On the other hand, if the variables are negative
directional, it is called ‘Child Deprivation Index’. Let us take an example that
child mortality is a negatively directional variable whereas percentage of
children immunized is a positively directional variable. Such index is very
useful in case of qualitative data. An index is more robust than a single
indicator or variable.
The choice of indicator is a big challenge for researchers. The major issues in
identifying measurable indicators are: whether data are available or not,
whether we have reliable data, whether the data are cross-section or time series,
and minimization of double counting raised from overlap or redundancy. Again
if the variables to be chosen are easy to understand, they are more acceptable to
a wider audience.
12.3 STEPS IN CONSTRUCTING COMPOSITE

INDEX
The following steps are required to follow in construction of composite index:
94
Index in Social Sciences
Step Why it is needed
1. Theoretical framework • To get clear understanding and

definition of the multidimensional
A theoretical framework need to be phenomenon to be measured.
developed because it provides the basis
for the selection and combination of • To structure the various sub-groups
variables into a meaningful composite of the phenomenon (if needed).
indicator under a fitness-for-purpose • To compile a list of selection
principle (involvement of experts and criteria for the underlying variables,
stakeholders is envisaged at this step). e.g., input, output, process.
2. Data selection • To check the quality of the

available indicators.
Indicators should be selected on the
analytical soundness, measurability, • To discuss the strengths and
country coverage, and relevance of weaknesses of each selected
the indicators to the phenomenon indicator.
being measured and relationship to • To create a summary table on data
each other. The use of proxy variables characteristics, e.g., availability
should be considered when data are (across country, time), source, type
scarce (involvement of experts and (hard, soft or input, output,
stakeholders is envisaged at this process).
step).
3. Imputation of missing data • To estimate missing values.

Consideration should be given to • To provide a measure of the reliability
different approaches for imputing of each imputed value, so as to assess
missing values. Extreme values the impact of the imputation on the
should be examined as they can composite indicator results.
become unintended benchmarks. • To discuss the presence of outliers
in the dataset.
4. Multivariate analysis • To check the underlying structure
of the data along the two main
Should be used to study the overall dimensions, namely individual
structure of the dataset, assess its indicators and countries (by means
suitability, and guid subsequent of suitable multivariate methods,
methodological choices (e.g., e.g., principal components analysis,
weighting, aggregation). cluster analysis).
• To identify groups of indicators or
groups of countries that are
statistically “similar” and provide
an interpretation of the results.
• To compare the statistically-
determined structure of the data set
to the theoretical framework and
discuss possible differences.
5. Normalisation • To select suitable normalization
procedure(s) that respect both the
Should be carried out to render the theoretical framework and the data
variables comparable properties.
95
Quantitative Methods-I • To discuss the presence of outliers Co
in the dataset as they may become
unintended benchmarks.
• To make scale adjustments, if
necessary.
• To transform highly skewed
indicators, if necessary.
6. Weighting and aggregation • To select appropriate weighting and
aggregation procedure(s) that
Should be done along the lines of the respect both the theoretical
underlying theoretical framework. framework and the data properties.
• To discuss whether correlation
issues among indicators should be
accounted for.
• To discuss whether compensability
among indicators should be
allowed.
7. Uncertainty and sensitivity • To consider a multi-modeling
analysis approach to build the composite
indicator, and if available,
Should be undertaken to assess the alternative conceptual scenarios for
robustness of the composite indicator the selection of the underlying
in terms of e.g., the mechanism for indicators.
including or excluding an indicator,
the normalization scheme, the • To identify all possible sources of
imputation of missing data, the choice uncertainty in the development of
of weights, the aggregation method. the composite indicator and
accompany the composite scores
and ranks with uncertainty bounds.
• To conduct sensitivity analysis of
the inference (assumptions) and
determine what sources of
uncertainty are more influential in
the scores and/pr ranks.
8. Back to the data • To profile country performance at
the indicator level so as to reveal
Is needed to reveal the main drivers what is driving the composite
for an overall good or bad indicator results.
performance. Transparency is
primordial to good analysis and • To check for correlation and
policymaking. causality (if possible).
• To identify if the composite
indicator results are overly
dominated by few indicators and to
explain the relative importance of
the sub-components of the
composite indicator.
9. Links to other indicators • To correlate the composite indicator
with other relevant measures, taking
Should be made to correlate the into consideration the results of
composite indicator (or its sensitivity analysis.
96
I dimensions) with existing (simple or • To develop data-driven narratives Construction of Composite
composite) indicators as well as to based on the results. Index in Social Sciences
identify linkages through regressions.
10. Visualisation of the results • To identify a coherent set of

presentational tools for the targeted
Should receive proper attention, audience.
given that the visualization can
influence (or help to enhance) • To select the visualization
interpretability. technique which communicates the
most information.
• To present the composite indicator
results in a clear and accurate
manner.
Source: OECD (2008), ‘Handbook on Constructing Composite Indicators Methodology and
User Guide’
Caution on choosing variable

1) Whenever we choose any variable or a particular dimension for the index,
we have to justify the inclusion of the variable into the index. This
justification should come from empirical evidences or policy based
research studies or from theoretical explanation.
2) If the variable is not unidirectional, the entire variables used should be

converted to unidirectional. For example, in construction of the food
security index, two variables like ‘proportion of agricultural worker to
total workers’ and per capita value of agricultural output’ is used for
index. Here the first variable is a negatively directional whereas the second
variable is positively directional. In this case we have to convert the entire
variables into either positive direction or negative direction. If we want to
convert this to positive direction, the first variable which has a negative
direction should be deducted from 100. On the other hand, if we want to
convert the entire variables into negative direction, we have to work out
the reciprocal of per capita value of agricultural output.
12.4 DEALING WITH MISSING VALUES AND

OUTLIERS
After selecting the indicators, you have to have a clear idea on missing values
for each selected variable. Data can be missing in a random or non random
fashion. In such situation, easy solution is to drop cases for which data is
missing. However, before dropping the variable, you have to see the number of
cases for which values are missing, because exclusion of such households for
which data is missing could significantly lower sample size. Further, deleting
such household may lead to some biases also. For example, in livelihood study
of households, if there is a missing value, we have to find out the frequency of
missing value because in many cases, the chances of missing value is high for
lower economic class as compared to upper economic class. In such cases if we
delete the households having missing value, it may result biases towards upper
economic classes of household. In case, inclusion or exclusion of households
having missing value put a little impact on the final result, we can delete such
cases.
97
Quantitative Methods-I The second solution is to impute the missing value by applying some Co
methodology or logic. But use of imputation is more common in academic
indices and datasets. Here you can try to impute the missing value which
should be as accurate as possible..For example, in calculating the consumption
expenditure of 50 households of which about 5 cases are missing. In this case
in imputing we can substitute the average value of consumption expenditure
and replace the missing value by the averages. But for a more accurate
estimation we can group theses 50 households on the basis of value of assets
holding and see to which asset category these ‘MPCE missing’ households
belonged to. Then we can substitute the average consumption expenditure of
the respected asset group.
Such approach has been used in a number of more recent indices such as the
Corruptions Perceptions Index (CPI) developed by Transparency International
and the World Bank’s Worldwide Governance Indicators (WGI). Such
approaches carry a dual advantage. They allow scores to be estimated for a
maximal number of countries, and can use a broader range of indicators to
triangulate indices for nebulous constructs.
Many a time,outliers may also disturb the analysis. For example if the income
of four persons is 18000, 17000, 18500 and 19000 then the average is 18125. If
we include the income of fifth person as 60000 then the average turns out to be
26500. In such matters, you have to drop the cases with high extreme value.
Both in case of missing value and outlier you are expected to document and
explain the selected imputation procedures and the results in detail.

1) What do you understand by the term Composite Index?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
2) Give some examples of Composite Index.

……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
3) List the steps involved in construction of Composite Index.

……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
98
I 4) State the alternative methods to deal with the missing values of the selected Construction of Composite
variables. Index in Social Sciences
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
12.5 DIFFERENT METHODS OF COMPOSITE

INDEX
The various methods are used in construction of composite index. These
include: Simple Ranking Method, Indices Method, Mean standardization
method, Range Equalization Method, and Principal Component Analysis
Method. Let us discuss one by one with simple examples.
12.5.1 Simple Ranking Method

Rank Method is one of the simplest methods to analyse the status of a
region/district/state/country. In this method, the first step is to arrange/convert
the variable into unidirectional for each district or state. Let us take an example
of development status of different districts in Orissa state. First, we have to
convert the relevant variables in either positive or negative direction. For
example, the variable says ‘the proportion of agricultural labour is negatively
associated with the development of a region. Hence we have to convert this
variable to positive direction by deducting this variable from 100 and change
the labeling of variable as ‘proportion of other than agricultural labour to total
labour’. Higher the value of this variable, higher the development of the district
or region. Depending on the value, each variable was ranked in the similar
manner. Highest rank (1st ) was given to the variables with high value and vice-
a-versa. Alternatively, we can do reverse ranking of the variable i.e. highest
rank is given to the variables with lowest value. The individual ranks are
added to get the total rank value for the district. This has been illustrated in the
example given in table 12.1. Our objective is to find out most backward tribal
district in Odisha (having more than 50 per cent of tribal population to total
population). A total of 11 districts qualify for tribal dominant area having 50
per cent or more proportion of population. We have selected 10 variables for
identifying most backward tribal areas. Each of the variables for 11 tribal
dominated districts was ranked according to total value of variable. The
individual rank of all the variables of all the districts is given in table 12.2. As
all the variable are unidirectional, rank 1 is given to district having higher value
and vice-a-versa. After doing so we can add together the value of all the
variables and find out the average rank for each district (Sum of total rank
divided by number of variable i.e. 10). The average value of variable shows the
status of district on development indicators.
99
Quantitative Methods-I Table 12.1: Districtwise Development Indicators of Tribal Concentrated Co
Districts in Odisha
Average casual wage rate

% of net irrigated area to
agricultural labour to all
Per capita consumption
% of household getting
having access to paved

Percentage of villages
having access to Phcs
Female Literacy rate
safe drinking water

women work force
Per capita value of
participation rate
agrictural output
% of other than
total sown area
expenditure
labour
road
District
Actual Value
Malkangiri 73 31 1304 201 18 67 82 21 16 36
Rayagada 50 23 479 201 18 67 78 34 13 37
Sundargarh 61 19 537 280 43 58 57 38 28 32
Nabarangapur 46 6 839 201 18 66 80 45 28 37
Kandhamal 62 14 410 201 33 68 32 21 19 36
Koraput 55 31 611 201 16 67 67 22 22 40
Mayurbhanj 60 30 572 280 35 62 44 42 21 31
Gajapati 52 25 529 331 24 77 43 29 20 34
Sambalpur 62 34 1075 280 50 62 56 31 13 39
Kendujhar 60 23 537 280 44 46 52 46 15 34
Jharsuguda 68 20 662 280 54 43 63 47 32 39
Average 59 23 687 249 32 62 60 34 21 36
Table 12.2: Districtwise Development Scenario of Tribal Orissa by Simple Ranking Method
women work force participation
access to primary health centre

% of net irrigated area to total
Percentage of villages having

Per capita value of agrictural
% of household getting safe

% of other than agrictural
Average Rank/indices
access to paved road

labour to all labour
drinking water
expenditure
sown area
output
rate
District
Malkangiri 1 2 1 7 8 5 1 10 8 7 5.0
Rayagada 10 7 10 7 9 3 3 6 11 4 7.0
Sundargarh 5 9 7 2 4 9 6 5 3 10 6.0
Nabarangapur 11 11 3 7 10 6 2 3 2 5 6.0
Kandhamal 3 10 11 7 6 2 11 11 7 6 7.4
Koraput 8 3 5 7 11 4 4 9 4 1 5.6
Mayurbhanj 7 4 6 2 5 7 9 4 5 11 6.0
Gajapati 9 5 9 1 7 1 10 8 6 8 6.4
Sambalpur 4 1 2 2 2 8 7 7 10 3 4.6
Kendujhar 6 6 8 2 3 10 8 2 9 9 6.3
Jharsuguda 2 8 4 2 1 11 5 1 1 2 3.7
100
I By this method, the most developed tribal district is Jharsuguda and most Construction of Composite
backward district is kandhamal. Index in Social Sciences
12.5.2 Indices Method

The indices method is another simple method for calculating the status of
development of an area/district or state. Like rank method, here also, at the first
stage the variables are converted into one direction. After converting the
variable into one (positive or negative) direction, we calculate the index. In this
method we have to convert the district figures based on the average figure for
the entire 11 districts as 100. Let us take an example of table 12.1. If the
proportion of other than agricultural labour to total labour in malkangiri district
is 73 and the average of that variable for all the 11 districts is 59, then the

indices of that variable for Malkangiri district will be: 73 124

In the same manner, we can find out the indices for all the variables for 11
districts as is given in Table 12.3. The final index of development for the
districts can be obtained by taking the arithmetic mean for all the indicators
(given in last column of table 12.3). After working out the average indices
value for all the districts, we can rank all the districts. The district with highest
indices will be most developed and the district with lowest indices will be least
developed.
Table 12.3: Districtwise Development Scenario of Tribal Orissa by Indices Method

% of net irrigated area to total

% of households getting safe

% of other than agrictural
Average Rank/indices
women work force

participation rate
drinking water
expenditure
sown area
output
District
Malkangiri 124 133 190 81 57 108 138 62 78 99 107.0
Rayagada 85 99 70 81 57 108 131 98 64 103 89.6
Sundargarh 104 83 78 113 135 93 96 111 135 90 103.6
Nabarangapur 78 25 122 81 56 107 135 132 137 102 97.4
Kandhamal 106 59 60 81 102 109 54 60 92 100 82.3
Koraput 93 133 89 81 49 108 113 65 108 113 95.1
Mayurbhanj 101 129 83 113 109 101 74 123 103 86 102.2
Gajapati 89 106 77 133 76 123 73 86 95 95 95.3
Sambalpur 105 147 157 113 155 100 95 90 65 108 113.3
Kendujhar 102 99 78 113 136 74 88 136 71 94 99.1
Jharsuguda 114 88 96 113 167 69 105 138 153 109 115.3

101
Quantitative Methods-I In our example Kandhamal is the most backward with indices (82.3) and Co
Jharsuguda (115.3) is the most developed district.
Let us make a comparison between the results obtained by rank method and
indices method. Results in Table 12.4 shows that Jharsuguda, Sambalpur, and
Malkangiri are most developed in both the methods whereas Raygada and
Kandhamal are most backward in both the methods. While comparing both
rank and indices method, we can also run correlation between the two set of
indices. If the correlation is high then we can say that the finding in both
methods is almost similar.
Table 12.4: A Comparison of Simple Ranking Method and Indices Method

Status of District in Status of District in
District Rank Method District indices Method
Jharsuguda 3.7 Jharsuguda 115
Sambalpur 4.6 Sambalpur 113
Malkangiri 5.0 Malkangiri 107
Koraput 5.6 Sundargarh 104
Sundargarh 6.0 Mayurbhanj 102
Nabarangapur 6.0 Kendujhar 99
Mayurbhanj 6.0 Nabarangapur 97
Kendujhar 6.3 Gajapati 95
Gajapati 6.4 Koraput 95
Rayagada 7.0 Rayagada 90
Kandhamal 7.4 Kandhamal 82
Note: Correlation of status of district in both rank and indices method is 0.927
12.5.3 Mean Standardization Method

Mean standardization Method is another simple method used both as a process
of normalization and also a composite index. In this method we normalize the
value of each variable and than work out the average of the normalized value
for all the variables. The average of normalized value will be the composite
index value. The normalization is done by dividing the actual value of variables
by their respective means. Let us take an example from Table 12.1. For
example the standardized value of the variable ‘% of other then agricultural
labour to all workers’ for Malkangiri district can be found out by
Normalized value MS Method=Actual value (73)/Mean value (59) =1.237
In the same process we can find out the normalized value of the entire variable
for 11 selected districts. The standardized value of each selected indicator for
different district is given in Table 12.5. The composite index value of each
district given in the last column of Table 12.5 has been worked out by taking
average of all the 10 selected indicators (row vector of the each district).
102
Table 12.5: Index Value in Mean Standardization Method
Percentage of villages having access to

Percentage of household getting safe
women work force participation rate
Per capita value of agrictural output
Per capita consumption expenditure

Percentage of other than agrictural
Percentage of net irrigated area

primary health centre
Composite Index
drinking water
paved road
District
Malkangiri 1.237 1.332 1.899 0.808 0.561 1.079 1.379 0.614 0.775 1.003 1.069
Rayagada 0.847 0.988 0.697 0.808 0.561 1.079 1.312 0.995 0.630 1.030 0.895
Sundargarh 1.034 0.816 0.782 1.126 1.340 0.934 0.959 1.112 1.357 0.891 1.035
Nabarangapur 0.780 0.258 1.222 0.808 0.561 1.063 1.346 1.316 1.357 1.030 0.974
Kandhamal 1.051 0.602 0.597 0.808 1.028 1.095 0.538 0.614 0.921 1.003 0.826
Koraput 0.932 1.332 0.890 0.808 0.499 1.079 1.127 0.644 1.066 1.114 0.949
Mayurbhanj 1.017 1.289 0.833 1.126 1.091 0.999 0.740 1.229 1.018 0.863 1.020
Gajapati 0.881 1.074 0.770 1.331 0.748 1.240 0.723 0.848 0.969 0.947 0.953
Sambalpur 1.051 1.461 1.565 1.126 1.558 0.999 0.942 0.907 0.630 1.086 1.132
Kendujhar 1.017 0.988 0.782 1.126 1.371 0.741 0.875 1.346 0.727 0.947 0.992
Jharsuguda 1.153 0.859 0.964 1.126 1.683 0.693 1.060 1.375 1.551 1.086 1.155
12.5.4 Range Equalization Method

Range Equalization (RE) method otherwise known as max-min approach is
adopted by UNDP in computation of Human Development Index. Under this
approach, an index is constructed for each variable by applying Range
Equalization formula derived by UNDP. This is worked out by subtracting an
indicator’s minimum value from each observation and then dividing it by its
range.

RE Index =
where Value of the variable, min X- Minimum value of X in the scaling ,

max X- Maximum value of X in the scaling. The RE index is also a
normalization technique. Without it, a composite index can be biased towards
an indicator with very high range. For example the per capita value of
agricultural output which is measured in rupees ranges from Rs. 410/- to
Rs. 1304/-. On the other hand, the variable proportion of female literacy rate is
measured in percentage. Different variables measured in different units
sometimes give upward or downward biases. To ensure the index value
biasfree, we convert all the variables into equal scaling from 0 to 1.
103
Quantitative Methods-I In RE method, as a first step we undertake scaling exercise. In undertaking the Co
scaling procedure, desirable norms are followed for each indicator. In some
cases, scaling of indicators is self-selected, while in others element of value
judgment is involved. This scaling exercise is called goal post where we
identify the maximum and minimum goal. The goalpost basically visualizes the
extent of minimum value and maximum value in the future time period (say
next 5 years). The scaling norm that we have adopted is given in Table 12.6. In
the table the third and fourth column shows the maximum and minimum
district value of selected 10 indicators. The first and second column is the
maximum and minimum goal post.
Table 12.6: Construction of Food Security Radar
Variable Description of Variables Goalposts District Value
Maximum
Minimum
Maximum
Minimum
District
District
Oth_agl Percentage of other than 30 85 46 73
agricultural labour to all labour
Irr Percentage of net irrigated area 2 55 6 34
to net cropped area
pcvao Per capita value of agricultural 200 2500 410 1304
output
Mpcce_ia Inequality adjusted per capita 150 450 201 331
consumption expenditure
Lit_f Female Literacy (adult) rate 10 70 16 54
Wfpr_f women work force participation 30 85 43 77
rate
Hhsdw Percentage of household having 20 90 32 82
access to safe drinking water
Paved_r Percentage of villages having 10 60 21 47
v_phcs Percentage of villages having 10 50 13 32
access to Primary health centre
Wage_c average Casual wage rate 25 70 31 40
Based on the goalpost, we normalize the value of the variable for all the
districts and all the indicators. In our example, the values have been presented
in Table 12.7.
104
I Table 12.7: Index Value in Range Equalization Method Construction of Composite

agrictural labour to all labour
getting safe drinking water

Percentage of net irrigated
Percentage of households
Percentage of other than
having access to primary

land to net sown area

women work force
Per capita value of
Composite Index
participation rate
agrictural output
health centre
expenditure
District
Malkangiri 0.782 0.547 0.480 0.170 0.133 0.673 0.886 0.220 0.150 0.244 0.429
Rayagada 0.364 0.396 0.121 0.170 0.133 0.673 0.829 0.480 0.075 0.267 0.351
Sundargarh 0.564 0.321 0.147 0.433 0.550 0.509 0.529 0.560 0.450 0.156 0.422
Nabarangapur 0.291 0.075 0.278 0.170 0.133 0.655 0.857 0.700 0.450 0.267 0.388
Kandhamal 0.582 0.226 0.091 0.170 0.383 0.691 0.171 0.220 0.225 0.244 0.300
Koraput 0.455 0.547 0.179 0.170 0.100 0.673 0.671 0.240 0.300 0.333 0.367
Mayurbhanj 0.545 0.528 0.162 0.433 0.417 0.582 0.343 0.640 0.275 0.133 0.406
Gajapati 0.400 0.434 0.143 0.603 0.233 0.855 0.329 0.380 0.250 0.200 0.383
Sambalpur 0.582 0.604 0.380 0.433 0.667 0.582 0.514 0.420 0.075 0.311 0.457
Kendujhar 0.545 0.396 0.147 0.433 0.567 0.291 0.457 0.720 0.125 0.200 0.388
Jharsuguda 0.691 0.340 0.201 0.433 0.733 0.236 0.614 0.740 0.550 0.311 0.485
After calculating the index of each variable, we have averaged them to give
each of the five dimensions of food security. The composite food security
index is again derived by averaging the five dimensions.
The normalized value for the variable percentage of other than agricultural
labour to total worker’ for Malkangiri district is found out in the following
manner.
(Actual value (73) – Minimum goalpost (30))

————————————————————— = 0.782
(Maximum goalpost (85) – Minimum goalpost (30))
Likewise we have converted and worked out the normalized value of all the
variables and for all the districts shown in Table 12.7.
Comparing MS Method and RE Method

The composite index value worked out by both the RE and MS methods can be
compared and analyzed (Table 12.8). We can also examine correlation between
two indices. Here the correlation between final index value worked out by both
the methods is very high (0.991).
105
Quantitative Methods-I Table 12.8: Comparison between RE and MS Methods Co
Composite Index Value by Composite Index Value by

District RE method MS method
Jharsuguda 1.155 0.485
Sambalpur 1.132 0.457
Malkangiri 1.069 0.429
Sundargarh 1.035 0.422
Mayurbhanj 1.020 0.406
Kendujhar 0.992 0.388
Nabarangapur 0.974 0.388
Gajapati 0.953 0.383
Koraput 0.949 0.367
Rayagada 0.895 0.351
Kandhamal 0.826 0.300
Note: Correlation of index value is 0.991
The high degree of the correlation between the values of composite index
computed by applying both methods separately indicate the high degree of
homogeneity between the two approaches. However, the Range equalization
method is preferred because it accounts for wider variations and strong
correlations to the PCA composite.
One important point should be kept in mind that the index value as discussed
above are based on equal weight. We can also give weights to variables/
components discussed in section 12.5.
1) State the procedure to workout Composite Index by way of Simple Ranking
Method.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) How will you compute the final value of Composite Index by Indices
Method?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
3) What is the Distinction between Range Equalization Method and Mean
Standarisation Method?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
…………………………………………………………………………….
106
12.6 PRINCIPAL COMPONENT ANALYSIS (PCA) Index in Social Sciences
PCA is one of the important composite index method used to analyse the various
problems in social sciences. PCA is a mathematical technique that transfers a
number of correlated variables into smaller number of uncorrelated variables
called the principal components. For our analysis we have an array of variables
which may have high correlation with each other. In case where there is high
relation between variables, the PCA is a suitable technique where the correlation
between different components is low.
Categorical data are not suitable for PCA analysis because in such cases
categories converted into quantitative scale has no meaning. To avoid this type
of problems, we should recode these variables into binary variables. Let us take
an example of social category of household that are Scheduled caste (1),
Scheduled tribe (2), Other backward caste (3) and General (4). In PCA analysis
such type of variables have no meaning. Such type of categorical variables can
be converted into a bianary variable. For example if we want to study the dalit
status, variable can be converted to a bianary variable as dalit (code 2) and non-
dalit (code 2, 3, and 4). Likewise, if we want to study the impact of reservation
for other backward castes we can categorize as OBC (code 3) and other than
OBC (code 1, 2 and 4).
The PCA is a data reduction technique. Under this technique, the original data
set is transformed into a new set of uncorrelated variables called principal
components. PCA reduces the number of variables in a data set to smaller
number of dimension/s. In a series of variables if amn is the weight of mth
principal component and nth variable then the matrix will be:
 PC 1   a11 a12 a13 ....... a1n   X 1 

 PC 2   a 21 a 22 a 23 ....... a 2 n   X 2 
    
 PC 3   .   X 3
    
 . = .  
 .  .  
    
 .    
    
 PCn  am1 am 2 am 3 ....... amn   Xn 
The matrix can be transformed into equation for principal component i.e.
PC1 = a11 x1 + a12 x2 + ..... + a1n xn
PC2 = a21 x1 + a22 x2 + ..... + a2 n xn
........
PCn = am1 x1 + am 2 x2 + ..... + amn xn
The components are ordered so that the first component (PC1) explains the
largest possible amount of variation in the original data, subject to the constraint
that the sum of the squared weights of the vectors is
equal to one.
107
Quantitative Methods-I In the output the eigen value is one of the important determinant of PCA. The Co
eigen value is a number that tell you how much variance there is in the data set.
In other words the eigen value is a number telling us how spread out the data is
on the line. The eigen vector with the highest eigen value is therefore the
principal component.
As the sum of the Eigen values equals the number of variables in the initial
data set, the proportion of the total variation in the original data set accounted
by each principal component is given by λi/n. The second component (PC2) is
completely uncorrelated with the first component, and explains additional but
less variation than the first component, subject to the same constraint. Each
subsequent component captures additional dimension in the data and adds
smaller and smaller proportion of variation in the original variables.
Before conducting PCA, we should run correlation between variables. If the
correlation between two variables is very high, we may remove one of those
variables. The reason being that the two variables seem to measure same thing.
The higher the degree of correlation among variables, the lower will be the number
of components required to capture common information. Sometimes two variables
may be combined in some ways (taking average). The PCA can be applied either
to the original values of variables or to the normalized values of the variables.
In general, normalization can be done by three methods, i.e (i) by deviation of
the variables from their respective means (i.e.); (ii) by dividing the actual
values by their respective means; (iii) and deviation of value of a variable from
the mean which is then divided by standard deviation {i.e. ()/σ}.We are
applying here the second method.
Let us try to apply and analyse PCA by using the database given in Table 12.1.
We are applying here the second method for normalization that is found out in
Table 12.5 column 2 to 10.
Table 12.9: KMO and Bartlett’s Test
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .394
Bartlett's Test of Sphericity Approx. Chi-Square 47.933
Df 45
Sig. .355
When the normalized database is prepared, you have to apply Kaiser-Meyer-

Olkin (KMO) technique and Measure of Sampling Adequacy and Bartlett's
Test of Sphericity (BTS). In KMO technique, the value varies between 0 and 1
and the value closer to1 is better. However, according to some statisticians, the
minimum value should be 0.6. The BTS tests the null hypothesis that the
correlation matrix is an identity matrix. Let us remember that identity matrix is
a matrice in which the main diagonal elements are 1 and off-diagonal elements
are 0.
The above two tests provide the minimum standard before conducting any PCA.
12.6.1 Conducting PCA and Analyzing Results
For the purpose of Principal Component analysis and interpreting its results,
SPSS is user friendly software. However, you can try to run the same in
STATA also. Let us take our previous example of Table 12.5. In the table,
there are 10 standardised variables and we can run our PCA model. Table
12.10 give the first output called ‘communalities’
108
I Table 12.10: (Output 1) Communalities: Extraction by Principal Component Construction of Composite
Analysis. Index in Social Sciences
Variable name Initial Extraction

% of other than agricultural labour to all labour 1.000 .721
% of net irrigated area to net sown area 1.000 .748
Per capita value agricultural output 1.000 .710
Monthly per capita consumption expenditure 1.000 .702
Female Literacy rate (adult) 1.000 .876
women work force participation rate 1.000 .858
% of household access to safe drinking water 1.000 .789
% of villages having access to paved road 1.000 .773
% of villages having access to primary health centres 1.000 .572
Average Casual wage rate 1.000 .491
Communalities represent the percentage of variance explained by the extracted

components. This explains the proportion of each variable’s variation
explained by PCA. The short notation of communalities is ‘h2’ and defined as
sum of squared factor loading. In table 12.10, the first column is the name of
variables used in our PCA. The second column is the ‘initials’. The initial value
of the communalities in PCA is 1. The second column ‘extraction’ shows the
proportion of each variable’s variation captured by PCA. This value ranges
between 0 and 1. When the value is near to 1 that means the variable is well
represented and the reverse for value is closer to ‘0’.
If the communality is low for an item, the reason might be that the item was
poorly designed. If the item has very little variance the communality is low. If
the different items usually resulting from large positive or negative skewed the
communalities is low. We can take an example that if everyone ticks strongly
agree, the variation within the variable is low and the communality is low. If
the communality is low, we can either remove the item from the analysis to
exclude it from any further analyses or we can treat it as a stand alone variable.
Table 12.11: (Output 2) Total Variance Explained: Extraction Method: Principal
Component Analysis.
Initial Eigen values Extraction Sums of Squared Loadings
% of Cumulative % of Cumulative
Component Total Variance % Total Variance %
1 3.141 31.413 31.413 3.141 31.413 31.413
2 2.182 21.820 53.233 2.182 21.820 53.233
3 1.919 19.195 72.428 1.919 19.195 72.428
4 .866 8.658 81.086
5 .750 7.502 88.587
6 .593 5.927 94.514
7 .365 3.645 98.160
8 .098 .976 99.135
9 .062 .625 99.760
10 .024 .240 100.000
10.0
109
Quantitative Methods-I The Table 12.11 shows ‘total variance explained’. The first column shows the Co
number of components. The number of components should be equal to number
of variables used for PCA. In our example, we have used 10 variables and
hence the total number of component is 10. The second part of the table reflect
initial Eigen value which consists of three columns i.e. Total, percentage of
Variance, Cumulative percentage. The initial Eigen values are the variances of
principal components. Here as we use the standardized values of variable, the
variance becomes equivalent to 1 and the total variance is equal to the total
number of variables i.e. ‘10’. The second column ‘total’ contains the Eigen values.
Here it can easily be seen that the Eigen values of subsequent components
gradually reduce implying that the successive components add less and less
variation. The third column shows the percentage variation on the components and
the fourth column explains the cumulative addition of percentage variation of
cumulative percentage components. In our example the first component explain
31.4 per cent of total variation and the second component explain 21.8 percent of
total variation and so on. The third row of fourth column shows that the first three
components explain 72 per cent of total variation.
The second part of the table is ‘Extraction Sums of Squared Loadings’ which
consists of three columns – Total, percentage of Variance and Cumulative
percentage. This has actually reproduced the three rows of figures of the first
part of table reflecting the components whose eigen value is greater than 1.
3
Eigenvalue
1 2 3 4 5 6 7 8 9 10
Component Number
Graph 12.1: Scree plot
110
I The ‘scree plot’ is a graph whose ‘X axis’ represent the component number and Construction of Composite
‘Y axis’ represents ‘Eigen values’. In other words scree plot demonstrates the Index in Social Sciences
graph of first two column of ‘output 2’ table. From the graph one can select the
number of components to be taken for analysis. It can be clearly seen from the
above graph that from the fourth component onward the line becomes flat
indicating that when the successive addition is less and less, the line of the
graph becomes more and more flatter.
Table 12.12: (Output 4) Component Matrix(a)
Componenta
1 2 3
Percentage of other than agricultural labour to all

.202 .807 .174
labour
Percentage of net irrigated area to total sown area. -.185 .793 -.293
Per capita value of agricultural output -.403 .596 .438
Monthly per capita consumption expenditure .717 .217 -.378
Female Literacy rate .837 .406 .105
women work force participation rate -.721 -.233 -.533
Percentage of households having access to safe

-.504 -.041 .730
drinking water
Percentage of villages having access to paved road .733 -.291 .389
Percentage of villages having access to primary

.441 -.417 .451
health centre
Average Casual wage rate -.430 .145 .535
Extraction Method: Principal Component Analysis.

a 3 components extracted.
Table 12.12 (output 4) shows Component Matrix of first three components.

The first column is the name of variable used in PCA. The values in the output
table are the component loading which are the correlation between the
variables and components. The values in the tables range from -1 to +1. One
can note that the sum of square of all three component matrix of a variable is
equal to the extraction value of the communalities of that variable. In this
example three component has been extracted as the eigen value for these
component is greater than 1.
111
Quantitative Methods-I Table 12.13: (Output 5) Reproduced Correlations Co
women work force participation

agricultural labour to all labour
Percentage of net irrigated area


Percentage of of households
getting safe drinking water

Percentage of other than

to total sown area
expenditure
rate
output
Reproduced Correlation
labour to all labour 0.721 0.554 0.477 0.253 0.512 -0.424 -0.007 -0.022 -0.171 0.120
Percentage of net irrigated to total sown
area 0.554 0.748 0.419 0.151 0.137 0.102 -0.152 -0.479 -0.543 0.032
Per capita value of agrictural output 0.477 0.419 0.710 -0.324 -0.049 -0.083 0.500 -0.297 -0.228 0.491
Per capita consumption expenditure 0.253 0.151 -0.324 0.702 0.648 -0.366 -0.645 0.314 0.055 -0.479
Female Literacy rate 0.512 0.137 -0.049 0.648 0.876 -0.754 -0.362 0.536 0.247 -0.246
women work force participation rate -0.424 0.102 -0.083 -0.366 -0.754 0.858 -0.015 -0.668 -0.460 -0.010
Percentage of households having access
to safe drinking water -0.007 -0.152 0.500 -0.645 -0.362 -0.015 0.789 -0.074 0.123 0.603
paved road -0.022 -0.479 -0.297 0.314 0.536 -0.668 -0.074 0.773 0.620 -0.144
Percentage of village access to primary
health centre -0.171 -0.543 -0.228 0.055 0.247 -0.460 0.123 0.620 0.572 -0.005
Average Casual wage rate 0.120 0.032 0.491 -0.479 -0.246 -0.010 0.603 -0.144 -0.005 0.491
Residual(a)
labour to all labour -0.149 -0.039 -0.161 -0.040 0.002 -0.087 -0.161 0.154 -0.134
Percentage of net irrigated area to total
sown area -0.149 -0.069 0.096 -0.066 -0.030 0.121 0.111 0.016 0.033
Per capita value of agrictural output -0.039 -0.069 0.139 -0.028 0.145 0.045 0.089 0.046 -0.180
Monthly per capita consumption
expenditure -0.161 0.096 0.139 -0.039 0.119 0.120 0.084 0.067 0.045
Female Literacy rate -0.040 -0.066 -0.028 -0.039 0.008 -0.082 -0.054 -0.052 0.151
women work force participation rate 0.002 -0.030 0.145 0.119 0.008 0.001 -0.016 0.147 -0.025
Percentage of households having access
to safe drinking water -0.087 0.121 0.045 0.120 -0.082 0.001 0.143 -0.070 -0.173
paved road -0.161 0.111 0.089 0.084 -0.054 -0.016 0.143 -0.183 -0.113
primary health centre 0.154 0.016 0.046 0.067 -0.052 0.147 -0.070 -0.183 -0.007
Average Casual wage rate -0.134 0.033 -0.180 0.045 0.151 -0.025 -0.173 -0.113 -0.007
Extraction Method: Principal Component Analysis.

a) Residuals are computed between observed and reproduced correlations. There are 29 (64.0%) no redundant residuals with absolute
values greater than 0.05.
b) Reproduced communalities
112
I Table 12.13 (Output 5) has two parts of analysis: reproduced correlation and Construction of Composite
residuals. The reproduced correlation is the correlation among the extracted Index in Social Sciences
components. From this table, we intend to ensure that the correlation between
the original variables and reproduced matrix should be as close as possible. The
lower part of the table ‘residual matrix’ shows the difference between original
variable and residual matrix. For a good PCA, it was expected that the
difference between original variable and extracted matrix should be near to
zero. Once the difference is near to zero, it can be said that the extracted
components accounted a larger variation in the original correlation matrix.
Here we can take an example from the table that the original correlation
between irrigation and other than agricultural labour is 0.403, Whereas the
extracted correlation between these two variable is 0.554 and the difference
between the two correlation given in residual part is 0.403-0.554= -0.149.
Final PCA Index Value

The final index is calculated by the addition of multiplication of normalized
value of the variable and the Eigen vector of that variable (first component). In
our example the PCA index value for all the 11 selected district can be
calculated by using two Table 12.5 and Table 12.12 given above. The final
PCA index of first district (Malkangiri) is
PCA index Malkangiri District = a11x1+a12x2+……+a1nxn

= (0.202*1.237) + (-0.185*1.332) + (-0.403*1.899) + (0.717*0.808) +
(0.837*0.561) + (-0.721*1.079) + (0.504*1.379) + (0.733*0.614) +
(0.441*0.775) + ( -.430*1.003)= 0.049
Likewise the PCA Index value of all other districts can be calculated as given
in Table 12.14
Table 12.14: PCA Index Value
District PCA Index
Malkangiri 0.049
Rayagada 1.273
Sundargarh 1.322
Nabarangapur 1.145
Kandhamal 1.641
Koraput 0.517
Mayurbhanj 1.333
Gajapati 1.757
Sambalpur 0.465
Kendujhar 0.290
Jharsuguda 1.551
113
Quantitative Methods-I The table 12.14 reveals that Gajapati is the most developed district followed by Co
Kandhamal district. On the other hand, Malkangiri is most backward district.
12.6.2 Use of Output Indicators

After finding out the Index findings need to be validated. This can be done by
comparing the result with some output indicators. Let us illustrate with
example discussed in section 12.6. District wise final index as arrived out by
PCA method is given in Table 12.14. This finding can be validated be taking
an output indicator say infant mortality rate (IMR). We can run a correlation
between PCA index and the output indicator ‘Infant mortality rate’. This has
been provided in Table 12.15 which is 0.893. The table shows that the most
backward district has a higher degree of mortality. Hence our final index is
validated. If suppose the correlation between these two variable is -0.120, then
our index is not validated as the correlation between the two variable is very
low.
Table 12.15: PCA Index and Comparing with Output Indicators
District PCA Index Infant Morality Rate*

Malkangiri 0.049 120
Kendujhar 0.290 100
Sambalpur 0.465 110
Koraput 0.517 92
Nabarangapur 1.145 89
Rayagada 1.273 100
Sundargarh 1.322 80
Mayurbhanj 1.333 75
Jharsuguda 1.551 65
Kandhamal 1.641 60
Gajapati 1.757 55
Note: Imaginary numbers
12.6.3 Use of Weight

Assigning weight i.e. all the indicators/dimensions are treated equally or
differentially is an important issue in construction of composite index..
Sometimes different dimensions are given differential weight but in
construction of overall index, an equal weight approach is followed. The
human development index assigns differential weights to indicators/dimension
but for overall index, an equal weight approach is followed. In construction of
human development index, in 2001 by Planning Commission, the overall index
was calculated by using equal weight whereas the different sub-indices like
Composite indicator on educational attainment, Composite indicator on health
attainment differential weights were given. In calculating educational
attainment, two variables i.e. ‘literacy rate for the age group 7 years and
‘adjusted intensity of formal education’ were used where the first and second
variable were given 1/3 and 2/3 weight respectively. The other ‘health
attainment was calculated by taking two variables ‘life expectancy at age one,’
114
I and ‘infant mortality rate’. The first variable has a weight of 2/3 whereas the Construction of Composite
second variable 1/3. But the overall weight is calculated by taking equal weight Index in Social Sciences
for both the variables. Take another example of our range equalization method
where we have given equal weight to all variables. In some other cases, the
individual variables of sub indices were given equal weights but the overall
index was calculated by giving differential weights to sub-indicators.
Let us denote the weight in mathematical notation:

I=XW,
where ‘X’ is a matrix with m rows and n column, ‘I’ is identity matrix and ‘W’
is weight. Hence the final indices can be arrived at by taking the weighted
component/variables. There are basically two ways in assigning weights.
According to Munda and Nardo (2005), we can define weight by taking the
importance of the particular variable or group of variables. In this process, the
weights are arrived at by the past knowledge or observation of individual and
the probable effects of that variable in the overall analysis. Let us illustrate
with an example. Suppose we want to find out the sanitation and hygiene
Index. For this we have selected some variables like ‘proportion of people
having access to toilet’, ‘proportion of people having drainage’, ‘proportion of
people with wash hand’, ‘proportion of people safely disposed child exgratia’
etc . Here either on the basis of literature or on the basis of our observation
from some villages, we found that not having toilet is very important then wash
hand before eating. In this case we can put higher weight for having toilet and
less weight for washing hand.
In case of literature, weights are derived from on the basis of theoretical or

statistical consideration.
12.6.4 Limitations of Principal Component Analysis

The major criticisms against the use of PCA is that the technique is arbitrary in
constructing indices. The number of components and the number of variables
used are not well defined. This method entirely depends on the first
component. If the first component does not explain large variation, this method
is not useful. Such possibility depends on nature of data and on relationship of
variable.
Alternative methods to PCA that can reduce the dimensionality of the data
include: correspondence analysis, multivariate regression or factor analysis.
The details about these methods have been provided in Unit 13 and Unit 16 of
Block 4.
12.7 MERITS AND LIMITATIONS OF

COMPOSITE INDEX
One of the important merits of a composite index is that this can summarize the
complex and multidimensional indicators into one indicator which helps the
policy makers for implementation of a particular programme or policy
decision. A composite index can easily be taken to interpret about the
development or backwardness of a particular region or area or sector. The
progress of a state in India say, e.g. Bihar can easily be accessed and we can
compare the status of Bihar at two or more points of time. The composite index
115
Quantitative Methods-I can reduce the number of indicators without reducing the underlying Co
information base. The composite index facilitates communication to general
public and promote accountability. The composite index is not free from
limitations. It has the risk of misleading policy message if it is poorly
constructed or misinterpreted. The calculation may be misleading and faulty if
sound statistical or conceptual principle is not applied. The assignment of
weight many times creates a debate. One of the principles of assigning weight
is on the basis of value judgment. This always being a source of criticism.
Again allocation of upper and lower bound (mainly in range equalization
method) is a point of criticism.
1) In the context of Composite Index, what are the uses of Principal

Component Analysis (PCA).
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) What is explained by the term ‘communality’?

……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
3) How is Final Index Value calculated?

……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
4) In construction of Human Development Index, how weights are assigned to

the different indicators?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
116
I 5) State the limitations of Composite Index. Construction of Composite
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
12.8 LET US SUM UP

Composite index is an important analytical technique to analyse the
developmental related issues like status of development, food security, human
development etc. at the district/region/state level. It is an expression of single
score made of different scores measuring the various dimensions and indicators
of particular issue. This Unit describes the process to derive composite index
by applying the various methods. These methods include simple ranking
method, indices method, mean standarisation method, and range equalization
method. The main advantage of the rank and indices method is that they are
easy to understand and interpret. The RE, MS and PCA methods uses all the
variables in reducing the dimensionality of the data. These methods are very
much useful in comparing across districts, states, countries or areas such as
rural and urban. Again these methods are useful to compare over time by
constructing composite index at two points of time by taking same
indicators/variables. The principal component analysis enables the students to
reduce the indicators that are uncorrelated to explain the variation in the
original data set. To run the PCA, SPSS and STATA softwares are preferred.
12.9 EXERCISES
1) Think and collect some gender related variables from the data sources like
women literacy, women workforce participation rate in the age group 15-
59, mean year schooling, mortality rate, immunization rate etc and find out
the gender development index.
2) Explain the process of using Principal Component Analysis (PCA) to

reduce the number of variables in a data set to smaller number of
dimensions.
3) By using range equalization method find out the food security index from
the report prepared by IHD-WFP given in website
‘http://122.180.7.122/displaymorePub.asp?itemid=84&subchkey=11&chna
me=Publications.‘
12.10 SOME USEFUL BOOKS/REFERENCES

A. Saltelli, G. Munda and M. Nardo (2006), ‘From Complexity to
Multidimensionality: the Role of Composite Indicators for Advocacy of EU
Reform’, Tijdschrift voor Economie en Management Vol. LI, 3
117
Quantitative Methods-I Downloaded from http://www.econ.kuleuven.be/rebel//jaargangen/2001- Co
2010/2006/TEM%202006-3/TEM_2006-3_03_Saltelli.pdf
Planning Commission (2001), ‘National Human Development Report’,

Government of India downloaded from http://planningcommission.
nic.in/reports/genrep/index.php?repts=nhdcont.htm
OECD (2008), ‘Handbook on Constructing Composite Indicators Methodology

and User Guide’, Organization For Economic Co-operation and Development.
12.11 ANSWER OR HINTS TO CHECK YOUR

PROGRESS EXERCISES
1) See Section 12.2
2) See Section 12.2
3) See Section 12.3
4) See Section 12.4

1) See the heading ‘conducting PCA and analyzing result’ under Section 12.6
2) See Sub-section 12.6.1 (Under the head conducting the PCA and analyzing
results)
3) See Sub-section 12.6.1 (Under the head final PCA index value)
5) See Section 12.7
118
MEC-109
Research Methods
Block
4
QUANTITATIVE METHODS-II
UNIT 13
Multivariate Analysis: Factor Analysis 5
UNIT 14
Canonical Correlation Analysis 31
UNIT 15
Cluster Analysis 43
UNIT 16
Correspondence Analysis 59
UNIT 17
Structural Equation Modeling 75
Expert Committee
Varanasi New Delhi
Bombay University and Former IGNOU
Chairman, ICSSR, New Delhi New Delhi
Faculty of Social Sciences IGNOU
Banaras Hindu University, Varanasi New Delhi
Former Joint Advisor Associate Professor in Economics
Niti Ayoug, New Delhi IGNOU, New Delhi
(Format and Language Editing)
13 Dr. Darvinder Kumar (Sr. Scale) Prof. Narayan Prasad
Assistant Professor in Statistics Professor of Economics
PGDAV College (University of Delhi), Delhi IGNOU, New Delhi Block Editor
14 Vivek N. Sharma Prof. Narayan Prasad Prof. G.K. Shukla
Assistant Professor in Mathematics IGNOU, New Delhi Rtd. Prof. of Statistics
SGTB Khalsa College, (Univ. of Delhi), Delhi IIT, Kanpur
Ms. Neha
Research Assist., IGNOU, New Delhi
15 Dr. Nausheen Nizami Prof. Narayan Prasad
Assistant Professor in Economics Professor of Economics
Gargi College (University of Delhi), Delhi IGNOU, New Delhi
Ms. Neha Bailwal
Research Assistant, IGNOU, New Delhi
16 Mr. Vivek N. Sharma Shri B.S. Bagla
Assistant Professor in Mathematics PGDAV College
SGTB Khalsa College (Univ. of Delhi) (University of Delhi)
17 Prof. Hari Shankar Asthana Prof. Narayan Prasad
Professor of Psychology, BHU, Varanasi Professor of Economics, IGNOU, New Delhi
Print Production Secretarial Assistance

Mr. Manjit Singh Shri Vinay Kumar
Section Officer (Pub.), SOSS, IGNOU, New Delhi
December, 2015
ISBN-978-93-85911-28-6
Printed at :
BLOCK 4 QUANTITATIVE METHODS-II
Many a times a social scientist or a policy maker is required to identify the
factors causing variation in any social or economic issue of policy importance
like food security, child deprivation etc. in order to take corrective policy
measures. To handle such issues, factor analysis technique, in case of interval
scale or ratio scale data and correspondence analysis method in nominal or
categorized variables are used. Another Multi-Variate statistical model namely
canonical correlation analysis is used to examine the interrelationships among
set of multiple dependent variables and multiple independent variables. Cluster
analysis enables us to classify the units or observations into sub-groups or
clusters based on characteristics or information contained in the variables.
Structural Equation Modeling (SEM) attempts to examine the causal relations
by including directly observed variables and latent variables. Keeping in view
the widening role of interdisciplinary nature of economics, the mixed methods
research is gaining popularity among researchers. All these five Multi-Variate
statistical techniques are increasingly being used in Mixed Methods Research
and hence have been covered in this block.
Unit 13 on ‘Multivariate Analysis: Factor Analysis’ throws light on various
concepts used in factor analysis, algorithm involved in factor analysis and the
process to interpret the results of factor analysis.
Unit 14 on ‘Canonical Correlation Analysis’ deals with the procedure to
compute the canonical correlation, its application in social sciences, interpretation
of its results and limitation.
Unit 15 entitled ‘Cluster Analysis’ discusses the concept and purpose of
cluster analysis, the various steps and algorithm in computation of its results
and the different approaches followed in it.
Unit 16 on ‘Correspondence Analysis’ covers the concept and special
features of correspondence analysis, its computation algorithm, interpretation
of its results and the process of application of this technique in social sciences.
Unit 17 entitled ‘Structural Equation Modeling (SEM)’ throws light on the
meaning and importance of SEM, its various models, steps involved in it, its
advantages and disadvantages.
-II Multivariate Analysis:
UNIT 13 MULTIVARIATE ANALYSIS: Factor Analysis
FACTOR ANALYSIS
Structure
13.0 Objectives
13.1 Introduction
13.2 Factor Analysis: Concept and Meaning
13.3 Historical Background of Factor Analysis
13.4 The Orthogonal Factor Model
13.4.1 Notations and Terminology
13.4.2 The Factor Model
13.4.3 Model Assumptions
13.5 Communalities
13.6 Methods of Estimation
13.6.1 Principal Component Method
13.6.2 Maximum Likelihood Method
13.7 Factor Rotation
13.8 Oblique Rotation
13.9 Factor Scores
13.10 Methods for Estimation of Factor Scores
13.10.1 Ordinary Least Square Method
13.10.2 Weighted Least Square Method
13.10.3 Regression Method
13.11 Let Us Sum Up
13.12 Key Words
13.15 Exercises
13.0 OBJECTIVES
• explain the terms used in factor analysis model like factor loadings, specific
variances, communalities, factor rotation and oblique rotation;
• know the various uses of factor analysis model;
• elucidate how to apply principle component analysis and maximum
likelihood methods for estimating the parameters of a factor model;
• discuss methods for estimating factor score; and
• learn how to interpret the results of factor analysis based on rotated factor
loadings, rotated eigenvalues and scree test.
5
Quantitative Methods-II
13.1 INTRODUCTION M
We have seen in Unit 12 that how the principal component analysis (PCA) is a
useful technique to reduce the indicators or data for construction of composite
index and identify the variables for policy decisions. Factor analysis is a further
extension of principal component analysis. Many a times, a researcher in social
sciences is confronted with entangled human behaviour, unknown
interdependencies and masses of qualitative and quantitative variables. In such
situations, he may intend to uncover major social, institutional and
international pattern. Factor analysis enables him to manage over a hundred
variables, explore a content area, structure a domain, map unknown concepts,
classify or reduce data, illuminate causal nexus, define relationships, test
hypotheses, formulate theories and make inferences. In short, a social scientist
can discern the regularity and order in a phenomena under investigation by
applying factor analysis. Factor analysis takes thousands of measurements and
qualitative observations and resolves them into distinct patterns of occurrence.
Factor analysis as a constituent of multi-variate analysis involves a large
number of issues ranging from conceptual, application, modeling, inference
and interpretation, factor rotation etc. Keeping in view, the limited scope of this
unit, we shall confine here to explain the concepts used in factor analysis,
methods of estimation in factor model and the techniques of extraction such as
principal component method and maximum likelihood method. Let us begin
with explaining the concept of factor analysis.
13.2 FACTOR ANALYSIS: CONCEPT AND

MEANING
The term factor analysis refers to anyone of a number of similar but distinct
multi-variate statistical models that model observed variables as linear
functions of a set of latent or hypothetical variables (also known as factors) not
directly observed.
Factor analysis models are similar to regression models in the sense that both
types of models possess dependent variables as linear functions of independent
variables. However, they are distinct to each other in the sense that in factor
analysis models, the independent variables are not observed independently of
the observed dependent variables.
The factor variables contained in factor analysis models may be determinate or
indeterminate. The determinate models encompass the various component
analysis models such as Principal Components Analysis, weighted principal
components and Gultman’s image analysis. Indeterminate models are
represented by the common factor model that seeks to account for co-variation
between the observed variables and a set of common factor variables.
Determinate factor analysis are more useful in data reduction role by finding a
smaller number of variables that capture most of the information of variation
and co-variation among the observed variables. In contrast, the common factors
in the indeterminate models are indeterminate from the observed variables and
are not linear combinations of them. In this unit, we shall confine our
discussion to determinate factor analysis.
Factor analysis is based on a model in which the observed vector is partitioned
into an unobserved systematic part and an unobserved error part. The
6
-II components of the error vector are considered as uncorrelated or independent, Multivariate Analysis:
while the systematic part is taken as a linear combination of a relatively small Factor Analysis
number of unobserved factor variables. The analysis separates the effects of the
factors, which are of the basic interest from the errors. From another point of
view the analysis gives a description or explanation of the interdependence of a
set of variables in terms of the factors without regard to the observed variability.
13.3 HISTORICAL BACKGROUND OF FACTOR

ANALYSIS
The early development of factor analysis was due to Charles Spearman in the
year 1904. He published a paper that is thought to be where factor analysis
originated. Spearman was working with examination scores of school children
and he was able to notice certain systematic effects in the matrix of correlation
between the scores the children made on different tests. For example, in one
case he obtained the following matrix of correlations for school children in a
preparatory school for their scores on tests in Classics (C), French (F), English
(E), Mathematics (M), Discrimination of pitch (D), and Music (Mu) as follows:
C F E M D Mu
C 1.00 0.83 0.78 0.70 0.66 0.63
F 0.83 1.00 0.67 0.67 0.65 0.57

E 0.78 0.67 1.00 0.64 0.54 0.51
M 0.70 0.67 0.64 1.00 0.45 0.51
D 0.66 0.65 0.54 0.45 1.00 0.40
Mu 0.63 0.57 0.51 0.51 0.40 1.00
It can be noted out that this matrix has the interesting property that any two
rows are almost proportional if the diagonals are ignored. Thus for rows C and
0.83 0.70 0.66 0.63

E there are ratios:
1.2
0.67 0.64 0.54 0.51
X l F ε
Spearman proposed the idea that the six test scores are all of the form
Where X is the ith standardized score with a mean of zero and a standard
deviation of one, l is a constant, F is a factor value which has mean zero and
standard deviation one for individuals as a whole and ε is the part of X that is
specific to the ith test only. On the basis of his work Spearman formulated his
two-factor theory of mental tests i.e each test result is made up of two parts,
one that is common to all tests is general intelligence and another that is
specific to the test. Later on this theory was modified to allow for each test
result to consist of a part due to several common factors plus a part specific to
X l F l F l! F! ε

test. This gives the general factor analysis model
13.4 THE ORTHOGONAL FACTOR MODEL

In this section, first we will define some notations and terminology along with
the underlying assumptions.
7
Quantitative Methods-II 13.4.1 Notations and Terminology M
Multivariate population and sample can be defined by mean vector µ and

covariance matrix Σ. These are defined as follows. Suppose that there are p
variables X1, X2,……Xp then the vector of sample mean is given by
X

X X
"
X
#
This can be thought of as the centre of the sample. It is an unbiased estimate of
the population vector of means
µ
µ $"%
µ
µ#
S S S!

The matrix of the sample variance and covariances is
S S S!

" " "
S# S# S#!
S
where S is called the sample covariance matrix or sometimes sample dispersion

matrix. It is an unbiased estimate of the population covariance matrix Σ.
Consider m unobserved common factors F1, F2,……Fm. The ith common factor
is Fi. Generally, m is substantially less than p.
F
The common factors are also formed into a vector as
F
F $ %
"
F!
13.4.2 The Factor Model

The factor model can be observed as a series of multiple regressions, predicting
each of the observable variable Xi from the unobservable common factors Fi. In
general the factor analysis model is given by
X µ l F l F l! F! ε

X µ l F l F l! F! ε
"
X# µ# l# F l# F l#! F! ε#
Here, the means µ , µ , … . . , µ# can be regarded as the intercept for the multiple
regression models. The regression coefficients l) (slopes) for all the multiple
regressions are called factor loadings. Here, l) is the loading of the ith variable
on the jth factor. These will be shown in the matrix form as follows:
8
-II Matrix of the factor loading Multivariate Analysis:
l l l!

Factor Analysis
l l l!

" " "
L
l# l# l#!
and finally the errors ε are called the specific factors. Here, ε is the specific
factor for variable i. The specific factors are also shown into a vector form as
Vector of specific factors

ε
ε$"%
ε
ε#
In general, the basic factor analysis model is like a regression analysis model.
Each of our response variable X is expressed as a linear function of the
unobserved common factors F1, F2,….Fm. Therefore, here one can say that we
have m unobserved factors that control the variation among the data.
XµLFε
We will write the above factor model in matrix notation as
(13.1)
In general we want m < p
13.4.3 Model Assumptions

Some assumptions are necessary to uniquely estimate the parameters.
Otherwise, an infinite number of equally well fitted models with different
values for the parameters may be obtained if these assumptions are not made.
Therefore, the following assumptions are made about the unobserved random
vectors F and ε and its certain covariance relationships:
E,ε - 0; i 1, 2, … . . , p
i) The specific factors or random errors all have mean zero i.e.,
ii) The common factors also have mean zero i.e.,

E1F) 2 0; j 1, 2, … . . , m
iii) The variance of specific factor i is Ψ , where Ψ is called the specific
variance, and they are independent, i.e. 567 1ε , ε′ 2 0. for 8 9 8′.
0 0

ψ
0 0
Cov ,ε - E1ε , ε′ 2 Ψ ; i 1, 2, … . . , p
ψ
" " "

0 0 ψ#
iv) The common factors F) have variance one and are uncorrelated, i.e.
<1F) 2 Cov 1F) , F)′ 2 0 for = 9 = ′ .
Thus co-variance matrix of F is an identify matrix I, i.e. cov (F) = I
9
Quantitative Methods-II v) Since the common factors and specific factors are independent, then the M
specific factors are uncorrelated with the common factors i.e
Cov 1ε , F) 2 E1ε , F)′ 2 0; i 1, 2, … . . , p; j 1, 2, … . . , m
Now, the factor analysis model implies a covariance structure for the random
,X > µ- ,X > µ- ,L F ε- ,L F ε-
variable X is given by equation
,L F ε- ,,L F- ε-
(13.1)
L F,L F- ε,L F- L F ε εε
Σ Cov ,X- E,X > μ- ,X > μ-

so that
L E,F F-L E,ε F-L LE,F ε- E,εε-

? ΣLLΨ (13.2)
This is the matrix of factor loadings, times its transpose plus a diagonal matrix
containing the specific variances. Hence a simplified model for the covariance
Also, on multiplying F in equation (13.1) on both sides, we have

matrix is obtained which is used for estimation.
,X > µ-F ,L F ε-F

LFFεF
so, Cov ,X, F- E,X > µ-F LE,F F- E,εF- L
or Cov 1X , F) 2 l)
Under this model, the variance for the ith observed variable is equal to the sum
of the squared loadings for that variables and specific variance. The variance of
variables Xi is:
σ ∑!
)A l) ψ

(13.3)
quantity ∑!
)A l) is called the communality for variable i.

This is a derivation which is based on the previous assumptions and the
1) The model X > µ L F ε assumes that the data is a linear function of

Important Remarks:
the common factors. However, since the common factors are not
observable, we cannot check the linearity of the model.
2) The covariance matrix is a symmetric matrix, that is the variance between
variables i and j is the same as the variance between j and i. For this
model:
Σ L L′ Ψ
The covariance matrix is going to have p(p+1)/2 unique elements of Σ
which are approximated by mp factor loadings in the matrix L and the p
specific variances ψ . When m= p, any covariance matrix Σ can be
reproduced exactly as L L′, so Ψ can be the zero matrix. However, when
m is smaller relative to p, that factor analysis is most useful. In this case,
10
-II the factor model provides a simple explanation of the covariation in X Multivariate Analysis:
with fewer parameters than the p(p+1)/2 parameters in Σ. Factor Analysis
For example, if X contains p=12 variables and the factor model with m=
terms of the mp+ p= 36 parameters l) and ψ of the factor model.

2 is appropriate, then the p(p+1)/2= 78 elements of Σ are described in
factored as L L′ Ψ, where the number of factors m is much less than p.

Unfortunately, for the factor analyst, most covariance matrices cannot be
consistent solution for the parameters l) and ψ from the variances and
There are some problems, where it is not possible to get a unique
covariances of the observable variables.

3) If m > 1, there is always some inherent ambiguity associated with the
factor model. To explain that, let T be any m × m orthogonal matrix. A
matrix is orthogonal if its inverse is equal to the transpose of the original
TT ′ T ′ T I
matrix.
Hence, we can write the factor model in matrix notation as:

X > µ L F ε L T T ′F ε LG F G ε
Note that this does not change the calculations since the identity matrix
times any matrix just gives back the original matrix. This results in an
alternative factor model, where the relationship between the new factor
LG L T
loadings and the original factor loadings is:
and the relationship between the new common factors and the original
FG T′F
common factor is:
This gives a model that fits equally well. Moreover, since there are
infinite numbers of orthogonal matrices, then there are an infinite number
of alternative models. This model, as it turns out, satisfied all the
E , F G - E , T ′ F- T ′ E , F- 0
assumptions that we discussed earlier as:
Cov ,F G - Cov ,T ′F- T ′Cov,F-T T ′I T T ′T I
Cov ,F G , ε- Cov ,T ′F, ε- T ′ Cov,F, ε- T ′0 0

and
loadings L from the loadings LG . That is, the factors F and F G T ′F have
On the basis of observations on X, it is impossible to distinguish the
the same statistical properties, and even though the loadings LG are, in
general different from the loadings L, they both have the same covariance
Σ L L′ Ψ L T T ′L′ Ψ ,LG - ,LG -′ Ψ

matrix Σ. That is
(13.4)
This ambiguity is going to be used to justify the factor rotation, since
orthogonal matrices correspond to rotation of the coordinate system for
X. We will use later to obtain more thrifty description of the data.
11
13.5 COMMUNALITIES M
The communalities are computed by taking the sum of squares of the loadings
for the ith variable on the m common factors. This can be shown as below
variance. It is denoted by h , we have from equation (13.2)

Variance due to the specific factor is often called the uniqueness or specific
σ l l l! Ψ
or σ h Ψ ; i = 1, 2,…….,p (13.5)
here, h l l l!
Now, to understand the computation of communalities in a simple way, let us

have the table of three factor loadings. The following dataset involves the
Places Rated Almanac (Boyer and Savageau) rates 329 communities according
to nine criteria:
These data are only recorded for the first three factors only. It should also note
that these factor loadings are the correlations between the factors and the
variables. For example, the correlation between the Arts and the first factor is
about 86%. Similarly the correlation between climate and that factor is only
about 28%.
Factor
Variable 1 2 3
Climate 0.286 0.076 0.841
Housing 0.698 0.153 0.084
Health 0.744 -0.410 -0.020
Crime 0.471 0.522 0.135
Transportation 0.681 -0.156 -0.148
Education 0.498 -0.498 -0.253
Arts 0.861 -0.115 0.011
Recreation 0.642 0.322 0.044
Economics 0.298 0.595 -0.533
For example, to compute the communality for climate, the first variable which
is given in the above table, we square the factor loadings for climate and then
h 0.286 0.076 0.841 0.795

add the results:
In the similar way we can obtain the communalities for all the 9 variables and
the total communality is obtained as 5.617. Therefore, all the communalities
computed are placed into a table shown below:
12
-II Variable Communality Multivariate Analysis:
Factor Analysis
Climate 0.795
Housing 0.518
Health 0.722
Crime 0.512
Transportation 0.510
Education 0.561
Arts 0.754
Recreation 0.517
Economics 0.728
Total 5.617
These values are as similar as multiple R2 values for regression models

predicting the variables of interest from the 3 factors. The communality for a
given variable can be interpreted as the proportion of variation in that variable
explained by the three factors. In other words, if we perform multiple
regression of climate against the three common factors, we obtain an R2=
0.795, indicating that about 79% of the variation in climate is explained by the
factor model. The results suggest that the factor analysis does the best job of
explaining variation in climate, the arts, economics, and health.
One assessment of how well this model is doing can be obtained from the
communalities. What you want to see is values that are close to one. This
would indicate that the model explains most of the variation for those variables.
In this case, the model does better for some variables than it does for others.
The model explains Climate the best, and is not bad for other variables such as
Economics, Health and the Arts. However, for other variables such as Crime,
Recreation, Transportation and Housing the model does not do a good job,
explaining only about half of the variation. If you take all of the communality
values and add them up you can get a total communality value:
# !
J h J λ
A A
Here, the total communality is 5.617. The proportion of the total variation
5.617
explained by the three factors is
0.624
9
This gives us the percentage of variation explained in our model. This might be
looked at as an overall assessment of the performance of the model. However,
this percentage is the same as the proportion of variation explained by the first
three eigenvalues, obtained earlier. The individual communalities tell how well
the model is working for the individual variables, and the total communality
gives an overall assessment of performance. These are two different
assessments that can be used.
Since the data are standardized in this case, the variance for standardized data
is going to be equal to one. Then the specific variances can be computed by
subtracting the communality from the variance as expressed below:
13
Quantitative Methods-II Ψ 1 > h M
Recall, that the data were standardized before analysis, so the variances of the
standardized variables are all equal to one. For example, the specific variance
Ψ 1 > 0.795 0.205

for Climate is computed as follows:
The specific variance for housing is Ψ 0.482 and the other values of specific
variances can be calculated accordingly.
Source: http://onlinecourses.science.psu.edu
EXAMPLE 1 Calculating Correlations from Factors
Factor analysis aims to explain the correlations among the observed variables
in terms of small number of factors. To estimate the success of factor solution,
we reproduce the original correlation matrix by using the loadings on the
common factors and see how large a discrepancy is there between the original
matrix and reproduced correlation matrix. The greater the discrepancy, the less
successful the factor solution in preserving the information in the original
correlation matrix. And, lesser the discrepancy, more successful the factor
solution in preserving the information in original correlation matrix.
In order to calculate the correlations from the factor solution the process is
two variables K and K is obtained by adding products of coefficients for these

quite simple when the factors are uncorrelated. The correlation between any
variables (K and K ) across all the common factors. Let us say for the three
factor solution L , L and LM the quantity would be ,N O N - ,N O N -
,NM O NM -. This process will become clearer in the description of the
hypothetical two-factor solution based on five observed variables as follows:
A Hypothetical Solution
Variables Loadings/ Correlations Communality Reproduced Correlations
Factor 1 Factor 2 X1 X2 X3 X4 X5
X1 .5 .2 .29 X1
X2 .6 .3 .45 X2 .36
X3 .7 .4 .65 X3 .43 .54
X4 .3 .6 .45 X4 .27 .36 .45
X5 .4 .7 .65 X5 .34 .45 .56 .54
∑ x = 1.35 ∑ x 1.14
The Coefficients: According to the above solution,

X1 = 0.5F1 + 0.2F2
"
X2 = 0.6F1 + 0.3F2
X5 = 0.4F1 + 0.7F2
14
-II As well as being weights, the coefficients above show that the correlation Multivariate Analysis:
between X1 and F1 is .50, and that of between X1 and F2 is .20, and so on. Factor Analysis
Variance Accounted For: The quantities at the bottom of each factor column
are the sums of the squared loadings for that factor, and show how much of the
total variance of the observed variables is accounted for by that factor. For
Factor 1, the quantity is .52 + .62 + .72 + .32 + .42 = 1.35. Because in the factor
analyses discussed here, the total amount of variance is equal to the number of
observed variables (the variables are standardized, so each has a variance of
one), the total variation here is five, so that Factor 1 accounts for (1.35/5)×100
= 27% of the variance.
The quantities in the communality column show the proportion of the variance
of each variable accounted for by the common factors. For X1 this quantity is
.52 + .22 = 0.29, for X2 it is .62 + .32 = 0.45, and so on.
Reproducing the Correlations: The correlation between variables K and K
the correlation between variables KM and KQ is equal to (.7 X .4) + (.4 X .7) =
as derived from the factor solution is equal to (.7 X .8) + (.2 X .3) = 0.36, while
.56. These values are shown in right-hand side of the above table.
EXAMPLE 2 To verify the relation Σ L L′ Ψ for two factors:
Consider the following covariance matrix
19 30 2 12
30 57 5 23
Σ$ %
2 5 38 47
12 23 47 68
19 30 2 12 4 1 2 0 0 0
The equality equation of the model is
30 57 5 23 7 2 4 7 >1 1 0 4 0 0
$ %$ %R S$ %
2 5 38 47 >1 6 1 2 6 8 0 0 1 0
12 23 47 68 1 8 0 0 0 3
Σ L L′ Ψ
or
may be verified by the matrix algebra. Therefore, Σ has the structure produced
by an m= 2 orthogonal factor model. Since
l l 4 1
l l 7 2
L$ %$ %
lM lM >1 6
lT lT 1 8
0 0 0 2 0 0 0

ψ
0 0 0 0 4 0 0
Ψ $ %
ψ
0 0 0 0 0 1 0

ψM
0 0 0 ψT 0 0 0 3
The communality of Xi can be obtained by using equation (13.5), we have
15
Quantitative Methods-II h l

l

4 1 17 M
σ ,l l - ψ h ψ

and the variance of Xi can be decomposed as

or
19 = 42 + 12 + 2 = 17 + 2
i.e. variance = communality+ specific variance
Similarly, we can find out the results for the other variables also.
1) What do you mean by a factor in the context of factor analysis?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
1.0 . 63 . 45
2) Show that the covariance matrix
Σ U. 63 1.0 . 35V
. 45 . 35 1.0
For the p=3 standardized random variables Z1, Z2 and Z3 can be generated
Z .9F ε
by the m=1 factor model:
Z .7F ε
ZM .5F εM
where, Var ,F - 1, Cov ,ε, F - 0
. 19 0 0
and Ψ Cov ,ε- U 0 . 51 0 V
0 0 . 75
Formulate the factor model for the given problem and also calculate
(a) Communalities h ; i= 1, 2, 3 and interpret the result.
(b) Cor (Zi, Fi) for i= 1, 2, 3. Which variable carry the greatest weight in
naming the common factor? And why?
……………………………………………………………………….......
……………………………………………………………………….......
……………………………………………………………………….......
……………………………………………………………………….......
σ l ψ
3) Let the factor model with p=2 and m=1. Show that

σ σ l l

σ l ψ
16
-II for given σ , σ and σ , there is an infinite choices for L and Ψ i.e., Multivariate Analysis:
the factor model need not be unique. Factor Analysis
……………………………………………………………………….......
……………………………………………………………………….......
……………………………………………………………………….......
……………………………………………………………………….......
……………………………………………………………………….......
1 .4 .9
4) Consider an m=1 factor model for the population with covariance matrix
Σ U. 4 1 . 7 V
.9 .7 1
Show that there is a unique choice of L and Ψ with Σ L L′ Ψ, but that
ψM \ 0, so the choice is not admissible i.e., the solution is unique but
improper.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
13.6 METHODS OF ESTIMATION

For the observations x1, x2,……,xn on p generally correlated variable, does the
factor model of equation (13.1) with a small number of factors, effectively
represent the data? The solution of such kind of statistical problems may be
tackled by using the covariance relationships. The sample covariance matrix S,
is an unbiased estimate of the unknown population covariance matrix Σ. If Σ
appears to deviate significantly from a diagonal matrix, then a factor model can
l) and specific variances ψ .

be entertained, and the initial problem is one of estimating the factor loadings
In this section, we shall consider two of the most popular methods of parameter
estimation, the principal component (and the related principal factor) method
and the maximum likelihood method. The solution from either method can be
rotated in order to simplify the interpretation of factors.
13.6.1 Principal Component Method

A vector of observations for the Xi variable may be defined as
X
X
X $ %
"
X#
The sample covariance matrix is denoted by S and is expressed as

17
1
^
S ] - ,X > X
J,X > X ] -′
n>1
Quantitative Methods-II M
A
We have p eigenvalues for the covariance matrix S as well as corresponding
λ_ , λ_ , … … . λ_ # and eigenvectors of the covariance matrix S are è , è , … … . è# .
eigenvectors for this matrix. Let the eigenvalues of the covariance matrix S are
The covariance matrix can be re-expressed in the following form as a function
Spectral Decomposition of a: The spectral decomposition allows us to express

of the eigenvalues and the eigenvectors:
the inverse of a square matrix in terms of its eigenvalues and eigenvectors, and
Σ have eigenvalues and eigenvectors pairs ,λ , e - with λ c λ c c λ# c 0,

this leads to a useful square-root matrix. Let the population covariance matrix
then
#
Σ λ e e′ λ e e′ λ# e# e′# J λ e e′

A
eλ e′

eλ e′
deλ e " eλ e " " eλ# e# f
"
(13.6)

eλ# e#
′
having as many factors as variables i.e., m= p and specific variances Ψ 0 for

This fits the prescribed covariance structure for the factor analysis model
all i. The loading matrix has jth column given by eλ) e) . Then we can write
Σ LL0 LL (13.7)
Apart from the scale factor eλ) , the factor loadings on the jth factor are the
coefficients for the jth principal component of the population. The factor
analysis representation of Σ in equation (13.7) is exact, it is not particularly
useful. It employs as many common factors as there are variables and does not
allow any variation in the specific factors ε in the main model of equation
just a few common factors. One approach when the last p > m eigenvalues are
(13.1). But we prefer models that explain the covariance structure in terms of
small is to neglect the contribution of λ!g e!g e′!g λ# e# e′# to Σ in

eq (13.6), we obtain the approximation
eλ e′

eλ e′
Σ deλ e " eλ e " " eλ! e! f LL
"
(13.8)

eλ! e′!
18
-II The approximation shown in eq (13.8) assumes that the specific factors ε in eq Multivariate Analysis:
(13.1) are of minor importance and can also be ignored in the factoring of Σ. Factor Analysis
Allowing for specific factors, we find that the approximation becomes

Σ L L′ Ψ
eλ e′
0 0
eλ e′
ψ
0 0
deλ e " eλ e " " eλ! e! f
"
ψ
" " h
"
0 0
(13.9)
ψ#

eλ! e′!
ψ σ > ∑!
where
)A l) ; i 1, 2, … … . , p

This yields the following estimator for the factor loadings
li) è) jλ_
When the units of the variables are not appropriate, it is better to use the
standardized variables. If the standardized measurements are used, we replace
sample correlation matrix S by R. This in turn suggests that the specific
variances, which are the diagonal elements of the matrix Ψ, can be estimated
k S > ∑! i
by using the expression:
ψ )A l) ; i 1, 2, … … . , p (13.10)
Here, the estimated specific variances are equal to the sample variance for the
ith variable minus the sum of the squares of factor loadings (i.e the
communality).
The equation (13.10), when applied to the sample covariance matrix S or the
sample correlation matrix R, is known as the principal component solution.
13.6.2 Maximum Likelihood Method

Maximum likelihood estimation requires that the data are sampled from a
multivariate normal distribution. But if the data have been collected on a likert
scale, which is most often the case in the social sciences, these kinds of data
cannot really be normally distributed. This is the main drawback of this
method. Therefore, using the maximum likelihood estimation method we must
assume that the data are independently sampled from a multivariate normal
distribution with mean vector µ and the covariance matrix Σ that takes the
Σ L L′ Ψ
particular form
where L is the matrix of factor loadings and Ψ is the diagonal matrix of

specific variances.
sample which is drawn from a multivariate normal population N# ,µ, Σ-, then
Again, let us assume that the p×1 vectors X1, X2,……Xn represents a random
the joint probability function or the likelihood function is given by
19
1
^
l,µ, Σ- m n eq1rsqµ2 Σ w
′ tu
1rs qµ2v
,2π-#⁄ |Σ|⁄
)A
,π-xy⁄z|Σ|x⁄z eq ∑s{u1rs qµ2 Σ

x ′ tu
1rsqµ2v
(13.11)
Maximum likelihood estimation involves estimating the mean, the matrix of

factor loadings and the specific variances.
L and the specific variances Ψ are obtained by finding the values of µ̀, L_ and Ψ
|
The maximum likelihood estimator for the mean vector µ , the factor loadings
that maximizes by taking logarithm of the likelihood function in equation
np
(13.11) which is given by the following expression:
Log l,µ, L, Ψ- > log 2π

2
n 1
^
> log~LL′ Ψ~ > J1x) > µ2 ,LL′ Ψ > 1- 1x) > µ2

2 2
′
)A
(13.12)
find the values of the population parameters µ, L and Ψ. As mentioned earlier,

The log of the likelihood function of the data is to be maximized. We want to
the solution for these factor models are not unique. Equivalent models can be
imposed is that L′ Ψq L is a diagonal matrix.

obtained by rotation. To obtain a unique solution an additional constraint to be
Computationally this process is complex. In general, there is no closed- form

solution to this maximization problem, so iterative methods must be applied.
Source: Adapted from Applied Multivariate Statistical analysis by the Johnson R.A. and
Wichern D.W. (2002).
13.7 FACTOR ROTATION

Broadly speaking, there are two kinds of rotations i.e. orthogonal rotation and
oblique rotation. Orthogonal rotation assumes that the factors are uncorrelated.
This is less realistic since factors generally are correlated with each other to
some extent or the other. Two commonly used orthogonal rotation techniques
are Quartimax and Varimax rotation. Quartimax involves the minimization of
the number of factors needed to explain each variable whereas Varimax
minimizes the number of variables that have high loadings on each factor and
works to make small loadings even smaller.
All factor loadings obtained from the initial loadings by an orthogonal
transformation have the same ability to reproduce the covariance matrix. From
matrix algebra, we know that an orthogonal transformation corresponds to a
rigid rotation of the coordinate axis. For this reason, an orthogonal
If L_ is the p×m matrix of estimated factor loadings obtained by any one of the
transformation of the factors is called factor rotation.
|LG L_ T where TT ′ T ′T I
method i.e., principal component or maximum likelihood, then
(13.13)
is a p×m matrix of rotated loadings. Moreover, the estimated covariance matrix
L_L_′ Ψ
| L_TT ′L_ Ψ
| L_G L_G′ Ψ
|
remains unchanged, since
(13.14)
20
Equation (13.14) shows that the residual matrix, S^ > L_L_′ > Ψ
| S^ > L_G L_G′ >
| , remains unchanged. Moreover, the specific variances Ψ | and hence the
communalities h_ are unaltered. Thus from a mathematical viewpoint, it is

Factor Analysis

Ψ
immaterial whether L_ or L_G is obtained.

The various factor rotation methods have, as a guiding principle, the simple
structure concepts. That is to say, the results after rotation should become
simple in their appearance. To put it another way, these simple structure
concepts should be considered when trying to determine whether or not a given
factor rotation has clarified the underlying structure of the data.
Factors are rotated for better interpretation since un-rotated factors are
ambiguous. The goal of rotation is to attain an optimal simple structure which
attempts to have each variable load on as few factors as possible, but
maximizes the number of high loadings on each variable. Ultimately, the
simple structure attempts to have each factor define a distinct cluster of
interrelated variables so that interpretation becomes easier. For example,
variables that relate to language should load highly on language ability factors
but should have close to zero loadings on mathematical ability.
Varimax Rotation: Perhaps the most widely used is the varimax criterion. It
seeks the rotated loadings that maximize the variance of the squared loadings
for each factor; the goal is to make some of these loadings as large as possible,
and the rest as small as possible in absolute value. The varimax method
encourages the detection of factors each of which is related to few variables. It
discourages the detection of factors influencing all variables. The quartimax
criterion, on the other hand, seeks to maximize the variance of the squared
loadings for each variable, and tends to produce factors with high loadings for
all variables.
the varimax (normal varimax) criteria. Let us define lGT iG _

) l) ⁄h to be the
Kaiser [5] has suggested an analytical measure of simple structure known as
rotated coefficients scaled by the square root of the communalities. Then the
(normal) varimax procedure selects the orthogonal transformation T that makes
# GT # G
V # ∑!
)A R∑A l) > 1∑A l) 2 vpS

(13.15)
as large as possible.
Scaling the rotated coefficients ilG) has the effect of giving variables with small
After the transformation T is determined, the loadings lG) are multiplied by h_ so
communalities relatively more weight in the determination of simple structure.
that the original communalities are preserved.

In the above equation, it has a simple interpretation. In words
!
V J1variance of the squares of ,scaled-loadings for j factor2

)A
One can maximize V that corresponds to the squares of the loadings on each
negligible coefficients in any column of the rotated loadings matrix L_G .

factor as much as possible. Therefore, we hope to find groups of large and
21
13.8 OBLIQUE ROTATION M
Orthogonal rotations are appropriate for a factor model in which the common
factors are assumed to be independent. Many investigators in social sciences
consider oblique (non-orthogonal) rotations, as well as orthogonal rotations.
The former rotations are often suggested after one view of the estimated factor
loadings that do not follow the assumption made in the factor model.
Nevertheless, an oblique rotation is frequently a useful aid in factor analysis.
In other words, oblique rotation is suitable when the factors are considered to
be correlated. Oblique rotation is more complex than orthogonal rotation, since
it can involve one of two coordinate systems, a system of primary axes or a
system of reference axes. Additionally, oblique rotation produces a pattern
matrix that contains the factor or item loadings and factor correlation matrix
that includes the correlations between the factors. The commonly used oblique
rotation techniques are Direct Oblimin and Promax. Direct Oblimin attempts to
simplify the structure and the mathematics of the output of the problem under
consideration, while Promax is used because of its speed in larger datasets.
Promax involves raising the loadings to a power of four which ultimately
results in greater correlations among the factors and achieves a simple
structure.
13.9 FACTOR SCORES

A factor score can be considered to be a variable describing how much an
individual would score on a factor. One of the methods to produce factor score
is called Bartlett method (or regression approach) which produces unbiased
scores that are correlated only with their own factor. Another method is called
the Anderson-Rubin method which produces scores that are uncorrelated and
standardized. In factor analysis, interest is usually centered on the parameters
in the factor model. However the estimated value of the common factors, called
factor scores, may also be required. These quantities are often used for
diagnostic purposes, as well as inputs to a subsequent analysis.
Factor scores are not estimates of unknown parameters in the usual sense.
Rather, they are estimates of values for the unobserved random factors Fj, j = 1,
2,……,m. We try to find out the vectors of m unobserved common factors that
X>µ LFε
underlie our model to estimate those factors. Therefore, for the factor model
We may wish to estimate the vectors of the factor scores F1, F2,…….,Fm.
13.10 METHODS FOR ESTIMATING FACTOR

SCORE
A number of different methods are used for estimating factor scores from the
data. These include
1) Ordinary Least Square Method
2) Weighted Least Square Method
3) Regression Method
Now, the above methods are discussed below in brief.
22
-II 13.10.1 Ordinary Least Squares Method Multivariate Analysis:
Factor Analysis
If the factor loadings are estimated by the principal component method, it is
customary to generate factor scores using an unweighted (ordinary) least
squares procedure. The L’s are factor loading and the f are unobserved
common factors. The vector of common factors for subject i, is found by
minimizing the sum of the squared residuals:
# #

J ε) J X) > µ) > l) F > l) F > > l)! F! ,X > µ > L F-′ ,X > µ
)A )A
> L F-
This is like a least squares regression, except in this case we already have
estimates of the parameters (the factor loadings), but we wish to estimate the
F) ,L′L-q L′1X) > µ2

explanatory common factors. In matrix notation the solution is expressed as:
(13.16)
In practice, the parameters are unknown in nature so we take the estimates of
the factor loadings as follows:
F_) 1L_′ L_2 L_′1X) > X
]2
q
F_) 1L_′ L_ 2 L_′ z)

q
or
(13.17)
For the standardized data, since L= Reλ_ e " eλ_ e " " eλ_ ! e! S, we have
1 ′
è 1X) > X]2
eλ_
1 ′
è 1X > ]2
X
F_) eλ_ )

"
1
è! ) > X2
′ 1X ]
e _
λ!
We see that the F_) are approximately same as that of obtained in the principal
component method with the unrotated factor loadings. Here, we have for the
∑^)A F_) 0
factor scores

^
(sample mean)
∑^)A F_) F_)′ I

^q
(sample variance)
13.10.2 Weighted Least Squares Method

The difference between weighted least square and the ordinary least square is
that the squared residuals are being divided by the specific variances as shown
below. Let us suppose that the mean vector µ, the factor loading L, and the
X>µ LFε
specific variance Ψ are known for the factor model
23
Further, regard the specific factors ε′ dε , ε , … … , ε# f as errors. Since,
Var,ε - Ψ ; i= 1, 2,……,p, need not be equal. Bartlett [2] has suggested that
weighted least squares be used to estimate the common factor values.

The sum of squares of the errors, weighted by the reciprocal of their variance is
# #
ε
J J ε′Ψq ε ,X > µ > LF-′Ψq ,X > µ > LF-
Ψ
A )A
(13.18)
The solution is given by the expression where Ψ is the diagonal matrix whose
diagonal elements are equal to the specific variances.
F_ 1L′Ψq L2 L′Ψq ,X > µ-
q
We take the estimates L_, Ψ

| and µ̀ ]
X as the true values and obtain the factor
th
scores for the j case as
F_) L_′Ψ
| q L_ L_′Ψ ]2
| q 1X) > X
q
(13.19)
When L_ and Ψ
| are determined by the maximum likelihood method, these
estimates must satisfy the uniqueness condition, L_′Ψ
| q L_ ∆
| , a diagonal
matrix.
13.10.3 Regression Method

This method is used when you are calculating maximum likelihood estimates
of factor loadings. A vector of the observed data, supplemented by the vector
In factor model X > µ L F ε, we initially treat the loadings matrix L and

of factor loadings for the ith subject, is considered.
specific variance matrix Ψ is known. When the common factors F and specific
combination X > µ L F ε has an N# ,0, LL′ Ψ- distribution. Moreover, the

factors ε are jointly distributed with means and covariances, the linear
joint distribution of ,X > µ- and F is N!g# ,0, ΣG -, where
Σ LL′ Ψ " L
Σ U… … … … … …
G
" … …V
L′ " I
(13.20)
of ,F|x- is multivariate normal with

and 0 is an (m×p)×1 vector of zeros. We find that the conditional distribution
mean = E,F|x- L′Σq ,x > µ- L′,LL′ Ψ-q ,x > µ- (13.21)

and
Covariance = E,F|x- I > L′Σ L I > L′,LL′ Ψ-q L
tu
The quantities L′ ,LL′ Ψ-q are the coefficients in a multivariate regression of

(13.22)
the factors on the variables. Estimates of these coefficients produce factor

scores that are analogous to the estimates of the conditional mean values in
observations xj, and taking the maximum likelihood estimates L_ and Ψ | as the
multivariate regression analysis. Consequently, given any vector of
th
true values, we see that the j factor score vector is given by
24
-II if) L_′ |
Σ 1x) > x2 L_′1L_L_′ Ψ
q
| 2q 1x) > x2; j 1,2, … … . n
The calculation of fi) in equation (13.23) can be simplified by using the matrix
(13.23) Multivariate Analysis:
Factor Analysis
identity
L_′ 1L_L_′ Ψ
| 2 I L_′ Ψ
| q L_ L_′ Ψ
| q
q q
(13.24)
This identity allows us to compare the factor scores in equation (13.23),
generated by the regression argument, with those generated by the weighted
by F_ j and the latter by F_ j . Then by using equation ( ), we have

least squares method in equation (13.19). Temporarily, If we denote the former
R LS
F_ j L
|′ Ψ
| q1 L
| |′ Ψ
| | F_ Rj I L
|′ Ψ
| q1 L
| F_ Rj
q1 q1
I L
LS q1
L (13.25)
|′ Ψ
| q1 L
| ∆ | q1 and if the elements of
q1
For maximum likelihood estimates L
this diagonal matrix are close to zero, the regression and generalized least
squares methods will give nearly the same factor scores.
There may be made an attempt to reduce the effect of a incorrect determination
using S (the sample covariance matrix) instead of |ΣL |L | . Then we have

|′ Ψ
of the number of factors, to calculate the factor scores in equation (13.23) by
fij L
|′ Sq1 1xj > x 2;
the following results:
j 1, 2, … … . n (13.26)
fij L
|′z Rq1 zj ;
or, if a correlation matrix is factored
j 1, 2, … … . n (13.27)
|z L
|z Ψ
|z
where
zj Dq1⁄2 1xj > x 2 and ρ̀ L
′
| T are used in place of the original loadings in

|G L
eq (13.26), the subsequent factor scores fij are related to fij by
G
Again, if rotated loadings L
ifGj T′fij ; j 1, 2, … . . , n
A numerical measure between the factor scores generated from two different
calculation methods is provided by the sample correlation coefficient between
scores on the same factor.
Source: Adapted from Applied multivariate statistical analysis by Johnson R.A. and Wichern
D.W. (2002)
1) The correlation matrix for chicken- bone measurements is

. 505 1.000

1.000
. 569 . 422 1.000

. 602 . 467 . 926
. 621 . 482 . 877 . 874
1.000
. 603 . 450 . 878 . 894 . 937 1.000

1.000
The following estimated factor loadings were extracted by the maximum

likelihood procedure:
25
Quantitative Methods-II Variable Estimated Varimax rotated M
factor loadings estimated factor loadings

F1 F2 F1G FG1
Skull length .602 .200 .484 .411
Skull breadth .467 .154 .375 .319
Femur length .926 .143 .603 .717
Tibia length 1.000 .000 .519 .855
Humerus length .874 .476 .861 .499
Ulna length .894 .327 .744 .594
Using the unrotated estimated factor loadings, obtain the maximum
likelihood estimates of the following:
a) The specific variances
b) The communalities
c) The proportion of variance explained by each factor
|z L
d) The residual matrix R > L |z > Ψ
′
|z
…………………………………………………………………………
…………………………………………………………………………
…………………………………………………………………………
…………………………………………………………………………
…………………………………………………………………………
…………………………………………………………………………
1) From the above exercise 1, Compute the value of the varimax criterion
using both unrotated and rotated estimated factor loadings Also interpret
the result.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) The following table provides data of the Students entering in a certain
MBA program must take three required courses in Finance, Marketing and
Business policy. Let X1, X2, and X3, respectively, represent a student's
grades in these courses. The available data consist of the grades of five
students (in a 10-point numerical scale above the passing mark), as shown
in Table 13.1. Using an appropriate statistical package (SPSS, Minitab,
SAS etc) run a factor analysis and interpret the results.
26
Table: 13.1: Students Grade
,-
Factor Analysis
Student No.,-
Finance (X1) Marketing (X2) Business Policy
(X3)
1 3 6 5
2 7 3 3
3 10 9 8
4 3 9 7
5 10 6 5
(Example adapted from Peter Tryfos, 1997, version printed: 14-3-2001).
13.11 LET US SUM UP

Factor analysis is something like an art. Factor analysis is used to study the
patterns of relationship among many dependent variables to have an idea about
the nature of the independent variables not measured directly.
Factor analysis is used to identify latent constructs or factors. It is commonly
used to reduce variables into a smaller set factors to save time and facilitate
easier interpretations. There are many extraction techniques such as Principal
component method and Maximum Likelihood. The factor analysis and the
principal component analysis are among the oldest of the multivariate
statistical methods of data reduction. Mathematically, factor analysis is
complex and the criteria used to determine the number and significance of
factors are vast. There are two types of rotation techniques – orthogonal
rotation and oblique rotation. Orthogonal rotation (e.g., Varimax and
Quartimax) involves uncorrelated factors whereas oblique rotation (e.g., Direct
Oblimin and Promax) involves correlated factors. The interpretation of factor
analysis is based on rotated factor loadings, rotated eigenvalues, and scree test.
In reality, investigators often use more than one extraction and rotation
technique based on pragmatic reasoning rather than theoretical reasoning.
13.12 KEY WORDS

Communality : It shows the variance of an observed variable
accounted for by the common factors; in an
orthogonal factor model. It is equivalent to
the sum of the squared factor loadings.
Common Factor : It implies the unmeasured (or hypothetical)
underlying variable which is the source of
variation in at least two observed variables
under consideration.
Confirmatory Factor : This technique of the factor analysis attempts
Analysis (CFA) to confirm hypotheses and uses path analysis
diagrams to represent variables and factors.
Exploratory Factor : The Exploratory Factor Analysis technique
Analysis (EFA) tries to uncover complex patterns by
exploring the dataset and testing predictions.
27
Quantitative Methods-II Diagonal Matrix : It is a square matrix in which the entries M
outside the main diagonal are all zero.
Eigenvalue : It is a mathematical property of a matrix; used
(characteristic root) in relation to the decomposition of a
covariance matrix, both as a criterion of
determining the number of factors to extract
and a measure of variance accounted for by a
given dimension.
Eigenvector : It is a vector associated with its respective
eigenvalue; obtained in the process of initial
factoring; when these vectors are appropriately
standardized, they become factor loadings.
Factor Extraction : It is the initial stage of factor analysis in
which the covariance matrix is resolved into a
smaller number of underlying factors or
components.
Factors : It implies hypothesized, unmeasured, and
underlying variables which are presumed to
be the sources of the observed variables; often
divided into unique and common factors.
Factor Loading : It is a general term referring to a coefficient in
a factor pattern or structure matrix.
Factor Pattern Matrix : It refers to a matrix of coefficients where the
columns usually refer to common factors and
the rows to the observed variables; elements
of the matrix represent regression weights for
the common factors where an observed
variable is assumed to be a linear combination
of the factors; for an orthogonal solution, the
pattern matrix is equivalent to correlations
between factors and variables.
Kaiser criterion : In this criterion, first we can retain only
factors with eigenvalues greater than 1 when
factors are calculated from correlation matrix.
In essence this is like saying that, unless a
factor extracts at least as much as the
equivalent of one original variable, we drop it.
This criterion was proposed by Kaiser (1960),
and is probably the one most widely used.
Orthogonal Factors : It indicates the factors that are not correlated
with each other; factors obtained through
orthogonal rotation.
Orthogonal Rotation : It refers to the operation through which a
simple structure is sought under the restriction
that factors be orthogonal (or uncorrelated);
factors obtained through this rotation are by
definition uncorrelated.
Principal Axis : It is a method of initial factoring in which the
28
-II Factoring adjusted correlation matrix is decomposed Multivariate Analysis:
hierarchically; a principal axis factor analysis Factor Analysis
with iterated commonalities leads to a least
squares solutions of initial factoring.
Principal Components : It reflects linear combinations of observed
variables, possessing properties such as being
orthogonal to each other, and the first
principal component representing the largest
amount of variance in the data, the second
representing the second largest and so on;
often considered variants of common factors,
but more accurately they are contrasted with
common factors which are hypothetical.
Scree Test : It is a rule of thumb criterion for determining
the number of significant factors to retain; it
is based on the graph of roots (eigenvalues);
claimed to be appropriate in handling
disturbances due to minor (unarticulated)
factors.
Varimax : It is a method of orthogonal rotation which
simplifies the factor structure by maximizing
the variance of a column of the pattern matrix.

1) Anderson, T. W. (1984); An Introduction to Multivariate Statistical
Analysis, John Wiley & Sons, Second Edition.
2) Bartlett, M.S. (1954); A note on Multiplying factors for various Chi-
squared Approximations, Journal of the Royal Statistical Society, 16, 296-
298.
3) Johnson R.A., & Wichern D. W. (2002); Applied Multivariate Statistical
Analysis, Pearson Education, Inc.
4) Johnson, Dallas E. (1998); Applied Multivariate Methods for Data
Analysis, International Thomson Publishing Inc.
5) Kaiser, H.F. (1958); The Varimax criterion for analytical rotation in factor
analysis, Psychometrica, 23, 187-200.
6) https://onlinecourses.science.psu.edu
7) https://en.wikipedia.org/wiki/Factor_Analysis
8) http://www.researchgate.net/file.PostFileLoader.html?id=53c50b3fd5a3f21
40c8b465e&assetKey=AS%3A271747469774848%401441801053551.

PROGRESS EXERCISES
1) See Section 13.1
2) See Section 13.4
29
Quantitative Methods-II 3) See Section 13.4 M
4) See Section 13.4
1) See Section 13.6
2) See Section 13.7
3) Hints: How to run Factor Analysis in SPSS?
After opening the SPSS windows sheet in your computer system, the
following are the steps which can be used for performing, evaluating and
analyzing the factor analysis solution:
Step 1 : Feed the entries in the SPSS work sheet.
Step 2 : Go to “Analyze” in the menu available at the top of the window
screen.
Step 3 : Click on “Data Reduction”
Step 4 : Click on “Factor”
Step 5 : Click on “Extraction” and select “principal axis factoring” and
press “Continue”
Step 6 : Click on “OK”
13.15 EXERCISES
1) What do you understand by the Factor score? Describe the different
methods for the estimation of factor score.
2) Write a short note on Factor rotation and Oblique rotation.
3) What do you understand by the methods of estimation? Describe in details,
the principal component and maximum likelihood methods.
4) With suitable example, explain communality in context to factor analysis.
5) Define the following terms:
a) Correlation matrix
b) Specific variance
c) Factor loadings
d) Specific factors
30
-II Canonical Correlation
UNIT 14 CANONICAL CORRELATION Analysis
ANALYSIS
Structure
14.0 Objectives
14.1 Introduction
14.2 Canonical Correlation Analysis (CCA): Concept and Meaning
14.3 Assumptions of Canonical Correlation
14.4 Canonical Correlation Analysis as Generalization of the Multiple
Regression Analysis
14.5 Steps and Procedure Involved in Computation of CCA Results
14.6 Illustration of CCA
14.7 Interpretation of CCA Results
14.8 Limitations of Canonical Correlation
14.9 Let Us Sum Up
14.10 Key Words
14.12 Answers or Hints to Check Your Progress
14.13 Exercises
14.0 OBJECTIVES
• explain the concept of canonical correlation analysis;
• state the similarity and difference between multiple regression and canonical
correlation;
• discuss the steps and procedure involved in canonical correlation;
• elucidate how to interpret the results of canonical correlation;
• point out the limitations of canonical correlation analysis.
14.1 INTRODUCTION
The conventional wisdom that the economic agents are rational and are guided by self-
interest in decision making is being questioned as it ignores the psychological and
social factors influencing decision making process. The research findings from many
disciplines like neuroscience, cognitive science, psychology, behavioral economics,
sociology, anthropology etc. indicate that the decisions made by the individuals in
many aspects of development (like savings, investment, energy consumption, health
etc.) are influenced by social contexts, local social networks, cultural factors, social
norms and shared mental models etc. (World Development Report, 2015). Hence,
inter-disciplinary perspective is being recognized as research approach to analyze
human behavior so as to improve the predictive power of economics. Canonical
31
Quantitative Methods-II correlation analysis is a powerful analytical tool to analyze the association between C
two sets of variables belonging to different disciplines. This method measures the
strength of association between the two sets of variables. An attempt is made in this
method to concentrate a high-dimensional relationship between two sets of variables
into a few pairs of canonical variables. Hence, in this unit, we shall throw light on
various issues relating to canonical correlation like concept and meaning of canonical
correlation, its similarity and differences with multiple regression, procedure involved
in the analysis of canonical correlation, and its advantages and limitations.
14.2 CANONICAL CORRELATION ANALYSIS (CCA):

CONCEPT AND MEANING
Canonical correlation analysis is a multivariate statistical model used to study the
interrelationships among sets of multiple dependent variables and multiple
independent variables. This technique is distinct from the multiple regression model in
the sense that multiple regression predicts a single dependent variable from a set of
multiple independent variables whereas canonical correlation simultaneously predicts
multiple dependent variables from multiple independent variables.
Let us understand the concept with an example. As a student of economics, you may
like to know the association between economic inequality X and political instability
Y . The economic inequality can be measured by five variables i.e. (i) the division of
farmland X , (ii) the gini coefficient X , (iii) the percentage of tenant farmers
X , (iv) the gross national product X , and (v) the percentage of farmers X .
Similarly the political instability can be measured by four variables (indicators) i.e. (i)
the instability of leadership Y , (ii) the level of internal group violence Y , (iii) the
occurrence of internal war Y , (iv) stability of democracy Y . These two
theoretical concepts X and Y can be called two sets of variables or canonical
variables. These can be shown in the following figure:
Y1
X1
X2 Y2
X* Y*
. .
. .
X5 Y4
Fig. 14.1: Canonical Correlation

Source: The sage encyclopedia of Social Sciences Research Methods vol. 1 (2004) page no. 83.
The first canonical variable X is measured by five variables (P = 5) and can be

considered as a linear combination (a weighted sum) of these X variables. The second
canonical variable Y is a linear combination of the q = 4 indicators, Y to Y . The
double side curved arrow indicates that the question of casualty remains open.
The purpose of canonical correlation analysis is to find the correlation between a

linear combination of the variables in one set and a linear combination of the variables
in another set. The idea behind this approach is first to determine the pair of linear
combinations having the largest correlation. Next, we determine the pair of linear
combinations having the largest correlation among all pairs uncorrelated with the
32
-II initially selected pair, and so on. As stated in the above para, the pairs of linear Canonical Correlation
combinations are called the canonical variables and their correlations are called Analysis
canonical correlations.
Thus, canonical correlation aims to (i) identify the dimensions among the dependent
and independent variables, and (ii) maximize the relationship between the dimensions.
In this manner, in canonical correlation, we can distinguish three types of correlations:
1) Correlation between X variables, the correlation matrix is R .
2) Correlation between Y variables, the correlation matrix is R .
3) Correlation between X and Y variables, the correlation matrix is R R′ .
14.3 ASSUMPTIONS OF CANONICAL

CORRELATION
1) The correlation coefficient between any two variables is based on linear
relationship.
2) The canonical correlation is the linear relationship between the variates.
3) The distribution of variables is normal.
4) Hetro-scedasticity, to the extent it decreases the correlation between variables.
14.4 CANONICAL CORRELATION ANALYSIS AS

A GENERALIZATION OF MULTIPLE
REGRESSION ANALYSIS
Multiple regression analysis is a multivariate technique which can predict the value of
a single dependent variable from a linear function of a set of independent variables.
But this is not always the case. There are real life problems, however, when interest
may not center on a single dependent variable. Rather, the researcher may be
interested in relationships between sets of multiple dependent and multiple
independent variables. Canonical correlation analysis is a multivariate statistical model
that facilitates the study of interrelationships among sets of multiple dependent
variables and multiple independent variables. Whereas multiple regression predicts a
single dependent variable from a set of multiple independent variables, canonical
correlation simultaneously predicts multiple dependent variables from multiple
independent variables. Therefore, canonical correlation analysis is said to be a
generalization of multiple correlation used in multiple regression problems.
The coefficient of determination R2, in regression problems is the proportion of the
variability in a dependent variable that is accounted for by a set of predictor variables
and √ is called the multiple correlation coefficient. The multiple correlation
coefficient can also be interpreted as a measure of the maximum correlation that is
attainable between the dependent variable and any linear combination of the predictor
variable. Canonical correlation places the fewest restrictions on the types of data on
which it operates. Because the other techniques impose more rigid restrictions, it is
generally believed that the information obtained from them is of higher quality and
may be presented in a more interpretable manner. For this reason, many researchers
view canonical correlation as a last effort, to be used when all other higher-level
techniques have been exhausted. But in situations with multiple dependent and
independent variables, canonical correlation is the most appropriate and powerful
multivariate technique. It has gained acceptance in many fields and represents a useful
tool for multivariate analysis, particularly while considering multiple dependent
variables.
33
14.5 STEPS AND PROCEDURE INVOLVED IN C
COMPUTATION OF CCA
In 1935-36, Hotelling proposed a method, known as Canonical Correlation Analysis to
investigate “linear” relationship between the two sets of variates.
James Press (2005) has expressed the whole idea of this Canonical Correlation
Analysis in the following words:
“The Canonical correlation” model selects weighted sums of variables from each of
the two sets to form new variables in each of the sets, so that the correlation between
the new variables in “different sets” is maximized while the new variables within each
set are constrained to be uncorrelated with mean zero and unit variance.
We shall adhere to Press’s approach.

Let
α: p 1
and
γ: p 1
be two unknown vectors to be determined such that the correlation between α′ Y and
γ′ Z be as large as possible.
So, let
U α′ Y
V γ′ Z
and
U , V Correlation coefficient between U and V .
The problem of correlation coefficient now amounts to the following:
maximize U , V
Subject to
Var U Var V 1
and
E U EV 0.
Hotelling solved this problem using the celebrated method of Lagrange Multipliers.
However, we shall just state the final results. Hotelling showed that this maximization
problem is equivalent to following algorithm:
Step 1: Solve for the equation
!Σ Σ
0
!Σ
----------------- (A)
Σ
Here, Σ Var Y

Σ Var Z
Σ Σ Var Y, Z
Let be the largest positive root of the above equation.
34
-II Step 2: Now, solve the system of equations: Canonical Correlation
! Σ Σ % Analysis
" #$ ' 0
Σ ! Σ &
for α and γ.
Mathematically, it is not so simple to solve the above system of linear equations.
However, there is an equivalent formulation. To compute α and γ, we solve the pair of
equations:
(Σ Σ ) Σ ! Σ *α 0
(Σ Σ ) Σ ! Σ *γ 0
Step 3: We now call these α and γ as α and γ respectively; and
U α′ γ
and
V γ′ Z
as First Canonical Variates.
It will turn out that will be the correlation coefficient between U and V. We,
therefore, write
ρU , V
and call this as First Canonical Correlation.
Step 4: We now proceed to the next iteration. We now define:
U α′ γ
V γ′ Z,
where % and & are to be determined.
maximize U , V
subject to
Var U Var V 1
and
E U EV 0.
We repeat the above procedure to compute α and γ as solution of
! Σ Σ % 0
" # $& ' $ '
Σ ! Σ 0
where is the second largest positive root of the equation (A).
Step 5: We continue this procedure uptil the smallest positive root.
Step 6: The result is now to be collected in a vector format:
 U1 
 
 U2 
 . 
 = (U … ./0 *
′
U =
 . 
 . 
 
 Up 
 1
35
Quantitative Methods-II and C
 V1 
 
 V2 
 . 
V=  V … VP ′ .
 . 
 . 
 
 Vp 
 1
The elements of U and V are called Canonical Variates and 2 is corresponding
canonical correlation.
14.6 ILLUSTRATION OF CCA

Let us now understand the canonical correlation analysis technique with the help of an
illustration.
This example is based on 416 observations collected through primary survey for the
purpose of research study entitled “Assessment of Human Well-being in Delhi:
Multidimensional Approach”. The researcher attempted to study the relationship
between economic well-being variables and overall life satisfaction variables. One of
the questions raised in this study was to examine how does economic well-being
influence the overall life satisfaction of the people or more specifically, to know
whether economic well-being indicators are predictive of overall life satisfaction of
people.
The main characteristic of canonical analysis is the investigation of the relationship

between two sets of variables. One set is the predictor set or, say, analytically, the set
of independent variables. The second consists of the criteria or dependent variables.
In our example, the set of economic well-being variables constitutes the set of
independent variables. This set of economic well-being variables consist of 5
indicators of economic well-being:
1) Annual Income
2) Movable Assets
3) Fixed Assets
4) Employment Status
5) Educational Attainment
Next, the set of criteria variables constitute the overall life satisfaction indicators:
1) My life closer to my ideal life
2) Circumstances of my life are best to my choice
3) I am satisfied with my life
4) I have achieved the things in my life I aspired
5) I would not prefer to make any drastic change in my rest of life till I survive
The data on these two sets of variables was collected through primary survey for the
above said study.
36
-II Let us now learn to run the Canonical Correlation Analysis for our example using Canonical Correlation
SPSS Analysis
Step 1: Click file, new and syntax sequence.
Step 2: Then type the following syntax :
37
Quantitative Methods-II Note that the criterion set of variables are listed before WITH and predictor variables C
are listed afterwords
Step 3: This command can be implemented using RUN button on toolbar menu.
The results of above illustration using SPSS has been presented in this section. The
table 1 below shows an overall multivariate test of the entire model using different
multivariate criteria.
Table 14.1: Multivariate Tests of Significance
Test Name Value Approx. F Hypoth. DF Error DF Sig. of F
Pillais .14258 2.40693 25.00 2050.00 .000
Hotellings .15845 2.56302 25.00 2022.00 .000
Wilks .86055 2.49153 25.00 1509.72 .000
Roys .11708 .000
Table 14.2: Eigenvalues and Canonical Correlations
Root No. Eigenvalue Pct. Cum. Pct. Canon Cor. Sq. Cor
1 .13260 83.69055 83.69055 .34217 .11708
2 .01720 10.85381 94.54437 .13003 .01691
3 .00698 4.40258 98.94695 .08323 .00693
4 .00167 1.05222 99.99917 .04080 .00166
5 .00000 .00083 100.00000 .00115 .00000
38
-II In this example, we are not only interested to know whether there is a relationship Canonical Correlation
between the predicator and criterion variables but also wanted to know which Analysis
indicators of economic well-being are more or less useful in explaining the
relationship between life satisfaction and economic well-being. This is exactly where
CCA comes into play. The interpretation and evaluation of the results has been
discussed in next section.
14.7 INTERPRETATION OF CCA RESULTS

We now come to the conclusive stages of the Canonical Correlation Analysis.
To draw the right conclusions amounts to the correct interpretation of the
results obtained by conducting the CCA.
The first step is to evaluate the overall statistical significance of the full
canonical model. This is done by testing the Null hypothesis. The Null
hypothesis is that there is no relationship between the two set of variables. The
alternative hypothesis is that the two sets of variables are related.
The interpretation of CCA is now accomplished by computing the F-statistic

value using the Wilks- test.
We now come to the actual interpretation of the CCA as conducted in our example.
Here, the value of the computed value of the F-statistic is high and significant. This
is given in table 1 as 0.860. This means that we can reject the Null hypothesis.
That is to say that we accept the alternative hypothesis. Thus, statistically significant
relationship exists between life satisfaction and economic well-being indicators.
Let us recall, that the first root (function) is created in such a manner that the canonical
correlation between the new variable is maximized and these new variables within
each set are uncorrelated with zero mean and unit variance.
Let us interpret only those functions which explain reasonable amount of variance
between variable sets. In our illustrations, we interpret only the first function as it
explains 12% variance within the function as shown in table no. 2. All other functions
explain less than 10% of variance in their functions, hence we can ignore them.
So, what we have concluded so far is that there is a statically significant relationship
between our two set of variables. Further, this relationship is largely captured by the
first root (function) in the canonical model. Next, we want to identify those variables
which contribute significantly to explain this relationship between economic well-
being and overall life satisfaction.
In multiple regression analysis, we often look at Beta weights to identify the relative
contribution of one independent variable to explain dependent variable. In CCA, we
look at the structure coefficient to decide which variables are useful for the model.
Therefore, we examine the standardized weights and structure coefficients to interpret
the first root (function). Let us underline the point that we are only concerned with the
first function and will ignore other functions as they are not significant.
To understand the pattern among two set of variables, we have created table 3 showing
coefficients which presents the standardized canonical function coefficients (i.e.
weights) and structure coefficients for all variables. The squared structure coefficient
(r3 2) are also given, which represent the percentage of shared variance between
observed variable and the new variable created from the observed variables set.
39
Quantitative Methods-II Table 14.3: Canonical Solution for Economic Well-being Indicator Predicting C
Overall Life Satisfaction for Function 1
Variable Coef 45 45 2 (%)
Life closer to ideal life .221 -.502 25.20
Circumstances of life best to my choice -.692 -.905 81.90
Satisfied with life -.223 -.729 53.14
Achieved things in life -.400 -.746 55.65
Want no drastic change in life -.054 -.427 18.23
Annual Income -.381 -.857 73.44
Fixed Assets .029 -.256 6.55
Movable Assets -.210 -.696 48.44
Educational Attainment -.668 -.895 80.10
Employment status .219 -.286 8.17
Coef: Standized Canonical Function Coefficient, r3 = Structure Coefficient, r3 2 =
Squared Structure Coefficient
Looking at the coefficient, we find that relevant dependent variables are: the
circumstances of life are best to my choice, I am satisfied with my life and I have
achieved things in my life I aspired, because all these variables high squared structure
coefficient which indicates the amount of variance the observed variable can
contributed to new latent criterion variable. Looking at the other side of the equation
on function 1 which involves predicator set, we find that Annual Income, Movable
Assets and Educational Attainment variables were primary contributors. You must
note that this process for interpretation of function is same as identifying the useful
predictors in regression analysis with the exception that in canonical correlation
analysis we have two set of equations for consideration.
Check Your Progress I

1) What do you mean by the term ‘inter-disciplinary perspective’
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
2) How is CCA useful to address the inter-disciplinary nature of research questions?
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
3) State the steps involved in computation of CCA through SPSS software.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
14.8 LIMITATIONS OF CANONICAL

CORRELATION
1) The canonical correlation express the variance shared by the linear composites of
the set of variables and not the variance extracted from the variables.
40
-II 2) Canonical weights derived in computing canonical functions are subject to great Canonical Correlation
deal of instability. Analysis
3) The interpretation of canonical variates is different due to the efforts to maximize
the relationship.
4) It is difficult to identify meaningful relationship between the subset of independent
and dependent variables because precise statistics is yet to be developed.
14.9 LET US SUM UP

Canonical correlation analysis is a useful and powerful technique for exploring the
relationships among multiple dependent and independent variables. The technique is
primarily descriptive, although it may be used for predictive purposes. In this
technique, weighted sums of variables are selected from each of the two sets to form
new variables in each of the sets so that the correlation between the new variables in
different sets in maximized while the new variables within each set are constrained to
be uncorrelated with mean zero and unit variance. Results obtained from a canonical
analysis suggest answers to questions concerning the number of ways in which the two
sets of multiple variables are related, the strengths of the relationships, and the nature
of the relationships defined. Canonical analysis enables the students to combine into a
composite measure what otherwise might be an unmanageably large number of
bivariate correlations between sets of variables. It is useful for identifying overall
relationships between multiple independent and dependent variables, particularly
when we have little a priori knowledge about relationships among the data for two sets
of variables. Essentially, we can apply canonical correlation analysis to a set of
variables that appear to be significantly related.
The CCA is based on two statistical assumptions. First, the correlation coefficient
between any two variables is based on a linear relationship. Second, the parent
population from which the sample has been drawn is normally distributed. CCA has
several advantages and limitations.
14.10 KEY WORDS

Canonical correlation : Measure of the strength of the overall relationships
between the linear composites (canonical variates)
for the independent and dependent variables. In
effect, it represents the bivariate correlation
between the two canonical variates.
Canonical correlation : It is a multivariate statistical analysis that facilitates
analysis the study of interrelationships among the sets of
multiple dependent variables and multiple
independent variables.
Canonical cross-loadings : Correlation of each observed independent or
dependent variable with the opposite canonical
variate. For example, the independent variables are
correlated with the dependent canonical variate.
They can be interpreted like canonical loadings,
but with the opposite canonical variate.
Canonical function : Relationship (correlation) between two linear
composites (canonical variates). Each canonical
function has two canonical variates, one for the set
of dependent variables and one for the set of
independent variables. The strength of the
relationship is given by the canonical correlation.
41
Quantitative Methods-II Canonical loadings : Measure of the simple linear correlation between C
the independent variables and their respective
canonical variates. These can be interpreted like
factor loadings, and are also known as canonical
structure correlations.
Canonical roots : Squared canonical correlations, which provide an
estimate of the amount of shared variance between
the respective optimally weighted canonical
variates of dependent and independent variables. It
is also known as eigenvalues.
Canonical variates : Linear combinations that represent the weighted
sum of two or more variables and can be defined
for either dependent or independent variables. It is
also known as linear composites, linear
compounds, and linear combinations.
Multiple regression : Multiple regression analysis predicts a single
analysis dependent variable from a set of multiple
independent variables.

1) Alissa Sherry, Robin K. Henson (2005); Conducting and Interpreting Canonical
Correlation Analysis in Personality Research. Journal of Personality Assessment.
2) Hotelling, H. (1936); Relations between two sets of variables, Biometrika 28, 321-377
3) Johnson, R.A. & Wichern, D.W. (2002); Applied Multivariate Statistical Analysis,
Pearson Education, Inc.
4) Johnson, Dallas E. (1998); Applied Multivariate Methods for Data Analysis,
International Thomson Publishing Inc.
5) Michael S. Lewis-beck Alan Bryman Tim Futing Liao (2004); The sage
encyclopedia of Social Sciences Research Methods vol. 1 page no. 83
6) Magnus Borga; Canonical Correlation. A tutotial www.cs.cmu.edu/~tom/10701-
spll/sides/cca-totorial.pdf
7) S. Press James (2005): Applied Multivariate Analysis. Dover publication.inc.

PROGRESS
Check Your Progress I
1) See Section 14.1
2) See Section 14.6
3) See Section 14.6
14.13 EXERCISES
1) Under what circumstances would you select canonical correlation analysis instead
of multiple regressions as the appropriate statistical technique?
2) Discuss in details the procedure for computation of the canonical correlation.
3) What are the limitations associated with canonical correlation analysis?
42
-II Cluster Analysis
UNIT 15 CLUSTER ANALYSIS
Structure
15.0 Objectives
15.1 Introduction
15.2 Cluster Analysis: Concept and Meaning
15.3 Steps and Algorithm Involved in Cluster Analysis
15.4 Methods of Cluster Analysis
15.5 Partitioning Cluster Methods
15.6 Hierarchical Cluster Methods
15.7 Other Approaches: Two-step Cluster Analysis
15.8 Interpretation of the Results
15.9 Let Us Sum Up
15.10 Key Words
15.13 Exercises
15.0 OBJECTIVES
• state the concept and purpose of cluster analysis;
• list the steps to be followed in cluster analysis;
• explain the different approaches to cluster analysis; and
• to learn how to apply cluster analysis in analyzing economic problems and
interpret its results.
15.1 INTRODUCTION
In social sciences, data set usually take the form of observations on unit of
analysis for a set of variables. You as a student of economics may be interested
to identify groups of similar objects like countries, enterprises, households on
the basis of selected variables like unemployment rate of men and women in
different countries, deprivation indicators of households etc. Similarly you may
need to segment customers into a small number of groups for additional
analysis and marketing activities. Further researcher may aim to find out spatial
or temporal pattern in human developments, educational development etc. In
such cases, a simple classification of units into sub-groups is required for the
purpose of analysis because classes or conceptually meaningful groups
(clusters) share common characteristics and play an important role in analysis
and description of the world. Cluster analysis is a useful technique in such
situations for data analysis. For example, cities can be grouped or clustered in
43
Quantitative Methods-II terms of their social, economic and demographic characteristics. Similarly,
people can be clustered in terms of their psychological profiles or other
attributes they possess. Hence, in this unit, we shall discuss various clustering
methods and algorithm involved in cluster analysis. Let us begin with
explaining the concept of cluster analysis.
15.2 CLUSTER ANALYSIS: CONCEPT AND

MEANING
Classes or conceptually meaningful groups of objects sharing common
characteristics, play an important role in how people analyze and describe the
world. The creation of groups of similar objects or variables based upon
measured characteristics is an important method of analysis which we call
cluster analysis. This is an important method of multi-variate analysis. After
collecting the data, we divide the observations into groups (clustering) and
assign the labels to these groups. Thus cluster analysis is the study of
techniques for automatically finding classes. Cluster analysis is a strong tool of
the multivariate exploratory data analysis. It involves a great amount of
techniques, methods and algorithms which can be applied in various fields of
social sciences including economics.
Let us remember that Grouping, or clustering, is distinct from the classification

methods. Classification pertains to known number of groups with the purpose
to assign new observations to one of these groups. Cluster analysis is a more
primitive technique in the sense that no assumptions are made concerning the
number of groups or the group structure. Grouping is done on the basis of
similarities or distances (dissimilarities). The inputs required are similarity
measures or data from which similarities can be computed. The general
practical application of cluster analysis is whether the investigator knows
enough about the problem to distinguish "good" groupings from "bad"
groupings. This can be done by enumerating all the possible groupings and
select the "best" ones for further study.
Cluster analysis aims to discover natural groupings of the items (or variables).
For this, we need to first develop a quantitative scale to measure the
association (similarity) between objects. Since clustering involves a collection
of data objects which are similar to one another within the cluster and
dissimilar to objects in other clusters, a cluster of data objects can be treated as
an implicit class. Due to this it is also sometimes referred to as ‘automatic
classification’. In some applications, it is also known as ‘data segmentation’
because clustering divides large data sets into groups based on their
similarities. Cluster analysis enables us to detect outliers. As a branch of
statistics, cluster analysis has been extensively studied with the main focus on
distance-based cluster analysis and has been built into statistical analysis
software packages or systems such as SPSS, SAS and S-Plus. An illustration of
cluster analysis using SPSS shall be discussed towards the end of the unit. It is,
however, first important to understand the steps and algorithm involved in
conducting cluster analysis and the main methods used under this analytic
technique in research.
44
-II
15.3 STEPS AND ALGORITHIM INVOLVED IN Cluster Analysis
CLUSTER ANALYSIS
The following five broad steps are involved in conducting cluster analysis:
1) Measuring the relevant variables (both quantitative and categorical) that
can be included,
2) Creating a (dis) similarity Matrix for an appropriate measure of (dis) similarity,
Distances are used to measure (dis) similarity between the samples of two
variables. To determine the extend of (dis) similarity we construct what is
known as (dis) similarity matrix. Every entry of (dis) similarity matrix
determine using the notion of distance, more formally metric. Various sort of
metric can be employed each being suitable to suit the particular sort of data set.
The metric that is usually employed is known as minskowi metric, the
formula of which is as under:

, | |
where , , , . . . , and , , , . . . , are two n –

dimensional data objects and k is a positive integer.
For k = 1 and k = 2 the Minskowi metric becomes equal to Manhattan and
Euclidean metric respectively.
3) Creating one or more clustering’s via a clustering algorithm. An example
for clustering algorithm:
For the sake of convenience, we shall consider a sub matrix from the table
of (dis) similarity matrix shown in table 1.
Let us take
0 
 
1 0 
0 1 0 
A=  
1 2 1 0 
0 1 0 1 0 
 
 16 17 16 9 16 0 
Now, we implement the algorithm.
Step 1: The entries in A are squared Euclidean distances. We first take their
positive square root and call the new matrix as D1
1 0 
 
2 1 0 
3 0 1 0 
D1 =  
4 1 2 1 0 
5 0 1 0 1 0 
 
6  4 17 4 3 4 0 
45
Quantitative Methods-II We label rows as 1, 2, 3, 4, 5, 6. Taking the row 1 as a fixed variable, we
now locate the variable closest to it. We observe that there are two such
choices (1,3) & (1,5) as both have distance zero. We take (1,3) and put 1 &
3 in the same cluster. Now we calculate the distance of (1,3) from
remaining rows with the understanding that
, 115
min" , # min"1,1# 1
$ min" $ , $ # min"1,1# 1
% min" % , % # min"0,0# 0
' min" ' , ' # min"4,4# 4
Hence, now D2 becomes
13  0 
 
2 1 0 
D2 = 4  1 2 0 
 
5 0 1 1 0 
6  4 17

3 4 0
We observe that because 1 & 3 were clustered together, therefore, 3rd row
so does the 3rd column from D1 gets deleted.
Step 2: We now see that taking (13) as the fixed variable, the variable
closest to (13) in the matrix D2 is row 5 as their distance is zero. Hence, we
cluster together (13) & 5.
Now we calculate distance of (13)5 from the remaining variable to

construct matrix D3.
Hence now D3 becomes
(13)5  0 
 
2  1 0 
D3 =
4  1 2 0 
 
6  17 − 1 17 3 0 
We can continue like this and it is entirely up to us when to stop depending

on how many clusters we want.This method of creating clusters is
applicable in both hierarchical as well as non hierarchical approach to
cluster analysis.
4) providing some assessment of the obtained clustering(s), and
5) Interpreting the clustering(s) in substantive terms.
Assessment and interpretation of the obtained clustering (s) has been explained
in section 15.8 with the help of an illustration.
46
-II
15.4 METHODS OF CLUSTER ANALYSIS Cluster Analysis
Imagine you are going to undertake fieldwork in Uttar Pradesh next week and
for doing so you wish to divide your group into 5 teams and assign 10 villages
under one team each. Strategically, you would like to identify those set of 10
villages to bear somewhat similar characteristics as much possible. Suppose
that the district information is also unavailable leaving manual grouping
difficult. You need a clustering tool to help you in this scenario. Clustering is a
process of grouping a set of objects into multiple groups or clusters so that
objects having high similarity are in same cluster but having dissimilarity in
other clusters. Similarities and dissimilarities are generally assessed based on
attribute values describing objects and involve distance measures.
Creation of simple group structure from a complex data set require a measure
of “closeness” or “similarity”. Often a great deal of subjectivity is involved in
the choice of a similarity measure. Important considerations include the nature
of the variables (discrete, continuous, binary), scales of measurement (nominal,
ordinal, interval, ratio), and subject matter knowledge. When items (units or
cases) are clustered, proximity is usually indicated by some sort of distance. On
the other hand, variables are usually grouped on the basis of correlation
coefficients or like measures of association.
An advantage of using Cluster analysis is to represent data in graphical way as
it offers several possibilities to do that. Dendrogram or other graphical
solutions can be used to visually examine linkages between observations. A
dendrogram is a tree-like structure commonly used to represent the process of
hierarchical clustering. It shows how objects are grouped together (in an
agglomerative method) or partitioned (in a divisive method) on a step-by-step
basis.
Two types of clustering algorithms are generally reported, namely non-
hierarchical and hierarchical. Partitioning is the most extensively used non-
hierarchical method and would be discussed in detail in this unit. While
partitioning method classifies objects into a specified number of groups (say k),
hierarchical method does not construct a single partition with k clusters, but
deals with all values of k in the interval [1,n] where n is the total number of
objects. Other types of non-hierarchical clustering methods are density-based
and grid-based which are beyond the scope of this unit.Non-hierarchical cluster
analysis tends to be used when large data sets are involved. It is sometimes
preferred because it allows subjects to move from one cluster to another (this is
not possible in hierarchical cluster analysis where a subject, once assigned,
cannot move to a different cluster). The two major disadvantages of non-
hierarchical cluster analysis are:
1) It is often difficult to know how many clusters you are likely to have and
therefore the analysis may have to be repeated several times, and
2) It can be very sensitive to the choice of initial cluster centres.
15.5 PARTITIONING CLUSTERING METHODS

As introduced earlier, a partitioning method classifies objects into say, k groups
which together must satisfy the following criteria:
47
Quantitative Methods-II Each group must contain at least one object and each object must belong to one
group. To put it simply, it implies that k should be less than or equal to n,
where n is the total number of objects. In other words, partitioning methods
conduct one-level partitioning on data sets. The basic partitioning methods
typically adopt ‘exclusive cluster separation’ i.e. each object must belong to
exactly one group. Stratified random sampling lend an interesting example of
partitioning methods. Suppose a household sample size of 100 has to be
collected from Delhi representing all five geographical zones of Delhi i.e. East,
West, North, South and Central zones. In this case the five zones are the 5
groups and are pre-specified in number. Each zone must contain at least one
household and each household in the sample must belong to one zone. Clearly
in this example, the number of groups k (here 5) are less than the number of
objects n (here 100).
In general, after initial partitioning, the ‘iterative relocation technique’ attempts
to improve the partitioning by moving objects from one group to another.
When it comes to judging a good partitioning, it can be inferred on the basis of
whether objects in the same cluster are ‘close’ or related to each other whereas
objects in different clusters are ‘far apart’.
We shall now discuss one of the most frequently used partitioning method in
social sciences and in statistical analysis and that is the K-means method. It is
a Centroid based technique. K-means clustering is used to split the
observations into K clusters/groups and test what are the main characteristics of
these clusters/groups. To understand it in simple words, let us assume that we
have a data set D which contains n objects in Euclidian space i.e. two
dimensional space. Partitioning methods first distribute the objects in data set
D into k clusters such as X1, X2,…Xk,wherein each Xi is an element of D and Xi
∩ Xj = Ø. Thereafter an objective function is defined to assess the partitioning
quality so that within a cluster there are similar objects while other clusters
have dissimilar objects.
As mentioned earlier, K-means is a centroid based method which uses the
centroid of a cluster to represent that cluster. Conceptually, centroid is nothing
but its center point or mean. It is, many a times, defined as the mean of the
objects assigned to the cluster. After calculating the mean, this methods
measures the distance of objects from the centroid in the Euclidian space. The
quality of a cluster is measured by the within-cluster variation, which is the
sum of squared error between all objects in Xi and the centroid ci.
How does the k-means algorithm work? Han, Kember and Pei have explained
the workings in a very simple manner. The k-means algorithm defines the
centroid of a cluster as the mean value of the points within the cluster.
a) First, it randomly selects k objects in D, each of which initially represent a
cluster mean or center.
b) For each of the remaining objects, an object is assigned to the cluster to
which it is most similar, based on the Euclidian distance between the object
and the cluster mean.
c) K-means algorithm thereafter improves the intra-cluster variation iteratively.
d) For each cluster, it computes the new mean using the objects assigned to
the cluster in the previous iteration.
e) All objects are then reassigned using updated means as the new cluster
centers.
48
-II f) Iterations continue until the assignment is stable i.e. clusters formed in the Cluster Analysis
current step are equal to the clusters formed in previous step.

1) Explain the concept and intuition of Cluster analysis. What are the steps
involved in carrying out cluster analysis?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Which are the broad methods used in cluster-analysis? Explain the k-means
method clearly indicating the steps involved in it.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
15.6 HIERARCHICAL CLUSTERING METHODS

In some situations, you may like to partition data into groups at different levels
as in a hierarchy. A Hierarchical clustering method works by grouping data
objects into a hierarchy or ‘tree’ of clusters. By doing so it becomes easy to
summarize data. This will enable you to visually interpret it. For example, the
employees of a college may be divided on the basis of their nature of
appointment: Permanent or Temporary and one may further partition them into
smaller sub-groups like Professor, Associate Professor, Reader, Assistant
Professor, etc. All these groups form a hierarchy. You can easily characterize
or summarize the data in a hierarchy and can use it to find the average working
experience or salaries of Professors and of Assistant Professors.
You may use hierarchical clustering methods to analyze the hierarchical
structure in an organization or dataset or may also use to discover an
underlying hierarchical structure. In the study of evolution, hierarchical
clustering has been used to group animals according to their biological features
which has enabled scientists to discover evolutionary paths, which are a
hierarchy of species.
Hierarchical algorithms do not construct a single partition with k clusters, but
they deal with all values of k in the interval [1, n]. Hierarchical clustering
methods are of two types: Agglomerative and Divisive. Agglomerative
hierarchical methods start with each unit in its own cluster and systematically
merges units and clusters until all units form a single cluster. Thus, there are
initially as many clusters as objects. The most similar objects are first grouped,
and these initial groups are merged according to their similarities. Eventually,
as the similarity decreases, all subgroups are fused into a single cluster. An
agglomerative hierarchical method uses a bottom-up strategy. Han, Kamber
49
Quantitative Methods-II and Pei explain that it typically starts by letting each object form its own
cluster and iteratively merges clusters into larger and larger clusters, until all
objects are in a single cluster. The single cluster becomes the hierarchy’s root.
For merging, it finds two clusters closest to each other and combines to form
one cluster. It requires at most n iterations as two clusters are merged per
iteration and each cluster contains atleast one object.
Divisive hierarchical methods work in the reverse direction. Initially all objects
belong to the same cluster; then, iteratively, a cluster is chosen according to a
selection criterion and bi-partitioned such that the objects in one subgroup are
“far from” the objects in the other. These subgroups are then further divided
into dissimilar subgroups; the process continues until there are as many clusters
as objects that is, until each object forms a group. Thus it employs a top-down
approach. It starts by placing all objects in one cluster, which is the hierarchy’s
root. It then divides the root cluster into several smaller sub-clusters and
repeatedly partitions them into smaller clusters.
In other words, hierarchical clustering techniques proceed by either a series of
successive mergers or a series of successive divisions. In either agglomerative
or divisive hierarchical clustering, a user can specify the desired number of
clusters as a termination condition. The results of both agglomerative and
divisive methods may be displayed in the form of a two- dimensional diagram
known as dendrogram. However, this graph is generally useful for relatively
small data sets only. In case of large data sets or sample size, individual objects
cannot be identified. As we shall see, the dendrogram illustrates the mergers or
divisions that have been made at successive levels.
Illustration: Suppose we take a sample of size 15 and wish to examine the inter-
relationship of employment status and educational attainment as shown in Figure
15.1. Cluster Analysis helps in clustering objects which are similar in attributes.
Fig. 15.1: Sample data for Cluster Analysis

Source: Assessment of Human wellbeing: A Multidimensional Approach un published study
undertaken by prof. Narayan Prasad
50
-II In Figure15.2, a dendrogram is constructed using SPSS which clusters sample Cluster Analysis
data. A dendrogram is a branching diagram that represents the relationships of
similarity among a group of entities. Each branch is called a clade. The
terminal end of each clade is called a leaf. Clades can have just one leaf (these
are called simplicifolious, a term from botany that means “single leafed”) or
they can have more than one. Two-leaved clades are bifolious, three-leaved are
trifolious, and so on. There is no limit to the number of leaves in a clade. The
arrangement of the clades tells us which leaves are most similar to each other.
The width of the branch points indicates how similar or different they are from
each other: the greater the width, the greater the difference. In this particular
dendrogram, we see that chunk 6, 10 and 13 are completely separate from all
the others as they are clustered through a separate leaf of the clade. We
interpret its placement as indicating that the distribution in that chunk is
substantially different from the distribution in the remaining chunks. By closer
inspection, it can be observed that these 3 individuals are post-graduates as
well as employed on either contractual basis or unemployed i.e. it reflects that
higher educational attainment has not improved their job status as well as job
security.
If we are trying to identify which individual segments are most similar to each
other, we read the dendrogram from the left to right, identifying the first clades
to join together as we move from left to right. By closer inspection we find 4
clusters which have most similar objects. The first cluster is of 2 and 15
indicating better job status after graduation, the second cluster is of 1 and 3
(both are employers and post-graduates), the 3rd cluster is of 4,8,9 and 12
which has linked post-graduates and the regularity of their employment while
the 4th cluster is made up of 6 and 10 indicating contractual work despite being
post-graduate. The further the clades away from the observations/objects, less
is the similarity between them.
Table 15.1: Similarity between level of education and level of employment
Proximity Matrix
Squared Euclidean Distance
Obser
vation 1:A 2:B 3:C 4:D 5:E 6:F 7:G 8:H 9:I 10:J 11:K 12:L 13:M 14:N 15:O
1:A .000 1.000 .000 1.000 .000 16.000 1.000 1.000 1.000 16.000 5.000 2.000 29.000 2.000 2.000
2:B 1.000 .000 1.000 2.000 1.000 17.000 2.000 2.000 2.000 17.000 4.000 5.000 26.000 1.000 1.000
3:C .000 1.000 .000 1.000 .000 16.000 1.000 1.000 1.000 16.000 5.000 2.000 29.000 2.000 2.000
4:D 1.000 2.000 1.000 .000 1.000 9.000 .000 .000 .000 9.000 2.000 1.000 20.000 1.000 1.000
5:E .000 1.000 .000 1.000 .000 16.000 1.000 1.000 1.000 16.000 5.000 2.000 29.000 2.000 2.000
6:F 16.000 17.000 16.000 9.000 16.000 .000 9.000 9.000 9.000 .000 5.000 10.000 5.000 10.000 10.000
7:G 1.000 2.000 1.000 .000 1.000 9.000 .000 .000 .000 9.000 2.000 1.000 20.000 1.000 1.000
8:H 1.000 2.000 1.000 .000 1.000 9.000 .000 .000 .000 9.000 2.000 1.000 20.000 1.000 1.000
9:I 1.000 2.000 1.000 .000 1.000 9.000 .000 .000 .000 9.000 2.000 1.000 20.000 1.000 1.000
10:J 16.000 17.000 16.000 9.000 16.000 .000 9.000 9.000 9.000 .000 5.000 10.000 5.000 10.000 10.000
11:K 5.000 4.000 5.000 2.000 5.000 5.000 2.000 2.000 2.000 5.000 .000 5.000 10.000 1.000 1.000
12:L 2.000 5.000 2.000 1.000 2.000 10.000 1.000 1.000 1.000 10.000 5.000 .000 25.000 4.000 4.000
13:M 29.000 26.000 29.000 20.000 29.000 5.000 20.000 20.000 20.000 5.000 10.000 25.000 .000 17.000 17.000
14:N 2.000 1.000 2.000 1.000 2.000 10.000 1.000 1.000 1.000 10.000 1.000 4.000 17.000 .000 .000
15:O 2.000 1.000 2.000 1.000 2.000 10.000 1.000 1.000 1.000 10.000 1.000 4.000 17.000 .000 .000
Source: Assessment of Human Wellbeing: A Multidimensional Approach, a research study

(unpublished) undertaken in 2015 by Prof. Narayan Prasad.
51
Quantitative Methods-II DENDROGRAM USING WARD METHOD
Rescaled Distance Cluster Combine
0 5 10 15 20 25
+---------+---------+---------+---------+---------+
N 14 ─┐
O 15 ─┼─┐
B 2 ─┘ ├───┐
K 11 ───┘ │
C 3 ─┐ ├─────────────────────────────────────────┐
E 5 ─┼─┐ │ │
A 1 ─┘ │ │ │
H 8 ─┐ ├───┘ │
I 9 ─┤ │ │
D 4 ─┤ │ │
G 7 ─┼─┘ │
L 12 ─┘ │
F 6 ─┬───┐ │
J 10 ─┘ ├───────────────────────────────────────────┘
M 13 ─────┘
Fig.15.2: Dendrogram
15.7 OTHER APPROACHES: TWO-STEP

CLUSTER ANALYSIS
Among the other widely used non-hierarchical approaches is the approach of
Two-step cluster analysis. The Two Step Cluster Analysis procedure is an
exploratory tool designed to reveal naturalgroupings (or clusters) within a data
set that would otherwise difficult to find. The Algorithm employed by this
procedure has several desirable features that differentiate it from traditional
clustering techniques:
• The ability to create clusters based on both categorical and continuous
variables.
• Automatic selection of the number of clusters.
• The ability to analyze large data files efficiently.
In order to handle categorical and continuous variables, the TwoStep Cluster

Analysis procedure uses a likelihood distance measure which assumes that
variables in the cluster model are independent. Further, each continuous
variable is assumed to have a normal (Gaussian) distribution and each
categorical variable is assumed to have a multinomial distribution.
The two steps of the Two Step Cluster Analysis procedure's algorithm can be
summarized as follows:
Step 1 The procedure begins with the construction of a Cluster Features (CF)
Tree. The tree begins by placing the first case at the root of the tree in a leaf
node that contains variable information about that case. Each successive case is
then added to an existing node or forms a new node, based upon its similarity
to existing nodes and using the distance measure as the similarity criterion. A
52
-II node that contains multiple cases contains a summary of variable information Cluster Analysis
about those cases. Thus, the CF tree provides a capsule summary of the data
file.
Step 2 The leaf nodes of the CF tree are then grouped using an agglomerative
clustering algorithm. The agglomerative clustering can be used to produce a
range of solutions. To determine which number of clusters is “best”, each of
these cluster solutions is compared using Schwarz's Bayesian Criterion (BIC)
or the Akaike Information Criterion (AIC) as the clustering criterion.
15.8 INTERPRETATION OF THE RESULTS

We shall now discuss this method with the help of an example based on 416
observations collected through primary survey for the purpose of research
study entitled “Assessment of Human Wellbeing in Delhi :Multidimensional
Approach”. The researcher in this study attempted to find out the clusters on
the basis of household level of income and their educational status. Four
income levels have been used to classify the sample into four categories:
income up to 2.5 lakhs, between 2.5 and 5 lakhs, between 5-10 lakhs and above
10 lakhs. The educational attainment has been take as : illiterate (no education),
upto middle education level, between middle to higher secondary education,
upto graduation, upto post-graduation and PhD and above. The hypothesis was
that higher the level of education, higher will the level of income meaning
thereby a positive association between level of education and the level of
income. Let us examine this hypothesis by applying the two-step cluster
analysis technique.
Figure 15.3 shows results of cluster 1 wherein educational attainment of post-

graduation has been found to be inter-linked with income category of 10 lakhs
and above. This is logically also true as higher levels of educational attainment
are associated with higher income levels. In Figure 15.4, it can be observed
that graduation is associated with income-levels of 5-10 lakhs and these two
sub-groups have been clustered together.
Fig.15. 3: Cluster 1
53
Fig.15.4: Cluster 2
Figure 15.5 finds another cluster linking illiterate individuals with the lowest
income category (i.e. less than 2.5 lakhs) while Figure 15.6 finds the fourth
cluster consisting of middle to higher secondary education and income level
upto 2.5 lakhs. Thus intuitively also, two-step cluster analysis has formed
clusters of similar attributes.
Fig.15.5: Cluster 3
54
-II Cluster Analysis
Fig.15.6: Cluster 4
Thus, by using Two step cluster analysis technique , we have separated 416
individuals into 4 broad clusters based on their income category and level of
educational attainment which, in turn , has enabled us to test the hypothesis
whether an association exist between the level of education and level of
income. Thus the hypothesis that higher the level of education, higher the level
of income has been accepted by the results of the cluster analysis.
1) What are the hierarchical methods used in cluster analysis? Critically
explain how dendrograms help in visual interpretation of clustered data.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) Explain the application of two-step cluster analysis in research. What are its
advantages?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
15.9 LET US SUM UP

The objective of cluster analysis is to assign observations to groups (clusters)so
that observations within each group are similar to one another with respect to
variables or attributes of interest and the groups themselves stand apart from
55
Quantitative Methods-II one another. Cluster analysis embraces a variety of methods aiming to group
observations or variables into homogeneous and distinct clusters. For
groupings based on variables, frequently used measures of the similarity of
observations are the Euclidean. Squared Euclidean are applied to the original,
standardized, or weighted variables. For groupings based on attributes, a
measure of the similarity of two observations is the ratio of the number of
matches (identical categories) to the number of attributes.
Hierarchical methods begin with as many clusters as there are observations and
end with a single cluster containing all observations. All clusters formed by
these methods are mergers of previously formed clusters. Other types of
clustering methods are the hierarchical divisive (beginning with a single cluster
and ending with as many clusters as there are observations) and the non-
hierarchical methods. In recent years, some other clustering methods have also
evolved like the two-step cluster analysis discussed above which are efficient
in handling large data sets of continuous and categorical variables as well.
Cluster analysis technique gives a different angle of viewing and interpreting
data which may help the researcher to discover significant new relationships
between the variables used in a study.
15.10 KEY WORDS

Cluster analysis : It is the task of grouping a set of objects in
such a way that objects in the same group
(called a cluster) are more similar (in some
sense or another) to each other than to those in
other groups (clusters).
Agglomerative method : In which subjects start in their own separate
cluster. The two ’closest’ (most similar)
clusters are then combined and this is done
repeatedly until all subjects are in one cluster.
At the end, the optimum number of clusters is
then chosen out of all cluster solutions.
Divisive method : In which all subjects start in the same cluster
and the above strategy is applied in reverse
until every subject is in a separate cluster.
Agglomerative methods are used more often
than divisive methods, so this handout will
concentrate on the former rather than the latter.
Dendrogram : When carrying out a hierarchical cluster
analysis, the process can be represented on a
diagram known as a dendrogram. This diagram
illustrates which clusters have been joined at
each stage of the analysis and the distance
between clusters at the time of joining.
Centroid method : Here the centroid (mean value for each
variable) of each cluster is calculated and the
distance between centroids is used. Clusters
whose centroids are closest together are
merged. This method is also fairly robust.
56
-II Ward’s Method : In this method all possible pairs of clusters are Cluster Analysis
combined and the sum of the squareddistances
within each cluster is calculated. This is then
summed over all clusters. The combination
that gives the lowest sum of squares is chosen.
This method tends to produce clusters of
approximately equal size, which is not always
desirable. It is also quite sensitive to outliers.
Despite this, it is one of the most popular
methods, along with the average linkage
method.
K-Means Clustering : The k-means algorithm was popularized and
refined by Hartigan (1975). The basic
operation of that algorithm is relatively simple.
Given a fixed number of (desired or
hypothesized) k clusters, assign observations to
those clusters so that the means across clusters
(for all variables) are as different from each
other as possible.
Euclidean Distance : The straight line distance between two points
in a Cartesian coordinate system. The
Euclidean distance can be determined using
the Pythagorean Theorem. In two dimensions,
the Euclidean distance is [(x1-x2)2 + (y1-y2)2]0.5.
Usually, the points represent samples and the
axes of the Cartesian coordinate system
represent the abundances of species.
Exploratory Analysis : A general term for an analysis in which the
chief objectives is to find pattern in the data.
Often, exploratory analysis conflicts with
hypothesis testing. For example, stepwise
regression is permissible in exploratory
analysis, but can cause serious problems if you
are interested in testing hypotheses.

1) Data Mining, Concepts and Techniques: Jiawei Han, Micheline Kamber,
Jian Pei, 3rd Edition, Morgan Kaufmann, 2012.
2) Cluster Analysis of Economic Data, Hana Rezankova, Statistika, 94 (1),
Czech Republic, 2014
3) http://wheatoncollege.edu/lexomics/files/2012/08/How-to-Read-a-
Dendrogram-Web-Ready.pdf
4) https://en.wikipedia.org/wiki/Cluster_Analysis
5) https://arifkamarbafadal.files.wordpress.com/2011/09/ebook-038-tutorial-
spss-two-step-cluster-analysis.pdf
57
PROGRESS EXERCISES
1) See Section 15.2 and 15.3
1) See Section 15.6
2) See Section 15.7
15.13 EXERCISES
1) For what purposes and situations would you choose to undertake cluster
analysis?
2) Explain in your own words briefly what hierarchical cluster analysis does.
Check your answer against the material above.
3) Explain the difference between hierarchical methods and non-hierarchical
methods of clustering citing their advantages and disadvantages.
4) Is clustering same as classification? What is the essential difference
between classification and clustering?
58
-II Correspondence
UNIT 16 CORRESPONDENCE ANALYSIS Analysis
Structure
16.0 Objectives
16.1 Introduction
16.2 Correspondence Analysis: Concept and Its Features
16.3 Steps and Algorithm Involved in Correspondence Analysis Technique
16.4 Basic Concepts and Definitions
16.4.1 Primitive Matrix
16.4.2 Profiles
16.4.3 Masses
16.4.4 Correspondence Matrix
16.4.5 Augmented Correspondence Matrix
16.4.6 Inertia
16.4.7 Distances
16.5 Reduction of Dimensionality
16.6 Biplots
16.7 Interpretation of the Results of Correspondence Analysis
16.8 Multiple Correspondence Analysis
16.9 Let Us Sum Up
16.10 Key Words
16.12 Answers or Hints to Check Your Progress
16.13 Exercises
16.0 OBJECTIVES
• state the concept of Correspondence Analysis (CA) as a special case of
principal component analysis;
• discuss the special features and applications of CA technique;
• explain the computational algorithm of Correspondence Analysis;
• elucidate the process of interpretation of Correspondence Analysis results; and
• learn how to apply Correspondence Analysis technique in conducting
research in social sciences.
16.1 INTRODUCTION
Understanding data is the ultimate problem of statistics and no one
technique/method can satisfactorily answer all our curiosities regarding any
particular set of data. Given a set of data, we often formulate hypothesis
regarding it. This formulation is motivated by our information regarding the
variables of the data. We then apply various statistical tools to test our
hypothesis.
59
Quantitative Methods-II In such traditional methods of hypothesis testing, we verify a priori hypothesis,
that is, the hypothesis pre-assumed regarding the relations between variables
involved in the data. However, these methods naturally suffer from a
limitation. The pre-assumed hypotheses may not be the only possibility. There
can be other hidden relationships among the variables of the data. How do we
find them?
You have studied in Unit 1 and 8 that from measurement scale point of view,
data can be of four types – nominal, ordinal, interval and ratio scale. Principal
Component Analysis (PCA) technique is used to identify the relative
importance of factors or components responsible for variance in the dependent
variable. PCA, however, pre-supposes tables consisting of continuous
measurement (i.e. interval or ordinal scale of data). If a situation arises when
one is interested to know association between two or more categorized
variables, search for an alternative analytic technique is needed. For example,
you may be interested to examine how the level of education is related to the
nature of employment (i.e. regular, casual, contract etc.). Similarly how the
quality of employment is associated with different levels of income? In such
situation, you need a data analytic technique which can help you in detecting
structural relationships among the categorized variables. Correspondence
Analysis (CA) is analytic technique which helps to examine such structural
relationship. This technique is very useful to analyze simple two way and
multi-way tables containing some measure of correspondence between the
rows and columns. Hence, in this unit, we shall throw light on various concepts
used in correspondence analysis, its pre-requisites, its computation algorithm,
interpretation of its results, etc. Let us begin with stating the meaning of
correspondence analysis.
16.2 CORRESPONDENCE ANALYSIS: CONCEPT

AND ITS FEATURES
Correspondence Analysis (CA) is a descriptive exploratory technique of
multivariate statistical analysis, allowing to define the nature and structure of
the relationship between qualitative variables measured in nominal and ordinal
scales. CA is a way of seeing the association in a two way – cross tabulation
rather than measuring it. In general, CA as a method is used to decomposes the
overall chi-square statistics by defining a system with a smaller number of
dimensions, in which the deviation from the expected values are presented. In
brief, CA may be defined as a special case of Principal Component Analysis of
the rows and columns of a table especially applicable to a cross tabulation. The
primary goal of CA is to transform a table of numerical information into a
graphical display, in which each row and each column is depicted as a point.
The important features of correspondence analysis are:
1) CA provides the multivariate treatment of the data through simultaneous

consideration of multiple categorical variables.
2) This technique can reveal relationships that would not be detected in a
series of pair wise comparison of variable.
3) It also provides graphical display of row and column points in biplots to
dectect structural relationships among the variable categories and objects.
60
-II 4) Except rectangular data matrix with non-negative entries, its data requirement Correspondence
is highly flexible. Analysis
16.3 STEPS AND ALGORITHM INVOLVED IN

CORRESPONDENCE ANALYSIS
TECHNIQUE
CA technique involves the following steps and algorithm:
1) Construction of the primitive matrix.
2) Calculation of the row profile and the column profile.
3) Computation of the correspondence matrix.
4) Obtain total row mass & total column mass.
5) Construction of the augmented correspondence matrix. Sometimes for the
sake of simplicity, it is also named as correspondence matrix.
6) Obtain the chi-square distance & inertia.
7) Reduction of the dimension of the data.
8) Obtain the biplots.
16.4 BASIC CONCEPTS AND DEFINITIONS

Let us understand the basic concepts and definitions of correspondence
analysis with the help of an illustration. We are taking up a worked out
example to understand some of the points involved in the correspondence
analysis technique.
Let us consider 2 qualitative variables: Educational Status and Employment

Status. Correspondence analysis technique will help us to examine the
association between these two qualitative variables. For the purpose of our
analysis, we have divided employment status into six categories: Employer,
regular, pensioner/earning income from assets, temporary, contractual/own
account worker and unemployed. Similarly, educational attainment of the
persons have been categorized as illiterate, upto medium education, between
medium and higher education, upto graduation, uptopost graduation or any
technical degree and Ph.D and above.
Before examining the association between these two categorical variables, let
us first understand the basic terms and definitions involved in the
correspondence analysis technique.
16.4.1 Primitive Matrix

Primitive Matrix is the information which has been collected and classified in
the tabular form. It is a cross tabulation of row and column variables. Table no.
16.1 offers a good example of Primitive Matrix. The information about
employment status and educational attainment with their sub categories in rows
and columns respectively has been provided in this table.
61
Table 16.1: Correspondence Between Educational Attainment and Employment Status
Primitive Matrix (A = aij)
Employment_Status
Active
Educational Attainment unemployed Contractual Temporary Pensioner Regular Employer Margin
Illiterate 22 16 1 1 0 1 41
Upto Middle 27 7 0 1 7 2 44
Middle to Higher
69 8 3 2 4 12 98
Secondary
Graduation 51 2 4 4 22 20 103
Post grad/Technical
13 11 7 0 46 14 91
degree
PhD & Above 1 3 4 0 28 3 39
Active Margin 183 47 19 8 107 52 416
Source: Assessment of Human Wellbeing: A Multidimensional Approach, a research study (unpublished)

undertaken in 2015 by Prof. Narayan Prasad.
16.4.2 Profiles
Profiles display the characteristics of a particular category which is determined
by their respective frequencies. As shown in the table no. 16.2a, row profile
describes the educational status while the column profile displays the
employment status.
Row profiles give the frequency of each row point such that summation for row
will be equal to 1. Row profile are calculated by taking individual row point
divided by their corresponding row total (active margin). That is,
A

row profile for the ith row =: A

=
∑
Column profiles are calculated in similar manner.
A

Column profile for the ithcolumn =: A

=
∑
The only difference is that in this case, each column point is divided by their
corresponding column total (active margin) as shown in table 16.2b.
62
-II Correspondence
Table 16.2a: Correspondence Between Educational Attainment and Employment Analysis
Status
Row Profiles
Employment_Status
Educational Active
Attainment unemployed Contractual Temporary Pensioner Regular Employer Margin
Illiterate .537 .390 .024 .024 .000 .024 1.000
Upto Middle .614 .159 .000 .023 .159 .045 1.000
Middle to
Higher .704 .082 .031 .020 .041 .122 1.000
Secondary
Graduation .495 .019 .039 .039 .214 .194 1.000
Post
grad/Technical .143 .121 .077 .000 .505 .154 1.000
degree
PhD & Above .026 .077 .103 .000 .718 .077 1.000
Mass .440 .113 .046 .019 .257 .125
Table 16.2b: Correspondence Between Educational Attainment and Employment Status

Column Profiles
Employment_Status
Educational Attainment unemployed Contractual Temporary Pensioner Regular Employer Mass

Illiterate .120 .340 .053 .125 .000 .019 .099
Upto Middle .148 .149 .000 .125 .065 .038 .106
Middle to Higher Secondary .377 .170 .158 .250 .037 .231 .236
Graduation .279 .043 .211 .500 .206 .385 .248
Post grad/Technical degree .071 .234 .368 .000 .430 .269 .219
PhD & Above .005 .064 .211 .000 .262 .058 .094
Active Margin 1.000 1.000 1.000 1.000 1.000 1.000
16.4.3 Masses
The mass of the ith row is the marginal frequency of ithrow divided by grand
total. Marginal frequency of the ith is the sum of the entries of the ith row.Mass
for each row and column points are depicted in table no. 16.2a & 16.2b
respectively.Grand total in this particular illustration is total sample size which
is 416. Similarly mass for jth column is marginal frequency of jth column
divided by grand total.
16.4.4 Correspondence Matrix

Correspondence matrix is defined as the original matrix divided by grand total.
The correspondence matrix shows how value of mass is distributed across the
cells. The row total and the column total of the correspondence matrix are the
row mass and column mass respectively as shown in table 16.3.
63
Table 16.3: Correspondence Between Educational Attainment and
Employment Status
Correspondence Matrix
Employment_Status
Educational
Attainment unemployed Contractual Temporary Pensioner Regular Employer
Illiterate 0.053 0.038 0.002 0.002 0.000 0.002
Upto Middle 0.065 0.017 0.000 0.002 0.017 0.005
Middle to Higher
Secondary 0.166 0.019 0.007 0.005 0.010 0.029
Graduation 0.123 0.005 0.010 0.010 0.053 0.048
Post grad/Technical
degree 0.031 0.026 0.017 0.000 0.111 0.034
PhD & Above 0.002 0.007 0.010 0.000 0.067 0.007
16.4.5 Augmented Correspondence Matrix

Having computed the correspondence matrix, we now append this
correspondence matrix by adding the column of row masses and the row of
column masses in the table 16.4a.
Table 16.4 a: Correspondence Between Educational Attainment and

Employment Status
Augmented Correspondence Matrix
Employment_Status
Educational Row
Attainment unemployed Contractual Temporary Pensioner Regular Employer Mass
Illiterate 0.053 0.038 0.002 0.002 0.000 0.002 0.099
Upto Middle 0.065 0.017 0.000 0.002 0.017 0.005 0.106
Middle to Higher
Secondary 0.166 0.019 0.007 0.005 0.010 0.029 0.236
Graduation 0.123 0.005 0.010 0.010 0.053 0.048 0.248
Post grad/Technical
degree 0.031 0.026 0.017 0.000 0.111 0.034 0.219
PhD & Above 0.002 0.007 0.010 0.000 0.067 0.007 0.094
Column Mass 0.440 0.113 0.046 0.019 0.257 0.125 1.000
16.4.6 Inertia
The term inertia, in the correspondence analysis explains the degree of
variation. The value of inertia is high when there is large deviation of row and
column profiles from their averages. The method to compute the total inertia is
as follows:
2
Total inertia = åå ( )
p ij - ri c j
i j ri c j
Where pij = (i, j)th entry of correspondence matrix
64
-II Correspondence
Ri = ith row profile Analysis
Cj = jth column profile

The inertia in table no. 16.3 gives the total variance explained by each
dimension in the model. In this illustration value of inertia is0.480. This value
of inertia indicates that in this model knowing something about education
attainment explains around 48% of something about employment status and
vice versa. This association is highly significant as indicated by chi square
statistic.
16.4.7 Distance
Independence and association between the two variables is measured by chi
square. Chi square distances are weighted Euclidean distance between the
profile points. Weights here refers to weights of dimension and not to the
weights of profile points. Chi square is calculated as follows:
Formula to compute the chi-square distance:
Suppose the correspondence matrix of the primitive matrix is mxn. Then, we
define:
Chi-square distance between rth and sth rows is:=
# ' 2
, ! "∑)*+# $ & (
%
% '
whereaj = marginal frequence of jth column

= ∑.
-+# ,-*
ar = marginal frequenceofrth column

= ∑)*+# ,/*
as = marginal frequence of sth column
= ∑)*+# ,0*
16.5 REDUCTION OF DIMENSIONALITY

Another way of looking at correspondence analysis is to consider it as a
method for decomposing the overall inertia by identifying a small number of
dimensions in which the deviations from the expected values can be
represented. This is similar to the goal of factor analysis, where the total
variance is decomposed, so as to arrive at lower – dimensional representation
of variables that allows one to reconstruct most of the variance/covariance
matrix of variables.
Notice the tables 16.4 given below. The educational attainment column shows
only 5 categories. Also notice that employment status has been condensed into
two categories only – employed and unemployed. This reduction of
dimensionalities is not arbitrary. It is a result of classification of sub categories
in distances from same point of reference. This will further help in pictorial
presentation of correspondence between the two variables – namely education
attainment and employment status. Correspondence analysis aims to
decompose the overall inertia so as to arrive at a lower dimensional space.
Only those dimensions are included in the model which can easily be
65
Quantitative Methods-II interpreted. That is why the value of inertia does not always add up to 1 as all
dimensions are not included in the model.
Employment Status
Summary
Confidence Singular
Proportion of Inertia Value
Correlation
Singular Chi Accounted Standard
Dimension Value Inertia Square Sig. for Cumulative Deviation 2
1 .594 .353 .735 .735 .035 .009
2 .329 .108 .226 .960 .054
3 .113 .013 .026 .987
4 .069 .005 .010 .996
5 .041 .002 .004 1.000
Total .480 199.660 .000a 1.000 1.000
a. 25 degrees of freedom
16.6 BIPLOTS
The biplot is concerned with data reconstruction in a joint map of the rows and
columns, rather than distance reconstruction. In the simple case of a two-way
table one can think of reconstructing different variants of the table depending
on the way we think of the table: either as a set of rows, or a set of columns, or
just a two-way table of entries where rows and columns are symmetric entities.
A biplot is a graphical representation of the information in an n x p data matrix.
The bi- refers to the two kinds of information contained in a data matrix. The
information in the rows pertains to samples or sampling units and that in the
columns pertains to variables. When there are only two variables, scatter plots
can represent the information on both the sampling units and the variables in a
single diagram. This permits the visual inspection of the position of one
sampling unit relative to another and the relative importance of each of the two
variables to the position of any unit.
With several variables, one can construct a matrix array of scatter plots, but
there is no one single plot of the sampling units. On the other hand, a two-
dimensional plot of the sampling units can be obtained by graphing the first
two principal components. The idea behind biplots is to add the information
about the variables to the principal component graph.
66
-II Correspondence
Analysis
Fig.16.1: Biplots showing correspondence between Educational Attainment and Employment

Status
Data has been graphically represented by the biplot in two dimensional space
as shown in figure 1. However, there is one caution in this particular model that
this model explains only 48% of employment status based on educational
attainment. From the above biplot some general trends can be seen. Two points
closer to each other shows greater degree of association between them as
compared to others. This plot is offering us significant insight into direction of
correspondence between the level of educational attainment and employment
status. For example, we see that illiterate individuals are more likely to be
employed on the contractual basis or they are mostly engaged in low level of
own account work. Those individuals who have studied upto middle school are
mostly found to be unemployed. Those with Ph.D and above education has
higher possibilities of regular employment.The status of employment can be
seen improving for individuals who have completed graduation and post-
graduation as they are one who provides employment opportunities to others
also.Therefore , this plot clearly reflects a greater degree of association
between educational attainment and employment status which is also
significant as confirmed by the value of chi-square.
16.7 INTERPRETATION OF THE RESULTS OF

CORRESPONDENCE ANALYSIS
The interpretation of the results of correspondence analysis comprises the
interpretation of numerical results and factor graphics yielded by CA. The
former implies selection of significant axes and significant points.
67
Quantitative Methods-II We are taking up a worked out example for Correspondence Analysis to
illustrate some of the points involved in the interpretation of the results of the
correspondence analysis. Table 16.1 has information on 416 persons, their
educational status and their employment status. The data originated from the
study entitled ‘Assessment of Human Wellbeing: A Multidimensional
Approach’ based on a sample survey conducted in Delhi in 2015. Through this
technique, an attempt was made to find out correspondence between
educational attainment and employment status among the respondents selected
for the study. Educational status is subdivided into 6 categories, ranging from
illiterate to Ph.D and above. The employment status varies from outright
unemployed to contractual, temporary, regular, pensioner and also where
respondents are employers rather than employees.
The table 16.4 shows summary of our calculation. The first column shows the
five dimension – notice that as we are looking at distance from some dimension
– the number of dimension declines from 6 to 5. The singular value and inertia
are reported in column 2 to 3 respectively. Column 6 shows proportion of
inertia accounted for by the first dimension which is highest at 0-735 of the
total. As we move through dimension 2 onwards, inertia falls rather sharply.
Notice the similarity with principal components analysis – there too, the first
component explained largest variation. The column 7 shows cumulative
proportion of inertia accounted for.
The table 16.5 gives an overview of row points. Column 2 shows mass &
column 3 & 4 respectively show scores in dimension 1 & 2. Column 5 records
inertia of respective categories in column 1. Column 6 & 7 record contribution
of a point to inertia of dimension 1 & 2 respectively, while column 8 & 9
describe contribution of two dimension to inertia of the point concerned last
column show total inertia.
Employment Status
Overview Row Pointsa

EducationAttainment Mass Score in Inertia Contribution
Dimension
1 2 Of Point to Of Dimension to
Inertia of Inertia of Point
Dimension
1 2 1 2 Total
.099 - .948 .179 .350 .321 .860 .137 .997
illiterate
1.534
upto middle .106 -.771 .275 .046 .095 .029 .912 .048 .960
middle to higher .236 -.458 - .050 .074 .216 .654 .328 .982
secondary .503
.248 .108 - .022 .004 .235 .085 .798 .883
graduation
.511
post graduation .219 .975 .295 .146 .313 .069 .941 .036 .977
PhD and above .094 1.074 .619 .085 .163 .130 .840 .116 .956
Active Total 1.000 .528 1.000 1.000
68
-II The table 16.6 sums up similar information for column points. Correspondence
Analysis
Table 16.6
Overview Column Pointsa

EmploymentStatus Mass Score in Inertia Contribution
Dimension
1 2 Of Point to Of Dimension to
Inertia of Inertia of Point
Dimension
1 2 1 2 Total
unemployed .440 -.684 .180 .129 .346 .043 .948 .036 .984
.113 -.292 - .091 .016 .782 .063 .926 .989
contractual
1.510
temporary .046 .766 .003 .019 .045 .000 .858 .000 .858
pensioner .019 -.679 .529 .011 .015 .016 .472 .158 .630
regular .257 1.151 .002 .204 .574 .000 .992 .000 .992
employer .125 .128 .645 .026 .003 .158 .046 .658 .705
Active Total 1.000 .480 1.000 1.000
a. Symmetrical normalization
The figure 16.1 presents symmetrical normalization of the row and the column
points in form of a picture in two dimension plane.
We must note that in our table 16.2a as we move up the educational attainment
ladder, initially unemployment proportion is rising from 53.7% for illiterates to
61.4% for middle school educated and then 70% for middle to higher
secondary level of education. For graduates, the proportion of unemployed
come down to 49.5%.For post graduates, it is merely 14.3% and for highest
educational category, it is barely 2.6%. Even, this data alone is offering us
significant insight into direction of correspondence between the levels of
educational attainments and employment status. The sub-categories of
employment – contractual, temporary, regular, employer are shown separately
and the data further tends to support the hypothesis that higher level of
education attainment leads to higher possibilities of regular employment.
However, the sub-category of ‘graduates’ also performs credibly as employers,
while Ph.D and above get more regular employment.
Our correspondence analysis tends to provide us ‘concrete’ evidence in support
of our tentative ‘conclusion’ from inspection of data described above.
16.8 MULTIPLE CORRESPONDENCE ANALYSIS

In a two-way contingency table, simple correspondence analysis provides
association between two categorical variables. On the other hand, multiple
correspondence analysis, an extension of the simple correspondence analysis,
tackles the more general problem of associations among a set of more than two
categorical variables. Generalization to more than two variables is neither
obvious nor well-defined. In other areas of multivariate analysis, such as
regression and log linear modelling, the situation is less complicated: for
example, the transition from the regression of a response variable on a single
predictor variable to the case of several predictors is quite straightforward.
69
Quantitative Methods-II The main problem is that the notion of association between two categorical
variables is already a complex concept and there are several ways to generalize
this concept to more than two variables.
Many different approaches exist in literature to define multiple correspondence
analysis. Important among them are two main approaches: first, the correlation
between sets of variables, known as canonical correlation, and second, the
geometric approach, which is directly linked to data visualization, and which
has many similarities to Pearson-style principal component analysis. In the
explanation of each approach, one can consider the case of two variables only
and then describe possible generalizations to more than two variables.
1) What is correspondence matrix?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) A sample of n= 416 people is given in the following table showing the
annual income and level of education:
Table:Income level and level of educational attainment
Educational Annual Income
Attainment
Upto 2.50 lakh 5.00 lakh to 10.00 lakh Above
2.50 to 5.00 10.00 lakh to 20.00 20.00 lakh
Lakh lakh lakh
Illiterate 40 1 0 0 0
Up to Middle 28 3 8 3 2
Middle to 44 18 23 11 2
higher
secondary
Graduation 26 18 25 27 7
Post 2 6 25 31 27
Graduation
Ph.D& Above 1 2 6 18 12
Compute the value of chi-square statistics and inertia from the above table.
Also interpret the results.
…………………………………………………………………………………..
…………………………………………………………………………………..
…………………………………………………………………………………..
…………………………………………………………………………………..
…………………………………………………………………………………..
70
-II
16.9 LET US SUM UP Correspondence
Analysis
Correspondence analysis is a data analytic technique that helps to detect

structural relationships among the categorized variables. It allows to define the
nature and structure of the relationship between qualitative variables measured
in nominal and ordinal scales. This technique can be considered a special case
of principal component analysis of the rows and columns of a table especially
applicable to a cross tabulation. Determination of the correspondence matrix,
row and column profiles and masses, calculation of the distances between the
rows and columns, presentation of row and column profiles, determination of
the average row and column profiles, reducing the dimension of spaces, and
plotting the correspondence map in terms of biplots are the various steps
involved in correspondence analysis technique.
Correspondence analysis facilitates the visualization of the associations

between the rows and columns of a contingency table. This technique can be
applied across the desciplines ranging from social sciences (including
economics, psychology etc.) to health sciences, biometry etc. It is a versatile
method of data analysis in the situations where exploratory or more indepth
data analysis of categorical data is required.
By virtue of being primarily a graphical technique designed to represent

associations in a low-dimensional space, correspondence analysis can be
regarded as a scaling method.This technique is also viewed as a complement to
other methods such as biplots and multidimensional scaling. Correspondence
analysis also has links to principal component analysis and canonical
correlation analysis.
16.10 KEY WORDS

Correspondence : It is a multivariate statistical technique by Hirschfeld and
Analysis later developed by Jean-Paul Benzecri. It is conceptually
similar to principal component analysis, but applies to
categorical rather than continuous data. In a similar
manner to principal component analysis, it provides a
means of displaying or summarizing a set of data in
two-dimensional graphical form.
Inertia : The inertia of the whole table is a function of the Chi-
square statistic (1 2 .
8
34 56 7
If 1 2 ! ∑- ∑* 6
where 9-* is the count of row iand column j of the
contingency table, :-* is the expected value under the
assumption of row-by-column independence, and N is
the total table count, then the total inertia of the table is
given by
;8
Total Inertia = )
The inertia value reported is the proportion of the total
inertia that is due to this profile. Another way to
interpret the inertia is that it is the weighted average of
71
Quantitative Methods-II the Chi-square distances between the row profiles and
their average profile.
Centroid : The (weighted) mean of a multivariate data set which
can be represented by a vector. For many ordination
techniques, the centroid is a vector of zeros (that is, the
scores are centered and standardized). In a direct
gradient analysis, a categorical variable is often best
represented by a centroid in the ordination diagram.
Categorical : A variable that is represented by several different types;
Variable for example: lake/river/stream, farm/pasture/ unmanaged,
pitfall trap/fence trap/direct sighting. For most multivariate
analyses, categorical variables must be converted to k-1
dummy variables (where k = the number of categories).
Classification : The act of putting things in groups. Most commonly in
community ecology, the “things” are samples or
communities. Classification can be completely subjective,
or it can be objective and computer-assisted (even if
arbitrary).
Correlation : A method which determines the strength of the
relationship between variables, and/or a means to test
whether the relationship is stronger than expected due to
the null hypothesis. Usually, we are interested in the
relationship between two variables, x and y. The
coefficient of correlation r is one measure of the
strength of the relationship.
Correlation : Usually abbreviated by r, and a number which reflects
Coefficient the strength of the relationship between two variables. It
varies between -1 (for a perfect negative relationship) to
+1 (for a perfect positive relationship). If variables are
standardized to have zero mean and a unit standard
deviation, then r will also be the slope of the relationship.
The value r2 is known as the coefficient of determination;
it varies between 0 and 1. The coefficient of determination
is loosely interpreted as “the proportion of variance
in y which can be explained by x”.
Multiple : Multiple correspondence analysis applied to the general
Correspondence problem of associations among a set of more than two
Analysis(MCA) categorical variables. We shall see that the generalization
to more than two variables is neither obvious nor well-
defined. In other areas of multivariate analysis, such as
regression and log linear modelling, the situation is less
complicated. The main problem is that the notion of
association between two categorical variables is already a
complex concept and there are several ways to generalize
this concept to more than two variables.
Singular Value : The Singular Value Decomposition is the fundamental
Decomposition mathematical result for Correspondence Analysis, as it is
for other dimension reduction techniques such as principal
component analysis, canonical correlation analysis etc.
This matrix decomposition expresses any rectangular
72
-II matrix as a product of three matrices of simple structure, Correspondence
⁄2 ⁄2 ⁄2
i.e. D5#
PD5#
. The columns of the matrices D5# and Analysis
5#⁄2
D are the left and right singular vectors respectively,
and the positive values of the diagonal of P, in descending
order are the singular values. The SVD is related to the
more well-known eigenvalue- eigenvector decomposition
of a square symmetric matrix. If a generalized form of the
SVD were defined, where the singular vectors are
normalized with weighting by the masses, then the CA
solution can also be easily obtained.

1) Greenacre, M. J. (1984); Theory and applications of Correspondence
analysis, London Academic Press.
2) Gifi, A. (1990). Nonlinear Multivariate Analysis. Chichester, UK: Wiley.
3) Johnson R.A., &WichernD. W. (2002); Applied Multivariate Statistical
Analysis, Pearson Education, Inc.
4) Manly, B.F.J. (2005), Multivariate Statistical Methods: A Primer, Third
edition, Chapman and Hall.
5) Rencher, A.C. (2002), Methods of Multivariate Analysis, Second edition,
Wiley.
6) https://onlinecourses.science.psu.edu
7) https://en.wikipedia.org/wiki/Factor_Analysis
8) Young, F.W, and R.M Hamer (1987); Multidimensional scaling: History,
Theory and Applications, Lawrence Associates Publishers.
9) http://www.unesco.org/webworld/idams/advguide/Chapt6_5.htm
10) Laura Doey and Jessica Kurta (2011); Correspondence Analysis applied to
psychological research, Tutorials in Quantitative Methods for Psychology,
2011, Vol. 7(1), Page 5-14.
11) Michael S. Lewis-Beck Alan Bryman Tim Futing Liao (Ed.) (2004); The Sage
Encyclopedia of Social Science Research Methods Vol. 1, Page 203-205.

PROGRESS
1) See Section 16.4.6
2) The value of chi square is 219.816 and the value of inertia 0.528.
16.13 EXERCISES
1) What is Correspondence Analysis? Why do we use Correspondence Analysis?
2) What do you understand by multidimensional scaling? How it differ from
correspondence analysis.
73
Quantitative Methods-II 3) A sample of 901 individuals was cross classified according to three
categories of income and four categories of job satisfaction. The results are
given in the following table:
Table: Income and Job Satisfaction
Income Job Satisfaction
Very Somewhat Moderately Very
Dissatisfied Dissatisfied Satisfied Satisfied
< $ 25,000 42 62 184 207
< $ 25,000 - $ 50,000 13 28 81 113
>$ 50,000 7 18 54 92
Perform a correspondence analysis of these data. Also interpret the results.
74
-II Structural Equation
UNIT 17 STRUCTURAL EQUATION Modeling
MODELING
Structure
17.0 Objectives
17.1 Introduction
17.2 History of Structural Equation Modelling (SEM)
17.3 Why do we Conduct Structural Equation Modelling?
17.4 Assumptions of SEM
17.5 Similarities between Traditional Statistical Methods and SEM
17.6 Differences between Traditional Statistical Methods and SEM
17.7 Concepts and Terminology used in SEM
17.7.1 Path Diagrams
17.7.2 Classical SEM Notations
17.8 SEM Models Spepecification
17.8.1 Based on Reflective Indicators
17.8.2 Based on Formative Indicators
17.8.3 Based on on both Reflective and Formative Indicators
17.9 Issues in SEM Tecnique
17.10.1 Sample Size and Power
17.10.2 Missing Data
17.10.3 Multivariate Normality and Outliers
17.10 Steps in SEM
17.10.1 Initial Model Conceptualization
17.10.2 Model Estimation
17.10.3 Model Evaluation
17.10.4 Model Modification
17.10.5 Reporting the Results
17.11 An Example of SEM
17.12 Software Programs for SEM
17.12.1 LISREL
17.12.2 EQS
17.12.3 Mplus
17.12.4 Amos
17.12.5 Mx
17.12.6 Others
17.13 Advantages and Disadvantages of SEM
17.13.1 Advantages of SEM
17.13.2 Disadvantages of SEM
17.14 Let Us Sum Up
17.15 Key Words
17.16 References
17.0 OBJECTIVES
After going through the unit, you will be able to:
• State the meaning and importance of Structural Equation Modeling (SEM);
• Describe the basic concepts and terminology of SEM;
• Explain why do we conduct the SEM;
75
•
Quantitative Methods-II • Discuss the various models of SEM;
• Learn the steps involved in SEM;
• Know the different softwares used in SEM; and
• Discuss the advantages and disadvantages of SEM.
17.1 INTRODUCTION
SEM techniques are considered today to be a major component of applied
multivariate statistical analysis and are used by education researchers,
economists, marketing researchers, medical researchers, and variety of other
social and behavioural scientists. Although the statistical theory that underlies
the techniques appeared decades ago, a considerable number of years passed
before SEM received widespread attention it hold today. There are two reasons
for the recent attention: (1) the availability of specialized SEM programme and
(2) publication of several introductory and advanced text books on SEM.
Structural equation modelling (SEM) does not designate a single statistical
technique but instead refers to a family of related procedures. SEM is a
collection of statistical techniques that allow a set of relations between one or
more independent variables (IVs), either continuous or discrete, and one or
more dependent variables (DVs), either continuous or discrete, to be examined.
Both IVs and DVs can be either measured variables (directly observed), or
latent variables (unobserved, not directly observed).
SEM is an attempt to model causal relations between variables by including all

variables that are known to have some involvement in the process of interest. It
can be viewed as a combination of factor analysis and regression or path
76
•
-II analysis. The interest in SEM is often on theoretical construct, which are Structural Equation
represented by the latent factors. The relationship between theoretical construct Modeling
are represented by regression or path coefficients between the factors.
The SEM implies a structure for the covariance between the observed variables
which provides the alternative names like Covariance structural model. SEM
provides a very general and convenient framework for statistical analysis that
includes several traditional multivariate procedure for example, factor analysis,
regression analysis, and canonical correlation as special case. Structural
Equation Models are often visualized by a graphical path diagram. The
statistical model is usually represented in a set of matrix equations.
Structural equation models, also called simultaneous equation models, refers to
multi equation systems that include continuous latent variables each
representing a concept or construct, multiple indicators of a concept or
construct that may be continuous, ordinal, dichotomous or censored, errors of
measurement and errors in equations. One may also view it as an interrelated
system of regression equations where some of the variables (latent or
observable) have multiple indicators and where measurement error is taken into
account when estimating relationships. From a different point of view, these
are factor analysis models in which factor loadings are restricted to zero or
some other constants, and the researcher allows factors to influence each other,
directly and indirectly. The most general form of the structural equation model
includes Analysis of Variance, Analysis of Covariance, Multiple Linear
Regression, Multivariate Multiple Regression, Recursive and Non-recursive
Simultaneous Equations, Path Analysis, Confirmatory Factor Analysis and
many other procedures as special cases.
Structural equation modeling (SEM) uses various types of models to depict
relationships among observed variables, with the same basic goal of providing
a quantitative test of a theoretical model hypothesized by the researcher. More
specifically, various theoretical models can be tested in SEM that hypothesize
how sets of variables define constructs and how these constructs are related to
each other. For example, an educational researcher might hypothesize that a
student’s home environment influences her later achievement in school. A
marketing researcher may hypothesize that consumer trust in a corporation
leads to increased product sales for that corporation. A health care professional
might believe that a good diet and regular exercise reduce the risk of a heart
attack.
In each example, the researcher believes, based on theory and empirical
research, sets of variables define the constructs that are hypothesized to be
related in a certain way. The goal of SEM analysis is to determine the extent to
which the theoretical model is supported by sample data. If the sample data
support the theoretical model, then more complex theoretical models can be
hypothesized. If the sample data do not support the theoretical model, then
either the original model can be modified and tested, or other theoretical
models need to be developed and tested. Consequently, SEM tests theoretical
models using the scientific method of hypothesis testing to advance our
understanding of the complex relationships among constructs.
SEM has been defined by different researchers in different ways.
According to Holey (1995) it is a comprehensive statistical approach to testing
hypotheses about relations among observed and latent variables.
77
•
Quantitative Methods-II Rigdon, (1998): SEM is a methodology for representing, estimating, and
testing a theoretical network of (mostly) linear relations between variables.
MacCallum & Austin (2000): it tests hypothesized patterns of directional and
non directional relationships among a set of observed (measured) and
unobserved (latent) variables.
So, the term “Structural Equation Model” (SEM) refers to a comprehensive
statistical methodology for testing and estimating causal relations using a
combination of cross-sectional statistical data and qualitative causal
assumptions. Unlike the usual multivariate linear regression model, the
response variable in one regression equation in a structural equation model may
appear as a predictor in another equation. Indeed, variables in a structural
equation model may influence one-another reciprocally, either directly or
through other variables.
Two goals in SEM are:
1) to understand the patterns of correlation/covariance among a set of
variables and
2) to explain as much of their variance as possible with the model specified
(Kline, 1998).
This is the family tree of SEM
78
•
-II
17.2 HISTORY OF STRUCTURAL EQUATION Structural Equation
Modeling
MODELLING (SEM)
To discuss the history of structural equation modelling, it is better to explain
the chronological order of following four models: regression, path,
confirmatory factor, and structural equation models.
The first model involves linear regression models that use a correlation
coefficient and the least squares criterion to compute regression weights.
Regression models were made possible because Karl Pearson created a formula
for the correlation coefficient in 1896 that provides an index for the
relationship between two variables (Pearson, 1938). The regression model
permits the prediction of dependent observed variable scores (Y scores), given
a linear weighting of a set of independent observed scores (X scores) that
minimizes the sum of squared residual error values. Regression analysis
provides a test of a theoretical model that may be useful for prediction.
Some years later, Charles Spearman (1904, 1927) used the correlation
coefficient to determine which items correlated or went together to create the
factor model. His basic idea was that if a set of items correlated or went
together, individual responses to the set of items could be summed to yield a
score that would measure, define, or infer a construct. Spearman was the first
to use the term factor analysis in defining a two-factor construct for a theory of
intelligence. D.N. Lawley and L.L. Thurstone in 1940 further developed
applications of factor models, and proposed instruments (sets of items) that
yielded observed scores from which constructs could be inferred. The term
confirmatory factor analysis (CFA) is used today based in part on earlier work
by Howe (1955), Anderson and Rubin (1956), and Lawley (1958). The CFA
method was more fully developed by Karl Jöreskog in the 1960s to test
whether a set of items defined a construct.
Factor analysis has been used for over 100 years to create measurement
instruments in many academic disciplines, while today CFA is used to test the
existence of these theoretical constructs.
Sewell Wright (1918, 1921, 1934), a biologist, developed the third type of
model, a path model. Path models use correlation coefficients and regression
analysis to model more complex relationships among observed variables. The
first applications of path analysis dealt with models of animal behavior.
Unfortunately, path analysis was largely overlooked until econometricians
reconsidered it in the 1950s as a form of simultaneous equation modeling (e.g.,
H. Wold) and sociologists rediscovered it in the 1960s (e.g., O. D. Duncan and
H. M. Blalock). In many respects, path analysis involves solving a set of
simultaneous regression equations that theoretically establish the relationship
among the observed variables in the path model.
The final model type is structural equation modeling (SEM). SEM models
essentially combine path models and confirmatory factor models, that is, SEM
models incorporate both latent and observed variables. The early development
of SEM models was due to Karl Jöreskog (1969, 1973), Ward Keesling (1972),
and David Wiley (1973). This approach was initially known as the JKW
model, but became known as the linear structural relations model (LISREL)
with the development of the first software program, LISREL, in 1973. Since
then, many SEM articles have been published; for example, Shumow and
79
•
Quantitative Methods-II Lomax (2002) tested a theoretical model of parental efficacy for adolescent
students. For the overall sample, neighbourhood quality predicted parental
efficacy, which predicted parental involvement and monitoring, both of which
predicted academic and social-emotional adjustment.
Jöreskog and van Thillo originally developed the LISREL software program at
the Educational Testing Service (ETS). The first publicly available version,
LISREL III, was released in 1976. Later in 1993, LISREL8 was released; it
introduced the SIMPLIS (SIMPLE LISREL) command language in which
equations are written using variable names. In 1999, the first interactive version
of LISREL was released. LISREL8 introduced the dialog box interface using
pull-down menus and point-and-click features to develop models, and the path
diagram mode, a drawing program to develop models.. The field of structural
equation modelling across all disciplines has expanded since 1994.
17.3 WHY DO WE CONDUCT STRUCTURAL

EQUATION MODELLING?
Why is structural equation modelling popular? There are at least four major
reasons for the popularity of SEM.
The first reason suggests that researchers are becoming more aware of the
need to use multiple observed variables to better understand their area of
scientific inquiry. Basic statistical methods only utilize a limited number of
variables, which are not capable of dealing with the sophisticated theories
being developed. The use of a small number of variables to understand
complex phenomena is limiting. In contrast, structural equation modelling
permits complex phenomena to be statistically modelled and tested.
The second reason involves the greater recognition given to the validity and
reliability of observed scores from measurement instruments. Specifically,
measurement error has become a major issue in many disciplines, but
measurement error and statistical analysis of data have been treated separately.
Structural equation modelling techniques explicitly take measurement error
into account when statistically analyzing data.
The third reason pertains the ability to analyze more advanced theoretical
SEM models. For example, group differences in theoretical models can be
assessed through multiple-group SEM models. In addition, analyzing educational
data collected at more than one level—for example, school districts, schools, and
teachers with student data—is now possible using multilevel SEM modelling.
These advanced SEM models and techniques have provided many researchers
with an increased capability to analyze sophisticated theoretical models of
complex phenomena, thus requiring less reliance on basic statistical methods.
Finally, SEM software programs have become increasingly user friendly. For
example, until 1993 LISREL users had to input the program syntax for their
models using Greek and matrix notation. Today, most SEM software programs
are Windows-based therefore, the SEM software programs are now easier to
use and contain features similar to other Windows-based software packages.
17.4 ASSUMPTIONS OF SEM

Four basic assumptions must be met for SEM to be appropriate. These are:
80
•
-II 1) The relationship between the coefficients and the error term must be linear. Structural Equation
Modeling
2) The residuals must have a mean zero, be independent, be normally
distributed, and have variances that are uniform across the variable.
3) Variables in SEM should be continuous, interval level data. This means
SEM is not appropriate for censored data.
4) No specification error. If necessary variables are omitted or unnecessary
variables are included in the model, there will be measurement error and
the measurement model will not be accurate. Variables included in the
model must have acceptable level of kurtosis.
17.5 SIMILARITIES BETWEEN TRADITIONAL

STATISTICAL METHODS AND SEM
SEM is similar to traditional methods like correlation, regression and analysis
of variance in many ways. First, both traditional methods and SEM are based
on linear statistical models. Second, statistical tests associated with both
methods are valid if certain assumptions are met. Traditional methods assume a
normal distribution and SEM assumes multivariate normality. Third, neither
approach offers a test of causality.
17.6 DIFFERENCES BETWEEN TRADITIONAL

METHODS AND SEM
Traditional approaches differ from the SEM approach in several areas. First,
SEM is a highly flexible and comprehensive methodology. This methodology
is appropriate for investigating achievement, economic trends, health issues,
family and peer dynamics, self-concept, exercise, self-efficacy, depression,
psychotherapy, and other phenomenon.
Second, traditional methods specify a default model whereas SEM requires
formal specification of a model to be estimated and tested. SEM offers no
default model and places few limitations on what types of relations can be
specified. SEM model specification requires researchers to support hypothesis
with theory or research and specify relations a priori.
Third, SEM is a multivariate technique incorporating observed (measured) and
unobserved variables (latent constructs) while traditional techniques analyze
only measured variables. Multiple, related equations are solved simultaneously
to determine parameter estimates with SEM methodology.
Fourth, SEM allows researchers to recognize the imperfect nature of their
measures. SEM explicitly specifies error while traditional methods assume that
measurement occurs without error.
Fifth, traditional analysis provides straightforward significance tests to
determine group differences, relationships between variables, or the amount of
variance explained. SEM provides no straightforward tests to determine model
fit. Instead, the best strategy for evaluating model fit is to examine multiple
tests (e.g., chi-square, Comparative Fit Index (CFI), Bentler-Bonett Non-
normed Fit Index (NNFI), Root Mean Squared Error of Approximation
(RMSEA)).
81
•
Quantitative Methods-II Sixth, SEM resolves problems of multicollinearity. Multiple measures are
required to describe a latent construct (unobserved variable). Multicollinearity
cannot occur because unobserved variables represent distinct latent constructs.
Finally, a graphical language provides a convenient and powerful way to
present complex relationships in SEM. Model specification involves
formulating statements about a set of variables. A diagram, a pictorial
representation of a model, is transformed into a set of equations. The set of
equations are solved simultaneously to test model fit and estimate parameters.
17.7 CONCEPTS AND TERMINOLOGY USED IN

SEM
The key elements of essentially all structural equation models are their
parameters (often referred to as model parameters or unknown parameters).
Model parameters reflect those aspects of a model that are typically unknown
to the researcher, at least at the beginning of the analyses, yet are of potential
interest to him or her.
Parameter is a generic term referring to a characteristic of a population, such
as mean or variance on a given variable, which is of relevance in a particular
study.
A model is a statistical statement about the relations among variables.
A path diagram is a pictorial representation of a model.
A measured variable (MV) is a variables that is directly measured whereas a
latent variable (LV) is a construct that is not directly or exactly measured.
A latent variable (LV) could be defined as whatever its multiple indicators
have in common with each other. LVs defined in this way are equivalent to
common factors in factor analysis and can be viewed as being free from error
of measurement.
Direct effect is a directional relation between two variables, e.g., independent
and dependent variables.
Indirect effect is the effect of an independent variable on a dependent variable
through one or more intervening or mediating variables.
Specification is formulating a statement about a set of parameters and stating a
model.
Fit indices indicate the degree to which a pattern of fixed and free parameters
specified in the model is consistent with the pattern of variances and
covariances from a set of observed data. Examples of fit indices are chi-square,
CFI, NNFI, and RMSEA.
Components of a general structural equation model are the measurement model
and the structural model.
The measurement model prescribes latent variables, e.g., confirmatory factor
analysis. The structural model prescribes relations between latent variables
and observed variables that are not indicators of latent variables.
Identification involves the study of conditions to obtain a single, unique
solution for each and every free parameter specified in the model from the
observed data.
82
•
-II The purpose of estimation is to obtain numerical values for the unknown (free) Structural Equation
parameters. Modeling
Exogenous variable—Independent variables not presumed to be caused by

variables in the model.
Endogenous variables— variables presumed to be caused by other variables in
the model.
Recursive models assume that all causal effects are represented as
unidirectional and no disturbance correlations among endogenous variables
with direct effects between them.
Non-recursive models are those with feedback loops Model Specification—
Formally stating a model via statements about a set of parameters.
Model Identification—Can a single unique value for each and every free
parameter be obtained from the observed data.
Evaluation of Fit—Assessment of the extent to which the overall model fits or
provides a reasonable estimate of the observed data.
Model Modification—adjusting a specified and estimated model by freeing or
fixing new parameters.
17.7.1 Path Diagrams

One of the easiest ways to communicate a structural equation model is to draw
a diagram of it, referred to as path diagram, using special graphical notation. A
path diagram is a form of graphical representation of a model under
consideration. Such a diagram is equivalent to a set of equations defining a
model (in addition to distributional and related assumptions), and is typically
used as an alternative way of presenting a model pictorially. Path diagrams not
only enhance the understanding of structural equation models and their
communication among researchers with various backgrounds, but also
substantially contribute to the creation of correct command files to fit and test
models with specialized programs.
For constructing the path diagrams, widely accepted set of conventions for the
graphical display of structural equation models can be used. According to these
conventions, unobserved variables, also known as latent variables or constructs
are represented as circles or ovals and measured variables also known as
observed variables, indicators or manifest variables are represented by squares
or rectangles. Lines indicate relations between variables; lack of a line
connecting variables implies that no direct relationship has been hypothesized.
Lines have either one or two arrows. A line with one arrow represents a
hypothesized direct relationship between two variables. The variable with the
arrow pointing to it is the DV. A line with an arrow at both ends indicates a
covariance between the two variables with no implied direction of effect.
Latent variables in SEM generally correspond to hypothetical constructs or
factors, which are explanatory variables, supposed to reflect a range that is not
directly observable. An example is the construct of intelligence. There is no
single, definitive measure of intelligence. Instead, researchers use different
types of observed variables, such as tasks of verbal reasoning or memory
capacity, to assess various facets of intelligence. Latent variables in SEM can
represent a wide range of phenomena. For example, constructs about attributes
of people (e.g., intelligence, neuroticism), higher-level units of analysis (e.g.,
83
•
Quantitative Methods-II groups, geographic regions), or measures, such as method effects (e.g., self-
report, observational), can all be represented as latent variables in SEM. An
observed variable used as an indirect measure of a construct is referred to as an
indicator. The explicit distinction between factors and indicators in SEM
allows one to test a wide variety of hypotheses about measurement.
1.7.2 Classical SEM Notations

These are the c “classical” SEM notation that needs some consideration. ξ (ksi)
represents a latent construct associated with observed xi indicators, η (eta)
stands for a latent construct associated with observed yi indicators, the error
terms δ (delta) and ε (epsilon) are associated with observed xi and yi indicators,
respectively. ζ (zeta) is the error term associated with the formative construct.
λij represents factor loading in the i-th observed indicator that is explained by
84
•
-II the j-th latent construct. γij represents weight in the i-th observed indicator that Structural Equation
is explained by the j-th latent construct. Modeling
Table 17.1: Summary of abbreviations and descriptions used in the SEM study
Symbol Name Description
ξ ksi A latent construct associated with observed xi indicators
η eta A latent construct associated with observed yi indicators
δi delta The error term associated with observed xi indicators
εi epsilon The error term associated with observed yi indicators
ζ zeta The error term associated with formative construct
λij lambda Factor loading in the i-th observed indicator that is explained by the
j-th latent construct
γij gamma Weight in the i-th observed indicator that is explained by the j-th latent
construct
xi An indicator associated with exogenous construct, i.e. vector of
observed exogenous variable
yi An indicator associated with endogenous construct, i.e. vector of
observed endogenous variable.
17.8 SEM MODELS SPEPECIFICATION

There are three diffeerent SEM models based on different specifications.
17.8.1 Based on Reflective Indicators
The following figure depicts the “classical” SEM case where the model is
specified in the reflective mode known as Type A model specification. The
case illustrates a path diagram between the two latent constructs (ξ –
exogenous and η – endogenous), with three indicators per construct (xi and yi).
This case can be represented by equations 1 and 2:
1) xi = λijξ + δi
2) yi = λijη + εi
This specification assumes that the error term is unrelated to the latent variable
COV(η, εi) = 0, and independent COV(εi, εj) = 0, for i ≠ j and expected.
Reflective indicators
17.8.2 Based on Formative Indicators

Type B model specification, is known as a formative (Fornell&Bookstein
1982; cf. Edwards 2001) or causal indicator (Bollen and Lennox 1991),
85
•
Quantitative Methods-II because the direction of causality goes from the indicators (measures) to the
construct and the error term is estimated at the construct level. This type of
model specification can be represented by equations 3 and 4:
3) ξ = γijxi + ζ
4) η = γijyi + ζ
x1 x2 x3 y1 y2 y3
Formative Indicators
This specification assumes that the indicators and error term are not related, i.e.
COV(yi, ζ) = 0, and E(ζ) = 0 . Formative indicators were introduced for the
first time by Curtis and Jackson (1962) and extended by Blalock (1964). This
type of model specification assumes that the indicators have an influence on (or
that they cause) a latent construct. In other words, the indicators as a group
“jointly determine the conceptual and empirical meaning of the construct.
The type B model specification would give better explanatory power, in
comparison to the type A model specification, if the goal is the explanation of
unobserved variance in the constructs. A model is identified if model parameters
have only one group of values that create the covariance matrix. In order to resolve
the problem of indeterminacy that is related to the construct-level error term, the
formative-indicator construct must be associated with unrelated reflective
constructs. This can be achieved if the formative construct emits paths to:
i) at least two unrelated reflective indicators;
ii) at least two unrelated reflective constructs; and
iii) one reflective indicator that is associated with a formative construct and
one reflective construct.
From an empirical point of view, the latent construct captures
i) the common variance among indicators in the type A model specification; and
ii) the total variance among its indicators in the type B model specification,
covering the whole conceptual domain as an entity (cf. Cenfetelli &
Bassellier 2009; MacKenzie et al. 2005).
Reflective indicators are expected to be interchangeable and have a common
theme. Interchangeability, in the reflective context, means that omission of an
indicator will not alter the meaning of the construct. In other words, reflective
measures should be unidimensional and they should represent the common theme
of the construct. Formative indicators are not expected to be interchangeable,
because each measure describes a different aspect of the construct’s common
theme and dropping an indicator will influence the essence of the latent
86
•
-II variable. The behavior of measures of the construct with regards to the same Structural Equation
antecedents and consequences is an important criterion for the assessment of Modeling
the construct-indicator relationship. Reflective indicators are interchangeable,
which means that measures are affected by the construct, and they must have
the same antecedents and consequences. The formative constructs are affected by
the measures, thus are not necessarily interchangeable, and each measure can
represent a different theme. For the formative indicators, it is not necessary to
have the same antecedents and consequences.
Internal consistency is implied in the reflective indicators because correlation
between variables is must. High correlations among the reflective indicators are
essential, because they represent the same underlying theoretical concept. This
means that all of the items are measuring the same phenomenon within the
latent construct. On the contrary, within the formative indicators, internal
consistency is not implied because the researcher does not expect high
correlations among the measures. Because formative measures are not required
to be correlated, validity of construct should not be assessed by internal
consistency and reliability as with the reflective measures, but with other
means such as nomological and/or criterion-related validity.
17.8.3 Based on both Reflective and Formative Indicators

The researcher can create a model that uses both formative and reflective
indicators. It is possible for a structural model to have one type of latent
construct at the first-order (latent construct) level and a different type of latent
construct at the second-order level (Fornell&Bookstein 1982; MacKenzie et al.
2005; Diamantopoulos & Siguaw 2006; Wetzels et al. 2009). In other words,
the researcher can combine different latent constructs to form a hybrid model
(Edwards & Bagozzi 2000; McDonald 1996; Tenenhaus et al. 2005).
Development of this model type depends on the underlying causality between
the constructs and indicators, as well as the nature of the theoretical concept.
The researcher should model exogenous constructs in the formative mode and
all endogenous constructs in the reflective mode,
i) if one intends to explain variance in the unobservable constructs (Fornell
& Bookstein 1982; cf. Wetzels et al. 2009); and
ii) in case of weak theoretical background (Wold 1980).
Conducting a VBSEM approach in this model, using a PLS algorithm, is equal
to redundancy analysis (Fornell et al. 1988; cf. Chin 1998b), because the mean
variance in the endogenous construct is predicted by the linear outputs of the
exogenous constructs.
87
•
17.9 ISSUES IN SEM TECHNIQUE
17.9.1 Sample Size and Power
SEM is based on covariance. Covariance is less stable when estimated from
small samples. Therefore, large sample sizes are needed for SEM analyses.
Parameter estimates and chi-square tests of fit are also very sensitive to sample
size; therefore, SEM is a large sample technique. However, it may be possible
to estimate small models with fewer participants if variables are highly reliable.
In addition, although SEM is a large sample technique and test statistics are
affected by small samples, promising work has been done by Bentler and Yuan
(1999) who developed test statistics for small samples sizes.
17.9.2 Missing Data
Problems of missing data are repeatedly exaggerated in SEM due to the large
number of measured variables employed. The researcher who relies on using
complete cases only is often left with an inadequate number of complete cases
to estimate a model. Therefore missing data attribution is particularly important
in many SEM models. When there is evidence that the data are missing at
random (MAR; missing data on a variable may depend on other variables in
the dataset but the missing data does not depend on the variable itself) or
missing completely at random (MCAR; missing data is unrelated to the
variable missing data or the variables in the dataset), the EM (expectation
maximization) algorithm to obtain maximum likelihood (ML) estimates, is
appropriate method of imputing missing data. Some of the software packages
now include procedures for estimating missing data, including the EM algorithm.
EQS 6.1 (Bentler, 2004) produces the EMbased maximum likelihood solution
automatically, based on the Jamshidian–Bentler (Jamshidian & Bentler, 1999)
computations.
LISREL and AMOS also produce EM-based maximum likelihood estimates.
but, if the data are not normally distributed, maximum likelihood test
statistics— including those based on the EM algorithm—may be quite
inaccurate.Multiple imputationis is another option for treatment of missing data.
17.9.2 Multivariate Normality and Outliers
The most commonly employed techniques for estimating models assume
multivariate normality in SEM. It is always useful to examine both univariate
and multivariate normality indexes to assess normality. Univariate distributions
can be examined for outliers and skewness and kurtosis. Multivariate
distributions are examined for normality and multivariate outliers. Multivariate
normality can be evaluated through the use of Mardia’s (1970) coefficient and
multivariate outliers can be evaluated through evaluation of Mahalanobis
distance. Mardia’s (1970) coefficient evaluates multivariate normality through
evaluation of multivariate kurtosis. Mardia’s coefficient can be converted to a
normalized score (equivalent to az score), often normalized coefficients greater
than 3.00 are indicative of non-normality. Mahalanobis distance is the distance
between a case and the centroid (the multivariate mean with that data point
removed). Mahalanobis distance is distributed as a chi-square with degrees of
freedom equal to the number of measured variables used to calculate the
centroid. Therefore, a multivariate outlier can be defined as a case that is
associated with a Mahalanobis distance greater than a critical distance specified
typically by a p < .001.
88
•
-II
17.10 STEPS IN SEM Structural Equation
Modeling
Many contemporary treatments introduce SEM not just as a statistical

technique but as a process involving several stages:
1) Model specification,
2) Model estimation,
3) Data-model fit assessment, and
4) Potential model modification,
5) Reporting the Results.
17.10.1 Initial Model Specification

The first step in SEM is model conceptualization or specification. This stage
consists of gaining knowledge of the underlying theory on which the particular
model will be based. This stage begins with stating the hypothesis in both
diagrams and equation form then statistically identifying the model and then
evaluating the statistical assumptions that underlie the model.
In this phase the model is specified through a diagram. The relations in the
diagram are directly translated into equations and the model is then estimated.
One method of model specification is the Bentler–Weeks method (Bentler&
Weeks, 1980). In this method every variable in the model, latent or measured,
is either an IV or a DV. The parameters to be estimated are the (a) regression
coefficients, and (b) the variances and the covariances of the independent
variables in the model (Bentler, 2001). In the Bentler–Weeks model only
independent variables have variances and covariances as parameters of the
model.
A normally tricky and often confusing topic in SEM is identification. In SEM,
a model is specified, parameters (variances and covariances of IVs and
regression coefficients) for the model are estimated using sample data, and the
parameters are used to produce the estimated population covariance matrix.
However, only models that are identified can be estimated. A model is said to
be identified if there is a unique numerical solution for each of the parameters
in the model. The first step consists of counting the number of data points and
the number of parameters that we want to estimate. The data in SEM are the
variances and covariances in the sample covariance matrix. The number of data
points is the number of non redundant sample variances and covariances,
where p equals the number of measured variables. The number of parameters is
established by adding together the number of regression coefficients, variances,
and covariances that are to be estimated. There should be more data points than
parameters to be estimated for the estimation of model.
Hypothesized models are said to be over identified if there are more data than
parameters to be estimated. The model is said to be just identified if, there are
the same numbers of data points as parameters to be estimated. In this case, the
estimated parameters perfectly reproduce the sample covariance matrix and the
chi-square test statistic and degrees of freedom are equal to zero, hypotheses
about the adequacy of the model cannot be tested. However, hypotheses about
specific paths in the model can be tested. The model is said to be under
identified if there are fewer data points than parameters to be estimated, and
parameters cannot be estimated. The number of parameters needs to be reduced
89
•
Quantitative Methods-II by fixing, constraining, or deleting some of them. A parameter may be fixed by
setting it to a specific value or constrained by setting the parameter equal to
another parameter. Examining the measurement portion of the model is the
second step in determining model identifiability. The measurement part of the
model deals with the relationship between the measured indicators and the
factors.
Factors, are hypothetical constructs and consist of common variance, as such
they have no intrinsic scale and therefore need to be scaled. Measured variables
have scales, for example, income may be measured in dollars or weight in
pounds. To establish the scale of a factor, the variance for the factor is fixed to
1.00, or the regression coefficient from the factor to one of the measured
variables is fixed to 1.00. Fixing the regression coefficient to 1 gives the factor
the same variance as the common variance portion of the measured variable. If
the factor is an IV, either alternative can be applied. If the factor is a DV the
regression coefficient is set to 1.00. Not setting the scale of a latent variable is
easily the most common error made when first identifying a model. Next, to
establish the identifiability of the measurement portion of the model the
number of factors and measured variables are examined. If there is only one
factor, the model may be identified if the factor has at least three indicators
with nonzero loading and the errors (residuals) are uncorrelated with one
another. If there are two or more factors, again consider the number of
indicators for each factor.
If each factor has three or more indicators, the model may be identified if errors
associated with the indicators are not correlated, each indicator loads on only
one factor and the factors are allowed to co-vary. If there are only two
indicators for a factor, the model may be identified if there are no correlated
errors, each indicator loads on only one factor, and none of the covariances
among factors is equal to zero.
17.10.2 Model Estimation

The next step in SEM is to evaluate and estimate population parameters. The
purpose of estimation is to minimize the difference between the structured and
unstructured estimated population covariance matrices.
F = (s – s(Q))W(s – s(Q)),
To accomplish this goal a function, F, is minimized where, s is the vector of
data (the observed sample covariance matrix stacked into a vector); s is the
vector of the estimated population covariance matrix (again, stacked into a
vector) and(Q) indicates that s is derived from the parameters (the regression
coefficients, variances and covariances) of the model. W is the matrix that
weights the squared differences between the sample and estimated population
covariance matrix.
In EFA the observed and reproduced correlation matrices are compared. This
idea is extended in SEM to include a statistical test of the differences between
the estimated structured and unstructured population covariance matrices. If the
weight matrix, W, is chosen correctly, at the minimum with the optimal, F
multiplied by (N – 1) yields a chi-square test statistic.
There are many different estimation techniques in SEM, these techniques vary
by the choice of W.
90
•
-II • Maximum likelihood (ML) is usually the default method in most programs Structural Equation
because it yields the most precise (smallest variance) estimates when the Modeling
data are normal.
• Generalized least squares (GLS) has the same optimal properties as ML
under normality.
• The asymptotically distribution free (ADF) method has no distributional
assumptions and hence is most general, but it is impractical with many
variables and inaccurate without very large sample sizes.
Satorra and Bentler (1988, 1994, and 2001) and Satorra (2000) also developed
an adjustment for non-normality that can be applied to the ML, GLS, or EDT
chi-square test statistics. Briefly, the Satorra–Bentler scaled χ2 is a correction
to the χ2 test statistic (Satorra & Bentler, 2001). EQS also corrects the standard
errors for parameter estimates to adjust for the extent of non-normality (Bentler
& Dijkastra, 1985).
Sample size and plausibility of the normality and independent assumptions
must be considered while choosing the suitable estimation technique. ML, the
Scaled ML, or GLS estimators may be good choices with medium (over 120)
to large samples and evidence of the plausibility of the normality assumptions.
ML estimation is currently the most frequently used estimation method in
SEM. In medium (over 120) to large samples the scaled ML test statistic is a
good choice with non-normality or suspected dependence among factors and
errors. In small samples (60 to 120) the Yuan–Bentler test statistic appears
best. The test statistic based on the ADF estimator (without adjustment) is
inappropriate choice under all conditions unless the sample size is very large (>
2,500).
17.10.3 Model Evaluation

A central issue addressed by SEM is how to assess the fit between observed
data and the hypothesized model ideally operationalized as an evaluation of the
degree of discrepancy between the true population covariance matrix and that
implied by the model’s structural and non structural parameters. As the
population parameter values are rarely identified, the difference between an
observed, sample-based covariance matrix and that implicit by parameter
estimates must serve to approximate the population discrepancy. The observed
data will fit the model perfectly for a just identified model. The system of
equations expressing each model parameter as a function of the observed
(co)variances is uniquely solvable; thus, the sample estimate of the model-
implied covariance matrix will, by default, equal the sample estimate of the
population covariance matrix. However, if a model is over identified, it is
unlikely that these two matrices are equal as the system of equations
(expressing model parameters as functions of observed variances and
covariances) is solvable in more than a single way.
Abiding by a universal desire for parsimony, over identified models tend to be
of more substantive interest than just identified ones because they characterize
simpler potential explanations of the observed associations. While data-model
fit for such models was initially conceived as a formal statistical test of the
discrepancy between the true and model-implied covariance matrices, such a
test now is often viewed as overly strict given its power to detect even trivial
deviations of a proposed model from reality. Hence, many alternative
91
•
Quantitative Methods-II assessment strategies have emerged and continue to be developed. Data-model
fit indices for such assessment can be categorized roughly into three broad
classes:
• Absolute indices assess the overall discrepancy between observed and
implied covariance matrices; fit improves as more parameters are added to
the model and degrees of freedom decrease: for example, the standardized
root mean square residual (SRMR), the chi-square test(recommended to be
reported mostly for its historical significance), and the goodness-of-fit
index (GFI).
• Parsimonious indices assess the overall discrepancy between observed and
implied covariance matrices while taking into consideration a model’s
complexity; fit improves as more parameters are added to the model, as
long as those parameters are making a useful contribution: for example, the
root mean square error of approximation(RMSEA) with its associated
confidence interval, the Akaike information criterion (AIC) for fit
comparisons across non-nested models, and the adjusted goodness-of-fit
index (AGFI).
• Incremental indices assess absolute or parsimonious fit relative to a
baseline model, usually the null model (a model that specifies no relations
among measured variables): for example, the comparative fit index (CFI),
the normed fit index (NFI), and the non-normed fit index (NNFI).
SEM software programs routinely report a handful of goodness-of-fit indices.
Some of these indices work better than others under certain conditions. It is
generally recommended that multiple indices be considered simultaneously
when overall model fit is evaluated. For instance, Hu and Bentler (1999)
proposed a 2-index strategy that is, reporting SRMR along with one of the fit
indices (e.g., RNI, CFI, or RMSEA). The authors also suggested the following
criteria for an indication of good model-data fit using those indices: RNI (or
CFI) ≥ .95, SRMR ≤ .08, and RMSEA ≤ .06. Despite the sample size
sensitivity problem with the chi-square test, reporting the model chi-square
value with its degrees of freedom in addition to the other fit indices is
recommended. If, after considering several indices, data model fit is deemed
acceptable (and judged best compared to competing models, if applicable),the
model is reserved as rational, and individual parameters may be interpreted. If,
however, evidence suggests unacceptable data-model fit, the next and often
final stage in the SEM process is considered: modifying the model to improve
fit in hopes of also improving the model’s correspondence to reality.
17.10.4 Model Modification

Any hypothesized model is only an approximation to reality; the question
remains is one of degree of that misspecification. With regard to external
specification errors—when irrelevant variables are included in the model or
substantively important ones remain left out—remediation can only occur by
respecifying the model based on more relevant theory. On the other hand,
internal specification errors—when unimportant paths among variables were
included or when important paths were omitted—can potentially be diagnosed
and remedied using Wald statistics (predicted increase in chi-square if a
previously estimated parameter were fixed to some known value, e.g., zero)
and Lagrange multiplier statistics (also referred to as modification indices;
estimated decrease in chi-square if a previously fixed parameter were to be
estimated). All are asymptotically equivalent under the null hypothesis but
92
•
-II approach model modification differently. As these tests’ recommendations are Structural Equation
directly motivated by the data and not by theoretical considerations, any Modeling
resulting respecifications must be viewed as exploratory in nature and might
not lead to a model that resembles reality any more closely than the one(s)
initially conceptualized.
Chi-square difference test
If models are nested, that is, models are subsets of each other, the χ2 value for
the larger model is subtracted from the χ2 value for the smaller, nested model
and the difference, also χ2, is evaluated with degrees of freedom equal to the
difference between the degrees of freedom in the two models. When the data
are normally distributed, the chi-squares can simply be subtracted. However,
when the data are non normal and the Satorra–Bentler scaled chi-square is
employed, an adjustment is required (Satorra, 2000; Satorra &Bentler, 2001) so
that the S–B χ2 difference test is distributed as a chi-square.
Chi-square difference tests are popular methods for comparison of nested
models especially when, as in this example, there are two nested a priori
models of interest; however, a potential problem arises when sample sizes are
small. Because of the relationship between sample size and χ2, it is hard to
detect a difference between models with small sample sizes.
Lagrange Multiplier (LM) test
The LM test also compares nested models but requires estimation of only one
model. The LM test asks would the model be improved if one or more of the
parameters in the model that are currently fixed were estimated. Or, equivalently,
what parameters should be added to the model to improve the fit of the model?
There are many approaches to using the LM tests in model modifications. It is
possible and often desirable to look only at certain parts of the model for
possible change although it is also possible to examine the entire model for
potential additions.
The LM test can be examined either univariately or multivariately. There is a
danger in examining only the results of univariate LM tests because overlapping
variance between parameter estimates may make several parameters appear as if
their addition would significantly improve the model. All significant parameters
are candidates for inclusion by the results of univariate LM tests, but the
multivariate LM test identifies the single parameter that would lead to the largest
drop in the model χ2 and calculates the expected change in χ2 if this parameter
was added. After this variance is removed, the next parameter that accounts for the
largest drop in model χ2 is assessed, similarly. After a few candidates for
parameter additions are identified, it is best to add these parameters to the model
and repeat the process with a new LM test, if necessary.
Wald test
While the LM test asks which parameters, if any, should be added to a model,
the Wald test asks which, if any, could be deleted. Are there any parameters
that are currently being estimated that could instead be fixed to zero? Or,
equivalently, which parameters are not necessary in the model? The Wald test
is analogous to backward deletion of variables in stepwise regression, in which
one seeks a non significant change in R2 when variables are left out. However,
if the goal had been the development of a parsimonious model, the Wald test
could have been examined to evaluate deletion of unnecessary parameters.
93
•
Quantitative Methods-II Finally, no matter how complex the model under study, it is always
conceivable that other variables or causal assumptions that the researcher had
failed to consider can account for the observed relationships. To enhance the
theoretical value of their investigations, researchers are well advised to
consider also models that are almost equivalent to the one reported. These are
models that fit the analyzed data set almost as well as the reported model and
that may offer interesting alternative theoretical interpretations.
Comparison of Nested Models
In addition to evaluating the overall model fit and specific parameter estimates,
it is also possible to statistically compare nested models to one another. Nested
models are models that are subsets of one another. When theories can be
specified as nested hypotheses, each model might represent a different theory.
These nested models are statistically compared, thus providing a strong test for
competing models. Some of the common variance among the items may be due
to wording and the general domain area (e.g., health problems, relationships
problems), not solely due to the underlying constructs. We can compare our
model to a model that also includes paths that account for the variance
explained by the general domain or wording of the item. The model with the
added paths to account for this variability would be considered the full model.
To test this hypothesis, the chi-square from the model with paths added to
account for domain and wording would be subtracted from the chi-square for
the nested model in that does not account for common domains and wording
among the items. The corresponding degrees of freedom for these two models
would also be subtracted. Given nested models and normally distributed data,
the difference between two chi-squares is a chi-square with degrees of freedom
equal to the difference in degrees of freedom between the two models. The
significance of the chi-square difference test can then be assessed in the usual
manner. If the difference is significant, the fuller model that includes the extra
paths is needed to explain the data. If the difference is not significant, the
nested model, which is more parsimonious than the fuller model, would be
accepted as the preferred model.
17.10.5 Reporting the Results

The final step is to precisely and completely describe the analysis in written reports.
How authors structure the results section obviously is dictated by the particular
model(s) and research questions under study. It is the researcher’s responsibility to
provide access to data in order to facilitate verification of the obtained results: If
moment-level data were analyzed, a covariance matrix (or correlation matrix with
standard deviations) should be offered in a table or appendix; if raw data were used,
information on how to obtain access should be provided.
Results from both the measurement and structural phases should be presented.
Judging the overall quality of a hypothesized model usually is presented early
in the results section or overidentified models. Given that available data-model
fit indices can lead to inconsistent conclusions, researchers should consider fit
results from different classes so readers can arrive at a more complete picture
regarding an model’s acceptability. Also, a comparison of fit across multiple, a
priori specified alternative models can help in weighing the relative merits of
preferring one model over others. As illustrated, when competing models are
nested, a formal chi-square difference test is available to judge if a more
restrictive-but also more parsimonious-model can elucidate the observed data
equally well, without a considerable loss in data-model fit (alternative models that
94
•
-II are not nested have traditionally been compared only descriptively—relative Structural Equation
evaluations of values are recommended, with smaller values indicating better fit— Modeling
but recent methodological developments suggest statistical approaches as well).
If post hoc model modifications are performed following unacceptable data-model
fit from either the measurement or structural phase of the analysis, authors should
give a thorough account of the nature and reasons (both statistical and
theoretical)for the respecification(s), including summary results from Lagrange
multiplier tests and revised final fit results. If data-model fit has been assessed and
believed to be satisfactory, with or without respecification ,more detailed results
are offered, usually in the form of individual unstandardized and standardized
parameter estimates for each structural equation of interest, together with
associated standard errors and/or test statistics and coefficients of determination.
When latent variables are part of a model, estimates of their construct reliability
should be presented, with values ideally falling above .70 or .80.
Researcher should provide a logic of what implications the results from the SEM
analysis have on the theory or theories that gave rise to the initial model(s). Claims
that a well-fitting model was “confirmed” or that a particular theory was proven to
be “true,” especially after post hoc respecifications, should be avoided. Such
statements are greatly deceptive given that alternative, structurally different, yet
mathematically equivalent models always exist that would produce identical data-
model fit results and thus would explain the data equally well. At most, a model
with acceptable fit may be interpreted as one reasonable explanation for the
associations observed in the data. From this perspective, a SEM analysis should be
evaluated from a disconfirmatory, rather than a confirmatory perspective. Based
on unacceptable data model fit results, theories can be shown to be false but not
proven to be true by acceptable data-model fit.
If evidence of data-model misfit was presented and a model was modified
based on statistical results from Lagrange multiplier tests, readers must be
made aware of potential model overfitting and the capitalization on chance.
Statistically rather than theoretically based respecifications are purely
exploratory and might say little about the true model underlying the data.
While some model modifications seem appropriate and theoretically justifiable
(usually, minor respecifications of the measurement portion are more easily
defensible than those in the structural portion of a model), they only address
internal specification errors and should be cross-validated with data from new
and independent samples. Finally, the interpretation of individual parameter
estimates can involve explicit causal language, as long as this is done from
within the context of the particular causal theory proposed and the
possibility/probability of alternative explanations is raised unequivocally.
Though some might disagree, we think that explicit causal statements are more
honest than implicit ones and are more useful in articulating a study’s practical
implications; after all, is not causality the ultimate aim of science.
In the end, SEM is a powerful disconfirmatory tool at the researcher’s disposal
for testing and interpreting theoretically derived causal hypotheses from within
an a priori specified causal system of observed and/or latent variables.
However, we urge authors to resist the apparently still popular belief that the
main goal of SEM is to achieve satisfactory data-model fit results; rather, it is
to get one step closer to the “truth.” If it is true that a proposed model does not
reflect reality, then reaching a conclusion of misfit between data and model
should be a desirable goal, not one to be avoided by careless respecifications
until satisfactory levels of fit are achieved.
95
•
17.11 AN EXAMPLE OF SEM
For example, we take some fictitious data in which anxiety was measured. We
performed an confirmatory factor analysis in which we defined anxiety as
consisting of four components, namely, affective, behavioral, cognitive and
physiological symptoms. Thus, we specified a model which looked like as the
figure below:
C1 fearfulness
C2 obsession
cognitive
Poor
C3 concentratio
n
A1 sadness
A2 anger
affective
A3 reduced
QOL
ANX
B1 avoidance
behavioural
compulsion
B2
isolation
B3
heart rate
p1
physiological
dry mouth
p2
sweating
P3
96
•
-II The next step is to estimate the parameters of the model. The software Structural Equation
performs this step and provides final results of parameters that produce best fit Modeling
to data possible.
.40
.60
fearfulness
C1
.35
0.63
.65C .35 cognitive
0.6
obsession
C2 0.6
.04
.96 0.21 0.73
C3
poor
concentration
.03
.48 .52 ANX

.97
sadness
0.72 0.24
.97
.7
.03 affective
0.18
A2 anger
.7
0.55
.3
A3 reduced
QOL
Figure 2 shows only 2 of the 4 components of anxiety. Above each arrow from
a latent variable to a variable is a number, called the path coefficient. This can
range from –1.0 to 1.0, with higher numbers (positive or negative) showing a
stronger association. As can be seen, the variable “obsessions” doesn’t fit very
well with the cognitive trait; and the variable “anger” doesn’t seem to go with
the affective trait. The numbers over the variable names are the SMRs, which
are the squared values of the path coefficients; they are interpreted in the same
way as R2 multiple regression-in terms of how much of the variance in one
variable is explained by, or are in common with, the other variable. Finally, the
numbers over the arrows between the error terms and the variables are the
variances of the errors. You’ll note that the sum of the SMR plus the error
variance for each variable is 1.0; that is, all the variance of a variable is divided
between that shared with the latent variable and error. The cognitive domain is
correlated more highly with the latent trait of anxiety than is the affective
realm, as reflected in their respective path coefficients.
The next step is model specification stage. Although no mathematics is
involved, it is probably the most difficult—and most important—part. It is the
most difficult because it requires the most thought and understanding of the
97
•
Quantitative Methods-II theoretical model of the purported influences. No computer program can help
us at this stage, only our knowledge of the field. It is the most important step
because everything depends on how well we specify the model. The computer
programs may help us in determining whether some variables aren’t important,
we can play with different paths to see whether they improve the model or not.
However, the primary cause of poorly fitting is the omission of crucial
variables.
The next step is relatively easy. The main focus is on the paths—from the
latent variable to the measured variables and from that latent variable to the
next one in the path. Once we have cleaned up the model by pruning
noncontributory paths, we can examine the fit of the overall model. The most
common index of how well the data match the model is RMSEA. it sees how
much the data deviate from the model. Values over 0.10 are considered to be a
bad fit, those less than 0.08 reflect a reasonable fit, and values less than 0.05
indicate a good fit.
Given is the table showing RMSEA of the model:
RMSEA
Model RMSEA LO 90 HI 90 PCLOSE
Independence model .100 .087 .115 .000
PCLOSE = .000 for the Independence model. Under the hypothesis of “close
fit” (i.e., that RMSEA is no greater than .05 in the population), the probability
of getting a sample RMSEA as large as .100 is .000.
17.12 SOFTWARE PROGRAMS FOR SEM

Most SEM analyses are conducted using one of the specialized SEM software
programs. However, there are many options, and the choice is not always easy.
Below is a list of the commonly used programs for SEM. Special features of
each program are briefly discussed. It is important to note that this list of
programs and their associated features is by no means comprehensive. This is a
rapidly changing area and new features are regularly added to the programs.
17.12.1 LISREL
LISREL (linear structural relationships) is one of the earliest SEM programs
and perhaps the most frequently referenced program in SEM articles. Its
version 9.1 has three components: PRELIS, SIMPLIS, and LISREL. PRELIS
(pre-LISREL) is used in the data preparation stage when raw data are available.
Its main functions include checking distributional assumptions, such as
univariate and multivariate normality, imputing data for missing observations,
and calculating summary statistics, such as Pearson covariances for continuous
variables, polychoricor polyserial correlations for categorical variables, means,
or asymptotic covariance matrix of variances and covariances (required for
asymptotically distribution-free estimator or Satorra and Bentler’s scaled chi-
square and robust standard errors). PRELIS can be used as a stand-alone
program or in conjunction with other programs.
Summary statistics or raw data can be read by SIMPLIS or LISREL for the
estimation of SEM models. The LISREL syntax requires understanding of
98
•
-II matrix notation while the SIMPLIS syntax is equation-based and uses variable Structural Equation
names defined by users. Both LISREL and SIMPLIS syntax can be built Modeling
through interactive LISREL by entering information for the model construction
wizards. Alternatively, syntax can be built by drawing the models on the Path
Diagram screen.
LISREL 9.1 allows the analysis of multilevel models for hierarchical data in
addition to the core models. A free student version of the program, which has
the same features as the full version but limits the number of observed
variables to12, is available from the web site of Scientific
SoftwareInternational,Inc. (http://www.ssicentral.com). This web site also
offers a list of illustrative examples of LISREL’s basic and new features.
17.12.2 EQS
Version 6.1 of EQS (Equations) provides many general statistical functions
including descriptive statistics, t-test, ANOVA, multiple regression,
nonparametric statistical analysis, and EFA. Various data exploration plots,
such as scatter plot, histogram, and matrix plot are readily available in EQS for
users to gain intuitive insights into modelling problems. Similar to LISREL,
EQS allows different ways of writing syntax for model specification.
The program can generate syntax through the available templates under the
“Build_EQS” menu, which prompts the user to enter information regarding the
model and data for analysis, or through the Diagrammer, which allows the user
to draw the model. Unlike LISREL, however, data screening (information
about missing pattern and distribution of observed variables) and model
estimation are performed in one run in EQS when raw data are available.
Model-based imputation that relies on a predictive distribution of the missing
data is also available in EQS. Moreover, EQS generates a number of alternative
model chi-square statistics for non normal or categorical data when raw data
are available. The program can also estimate multilevel models for hierarchical
data. Visit http://www.mvsoft.com for a comprehensive list of EQS’s basic
functions and notable features.
17.12.3 Mplus
Version 7.3 of the Mplus program includes a Base program and three add-on
modules. The Mplus Base program can analyze almost all single-level models
that can be estimated by other SEM programs. Unlike LISREL or EQS, Mplus
version 7.33 is mostly syntax-driven and does not produce model diagrams.
Users can interact with the Mplus Base program through a language generator
wizard, which prompts users to enter data information and select the estimation
and output options. Mplus then converts the information into its program-
specific syntax. However, users have to supply the model specification in
Mplus language themselves.
Mplus Base also offers a robust option for non-normal data and a special full-
information maximum likelihood estimation method for missing data (see
footnote 4). With the add-on modules, Mplus can analyze multilevel models
and models with latent categorical variables, such as latent class and latent
profile analysis. The modelling of latent categorical variables in Mplus is so far
unrivalled by other programs. The official web site of Mplus
(http://www.statmodel.com) offers a comprehensive list of resources including
99
•
Quantitative Methods-II basic features of the program, illustrative examples, online training courses,
and a discussion forum for users.
17.12.4 Amos
Amos (Analysis of Moment Structure) version 22 is distributed with IBM
SPSS. It has two components: Amos Graphics and Amos Basic. Similar to the
LISREL Path Diagram and SIMPLIS syntax, respectively, Amos Graphics
permits the specification of models by diagram drawing whereas Amos Basic
allows the specification from equation statements. A notable feature of Amos is
its capability for producing bootstrapped standard error estimates and
confidence intervals for parameter estimates.
An alternative full-information maximum likelihood estimation method for
missing data is also available in Amos. The program is available at
http://www.smallwaters.com or http://www.spss.com/amos/.
17.12.5 Mx
Mx (Matrix) version 6 is a free program downloadable from
http://www.vcu.edu/mx/. The Mx Graph version is for Microsoft Windows
users. Users can provide model and data information through the Mx
programming language. Alternatively, models can be drawn in the drawing
editor of the Mx Graph version and submitted for analysis. Mx Graph can
calculate confidence intervals and statistical power for parameter estimates.
Like Amos and Mplus, a special form of full-information maximum likelihood
estimation is available for missing data in Mx.
17.12.6 Others
In addition to SPSS, several other general statistical software packages offer
built-in routines or procedures that are designed for SEM analyses. They
include the CALIS (covariance analysis and linear structural equations)
procedure of SAS (SAS Institute Inc., 2000; http://www.sas.com/), the
RAMONA (Reticular Action Model Or Near Approximation) module of
SYSTAT (Systat Software, Inc., 2002;http://www.systat.com/), and SEPATH
(structural equation modeling and path analysis) of Statistica (StatSoft,
Inc.,2003; http://www.statsoft.com/products/advanced.html).
17.13 ADVANTAGES AND DISADVANTAGES OF

SEM
17.13.1 Advantages of SEM
1) SEM allows researcher to identify direct and indirect effects. Direct effects
are principally the relationship between a dependent variable we are
interested in explaining and an independent variable we think is causing or
related to the dependent variable. Thus this establishes links directly from a
cause to an effect. An indirect effect occurs when one variable goes through
another variable on the way to some dependent or independent variable.
2) It is possible to have multiple indicators of a concept. With multiple
regression analysis we only evaluate the effect of individual independent
variables on the dependent variable. It creates farily a simple path model
100
•
-II while SEM allows for the estimation of the combined effects of Structural Equation
independent variable into concepts/ construct. Modeling
3) SEM includes measurement error into the model. A problem for path
analyses using regression as its underlying analysis not adequately control
for measurement error. SEM analyses do account for measurement error,
therefore providing a better understanding of how good the theoretical
model predicts actual behaviour.
17.13.2 Disadvantages of SEM

1) Most of the procedures that have been suggested involve nonstandard and
complex model specifications that are challenging for the average user and
thus susceptible to error. Indeed, errors have even been noted in the
specifications developed by SEM specialists.
2) Because products of normally distributed observed and latent variables are
themselves not normally distributed, standard errors and estimates of fit
might not be accurate. This problem will be more severe to the extent that
the latent exogenous variables used to form the product term are highly
correlated.
3) If the latent variables that denote main effects are not normally distributed,
the parameter estimates yielded by several procedures are not consistent.
4) Most of the methods proposed in the literature are applicable to a restricted
class of measurement models.
5) Although a number of alternative procedures and options have been
proposed, the selection of an optimal approach is made difficult by the
absence of any one study or set of studies that compares all the viable
alternatives that have been proposed to date under a variety of conditions.
6) Some of the more promising approaches are not easily available in
conventional SEM software.
17.14 LET US SUM UP

SEM techniques are considered today to be a major component of applied
multivariate statistical analysis and are used by education researchers,
economists, marketing researchers, medical researchers, and variety of other
social and behavioural scientists. SEM is a collection of statistical techniques
that allow a set of relations between one or more independent variables(IVs),
and one or more dependent variables (DVs). SEM is an attempt to model
causal relations between variables by including all variables that are known to
have some involvement in the process of interest. It can be viewed as a
combination of factor analysis and regression or path analysis. Structural
equation modeling (SEM) uses various types of models to depict relationships
among observed variables, with the same basic goal of providing a quantitative
test of a theoretical model hypothesized by the researcher. Two goals in SEM
are (1) to understand the patterns of correlation/covariance among a set of
variables and (2) to explain as much of their variance as possible with the
model specified.
101
•
Quantitative Methods-II To discuss the history of structural equation modeling, it is better to explain the
chronological order of following four models : regression, path, confirmatory
factor, and structural equation models.
One of the easiest ways to communicate a structural equation model is to draw
a diagram of it, referred to as path diagram, using special graphical notation. A
path diagram is a form of graphical representation of a model under
consideration. Such a diagram is equivalent to a set of equations defining a
model (in addition to distributional and related assumptions), and is typically
used as an alternative way of presenting a model pictorially. Path diagrams not
only enhance the understanding of structural equation models and their
communication among researchers with various backgrounds, but also
substantially contribute to the creation of correct command files to fit and test
models with specialized programs. Five steps are involed in SEM: Initial
model conceptualization, Model estimation, Data-model fit assessment,
Potential model modification and Reporting the Results. Most SEM analyses
are conducted using one of the specialized SEM software programs. These
include LISREL, EQS, Mplus, Amos and Mx. The SEM has certain advantages
and disadvantages also.
17.15 KEY WORDS

SEM : Collection of statistical techniques that allow a set
of relations between one or more independent
variables (IVs), and one or more dependent
variables (DVs).
Path Diagrams : Pictorial representation of a model.
Observed variables : Variables that are directly measured.
Unobserved : Latent variable in terms of multiple indicators that
variable are in common with each other.
Residual or error : Can be associated with either observed variables or
terms factors specified as outcome (dependent) variables.
Measurement : The part of the model that relates the measured
model variables to the factors.
ξ (ksi) : Latent construct associated with observed xi
indicators.
η (eta) : Latent construct associated with observed yi
indicators.
δ (delta) : Error term associated with observed xi indicators.
ε (epsilon) : Error term associated with observed xi, yi
indicators.
ζ (zeta) : Error term associated with the formative construct.
Λij : Factor loading in the i-th observed indicator that is
explained by the j-th latent construct.
Γij : Weight in the i-th observed indicator that is
explained by the j-th latent construct.
102
•
-II Reflective indicator : The construct is the cause of the observed Structural Equation
measures, so a variation in the construct leads to a Modeling
variation in all its measures.
Formative or : The direction of causality goes from the indicators
causal indicator (measures) to the construct and the error term is
estimated at the construct level.
Model specification : Involves using all available relevant theory,
research, and information to construct the
theoretical model.
Model : A model is said to be identified if there is a unique
identification numerical solution for each of the parameters in the
model.
Underidentified : If one or more parameters cannot solved uniquely
on the basis of the covariance matrix S (more
unknown parameters than equations).
Just-identified : There is just enough information in the covariance
matrix S to solve the parameters of the model
(equally number of equations as unknown
parameters).
Over-identified : More than one way to estimate the unknown
parameters (more equations than unknowns).
Model estimation : Minimize the difference between the structured and
unstructured estimated population covariance
matrices.
Model evaluation : Assess the fit between observed data and the
hypothesized model ideally operationalized as an
evaluation of the degree of discrepancy between the
true population covariance matrix and that implied
by the model’s structural and non structural
parameters.
Model modification : Respecifying the model based on more relevant
theory.
Nested models : Models that are subsets of one another.
17.16 REFERENCES
Anderson, J.C., & Gerbing, D.W. (1988). Structural Equation Modeling in
Practice: A Review and Recommended Two-Step approach. Psychological
Bulletin, 103(3), 411-423.
Bentler, P.M. (1990). Comparative fit indexes in structural models. Psychology
Bulletin, 107, 256–259.
Bentler, P.M. (2001). EQS 6 structural equations program manual.
Encino,CA: Multivariate Software.
Bentler, P. M., & Dijkstra, T. (1985). Efficient estimation via linearization
instructural models. In P. R. Krishnaiah (Ed.), Multivariate analysis 6 (pp.9–
42). Amsterdam: North-Holland.
103
•
Quantitative Methods-II Bentler, P. M., & Raykov, T. (2000). On measures of explained variance
innonrecursive structural equation models. Journal of Applied Psychology, 85,
125–131.
Bentler, P. M., &Weeks, D. G. (1980). Linear structural equation with latent
variables. Psychometrika, 45, 289–308.
Bentler, P. M., & Yuan, K.-H. (1999). Structural equation modeling withsmall
samples: Test statistics. Multivariate Behavioral Research, 34,181–197.
Fox, J. (2006). Structural equation modelling with the sem package in R.
Structural equation modelling, 13(3), 465-486.
Hu, L., & Bentler, P.M. (1998). Fit indices in covariance structural equation
modeling: Sensitivity to underparameterized model misspecification.
Psychological Methods, 3, 424–453.
Hu, L., &Bentler, P. M. (1999). Cut off criteria for fit indexes in covariance
structure analysis: Conventional criteria versus new alternatives. Structural
Equation Modeling, 6, 1–55.
Kline RB. (1998). Principles and Practice of Structural Equation Modeling.
New York: Guilford
Lei, P., & Wu, Q. (2007). Introduction to Structural Equation Modeling: Issues
and Practical Considerations. Instructional topics in educational
measurement, fall, 33-43.
Paxton, P., Curran, P.J., Bollen K.A., Kirby, J.,Chen, F.(2001). Monte carlo
experiments: design And implementation. Structural Equation
Modeling8:288–312
Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample
analysisof moment structures. In D. D. H. Heijmans, D. S. G. Pollock, &
A.Satorra (Eds.), Innovations in multivariate statistical analysis: Afestschrift
for Heinz Neudecker (pp. 233–247). Dordrecht, The Netherlands: Kluwer
Academic.
Satorra, A., & Bentler, P. M. (1988). Scaling corrections for chi-square
statistics in covariance structure analysis. Proceedings of the American
Statistical Association, 308–313.
Satorra, A., &Bentler, P. M. (1994). Corrections to test statistics and standard
errors in covariance structure analysis. In A. von Eye & C. C. Clogg (Eds.),
Latent variables analysis: Applications for developmental research (pp. 399–
419). Thousand Oaks, CA: Sage.
Satorra, A., &Bentler, P. M. (2001). A scaled difference chi-square test
statisticfor moment structure analysis. Psychometrika, 66, 507–514.
Tomarken, A. J., & Waller, N. G. (2005). Structural equation modeling:
strengths, limitations, and misconceptions. Annual review of clinical
psychology, 1, 31-65.
Ullman, J. B., (2006). Structural Equation Modeling: Reviewing the Basics and
Moving Forward. Journal of personality assessment, 87(1), 35–50.
Valluzi, J.L., Larson, S.L., Miller, G.E. (2003). Indications and Limitations of
Structural Equation Modeling in Complex Surveys: Implications for an
Application in the Medical Expenditure Panel Survey (MEPS). Joint Statistical
Meetings - Section on Survey Research Methods.
104
•
MEC-109
Research Methods
Indira Gandhi
in Economics
Block
5
QUALITATIVE METHODS
UNIT 18
Participatory Method 5
UNIT 19
Content Analysis 21
UNIT 20
Action Research 37
Expert Committee
University of Hyderabad, Hyderabad University of Mumbai
Mumbai
Prof. Harishankar Asthana
Professor of Psychology Dr. Manish Gupta
Banaras Hindu University, Varanasi Sr. Economist
National Institute of Public
Prof. Chandan Mukherjee
Finance and Policy
Professor and Dean
New Delhi
School of Development Studies
Ambedkar University, New Delhi Prof. Anjila Gupta
Prof. V.R. Panchmukhi
IGNOU, New Delhi
Rtd. Professor of Economics
Bombay University and Former Prof. Narayan Prasad (Convenor)
Chairman ICSSR, New Delhi Professor of Economics
IGNOU, New Delhi
Prof. Achal Kumar Gaur
Professor of Economics Prof. K. Barik
Faculty of Social Sciences Professor of Economics
Banaras Hindu University, Varanasi IGNOU, New Delhi
Professor, Indian Institute of Public Associate Professor in Economics
Administration, New Delhi IGNOU, New Delhi
Units Resource Person IGNOU Faculty
(Format, Language and Conent Editing)
18 Prof. Shalina Mehta
Professor Prof. Narayan Prasad
Department of Anthropology Professor of Economics
Panjab University IGNOU, New Delhi
Chandigarh
Prof. Narayan Prasad
19-20 Prof. H.S. Asthana
Professor
IGNOU, New Delhi
Department of Psychology
Banaras Hindu University
Varanasi
Print Production
Mr. Manjit Singh
October, 2015
 Indira Gandhi National Open University, 2015
ISBN-978-81-266-
Printed at :
BLOCK 5 QUALITATIVE METHODS
As explained in Block 1 of this course, the major difference between the
quantitative approach and qualitative approach is not the type of data used or
preferred but is much broader and deeper. The quantitative research is associated
with positivist and post-positivist paradigm whereas qualitative research is
associated with interpretative paradigm and critical theory paradigm.
Notwithstanding with the debate about quantitative research vs. qualitative

research, a researcher may use either quantitative methods or qualitative methods
or mixed methods blending the techniques of the two. In Block 3 and 4 we have
covered various quantitative methods of data analysis. In this block you will
find the qualitative methods exclusively associated with interpretative paradigm
and critical theory paradigm. This block comprises of 3 units.
Unit 18 on Participatory Method throwing light on the limitations of the

methods to collect quantitative data either through primary survey or secondary
sources covers the various tools and techniques used in collection and analysis
of qualitative data under participatory method and its advantages.
Unit 19 on Content Analysis deals with the conceptual and methodological

issues related to content analysis as a research method to undertake research
studies in social sciences in general and in economics in particular.
Unit 20 entitled Action Research throws light on Action Research as an offshoot

to emancipatory method associated with critical theory paradigm wherein
knowledge is generated in the process of knowing through doing rather through
conceptualization and theorizing. This unit covers the various models of action
research and the different steps involved in conducting action research.
Qualitative Methods
4
Participatory Method
UNIT 18 PARTICIPATORY METHOD
Structure
18.0 Objectives
18.1 Introduction
18.2 What is Participatory Research?
18.3 Methods of Participatory Research: Observation Method
18.3.1 Participant Observation
18.3.2 Non-Participant Observation
18.3.3 Quasi-participant Observation
18.4 Focused Interview
18.5 Oral Histories
18.6 Life History
18.7 Case Study Method
18.8 Narratives
18.9 Focus Group Discussion
18.10 Grounded Theory
18.10.1 Data Sources and Sampling
18.10.2 Subjectivity
18.10.3 Transcription
18.11 Analysis of Qualitative Data
18.12 Report Writing
18.13 Criticism of Participatory Methods
18.14 Advantages of Participatory Research
18.15 Let Us Sum Up
18.16 Key Words
18.17 Some Useful Books/Readings
18.0 OBJECTIVES
• state the concept of participatory research;
• state the distinction between data collection and data generation;
• explain the various tools of participatory research;
• discuss the grounded theory as a method of inductive theorising;
• identify the various steps involved in the analysis of qualitative data;
• pinpoint the major criticisms against participatory methods; and
• draw the lessons for participatory researchers.
5
Qualitative Methods
18.1 INTRODUCTION
As discussed in Unit 6, traditionally two approaches – quantitative research
associated with post positivist paradigm and qualitative research associated with
interpretative paradigm have been followed in undertaking research in social
sciences. The researchers working in the area of economics (both in micro and
macro) generally rely upon the inferences drawn from the large pool of data. The
data for these studies are collected either by the data compiling agencies or by
the individuals hired for data collection without involvement of individual
researchers in the field. The results of such quantitative research are insensitive
to context and ignore the people’s voices or their views about which studies
have been undertaken. Instead researcher’s opinion is imposed on what people
say and want. The collection of quantitative data through sampling survey and
its analysis involving several steps is time consuming. In the situation of natural
calamities and disasters like tsunami, we cannot wait for the outcome of
quantitative research. Further, quantitative research has not been very effective
in evaluation of poverty alleviation schemes and programmes. In such situation,
alternative method i.e. participatory method is increasingly being used in social
sciences to generate and analyse the data.
Hence, in this unit, we shall take up the various issues related to participatory
methods like concept of participatory research, the various tools to collect
qualitative data, analysis of qualitative data, advantages, limitations and important
lessons for participatory research etc. Let us begin with explanation of the concept
of participatory research.
18.2 WHAT IS PARTICIPATORY RESEARCH?

The fundamental premise of participatory research involves the direct participation
of researcher in the process of data generation. Important distinction is made
here between data collection and data generation. When data is collected by an
agency on large sample populations, standardizations are made with the
assumption that all those being interviewed will understand and respond to the
questions in the manner in which the primary researcher has conceptualized it.
In the field situation this may or may not be the case. Each field researcher
asking those questions may convey a different meaning and the respondent may
give answer that may not fit into any of the standard categories, but the field
researcher will reduce the answer and fit it in any of the given categories in a
structured schedule. Hence, the results obtained may not necessarily reflect the
market sentiments or the opinion of the people involved in the study. It is for this
reason that this process of procuring data using survey method and questionnaire
is called data collection.
In participatory research, the primary researcher is always in contact with the

respondent and has a face-to-face interaction. If the researcher thinks that the
respondent has not understood the query, he has the freedom to change the
language or reconstruct his probe question or collect information from other
indirect source. In this approach called data generation, the researcher has the
flexibility to generate multiple answers to a single query and then use his/her
interpretative skills to draw inference or meaning out of it to arrive at a
generalization. In participatory method of data generation, the process of data
6
collection and analysis proceeds simultaneously, making it more reliable and Participatory Method
presenting plurality of responses and possibilities. It is this flexibility and its
ability to generate reliable generalizations that participatory methods of research
have acquired importance in research methodology being used by present day
economists.
Let us first try and understand what participatory research is and how and where
it began? Methods of participatory research first find mention in anthropological
empirical research commonly known as ethnography. Ethnography is defined as
a method used to generate qualitative data from an insider’s perspective known
as EMIC view. In simple terms it implies that the people about whom we are
writing, must tell their stories and the researcher simply interprets them as “thick
description” (Geertz, 1973). People’s voices are critical to research instead of
researcher’s imposition of his opinion on what people say and want.
18.3 METHODS OF PARTICIPATORY RESEARCH:

OBSERVATION METHOD
Bronislaw Malinowski is regarded as the founding father of this methodology.
He described this method as “participant observation”. He stayed for more than
three years among the “Trobriand Islanders” and described their lifeway’s in a
classic ethnographic work titled Argonauts of Western Pacific first published in
1922. Methods of participatory research in the last ninety years or more have
been defined and redefined in many different ways. Malinowski’s method of
participant observations demands the researcher’s stay in the field area for
prolonged periods and become a part of the society that he/she is researching.
Later researchers realized that it was always not possible to stay for extended
duration (from three months to three years and in some cases even longer) in the
field and some problems can require immediate and urgent interventions. Another
critique of participant observation is that an interviewer is always an outsider
and even when he lives at a particular field site for long periods, there is no
guarantee that the community will view him as an insider and share every aspect
of their private lives with the researcher.
Nonetheless, none of these limitations influenced the importance of using

observations as one of the most important tools of participatory researches. Balki
(2009:206) calls it as “qualitative method par excellence”. Bryman (1998, 47-9)
puts forward a case for method of observation by saying that it is not a single
method but can combine several ways of doing observations. Given the significance
of method of observation, the techniques went in for some modifications. There
are three important ways of doing observations that are now recognized: (i)
Participant Observations, (ii) Quasi-Participant Observations, (iii) Non-
Participant Observations.
18.3.1 Participant Observations

Uncontrolled Participant observation is defined by Goode & Hatt (1981: 121) as
a procedure in which “the investigator can so disguise himself as to be accepted
as a member of the group”. When researchers live at a field site for long duration,
the assumption is that they are regarded as member of the community or the
group. A researcher wanting to study consumer behaviour may not disclose his
identity as a researcher and pretend to be a consumer, sharing their experiences
7
Qualitative Methods and concerns. While doing participant research, researchers often assume insider
identity. Sometimes researchers take neutral positions without disclosing the
real purpose of their being there.
Advantages of Participant Observation Method

Researcher can record all the information relevant for his work without disturbing
the community and intruding upfront into their privacy. Instead of sitting in a
corner in the market, an organization or any other work or field situation writing
notes, the researcher participates in these activities. In doing so, he collects relevant
information without disturbing the normal activities and records his observations
in a work diary. Thereafter, he makes a mental note of it and returns to his/her
workplace and records these observations immediately. The community or people
may not be aware that they are being researched. This helps obtain unhindered
and unbiased information.
Participant observation helps to obtain detailed information without any subject

bias. Many researchers regard it as a better tool for data gathering vis-à-vis
questionnaire and in-depth interviews.
Drawbacks of Participant Observation Method

One of the major drawbacks of participant observation is its inability to develop
procedures for standardization. Researcher doing participant observation acquires
a specific position in the community and observes from this vantage point. His
recordings are personal or individual specific. Another observer observing the
same situation may not be able to view it with the same perspective.
In participant observation method, the researcher may get emotionally involved

with the issues and tends to loose his/her objectivity. This often happens on
issues of dowry, bride burning, female foeticide, student agitations, price rise,
public-private partnership, trade union activities etc.
To overcome some of these drawbacks, researchers proposed method of non-

participant observations.
18.3.2 Non-Participant Observation

The qualitative researchers innovate in several ways to do non-participant
observations. Many of them would participate in some activities and observe
others from the outside. Some others would use video- recordings and analyse
these recordings later for generalizations. While doing so, it was possible to take
into account people’s gestures and non-verbal movements for detailed scrutiny.
However, critics of the method thought that as one is not fully participating in
this situation, he/she is not in a position to attribute right meaning to these
observations.
Sometimes, researchers use hidden cameras and record individual observations

separately. But there are several problems with this method of observations too.
Many regard use of hidden camera for research unethical. Participating in some
activities and not in others amounted to what researchers later called as Quasi-
participant observation.
8
18.3.3 Quasi- Participant Observation Participatory Method
Most researchers these days prefer to use quasi-participant method of

observations. In this method, the researcher can follow the dual entry system.
For instance, if you are trying to examine labour management relations, it is
imperative that you obtain permission from the management and also establish
rapport with the trade union leaders to have access to the union meetings. Many
a times, one has to conduct interviews in different locations. In one such study in
Punjab and Uttar Pradesh, we found that the workers preferred to talk to the
researcher outside the factory. The researcher would talk to the mangers and
supervisors in the factory and had lunch in the workers canteen. Appointments
were fixed with the workers while having lunch and then researcher went to
their houses, often spent hours there and after completing the interviews with
the supervisors and mangers hired a small room in the same slum area, where
most of the workers lived and shared their daily experiences.
In addition to the important research tool of observations, qualitative researchers
often use Focused interview, in-depth interviews, narratives and oral /Life case
histories, Focus group discussion/interviews’, and content analysis. These tools
have been discussed in next sections.
1) What is the distinction between data collection and data generation?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) What do you mean by the term EMIC view?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
3) How is participant observation method different from non-participant
observation method?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
18.4 FOCUSED INTERVIEW

In a focused interview, the researcher and the respondent are engaged in face-to-
face or one to one conversation in which the researcher poses questions and the
respondent provides detailed or precise answers. Talking to people or a specified
group of people selected randomly, using an interview schedule, interview guide
or a structured questionnaire is standard method of data collection. For qualitative 9
Qualitative Methods research, focused interviews are often unstructured and questions are invariably
open ended. This gives the respondent liberty to provide extensive answers.
It is important to remember that interview is “fundamentally a process of social

interaction” and involves developing a social relationship. In an interview situation
it is not only the interviewer but also the interviewee who makes an opinion
about the researcher. Respondent’s responses can be determined by what he thinks
what the researcher may expect from him and also the amount of trust and faith
he invests in the researcher. We will repeatedly stress the importance of
confidentiality and researcher’s ability to convey the same as hallmarks of
participatory research.
18.5 ORAL HISTORIES

Uninterrupted long interviews giving space to the respondents to express his
opinion or experiences at length is the preferred method by several qualitative
researchers. The interviewee is encouraged to share his life experiences, to express
his opinion, to recollect and recount their antecedents or lives of their
contemporaries, “and to discuss their perceptions of the processes involved and
the changes they have seen”(Balki, 2010:207). Collecting oral histories or
narratives requires skill of being a good listener. Many a times, you may assume
that the information provided by the respondent is not directly related to the area
of your focused interest. But that subsidiary information can provide certain
reference for a better understanding of the problem, issues or question that you
are exploring. In oral histories, one may pose probing questions in the beginning
and then occasionally intervene to ask a lead question for seeking any detailed
answer to specific query.
18.6 LIFE HISTORY

Another important tool that ethnographic and qualitative researchers have in
their kit is called the life history method. In this method life histories of individuals
are reconstructed to understand the historical events of that period. Many a times
researchers would use diaries and autobiographies to construct chronology of
events of that period. Recording experiences of individuals facilitates
understanding of how and what happened in certain circumstances. Life history
approach is largely dependent on the ability of the respondent to recall and share
that memory with the researcher. With the help of the respondent, researcher
gradually determines the social processes. In the context of economics, rise and
fall of a particular model is determined by prevalent social processes of that
period, e.g. popularity and weakening of models of conservative and liberal
economics in the last fifty years. It also helps to understand different perspectives
based on gender and class. While recollecting their life history, many a times
respondents may fall silent. This silence also provides data as it tells the researcher,
how that particular event may have caused anxiety, anguish or an intense
emotional experience to the respondent.
18.7 CASE STUDY METHOD

To study firms and organization’s behaviour, Business economists use the case
study method. The researcher can focus on one company or organization and
10
collect a detailed case history of the organization. Yin (1984:23) defines case Participatory Method
study research method “as an empirical inquiry that investigates a contemporary
phenomenon within its real life context; when the boundaries between
phenomenon and context are not clearly evident and in which multiple sources
of evidence are used”. In an organization, there are multiple players and a detailed
case study will take into account perspective of each stakeholder there. You can
use more than one method to collect evidence from different stakeholders e.g.
the supervisors, workers etc. A schedule or a questionnaire can be used to talk to
the workers. The interview method can be used, to collect history of the
organization and organogram depicting position and functions of each member
of the organization. Content analysis can be done by examining available records
of the organization. This provides a holistic picture of the organization. Case
study as an approach is holistic, inductive and idiographic. It takes complete
account of the organization from all perspectives; theorizing moves from general
to specific and individual cases are located in their particular context. Case study
can take cross-sectional or longitudinal data.
The researcher must remember that generalizations drawn for one organization
are not necessarily applicable to the other. Each organization is a distinct unit.
Six important steps are involved in conducting a study by using case study method:
• Determine and define the research question
• Select the cases and determine data gathering and analysis techniques
• Prepare to collect the data
• Collect data in the field
• Evaluate and analyse the data
• Prepare the report
18.8 NARRATIVES
Social science research in anthropology, sociology, political science, history and
now economics is moving from the positivist position of research to narrative
approach. The positivist position also popularly known as scientific approach
was the tradition in which researchers from 50’s to 80’s were generally trained.
The standardized tools of survey research, questionnaire approach or even
interviews with a standardized set of questions was expected to produce if not
same, similar responses. Narrative research emphasizes that this is not true.
Different respondents may interpret each question asked in the same words
differently. In structured interviews both meaning and plurality of responses are
lost. Researcher-using method of narrative understands that there can be
standardized questions but there are no ‘standardized meanings’.
18.9 FOCUS GROUP DISCUSSION

One of the methods of qualitative research that found ready acceptance by
economists was that of focus group. It is frequently used in market surveys for
quick appraisal of response to an economic programme, for evaluation studies
and for media and health research. In focus group discussion, a small group of
six to eight (market researchers prefer groups of 10-12 participants in each group)
targeted audiences is gathered at a suitable location. They are encouraged to
11
Qualitative Methods have a frank discussion on a given topic and a recorder sitting outside the group
will either audio record or take notes on the discussion. The researcher acts as a
prompter posing questions on the subject of inquiry. Kruger (1988: 18) defines
focus group as a “carefully planned discussion designed to obtain perceptions
on a defined area of interest in a permissive, non-threatening environment”.
Bloor et al. (2001) cautions that ‘focus groups are the method of choice only
when the purpose of the research is ‘to study group norms, group meanings and
group processes’ (cf. Barbour, 2008:133)
Steps for focus group
• Topic for focus group should be of general interest to all the participants in
the group.
• Number of participants can vary depending on the requirements of your
research. Sampling is the key to data generation.
• ‘Stimulus material’ for initiating the discussion should be brief and encourage
a free flowing conversation. Use of video clips, leaflets, cartoons, brief
recollection of event or topic being discussed etc.
• Moderator must mange the discussion in a manner that without sounding
intrusive, long extensive talks are avoided.
• Barbour (2008) suggests ‘anticipate the analysis through attentive moderating’.
• Locate themes visible in the form of repetitive patterning of the responses.
• Second stage sampling could be considered to further explore any emerging
theories, paradigms or even trends from the data.
These are some of the tools that participatory researchers frequently use to gather
primary data. In this brief discussion, there are two more important steps that a
participatory researcher must remember. First is Grounded Theory and second
is Data Sources and Sampling procedure. We shall discuss these steps in next
sections.
1) Make a distinction between focused interview and focused group discussion.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) State the uses of Case Study Method.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
12
3) Identify the steps involved in focus group discussion. Participatory Method
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
18.10 GROUNDED THEORY

Quantitative researchers are familiar with the procedure of preparing a research
design that lists various steps of research including hypothesis and sample
selection. Average participatory researcher will first go to the field, explore various
research questions that he/she may have problematized and engage in casual
conversations with the people on the ground. A research design is generally not
prepared before empirically exploring the field situation. The research problem
is then structured keeping in view the perspective, people’s priorities and that of
the researcher.
Glaser and Strauss (1967) proposed grounded theory as a method of inductive

theorizing. In their opinion, theories are not to be tested but to be generated as
research proceeds. They were of the opinion that “‘good theory’ is systematically
discovered from and verified with the data of social research”. Researchers further
clarified ‘generating a theory from data means that most hypotheses and concepts
not only come from the data, but are systematically worked out in relation to the
data during the course of research’ (cf. Blaikie 2010:141). In this kind of analysis,
comparative analysis is continuously generated along with theory building. Two
kinds of theories are generated namely substantive and formal. Substantive
theories are located in a specific context and are related to a specific process.
Formal theories are generated at a higher level of generalization and can apply to
number of substantive areas. Blaikie sums it brilliantly “research conducted from
a grounded theory point of view is not a pre-planned linear process of testing
hypotheses, but rather an evolving process in which what has been ‘discovered’
at any point will determine what happens next. An understanding of any
phenomenon is seen as a developing process involving the collection of variety
of data, by a variety of methods’. In simple terms, it suggests that you do not
have to go to the field with a pre-decided hypothesis with the intent of verifying
it but evolve your theory from the empirical evidence that was collected.
18.10.1 Data Sources and Sampling

Three types of data sources are recognised:
Primary Data – Data collected by researcher or researchers working on the
project and they are responsible for collecting/generating, analysis and reporting.
Majority of participatory researchers work with primary data.
Secondary Data – Data that is already collected by some other researchers,
agencies, individuals (raw data) and is available in the public domain e.g. Census,
National Sample Survey, NCAER Data, Economic Survey Data etc.
Tertiary Data – Analysed data either by the researcher or other researchers,
institutions, agencies and its results are available for comparisons or reinterpretation.
13
Qualitative Methods Data for research can come from number of sources: (i) Individuals, (ii)
Populations.
Samples from populations

Once data source is identified and it happens to be a population, it is not always
possible to cover the entire population. In these situations a representative sample
is identified. There are several methods of sampling. Probability and
nonprobability sampling are commonly used. A sample is expected to contain
one or more elements of the population/universe of study. Sampling frame using
probability sampling is able to ensure elements of representation but that may
not be possible in nonprobability sampling. Probability samples are selected using
simple random and systematic random sampling. Non-probability sampling
methods frequently used by participatory researchers are:
• Accidental or Convenience sampling – interviewing any individual, any
where at any time without keeping research population in focus. This method
is to be used only when no other method is feasible. This is often used in
market research for quick feedback or for consumer surveys.
• Quota sampling – frequently used after selecting certain criterion to be
explored in the research study. Quota is taken to represent the population
and problem being researched. For doing an evaluation study for a product
to be launched target the population e.g. for the launch of a new video game-
target audience can be adolescents in the age group of 15-20 that come to a
video store enquiring about new video games.
• Judgemental or purposive sampling – is another popular non-probability
sampling method used for studying previously identified sub-set of population
e.g. to analyse models for successful organizational management, a sample
is selected from organizations that have shown successful management
background.
• Snowball method – as the term suggests, this method of sampling works on
the principle of networks, chain referral or reputational sampling. Researcher
identifies an individual or a key respondent and he/ she may lead him to
others having similar interests – the focus of the research. Internet networks,
members of mac-book or Mercedes clubs etc. are few illustrative examples.
Participatory researchers are often criticized for using these sampling procedures
arguing that these are not necessarily representative, as the sample is not
objectively selected. One must remember that qualitative participatory research
is invariably resource intensive and requires smaller samples as compared to
large quantitative surveys.
18.10.2 Subjectivity
Qualitative researchers are not shy of admitting that there is always an element
of subjectivity in research. Hollway and Jefferson (2000:3) draw attention to
this important element of research: “ As researchers, we cannot be detached but
must examine our subjective involvement because it will help us to shape the
way in which we interpret interview data”.
14
18.10.3 Transcription Participatory Method
One of the important components of qualitative method of research or

participatory approaches is rewriting the data procured from the field. The data
is usually recorded in memory or sometimes on paper, and if circumstances permit
it can be recorded using an audio or wherever possible a video recording. The
data has to be transcribed for the purpose of analysis. This is one of the most
difficult steps that many qualitative researchers have to learn to deal with.
18.11 ANALYSIS OF QUALITATIVE DATA

The findings collected with the help of qualitative tools involve a continuous
process of data generation and analysis. ‘It is iterative rather than being linear’
(Barbour, 2008:189). Unlike quantitative methods in which first data is collected
and subjected to statistical tool to obtain results and analysis follows the outcomes,
in qualitative data, we do not generate results but detail ‘findings’. While talking
to people either in focused or in-depth interviews or in focus group discussions
and case studies, people provide the answers through the process of reconstruction
of these narratives. Qualitative data analysis uses ‘constant comparative
method’. It involves constant comparing and contrasting of notes meaning thereby
that the researcher must focus on ‘who is saying what and in what context’ (ibid:
217).
Steps for doing analysis of qualitative data:
• Compilations of field notes and observations recorded in the field diary.
• Some scholars prefer to have complete verbatim account of data transcribed

but others opt to use ‘indexed recordings and notes’ (ibid: 192).
• Interrogation of the data and diligence shown by the researcher is the key to
producing good transcripts and reliable generalizations.
• Codes and themes have to be developed to capture meaning.
• Grounded theory that uses concepts developed in the field by the respondent
is an important way of theorizing and analysing in qualitative research.
• Both computer and Manual analysis can be done.
• NUD*IST, ATLAS/TI are some of the Computer software used for qualitative
analysis.
18.12 REPORT WRITING

Richardson (2000: 923) refers to writing as “a method of inquiry, a way of finding
out about yourself and your topic”. A “personal tale of what went on in the
backstage of doing research” (Ellis & Bochner, 2000: 741). Writing research
differs in style depending on whether one is writing for a project agency for a
journal or a dissertation/ thesis. Standard format comprises of introduction/
background, methods, sample, results, discussion, and conclusion/
recommendations. In qualitative research writing discussion is a challenge
because the data collected is iterative or repetitive. Many qualitative researchers
15
Qualitative Methods prefer to use subtitle ‘findings’ to combine discussion and conclusions. As young
researcher, you can always evolve your own style of writing but remember that
when you are writing for an indexed journal, you are expected to conform to the
writing style and referencing style of that particular journal which is always
given at the back of the journal as instructions to the authors.
18.13 CRITICISM OF PARTICIPATORY METHODS

Opinions on the efficacy of participatory methods are widely divided. There is a
growing demand for bringing in participatory methods of research in disciplines
like economics, political science, management studies and law because standard
methods of quantitative research have not yielded the desired results. There are
others opposing reliance on participatory research arguing that there is every
possibility of personal bias impacting the objectivity of research. It is restricted
to a very small area and generalizations arrived on the basis of data generated by
it may not be applicable to other areas. The methods of participatory research are
slow. Many others call these studies impressionistic and unreliable and call it a
‘soft’ method of doing research.
To overcome some of the limitations of participatory research mentioned above,

many researchers prefer to use a mixed methods approach. Let us remember that
the details on mixed methods research have already been covered in Unit 6.
18.14 ADVANTAGES OF PARTICIPATORY

RESEARCH
One of the important advantages of qualitative method of research is that it is
“reflexive, flexible and iterative” (Cornwell & Jewkes: 1668; Chambers 1992).
Diane Watt (2007:82) rightly points out that “since the researcher is the primary
instrument of data collection and analysis, reflexivity is deemed essential”– this
is a position supported by almost all researchers using participatory methods.
Reflexive means that the researcher can take stock of his/her own reservations,
prejudices and resolve them before recording responses obtained from the field.
In simpler terms, it implies that in participatory researche, we deal with other
human beings that look like us but may have differently acquired social values.
The answer that they may give to an enquiry posed by us may be very different
from the response that we expect or believe is the correct answer. Reflexivity in
participatory research methods helps to distinguish qualitative research methods
from quantitative methods of research. It teaches us to delineate subjectivity
from objectivity. One of the strongest planks of quantitative researchers and
scientific (positivist) methods of research was their claim that the results derived
by these methods are ‘objective’. Qualitative researchers accepted the charge of
being ‘subjectively objective’ arguing that when a researcher interacts in the
field with his respondents, he always carries a sense of subjectivity with him/her
but in quantitative researches it is not admitted and dismissed by counting in
sampling error. It was thus considered imperative to admit that bias by the
qualitative researchers. They accepted that one must recognise it in terms of
one’s reflexivity. Qualitative/Participatory researchers have recognised the
limitations and persistence practice that is required in recognising, admitting
and becoming aware of this bias, particularly while writing their research report
(Borochowitz, 2005; Denzin and Lincoln, 2005; Harding, 1991).
16
We discussed ‘reflexivity’ and ‘flexibility’ given in Cornwell and Jewkes Participatory Method
statement describing advantages of participatory research. We now examine the
third attribute of participatory research ‘iterative’ that means ‘ability to repeat’–
this can be inferred in several ways. One can repeatedly seek answers to the
same query by enquiring about it in different ways to establish the validity of the
response. Detailed narratives generated from the respondents may have responses
that are repetitive of the statements made by other respondents and the common
thread running through these conversations provides the window to ascertain
the public opinion. It facilitates “ ‘bottom up’ approach with a focus on locally
defined priorities and local priorities”(Cornwell & Jewkes, 1995:1667).
Prior to the popularity and acceptance of participatory approaches, most planning

was based on the philosophy of ‘trickle down effect’– implying that the experts
plan and enrich the upper crust and its impact will automatically get transferred
to the lower stratum of the society. Felt needs of the people were not taken into
consideration and the planning was top heavy. Many intervention projects with
large funding failed because they failed to meet the immediate requirements of
the local people. For example – several housing projects for the adivasi in remote
parts of tribal heartland remained unutilized or underutilized, as the houses were
not built in accordance with the tribal worldview and customary lifestyle.
Participatory approach first explores people’s immediate needs as perceived by
the ‘local people’ and then suggests policies for meeting those needs.
Participatory researchers argue, “that the key element of participatory research

lies not in methods but in the attitudes of researchers, which, in turn, determine
how, by and for whom research is conceptualized and conducted” (ibid) If the
researcher is not able to establish rapport with the respondent, he will not be able
to generate good data. It is also important in participatory researches that the
researcher takes keen interest in the problems that are addressed in his research
query.
It is important to remember that you cannot become a good participatory

researcher by learning these methods only through a formal lesson in a classroom.
It requires personal experience in the field and repeated interactions with a large
number of individuals to master the craft of participatory research. A lesson on
participatory research empowers you with tools like the necessity to respect
confidentiality of your respondents, to refrain from encroaching on their privacy,
to respect their distinct traditions and cultural values, to conduct your interviews
without offending or getting into an argument with them, respecting their opinion
and recording it objectively and being ethical at all times. Your honesty in data
generation and data analysis will give the necessary quality weightage to your
research. In this context, it is worth to remember that one may encroach on the
respondent’s time and privacy. While writing your research never disclose the
name and location of your respondents. Seeking their consent for interviewing
and writing about them is equally important.
Participatory Research: An Illustration

You enquire from an individual in the market about the price of onions in the
month of July 2014, when the market sentiment suggests, that the price of
onions is very high. Contrary to the assumed answer the respondent suggests
that “no, it is not as high as last year because last year in the same month, I
17
Qualitative Methods
bought onions for Rupees ninety a kilogram but this year it is much cheaper,
as I have paid only rupees forty per kilogram”. His response is relative to his
experience in the past and similarly other responses may differ depending on
the experiences of different individuals. But when you ask standardized
question in a yes or no format and ask a direct question with only two options
of ‘yes’ or ‘no’, your results will not reflect ‘actual sentiments’ in the market.
In the participatory approach, you will have flexibility of questioning as also
response. On the contrary in the non-participatory approach you will get only
two kinds of responses either saying ‘the price of onions is very high’ or
some saying ‘no it is not high’. Thus there is inbuilt flexibility in the
participatory approaches. They provide a better understanding of the real
market situation, giving more options for planning and interventions.

1) Enlist the various methods of sampling used in participatory research.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) What is inductive theorizing?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
3) State the steps involved in analysis of qualitative data.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
18.15 LET US SUM UP

In participatory research, data is generated from an insider’s perspective known
as EMIC view. Unlike quantitative research, in this approach, people provide
answers through the process of reconstruction of the narrations. The tools used
in data collection are – participation observation, non-participation observation,
quasi participation observation, in-depth interview, oral and life histories,
narratives, case study method and focus group discussion etc. Qualitative
researchers use various non-random sampling techniques, these include –
convenience sampling, quota sampling purposive sampling, and snowball
sampling. The qualitative data is analysed by ‘constant comparative method’.
The personal bias is the important criticism levelled against this method.
Reflexivity, flexibility and iterative are the key advantages of this method and
hence, participatory research by using mixed methods approach is increasingly
being used to address complex issues in social science research.
18
18.16 KEY WORDS
Focus Group Discussion : It is a technique for a eliciting descriptive
information/data from specific population
subgroups (a group of 8 to 12 persons).
Reflexivity : Reflexivity refers to a situation where an

individual becomes a self by being able to take
the attitude of others and thereby reflect on his
or her own behaviour.
Snowball Sampling : A technique for gathering subjects through the

identification of an initial subject who is used
to provide the names of other actors.
Quota Sampling : A method of selecting respondents for service

by assigning quotas to the interviewers that
define groups of respondents by using a few key
demographic characteristics.
Iteration : It is an act of repeating a process with the aim of

approaching a desired goal, target or result.
18.17 SOME USEFUL BOOKS/READINGS

Borochowitz, Dalit Yassour. (2005). “Teaching a Qualitative Research Seminar
on Sensitive Issues”. Qualitative Social Work 4:347-62.
Bogdan, R.C. and Taylor, S.J. (1975). Introduction to Qualitative Research
Methods: A Phenomenological Approach to the Social Sciences. New York: John
Wiley.
Barbour, Rosaline. (2008). Introducing Qualitative Research: A Student Guide
to the Craft of Doing Qualitative Research. London: Sage Publications.
Denzin, Norman. K. & Lincon, Yvonna. S. (eds) (2000). Handbook of Qualitative
Research (2nd ed.). Thousand Oaks, CA: Sage Publications.
Denzin, Norman K. and Yvonna S. Lincoln (2005). “Introduction: The Discipline
and Practice of Qualitative Research.” Pp. 1-33 in The SAGE Handbook of
Qualitative Research edited by Norman K. Denzin, and Yvonna S. Lincoln,
California: Sage Publications.
Harding, Sandra. (1991). Whose Science? Whose Knowledge?: Thinking from
Women’s Lives. Ithaca, New York: Cornell University Press.
Watt, Daine. (2007). On Becoming a Qualitative Researcher: The Value of
Reflexivity. The Qualitative Report. Volume 12 Number 1 March 2007. Pp 82-
101.
Yin, R.K. (1984) Case Study Research: Design and Method. Newbury Park,
CA: Sage.
Websites:
http://www.nova.edu/ssss/QR 19
Qualitative Methods
18.18 HINTS/ANSWERS OR HINTS TO CHECK
YOUR PROGRESS EXERCISES
1) See Section 18.2
2) See Section 18.2
3) See Section 18.3
2) See Section 18.7
3) See Section 18.9
2) See Section 18.10
3) See Section 18.12
20
UNIT 19 CONTENT ANALYSIS
Structure
19.0 Objectives
19.1 Introduction
19.2 Historical Background of Content Analysis
19.3 Content Analysis: Concept and Meaning
19.4 Terms Used in Content Analysis
19.5 Approaches of Content Analysis
19.5.1 Conceptual Content Analysis
19.5.2 Relational Content Analysis
19.6 Procedure Involved in Content Analysis
19.6.1 Formulation of Research Question
19.6.2 Determining Materials to be Included
19.6.3 Developing Content Categories
19.6.4 Selecting Units of Analysis
19.6.5 Code the Materials
19.6.6 Analyze and Interpret the Results
19.7 Uses of Content Analysis
19.8 Advantages and Disadvantages of Content Analysis
19.8.1 Advantages
19.8.2 Disadvantages
19.9 Let Us Sum Up
19.10 Key Words
19.11 References and Suggested Books
19.0 OBJECTIVES
• explain the concept and meaning of the content analysis;
• identify the areas of research in social sciences wherein content analysis can
be used;
• discuss the various types of content analysis;
• state the various approaches of content analysis;
• describe the various steps involved in using the content analysis;
• enable you to analyse and interpret the results; and
• discuss the advantages and disadvantages of the content analysis.
19.1 INTRODUCTION
Many research Scholars take up historical overviews of economics as an area of
research. They probe the research questions like how the different streams of
21
Qualitative Methods economics like institutional economics, Neo-classical economics, behavioural
economics etc. have evolved over a period of time and how researcher’s focus
had shifted from one area to another isolating a particular paradigm change.
Sometimes a researcher aims to lay concepts and meanings from the social actors’
account of realities related to his identified research problem. They use the sources
like recorded interviews, letters, journals, newspaper stories and verbal materials.
In such situations, content analysis as data (information) analysis technique is
used by the social scientists. Much of the subject matter of social sciences
including consumer studies is in the form of verbal and nonverbal behaviour.
The exchange process in the market place and the communication of the values
of the exchange depends upon the written or spoken words. Much of consumer
research has concentrated on the characteristics, opinion, or behaviour of the
interpreter of communication messages. In media economics, issues like
monopolistic competition in TV channels, competition and diversity in newspaper
industry, effects of group ownership on daily newspaper content etc. can be probed
by content analysis alongwith other statistical techniques. Similarly content
analysis can also be used to analyse the cultural issues in cultural economics to
capture pattern form of explanation. Hence, in this unit, we will discuss the
concept of content analysis, its various types and approaches, the procedures
involved in its application, and its advantages and disadvantages. Let us begin
with historical background of content analysis followed by describing its concept
and meaning.
19.2 HISTORICAL BACKGROUND OF CONTENT

ANALYSIS
We find different approaches to analysis and comparison of texts in hermeneutic
contexts (e.g. bible interpretations) like newspaper analysis, graphological
procedures, content analysis, and the dream analysis by Freud etc. The basis of
quantitative content analysis had been laid by Lazarsfield and Lasswell in the
USA during 1920’s and 30’s. Historically, content analysis was a time consuming
process. Analysis was done either manually, or by slow mainframe computers to
analyze punch cards containing data punched by human coders. Single study
could employ thousands of these cards. Human error and time constraints made
this method impractical for large texts. However, despite its impracticality, content
analysis was utilized as research method by 1940’s. Although initially limited to
studies that examined texts for the frequency of the occurrence of identified
terms (word counts), by mid-1950’s researchers started to consider the need for
more sophisticated methods of analysis, focusing on concepts rather than simply
words, and on semantic relationships rather than just presence. The first textbook
about this method was published by Berelson in 1952. In the sixties of 20th
century, content analysis found its way into linguistics, psychology, sociology,
history, arts etc. The procedures involved in content analysis were refined by
incorporating models of communication; analysis of non-verbal aspects,
contingency analysis, computer applications. Since the middle of 20th century
objections had been raised against a superficial analysis without respecting latent
contents and contexts, working with simplifying and distorting quantification
(Kracauer, 1952). Subsequently qualitative approaches to content analysis had
been developed. Today, the use of modern computers makes content analysis
much simpler and the data less vulnerable to human error.
22
Content Analysis
19.3 CONTENT ANALYSIS: CONCEPT AND
MEANING
Content analysis is a set of procedures for collecting and organizing information
in a standardized format that allows analysts to make inferences about the
characteristics and meaning of written and other recorded materials.
Content analysis is a multipurpose research method developed specifically for

investigating a broad spectrum of problems in which the content of communication
serves as the basis of inference. Content analysis covers both the content of the
material and its structure. Content refers to the specific topics or themes in the
material. Structure refers to form. Whether an article is prominently featured on
the first page of a newspaper or buried in the middle section is a structural question.
Careful reading of the written materials is necessary for all kinds of researches
in social sciences. But content analysis is something different from careful reading
of written materials in many ways.
According to Berelson, content analysis is a research technique for the objective,

systematic, and quantitative description of the manifest content of communication.
The term content analysis is used here to mean the scientific analysis of
communication messages. The method being “scientific” catholic in nature,
requires that the analysis be rigorous and systematic. According to Paisley, content
analysis is a phase of information processing in which communication content is
transformed through objective and systematic application of categorization rules
into data that can be summarized and compared.
This definition underlines the point that content analysis is a systematic technique
for analyzing message content and message handling. It is a tool for observing
and analyzing the overt communication behaviour of selected communicators.
Content analysis , while certainly a method of analysis, is more than that, it is a

method of observation. Instead of observing people’s behaviour directly, or asking
them to respond to scales or interviewing them, the investigator takes the
communications that people have produced and asks questions of the
communication ( Kerlinger, 1968, p.544).
Content analysis refers to any procedure for answering the relative extent to
which specified references, attitudes or themes permeate a given message or
document ( Stone, 1964).
Neuendorf defines content analysis as “ the systematic, objective, quantitative

analysis of message characteristics”.
Based on the above definitions, the following characteristics of content analysis

emerge: objectivity, systematic and generality.
Objectivity: To have objectivity, the analysis must be carried out on the basis of
explicitly formulated rules which will enable two or more investigators to obtain
the same results from the same documents. This requirement of objectivity gives
scientific standing to content analysis and differentiates it from literary crtiticism.
Systematic: In a systematic analysis, the inclusion and exclusion of content or

categories is done according to consistently applied criteria of selection. This
23
Qualitative Methods requirement is meant to eliminate elements in the content which do not fit in the
analyst’s thesis.
Generality: By generality, we mean that the findings must have theoretical

relevance, purely descriptive information about content, unrelated to other
attributes of content or to the characteristics of communicator or recipient of the
message, is of little scientific value.
Thus, content analysis can be defined as a research technique for collecting and
organizing information in a standardized format for making inference
systematically and objectively about the characteristics of message.
19.4 TERMS USED IN CONTENT ANALYSIS

The following concepts and terms are frequently used in content analysis. Let us
understand these terms.
Manifest Content Analysis: It involves simply counting words, phrases, or

“surface” features of the text itself. It provides reliable quantitative data that can
easily be analyzed using inferential statistics.
Latent Content Analysis: It involves interpreting the underlying meaning of

the text. Latent analysis is different to manifest analysis because researcher must
have a clearly stated idea about what is being measured. The value of the latent
content analysis actually depends upon the researcher’s ability to expose
previously marked themes, messages and cultural values within the text. This
analysis is widely used in qualitative content analysis.
Content Units: There are two types of content units i.e. the unit of analysis and
unit of observation. The unit of analysis concerns the general idea or phenomenon
being studied. The unit of observation concerns the specific item measured at an
individual level.
Coding: Coding is the process whereby raw data are systematically transformed
and aggregated into units which permit precise description of relevant content
characteristics. There are two methods of coding: (i) deductive measurement,
(ii) inductive measurement.
Deductive Measurement: It requires the development of specific coding

categories before a researcher starts a content analysis. Deductive measurement
is useful with an established set of coding categories or if a clear hypothesis or
research question exists at the outset of the analysis.
Inductive Measurement: This method supports the practice of emergent coding,

which means that the basic research question or hypothesis for a formal content
analysis emerges from the units of observation. It entails creating coding categories
during the analysis process. Emergent coding is useful in exploratory content
analysis.
Coding Scheme: Every analysis begins with an existing coding scheme. Coding
scheme is another phrase of coding categories and coding book, within which all
instances of the content are analyzed or noted. Every coding scheme consist of
the following statements.
24
Master Code Book: It provides the coders with explicit instructions and defines Content Analysis
each word/ phrase/ aspect to be analyzed. It explains how to use code sheet.
Code Sheet: Code sheet provides the coders with a form on which they note
every instance of every word/ phrase/ aspect being analyzed. The code sheet lists
all the coding categories.
Coding Dictionary: It is used with computer based content analysis. A set of

words/ phrases, part of speech …. That is used as the basis for a search text.

1) Identify the areas of research in economics wherein content analysis can be
used.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) What is the distinction between manifest content analysis and latent content
analysis?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
19.5 APPROACHES OF CONTENT ANALYSIS

Instead of being a single technique, content analysis is a collection of different
approaches to the analysis of texts or more generally of messages of any kind.
Important approaches frequently used in Social Sciences are discussed below:
19.5.1 Conceptual Content Analysis

Traditionally, content analysis has most often been thought in terms of conceptual
analysis. In conceptual analysis, a concept is chosen for examination and the
analysis involves quantifying and tallying its presence. It is also known as thematic
analysis. The focus here is on looking at the occurrence of selected terms within
a text or texts, although the terms may be implicit as well as explicit. While
explicit terms obviously are easy to identify, coding for implicit terms and deciding
their level of implication is complicated by the need to base judgments on a
somewhat subjective system.
19.5.2 Relational Content Analysis

Relational content analysis, like conceptual analysis, begins with the act of
identifying concepts present in a given text or set of texts. However, relational
analysis seeks to go beyond presence by exploring the relationships between the
concepts identified. Relational analysis has also been termed semantic analysis.
25
Qualitative Methods In other words, the focus of relational analysis is to look for semantic, or
meaningful relationships. Individual concepts, in and of themselves, are viewed
as having no inherent meaning. Carley (1992) asserts that concepts are ‘ideational
kernels.’ These kernels can be thought of as symbols which acquire meaning
through their connections to other symbol. Relational content analysis approach
can be of three types:
i) Affect extraction: This approach provides an emotional evaluation of

concepts explicit in a text. It is problematic because emotion may vary across
time and populations. Nevertheless, when extended, it can be a potent means
of exploring the emotional/psychological state of the speaker and/or writer.
ii) Proximity analysis: This approach, on the other hand, is concerned with
the co-occurrence of explicit concepts in the text. In this procedure, the text
is defined as a string of words. A given length of words, called a window, is
determined. The window is then scanned across a text to check for the co-
occurrence of concepts. The result is the creation of a concept determined
by the concept matrix. In other words, a matrix, or a group of interrelated,
co-occurring concepts, might suggest a certain overall meaning.
iii) Cognitive mapping: This approach is one that allows for further analysis of
the results from the two previous approaches. It attempts to take the above
processes one step further by representing these relationships visually for
comparison. Whereas affective and proximal analysis function primarily
within the preserved order of the text, cognitive mapping attempts to create
a model of the overall meaning of the text. This can be represented as a
graphic map that represents the relationship between concepts. In this manner,
cognitive mapping lends itself to the comparison of semantic connections
across texts. This is known as map analysis which allows for comparisons to
explore “how meanings and definitions shift across people and time”.
19.6 PROCEDURE INVOLVED IN CONTENT

ANALYSIS
The various steps involved in content analysis are: (i) Formulation of research
questions, (ii) Determining materials to be included, (iii) Developing content
categories, (iv) Selecting and Finalizing units of analysis, (v) Code the materials,
(vi) Analyze and interpret the results. These are discussed below:
19.6.1 Formulation of Research Questions

Content analysis begins with a specific statement of the objectives or research
questions. The objective of content analysis is to convert recorded “raw”
phenomenon into data, which can be treated essentially in a scientific manner so
that body of knowledge may be built up. Objectives are precisely worded questions
that investigator/s is/are trying to answer. By making a clear statement of the
research question, the researcher can ensure that the analysis focuses on those
aspects of content which are relevant for the research. The question should be
based on a clear understanding of research needs and the available data. Therefore,
the selection of the topic should be such that can be answered by analyzing the
appropriate communication content.
26
In general content analysis can be used to answer “ What” but not “ Why” question. Content Analysis
It helps analysts to describe or summarize the content of written material , the
attitudes or perceptions of its writers, or its effects on the audience. For example,
if analysts want to assess the effects of different women empowerment programmes
on the lives of younger and middle aged women in rural and urban area. Content
analysis of open–ended interview / responses could be used to identify their
outlook, attitudes and security about their life.
19.6.2 Determining Materials to be Included

The next step of content analysis is to decide relevant communication content to
answer the research question and to determine the time period to be covered.
Content analysis can be used to study any recorded materials as long as the
information is available to be reanalyzed for reliability checks. It is used most
frequently to analyze written materials to study any recorded communication
including TV programmes, movies, photographs, regulations, other public
documents, workplaces, case studies, reports, answer to survey questions,
newspapers, news release, books, Journals article, letters etc. Speeches and
discussions can also be analyzed.
Sampling: Sampling is necessary if the population is too extensive to be analyzed.

Thus a sample should be selected from the population in order to make valid
conclusion and generalization about a population. Simple random sampling,
interval sampling, cluster sampling and multistage sampling techniques are used
in content analysis.
19.6.3 Developing Content Categories

Content analysis is no better than its categories., since they reflect the formulated
thinking , hypotheses, and the purpose of the study. Categories provide structure
for grouping and recording units. Developing the category system to classify the
text is the heart of content analysis. Berelson (1952) has emphasized the
importance of formulating coding categories by quoting that “ content analysis
stands or falls by its categories. Particular studies have been productive to the
extent that categories were clearly formulated and well adequate to the problem
and to the content”.
To be useful, every content category must be thoroughly defined, indicating what
type of material be and not to be included. Chadwick et.al. (1984) have also
emphasized the following three characteristics of content categories i.e. (i)
Categories should be exhaustive so that all relevant items in the material being
studied can be placed within a category. (ii) It should be mutually exclusive so
that no item can be coded in more than one category. (iii) Thirdly categories
should be independent so that recording of unit’s category assignment is not
affected by the category assignment of other recording units.
Category format: Categories can be conceptualized in many ways. Some
common category formats are grouping, scales, and matrices. Structured category
format increase coding efficiency especially when the number of categories are
large. Scales provide the rank ordering information. Matrices are useful formats
when analysts seek more information about issues than simply whether they are
present or absent.
27
Qualitative Methods 19.6.4 Selecting Units of Analysis
Once the categories have been identified, the analyst would be interested in
determining the unit of content for classification under the content categories
and the system of enumeration for the same. The unit of analysis is the smallest
unit of content that is coded under the content category. The unit of analysis vary
with the nature and objective of the analysis. Thus, the unit of analysis might be
a single word, a theme, a letter, a symbol, a news story, a short story, a character
or an entire article etc. There are two kinds of unit of analysis : Recording unit
and Context unit.
The Recording Unit is the specific segment of content in which the occurance
of a referene /fact is counted or the unit can be broken down so that reference/
facts can be placed in different categories.
The Context Unit is the larger body of the content that may be searched to
characterize the recording unit. Context units set limits on the portion of written
material that is to be examined for categories of words or statements. A recording
unit is the specific segment of context unit in written material that is placed in a
category. For example, if the coding unit is the word, then the context unit may
be the sentence or the pargraph in which the word appears and characterizes the
recording unit.
Word: The smallest unit generally used in the content analysis as a unit is a
word. Lasswell(1952) calls word as a symbol and may include word compounds
e,g, phrases as well as single word. In this type of research one might study the
relative occurrence of key symbols or value laden terms until the content has
been systematically examined relevant to the hypotheses of the study.
Theme: The theme is a single assertion about a subject. The theme is among the
most useful units of content analysis because issues, values, beliefs, and attitudes
are usually discussed in this form.
Character: Character may be defined as a use of fictional or historical character

as the recording unit is also employed.
Item: The item is the whole natural unit employed by procedures of symbolic
material. It may be the entire speech, radio programme, letter to the editors, or
news story.
Space and Time Measures: Some studies have classified content by physical
division such as column inch, the line or paragraph, the minutes or the foot film.
Finalizing units of analysis: In content analysis, the counting or quantification

of the units is performed by using three methods of enumeration : 1) Frequency,
2) Intensity or Direction, and Space/ Time.
Frequency: Frequency simply means counting whether or not something occurs

and how often (how many times).
Direction: Direction is nothing but the direction of messages in the content

along some continuam e,g, positive, negative, supporting or opposed.
28
Intensity: Intensity is the strength or power of a message in a direction. For Content Analysis
example, the characteristic of forgetfulness can be minor (e.g. not remembering
to take the keys when leaving home, taking time to recall the name of someone
whom you have not seen in years) or major (e.g. not remembering your name,
not recognizing your children).
Space: A researcher can record the size of the text messages or the amount of
space or volume allocated to it. Space in written text is measured by counting
words, sentences, paragraphs, or space on a page (e.g. square inches) for video
or audio text. Space can also be measured by the amount of time allotted.
To explain the differences among the above described quantification levels and
how they relate to constructing categories, let us give a hypothetical example of
‘Importance of FDI in public sector organizations’. The analyst has a major source
of information as newspaper, articles, public documents, transcripts of interview
with political leaders, and public officials. For each issue of each newspaper in
the sample, the analysts add together number of column inches from all news
articles/editorials to find he total number of space for each position in addition
to the coding, the name , location, and date of each newspaper. The analyst who
use this level of quantification have to assure that the differences they find in
amounts of space are valid indicators of relative emphasis or importance. At the
next level of quantification, the analyst can code the frequency of recording
units by tallying the number of times each issues or statements occurs in the text.
At the third level of quantification, analyst provide code for intensity. Frequencies
are counted but each coded statement or issues is also adjusted by a weight that
measures related intensity.
19.6.5 Code the Materials

Coding the unit of analysis into content category is called coding. Defining
categories and preparing coding schedule for analysis and coding of the content
are almost simultaneous steps. Material can be coded either manually or by
computers, depending on the sources available and format of the material. After
deciding how the material will be coded, the analyst writes the instructions for
coding. According to Kripendorff (1980), the guidelines for coding instructions
include definition of recording units and procedures for identifying theme,
descriptions of variables and categories, outline of cognitive procedures used in
placing data in categories, and instructions for using and administering data sheets.
Pre-testing: Pretesting is an important step before actual coding begins. It involves

coding a small portion of the material to be analyzed. A pretest enables the analyst
to determine whether (1) the categories are clearly specified and meet the necessary
requirement, (2) the coding instructions are adequate and, (3) the coders are
suitable for coding. These are determined on the basis of reliability among coders
band consistency in individual coding decision. If analyst find that material can
be coded with high reliability then actual coding begins.
19.6.6 Analyzing and Interpreting the Results

The main objective of content analysis is to analyze the collected information
with regard to the proposed objective of the analysis. The analysis involves
summarizing the coded data, discovering patterns and relationships within the
data, testing hypotheses about the patterns and relationship to assess the validity
of the analysis. 29
Qualitative Methods The absolute and relative frequencies are most commonly used for summarizing
the data. Absolute frequency might be the number of times statements or issues
are found in the sample. A relative frequency might be represented by a percentage
of the sample size.
Another way of analyzing content analysis is to examine relations among variables

by cross tabulating the cooccurrence of variables. Other techniques for discovering
patterns of relationship in data include contingency analysis, cluster analysis
and factor analysis.
One important development in analyzing content analysis of the data is the use
of computer. Computer programme like “General Inquirer” can identify within a
body of text , those words and phrases that belong to specified categories (Stone
et.al., 1996). Other computer programme like SPSS, Atlas titi, Minitab etc. are
also very useful in both quantitative and qualitative content analysis.
Whatever the technique used, a final and important task is to assess the validity
of the result by relating them to other data that are known to reasonably valid.
The inference a researcher can or cannot make on the basis of research is critical
in content analysis. Content analysis describes what is in the text. It cannot reveal
the intentions of those who created the text or the effects that messages in the
text have on those who receive them.

1) State the different sources of content analysis.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) Which methods are used to quantify the units of analysis?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
3) Briefly describes the different steps involved in the process of content
analysis.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
30
Content Analysis
19.7 USES OF CONTENT ANALYSIS
Content analysis is a scientific, objective, systematic, quantitative and generalizable
description of the communication content. This method can be used to understand
a wide range of themes such as social change, cultural symbols, changing trends
in the theoretical content of different disciplines, changes in mass media content,
nature of news coverage of social issues, election issues as reflected in mass
media. Content analysis can be used to examine anything that you see around
you.
It is used in several mass media and literature to cultural studies, psychology,

economics, political science, gender, age issues, as well as many other fields
where inquiry is made. Written documents, pictures, videos, can also be used for
content analysis.
In consumer behaviour and marketing, content analysis has been used to study
the following questions (Cited in Kassarjian, 1977).
1) What are product and company images of selected consumer goods as
reflected in the mass media. ( Stone, Dunphy & Bernstein, 1966).
2) Which of the several decision making models ,for example conmpensatory,
lexicographic, risk etc. are used by magazine and television advertisers
(Wright and Barbour, 1975).
3) What is the ease of readability of various marketing, advertising and consumer
research Journals. ( Lacho, Stearns, & Villere, 1975).
4) What are the changing values in society as reflected in the analysis of mass
periodical fiction (Johns-Heine, & Gerth, 1949)
The above example indicates the wide applicability of content analysis. Many
researchers have explored changes in women’s role, sexual behaviour and health
and violence by analyzing the content in television and movies messages (Olson,
1994).
Application of Content Analysis: An Illustration

A Research Study entitled “Gandhi’s Constructive Programme: A Study
of Its Contemporary Relevance”: Undertaken by Ms. Raunak Ahmad,
Research Scholar, (Gandhi and Peace Studies) IGNOU, New Delhi for
award of the Ph.D Degree.
The main objectives of the study were to:
(i) find out the gaps in the value and principles of social work profession and
its applicability in the Indian social setup, (ii) examine the perception of
social work students in India towards Gandhian principles of Constructive
Programme, (iii) study the relevance of principles of Gandhian Constructive
Programme from the perspective of students of social work discipline in
India and (iv) to pin point the importance of Gandhian Constructive
Programme to strengthen the indigenous base of contemporary social work
education.
31
Qualitative Methods
Research Methodology: Keeping in view the objectives and issues to be
probed in the study, the method of content analysis was considered appropriate
to conduct the study. The study was undertaken for an in-depth and systematic
analysis into the convolution of the phenomenon of Gandhi’s Constructive
Programme and its contemporary relevance.
In order to interrogate the values and principles of social work profession
and its subsequent applicability in Indian social setup, a detailed study of the
comprehensive MSW (Master of Social Work) syllabus of seven central
universities was undertaken. These seven universities were: University of Delhi,
Jamia Millia Islamia, Pondicherry University, Mahatma Gandhi Antarrashtriya
Hindi Vishwavidyalaya, Banaras Hindu University, Central University of
Rajasthan and Maulana Azad National Urdu University. Through conceptual
content analysis, the occurrence of the selected terms in the current social work
curriculum was studied. The text in the curriculum was carefully examined
and was classified under these categories: foundation of social work, theory
and practice of group work, working with communities, social work research,
social group work, social casework and counseling, history and philosophy
of social work etc. These categories were further subcategorized for deeper
analysis. Texts depicting similar meaning were put under single category and
none of the category was repeated. The process of content analysis further
went into detail when the common foreign scholars work and common Indian
scholars works prescribed under the curriculum of universities were created
as two separate categories. It helped to present a comparative picture. Categories
were then coded and the coded data were summarized and analyzed.
Main Results
1) Methods of social work such as community work, casework, group work,
social welfare management have been covered in the curriculum of all
the seven universities.
2) Majority of universities in the list are also offering social work history
and foundation as well as social work research.
3) The social action and social policy and planning being development
oriented are more meaningful in Indian context but are offered by
minimum number of universities.
4) The Gandhian ideology of social work is marginally touched in social
work curriculum of most of the universities under study.
5) 67 per cent of the suggested readings recommended in the curriculum
are of foreign scholars work.
6) The books on Gandhi, Gandhian ideology of social work written by
Gandhi are barely included in the suggested reading list of the curriculum
of these universities.
Apart from above, this study also made use of the method of content analysis
in studying the Gandhian principles of Constructive Programme. For this
purpose, the various documents introduced by Gandhi in the form of vows,
pledges, principles etc. were used. In this manner, the method helped to draw
the final set of principles of Constructive Programme which can further be
studied in the context of its relevance in social work education.
32
Content Analysis
CONTENT ANALYSIS
19.8.1 Advantages
The Advantages of content analysis as a research technique can be summarized
as follows:
1) The greatest advantage of content analysis is its economy in terms of time
and money. There is no requirement for a large research staff. No special
equipment is needed.
2) The methods allows the correction of errors. In content analysis, it is usually
easier to repeat a portion of the study than it is in other research method.
3) Content analysis permits the study of processes occurring over a long time.
4) Content analysis has the advantage of all unobtrusive measures that it has
any effect on the subject being studied.
5) It can present an objective account of events, themes, issues, and so forth,
that may not be immediately apparent to a reader or viewer .
6) It deals with large volume of data. Processing may be laborious but of late
computer has made it easy.
1) It is limited to the examination of recorded communication. Such
communication may be oral, written, or graphic, but they must be in some
fashion which permits analysis.
2) Content analysis may not be as objective as it claims since the researcher
must select and record data accurately. In some instances the researcher must
make choices about how to interpret particular form of behaviour.
3) It has both advantages and disadvantages in terms of reliability and validity.
Problems of validity is unlikely unless researcher happen to be studying
communication process itself.
4) It describes, rather explains people’s behaviour. It does not tell us what
behaviour means to those involved and those watching.
5) By attempting to quantify behaviour, this method may not tell us very much
about the quality of people’s relationship,

1) What are the different areas where the content analysis can be used as a
research technique?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
33
Qualitative Methods 2) Do you think that the technique of content analysis maintain objectivity?
Give reasons.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
19.9 LET US SUM UP

Content analysis is a multipurpose research method developed specifically for
investigating a broad spectrum of problems in which the content of
communication serves as the basis of inference. Content analysis is a set of
procedures for collecting and organizing information in a standardized format
that allows analysts to make inferences about the characteristics and meaning of
written and other recorded materials. Broadly content analysis is observed in
two forms – manifested content analysis and latent content analysis. Mainly two
approaches are followed in application of content analysis: conceptual analysis
and Relational analysis. In conceptual analysis, a concept is chosen for
examination, and the analysis involves quantifying and tallying its presence.
Relational analysis also termed as semantic analysis too begins with the act of
identifying concepts present in a given text or set of texts. However, relational
analysis seeks to go beyond presence by exploring the relationships between the
concepts identified. The various steps involved in content analysis include:
objective or formulation of research question, determining materials to be included
developing content categories, selecting and finalizing units of analysis, code
the materials, and analyze and interpret the results. Content analysis can be used
to understand a wide range of themes such as social change, cultural symbols,
changing trends in the theoretical content of different disciplines, election issues
as reflected in mass media. It can be used to examine anything that you see
around you. Content analysis has several advantages and limitations.
19.10 KEY WORDS

Objectivity : To have objectivity, the analysis must be carried
out on the basis of explicitly formulated rules
which will enable two or more investigators to
obtain the same results from the same documents.
Systematic : In a systematic analysis the inclusion and exclusion
of content or categories is done according to
consistently applied criteria of selection.
Generality : By generality, we mean that the findings must
have theoretical relevance unrelated to other
attributes of content or to the characteristics of
communicator or recipient of the message.
Manifest Content Analysis: It involves simply counting words, phrases, or
“surface” features of the text itself.
34
Latent Content Analysis : It involves interpreting the underlying meaning of Content Analysis
the text.
Unit of Analysis : The unit of analysis concerns the general idea or
phenomenon being studies.
Unit of Observation : Unit of observation concerns the specific item
measured at an individual level.
Word : The smallest unit generally used in the content
analysis as a unit.
Theme : The theme is a single assertion about a subject.
Character : Character may be defined as a use of fictional or
historical character.
Coding : Coding is the process whereby raw data are
systematically transformed and aggregated into
units which permit precise description of relevant
content characteristics.
Coding Scheme : Coding scheme is another phrase of coding
categories and coding book, within which all
instances of the content being analyzed or noted.
Master Code Book : It provides the coders with explicit instructions
and defines each word/ phrase/ aspect to be
analyzed.
Code Sheet : Code sheet provides the coders with a form on
which they note every instance of every word/
phrase/ aspect being analyzed.
19.11 REFERENCES AND SUGGESTED BOOKS

Altheide, David L. (1996). Qualitative Media Analysis. Qualitative Research
Methods Vol. 38. Thousand Oaks: Sage Publications.
Barcus, F.E., (1959). Communication analysis : analysis of the research, 1900-
1958 Unpublished doctoral dissertation, University of Illinois.
Berelson, B. (1952). Content Analysis in Communication Research, New York:
The Free Press
Budd, R.W., Thorp, R.K. & Donohue, L. (1967). Content Analysis of
Communications, New York: The Macmillan Co.
Carley, K. (1992). Coding choices for textual analysis: A comparison of content
analysis and map analysis. Unpublished Working Paper.
Chadwick, B.A., Bahar, H.M. & Albrecht, S.L. (1984). Content analysis. In
B.A. Chadwick et.al., Social Science Research Methods (pp. 239-257), New
de Sola Pool, I. (1959). Trends in Content Analysis. Urbana, Ill: University of
Illinois Press.
Gerbner, G., Holsti, O. R., Krippendorff, K., Paisley, W.J., and Stone, P.J., eds.
(1969). The analysis of Communications Content : Developments in Scientific
Theories and Computer Techniques, New York : Wiley. Jersey: Prentice –Hall. 35
Qualitative Methods Kerlinger, F.N. (1986). Foundations of Behavioural Research (3rd ed), New
York: Holt, Rinehart and Winston.
Kracauer, S. (1952). The challenges of qualitative content analysis. Public
Opinion Quarterly ,16,631-642.
Krippendorff, K. (1980). Content Analysis: An introduction to its Methodology,
Lacho, K.J., Stearns, G. K., and Villere, M.F. (1975).An analysis of the reliability
of marketing journals. Combined Proceeding, American Marketing Association,
489-497. London: Sage Publications.
Mostyn, Barbara (1985). The content analysis of qualitative research data: A
dynamic approach. In Michael Brenner, Jennifer Brown & David Canter (Eds.),
The research interview, uses and approaches (pp.115-145). London: Academic
Press.
Olson, B. (1994). Sex and the soaps: A comparative content analysis of health
issues, Journalism Quarterly, 71(4): 840-850.
Paisley, W.J., (1969). Studying style as deviation from encoding norms. In The
Analysis of Communications Content : Developments in Scientific Theories and
Computer Techniques,, eds. G. Gerbner, et al., New York : Wiley, 133-146.
Stone, P. J., Dunphy, D. C.m and Bernstein, A. (1966). The analysis of product
image. In The General Inqurier : A Computer Approach to Content Analysis,
eds. P.J. Stone, et al , Cambridge , Mass : The MIT Press.
Wittkowski, Joachim (1994). Das Interview in der Psychologie. Interviewtechnik
und Codierung von Interviewmaterial. Opladen: Westdeutscher Verlag.
Wright, P. And Barbour, F . (1975). The relevance of decision process model in
structuring persuasive messages.

PROGRESS EXERCISES
1) See Section 19.1 and Section 19.7
2) See Section 19.4
1) Written materials, recorded communications including TV programmes,
movies, photographs, regulations & other public documents, work places,
case studies, reports, newspapers, new press releases, books, journals, articles,
letters etc.
1) See Section 19.7
36
Content Analysis
UNIT 20 ACTION RESEARCH
Structure
20.0 Objectives
20.1 Introduction
20.2 Historical Background of Action Research
20.3 Definition of Action Research
20.4 Principles of Action Research
20.5 Characteristics of Action Research
20.6 Models of Action Research
20.6.1 Kenmis and McTaggart’s Spiral Model
20.6.2 Elliot’s Action Research Model
20.6.3 O’ Leary’s Cycles’ of Action Research Model
20.6.4 Stringer’s Interacting Spiral Action Model
20.6.5 Kurt Lewin’s Action Research Spiral Model
20.6.6 Calhoun’s Action Research Cycle Model
20.6.7 Bachman’s Action Research Spiral Model
20.6.8 Riel’s Progressive Problem Solving Model
20.6.9 Hendricks’s Action Research Model
20.7 Steps Involved in Action Research
20.8 Advantages and Disadvantages of Action Research
20.8.1 Advantages
20.9 Let Us Sum Up
20.10 Key Words
20.11 References
20.0 OBJECTIVES
• state the meaning, type and importance of action research;
• describe the principles guiding the action research;
• explain the various models of action research;
• learn the steps involved in action research; and
• discuss the advantages and disadvantages of action research.
20.1 INTRODUCTION
We have learned in Unit 1, 5, and 6 that two approaches – quantitative research
associated with dominant post positivist paradigm and qualitative research
associated with interpretative paradigm and critical theory paradigm have been
followed in undertaking research in Social Sciences. Under critical theory
paradigm, research and practice are integrated activities and both (researcher 37
Qualitative Methods and practitioner) guide each other. With emergence of critical theory approach
as an alternative to positivism and post positivism, participatory method and
emancipatory methods are also used to conduct the research studies in economics.
You have already studied participatory method in Unit 18. Emancipatory method
relies upon the process of knowing through doing rather than generation of
knowledge through conceptualization and theorizing. This strategy of knowledge
generation is termed as action research wherein it is believed that theory is only
really useful if it is put in service of a practice focusing on achieving social
change.
Action research is a powerful tool for change and improvement at the local level.
Kurt Lewin’s , one of the pioneer person for action research, work was intended
to change the life of disadvantaged groups in terms of housing, employment and
their working conditions. Action research is essentially research through action.
It is usually a collaborative activity which involves input from people who are
likely to be affected by the research. It involves deep inquiry into one’s
professional action. The researchers examine their work and look for opportunities
to improve. They work with others to propose a new course of action to help
their community, improve its work practices. They seek evidence from multiple
sources to help them, analyze reactions to the action taken. They recognize their
own view as subjective and seek to develop their understanding of the event
from multiple perspectives. The researcher uses data collection to characterize
the forces in ways that can be shared with practitioner. This leads to a new plan
for action.
Thus, action research aims to contribute both to the practical concerns of people
in an immediate problematic situation and to further the goals of social sciences
simultaneously. There is a dual commitment in action research to study a system
and concurrently to collaborate with members of the system in changing it in
what is together regarded as a desirable direction. Accomplishing this twin goal
requires the active collaboration of researchers and client and thus it stresses the
importance of co-learning as a primary aspect of the research process.
Hence, in this unit, we shall discuss the concept and type of action research, its
various models, steps involved in action research, its uses, advantages and
disadvantages. Let us start with the historical evolution of action research.
20.2 HISTORICAL BACKGROUND OF ACTION

RESEARCH
It is not certain who invented action research. Origin of Action research is
considered to begin from the work Kurt Lewin in 1940’s. He was strong exponent
of research action in its concern with power relations between researcher and
researched and the rights of the individuals. He was concerned with the social
problems and focused on participative group process for addressing conflict,
crises and change within organization. Lewin first coined the term action research
in 1946 in his paper “Action Research and Minority Problems”. He characterized
action research as: “A comparative research on the conditions and effect of various
forms of social action and research leading to social action, using a process of a
spiral steps, each of which is composed of a circle of planning, action and fact
– finding about the result of the action”.
38
Eric Trist was another major contributor to the field. He was a social psychiatrists. Action Research
Both, Lewin and Trist applied their research to systematic change in and between
organizations. They emphasized direct professional – client collaboration and
affirmed the role of group relations as basis for problem solving. Both were
strong proponents of the principle that decisions are best implemented by those
who help them. Alternatively, Deshler and Ewart (1995) suggested that action
research was first used by John Collier to improve race relations at the community
level when he was the Commissioner of Indian Affairs prior to and during the
Second World War, and Cooke (undated) appears to provide strong support for
this. It is, therefore, unlikely that we will ever know when or where the method
originated, simply because people have always investigated their practice in order
to make better or improve. Rogers’ (2002) account of John Dewey’s (1933)
notion of reflection, for instance, shows that it is very similar, and one could also
point to the ancient Greek empiricists as using an action research cycle. Action
research is difficult to define for two linked reasons: first, it is such a natural
process that it comes in many different guises, and second, it has been developed
differently for different applications. Almost immediately upon Lewin’s coining
of the term in the literature, action research was seen as a general term for four
different processes: diagnostic, participant, empirical and experimental.
20.3 DEFINITION OF ACTION RESEARCH

According to Frost (2002) ‘Action research is a process of systematic reflection,
enquiry and action carried out by individuals about their own professional
practice’. Hopkins (1985) suggests that action research is the combination of
action and research rendering action as a form of disciplined inquiry, in which a
personal attempt is made to understand, improve and reform practice. Ebbutt
(1985), too, regards action research as a systematic study that combines action
and reflection with the intention of improving practice.
Corey (1953) views action research as a process in which practitioners study

problems scientifically so that they can evaluate, improve and steer decision-
making and practice. ‘Action research is a term used to describe professionals
studying their own practice in order to improve it’. Kemmis and McTaggart
(1992) also argued that ‘to do action research is to plan, act, observe and reflect
more carefully, more systematically, and more rigorously than one usually does
in everyday life’. ‘Action research is ... usually described as cyclic, with action
and critical reflection taking place in turn. ‘Action research is a flexible spiral
process which allows action (change, improvement) and research (understanding,
knowledge) to be achieved at the same time’ (Dick, 2002). Cohen and Manion
(1980) described it as “essentially on the spot procedure designed to deal with a
concrete problems located in an immediate situation. This means that a step by
step process is constantly monitored over varying periods of time and by a variety
of techniques ….. diaries, interviews, case studies, etc., so that ensuing feedback
may be translated into modifications, adjustments, directional changes,
redefinitions as necessary”.
A more philosophical stance on action research is taken by Carr and Kemmis,

who regard it as a form of ‘self-reflective enquiry’ by participants, which is
undertaken in order to improve their understanding of their practices in context
with a view to maximizing social justice... Kemmis and McTaggart (1992)
suggest that: Action research is concerned equally with changing individuals, on 39
Qualitative Methods the one hand, and, on the other, the culture of the groups, institutions and societies
to which they belong. Action research is designed to bridge the gap between
research and practice (Somekh 1995) thereby striving to overcome the perceived
persistent failure of research practice.
Thus, Action research combines diagnosis, action and reflection, focusing on

practical issues that have been identified by participants and which are somehow
both problematic yet capable of being changed. Over the last decade, action
research has begun to capture the attention of teachers, administrators, and
policymakers around the country. Kemmis and McTaggart (1988) defined it very
effectively by encompassing the different definitions of action research.
Action research is a form of collective self-reflective enquiry undertaken

by participants in social situations in order to improve the rationality
and justice of the own social or educational practices, as well as their
understanding of these practices and the situations in which these
practices are carried out. . . . The approach is only action research when
it is collaborative, though it is important to realize that the action research
of the group is achieved through the critically examined action of
individual group members.
(Kemmis and McTaggart 1988)
Most simply put, action research (AR) implies a process of research where the
purpose of the research is not only to study the existing reality but to engage in
an effort to transform it. In this model, research assumes a catalytic role and
produces both a new dynamic and concrete, suitable changes in the reality that
can then be actively inducted into the process of knowledge creation. A
requirement of AR is that the separation between “researcher and researchee” is
dissolved—so as to avoid the weakness of conventional methods which view
“affected persons and groups” as being passive and incapable of analysing their
own situation and identifying solutions to their own problems. (Mukherjee, 2006.)

1) How action research is distinct from pure and simple applied research?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) What are the aims of action research?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
40
3) State the concept of action research coined by Kurt Lewin. Action Research
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
4) Do you think that action research play a catalysist role? Give reasons for
your answer in two sentences.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
20.4 PRINCIPLES OF ACTION RESEARCH

A set of six principles according to winter guide an action research. These
principles are discussed below:
1) Reflective Critical Analysis

It is the process of becoming aware of our own perceptual biases. For example,
description or narration of a situation in any notes, transcripts, or official
documents imply that it is factional and true. It claims to be reliable and
authentic. But truth in any social setting or social reality is in relative to the
teller. The principle of reflective critical analysis ensures people to reflect
on issues and practices and make explicit interpretations, biases, assumptions
and concern upon which judgments are made. In this way, practical
description can give rise to theoretical considerations.
2) Dialectical Critical Analysis

It is a way of understanding the relationships between the elements that
make up various phenomena in our context. Social reality is validated or
shared through language. Phenomena are conceptualized in dialogue.
Therefore a dialectical critical analysis is required to understand a set of
relationship both between the phenomenon and its context, and between the
elements constituting the phenomenon. The important point is to focus
attention on those constituent elements that are unstable or opposing one
another simply because these elements are most likely to create changes.
3) Collaborative Resource
The principles of collaborative resources presuppose that each person’s ideas
are equally significant as potential resources for creating interpretive
categories of analysis, which are negotiated among the participants. It makes
possible the insight gathered between many viewpoint and within a single
viewpoint. Hence everyone’s view is taken as a contribution to understand
the situation. Participants in action research are co-researchers.
4) Risk
It is an understanding of our own taken-for-granted processes and willingness
to submit them to critique. Normally, the change process is considered a
threat to all previously established ways of doing things, thus creating 41
Qualitative Methods psychiatric fears among the participants. One of the most prominent fears
emanates from the risk of ego generation from open discussion of one’s
interpretations, ideas and judgment. Initiator of action research will use this
principle to reduce other’s fears and invite participation by assuring that
they, too, will be the subject to the same process and that whatever be the
outcome, learning will take place.
5) Plural Structure
It involves developing various accounts and critiques, rather than a single
authoritative interpretation. The action research comprises of a multiplicity
of views, Commentaries and critical analysis which lead to multiple possible
actions and interpretation. This inquiry requires much texts for reporting.
Thus, there will be many explanations made clear and range of options for
action presented. A report, therefore, acts as a support for ongoing discussion
among collaborators rather than a final conclusion of fact.
6) Theory, Practice and Transformation

Theory and practice are considered two inter-dependent yet complementary
phases of the change process. For action researchers, theory informs practice
and practice refines theory in a continuous transformation. In any setting,
people’s actions are based on implicitly held assumptions, theories and
hypotheses. With every observed result, theoretical knowledge is enhanced.
Thus, a single change process has two interwined aspects. It is up to the
researchers to make the theoretical justifications explicit for action, and to
question the basis of that justification. The ensuring practical applications
as a follow up are subjected to further analysis in five cycles that continuously
alternates emphasis between theory and practice.
20.5 CHARACTERISTICS OF ACTION RESEARCH

Action research has the following characteristics:
1) It is collaborative, i.e. everyone’s view is taken as a contribution in
understanding the situation. Moreover, if a problem is faced by a practitioner
in a particular situation (say a school), action research can be collaborative
where practitioners facing similar problems in nearby schools can collaborate
to find solutions of a problem.
2) Action research helps in systems planning and restructuring. For
example, if a primary teacher finds that in his/her class the students are not
able to concentrate, the teacher starts finding the reasons for the same. After
analyzing the situation, the teacher finds that most of the children often
observe other children playing in the playground because of opening of
classroom window towards playground. They, therefore, are not able to
concentrate in their studies in the class. Now, what do you think a teacher
should do? Well, in such a case a teacher can change the seating plan of the
classroom. This is the way a teacher gets involved in restructuring the class.
3) Action research is a small-scale intervention. Its objective is to bring out
changes in the functioning of the practitioner himself/herself. It may or may
not have applicability for others. Action research is a narrowly focused
research undertaken by teachers and other practitioners in a given specific
42 situation and context.
4) Contextual nature is an important characteristics of action research. For Action Research
example, a teacher of a particular school may face a particular problem in
the form of errors committed by fifth grade students in English comprehension
in a school but the same problem may not be observed by him/her in other
schools.
5) It enhances the competencies of the practitioners. Action research enables
practiceners to have a clear vision of the problematic situation. This becomes
helpful in identifying ways and means to tackle the problem.
6) Action research seeks to improve the quality of human relationships.
7) Action research seeks to understand particular complex social situations
whether it is a class, school or community.
8) Action research allows us to identify remedial measures for improvement.
It is specific in nature, i.e. specific to a particular class, school or situation.
Therefore, results cannot be generalized.
9) It is a systematic and scientific process but not very rigorous.
10) Action research is participatory: it is research through which people work
towards the improvement of their own practices.
11) Action research makes use of quantitative and qualitative methodologies.
12) Action research is open-minded about what counts as evidence (or data) – it
involves not only keeping records which describe what is happening as
accurately as possible . . . but also collecting and analysing our own
judgements, reactions and impressions about what is going on.
13) Action research is a political process because it involves us in making changes

that will affect others.
14) It is a systematic learning process wherein people act deliberately, but

remaining open to surprises and responsive to opportunities.
15) Action research allows us to build records of our improvements and changing
activities and practices. It records the changes in the language and discourse
we describe, explain and justify our practices. It also documents changes in
the social relationships and forms of organization which characterize and
constrain our practices.
20.6 MODELS OF ACTION RESEARCH

Numerous authors and researchers have proposed models for undertaking the
action research. Because the process involved in action research is somewhat
dynamic, various models look a bit different from one another but possess
numerous common elements. Starting with a central problem or topic, action
research models involve some observation or monitoring of current practice,
followed by the collection and synthesis of information and data. Finally, some
sort of action is taken, which serves as the basis for next stage of action research.
Some models are simple in their design, while others appear relatively complex.
Let us discuss some models of action research as an illustration.
43
Qualitative Methods 20.6.1 Kemmis and McTaggart’s Spiral Model
These two authors describe this model as participatory research. In this model,
action research involves a spiral of self-reflective cycles of:
• Planning a change.
• Acting and observing the process and consequences of the change.
• Reflecting on these processes and consequences and then re-planning.
• Acting and observing.
• Reflecting, etc.
Fig. 20.1: Kemmis and McTaggart’s Action Research Spiral
Kemmis and McTaggart’s (2000) do not recommend to use it as a rigid structure.

They maintain that in reality the process may not be as neat as the spiral of self-
contained cycles of planning, acting, observing, and reflecting suggests. These
stages, they maintain, will overlap, and initial plans will quickly become obsolete
in the light of learning from experience. In reality the process is likely to be more
fluid, open, and responsive.
We find the spiral model appealing because it gives an opportunity to visit a

phenomenon at a higher level each time and so to progress towards a greater
overall understanding. By carrying out action research using this model, you can
understand a particular issue within a healthcare context and make informed
decisions with an enhanced understanding. It is therefore about empowerment.
However, Winter and Munn-Giddings (2001) point out that the spiral model
may suggest that even the basic process may take a long time to complete.
44
20.6.2 Elliot’s Action Research Model Action Research
The model employed by Elliot (1991) shares many of the features of that of
Kemmis and McTaggart and is based on Lewin’s work of the 1940s. It includes
identifying a general idea, reconnaissance or fact-finding, planning, action,
evaluation, amending plan and taking second action step, and so on, as can be
seen in Figure 20.2.
Identifying
Initial Idea
Reconnaissance
(Fact-finding and analysis)
CYCLE 1
General Pain
Action Steps 1
Action Steps 2 Implement
Action Steps 1
Action Steps 3
Monitor Implimentation
and effects
‘Reconnaissance’
(explain any failure to Revise general idea
implement, and effects)
Amended plan
CYCLE 2
Action Steps 1
Action Steps 2
Action Steps 3 Implement next
action Steps
Monitor Implimentation
and effects
Revise general idea
(explain any failure to
Amended plan
Action Steps 1
Action Steps 2
CYCLE 3
Action Steps 3
Implement next
Monitor Implimentation action Steps
and effects
(explain any failure to
Fig. 20.2: Elliot’s Action Research Model
Source: Ellliot, J. Action Research for Educational Change
20.6.3 O’Leary’s Cycles of Action Research Model

Cycles of action research shown in Figure 20.3, portray action research as a
cyclic process which takes shape as knowledge emerges. In this model, it is
stressed that ‘cycles converge towards better understanding of situation and
improved action implementation. Cycle of actions are based in evaluative practice
that alters between action and critical reflection’. O’Leary sees action research
as an experiential learning approach to change, where the goal is to continually
45
Qualitative Methods refine the methods, data, and interpretation in the light of understanding developed
in each earlier cycle.
Fig. 20.3: O’Leary’s Cycles of Research
20.6.4 Stringer’s Interacting Spiral Action Research Model

“Simple, yet powerful framework” consisting of a “look, think, and act” routine.
During each stage, participants observe, reflect, and then take some sort of action.
This action leads them into the next stage (see Figure 20.4).
Fig. 20.4: Stringer’s Interacting Spiral Action Research Model
20.6.5 Kurt Lewin’s Action Research Spiral Model

Stringer propounded interacting spiral model and viewed that “action research”—
also depicts an action research spiral, which includes fact finding, planning, taking
action, evaluating, and amending the plan, before moving into a second action
step (see Figure 20.5)
46
Action Research
Fig. 20.5: Lewin’s Action Research Spiral

Source: Adapted from Encyclopedia of Informal Education. (www.infed.org).
20.6.6 Calhoun’s Action Research Cycle Model

While not appearing as a “spiral,” still represents a process that is built around a
cyclical notion. As she describes, the solid lines indicate the primary direction of
the action research cycle through the phases in numerical order. The dotted lines
indicate backward and forward movement within the cycle as refinement or
clarification of information is warranted (see Figure 20.6).
Fig. 20.6: Calhoun’s Action Research Cycle
20.6.7 Bachman’s Action Research Spiral Model

Action research spiral continues this notion of the cyclical nature of action research
(see Figure 20.7). His downward spiral suggests that participants gather
information, plan actions, observe and evaluate those actions, and then reflect
and plan for a new cycle of the spiral, based on the insights that were gained in
the previous cycle.
47
Qualitative Methods
Fig. 20.7: Bachman’s Action Research Spiral Model
20.6.8 Riel’s Progressive Problem Solving Model

Through action research model takes the participants through four steps in each
cycle: planning, taking action, collecting evidence, and reflecting (see Figure
20.8).
48
Action Research
Fig. 20.8: Riel’s Action Research Model
20.6.9 Hendricks’s Action Research Model

It is shown in Figure 20.9. In her model, which she has placed in a school-based
context, focuses on acting, evaluating, and reflecting.
Fig. 20.9: Hendricks’s Action Research Process

Source: Adapted from Improving Schools Through Action Research: A Comprehensive Guide
for Educators, by Cher Hendricks, 2009, Boston: Allyn& Bacon.
49
Qualitative Methods From the above different models of action research, it can be inferred that the
following four stages are essential features of the model. However, this does not
mean that this is how all action research projects will work. The flexibility of
action research based on constant evaluation and reflection means that the cycles
may be truncated as new ways to proceed further.
i) Planning
Identifying the issue to be changed looking elsewhere for information. Similar
projects may be useful, for developing the questions and research methods
to be used to develop a plan related to the specific environment. In the school
setting this could involve personnel, budgets and the use of outside agencies.
ii) Acting
Trialling the change following your plan, collecting and compiling evidence,
questioning the process and making changes as required.
iii) Observing
Analysing the evidence and collating the findings, discussing the findings
with co-researchers and /or colleagues for interpretation, writing the report,
sharing your findings with stakeholders and peers.
iv) Reflecting
Evaluating the first cycle of the process, implementing the findings or new
strategy, revisiting the process.
These features might be represented diagrammatically as given below
Fig. 20.10
While working through action research remember that:

• It is cyclical and progress is made in small chunks,
• It is heavily based on critical reflection you can use.
50
• A wide range of methods for collecting data but it may be advisable to limit Action Research
these to a manageable number,
• Participants should have meaningful roles in the collection and presentation
of data because of the flexibility of the process and the constant reflection,
• Not every cycle will be complete.
• There may be times when it is advisable to stop mid stream and start a new
cycle.
1) What do you understand by dialectical critical analysis?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
2) How do multiplicity of views are accommodated in action research?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
3) Match the followings:
1) Kemmis and Mc Taggart a) Cycles of Action Research Model
2) Stringer b) Action Research Cycle Model
3) Kurt Lewin c) Spiral Model
4) Rial d) Intracting Spiral Action Model
4) Enlist the features of flexibility of action research.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
20.7 STEPS INVOLVED IN CONDUCTIONING

ACTION RESEARCH
Action research is a systematic process for finding the solution of the problem. It
can be conducted either individually or in collaboration with others. Different
models of action research have envisioned different stages. In general, following
steps are involved in conducting action research:
51
Qualitative Methods
Fig. 20.11
1) Identification of the Problem Area and Developing a Focus

Suppose a teacher may have several questions or problems to encounter
such as poor reading ability among his/her students, pronunciation problem
among students, effective monitoring of the various programs and many
more. Therefore, the focus of action research is on what students are
experiencing or have experienced? For example, a teacher can study how to
improve problem-solving skills in mathematics among the students or
increase reading ability among students and so on. It is therefore crucial to
choose the problem which is meaningful so that it can be solved in the
stipulated time.
The need for action research arises because of perceived dissatisfaction with
an existing situation. It is followed with the idea of bringing out improvement
in the situation. The focus is on the following: (i) what is the cause of problem?
(ii) Why is it happening? (iii) As a practitioner or a researcher, what can I do
about it? (iv) Which steps can I take to solve the problem? The answers to
all such questions are helpful in perceiving a problem as it exists which is a
pre-requisite for undertaking any action research problem.
2) Formulating the Problem

Once, the problem is identified, the next step is to formulate it. The researcher
tries to find causes underlying the problem along with various issues that are
related to causes. These probable causes need to be stated in concise and
unambiguous terms.
3) Stating the Research Questions and Development of Propositions

After formulating the problem, the researcher needs to state the research
questions and develop a tentative theory in the form of propositions keeping
in view the genesis of the problem. It is necessary to develop a conceptual
and functional relationship, tentatively to understand and explain the given
situation.
4) Data Collection
The collection of data is the most important step in deciding what action is
52 needed for solving the problem. For example, in the school, there could be
multiple sources of data, which a practitioner can use to identify causes and Action Research
developing, and implementing remedial measures. These include.
• Videotapes and audio tapes,
• Report cards,
• Attendance, samples of student work, projects, performances,
• Interviews with the parents, students etc.,
• Cumulative records and Anecdotal records,
• School Diaries,
• Photos,
• Questionnaires,
• Focus groups discussions,
• Checklists,
• Observation schedules.
Select the data that are most appropriate for the study. After collecting the
data, these are arranged on the basis of gender, classroom, grade level, school,
etc. The practitioner may use purposive samples of students or teachers from
each grade level in case of larger groups.
5) Analysis and Interpretation of Data

After the data has been gathered, the next step is to analyze the data in order
to identify trends and themes. The qualitative data obtained can be reviewed
to take out the common elements or themes and may be summarized in the
suitable table formats. The quantitative data can be analyzed with the use of
simple statistics such as percentages, simple frequency tables, or by
calculating simple, descriptive statistics.
At this step, the data is turned into information, which can help the practitioner
or the researcher in making decisions. Therefore, this stage requires maximum
time. After the analysis, it becomes clear what important points do these
data reveal and which important patterns or trends are emerging.
6) Discussion and Evaluating Actions

After the careful analysis of the data, review of current literature is done for
taking decisions and necessary actions. Following points need be kept in
mind while conducting the literature review:
i) Identifying topics that relate to the area of the study and would most
likely yield useful information.
ii) Gather or collect research reports, research books and videotapes relating
to the problem.
iii) Organise these materials for drawing inferences in the light of result of
the action research study.
Suggesting a plan of action will allow the practitioner to make a change.
This is well informed decision-making. It is advisable to suggest one action
at a time and then observe its outcome in improving the situation.
53
Qualitative Methods
ACTION RESEARCH
20.8.1 Advantages
1) Action research lends itself to use in work or community situations.
Practitioners, people who work as agents of change, can use it as part of
their normal activities. Mainstream research paradigms in some field
situations can be more difficult to use. Action research offers such people a
chance to make more use of their practice as a research opportunity.
2) When practitioners use action research, it has the potential to increase the
amount they learn consciously from their experience. The action research
cycle can also be regarded as a learning cycle. Systematic reflection is an
effective way for practitioners to learn.
3) Action research is usually participative. This implies a partnership between
you and your clients. You may find this more ethically satisfying. For some
purposes it may also be more occupationally relevant.
4) Action research is helpful in determining policy related to the problem.
1) Excuting an action research project is more difficult than conventional
research. Here a researcher takes responsibilities for change as well as for
research. In addition, as with other field of research, it involves you as a
researcher in more work to set it up, and you don’t get any credit for that.
2) It doesn’t accord with the expectations of some examiners. Deliberately and
for good reasons it ignores some requirements which have become part of
the ideology of some conventional research. In that sense, it is counter-
cultural.
3) As a researcher, one is not exposed much about action research. Action
research methodology is something that a learner has to learn almost from
scratch.
4) The library work for action research is more demanding. In conventional
research you know ahead of time what literature is relevant. In most forms
of action research, the relevant literature is defined by the data you collect
and interpret. That means that you begin collecting data first, and then go to
the literature to challenge your findings.
5) Action research is more difficult to report, at least for thesis purposes. If you
stay close to the research mainstream, you don’t have to take the same pains
to justify what you do. For action research, you are obliged to justify your
overall approach.
1) What does motivate you to undertake action research?
.......................................................................................................................
.......................................................................................................................
54 .......................................................................................................................
2) How do research questions enable to formulate the problem? Action Research
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
3) Identify the possible sources of data for an action researcher.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
4) What are the advantages of action research over conventional research?
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
5) Do you think that action research is a tough challenge before a conventional
researcher? Give two reasons in support of your answer.
.......................................................................................................................
.......................................................................................................................
.......................................................................................................................
20.9 LET US SUM UP

Action Research is associated with emancipatory method of critical theory
paradigm approach wherein knowledge is generated in the process of knowing
through doing rather through conceptualization and theorizing. In this process,
it becomes difficult to make a distinction between researcher and practitioner as
their activities become integrated.
Action research is a powerful tool for change and improvement at the local level.
Researcher in action research aims to contribute both to the practical concerns of
people in an immediate problematic situation and to further the goals of social
sciences simultaneously.
Origin of Action research is traced in the work Kurt Lewin in 1940’s. He was
strong exponent of research action in its concern with power relations between
researcher and researched and the rights of the individuals. Action research is a
form of collective self-reflective enquiry undertaken by participants in social
situations to improve the conditions of researched.
A set of six principles guide action research. These include: reflective critical
analysis, dialectical critical analysis, collaborative resource, risk, plural structure,
theory, practice and transformation. Different models of action research discussed
in this unit are: Kemmis and McTaggart action research model, Elliot’s Action
Research Model: O’Leary’s Cycles of Research (2004), Stringer’s model (2007),
Lewin’s Action Research Spiral, Calhoun’s Action Research Cycle, Bachman’s
55
Qualitative Methods Action Research Spiral, Riel’s Action Research Model and Hendricks’s Action
Research Process.
The steps followed in action research include: Identification of the Problem

Area and Developing a Focus, Formulating the problem, Stating the Research
Questions and Development of Propositions, Data Collection, Analysis and
Interpretation of Data, Discussions and Evaluating Actions. Action research has
several advantages over conventional research. However, it has several limitations
as well.
20.10 KEY WORDS

Action Research : Learning by doing. A group of people
identifies a problem, do something to resolve
it, see how successful their effects were and if
not satisfied, try again.
Reflective Critical Analysis : It is the process of becoming aware of our
own perceptual biases.
Dialectical Critical Analysis : It is a way of understanding the relationships
between the elements that make up various
phenomena in our context.
Collaborative Resource : Which is intended to mean that everyone’s
view is taken as a contribution to understand
the situation.
Risk : It is an understanding of our own taken-for-
granted processes and willingness to submit
them to critique.
Plural Structure : It involves developing various accounts and
critiques, rather than a single authoritative
interpretation.
Collaborative : Everyone’s view is taken as a contribution in
understanding the situation.
Reconnaissance : Fact finding and explanation.
Planning : Identifying the issue to be changed, developing
the questions and research methods to be used
developing a plan related to the specific
environment.
Acting : Trialling the change following your plan,
collecting and compiling evidence.
Observing : Analysing the evidence and collating the
findings, discussing the findings with co-
researchers and/or colleagues for the
interpretation.
Reflecting : Evaluating the first cycle of the process,
implementing the findings or new strategy,
56 revisiting the process.
Action Research
20.11 REFERENCES
Calhoun, E. F., (1994). How to use Action Research in Self Renewing School.
Alexandria. V A, ASLD.
Carr, W. and Kemmis, S. (1986). Becoming Critical: Education, Knowledge
and Action Research. London: Falmer.
Ebbutt, D. (1985). Educational Action Research: Some General Concerns and

Specific Quibbles. In R. Burgess (ed.) Issues in Educational Research: Qualitative
Methods. Lewes: Falmer, 152–74.
Elliot, J. (1991). Action Research for Educational Change. Buckingham: Open

University Press.
Ernest T. Stringer (2007). Action Research (Third Edition). London: Sage

Publications, 279pp (pb), ISBN: 978-1-4129-5223-1, $46.95.
G. E. Mills (2011). Action Research: A Guide for the Teacher-Research (4th ed.).
Pearson.
Hendricks. C. (2009). Improving School Through Action Research; A

Comprehensive Guide for Educator. (2nd ed.). Upper Saddle River, N J: Pearson.
Hopkins, D. (1985). A Teacher’s Guide to Classroom Research, Philadelphia:

Open University Press.
Kemmis, S. and McTaggart, R. (1992). The Action Research Planner (third

edition) Geelong, Vic.: Deakin University Press.
Kemmis, S. and McTaggart, R. (2000). “Participatory Action Research”, in N.K.

Denzin and Y.S. Lincoln (eds) Handbook of Qualitative Research (2nd ed.). Sage,
CA, pp. 567–605.
McNiff, J., and Whitehead, J. (2002). Action Research: Principles and Practice
(second edition). London: RoutledgeFalmer.
O’Leary, Z. (2004). The Essential Guide to Doing Research. London: Sage

Publications.
Richard. Winter and Carol Munn-Giddings(2001). A Handbook for Action

Research in Health and Social Care. by Routledge, New York NY 10017.
Selener, D. (1997). Participatory Action Research and Social Change. New

York: Cornell Participatory Action Research Network.
Zuber-Skerritt, O. (ed.) (1996). New Directions in Action Research, London;

Falmer Press.
57
Qualitative Methods
PROGRESS EXERCISES
1) See Section 20.1
2) See Section 20.1
3) See Section 20.2
4) See Section 20.3
3) See Section 20.6
1) See Section 20.7
2) See Section 20.7
3) See Section 20.7
58
MEC-109
Research Methods
Block
6
DATA BASE OF INDIAN ECONOMY
UNIT 21
Macro-Variable Data: National Income, Saving and
Investment 9
UNIT 22
Agricultural and Industrial Data 37
UNIT 23
Trade and Finance 63
UNIT 24
Social Sector 85
Expert Committee

Varanasi New Delhi


Former Professor of Economics Professor of Economics
Bombay University and Former IGNOU, New Delhi
Chairman ICSSR, New Delhi

Faculty of Social Sciences IGNOU, New Delhi
Banaras Hindu University, Varanasi



(Format, Language and Content Editing)
21-24 Shri S.S. Suryanarayan Prof. Narayan Prasad
Ex Joint Advisor Professor of Economics
Planning Commission IGNOU, New Delhi
Renamed as Niti Ayog
New Delhi
Updated by
Mr. Purnendu Kishore Banerjee
Dy. Registrar General and Census Commissioner
New Delhi
Print Production
Mr. Manjit Singh
October, 2015
ISBN-978-81-266-
Printed at :
BLOCK 6 DATA BASE OF INDIAN ECONOMY
For undertaking any meaningful research in terms of situational assessment,
testing of models, development of theory, evolving economic policy, assessing
the impact of such policy etc., data is crucial. Data on different variables is thus
an essential input for assessment and analysis of economic situations. The
availability or otherwise of data, therefore, determines the scope of analysis.
The reliability of the conclusions arrived at also depends on the availability and
veracity of data. Hence, researcher’s knowledge about the availability of data is
important for conducting a meaningful research. Keeping in view that a student
of economics at postgraduate level is expected to know the available databases,
the present block deals with the different databases of Indian Economy.
The block comprises of 4 units. Unit 21 deals with the data available on major
macro variables relating to the economy – national income, saving and
investment, etc. Unit 22 throws light on the kind of agricultural and industrial
databases. Unit 23 deals with data on trade and finance.
Unit 24 discusses the availability of data on other social sectors like

employment and unemployment, education, health, shelter and amenities,
environment, quality of life etc. These four units lay particular emphasis on the
different concepts used in data collection including the agencies involved in the
compilation of data.
Abbreviations Used in the Block
ADR - American Deposit Receipt

AHSD - Animal Husbandry Statistics Division
AIES - All India Educational Survey
ANM - Auxiliary Nurse-Midwife
ASI - Annual Survey of Industries
ASDR - Age – Specific Death Rate
ASG - Agricultural Statistics at a Glance
BAHS - Basic Animal Husbandry Statistics
BC - Backward Classes
BD - Budgetary Deficit
BMI - Body Mass Index
BOD - Biological Oxygen Demand
BoP - Balance of Payments
BR - Birth Rate
CACP - Commission on Agricultural Costs and Prices
CAD - Current Account Deficit
CAS - Current Account Surplus
CB - Commodity Boards
CBHI - Central Bureau of Health Intelligence
CCPS - Cumulative Convertible Preference Shares
cfc - Consumption of Fixed Capital (depreciation)
CGHS - Central Government Health Scheme
CGWB - Central Ground Water Board
CHC - Community Health Centre
CIF - Cost, Insurance and Freight
CIS - Commonwealth of Independent States
CMI - Census of Manufacturing Industries
CMIE (EIS) - Centre for Monitoring Indian Economy (Economic
Intelligence Services)
CPI-AL - Consumer Price Index – Agricultural labour
CPI-IW - Consumer Price Index – Industrial Workers
CPI-UNME - Consumer Price Index – Urban Non-Manual Employees
CSO - Central Statistical Organisation
CWC - Central Water Commission
CWS - Central Weekly Status
DCI - Dental Council of India
DDP - District Domestic Product
DE - Directory Establishments
DEs - Directorates of Employment
DESMOA - Directorate of Economics & Statistics, Ministry of
Agriculture
DES(s) - Directorate(s) of Economics & Statistics
DGE&T - Directorate General of Employment & Training
DIPP - Department of Industrial Policy and Promotion
DR - Death Rate
EBs - Electricity Boards
EMI - Employment Market Information
EPC - Export Promotion Council
EPFO - Employees Provident Fund Organisation
EPW - Economic and Political Weekly
EPWRF - Economic and Political Weekly Research Foundation
ESI Act - Employees’ State Insurance Act
ESIC - Employees State Insurance Corporation
EXIM Bank - Export Import Bank
FDI - Foreign Direct Investment
FIs - Financial Institutions
FIIs - Foreign Institutions Investors
FIPB - Foreign Investments Promotion Board
FISIM - Financial Intermediation Services Indirectly Measured
f.o.b. - free on board
FRG - Federal Republic of Germany
FSI - Forest Survey of India
FSS - Farmers’ Service Societies
GDI - Gender Development Index
GDCF - Gross Domestic Capital Formation
GDFCF - Gross Domestic Fixed Capital Formation
GDI - Gender Development Index
GDP - Gross Domestic Product
GDR - Global Deposit Receipt
GDS - Gross Domestic Saving
GFCE - Government Final Consumption Expenditure
GFD - Gross Fiscal Deficit
GNP - Gross Primary Deficit
GPD - Gross Primary Deficit
GSDP - Gross State Domestic Product
GTT - Gross Terms of Trade
HA - Health Assistant
HDFC - Housing Development Finance Corporation
HDI - Human Development Index
IAMR - Institute of Applied Manpower Research
IBM - Indian Bureau of Mines
ICICI - Industrial Credit and Investment Bank of India
ICSSR - Indian Council of Social Science Research
IDBI - Industrial Development Bank of India
IIP - Index of Industrial Production
IIPS - International Institute for Population Sciences
IMF - International Monetary Fund
IMR - Infant Mortality Rate
I-O TT - Input Output Transaction Table
IPO - Initial Public Offering
IPP - Index of Prices Paid
IPR - Index of Prices Received
ISCED - International Standard Classification of Education
ISCO - International Standard Classification of Occupations
ISIC - International Standard Industrial Classification
ISIC – Rev. 3 - International Standard Industrial Classification –
Revision 3
ISM - Indian Systems of Medicine
ITC (HS) - Indian Trade Classification (based on Harmonised
Commodity Description and Coding System)
ITC – Rev. 2 - Indian Trade Classification – Revision 2
ITES - Information Technology Enabled Services
ITT - Income Terms of Trade
KVIC - Khadi and Village Industries Commission
LAMPS - Large – sized Adivasi Multi-Purpose Societies
LHV - Lady Health Visitor
LIC - Life Insurance Corporation of India
MCI - Medical Council of India
MF - Mutual Fund
MHA - Male Health Assistant
MHRD - Ministry of Human Resource Development
MI - Minor Irrigation
MMR - Maternal Mortality Rate
MOEF - Ministry of Environment and Forests
MoLE - Ministry of Labour & Employment
MoSPI - Ministry of Statistics & Programme Implementation
MPCE - Monthly Percapita Consumption Expenditure
MSP - Minimum Support Price
NABARD - National Bank for Agricultural and Rural
Development
NABS - National Advisory Board on Statistics
NAS - National Accounts Statistics
NBFC - Non-Banking Finance Companies
NBNFC - Non-Banking Non-Finance Companies
NCERT - National Council for Educational Research and
Training
NCO - National Classification of Occupations
NCI - Nursing Council of India
NCS - Net Capital Stock
NDC - National Development Council
NDE - Non-Directory Establishments
NDCF - Net Domestic Capital Formation
NDFCF - Net Domestic Fixed Capital Formation
NDP - Net Domestic Product
NDS - Net Domestic Saving
NEER - Nominal Effective Exchange Rate
NFCS - Net Fixed Capital Stock
NFD - Net Fiscal Deficit
NFHS - National Family Health Survey
NFE - Non-Formal Education
NHB - National Housing Bank
NIC - National Industrial Classification
NNP - Net National Product
NNR - Nano-Natal Rate
NPA - Non-Performing Assets
NPD - Net Primary Deficit
NPISH - Non-Profit Institutions Serving Households
NRI - Non-Resident Indian
NSDP - Net State Domestic Product
NSE - National Stock Exchange
NSRC - National Statistical Commission Report
NSSO - National Sample Survey Organisation
NTMIS - National Technical Manpower Information System
NTT - Net Terms of Trade
OAE - Own Account Enterprises
OECD - Organisation for Economic Co-operation and
Development
OPEC - Oil and Petroleum Exporting Countries
PACS - Primary Agricultural Credit Society
PCI - Pharmacy Council of India
PFCE - Private Final Consumption Expenditure
PHC - Public Health Centre
PHSC - Public Health Sub-Centre
PISS - Priced Information Service Systems
PLR - Prime Lending Rate
PNR - Post – Natal Rate
PPF - Public Provident Fund
PRD - Primary Revenue Deficit
QI - Quantum Index
RAS - Regional Accounts Statistics
RBI - Reserve Bank of India
REC - Rural Electrification Corporation
REER - Real Effective Exchange Rate
RD - Revenue Deficit
RGI - Registrar General of India
RRBs - Regional Rural Banks
RS - Remote Sensing
SAS - State Accounts Statistics
SCBs - Scheduled Commercial Banks
SBR - Still Birth Rate
SC & ST - Scheduled Casts & Scheduled Tribes
SDDS - Special Data Dissemination Standards
SDP - State Domestic Product
SDR - Special Drawing Rights
SEBI - Securities Exchange Board of India
SIA - Secretariat for Industrial Approvals
SIC - Standard Industrial Classification
SIDBI - Small Industries Development Bank of India
SITC – Rev. 2 - Standard International Trade Classification –
Revision 2
SLR - Statutory Liquidity Ratio
SNA - System of National Accounts
SPM - Solid Particulate Matter
SRS - Sample Registration Scheme
SSMI - Sample Survey of Manufacturing Industries
UGC - University Grants Commission
UN - United Nations
UPS - Usual Principal Status
UPSS - Usual Principal and Subsidiary Status
UTI - Unit Trust of India
UVI - Unit Value Index
VRC - Vocational Rehabilitation Centre
WC Act - Workmen’s Compensation Act
WPI - Wholesale Price Index
WPR - Worker-Population Ratio
WTO - World Trade Organisation
Macro-Variable Data:
UNIT 21 MACRO-VARIABLE DATA: National Income, Saving
and Investment
NATIONAL INCOME SAVING
AND INVESTMENT
Structure
21.0 Objectives
21.1 Introduction
21.2 The Indian Statistical System
21.3 National Income and Related Macro Economic Aggregates
21.3.1 System of National Accounts (SNA)
21.3.2 Estimates of National Income and Related Macroeconomic Aggregates
21.3.3 The Input-Output Table
21.3.4 Regional Accounts – Estimates of State Income and Related Aggregates
21.3.5 Regional Accounts – Estimates of District Income
21.3.6 National Income and Levels of Living
21.4 Saving
21.5 Investment
21.6 Let Us Sum Up
21.7 Exercises
21.0 OBJECTIVES
• know the various approaches followed by the Indian statistical system in
generating data on various economic and social phenomena;
• explain the SNA (System of National Accounts) methodology of data
compilation;
• identify the various aggregates on which estimates are made by the
National Accounts System;
• compare the interrelationship among different sectors of the economy
through input-output transaction table (I-OTT) compiled by CSO; and
• describe the data on National Income, Saving, Investment and other
related macro economic variable and the estimates of the state income
which enable analysis of various aspects of economy.
21.1 INTRODUCTION
You would soon be a professional economist and would possibly be engaged in
research or would function as an economic analyst or adviser in a Government
or non-Government organization. As a professional economist, you will have
to analyse situations and arrive at conclusions about them or develop solutions 9
Data Base of Indian for further action to tackle such situations. You will, for this purpose, identify
Economy the variables that, in your opinion, shape or affect the problem at hand, N
incorporate these in a suitably chosen model, test the efficacy of the model in
reflecting the situation under examination and proceed further with the
analysis. A basic input for such activity is data on the variables incorporated in
the model. The availability of data relevant to the proposed analysis enhances
or restricts the scope of your analysis and the reliability of the conclusions
arrived at on the basis of the analysis. Indeed, it might even restrict the kind of
variables and model you would like to use in your analysis, unless you want to
conduct a survey yourself to collect the data that are not available. Your
technical and professional equipment as an economist is, therefore, incomplete
without knowledge of where data on the variables in question are available and
how good they are for the purpose of the analysis at hand. This knowledge
should naturally cover the entire data base of the Indian economy since an
economist’s work can cover any part or the whole of the Indian economy.
We shall look at the data base of the Indian economy in this Block. This Unit
deals with data available on National Income, Saving and Investment – the
major macro variables relating to the economy and associated macroeconomic
aggregates. Agricultural and Industrial data, data relating to the Social Sector,
Trade and Finance will be discussed in the subsequent units of this block.
Let us begin by taking a look at the Indian statistical system.
21.2 THE INDIAN STATISTICAL SYSTEM

The Indian Statistical System generates data on a variety of economic and
social phenomena, essentially through six approaches. First, Central Acts like
the Census Act, 1948; and the Collection of Statistics Act, 2008, etc. enable the
Government agencies to conduct large-scale sample surveys for collection of
data at regular intervals. Second, statutory returns prescribed under several
other Acts like the Factories Act, 1948 the Companies Act, 1956, the Reserve
Bank of India Act, 1953; the Registration of Births and Deaths Act, 1969 and
the implementation of these Acts generate data on matters not covered by the
surveys. Third, data collected by individual Ministries, Departments and
organizations of the Central and State Governments as part of their specific
functions reflect the emerging situation in different sectors and sub-sectors of
the economy and administrative divisions of the country. Fourth, the
administrative reports of these organizations supplement such data. Next,
information derived from the data flows mentioned above, like the National
Accounts Statistics (NAS), index numbers of prices and production and
indices like the Human Development Index and Gender Development Index
provide readily usable inputs for research and policy and evaluation of the
impact of policies and programmes of development in terms of the health of the
economy and the well-being of the society. Lastly, a large number of surveys
and research studies conducted by various institutions, public and private, on a
variety of subjects constitute another flow of data and information.
The Ministry of Statistics and Programme Implementation of the Government

of India is the apex body in the official statistical system of the country. The
Ministry is headed by the Chief Statistician of the country. The Ministry
consists of the Central Statistics Office (CSO), the National Sample Survey
Office (NSSO) and the Computer Centre (CC). The Directorates of Economics
and Statistics function at the level of the State/UT Governments. The CSO
10
coordinates the statistical activities in the country, lays down and maintains Macro-Variable Data:
statistical norms and standards and provides liaison with Central, State and National Income, Saving
and Investment
International statistical agencies. There is also a National Statistical Commission
(NSC), comprising of Chairman and Members, who are eminent economists
and statisticians from research institutions, representatives of Central Ministries
and Departments and the State Directorates of Economics and Statistics, to
provide (i) guidance for an overall perspective for statistical development in the
country, (ii) guidance to Government on policy issues, and (iii) ensure effective
coordination of all statistical activities of the Government of India. CSO is the
Secretariat of the NSC.
The CSO, as the Central Statistical Authority, is responsible for coordination of

statistical activities in the country and for evolving and maintaining statistical
standards. Mention should be made in this connection of three publications of
CSO, namely, the Sources and Methods of National Accounts Statistics, the
National Industrial Classification and the Consumer Price Index. The first two
are adhoc while the third a monthly one. The first one covers the methods and
classification principles to be followed for preparation of the macro economic
aggregates, the second one gives the classification principle to be followed for
national and international comparability and the third one provides State/UT
wise price indices.
Certain other agencies concerned with economic development also bring out
publications that cover data relating to most of all the sectors of the economy at
one place. These are the Reserve Bank of India (RBI) Bulletin (monthly),
the Financial Stability Report (including Trend and Progress of Banking
in India), RBI’s Handbook of Statistics on the Indian Economy and the
website of the RBI, the pre-Budget Economic Survey and the Budget presented
to Parliament by the Finance Minister every year. The Five-Year Plan
documents of the Planning Commission of the Government of India used to
make up yet another set of data sources that provides important data across
sectors. The Human Development Report was prepared by the Planning
Commission for the first time in the year 2001. This and similar reports
prepared subsequently by many of the State Governments for individual
States and for district constituted another set of sources for data across sectors
at the national, State, district and even sub-district levels. The District Census
Handbooks published after successive population censuses by the Office of
the Registrar General of India (ORGI) and the Directorates of Census of
different States and Union Territories provide comprehensive information
across sectors on individual districts, sub-districts, towns and villages. The
ORGI, in the year 2014 has also launched a digital library of all census tables
published since Census 1991 in their website www.censusindia.gov.in. These
tables together provide a very useful time series data on demographic
characteristics of the country and its States/ UTs. Most of the data and reports
published by Government agencies are available in electronic format, that is, in
DVDs/CDs and are accessible at the websites of the agencies concerned. In the
year 2013, the Planning Commission has also set up a central web-based
repository, namely, www.data.gov.in to bring published data of different
Government agencies under one roof.
The Indian statistical system has also recently been reviewed so as to bring
about improvements in it. A National Statistical Commission was appointed by
the Government of India to examine critically the deficiencies of the existing
statistical system from the point of view of timely availability, reliability and
adequacy of data and to recommend measures to correct these deficiencies and 11
Data Base of Indian revamp the statistical system to generate reliable statistics for the purpose of
Economy policy and planning at different levels of Government. The Commission N
submitted its report to Government in September, 2001. The Report of
National Statistical Commission (2001), published by the Ministry of
Statistics and Programme Implementation (MoSPI), is itself a useful
reference book on the database of the Indian economy. This can be viewed or
downloaded from the website www.mospi.nic.in/nscr/hp.htm. Most of data
available in Government websites can be downloaded free of cost, either
directly or as a registered user. In most cases, this registration is free of charge.
For example, one can make a free registration in the website of the MoSPI and
with the username, freely download all the reports and tables published by
different offices under the MoSPI. For the Office of the RGI, one can simply
visit its website www.censusindia.gov.in click on the link digital library and
download all the census tables published from 1991 census onwards.
There are also non-Government sources that publish these for the use of
researchers and other data users (Certain private websites like www.indiastat.
com provide India centric, sector specific and state specific data for the
research fraternity, usually for a fee). The Centre for Monitoring Indian
Economy (CMIE), Mumbai and its Economic Intelligence Services (EIS)
provide detailed and up-to-date information on the Indian economy through a
Monthly Review of Indian Economy and annual documents each covering a
specific sector or subject in great detail. These publications bring together at
one place not only data available with Government statistical agencies but also
data collected by CMIE itself. The Economic and Political Weekly Research
Foundation (EPWRF), Mumbai is another organization that publishes
detailed time series data on a variety of subjects along with a description of
related conceptual and methodological issues and quality of data, besides
presenting statistics on specific areas in subject-specific Special Issues of the
Economic and Political Weekly. The search for data by an analyst like you
can well start from one or more of the comprehensive sources enumerated
above and move on, as necessary, to the primary agency concerned with the
specific area of your interest.
Data on any variable have to be classified suitably for facilitating analysis.

And the classification system needs to be standardized when data on any
variable are collected by several agencies and utilized by many in order to
ensure meaningful collection of data and comparability of data across space
and time. At the international level, different agencies of the United Nations,
like, the United Nations Statistics Division (UNSD), the OECD, the
International Labour Office (ILO), etc. prepares these classifications for cross-
country comparison. For individual countries, these classification systems are
recommended. The countries are advised to maintain full compatibility up to a
certain level and then make additions/ alterations to suit the country specific
needs. In India, usually, such classifications are expanded to suit the need of
the country, while maintaining international comparability. Two important
classificatory systems developed for use may be referred to here, as its utility
cuts across sectors. One is the National Industrial Classification (NIC),
developed by the Central Statistics Office (CSO) in consultation with all
concerned. It groups the entire spectrum of economic activities by means of a
five-digit code structure. NIC has been revised from time to time in keeping
with changes in the structure and variety of economic activities over the years.
A Standard Industrial Classification, 1962 (SIC 62) developed by CSO was
being used in the Sixties until NIC 70 was introduced. This was replaced by
12 NIC 1987, which in turn was replaced by NIC 98 in December 1998. The most
recent revision of the NIC series is NIC 2008. All the NIC classifications are Macro-Variable Data:
based on the relevant UN International Standard Industrial Classification National Income, Saving
and Investment
(ISIC). NIC 2008 was, for instance, based on ISIC- Revision 4. NIC 2008 has
been developed and released for use in the place of NIC 04 in September,
2008. NIC 2008 classifies the whole range of economic activity into 21
Sections, 88 Divisions, 238 Groups, 403 Classes and 1,304 Sub-classes. NIC
2008 is comparable completely with ISIC-Rev.4 up to the four-digit level. The
fifth digit (sub-class) inside each of the class (4thdigit) takes care of the special
features of the Indian economy. NIC is used for classifying data on variables
like output, employment and income by economic activity. In each release of
the NIC, concordance tables are made available to facilitate recasting of past
data as per the new classification, to the extent possible.
The other classificatory system is the National Classification of Occupations

(NCO), developed by the Directorate General of Employment & Training
(DGE&T) of the Ministry of Labour & Employment. NCO 1958, a five-
digit code structure, was developed in the Fifties. This was revised to NCO
1968. This was in use till recently. This has now been replaced by NCO 2004
(NCO04). The Indian NCOs have been based on the international occupational
classification system developed by the International Labour Organisation (ILO)
and revised from time to time. NCO 1968 was patterned on the International
Standard Classification of Occupations 1966 – ISCO 66 – of the ILO.
While revising ISCO 66 to prepare ISCO 88, ILO incorporated the concept of
skill for effective performance in the relevant occupation, utilizing for the
purpose the International Standard Classification of Education (ISCED).
NCO 04 has followed the approach adopted in ISCO 88. It takes note of the
skill required for satisfactory performance in the relevant occupations and has a
six-digit code structure. NCO 04 has 8 occupational divisions (the first digit of
the code), 95 occupational groups (the second digit), 462 occupational families
(the third digit) and 2484 occupations (the fifth and the sixth digits). NCO is
usuful for classification of the job seekers, the jobs and employment, thereby
facilitating analysis of trends in the labour market and in the employment
potential of the economy. Past data can be recast as per NCO 04 with the help
of concordance tables provided in NCO 04.
A word of caution is necessary while using data assembled from sources other
than one’s own. Data extracted from printed publications or websites have to
be checked for printing errors/errors that occur while data are being posted on
the websites. A check of the row and column totals of the tables and other
identity relations relevant to the data under reference in the tables should take
care of such errors. More importantly, you should also look into aspects of the
data, like the concepts, definitions and methodology adopted by the source-
agency for collecting and compiling the data, coverage of data (like, whether
data was collected from entire country or some specific areas, whether the
sample drawn was based on a statistically reliable random sample or it was a
purposive sample of a few selected locations, etc.), the methods used for
collecting data, reliability and so on periodicity and timeliness in the
availability of data, integrity (confidentiality), the data are firmly moored to
what the data are actually supposed to reflect. These aspects are collectively
referred to as “metadata”. All these details are generally available in the
publication containing the data or in a related publication of the agency
publishing the data or in the relevant website of the agency.
It would sometimes be useful and instructive to judge the Indian situation in an

international setting. Publications of the United Nations and its agencies as also 13
Data Base of Indian agencies like the International Monetary Fund (IMF) and the World Bank and
Economy regional agencies such as the Organisation for Economic Cooperation and N
Development (OECD)1 provide data on different facets of the economies of the
member countries including India along with information regarding
comparability of country-wise data. The IMF has formulated a “Special Data
Dissemination Standards” (SDDS) covering the real sector (national
accounts, production index, price indices, etc.), Fiscal Sector, Financial Sector,
External Sector and Socio-demographic data for different countries to facilitate
transparency in the compilation and dissemination of data on important aspects
of the economy of individual countries and cross-country comparison of such
data. Countries like India who have accepted the SDDS provide to IMF a
National Summary Data Page in respect of each of the areas and sub-areas
listed in SDDS and metadata relating to such data, as per a Dissemination
Format prescribed in SDDS. In addition, individual country agencies concerned
disseminate an advance release calendar, which gives notice of the precise
dates of release of data three months ahead of the date of release of the data, on
the internet of the IMF’s Data Dissemination Bulletin Board (DSBB). Such
information, as provided by any country covered by SDDS, can be accessed on
the internet with the help of the Google search engine using the search
parameter Special Data Dissemination Standards IMF. You can also visit the
site of the United Nation’s Statistical Division databases at http://unstats.un.org/
unsd/databases.htm to access data compiled by different UN agencies on a host
of economic and social parameters.
Your capacity to utilize the vast amount of data available in the Indian
Statistical System and examine critically the state of the society can be
enhanced considerably by an intelligent combination of data across data
domains. You have already come across some examples of such efforts –
poverty ratios, concentration ratios for measuring the level of inequality,
employment elasticity, labour productivity and capital productivity. (Unit 4 on
Poverty and Inequality – Policy Implications and Unit 5 on Employment and
Unemployment – Policy Implications in Block 1 of MEC-005 on Indian
Economic Policy in your first year Course). Other examples are the Human
Development Index and Gender Development Index (Units 3 and 4 in the same
Block of MEC-005).

1) State the major functions of CSO.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
1
India is not a member of OECD, which is an organization in which several countries in
14 Europe are members.
2) Name any three web sources where you can get data for your research, Macro-Variable Data:
free of cost. National Income, Saving
and Investment
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Which precautions should be taken care of while using data assembled
from various sources?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) What do you understand by the term Special Data Dissemination
Standard?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
Let us now turn our attention to data on macroeconomic variables like national
income, saving and investment.
21.3 NATIONAL INCOME AND RELATED

MACRO ECONOMIC AGGREGATES
21.3.1 System of National Accounts (SNA)
How do you assess the performance of an economy? You may be able to look
at the trend in the production of rice or wheat to be able to say something about
the performance of paddy or wheat crop. You can make a similar assessment
about the production of steel. If you want to say something on agricultural
production, where you find that some crops have done well and others have
not, you think of making an overall assessment of agricultural performance by
constructing an index of agricultural production to review the performance of
agricultural production. But we should like to go beyond levels of output or
production and look at performance in terms of incomes flowing from output in
the form of rent, wages, interest and profit to those participating in the creation
of output namely the factors of production – land, labour, capital and
entrepreneurship (Since 1993, the role of the Government in influencing the
economy, through various taxes and subsidies have also become an important
item of analysis). Alternatively, we would like to base our judgment of 15
Data Base of Indian performance on value addition made by the production system namely, value
Economy of output net of the (intermediate) costs incurred in creating the output. It is (i) N
this overall value addition computed for all sectors/activities of the economy,
that is referred to as the National Product, (ii) macro-aggregates related to it,
and (iii) trends in (i) and (ii), that can help you in analyzing the performance of
an economy.
As you know, National Income (NI) is the Net National Income (NNI). It is
also used to refer to the group of macroeconomic aggregates like Gross
National Income (GNI), Gross Domestic Product (GDP), Gross Value Added
(GVA) and Net Value Added (NVA). All these of course refer to the total
value (in the sense mentioned above) of the goods and services produced
during a period of time, the only differences between these aggregates being
depreciation and/or net factor income from abroad. There are other
macroeconomic aggregates related to these that are of importance in relation to
an economy. What data would you, as an analyst, like to have about the health
of an economy? Besides a measure of the National Income every year or at
smaller intervals of time, you would like to know how fast it is growing over
time. What are the shares of the national income that flow to labour and other
factors of production? How much of the national income goes to current
consumption, how much to saving and how much to building up the capital
needed to facilitate future economic growth? What is the role of the different
sectors and economic activities – in the public and private sectors or in the
organized and unorganized activities or the households in the processes that
lead to economic growth? How does the level and pattern of economic growth
affect or benefit different sections of society? How much money remains in the
hands of the households for consumption and saving after they have paid their
taxes (Personal Disposable Income) – an important indicator of the economic
health of households? What is the contribution of different institutions to
saving? How is capital formation financed? Such a list of requirements of data
for analyzing trends in the magnitude and quality of, and also the prospects of,
efforts for economic expansion being mounted by a nation can be very long.
Such data, that is, estimates of national income and related macroeconomic
aggregates from part of a system of National Accounts that gives a
comprehensive view of the internal and external transactions of an economy
over a period, say, a financial year and the interrelationships among the
macroeconomic aggregates. National Accounts thus constitute an important
tool of analysis for judging the performance of an economy vis-à-vis the aims
of economic and development policy.
The United Nations (UN) has been recommending guidelines in the form of a
System of National Accounts (SNA) in order to promote international
standards for the compilation of national accounts as an analytical tool and
international reporting of comparable national accounting data. The first
version of SNA was recommended in 1953. This was revised in 1968 and then
in 1993. The fourth version (2008 SNA) was prepared under the auspices of
the Inter-Secretariat Working Group on National Accounts consisting of the
Commission of the European Communities (Eurostat), International Monetary
Fund (IMF), Organisation for Economic Cooperation and Development
(OECD), the UN and the World Bank and adopted by the UN Statistical
Commission in 2008. This system has been harmonized with other major
statistical systems like the Balance of Payments Statistics and Government
Finance Statistics of the IMF. The 2008 SNA contains a coherent, consistent
and integrated set of macroeconomic accounts based on a set of internationally
16 agreed concepts, definitions, classifications and accounting rules. It provides a
comprehensive accounting framework within which data can be compiled and Macro-Variable Data:
presented in a format that is designed to facilitate economic analysis, formulation National Income, Saving
and Investment
of policy and decision-making. You can read the entire SNA by accessing it at
http://unstats.un.org/unsd/nationalaccount/docs/SNA2008.pdf.
The informal sector or the unorganized sector occupies an important place in

our economy, in terms of contribution to national income, employment and
exports. It is necessary, therefore, to bestow adequate attention to unorganized
sector in the compilation of National Accounts and in other data domains and
in international efforts at promoting collection of data on informal sector
activities. The 15th International Conference of Labour Statisticians (ICLS)
(January, 1993) adopted a resolution on informal sector statistics with a view to
helping member countries of the International Labour Organisation (ILO) in
reporting comparable statistics of employment in the informal sector. The
resolution was endorsed by the UN Statistical Commission and by 1993 SNA.
National Accounts are compiled in India using a mix of 1968 SNA and 1993
SNA. India is also in the process of moving towards the full implementation of
SNA methodology to the extent feasible.
21.3.2 Estimates of National Income and Related

Macroeconomic Aggregates
a) Estimates Released by CSO
The Central Statistics Office (CSO) in the Ministry of Statistics and Programme
Implementation (MoSPI), Government of India compile and publish National
Accounts, which include estimates of National Income and related
macroeconomic aggregates like NNP, GNP, GDP and NDP, PFCE (private
final consumption expenditure), saving, capital formation and so on for the
country and for the public sector for every financial year. Quarterly estimates
of GDP are also prepared and released. Estimates prepared for any year at the
prices prevailing in that year are called estimates at current prices. Estimates
of national income and other aggregates at current prices is generally not used
in analyzing changes in the magnitude of these aggregates over time, as price
levels over time gets affected by inflation or deflation of an economy. Suppose
we need to compare the performance of the economy in terms of national
income or other macroeconomic aggregates, over a period of time, say, five
years. A comparison of estimates of, say, national income at current prices in
the opening year and the final year of the five-year period will give us the
increase or decrease, as the case may be, in national income at current prices.
We can easily note that the quantum of change observed in national income is
the composite measure of the change in national income and the changes in the
prices of goods and services between the two points of time. How then to get at
the actual change, or the change in real terms in national income (or any other
macroeconomic aggregate) over the period? This is done by removing the
effect of changes in prices while comparing the aggregate in question at
different points of time. Estimates of the aggregate for different years are,
therefore, prepared at the estimates of the aggregate at constant (base year)
prices. The comparison of estimates of the aggregate at constant (base year)
prices at different points of time is comparison of the magnitude of the
aggregate in real terms and measures the real change over time in the
magnitude of the aggregate. The base year now in use is 2011-12.
17
Data Base of Indian CSO changes the base year from time to time in order to take into account the
Economy structural changes that take place in the economy and depict a true picture of N
the economy. It was 1948-49, initially, when estimates of National Income
were first prepared by CSO and published in 1956. It was shifted to 1960-61 in
August, 1967, to 1970-71 in January, 1978, to 1980-81 in February,1988, to
1993-94 in February,1999 to 1999-2000 in January, 2006; to 2004-05 in
January 2010and to 2011-12 in January 2015. Note that, in the beginning, the
base years were the years in which the decennial population census was
conducted, that is, the first year of every decade from 1961 to 1981. The
choice of the subsequent base years, namely, 1993-94, 1999-00 and 2004-05
were a departure from this practice. What are the reasons for this? Estimates
of workforce participation rates (WPR), that is, the proportion of workers to
population for the benchmark years play an important role in the compilation
of estimates of national income. CSO had been making use of data on WPR
available from decennial population censuses till 1981 while compiling
national income estimates and for this reason the base year was the relevant
decennial census year. It was, however, observed that the Employment and
Unemployment Surveys of the National Sample Survey Organisation (NSSO)
captured WPR better than the population censuses. The base years chosen after
1981 were thus 1993-94, 1999-2000 and 2004-05, making use of data on
employment based on the NSSO 50th, 55th, 61stRound and 68th round Surveys
on Employment and Unemployment2 conducted during July ‘93 – June ’94,
July ’99 – June 2000, July ‘03 – June ‘04 and July 2011- June 2012 respectively.
After every revision of base year, the CSO publishes a set of results which
provide comparable estimates of the years prior to the base year, so that a long
time series data on national income aggregates can be analysed by the
researchers and policy makers.
Base year revisions differ from annual revisions in National Accounts

primarily because of nature of changes. In annual revisions, changes are made
only on the basis of updated data becoming available without making any
changes in the conceptual framework or using any new data source, to ensure
strict comparison over years. In case of base year revisions, apart from a shift
in the reference year for measuring the real growth, conceptual changes, as
recommended by the international guidelines, are incorporated. Further,
statistical changes like revisions in the methodology of compilation, adoption
of latest classification systems, and, inclusion of new and recent data sources
are also made. Changes are also made in the presentation of estimates to improve
ease of understanding for analysis and facilitate international comparability.
Estimates of national accounts aggregates for different years are published in

considerable detail in CSO’s Annual Publication “National Accounts Statistics
(NAS)”. The latest available NAS estimates are detailed in the publication
NAS 2014 and the Press Notes of MoSPI dated the 31st January, 20153. The
prerss note of January 2015 provides, for the first time, estimates of macro-
economic aggregates using the base year 2011-12. Note that in this release, for
the first time, Gross Domestic Product (GDP) “at factor cost” has been
discontinued. As is the practice internationally, industry-wise estimates have
been presented as Gross Value Added (GVA) at basic prices, while “GDP at
market prices has been referred to as GDP. Estimates of GVA at factor cost
2
Some employment data from Census 1991/2001 have also been used to supplement the data
from NSSO.
3
This publication and these Press Notes can also be accessed in the Ministry of Statistics
18 and Programme Implementation website www.mospi.nic.in
(earlier called “GDP at factor cost”) can be compiled by using the estimates of Macro-Variable Data:
GVA at basic prices and production taxes less subsidies. These have been National Income, Saving
and Investment
given in Statement 3.1 of this note. For the years 2011-12, 2012-13 and
2013-14, GVA at factor cost have been compiled and are presented in
Statements 10.1 & 10.2 of the press release.4Note that GVA at basic prices =
GVA at factor cost + net taxes on production and GDP at market prices = GVA
at basic prices + net taxes on products.
CSO releases, every January, “first revised estimates”, earlier called the “Quick
Estimates” of GDP, National Income, per capita National Income, Private Final
Consumption Expenditure, Saving and Capital Formation by broad economic
sectors5 for the financial year that ended in March of the preceding year. Quick
Estimates for any financial year are thus available with a time lag of ten
months. It also releases, along with Quick Estimates for any financial year,
revised estimates of national accounts aggregates for earlier financial years.
Further, CSO brings out Advance Estimates of GDP, GNI, NNI and per
capita NNI. The GDP, from the base year 2011-12, is published at market
prices, as per the latest international norms. Previously, it was published
at factor cost. Advance estimates of the GDPfor the current financial year is
published in February, two months before the close of the financial year.
These advance estimates are revised thereafter and the undated advance
estimates are released by the end of June, three months after the close of the
financial year. Meanwhile, by the end of the March, Quarterly Estimates of
GDP for the quarter ending December of the preceding year are also released.
Thus by the end of March every year, that is, by the end of every financial year,
advance estimates of national income for the financial year that just ended, first
revised estimates of national income for the preceding financial year and the
quarterly estimates of national income up to the preceding quarter, that is, up to
the quarter that ended in December of the financial year that just ended become
available.
CSO sets before itself an advance release calendar for the release of national
accounts statistics over a period of two years. 6 For instance, the Advanced
Release Calendar for the years 2015 indicates that the Quarterly Estimate of
GDP for any quarter during the period January, 2015 to end of September,
2015 would be available at the end of the next quarter. It also indicates the
dates on which Advance Estimates, Quick Estimates and Revised Advance
Estimates pertaining to the financial years ending on the 31st March, 2006, 30th
June, 2006, 29th September, 2006 and 29th December, 2006, respectively would be
released. Similarly, Revised Advance Estimates for the financial year 2004-05,
Quick Estimates for the financial year 2004-05, Advance Estimates for the
financial year 2005-06 and Revised Advance Estimates for the financial year
2005-06 would be released on the 30th June, 2005, the 31st January, 2006, the
7th February, 2006 and the 30th June, 2006, respectively. CSO has already (30th
January, 2015) released the following estimates, as proposed:
i) First Revised Estimates for 2013-14 at current and constant (2011-12) prices,
4
These 2 paragraphs have been taken from the Press release of the CSO dated 30th January
2015.
5
1. Agriculture, Forestry & Fishing. 2. Mining & Quarrying, 3. Manufacturing, 4. Electricity,
Gas & Water Supply, 5. Construction, 6. Trade, Hotels & Restaurants, 7. Transport, Storage
& Communication, 8. Financing, Insurance, Real Estate & Business Services, 9. Community,
Social & Personal Services.
6
This also takes care of SDDS referred to earlier. 19
Data Base of Indian ii) First revised Estimates of GDP, National Income, per capita National
Economy Income, Consumption Expenditure, Saving and Capital Formation for N
2013-14 (along with revised estimates of the earlier two years)at current
and constant (2011-12) prices,
iii) Advance Estimates of GDP, GNP and NNP and per capita NNP for
2014-15 at current and constant (2011-12) prices (on the 9th February, 2015).
The NAS 2014 presents estimates of GNP, NNP and the corresponding
domestic products GDP andNDP at factor cost and market prices at current and
constant (2004-05) prices7. Besides giving estimatesof the components of GDP,
estimates like Government Final Consumption Expenditure(GFCE), Private
Final Consumption Expenditure (PFCE) in the domestic market, Gross
Domestic Saving (GDS), Gross Domestic Capital Formation (GDCF), Exports,
Imports, net factor income from abroad and the share ofthe public sector in
GDP it presents8, estimates of the following aggregates at current and constant
prices(2011-12 prices) for 2013-14 and the period 2011-12 to 2013-14:
i) The contribution to GDP and NDP of different economic activities9.

ii) Quarterly estimates of GDP by broad economic sectors;
iii) GDP of sub-groups of each of the economic activities (see footnote 9) 1.1
and also separately for its livestock sub-sector10, 1.2,1.3,3.1,3.2,4, the sub-
sectors of 4 namely, electricity, gas and water supply.
iv) NDP for the sub-group 10, namely, Public Administration & Defence, by
(a) Central Government and Union Territories; (b) individual State
Governments; (c) local authorities; and (d) quasi government bodies.
v) Factor incomes (compensation for employees and operating surplus/mixed
income) of NDP for each economic activity (at current prices).
vi) Factor incomes of NDP of the unorganized segment of each economic
activity (at current prices).
vii) Property incomes (rent and interest) and Financial Intermediation Services
Indirectly Measured (FISIM) in the organized and unorganized segments
of each economic activity (at current prices).
viii) Net National Disposable Income, Private Income, Personal Income and
Personal Disposable Income.
7
As mentioned earlier, the shift in the base for constant prices to 1999-00 was made in
January, 2006, that is, after the publication of NAS 2005 in May, 2005. NAS 2006, due in
June, 2006, would be presenting estimates at constant prices at 1999-00. Besides the press
note of MoSPI referred to in the text, the Economic Survey 2006 of the Ministry of
Finance presented by the Finance Minister to Parliament on the 27th February, 2006 also
contains estimates released by CSO at current and constant (1999-00) prices for the years
2005-06 (Advanced Estimates), 2004-05 (Quick Estimates), 2003-04 and 2002-03.
8
Estimates of some of these macro aggregates are also published in the Monthly Bulletin of
RBI.
9
1. Agriculture and Allied Activities; 1.1. Agriculture; 1.2. Forestry& Logging; and 1.3.
Fishing; 2. Mining & Quarrying; 3. Manufacturing; 4. Electricity, Gas, Water Supply and
other utility services; 5. Construction; 6. Trade, Repair, Hotels & Restaurants; 7. Transport,
Storage, Communication and services related to broadcasting; 8. Financial services, 9. Real
Estate Ownership of dwelling &Professional Services; 10. Public administration and defence
and 11 Other Services;
10
The livestock sub-sector is a sub-sector of the sub-group ‘agriculture’ (1.1 in the preceding
20 footnote) and includes animal husbandry, (fuel & manure), silk worm cocoons and honey.
ix) GDP and NDP of the Public Sector by type of public sector institutions11 Macro-Variable Data:
and economic activity. National Income, Saving
and Investment
x) Factor incomes of Public Sector NDP for each economic activity by type
of institutions (current prices).
xi) Property incomes and FISIM in each economics activity of Public Sector
by type of institutions.
xii) Value of output, inputs and consumption of fixed capital (CFC) for a
number of economic activities.
xiii) Financial Assets and Liabilities of the Household sector.
xiv) Economic and Purpose Classification of Current and Capital Expenditure
of Administrative Departments (current prices).
xv) Production Accounts of (a) Railways, (b) Communication, (c) Departmental
Enterprises other than (a) and (b), and (d) Producers of Government
Services.
xvi) Detailed external transactions accounts.
xvii) Depreciation as provided in the books of accounts of public sector
institutional groups12 and private corporate sector; and
xviii) Time series of GNI, GDP NNP, NDP, Per capita NNI and other
macroeconomic aggregates like CFC, net factor income from abroad,
PFCE in the domestic market, GFCE exports, imports and mid-year
population from 1950-51 to 2013-1413.
Estimates of the aggregates referred to at (i), (vii), (xiii), (xvi) and PFCE and
its distribution by object and by type of goods, GFCE for the period 2004-05 to
2012-13 at current and constant (2004-05)Prices and the time series estimates
referred to at (xvii) above at current and constant (2004-05) prices have been
released by CSO in NAS 2014 by the end of May 2014 and posted on the
MoSPI website.
What about separate estimates of National Income aggregates for rural and
urban areas? CSO publishes estimates of NDP for rural and urban areas for
each economic activity (see footnote 8) at current prices for each of the base
year, namely, 1970-71, 1980-81, 1993-94, 1999-2000 and 2004-05.
b) Other Publications Giving CSO’s Estimates

The Handbook of Statistics on the Indian Economy14 – 2013-14, published
by the Reserve Bank of India (RBI) also publishes time series from 1952-53
of macro economic aggregates (that would otherwise have to be compiled from
different issues of the NAS document) like GDP at factor cost, CFC, NDP at
factor cost, indirect taxes less subsidies, NDP at market prices, net factor
income from abroad, GNP at factor cost, NNP at factor cost, GNP at market
price, NNP at market price, personal disposable income, GDP and NDP of
11
Administrative Departments, Departmental Enterprises, Non-departmental Enterprises and
Quasi-Government Bodies.
12
Administrative Departments, Public Sector departmental Enterprises subdivided into (i)
Railways, (ii) Communication; and (iii) others and Public Sector Non-Departmental
Enterprises subdivided into (i) financial companies; and (ii) non-financial companies.
13
The Pre-budget Economic survey also presents the time series on GNP, GDP, NNP and per
capita NNP and mid year population.
14
Published annually. This can downloaded from the RBI WEBSITE http://rbidocs.rbi.org.in/
rdocs/Publications/PDFs/000HSE13120914FL.pdf 21
Data Base of Indian public sector (from 1960-61), gross and net domestic capital formation, per
Economy capita GNP and NNP, at one place. The Economic and Political Weekly N
research Foundation (EPWRF) also presents comprehensive national income
statistics in a time series format from time to time along with details of
concepts, methodology and data sources for ready use by researchers. It
provides data from 1950-51 to 2011-12 with 2004-05 as the base year
(http://www.epwrfits.in/TimeSeriesDataResearch.aspx)” is the latest. This
can be obtained from EPWRF either as a subscriber or through a pay per
module mode. National Accounts Statistics of most countries and areas of the
world can be accessed at the UNSD website http://unstats.un.org/
unsd/nationalaccount/madt.asp. At the time of writing the module, the latest
year for which this data is available is 2013.
c) Limitations of the Estimates

The concepts and methodology used and the data sources utilized for making
these estimates are set out in two publications of the CSO, namely, “National
accounts statistics : Sources and Methods” (available for the base 2004-05
series, published by the CSO in 2012) and “New Series on National
Accounts (Base 2004-05)” (CSO, 2010). The MoSPI Press Note dated the
31st January, 2006 releasing Quick Estimates of National Income, etc., for
2004-05 indicates briefly the changes in the new series of National Accounts
Statistics with base year 2004-05. The publication NAS- Manual of estimating
state and District Income, 2008 provide the methodology for making estimates
of GSDP and GDDP. Besides, the publication “National Accounts Statistics”
brought out every year has a chapter titled “Notes on Methodology and
Revision in the Estimates”. This chapter in the NAS also contains tables
explaining, sector-wise, reasons for the revision in GDP growth rate as per the
different types of estimates (like revised advance, quick, etc., estimates)
released in the preceding year. For instance, NAS 2014 gives the reasons for
revision in GDP growth rate during.
i) 2012-13 between the provisional Estimates released in May, 2013 and first
revised estimates released in January, 2014;
ii) 2011-12 between First Revised Estimates released in January, 2013 and
Estimates released in January, 2014; and
iii) 2010-11 between Estimates released in January, 2013 and Estimates
released in January, 2014.
Limitations of estimates of national income aggregates arise from insufficiency

of data or the choice of data in capturing adequately income flows. Estimates
of GDP/NDP at factor cost by sectors can be classified into two broad
categories from the point of view of differences in the data base – direct
estimates and indirect estimates. Direct estimates are based on statistics
available annually on a regular basis so that these reflect year-to-year variations
in the volume of the economic activities concerned. The translation of such
annual statistics into National Accounts aggregates requires the use of certain
norms and ratios or other assumptions. Resort to readily available indications
of the economic activity in question has often to be made when and if there are
delays on the part of the data-generating agencies in supplying the required
regular annual data. As a result, revisions made in the estimates of national
accounts aggregates when regular annual data become available later lead to
major changes in the provisional estimates released earlier. Direct estimates
mostly cover the institutional groups, (i) the public sector (and the Government
22
component within it), and (ii) the private corporate sector, or what is usually Macro-Variable Data:
referred to as the ‘organised’ segment of the economy. On the other hand, National Income, Saving
and Investment
indirect estimates require to be made when regular annual statistics are not
available in respect of any economic activity. How are these made? Estimates
based on periodic benchmark surveys are first derived for the survey year.
These are then extrapolated forward or backward, as required, on the basis of
physical indicators of the economic activities concerned. The degree of
approximation involved in this procedure depends critically on the sensitivity
of the indicator in reflecting year-to- year variations in the volume of the
economic activity concerned. Indirect estimates are used in respect of the
institutional groups, (a) households, and (b) non-profit institutions serving
households (NPISH) and the “unorganized” segment of the economy. The
share of direct estimates in aggregate GDP has increased from 57.6 per cent in
the series with 1970-71 as the base year to 63.7 per cent in the series with
1980-81 as the base year and to 89.6 per cent in the series with 1993-94 as the
base year. The share of direct estimates in different sectors in the GDP series
with 2004-05 as the base year varied from 100 per cent in Mining, Registered
Manufacturing, Electricity, Gas & Water Services, Railways and Public
Administration &Defence sector to about 21 per cent in trade sector15. As such,
the share of organized sector in 2012-13 had become only 44.7 per cent of the
economy (with base year 2004-05, page 1i, NAS 2014).
The intention of this sub-section is not to comment adversely on the admittedly

complex task of compiling national accounts statistics but to draw attention to
limitations that need to be kept in mind by users as economists/statisticians/
analysts utilizing these data while carrying out their research, analytical and
advisory work. Improvements in data and methodology for estimating national
accounts aggregates constitute a continuous process, calling for efforts across
sectors, in which Ministries in the central Government, regulatory agencies, the
CSO, the State Directorates of Economics & Statistics and other research
institutions participate. The interesting reader may like to familiarize
himself/herself with the work being done by the Indian Association for
Research in National Income and Wealth, New Delhi. Their Bulletin would
show the efforts underway to bring about further improvements in national
accounts methodologies and statistics. The recommendations of the National
Statistical Commission (2001) will also pave the way to further enhancement
of the quality of these estimates.
21.3.3 The Input-Output Table

Any economic activity is dependent on inputs from other economic activities
for generating its output and the output from this economic activity serves as
inputs for producing the output from other activities. Data relating to the
interrelationship among different sectors of the economy and among different
economic activities are thus important for analyzing the behavior of the
economy and, therefore, for formulation of development plans and setting
targets of macro variables like output, investment and employment. Such an
input-output table will also be useful for analyzing the impact of changes in a
sector of the economy or economic activity on other sectors of the economy
and indeed the entire economy. An Input-Output Transactions Table
(IOTT) is prepared by CSO from time to time. The latest to be published is the
one relating to 2007-0816. It gives, besides the complete table, the methodology
15
This paragraph is based on Section 13.2 Chapter 13 of the Report of the National Statistical
Commission.
16
This can be accessed at http://mospi.nic.in/Mospi_New/upload/iott-07-08_6nov12.htm 23
Data Base of Indian adopted, the database made use of, analysis of the results and the
Economy supplementary tables derived from the IOTT giving the input structure and the N
commodity composition of the output. The Planning Commission updates and
recalibrates the IOTT and prepares an Input-Output table for the base period of
a Five Year Plan being formulated and another Input-Output table for the
terminal year of the Five Year Plan. The detailed results of this exercise,
carried out in the Planning Commission for the formulation of any Five Year
Plan, are published by the Planning Commission as the Technical Note to the
Five Year Plan. The Technical Note contains the relevant Input-Output Table,
the methodology adopted and related material. The latest available in this series
is the Technical Note tothe Tenth Five Year Plan. The IOTT of CSO and the
Input-Output Table of the Planning Commission would be useful to researchers
interested in exploiting the power of the input-output technique in economic
and econometric analysis in their research work.
21.3.4 Regional Accounts – Estimates of State Income and

Related Aggregates
a) Estimates of States Domestic Product (SDP) Prepared and Released
by State Governments and Union Territory Administrations
State Accounts Statistics (SAS) consist of various accounts showing the flows
of all transactions between the economic agents constituting the State economy
and their stocks. The most important aggregate of SAS is the State Domestic
Product (SDP) (State Income). Estimates of GSDP and NSDP at constant and
current prices are being prepared and published by the Directorates of
Economics and Statistics (DES) of all State Governments and Union
Territory Administrations except the Union Territory Administrations of
Dadra & Nagar Haveli, Daman & Diu and Lakshadweep.
b) Other Publications Giving Estimates of SDP Made by State/UT DES

The State Governments and Union Territory Administrations send their
estimates of SDP to the CSO. These estimates at current and constant prices are
available in the CSO website. The pre-Budget Economic Survey presented to
Parliament by the Finance Minister every year contains these estimates (as a
time series from 1993-94) of Net State Domestic Product (NSDP) at current
prices. The RBI Handbook referred to above provides more detailed data
drawn from the CSO website. It presents:
i) estimates of NSDP at current and constant prices – one set with the
constant prices base 1980-81 for the period 1980-81 to 1998-99 and the
other with the constant prices base 1993-94 for 1993-94 to 2003-04; and
ii) estimates of NSDP by economic activity and per capita NSDP at current
and constant (1993-94) prices for the period 1993-94 to 2003-04; (new
series).
Estimates of State Income (gross State Domestic Product) made by States and
Union Territories by economic activity at current as well as constant prices are
readily available in a (short) time series format at one place in the publication
of the CMIE (EIS), Mumbai referred to above. These estimates at current and
constant prices by sectors for the period 1960-61 to 2000-01 have also been
brought together at one place for the use of research scholars by the EPWRF in
their publication Domestic Product of States of India: 1960-61 to 2000-01
24
(EPWRF, June, 2003). These are also available on the website of the Macro-Variable Data:
EPWRF and in user-friendly and interactive CD ROMS.17 National Income, Saving
and Investment
c) Limitations of Estimates of SDP

The preparation of estimates of SDP call for much more detailed data than for
the preparation of national level estimates, especially on flows of goods and
services and incomes across geographical boundaries of States/ Union
Territories. Conceptually, estimates of SDP can be prepared by adopting two
approaches. These are the income originating approach and the income
accruing approach. In the former case, the measurement relates to the income
originating to the factors of production physically located within the area of a
state. In other words it is the net value of goods and services produced within a
state. In the latter case, the measurement relates to the income accruing to the
factors of production physically located within the area of a State. Here, the
measurement relates to the income accuring to the normal residents of a State.
The income accruing approach provides better measure of the welfare of the
residents of the State. Preparation of estimates of SDP using the income
accruing approach is not possible because data on inter-State flow of goods and
services are not available. Compilation of other aggregates of State accounts is
also problematic because the data required for the purpose, especially on inter-
State flows of incomes, are not available.18 Thus only the income originating
approach is used in preparing estimates of SDP. This has to be kept in mind
while using estimates of SDP.
Efforts have been made by the CSO over the years to bring about a good
degree of uniformity across States and Union Territories in SDP concepts and
methodology. The major milestones in these efforts are:19
i) the recommendations of the First report of the Committee on Regional

Accounts (CSO, 1974) set up by the Government of India and the Final
Report of the Committee (CSO, 1976) in the form of a set of Standard
Tables to be prepared and a system of regional accounts, besides
suggestions regarding steps to fill gaps in data vis a vis those needed for
compilation of SDP, for the guidance of DES of States/Union Territories;
ii) the article “Mahabaleshwar Accounts of States” in the October, 1976
issue of the Journal of Income & Wealth;
iii) the article “The Status of State Income Estimates” appearing in the
October, 1980 issue of the Monthly Abstract of Statistics (CSO); and
iv) the article “Comparable Estimates of SDP – 1970-71 to 1975-76” appearing
NAS, January, 1979 of CSO, which described the methodology of
preparing estimates of SDP.
v) the publication “National Accounts Statistics: Manual of Estimating
State and District Income, 2008” published by the CSO in 2008.
Are the estimates of SDP of different States/Union Territories comparable?
Probably not. The successive Finance Commissions have found it necessary to
get comparable estimates of NSDP and per capita NSDP made for their use for
periods relevant to their work. These comparable estimates of SDP are
published in the Reports of the successive Finance Commissions from the
17
E-mail id for queries to EPWRF: epwrf@vsnl.com .
18
The Report of the National Statistical Commission, Chapter 13 section 13.7.
19
ibid. See also the Preface to EPWRF publication on SDP in the preceding sub-section (b) on
other publications giving SDP estimates. 25
Data Base of Indian sixties. The EPWRF publication on SDP referred to earlier also provides
Economy comparable estimates of SDP and evaluates these with reference to those made N
for the periods 1960-61 to 1964-65, 1976-77 to 1978-79, 1982-83 to 1984-85,
1987-88 to 1989-90 and 1994-95 to 1996-97 for the use of the Finance
Commissions. The question of comparability of estimates of SDP of different
States/Union Territories is important from the point of view of their use in
econometric work involving Inter-State or regional comparisons or the
contribution of the regions or the States to the national product.
21.3.5 Regional Accounts – Estimates of Districts Income

The need for preparing estimates of district income has become urgent in the
context of decentralization of governance and the importance of, and the
emphasis on, decentralized planning or planning from below. Of late, it has
gained a special importance, as this is one of the three indicators to compile the
district-wise Human Development Index, prepared by some of the States. The
State Governments have realized this need and estimates of District Domestic
Product (DDP) are being prepared by a number of State DES and published by
them in State Statistical Handbooks/Abstracts/Economic Surveys and are
also posted on their websites. The position in different States and Union
Territories (zone-wise) is as follows:
Type of estimate of DDP State/ UT Base

year
1. for all years till 2013-14 Rajasthan 2004-05
2. for all years till 2012- Andhra Pradesh, Arunachal Pradesh, 2004-05
13 Assam, Madhya Pradesh, Maharashtra,
Meghalaya, West Bengal
3. for all years till 2011- Bihar, Haryana, Himachal Pradesh, 2004-05
12 Karnataka, Kerala, Chhattisgarh, Punjab,
Tamil Nadu, Telangana, Uttar Pradesh
4. for all years till 2010-11 Nagaland, Odisha, Uttarakhand 2004-05
5. Up to 2005-06/ 2006- Jharkhand, Mizoram, Andaman &Nicobar 1999 –
07/ 2007-08 Islands/ Manipur / Sikkim 2000
6. Not available Goa, Gujarat, Jammu & Kashmir, Tripura,
Delhi, Puducherry
7. Not required Chandigarh, as the UT has only 1 district
For some of the States in 6, it might be available with the State Directorate of
Economics and Statistics, but not available centrally at one location.
The methodology for preparing estimates of district income at present is based
on20
i) The recommendations of the Report21 of the Technical Group set up by
the Department of Statistics for recommending a suitable methodology
for preparing estimates of district income;
ii) The methodology developed in August, 1996 jointly by the DES of Uttar
Pradesh and Karnataka – a task entrusted to these organization by CSO in
1995.
20
See Methodology of estimating State and District Domestic Product, CSO, 2008.
21
26 Submitted in January, 1987.
The income originating approach is adopted for compiling district income Macro-Variable Data:
estimates in the light of the kind of data that is available. As mentioned earlier, National Income, Saving
and Investment
the income accrual approach can measure the welfare of the normal residents
of the district in a more reliable manner. However, as flow of funds from
district to district is not available and difficult to estimate, the income
originating approach is followed. It is necessary to make adjustments, in the
estimates based on the income originating approach, for flow of incomes
across territories of districts that are rich in resources like minerals or forest
resources and where there is a daily flow of commuters from other districts.
There are a lot of problems in the context of data availability separately for
each district. For most of the commodity producing sectors, like agriculture,
mining and registered manufacturing, data is fairly available. But, in remaining
sectors, it is very scanty. Even within the agricultural sector, even if one can
get the district-wise production, prices at each district sometimes may not be
available. In case of entire unorganized sector, where national and State level
estimates are based on periodic surveys, reliable district level estimates from
these data cannot be prepared due to limitations in sample size.
21.3.6 National Income and Levels of Living

The foregoing paragraphs have talked about the availability of estimates of
macroeconomic aggregates at the national and state levels, in different
economic activities, in segments of economic and geographic spaces like
public and private sectors, and households and rural and urban areas. What do
trends in the magnitude of these variables say about the welfare of different
sections of society? Precious little, perhaps, especially when these are
considered without information on the manner in which the macroeconomic
aggregates are distributed among these sections of society. Per-capita national
income or even per-capita personal disposable income can only indicate overall
(national) averages. Distribution of population by levels of income can be a big
step forward in understanding how well the performance in the growth of GDP
has translated into or has not translated into, improvements in levels of living
for sections of society below levels considered the minimum desirable level. It
would also help us analyse whether inequalities in levels of living have
worsened or abated. What about the level of unemployment or levels and
quality of employment? What about the health status of people? Or, to consider
all these indicators of the state of a society and different sections of society,
what are the levels of human development and gender discrimination? Such
lines of analysis and the data required for the purpose are important from the
point of view of planning for a strategy of growth with equity.
The Quinquennial Consumer Expenditure Surveys of the NSSO provides

such data. These provide the distribution of households by monthly per capita
consumption expenditure (MPCE) classes. You have already looked at such
data and their implications for planning and policy in Unit 4 on “Poverty and
Inequality – Policy Implications” in Block 1 on “Framework of the Indian
Economy” in Course MEC-105 on “Indian Economic Policy” in your first year
course. As you have seen there, the latest available data on this relate to
2011-12, flowing from the 68th round NSSO survey of 2011-12. The next
such comprehensive survey (74thRound of NSS)is planned in 2016-17. You
have also looked at the data on recent trends on growth rate of employment and
the growth rate of the economy (employment elasticity) in Unit 5 on
“Employment and Unemployment: Policy Implications” and comprehensive
indicators like Human Development Index and the Gender Development Index 27
Data Base of Indian in Units 3 and 4 in Block 1 of Course MEC-005. You would have noted that
Economy the Human Development Reports prepared by the Planning Commission N
(available on its website) and several State Governments contain detailed
data on these questions.

1) List the various components of the system of National Accounts.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Which have been the base years for compilation of national accounts
statistics (NAS)? What is the difference between base year revision and
revision of estimates in annual series of the NAS?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) In addition to CSO, which other agencies bring out the data on macro
economic aggregates?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) By whom the estimates of state income and related aggregates are made?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
28
21.4 SAVING National Income, Saving
and Investment
As you are aware, broadly speaking, GNP is made up of consumption, saving
exports net of imports, besides net factor income from abroad. Saving is
important in as much as it goes to finance investment, which in turn brings
about growth of GNP. What is the volume of Saving relative of GNP? How
much of it is consumed by the needs of depreciation? Who all contribute, and
how much, to the total volume of Saving? Let us see what kind of data is
available on such questions.
Estimates of Gross Domestic Saving (GDS) and Net Domestic Saving (NDS)
in current prices and the Rate of Saving are made by CSO and published in the
National Accounts Statistics (NAS) every year and the Press Note of
January of every year releasing First Revised Estimates. These are first
made for any year along with First Revised Estimates of GDP, etc., and revised
and finalized along with its subsequent revisions. The structure of Savings, that
is, the distribution of GDS and NDS by type of institution – household sector,
private corporate sector and public sector are also available in this publication.
Also presented in this document is the distribution of:
i) GDS, NDS and consumption of fixed capital (depreciation) by public

sector, private corporate sector and household sector;
ii) Public sector GDS, consumption of public sector fixed capital and public
sector NDS by type of institutions22;
iii) Public sector GDS – by public authorities [sub-divided further into
government administration and departmental (commercial) enterprises],
non-departmental enterprises (broken up further into government
companies and statutory corporations, the latter including port trusts
also);
iv) Private sector GDS, consumption of fixed capital and NDS by private
corporate sector and household sector;
v) Private corporate sector GDS by joint stock companies (distributed
further into financial and non-financial companies) and cooperative banks
and societies;
vi) GDS of the household sector into (net) financial saving and saving in
physical assets and that of (net) financial saving, that is, currency, net
deposits, shares and debentures, net claim on Government, life insurance
funds and provident and pension funds;
vii) Financial assets [in greater detail than in (vi) above] and liabilities of the
household sector leading to net financial saving of the household sector;
and
viii) Item (xiii) in the list in sub-section (a) of Section 20.3.2 – these indicate
estimates of net saving in these entities.
NAS 2014 presents the above estimates for 2012-13 and for the period 2004-05
to 2012-13. The Press Note dated the 31st January, 2014 of MoSPI releasing
Quick Estimates provides estimates of GDS, NDS, and all the estimates
referred to at (i) to (vii) above for 2012-13. Time series of estimates of GDS
and NDS in current prices are, however, available from 1950-51 onwards. The
22
See footnote 12. 29
Data Base of Indian time series on the structure of GDS by institutions (public, private corporate
Economy and household sectors) are presented in the Pre-Budget Economic Survey N
(2014) of the Ministry of Finance. The Handbook of Statistics on the Indian
Economy of RBI presents time series data from 1952-53 on GDS and NDS
and from 1970-71 onwards on (i) their components by institutions, and (ii)
financial assets and liabilities of the household sector, item-wise, at current
prices. The EPWRF publication on NAS (fifth edition) gives time series data
from 1950-51 on all these variables and on Domestic Saving in public sector by
type of institutions. The CMIE publication referred to earlier also contains time
series data on GDS and NDS, though for a shorter period of time.
Estimates of Gross and Net Domestic Saving at the State and Union Territory
levels are not being made at present by DESs of State/Union Territories.
The CSO publishes “Sources and Methods”, an additional publication

showing the detailed methodology followed and data sources used for
preparation of the National Income aggregates in the base year and the
methodology to be followed since the change in base year. It may be added that
estimates of Savings suffer from a number of limitations due largely to
deficiency of data. This is particularly so in respect of estimates of various
financial instruments from the private corporate sector and the household
sector. Saving being the excess of income over expenditure, the major gaps in
data relate to the household sector, non-profit institutions serving households
(NPISH) and local bodies. Quality of data relating to the private corporate
sector poses an additional problem. The estimate of Savings of the household
sector in physical assets is thus derived as a residual, in the absence of direct
data.23
21.5 INVESTMENT
Investment is Capital Formation (CF). Investment of money in the shares of a
company or purchase of land is not investment but buying a house or
machinery is investment. In other words, investment is creation of physical
assets like machinery, equipment, building, inventories and so on, called
“produced assets”, adding to the capital stock (of such assets) in the economy,
enhances the productive capacity of the economy. Investment or CF is another
important component of GNP and the rate of investment – expressed as a
proportion of GNP – largely determines the rate of growth of the economy.
How is capital formation financed by the economy? What is the contribution of
different sectors to capital formation or, how much is used up by different
sectors? What is the capital stock available in the economy? These are all the
questions that rise in one’s mind when considering strategies for economic
growth. What kind of data is available?
The annual NAS publication of the CSO presents estimates of Gross

Domestic Capital Formation (GDCF), Gross Domestic Fixed Capital
Formation (GDFCF), Change in Stocks, Consumption of Fixed Capital (CFC),
Net Domestic Capital Formation (NDCF) and Net Domestic Fixed Capital
Formation (NDFCF) in current prices and at constant prices. These estimates
are made along with First Revised Estimates of National Income in January
every year and the process of revision of these estimates proceeds along with
that of the estimates of national income aggregates. The estimates of capital
formation are presented in the NAS through two alternate approaches, namely,
23
30 Report of the National Statistical Commission, Chapter 13, Section 13.6
through the flow of funds approach and through the commodity flow approach. Macro-Variable Data:
The flow of funds approach is gross domestic savings plus net capital inflows National Income, Saving
and Investment
from abroad, while the commodity flow approach, derived by the types of
assets. The successive issues of NAS and the Press Note of MoSPI of
January every year also present estimates of the distribution of the following
aggregates at current and constant prices:
i) GDCF, GDFCF, Change in Stocks, CFC, NDCF and NDFCF by type of

institutions (public sector, private corporate sector and household sector);
ii) GDFCF in the public sector, private corporate sector and household
sector by type of assets (construction and machinery & equipment);
iii) GDCF, GDFCF, Change in Stocks, CFC, NDFCF and NDCF by
economic activity24;
iv) Net Capital Stock (NCS), Net Fixed Capital Stock (NFCS) and Inventory
as on the 31st march of each calendar year by type of institutions
(Government administrative departments, Departmental enterprises, Non-
departmental enterprises, joint stock companies of the private corporate
sector, cooperative and household sector);
v) NCS, NFCS and Inventory as on the 31st March of each calendar year by
economic activity;
vi) Consolidated Account of the Nation – How Capital Formation is financed
(current prices);
vii) Consolidated Account of the Nation – External Transactions showing
current and capital transactions (in current prices);
viii) GDCF, GDFCF and Change in Stock in the public sector by economic
activity and type of institutions;
ix) NDCF, NDFCF and Change in Stock in the public sector by economic
activity and type of institutions;
x) GDFCF in each type of public sector institutions by type of assets (at
current prices);
xi) (v) above for the public sector;
xii) Financing of Capital Formation in (a) Railways, (b) Public Sector
Communication and (c) Administrative Departments including Departmental
Enterprises other than (a) and (b), (d) Non-Departmental Financial and
Non-Financial Enterprises (in current prices); and
xiii) Time series of GDCF and GFCF from 1951 to latest financial year.
NAS 2014 presents estimates of aggregates listed above at current and constant
(2004-05) prices for the period 2004-05 to 2012-13. The MoSPI Press Note of
the 31st January, 2015 provide estimates of GCF, GFCF and consumption of
fixed capital in current and constant (2011-12) prices as also of the distribution
of these by industry of use and by institutional sectors of the economy, as
defined in the SNA, namely, Public non-financial corporations, private non-
financial corporations, public financial corporations, private financial corporations,
general Government and Households for the period 2011-12 to 2013-14.
24
See footnote 10. 31
Data Base of Indian Other publications providing time series of CSO data on capital formation are:
Economy N
Sl. Publication Variable Time Series

No. From
1) Pre-Budget Economic #GDCF, GFCF and 1950-51
Survey (2006) Change in Stocks and their
structure by by institutions
2) RBI Handbook of *i) GDCF and NDCF 1950-51
Statistics on the Indian
Economy – 2005
*ii) GDCF, GFCF and Change 1970-71
in Stocks and their
structure by institutions.
#iii) Financial assets and 1970-71
liabilities of the household
sector (item-wise)
3) EPWRF publication *i) data of the type in items (i) 1950-51
on NAS (Fifth to (ix) and (xi) of NAS
Edition) listed in the preceding
paragraph of this section
are generally given.
*ii) Capital-output ratios by 1950-51
institutions;
*iii) Average NFCS to Output 1950-51
4) CMIE *GDCF and GFCF 1960-61
# at current prices *at current and constant (1993-94) prices
The publications/documents of CSO and other documents like the Report of

the National Statistical Commission referred to in the section on National
Income contain the methodology and the data utilized for preparing estimates
of Capital Formation and related aggregates also.
Estimates of GFCF at the State level are being prepared by 14 States, namely,
Punjab, Haryana, Himachal Pradesh, Madhya Pradesh, Orissa, Meghalaya,
Assam, Andhra Pradesh, Tamil Nadu, Kerala, Karnataka, Maharashtra, Gujarat
and Rajasthan. The status of these estimates are indicated in the EPWRF
publication on NAS, referred to in the sub-section 20.3.4.
Gaps in data for the estimation of capital formation exist in a number of

relevant areas25. Some of these are:
i) Non-availability of regular data on equipment and machinery (part of

GFCF) in domestic production in unregistered manufacturing;
ii) Lack of annual enterprise surveys and surveys on NPISHs for estimating
GFCF by institutional sectors of the economy;
iii) Complete accounts of local bodies are not available for some States/ UTs;
iv) Bank advances and other indirect approaches form the basis of estimation
of Changes in Stocks, with benchmark estimates coming from
quinquennial enterprise surveys; and
25
32 Report of the National Statistical Commission, Chapter 13, Section 13.6.
National Income, Saving
v) Inadequacies in (a) the representative character of the sample of and Investment
companies and (b) the blow-up factor used for estimation of saving and
investment of the private corporate sector.
1) What approaches are followed by CSO to compile the saving data in
National Accounts Statistics? Which estimates of investment are prepared
by the CSO?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Go through the methodology contained in NAS 2015 and the data utilized
for preparing estimates of capital formation and related aggregates.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) From where can you get the data on contribution of different sectors to
capital formation?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
21.6 LET US SUM UP

The Indian Statistical System with the Ministry of Statistics and Programme
Implementation (MoSPI) at its apex collects, compiles and disseminates an
enormous amount of data on diverse aspects of the Indian economy, to the
extent feasible, international standards prescribed by the United Nations and
other international bodies. MoSPI sets standards to be followed by statistical
agencies in the country and coordinates statistical activities carried out by these
agencies. It also developes standardized classification of economic activities,
namely, NIC, revising it from time to time keeping pace with the structural
changes taking place in the economy. Data published by various Government
agencies can be accessed through the portal www.data.gov.in . Other agencies
also bring together and publish comprehensive data on some or all sectors of
the economy, these publications being, the annual Economic Survey of the
Ministry of Finance, the Budget documents, the Five-Year Plan documents,
RBI’s Statistical Handbook on the Indian Economy, its monthly bulletin,
the website on the database of the Indian Economy, and the Human
Development Reports prepared by the Planning Commission and the State
Governments. Most of these data are available in electronic format – and/or on
33
Data Base of Indian the websites of the agencies concerned. One can usually download the
Economy published reports of the Government agencies from their respective websites N
free-of-charge. Various agencies also give anonymized data for analysis
purposes. The CMIE (EIU), Mumbai and the EPWRF, Mumbai also bring
together and publish comprehensive data on different sectors/aspects of the
economy regularly. The Statistical Handbooks of the State/Union Territory
DES and the District Census Handbooks published by the Census
Directorates/Registrar general of India provide comprehensive information
across sectors at the State and sub-State levels. Finally, the website of the
United Nations Statistics Division provide access to methodological
documents, data and report of various UN organisations which can be used for
cross-country comparisons. The search for any kind of data can thus very well
start from one of these sources and move on, if necessary, to the primary
source.
National Accounts Statistics (NAS), which consists of estimates of National

Income (NNI), GNI, GDP, NDP and related macroeconomic aggregates like
Consumption Expenditure, Personal Disposable Income and Factor Incomes
constitute an important tool for analyzing the performance of an economy.
These compiled by CSO, MoSPIand published every year. In compiling these,
CSO follows the guidelines given in 1993 SNA of the United Nations to the
extent feasible. These estimates are published in great detail in CSO’s annual
publication National Accounts Statistics (NAS). The latest issue of this
publication – NAS 2014 – is also available on the MoSPI website. In addition,
Advance Estimates of GNI, GDP, NNI and per capita income for a financial
year are released two months before the end of the financial year through a
Press Note and in the MoSPI website. First Revised Estimates of these
aggregates relating to a financial year and revised estimates of earlier years are
released by January of the next year, that is, within ten months after the close
of the financial year through a Press Note and on the MoSPI website. Quartly
Estimates of GDP are also prepared and released, generally within 45 days
from the close of a quarter.
NAS estimates are presented in current and constant (base year) prices. CSO
changes the base year from time to time to take note of structural changes in
the economy and to depict a true picture of the economy. The base year was
changed to 2004-05 in January, 2010 and to 2011-12 in January, 2015. Thus, as
of the 1st March, 2015, Advance Estimates of GDP, etc., for 2014-15, First
Revised Estimates of GDP, etc., for 2013-14, estimates of GDP for the quarter
ending December, 2014 and for the period April-December, 2014, all in current
and constant (2011-12) prices, have been released.
Estimates of Saving and Capital Formation form part of the National Accounts
Statistics. These estimates are presented in detail in the publication National
Accounts Statistics, published in May of each year. The press note of
January, 2015, has provided the estimates referred to above with a changed
base year 2011-12. The detailed estimates presented in NAS enable an analysis
of various aspects of Saving and Capital Formation like the structure of Saving
and Capital Formation, the manner in which Capital Formation is financed, the
role of the public sector in Capital Formation and trends in the rate of Saving
and Capital Formation and net capital flow from abroad.
Estimates of State Income are made and released by DES of all States and
Union Territories except Dadra & Nagar Haveli, Daman & Diu and
Lakshadweep. Estimates of GFCF are being made by 14 state DESs. District
34
Income estimates are made and released by DES of States. Estimates of State Macro-Variable Data:
Income are also available in MoSPI website, besides the Economics Survey, National Income, Saving
and Investment
RBI Statistical Handbook on the Indian Economy and the CMIE and
EPWRF (State Income) publications referred to earlier. The last publication
also give the State level GFCF estimates. The RBI Handbook, the Economic
Survey, the EPWRF publication on NAS and the CMIE publication referred
to earlier also present the national level estimates of GDP, etc., estimates of
Saving and Capital Formation made by CSO. The CMIE and the EPWRF
publications also contain almost as much detail as the publication NAS and in a
time series format.
We have also noted the limitations and inadequacies to which estimates of

national and state income and those of capital formation and saving are subject.
Improvement in methodology and efforts to fill gaps in data constitute a
continuous process.
21.7 EXERCISES
1) What type of data will you need to assess the performance of Indian
Economy? Explain the various sources of such data?
2) State the difference between “Gross” and “Net” in the context of domestic
and national income. How can one arrive at “Net National Income” from
“Gross Value Added at basic prices”?
3) What are the relationships depicted in Input-Output Table? How are these
tables useful in research work?
4) With the help of NAS 2014, explain the structure of saving in India.
5) What do you mean by the term capital formation? What is the capital stock
available in India?

Central Statistics Office (2012): National Income Statistics – Sources and
Methods, Ministry of Statistics and Programme Implementation, Government
of India, New Delhi.
_________(2008): Methods of estimating State and District Domestic Product.
_________(2014): National Accounts Statistics, 2014,
_________(2007): Input Output Transactions Table2007-08
_________(2015): Press Note New Series estimates of national income,
consumption expenditure, saving and capital formation (base year 2011-12)
International Labour Organisation (ILO) (1993) :Resolution Concerning
Statistics of Employment in the Informal Sector, 15thInternational Conference
of Labour Statisticians (ICLS), January, 1993. Geneva.
Indian Association for Research in National Income and Wealth &

Institute of Economic Growth (1998): Golden Jubilee Seminar on Data Base
of the Indian Economy, Delhi. A look at the Bulletins of the Association would
also be useful.
http://mospi.gov..in : website of the Ministry of Statistics and Programme
Implementation, Govt. of India.
http://rbi.org.in : website of the Reserve Bank of India. 35
Data Base of Indian http://unstats.un.org : website of the United Nation’s Statistics Division.
Economy N
Katyal, R.P., Sardana, M.G. and Satyanarayana, J. (2001): Estimates of
DDP, Discussion Paper 2, National Workshop on State HDRs and the
Estimation of District Income Poverty Under the State HDR Project Executed
by the Planning Commission (GOI) with UNDP Support held in Bangalore in
July, 2001, UNDP, New Delhi.
Ministry of Statistics & Programme Implementation (2001): Report of the

National Statistical Commission, Ministry of Statistics and Programme
Implementation, New Delhi.
System of National Accounts 1993 (SNA 1993) Commission of the European

Communities, International Monetary Fund, Organisation for Economic
Cooperation and Development United Nations, World Bank, Brussels/
Luxembourg, New York, Paris, Washington, D.C. www.mospi.nic.in :
Website of the Ministry of Statistics and Programme Implementation.
21.9 ANSWERS ON HINTS TO CHECK YOUR

PROGRESS EXERCISES
1) Coordination of statistical activities in the country maintenance of
statistical norms and standards, providing liaison with control, state and
international statistical agencies.
2) See Section 21.2
3) See Section 21.2
4) See Section 21.2
1) The data like estimates of national income and related macro economic
aggregates (for example Personal Disposable Incomes, Saving, Consumption,
Capital Formation etc.) which provide a comprehensive view of the
internal and external transactions of the economy over a period form the
system of National Accounts.
2) 1948-49, 1960-61, 1970-71, 1980-81, 1993-94, 1999-2000.
3) RBI – The Handbook of Statistics on the Indian Economy, CMIE –
National Income Statistics, EPWRF publication on NAS and SDP etc.
4) Directorates of Statistics and Economics.
1) See Section 21.4
2) The annual document of NAS published by CSO.
3) See Section 21.5
36
Agricultural and
UNIT 22 AGRICULTURAL AND Industrial Data
INDUSTRIAL DATA
Structure
22.0 Objectives
22.1 Introduction
22.2 Agricultural Data
22.2.1 Agricultural Census
22.2.2 Studies on Cost of Cultivation
22.2.3 Annual Estimates of Crop Production
22.2.4 Livestock Census
22.2.5 Data on Production of Major Livestock Products
22.2.6 Agricultural Statistics at a Glance (ASG)
22.2.7 Another Source of Data on Irrigation
22.2.8 Other Data on the Agricultural Sector
22.3 Industrial Data
22.3.1 Data Sources Covering the Entire Industrial Sector
22.3.2 Factory Sector – Annual Survey of Industries (ASI)
22.3.3 Monthly Production of Selected Industries and Index of Industrial Production (IIP)
22.3.4 Data on Unorganised Industrial Sector
22.3.5 Industrial Credit and Finance
22.3.6 Contribution of Industrial Sector to GDP
22.4 Let Us Sum Up
22.5 Exercises
22.0 OBJECTIVES
• know the kind of agricultural data available;
• explain the different sources of agricultural data and different data
generating agencies;
• describe the characteristics of multivariable industrial data; and
• know how to access the different types of agricultural and industrial data.
22.1 INTROCUTION
In the previous unit, we have looked at data on macro variables like National
Income, Saving and Investment, which tell us about the dimension and broad
structure of the economy and the direction in which the economy is moving.
37
Data Base of Indian We shall now look at two of the sectors of the economy in this unit. One is the
Economy
agricultural sector that provides livelihood to roughly two-thirds of the total
workforce in the country. The other is the industrial sector, which consists of
manufacturing, power, gas and water supply.
22.2 AGRICULTURAL DATA

You are already aware of the importance of agriculture to the Indian economy
and indeed to the Indian way of life. You would, therefore, like to examine
several aspects of agriculture like the level of production of different crops and
commodities, the availability and utilization of important inputs for agricultural
production, incentives, availability of post-harvest services and the role of
agriculture in development. Similarly, you would like to know about the livestock
and their products, fisheries and forestry, people engaged in these activities and so
on. All these analyses require enormous amount of data over time and space. Let us
have a look at what kind of data are available and where.
The Directorate of Economics and Statistics (DESMOA) in the Department of

Agriculture and Cooperation (website http://eands.dacnet.nic.in ) and the
Animal Husbandry Statistics Division (AHSD) of the Department of Animal
Husbandry, Dairy and Fisheries (website http://dahd.nic.in/dahd/statistics/
animal-husbandry-statistics.aspx) in the Ministry of Agriculture are the major
sources of data on agriculture and allied activities. Among the major efforts at
collection of data mounted at regular intervals, the following are the major
ones:
i) the quiquennial agricultural census and the input survey;

ii) the cost of cultivation studies;
iii) annual estimates of crop production;
iv) the quinquennial livestock census; and
v) integrated sample survey to estimate the production of major livestock
products.
DESMOA releases statistics on agriculture and allied activities flowing from

its work and also collected from other division (like the Agricultural Census
Division) and Departments of the Ministry of Agricultural and other Ministries
and agencies in their comprehensive annual publication Agricultural Statistics
at a Glance (ASG)1. Other publications include, “Cost of Cultivation in
India”, the monthly bulletin “Agricultural Situation In India”. AHSD
publishes the data flowing from its activities [which include the activities
mentioned at (iv) and (v) above] and those collected from other divisions of the
Department of Animal Husbandry, Dairy and Fisheries and other agencies in
its biennial publication Basic Animal Husbandry Statistics (BAHS).
22.2.1 Agricultural Census

Agriculture Census forms part of a broader system of collection of Agricultural
Statistics. It is a large-scale statistical operation for the collection and derivation of
quantitative information about the structure of agriculture in the country. An
1
This can be accessed in the Ministry of Agriculture (Department of Agriculture) website .
38 One can also access www.data.gov.in and click on
agricultural operational holding is the ultimate unit for taking decision for Agricultural and
Industrial Data
development of Agriculture at micro level. It is for this reason that an
operational holding is taken as the statistical unit of data collection for
describing the structure of agriculture. Through Agriculture Census it is
endeavored to collect basic data on important aspects of agricultural economy
for all the operational holdings in the country. Aggregation of data is done at
various levels of administrative units. The agricultural census started in
1970-71. The ninth of these censuses relates to the reference year 2010-11.
Periodic Agriculture Censuses are important as these are the main source of
information on basic characteristics of operational holdings such as land use
and cropping patterns, irrigation status, tenancy particulars and the terms of
leasing. This information is tabulated by different size classes and social
groups including Scheduled Castes/Scheduled Tribes which are needed for
development planning, socio-economic policy formulation and establishment
of national priorities. The census also provides the basis for the development of
a comprehensive integrated national system of agricultural statistics and has
links with various components of the national statistical system. The whole
project of Agriculture Census in the country is implemented in three distinct
phases, which are statistically linked together but focus on different aspects of
agricultural statistics. In Phase-I, a list of holdings with their area and social
characteristics and gender of the holders is prepared. In Phase-II, detailed data
on agricultural characteristics of holdings are collected from selected villages.
Thus the whole operation of Agriculture Census in India is not really a
complete Census. In fact, it is a combination of Census and Sample Survey.2
The third phase, which is better known as the Input Survey and conducted in
the year following the census year3, relates to collection of data on the pattern
of input-use across crops, regions and size-groups of holdings. The main
objective of the input survey is to generate data on consumption of various
agricultural inputs, according to major size-groups of operational holdings,
viz., marginal (below 1 ha.), small (1- 1.99 ha.), semi-medium (2- 3.99 ha.),
medium (4- 9.99 ha.) and large (10 ha. and above), for getting an insight into
the consumption pattern of inputs by various categories of farmers. This
information is vital for planning production, imports and distribution of
fertilizers. The inputs covered in the survey include chemical fertilizers, HYV
seeds, pesticides, farmyard manures/compost, bio-fertilizers, agricultural
implements and machinery, livestock and agricultural credit besides data on
input use including use of certified/notified seeds, high yielding variety seeds,
pest control measurements adopted by cultivators, educational qualification,
age and size of households of operational holders are captured through Input
Survey. The results of the first phase, including the all-India report of the first
phase of the ninth agricultural census can be accessed at http://agcensus.nic.in/
agcen201011.html. The provisional results of the second phase have also been
released. Data collection and processing of the third phase is progressing in the
different States and UTs. Selective information is also available on the
internet4 through a query-based facility and can provide data on operational
2
This part is reproduced from http://agcensus.dacnet.nic.in/, which can be accessed to get the
latest data, up to tehsil level from Agricultural Census 2013.
3
The input survey was initiated in the 1975-76 Agricultural Census and conducted in 1976-77.
4
On the website http://agcensus.dacnet.nic.in/nationalholdingtype.aspx, one can access tables
related to agricultural census years 1995-96 to 2010-11. Further, On the website
http://inputsurvey.dacnet.nic.in/nationalsummary.aspx one can access the tables related to
1996-97 to 2006-07 so far. 39
Data Base of Indian holdings at the national, State, district and tehsil levels by type, characteristics
Economy
and size class/group. You can, for example, get information on operational
holdings by size class of the holding, social group (Scheduled Caste, Scheduled
Tribe, Other and all groups) and tenancy characteristics of the holding (lease
terms, land use, irrigation status, irrigation by sources, wells and tube wells,
gross cropped area, cropping pattern, and area dispersal. Clearly, there is a time
lag in the availability of agricultural census and Input Survey results.
22.2.2 Studies on Cost of Cultivation

DESMOA implements a comprehensive scheme for studying the cost of
cultivation of principal crops in India and this results in the collection and
compilation of field data on the cost of cultivation and production in respect of
29 principal crops leading to estimates of crop-wise and State-wise costs of
cultivation and also computation of the index of the terms of trade between
agriculture and non-agricultural sectors. The scheme covers 19 States and
foodgrain crops, oil seeds and commercial crops and selected vegetables.
Statistics of cost of cultivation of principal crops are published by DESMOA
in its publication “Cost of Cultivation in India” and also in ASG. The
Commission for Agricultural Costs and Prices (CACP) makes use of the
estimates of cost of cultivation and production and the structure of cost of
cultivation flowing from these studies, along with an analysis of a wide spectrum
of data on variables like market prices, productivity of the crops concerned,
domestic and global inter-crop price parity, the terms of trade between the
agricultural and the non-agricultural sectors and the supply-demand situation in
arriving at their recommendations to Government on Minimum Support Prices
(MSP). The CACP reports thus contain not only the data thrown up by the cost of
cultivation studies but also those assembled by it from various sources for its work.
For example any of the CACP report presents:
i) estimates of cost of cultivation per hectare, cost of production per quintal,

and the value of the main product and the byproduct;
ii) the breakup of the cost by inputs like human labour, bullock labour,
machine labour, seed, fertilizers and manures, insecticides, irrigation
charges and so on;
iii) a variable input price index; and
iv) month-end wholesale prices of the commodities concerned by variety and
location.
for all the commodities dealt with in the report and for all states. In addition, it
gives information on the index number of whole sale prices for farm inputs,
index numbers of whole sale prices of cereals, pulses, etc., and data on a host
of aspects relevant to the subject matter of the report like area, production,
yield of major crops, buffer stock of cereals and month-wise and area-wise
daily wage rates for agricultural labour5. MSPs for different commodities are
also published in ASG.
5
The cost of cultivation studies reports and related data can be accessed at
http://eands.dacnet.nic.in/Cost_of_Cultivation.htm. Now, the data is available for reference
years 2004-05 to 2011-12. The details regarding the work of CACP and the data presented in
CACP reports are based on the information available in the CACP website
http://cacp.dacnet.nic.in . The MSP up to agricultural year 2014-15 can be viewed in the
40 link http://cacp.dacnet.nic.in/ViewContents.aspx?Input=1&PageId=36&KeyId=0A .
The other useful outcome of the cost of cultivation studies, as already Agricultural and
Industrial Data
mentioned, is the index of the terms of trade between the agricultural and
non-agricultural sectors. This is defined as
IPR [Index of Prices Received]

Index of the Terms of Trade = ———————————————
IPP [Index of Prices Paid]
(Base Year : the triennium ending 1991-92)
The numerator relates to the prices that the agricultural sector gets for its
produce while the denominator relates to the prices that the sector pays for
products it purchases for final consumption, intermediate consumption and
capital formation. The time series of ITT from 1981-82 onwards, complete with
individual values of IPR and IPP and those of the component indices of IPP,
namely, final consumption (weight in IPP 73.54), intermediate consumption
(weight in IPP 21.63) and capital formation (weight in IPP 4.83) are published
in ASG.
The Terms of Trade, specifically, the terms of trade between agriculture and
industry is the ratio of agricultural prices to industrial prices, both measured as
price indices. The terms of trade between agricultural and industry lies at the
heart of the government's policy in the agricultural sector. A rise in the ratio
(agricultural prices divided by industrial prices) means that the agricultural
sector is better off in terms of its purchasing power of industrial goods.
Suppose there are just two goods in the economy, bananas and pins and you
produce and sell bananas. Both initially cost Rs. 10. If now both prices go up
by 100 per cent, nothing happens to the terms of trade and purchasing power
remains unchanged. If instead, banana prices go up by 100 per cent and the
price of pins doesn't, then as a banana manufacturer you are better off since one
banana can now buy two pins instead of one. The ratio or the terms of trade
between bananas and pins has doubled. Substitute agriculture for bananas and
industry for pins and you can get what a rise in agriculture's terms of trade does
for farmers.6
The National Statistical Commission has pointed out that there is a sizeable
time lag in the availability of results for use in the work of CACP and that steps
should be taken to cut down delays in the release of the results of these
studies7.
22.2.3 Annual Estimates of Crop Production

DESMOA makes and releases annual estimates of area, production and yield
in respect of principal crops of foodgrains, oil seeds, sugar cane, fibres and
important commercial and horticulture crops. These crops account for about
87 per cent of the total agricultural output. Estimates of area are based on a
reporting system that is a mix of complete coverage and coverage by a sample,
those of yield are based on a system of crop cutting experiments and General
Crop Estimation Surveys. The preparation of these estimates takes time.
Estimates of crop production are, however, needed earlier and in fact even
6
This part is taken from an article published in the economic times ET classrooms series,
which can be accessed at http://articles.economictimes.indiatimes.com/2002-02-04/news/
27340862_1_bananas-issue-prices-trade.
7
Report of the national Statistical Commission, Chapter 4. 41
Data Base of Indian before the crops are harvested, for policy purposes. Advance estimates of crop
Economy
production are, therefore, made. The first assessment of the kharif crop is made
in the middle of September, that is, when the South-West Monsoon is about to
be over. The second advance estimate – a second assessment of the earlier
estimate of the kharif crop and the first assessment of the rabi crop – is made in
January. The third advance estimate is made at the end of March or the
beginning of April and the fourth in June. The methodology for estimating area
and yield of crops and for making advance estimates are given in ASG.
Estimates of annual8 production of the principal crops mentioned above, gross
area under different crops and yield per hectare of these crops and Index
Numbers on these variables with base year the triennium ending 2007-08 =
1009 are also available in ASG in the form of a time series. So are estimates of
production of crops by States.

1) Give two major sources of data on Agriculture and allied activities like
dairy and fisheries.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) How can you retrieve the agricultural census data at district and tehsil
level?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) How are annual estimates of crop production made by the Directorate of
Statistics and Economics?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
8
Season-wise in some cases.
9
The base of these index numbers is a triennium in order to even out year to year to
fluctuations of the underlying variable(s). The earlier base TE 1981-82 for these index
number series was changed to TE 1993-94 in 2000-01 in order to update the index numbers
and also for these to be in harmony with the index numbers of wholesale prices and
42 industrial production with base 1993-94 = 100.
22.2.4 Livestock Census Agricultural and
Industrial Data
The last quinquennial livestock census, the nineteenth in the series, was
conducted in 2012. The census has collected information, district-wise on
livestock, poultry, fishery and also agricultural implements. Livestock covers
cattle, buffaloes, sheep, goats, pigs, horses and ponies, mules, donkeys, camels,
yak, mithun and also, dogs and rabbits. These are classified by age (appropriate
for the species). sex, breed, function (‘breeding’, ‘work’, ‘both’ and ‘others’ for
males and ‘in milk’ or ‘dry’ in the case of females). Poultry covers cock, hen,
duck, and drake, which are classified as desi and ‘improved’ varieties. Fishery
covers fishing activity (inland capture, inland culture, marine capture and
marine culture), members involved in fishing (male, female and child), persons
engaged in fishing (part time male, full time male, part time female and full
time female) craft/gear, namely, trawlers, gin-netters, liners, seiners, motorized
canoe/maran, non-motorized canoe/maran, and miscellaneous gears like trawler-
net, gill netter-net, cast-net, drag-net, hook lines, set barriers and others. Each
of these classes except the last one is further classified into size (length) and
horsepower. Agricultural implements cover annually operated implements,
animal operated implements, plant protection equipment, water lifting devices,
tractor and power operated implements, equipment for livestock and poultry
and horticulture tools. Each group provides further data on four or five specific
equipments like for example, thresher under the first group, ST Plough under
the second, TPO Spray under the third, diesel under the fourth, Crawler Tractor
under the fifth, incubator under the sixth and power operated tools under the
seventh group.
The provisional results of the 2012 census are available on the website of the
Department of Animal Husbandry, which can be accessed at http://dahd.nic.in/
dahd/WriteReadData/Livestock.pdf. ASG and BAHS also present some data
from the livestock census.
22.2.5 Data on Production of Major Livestock Products

AHSD is responsible for collection of statistics on animal husbandry, dairy and
fisheries. These are published, as already mentioned, in BAHS. The latest
relates to 2004. This presents data up to 2003-04, on
i) long time series of estimates of production of milk, eggs, meat and wool;
ii) State-wise short time series of such estimates;
iii) State-wise and national estimates of per capita availability of milk and
eggs;
iv) contribution of cows, buffaloes and goats to milk production and of fowls
and ducks to egg production in different States in 2003-04;
v) average annual growth rates of production of major livestock products,
milk, eggs and wool;
vi) short time series on imports and exports of livestock and livestock
products;
vii) short time series on area under fodder crops, pastures and grazing in
different States;
viii) estimates of dry and green fodder production in different States during the
three-year period ending 2002-03;
43
Data Base of Indian ix) a short time series on the no. of artificial inseminations performed in
Economy
different States;
x) achievements in key components of dairy development;
xi) livestock and poultry (numbers) in India and in different States – 1992,
1997 and 2003; and
xii) time series on world livestock production and milk production.
22.2.6 Agricultural Statistics at a Glance (ASG)

The pocket book on agricultural statistics and the agricultural statistics at a
glance are two annual publications available in the website http://eands.dacnet.
nic.in/latest_2006.htm. One of the most important statistics, which is the
statistics on Land utilization, is available in the ASG. As for inputs and access
of individual operational holdings to such inputs, ASG provides data on:
i) Methodology of crop estimation,

ii) Socio-economic indicators: population and growth rates, State-wise
classification of cultivators, Gross State Domestic Product for agriculture
and allied activities,
iii) Outlays, expenditure and capital formation in agriculture,
iv) Area, production and yield of principal crops: target and achievement of
production of major crops, three largest producing States of important
crops, season-wise area, production and yield of food grains (rice, wheat,
jowar, bajra, maize, total pulses, gram, tur, masur, nine oilseeds, groundnut,
rapseed and mustard, soyabean, sunflower), cotton, jute, mesta, sugarcane,
tobacco, guarseeed, etc.; yield rates of principal crops, area and yield under
the high yielding varieties, area and production of horticulture and plantation
crops (potato, onion, coconut, cashewnut, arecanut, garlic, ginger, sweet
potato, turmeric, chillies, cardamom, pepper);
v) All-India index numbers of area, production and yield of principal crops;
vi) Area, production and yield of major crops in different countries;
vii) Minimum support prices/marketed surplus ratios of various agricultural
commodities;
viii) Procurement by public agencies, per capita net availability of foodgrains,
consumption and stocks of foodgrains and raw jute,
ix) Import and export .of agricultural commodities, tariffs and bound rates of
major agricultural commodities, trends in wholesale price index of
foodgrains and commercial crops;
x) Land use statistics: agricultural land by use in India, selected categories
of land use;
xi) Inputs: production and use of agricultural inputs (fertilizers, crop-wise
distribution of certified/quality seeds, consumption of electricity for
agricultural purposes, flow of institutional credit for agricultural sector,
state-wise number of kisan credit cards, state-wise/season-wise national
agricultural insurance scheme, etc.
44
xii) Agricultural census data: number and area of operational holdings by size Agricultural and
Industrial Data
group – marginal (size less than one HA10 area), small (1 to 2 HA), semi-
medium (2 to 4 HA), medium (4 to 10 HA) and large (more than 10 HA),
area irrigated by different source of irrigation and by size class and the
position in different States;
xiii) Estimated number of rural households, farmer households, indebted
farmer households by size class of land possessed;
xiv) Ceiling on land holding and wages for agricultural work;
xv) Livestock population in India, all-India and State-wise production of
milk, eggs, meat and wool;
xvi) Fish production in India and States;
xvii) Rainfall scenario and management of natural disaster.
The subjects Subsidies is a much-debated subject nowadays and agricultural

subsidies in developing countries and developed countries of Europe and in
USA are also much in the news. The Ministry of Agriculture provides a time
series of the amount of subsidy given to agricultural with its break-up into
subsidy for (i) fertilizers (ii) electricity, and (iii) irrigation11, (iv) other
subsidies given to marginal farmers and Farmers Cooperative Societies in the
form of seeds, development of oil seeds, pulses, etc. One can easily derive the
size of the total subsidy relative to GDP in current prices. The limitations to
which these figures are subject need to be kept in mind.
22.2.7 Another Source of Data on Irrigation

The data on irrigation presented in ASG (referred to the proceeding sub-
section) is based on crop statistics collected by DESMOA from village through
a mix of a reporting system and sample surveys. The Central Water
Commission (CWC) under the Ministry of Water Resources collects
hydrological data on the important river systems in the country through 877
hydrological observation sites. The Ministry also conducts periodic Censuses
of Minor Irrigation Works along with a sample check to correct the Census
data. The five Census conducted so far related to 1986-87, 1993-94, 2000-01,
2006-07 and 2013-14. This is conducted in all States/UTs except in Daman &
Diu and Lakshadweep. The reports can be seen in the website of the Ministry
http://micensus.gov.in.The Report of the Minor Irrigation Census 2006-07
provides information on the 20.7 million minor irrigation works in the country
like the share of ground water and surface water works, crop-wise utilization of
the irrigation potential created, the manner in which the water is distributed in
the field – sprinkler, drip, open channel or underground water irrigation and
state-wise distribution of these variables.
The National Commission on Statistics had in their report made the following
observations12 on irrigation data available from DESMOA and the Ministry of
Water Resources:
10
HA: abbreviation for hectare.
11
The rates for supply of water to farmers are kept low as a matter of policy. This results in a
loss to the Government Irrigation System. The excess of operating costs over gross revenue
is treated as imputed irrigation subsidy.
12
Chapter 4 of the Report. 45
Data Base of Indian i) The CWC had a large volume of data on various aspects of irrigation
Economy
without any statistical analysis of such data being carried out. The CSO
and the CWC need to get together and ensure that the data collected in
analysed and put to use by the statistical machinery.
ii) There is a large variation between the statistics on the “area irrigated”
published by DESMOA and the “irrigation potential actually utilized”
published by the Ministry of Water Resources.
iii) Data users should be made aware of the reasons for such variation –
differences (if any) in concepts, definition etc.
iv) The reluctance on the part of the State Governments to furnish data in
view of their vested interest in the sharing of water is an added problem.
v) Data from both the sources – DESMOA and the Ministry of Water
Resources – are available after a large time lag and this needs to be
reduced.
22.2.8 Other Data on the Agricultural Sector

Data on forest cover is, already mentioned, as part of land-use statistics
presented on the basis of nine fold land-use classification in ASG. Forest
Survey of India (FSI) also collects data on forest cover (dense forest, open
forest and mangroves) through a biennial survey by making of Remote Sensing
(RS) technology since 1987. The latest survey relates to 2013. Digital
interpretation has reduced the time lag in the availability of such data obtained
earlier through periodic reports from field formations. There are discrepancies
between the data on forest area given by ASG and those given by FSI due to
differences in the concepts and definitions used by the two organizations in
collecting data. Data on production of industrial wood, minor forest produce
and fuel wood are available with the Principal Chief Conservator of Forests
in the Ministry of Environment and Forests.
The National Bank for Agriculture and Rural Development (NABARD) is

another agency that is closely involved in agriculture, especially in the matter
of credit to the sector through cooperatives. The annual reports of NABARD
and its publication like “Statistical Statements Relating to the Cooperative
Movement in India” (Part I deals with credit societies and the other with non-
credit societies) and “Key Statistics on Cooperative Banks”, besides
NABARD’s website (https://www.nabard.org/english/home.aspx) would be
useful source of information on agricultural credit.
Statistics on area, production and yield of crops and subsidies given to

agriculture are also available as time series in the Economic Survey,
expenditure budget documents of the Ministry of Finance and the RBI
Handbook of Statistics on Indian Economy (http://rbidocs.rbi.org.in/
rdocs/Publications/PDFs/000HSE13120914FL.pdf). The third source is
important in the context of the policy direction to banks and all financial
institutions to ensure that 40 per cent of the total institutional credit should flow
to the priority sector of which agricultural sector forms a part. The data
warehouse of the RBI can be accessed at http://dbie.rbi.org.in. The RBI
Handbook of Statistics on Indian Economy – 2013-14 publishes time series
data on:
i) National income, saving and employment from 1952-53 onwards;

46
ii) Output and prices of agricultural production, from 1956-57 onwards; Agricultural and
Industrial Data
iii) Assets and liabilities of the RBI, components of money stock, etc. from
1962-63 onwards;
iv) Select aggregates of Scheduled Commercial Banks (SCBs), like demand
deposit, time deposit, borrowings from RBI, liabilities to banks,
investment in Government securities, food credit, non-food credit, cash in
bank, balance with RBI, assets with banks from 1956-57 onwards;
v) Sectoral deployment of non-food gross bank credit; industry-wise
deployment of gross bank credit;
vi) Direct and indirect institutional credit to agriculture and allied activities,
1975-76 onwards;
vii) SCB’s direct finance to farmers according to size of land holdings, 1985-
86 onwards;
viii) the distribution of non-food credit by priority sector credit and credit for
other sectors;
ix) the share of agriculture in priority sector credit;
x) Financial market: structure of interest rate, etc.;
xi) Public finances: key deficit indicators of central Government, major
components of receipts of central Government, key deficit indicators of
state governments, pattern of receipts of state governments, etc.;
xii) Trade and balance of payment, etc.
Direct credit to agriculture is also available from other institutions and the
Handbook furnishes data on:
i) the short and long term loans issued by each of the institutions –
Cooperatives, Scheduled Commercial Banks (SCBs), Regional Rural
Banks (RRBs) and the State Governments directly to beneficiaries or
borrows for agriculture and activities allied to it during a year; and
ii) the loans outstanding at the end of each year.
Institutions like SCBs, Cooperative, RRBs and the Rural Electrification

Corporation (REC) provide indirect credit to agriculture and allied activities.
This is routed through some other agency, conduit or tier like the Electricity
Boards (EBs), the Primary Agricultural Credit Societies (PACS), the Farmers’
Service Societies (FSS), the Large-sized Adivasi Multi-Purpose Societies
(LAMPS) and State-supported corporations and agencies. Such indirect credit
is designed to promote agricultural productivity or to increase the agricultural
income of the ultimate beneficiary, through a number of steps like distribution
of fertilizers and other inputs to these activities, loans to EBs and loans to
farmers through PACS, FSS and LAMPS. Other types of indirect credit include
advances to State-supported corporations and agencies for onward lending of
funds (up to Rs. 10,000/-) to weaker sections of society engaged in agriculture,
namely, small and marginal farmers and those engaged in activities allied to
agriculture.
We have already seen in the preceding Unit (on Macro-variable like national
income) that National Accounts Statistics (NAS) publishes data on the
contribution of agriculture and its sub-sectors to GDP and other measures of
47
Data Base of Indian national/domestic product. NAS also provides information on value of output
Economy
of various agricultural crops including those of drugs and narcotics, fibers,
fruits and vegetables, livestock products like milk group products, meat group
products, eggs, wool and hair, dung, silkworm cocoons and honey, forestry
products like industrial wood, firewood and minor forest produce, inland fish
and marine fish. It also furnishes data on capital formation in agriculture and
animal husbandry, forestry and logging and fishing.

1) How can you access the data on per-capita availability of milk, egg, wool,
cow and buffaloes?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
2) What do you mean by the term ‘cropping intensity’?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
3) Which document contains the data on subsidy given to agriculture?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
4) Name the document which contains the data on food and non food credit
provided by Scheduled Commercial Banks.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
48
Agricultural and
22.3 INDUSTRIAL DATA Industrial Data
The industrial sector can be divided into a number of subgroups. This grouping
is different from grouping by economic activity. Such divisions arise from
framework factors like applicability or coverage or certain laws, employment
size of establishments of groupings for purposes of promotional support by
Government. Such groupings are the organized and unorganized sectors, the
factory sector (covered by the Factories Act, 1948 and the Bidi and Cigar Workers
Condition of Employment Act, 1966), small-scale industries, cottage industries,
handicrafts, khadi and village industries (KVI), directory establishments (DE), non-
directory establishments (NDE), own account enterprises (OAE)13. Attempts have
been made to get at a detailed look at the characteristics of some of these sub-
sectors of the industrial sector, as the data sources covering the whole sector
often do not provide information in such detail. Let us now examine the kind of
data available for such individual subgroups of the industrial sector and those
that cover the entire industrial sector.14
22.3.1 Data Sources Covering the Entire Industrial Sector

There are five sources that cover the whole spectrum of economic activity or
non-agricultural activities and therefore, the entire industrial sector. However,
four of these sources give data only on levels of industrial employment:
i) The first source is the decennial Population Census. This provides data
on the levels of employment in various economic activities across the
economic spectrum, and therefore, the industrial sector, down to the latest
NIC15 three-digit code levels. Thus, Census 1991 provides data at 3-digit
level of NIC 1987, Census 2001 provides data at 3-digit level of NIC 1998
and Census 2011 will be providing data at 3-digit level of NIC 2008. It
also provides (i) employment levels in each of the three-digit NIC code
classified by Occupational Divisions [the first digit of the National
Classification of Occupations 1968 (NCO 1968)], (ii) industrial
employment in broad industrial sectors (a nine-fold classification of
economic activities is considered16) classified by broad age groups, and
(iii) industrial employment in broad sectors (nine-fold classification, as
above) by levels of education. Such details are available up to district
levels. These data, however, become available after a time lag of 4-5 years
after completion of the Census field-work.
ii) The second source consists of the quinquennial sample surveys relating
to labour force,employment and unemployment conducted by the National
Sample Survey Office (NSSO). These surveys also provide similar type of
data on industrial employment, down to State levels, separately for rural
and urban and males and females. Data by the districts can also be
tabulated, with a suitable caution for the district level sample sizes, as the
13
DEs are those establishments that employ at least six persons. NDEs are establishments that
employ at least one but not more than five employees. OAEs are the self-employed.
14
One distinction between the economic activity classification and the industrial sector
classification is the placement of Mining and Quarrying industries. Although this is a part
of “primary sector” under the economic activities, it is considered as a part of industrial
sector.
15
See the section on India Statistical System in Unit 20 on Macro-Variable Data: National
Income etc.
16
Same as footnote 15 given above. 49
Data Base of Indian primary data collected (unit record data) can be obtained on CDs from
Economy
NSSO17. Key results from the NSSO surveys are available after about six
months, while the final report becomes available one year after the surveys
are completed.
iii) The third source is the Economic Census, which has been conducted in
1977, 1980, 1990, 1998, 2005 and 2012. The Economic Census covers all
economic enterprises in the country except those engaged in crop
production and plantation and provides data on employment in these
enterprises18. This provides a frame for the conduct of more detailed
follow up (enterprise) surveys covering different segments of the
unorganized non-agricultural sectors, which in turn throw up data on
production and employment in these segments, useful for an analytical
study of these segments and in the compilation of national accounts.
iv) The fourth source is the Employment Market Information (EMI)

programme of the Directorate General of Employment & Training
(DGE&T), Union Ministry of Labour & employment and the
Directorate of Employment under the State Governments. This is based
on the statutory quarterly employment returns furnished by non- agricultural
establishments in the private sector employing 10 or more persons and all
public sector establishments. This source of industrial employment
provides data at quarterly intervals down to district level. The quarterly
data are available with a time lag of about a year. Detailed data at the level
of three digit NIC 1987 codes are available in the Annual Employment
Reviews based on this programme with a time lag of about two years.
While data at national and State levels would be available from the
National reports, district level data would be available from the State
reports. This source or course covers only the organized sector.
v) The fifth source consists of the periodic unorganized sector surveys of the
NSSO, which covers the unorganized non-agricultural enterprises. These
surveys, usually conducted as follow-up surveys of the Economic Census,
provide data on characteristics of the enterprises, its 5-digit NIC code,
operating expenses, receipts, gross value added, fixed assets, liabilities and
the number of workers working on a regular basis in the enterprise.
Similar characteristics for the informal sector enterprises are also tabulated
and published from these survey data by the NSSO. These data are
tabulated and published at State levels, separately for rural and urban. Data
by the districts can also be tabulated, with a suitable caution for the district
level sample sizes, as the primary data collected (unit record data) can be
obtained on CDs from NSSO19. Key results from the NSSO surveys are
available after about six months, while the final report becomes available
one year after the surveys are completed.
17
You can see the rate list of NSSO unit level survey data from http://mospi.nic.in/
Mospi_New/upload/nsso/ratelist_UnitData.pdf.
18
The report on the Economic Census 1998 is available on the website of the Ministry of
Statistical and Programme Implementation (MoSPI) www.mospi.nic.in . The website of
MoSPI announces that the site’s new address is www.mospi.gov.in . The report on the
Economic Census 2005 has been released by the Ministry on the 12th June, 2006, according
to news item “Rural enterprises growing faster – Employment rate high in J & K: Economic
Census 2005” appearing on page 11 of The Hindu, Chennai Edition, dated Tuesday the 13th
June, 2006. The Ministry’s website does not say anything about the 2005 report till 14/6/06.
19
You can see the rate list of NSSO unit level survey data from http://mospi.nic.in/
50 Mospi_New/upload/nsso/ratelist_UnitData.pdf.
There are, however, other sources that provide data on production or on a Agricultural and
Industrial Data
variety of interrelated aspects of industry like employment, output, inputs and
so on. Let us consider these sources.
22.3.2 Factory Sector – Annual Survey of Industries (ASI)

Collections of industrial statistics on aspects like output, employment, input
and value added was initiated as early as 1944, when an annual Census of
Manufacturing Industries (CMI) was launched, covering factories registered
under the Indian Factories Act, 1934 and employing 20 or more persons and
using power. Subsequently, a Sample Survey of Manufacturing Industries
(SSMI) was started from 1949. While the former was restricted to only 29 of
the 63 groups of industries, SSMI covered the remaining 34 groups of
industries. As soon as the Factories Act, 1948 came into force, the coverage of
both CMI and SSMI was extended to cover factories registered under the
Factories Act, 1948 employing 10 or more persons and using power and those
employing 20 or more persons without using power. The CMI and SSMI
reports constituted the main source of data on different aspects of the factory
sector up to 195820. A brief coverage of the ASI can be seen at the site
http://www.csoisw.gov.in/CMS/En/1024-asi-manual.aspx and at http://www.
csoisw.gov.in/cms/cms/Files/572.pdf.
The annual Survey of Industries (ASI) replaced the CMI and the SSMI in 1960
after the Collection of Statistics Act, 1953 and the Collections of Statistics
Rules, 1953 came into force. Detailed industrial statistics relating to industrial
units in the country21 like capital, output, input, value added employment and
factor shares are collected in ASI. The survey was launched in 1960 and has
been conducted every year since then except in 1972. From 2011, the ASI is
conducted under the Collection of Statistics Act 2008 and the Collection of
Statistics Rules 2011, except in the State of Jammu and Kashmir, where it is
conducted under the State Collection of Statistics Act, 1961 and the rules
framed thereunder in 1964. The frames used for the survey consists of:
a) all factories registered under Sections 2m(i) and 2m(ii) of the Factories
Act, 1948 employing 10 or more workers using power as well as those
employing 20 workers but without using power;
b) bidi and cigar manufacturing establishments registered under the Bidi and
Cigar Workers (Conditions of Employment) Act, 1966 with coverage of
units as in (i) above;
c) certain servicing activities like water supply, cold storage, repair of motor
vehicles and other consumer durables like watches etc. Though servicing
industries like motion picture production, personal services like laundry
services, job of dyeing, etc., are covered under the Survey but data are not
tabulated separately, as these industries do not fall under the scope of
industrial sector defined by the United Nations. Defence establishments,
oil storage and distribution depots, restaurants, hotels, café and computer
services and the technical training institutes, etc., are excluded from the
purview of the Survey.
20
The ASI section of the CSO website.
21
The coverage was extended to Jammu & Kashmir with the passing of the Collection of
Statistics Act, 1961 (Jammu & Kashmir) and the notification of the Collection of Statistics
Rules, 1964 under that Act. 51
Data Base of Indian The frames for the ASI are revised every year. At the time of revision, the
Economy
deregistered factories are removed and newly registered factories are added in
the frame. In ASI 2013-14, more than 2.65 lakh units were in the ASI frame, up
from about 2.52 lakh units in ASI 2012-13. The “factory sector” thus consists
of factories, bidi and cigar manufacturing units, electricity sector and some
service units. Electricity units registered with the CEA, the departmental units
such as railway workshops, Road Transport Corporation workshops, Government
Mints, sanitary, water supply, gas storage, etc., are not covered from 1998-99
as there are alternatives sources of data for CSO for purposes of compiling
GDP estimates in respect of these segments of industry.
The reference period for the survey is the accounting year April to March
preceding the date of the survey. The sampling design and the schedules for
the survey were revised in 1997-98, keeping in view the need to reduce the
time lag in the availability of the results of the survey. The survey will not,
however, attempt estimates at the district level. The survey used NIC 1987 for
classifying economic activities before ASI 1997-98 and shifted to NIC 1998
with effect from ASI 1998-9922. From ASI 2009-10, the NIC 2008 codes are
being used.
CSO has released the final results of ASI 2011-12 and provisional results
of ASI 2012-13. Results in respect of selected characteristics are available in
electronic media at various levels of aggregation23:
• All India by 4-digit level of NIC 2008,

• State by 3-digit level of NIC,
• Unit level with suppressed identification, etc.
All the reports of previous years can be accessed on the website of the
Ministry of Statistics and Programme Implementation (MoSPI). These
results are:
1) Time series data (1980-81 to 2007-08) of principal characteristics24 of the

factory sector;
2) principal characteristics by major industry group;
3) principal characteristics by major States;
4) estimates of some important characteristics by States for 2002-03;
5) estimates of some important characteristics by 3-digit code of NIC 1998;
6) rural-urban break-up of principal characteristics; and
22
NIC 1987 and NIC 1998 contain concordance tables to enable data users to recast data of
earlier years from one NIC structure to another.
23
You can download the ASI 2011-12 reports from the ASI portal of the CSO at
http://www.csoisw.gov.in/CMS/cms/Home.aspx.
24
The principal characteristics are: 1) no. of factories, 2) fixed capital, 3) working capital, 4)
invested capital, 5) outstanding loans, 6) no. of workers, 7) mandays of workers, 8) no of
employees, 9) mandays – employees, 10) total persons engaged, 11) wages to workers, 12)
total emoluments, 13) old age benefits, 14) social security benefits, 15) other benefits, 16)
fuels consumed, 17) total inputs, 18) products, 19) value of output, 20) depreciation, 21) net
value added, 22), rent paid, 23) interest paid, 24) net income, 25) gross fixed capital
formation (GFCF) 26) (a) material, fuels, etc., (b) semi-finished goods, (c) finished goods,
52 27) total, 28) gross capital formation (GCF), 29) profits.
7) principal characteristics by type of organization25. Agricultural and
Industrial Data
8) estimates for structural ratios and technical coefficients for factory sector
from 2006-07.
Data users can find out from the website (click on ASI DATA COST) to find
out what data – tables or unit record data –are available and the cost of
obtaining the same from CSO.
CSO had earlier published ASI 2010-11 final (detailed) results in two
volumes. These volumes present data at the 3-digit and 4-digit level of NIC
1998 codes for the national level and at the 3-digit level of NIC 1987 codes for
individual States. Data on the following aspects of the industries forming part
of the factory sector are presented:
i) Capital employed, input, output and gross value added (GVA),

ii) Employment, mandays and wages,
iii) Fuel consumed,
iv) Material consumed,
v) Products and by-products.
These volumes as well as those relating to earlier years are available also on
electronic media. In addition, CSO has also released time series data on ASI
in 5 parts – Volumes I relates to ASI 1959-71, Volume II, Series A to ASI
1973-74 to 1981-82, Volume II, Series B to ASI 1982-83 to 1988-89, Volume
III, Series A to ASI 1989-90 to 1993-94 and Volume III, Series B to
1994-95 to 1997-98. These volumes present data on important characteristics
for all-India at two-digit and three-digit NIC code levels and for the States at
two-digit NIC code levels. These publications are also available in electronic
media on payment.
The database on the Annual Survey of Industries, 1973-74 to 2009-10 also

provides time series ASI data on the principal characteristics of the factory
sector, along with concepts and definitions used. The data can be procured
from the EPWRF on payment.
The data available from ASI can be used to derive estimates of important
technical ratios like capital-output ratio, capital-labour ratio, labour cost per
unit of output, factor shares in net value added and productivity measures for
different industries as also trends in these parameters. The most important use
of the detailed results arises from the fact that these enable derivation of
estimates of
i) the input structure per unit of output at the individual industry; and
ii) the proportions of the output of each industry that are used as inputs in
other industries, enabling us to use the technique of input-output analysis
to evaluate the impact of a change effected in (say) the output of an
industry on the rest of the economy. The construction of the Input-Output
25
Types of organization are 1) proprietorship, 2) joint family (HUF), 3) partnership, 4) public
limited company, 5) private limited company, 6) govt departmental enterprises, 7) public
corporations, 8) corporate sector (4+5+6+7), 9) Khadi& Village Industry, 10) handloom
industry, 11) cooperative sector, 12) others (including ‘not recorded’). 53
Data Base of Indian Transaction Tables (I-OTT) for the Indian economy is largely based on
Economy
ASI data. As we know, I-OTT provides the basic ratios required for Input-
Output Analysis, which has a wide range of uses, from analysis of economic
structure and backward and forward linkages of different industries to the
setting up of targets of production, investment, employment and so on at the
macro level.
22.3.3 Monthly Production of Selected Industries and Index of

Industrial Production (IIP)
CSO prepares and releases monthly indices of industrial production (IIP)
and also the monthly use-based index of industrial production (base year
2004-05). IIP with base year 2004-05 was first released in April 2005. The base
year for IIP was earlier 1993-9426. IIP with the new base year has (i) a wider
coverage of items than the IIP with base year 1993-94, (ii) the weighting
diagram of the new index has taken into account the contribution of the
unorganized manufacturing sector, (iii) small scale industry items have been
given individual weights in the weighting diagram of the new index, (iv) the
revised IIP series follows NIC 2004 (the earlier one followed NIC 1987) and
(v) the new IIP is released within six weeks of the reference month. The IIP
(2004-05) is a quantitative index based on production data received from 16
source agencies covering 682 items clubbed into 399 item groups in the basket
of items of the index.
The Directorates of Statistics and Economics of the State Governments

and Union Territory Administrations (DESs) had been preparing IIPs for
their respective areas. These IIPs were not comparable with each other or
with that prepared by CSO for the country as a whole because of differences in
the base year, basket of items, data and methodology used for the construction
of the indices. The question of preparing State-wise IIPs that are comparable
with the national IIP was, therefore, examined and State Governments and
Union Territory Administrations have initiated work I this regard on the basis
of the recommendations made in this regard. The preparation and release of IIP
is at different stages in different States and Union Territories. For example the
Governments of Tamil Nadu, Andhra Pradesh, West Bengal, Assam, Goa,
Punjab, Haryana, Rajasthan and Maharashtra have released the new IIP
prepared as a result of these efforts and others are in the process of
preparing such IIPs.
CSO releases Quick Estimates of IIP within six weeks of the closing of the
reference month. CSO has released the IIP for December, 2014 through the
Press Release dated 12th February 2015. The Release provides estimates of
• monthly IIP – overall index and for sectors (mining, manufacturing and
electricity) – for the period April, 2014 to December, 2014;
• monthly IIP at 2-digit level NIC 2004 code for December 2013 and
December 2014 and the cumulative index for the period April-December
2014; and
26
Change in the base of IIP is necessary to measure the real growth in the industrial sector.
The United nations Statistical Office (UNSO) recommends that the base year should be
changed quiquennially. IIP was first prepared in India with the base year as 1937. The base
54 year was then changed to 1946, 1951, 1956, 1960, 1970, 1980-81, 1993-94 and 2004-05.
• monthly IIP use-based index (basic goods, capital goods, intermediate Agricultural and
Industrial Data
goods, consumer goods, consumer durables, consumer non-durables) for
the period April, 2014 to December, 2014.
CSO follows the SDDS norms of the International Monetary Fund (IMF)27 in
respect of estimates of IIP and, accordingly, the Press Release of the 12th
February 2015 announces that IIP for January 2015 would be released on the
12thMarch 2015. This Press Release is accessible in the MoSPI website. The
website also has time series data at 2-digit level NIC code on (a) monthly IIP
from April, 1994 (b) annual averages from 1994 and also on the monthly use-
based IIP. Similar time series data on (a) monthly IIP with base year 1980-81 at
2-digit level of NIC 1970, and (b) monthly use-based IIP (base year 1980-81)
are also available from April,1990.
The Reserve Bank of India’s (RBI) “Handbook of Statistics on the Indian

Economy” also presents IIP at two-digit-level, the use-based IIP and index
numbers of Infrastructure Industries28 as also data on production in selected
industries. Monthly statistics of mineral production are published by the
Indian Bureau of Mines (IBM), Nagpur in their publication “Monthly
Statistics of Mineral Production”. It also releases mineral group-wise and
State-wise value of mineral products. The latest data available while writing
this chapter was from March 201329. The CSO publication “Energy
Statistics” bring together important data on different sources of energy in the
country at one place. It presents time series data that give a broad picture of the
trends in production, consumption and price indices of major sources of
conventional energy in the country for last 30 years. The ministries of
Petroleum and Gas, Power and Non-Conventional Energy provide
information in their respective spheres of activity. The Economic Intelligence
Services (EIS) of CMIE, Mumbai has volumes on Energy and
Infrastructure that present detailed data on the trends in these sectors. A visit
to the EPWRF website (click with words EPWRF through the Google search
engine) would also be rewarding in terms of time series data on industrial
production. At the global level, the International Energy Agency (IEA) is
considered as one of the most authoritative organisations in dealing with
energy issues. The IEA publishes the world Energy Outlook. The latest one,
World Energy Outlook 2014, was published in November 2014. The IEA
portal http://www.iea.org/statistics/ also provides for search of statistics by
country. IEA online data services provide monthly data on oil and gas,
quarterly energy prices, etc. However, data from this site are to be purchased,
be it hard copy, CD-RoM or pdf files. The UNSD also compiles and publishes
energy statistics in a comparable manner for 224 countries of the world. The
reports, although a bit dated, can be downloaded free of charge from their
website30.
27
See Indian Statistical System in Unit 18 on “Macro Variable Data: National Income,
Saving, Investment”.
28
Infrastructure industries are electricity, coal, steel, cement, crude petroleum and refined
petroleum products. These industries are part of the basket of items of IIP and the Index
for Infrastructure Industries can be computed from the indices for the individual
infrastructure industries.
29
These can be accessed from the IBM website http://ibm.nic.in/msmpmar13.htm.
30
You can visit the link http://unstats.un.org/unsd/energy/yearbook/2011/004-10.pdf and
read the introduction to the Yearbook to have an idea of the data being made available by
them. 55
Data Base of Indian Check Your Progress 3
Economy
1) Indicate the major sources of data on levels of industrial employment.
Which one is the most and which one the least frequent? Which of the
sources shall you use for your analysis, if the analysis are to be done i) for
backward districts and ii) for major States of India?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) How can you access the report of the economic census 1998?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) What do you understand by quarry based system developed by CSO?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) List two important uses of ASI data.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
5) From which document can you have data pertaining to various sources of
energy?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
56
22.3.4 Data on the Unorganised Industrial Sector Agricultural and
Industrial Data
The Development Commissioner for Small Scale Industries (DCSSI) in the

Ministry of Small Scale Industries, in the Central Government and the
Directorates of Industries of State Government and Union Territory
Administration provide data on small-scale industrial units registered with the
latter set of agencies. The DCSSI has conducted a census of small scale
industrial units thrice – in 1973-74 (reference year 1972), 1990-91 (reference
year (1987-88) and in November, 2002 (reference year 2001-02). The results of
the third census can be seen in the publication “Final Results: Third all India
Census of SSI – 2001-02” released by the Ministry of Small Scale
Industries. Broad details of the performance of small-scale industries are
available in the Annual Reports of the Ministry of Small Scale Industries. Time
series data on employment, production, labour productivity in small-scale
industries (SSI) and value of exports of the products of small-scale industry
are also available in the RBI Handbook. Data on some parts of Khadi and
Village Industries Commission (KVIC), handlooms and handicrafts do get
included in ASI but data relating exclusive to these sub-sectors are available in
the Annual Reports of these organizations or in the Annual Reports of the
Ministries under which these Boards/Commissions function.
Surveys of unorganized sector enterprises conducted once in five or six years

contribute to the strengthening of the database of the unorganized sector. The
following rounds of the NSSO provide data on input, output, value added,
fixed assets, liabilities, number of working owner, hired worker and other
worker (most of whom would be unpaid family workers) and related data on
the unorganized non-agricultural sector. You can become a user in the website
of the MoSPI and download relevant reports for your research, free of cost.
You can also purchase relevant micro data from the MoSPI.
NSS Round Reference Year Subject coverage

No.
67th July 2010 – June 2011 Unorganised Non-agricultural Enterprises
(excluding cconstruction)
63rd July 2006 – June 2007 Unorganised service sector enterprises
(excluding construction)
62nd July 2005 – June 2006 Unorganised manufacturing Enterprises
57th July 2001 – June 2002 Unorganised service sector enterprises
(excluding construction and finance)
56th July 2000 – June 2001 Unorganised manufacturing Enterprises
55th July 1999 – June 2000 Informal sector non-agricultural enterprises
(excluding finance)
53rd January – December Unorganised trading Enterprises
1997
51st July 1994 – June 1995 Unorganised manufacturing Enterprises
th
46 July 1990 – June 1991 UnorganisedNDTE and DTE (trading)
Enterprises
45th July 1989 – June 1990 Unorganised manufacturing Enterprises
57
Data Base of Indian At the global level, the UN International Labour Organisation (ILO) supports
Economy
surveys on employment and child labour related issues. The Laborsta database
and the Key Indicator of Labour market (KILM) of the ILO contains
comprehensive country level comparable estimates on labour force31,
employment and unemployment, forced labour, informal economy and labour
migration related data. The UN Industrial Development Organisation (UNIDO)
has recently developed the world productivity database, providing total factor
productivity and related indicators.
We have looked at data sources that provide data on production and also on
certain related variables like capital employed, employment etc. How is the
capital financed? Let us look at some of the sources that throw light on these
matters.
22.3.5 Industrial Credit and Finance

The RBI Handbook on Statistics of the Indian Economy provides time
series data on the sectoral deployment of non-food gross bank credit provided
by Scheduled Commercial Banks to different sectors of the economy which
enables you to study trends in the flow of bank credit to small-scale industry,
medium and large industries, wholesale trade other than food production and
export credit. This would enable you to evaluate the implementation of the
policy regarding allocation of bank and institutional credit to the priority
sector, which includes small-scale industries. It also provides time series data
on deployment of bank credit to some 25 industrial categories and power. In
addition, it presents time series data on the health of SSI and non-SSI units –
(a) the number of SSI units that are sick, (b) the number that are weak, (c) the
number of non-SSI units that are sick, and (d) the amounts outstanding (loans)
from each of these categories of units.
The ASI provides, as you have seen, some data on financial aspects of
industries – fixed capital, working capital, invested capital, loans outstanding
and also the interest burden of industrial units (up to the 4-digit NIC code
level). From where and how have the industries raised capital needed by them?
We have looked at one source of capital or working capital, namely, bank
credit. Time series data on new capital issues and the kinds of
shares/instruments issued (ordinary, preference or rights share or debentures,
etc.,) and the composition of those contributing to capital (like promoters,
financial institution, insurance companies, government, underwriters and the
public) are also presented. Data on assistance sanctioned and disbursed by
financial institutions like Industrial Development Bank of India (IDBI),
Industrial Credit and Investment Corporation of India (ICICI), Small Industrial
Development Bank of India (SIDBI) and Life Insurance Corporation (LIC) of
India can also be obtained from the Handbook. Similarly, some data on the
financing of project costs of companies are available in the Handbook. All
these data of course relate to companies, which may include non-industrial
ventures too. The primary source of such data is the Ministry of Company
Affairs. The publication of the Securities Exchange Board of India (SEBI)
“Handbook of Statistics on the Indian Securities Market”(the latest one
published in 2012) provides annual and monthly time series data on industry-
wise classification of capital raised through the securities market – the
industrial sector activities by which the classification is being made are, cement
31
58 See the website http://labordoc.ilo.org/ .
and construction, chemical, electronics, engineering, entertainment, finance, Agricultural and
Industrial Data
food processing, health care, information technology, leather, metal mining,
packaging, paper and pulp, petrochemical, plastic, power, printing and rubber.
A reference to the two volumes of EIS, CMIE, one “Industry: Financial
Aggregates” and the other “Industry: Market Size and Shares” would be
rewarding.
Foreign direct investment (FDI) is another important source of capital finance

for industrial expansion. Availability of data on FDI is dealt with in Unit 22 on
Trade and Finance.
22.3.6 Contribution of Industrial Sector to GDP

The National Accounts Statistics (NAS), as we have seen in UNIT 18 on
National Income, Saving, Investment, furnishes information on the contribution
of the industrial sector to GDP. NAS presents a short time series of estimates of
(i) value of output and GDP of each two-digit NIC code level industry in the
registered and the unregistered sub-sector of the manufacturing sector, (ii)
value of output of major and minor minerals and GDP and NDP of the mining
& quarrying sector, and (iii) GDP and NDP of the sub-sectors electricity, gas
and water supply.
1) Name the agencies which generate the data on small scale industries.
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
2) How can you get the data on deployment of bank credit to SSI units?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
3) From which sources you can get the data on credit flowing to small scale
sector?
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
……………………………………………………………………………..
59
Data Base of Indian
Economy 22.4 LET US SUM UP
The major efforts to collect data on different aspects of the agricultural sector
are the quinquennial agricultural census, the quinquennial livestock census, the
cost of cultivation studies, annual estimates of crop production and the
integrated sample survey to estimate the production of major livestock
products. Most of the data flowing from these efforts relate to production,
inputs for production, land utilization patterns, intensity of land use, irrigation,
costs and pricing of agricultural produce, procurement and distribution,
subsidies and credit are found in the two publications “Agricultural Statistics
at a Glance” (ASG) and “Basic Animal Husbandry Statistics” (BAHS).
Statistics on the characteristics of land holdings of different size with reference
to the parameters referred to above are also available in ASG. More detailed
information in this regard, collected in the succcessive agricultural census are
available in the Ministry’s website. Similarly, detailed information relating to
sub-sectors – livestock, poultry and fishery as also agricultural implements,
collected in the livestock censuses, are available in the website of the
Ministry, though some information relating to the earlier census are available
in ASG and BAHS of relevant years. Data on credit flowing to the
agricultural sector are available in the RBI Handbook of Statistics on the
Indian Economy and NABARD’s Publications. Another source containing
comprehensive information on agriculture at the global level is the FAO
statistical yearbooks published by the UN Food and Agriculture Organisation.
The most detailed data available in the industrial sector relates to the factory
sector provided by ASI. The multi-variable data covers about 30 characteristics
– investment, inputs, output, gross and net value added, employment, factor
shares, net value added at 3 or 4-digit NIC code levels by States and Union
territories. Quick estimates of monthly production in selected industries and the
Index of Industrial Production are available within six weeks of the reference
month from the CSO and revised in the next two months. The State
Governments and Union Territory Administrations are also taking steps to
prepare and release monthly IIPs that are comparable to the national level IIP.
Some have already started releasing such estimates. The fourth census of
micro, small and medium enterprises (MSME) relating to 2006-07 is the
most recent and comprehensive source for data on production, employment and
exports of the small-scale sector. Reports from the NSSO on surveys of the
unorganized sector carried out once in five or six years provide useful data on
the characteristics of this sector. Enquiries covering the entire industrial sector
or the organized industrial sector such as the population census, NSSO
surveys on employment, the Economic Census and the Employment
Market Information Programme of the Ministry of Labour & Employment
provide data only on employment. The RBI handbook gives data on industrial
credit and financing of industrial capital. The CMIE (EIS) volumes on
industry, one on financial aggregates and the other on market size and
shares provide a fund of data on finance and market competition aspects of
industry. At the global level, the International Labour Organisation (ILO)
provides statistics on employment and the UNIDO provides statistics on total
factor productivity.
60
Agricultural and
22.5 EXERCISES Industrial Data
1) Which major efforts have been made to collect the data on different
aspects of agricultural sector?
2) Discuss the characteristics of data flowing from the agencies involved in
compilation of agriculture data.
3) “The detailed data available in the industrial sector relates to the factory
sector” explain.
4) Do you think that data available on unorganized sector is inadequate?
What suggestion would you like to make in this regard?

Amitabh Kundu & : Informal Sector in India – Perspectives &
Alakh Sharma (Ed.) Policies, Institute for Human Development &
(2001) Institute of Applied Manpower Research, IP
Estate, Ring Road, New Delhi.
____National Sample : Survey Reports of unorganized sector
Survey Office (NSSO) enterprise surveys NSSO, Ministry of
(Various issues) Statistics & Programme Implementation, Govt
of India, Sardar Patel Bhavan, New Delhi.
Department of : Agricultural Statistics At a Glance, Directorate
Agriculture & Coop of Economics & Statistics (DES), Ministry of
Agriculture, Govt. of India, Krishi Bhavan,
New Delhi. Cost of Cultivation in India, DES,
Ministry of Agriculture & Cooperation, New
Delhi.
Reports of the Commission on Agricultural
Costs and Prices (CACP), DES, Ministry of
Agriculture & Cooperation, New Delhi.
Indian Horticulture Data Base, Ministry of
Agriculture & Cooperation, New Delhi
___Ministry of Finance : Economic Survey, Ministry of Finance, North
Block, New Delhi.
___Reserve Bank of : RBI Handbook of Statistics on the Indian
India (various issues) Economy. The Reserve Bank of India, Mumbai
www.mospi.nic.in : Website of the Ministry of Statistics &
Programme Implementation. (even the letters
cso in google search will lead you to the (CSO
website).
eands.dacnet.nic.in : Directorate of Economics and Statistics,
Department of Agriculture and Cooperation,
Ministry of Agriculture, Government of India
www.wrmin.nic.in : Website of the Ministry of Water Resources.
www.rbi.org.in : Website of the Reserve Bank of India (RBI).
www.fao.org : UN Food and Agriculture Organisation (FAO)
website. 61
Data Base of Indian
Economy
www.ilo.org : UN International Labour Organisation (ILO)
website.
www.unido.org : UN Industrial Development Organisation
(UNIDO) website.

PROGRESS EXERCISES
1) The Directorate of Economic and Statistics in the Department of
Agriculture and Statistics and Animal Husbandry Statistics Division of the
Department of Animal Husbandry.
2) See foot note no. 2 and 4.
3) See Sub-section 22.2.3 and foot note no. 6.

2) See foot note no. 8.
3) Agricultural Statistics at a glance.
4) RBI Handbook of Statistics on Indian Economy 2005.

2) See foot note no. 18
3) A system that enables the data uses to access specific data relating to ASI.
4) Derivation of important technical ratios for using the principal
characteristics of factory sector.
5) Energy Statistics published by CSO.

1) Development commissioner for Small Scale Industries (DCSSI).
2) Directorate of Industries of State Government and Union Territory
Administrations.
3) SIDBI, LIC, Handbook of statistics on the Indian Security Market.
62
Trade and Finance
UNIT 23 TRADE AND FINANCE
Structure
23.0 Objectives
23.1 Introduction
23.2 Trade
23.2.1 Merchandise Trade
23.2.2 Services Trade
23.3 Finance
23.3.1 Public Finances
23.3.2 Currency, Coinage, Money and Banking
23.3.3 Financial Markets
23.4 Let Us Sum Up
23.5 Exercises
23.0 OBJECTIVES
• describe the kind of data available in India in the area of trade and finance;
• state the various sources of data on trade and finance;
• explain the reasons for divergence between merchandise trade
deficit/surplus data provided by DGCI&S and RBI’s BOP data; and
• discuss which agencies are involved in compilation of data on trade and
finance.
23.1 INTRODUCTION
Trade is the means of building up an enduring relationship between countries
and the means available to any country for accessing goods and services not
available locally for various reasons like the lack of technical know-how. It is
also the means of earning foreign exchange through exports so that such
foreign exchange could be utilized to finance essential imports and to seek the
much-needed technical know-how from outside the country for the
development of industrial and technical infrastructure to strengthen its
production capabilities. Trade pacts or agreements between countries or groups
of countries constitute one way of developing and expanding trade, as these
provide easier and tariff-free access to goods from member countries. While
efforts towards such an objective will be of help in expanding our trade,
globalization and the emergence of World Trade Organization (WTO) have
only sharpened the need to ensure efficiency in the production of goods and
services to compete in international markets to improve our share of world
merchandise trade and trade in services. Trade is also closely tied up with our
development objectives. Trade deficit or surplus, made up of deficit/surplus in
merchandise trade and trade in services, contributes to current account deficit 63
Data Base of Indian or surplus. Data on trade in merchandise and services would enable us to
Economy
appreciate the trends and structure of trade and identify areas of strength and
those with promise but need sustained attention.
While trade and finance have been closely bound up with each other ever since
the time money replaced barter as the means of exchange, finance is the lifeline
of all activities – economic, social and administrative activities. It flows from
the public as taxes to Government, as savings to banking and financial
institutions and as share capital or bonds or debentures to the entrepreneur. It
then gets used for a variety of developmental and non-developmental activities
through Government and other agencies and flows back to members of the
public as incomes in various ways, as factor incomes. It would therefore, be of
interest to know how funds get mobilized for various purposes and get used.
This Unit looks at the kind of data available that could enable us to analyse this
mobilization process and the flows of funds to different areas of activity. Let us
begin to discuss the data on trade.
23.2 TRADE
23.2.1 Merchandise Trade
The Directorate General of Commercial Intelligence and Statistics
(DGCI&S) collects and compiles statistics on imports and exports. It releases
these data at regular intervals through their publications and through CDs. It
prepares “Quick Estimates” on aggregate data of exports and imports and
principal commodities within two weeks of the reference month and releases
these in the monthly press release. It publishes
1) A monthly brochure “Foreign Trade Statistics of India (Principal

Commodities and Countries)” containing provisional data issued to meet
the urgent needs of the Ministry of Commerce, other government
organizations, Commodity Brands (CBs), Export Promotion Councils
(EPCs) and research organizations. It contains commodity-wise, country-
wise and port-wise foreign trade information;
2) “Monthly Statistics of Foreign Trade of India, Volume I (Exports

including re-exports) & Volume II (Imports) containing detailed data
on foreign trade at the 8-digit level codes of the ITS (HTS) (see below);
3) Quarterly publications :
a) Statistics of the Foreign Trade of India by Countries – VolI

(Exports including re-exports) and
b) Statistics of the Foreign Trade of India by Countries – Vol II
(Imports)
4) Annual publications:
a) Statistics of the Customs and Excise Revenue Collections of the
Indian Union,
b) Statistics of the Inland Coasting Trade Consignment of India,
c) Inter-State Movements/ Flows of Goods by Rail, River and Air,
d) Statistics of Foreign and coastal Cargo Movements of India,
e) Selected Statistics of the Foreign Trade of India;
64
5) The DGCI&S website (www.dgciskol.nic.in) has two parts, one consists Trade and Finance
of static pages and the dynamic pages. The first contains the history and
the activities of DGCI&S and summary data on principal commodities,
and countries and is updated regularly. The dynamic pages are mainly for
on-line data dissemination and provide access on free and payment basis.
This contains at least 24 months final foreign trade data at 8-digit
commodity and principal commodity level and updated regularly; and
6) The Priced Information Service System (PISS) provides information to

private parties, EPCs, CBs, Foreign Embassies etc., on payment basis @
Re. 1 /- per unit record of information. It does not give the whole basket of
8-digit commodity-country data for any particular period for reasons of
‘copyright’ provision of the DGCI&S. It however, provides aggregate and
detailed data to Centre for Monitoring Indian Economy (CMIE),
Mumbai for an efficient trade intelligence service.
7) It also has a web-based data dissemination system for online data

transaction through advance payment. The export and import trade data
are available countrywise and economic regionwise. The detailed data on
India’s foreign trade in merchandise are made available after completion
of the third month from a particular month. Here the commodities referred
to in these data stand for the ones specified in the ITC(HS) against 8-digit
codes which have been developed by the DGCI&S by sub-dividing the 6-
digit codes of Harmonised System as internationally standardised. If you
want to get data online, you can visit http://www.dgciskol.nic.in/new_
registration.asp and see the terms, conditions and procedures.
The DGCI&S data are also presented as time series data in the Reserve Bank
of India Handbook of Statistics on the Indian Economy. As mentioned
above the CMIE is another source – their volume on Foreign Trade and
BoP. The EPWRF (see its website – google search with EPWRF will enable
you to access it) also publishes time series data on foreign trade – it is one of
the 35 sets of special statistics on which it publishes long time series data along
with information on conceptual and methodological issues.
Foreign trade data published by DGCI&S relates to merchandise trade through

all recognized seaports, airports, land customs stations and inland container
depots located all over India. Data on exports include re-exports and relate to
the free on board (f.o.b.) values and imports relate to cost, insurance and freight
(C.I.F.) values. Exports and imports are based on a general system of
recording. According to this, exports relate to Indian merchandise and re-
export relates to foreign merchandise previously imported into India. Imports
relate to foreign merchandise, whether these are intended for consumption in
India, bonding or re-exportation.
The commodities comprising merchandise imports and exports are classified

according to a standard classification. The trade classification in vogue in India
between April, 1977 and March, 1987 was the Indian Trade Classification,
Revision 2 (ITC – Rev. 2) – one that was based on the Standard International
Trade Classification Revision 2 (SITC – Rev.2). A new system of commodity
classification, known as Indian Trade Classification (based on the Harmonized
Commodity Description and Coding System), or ITC(HS), has been adopted
since April, 1987. ITC (HS) is an extended version of the International
Classification System called “Harmonized Commodity Description and Coding
65
Data Base of Indian System” evolved by the World Customs Organization, Brussels1. These
Economy
changes in the trade classification of commodities mean that time series data on
export and import relating to some commodities may not be strictly
comparable. Another element of non-comparability of time series data arises
from changes in the definition of countries and/or groups. For example, data
for Russia prior to 1993-94 relate to erstwhile USSR, with the exception of
1992-93, the data for which relate to the Commonwealth of Independent States
(CIS) representing a group of 15 countries2. Similarly, Indian trade with
Germany relates to Federal Republic of Germany (FRG) till 1989-90 and to the
unified Germany thereafter. There are also changes in the membership of
groups like the European Union, and Oil and Petroleum Exporting Countries
(OPEC) from time to time.
What would we like to know about foreign trade? The volume of trade, that is,
the volume of exports and imports, the size of export earnings, the expenditure
on imports, the size of exports relative to imports, earning from exports
compared to expenses incurred on imports since exports earn foreign exchange
while imports imply outflow of foreign exchange. We should like to know
about the trends in these variables. Besides looking at the trends in the quantum
and value of imports and exports, it is important to analyse the growth in
foreign trade both in terms of value and volume, since both are subject to
changes over time. Exports and imports are made up of a large number of
commodities and fluctuations in the export and imports of individual
commodities contribute to overall fluctuations in the volume and value of
exports and imports. We, therefore, need a composite indicator of the trends
in trade. The index number of foreign trade of a country is a useful indicator of
the temporal fluctuations in exports and imports of the country in the term of
value, quantum and unit price and so on. Similarly, measures of the terms of
trade could be derived from such indices relating to imports and exports.
The index number of foreign trade is computed and presented as Unit Value
Index (UVI) and Quantum Index (QI). These are defined as follows:
UVI = [ ∑ ] / [∑ ] …………………………. (1)
QI = [ ∑ ] / [∑ ] …………………………. (2)
where P1 Is the unit value of an item in the current period and Qt is the quantity
of the same item in the current period, P0 and Q0 are the unit value and the
quantity respectively of the same item during the base period and Σ is the
summation over all commodities. These indices have been computed with
1979-80 as the base year.
Three types of Terms of Trade are computed from these indices:
1) Gross Terms of Trade (GTT) = 100 × [(QI of imports)/QI of exports)]
2) Net Terms of Trade (NTT) = 100 × [(UVI of exports)/(UVI of imports)]
3) Income Terms of Trade (ITT) = [NTT × QI of exports]/100
= [UVI(exports)× QI(exports)]/ [UVI (imports)].
The existing index numbers have the base year 1979-80. Changes in the ITC
since April, 1987 and the recasting of the basket of commodities to suit the new
1
This organization was previously known as Customs Cooperation Council.
2
These were Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyz republic,
66 Latvia, Lithuania, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine and Uzbekistan.
classification system, the base year 1979-80 has become too old to serve the Trade and Finance
purpose of temporal comparability and the new ITC adopted in April, 1987.
The DGCI&S, therefore, decided to construct new index numbers with the
base year 1998-99. It also decided to bring out global monthly indices and for
selected countries. The preparation of these index numbers is in progress.
What kinds of data are available on foreign trade? Time series on the following
data are published in the RBI Handbook of Statistics of the Indian Economy
2013-14, under the sub-heading “Trade and Balance of Payments”:
a) India’s foreign trade (in US $ and Indian Rupees) from 1975-76 to 2013-14,
The exports and imports have been shown as total and in two groups,
namely, oil and non-oil. The “Trade Balance”, i.e., exports – imports is
also given for each of these three groups. A negative figure meaning exports
are less than imports. You can observe from this Table that India had a positive
non-oil trade balance, albeit sporadically, till 2003-04. However, after that,
trade balance for both oil and non-oil has become negative.
b) Exports (in US$ and Indian Rupees) of principal commodities, from 1998-
99 to 2013-143;
c) Imports (in US$ and Indian Rupees) of Principal Commodities4, from
1998-99 to 2013-14;
d) Exports (in US$ and Indian Rupees) of selected commodities to principal
countries;
e) Direction of Foreign Trade (in US$ and Indian Rupees) showing exports
and imports for each year by trade areas, groups of countries and
countries; and within each group of countries or trade area, data are
presented only for selected countries5;
f) Year-wise indices, both UVI and QI, for imports and exports from 1976-
77 to 2007-08 (base 1978-79 = 100) and 1999-2000 to 2013-14 (base
1999-2000 = 100) the three terms of trade measures; GTT, NTT and ITT;
3
Commodities were classified into 4 groups (including total), namely, primary products,
manufactured goods, petroleum products and others; sub-divided further in into 11 sub-
groups and 44 items: A. Agriculture and allied Products (15 items), 2. Ores and Minerals (3
items), 3. Leather and manufactures, 4. Chemicals and related products (4 items), 5.
Engineering goods (6 items), 6. Textile and textile products (11 items), 7. Gems and jewelry,
8. Handicrafts, excluding handmade carpet; 9. Other Manufactured Goods; 10. Petroleum
products; and 11. Others.
4
The classification of imports consisted of three broad groups, I Bulk Imports, II Non-bulk
Imports and III Total Imports; Bulk Imports were further divided into three groups, namely,
A. Petroleum, Crude and Products; B. Bulk Bulk Consumption Goods (consisting of 4 sub-
items); C. Other Bulk Items (9 sub-items); Non-bulk Import were A. Capital Goods (8 sub-
items), mainly export related items ( 4 sub-items) and Others (9 sub-items).
5
The trade areas for which data are presented are: I OECD countries, II OPEC, III Eastern
Europe, IV Developing Countries, V Others (Unspecified). Each of these trade areas is
further divided into country-group and within each group figures are given only for some
selected countries. Trade area I is divided into A. European Union (within which data for 6
selected countries are given). B. North America (2 selected countries). C. Asia & Oceania (2
selected countries) and D. Other OECD Countries (1 selected country). II and III have no
further grouping of countries within them but figures for 6 selected countries under II and for
2 selected countries under III are presented. Trade area IV is divided into A. Asia, B. Africa
and C. Latin America Countries. A gets further grouped into (a) SAARC (figures for all
countries are given under this), (b) Other Asian Developing Countries (figures for 6 selected
countries are given). Figures for 7 selected countries under Africa are given and for Latin
America no further break up by countries are given. 67
Data Base of Indian g) Index numbers of Exports – QI and UVI for 9 commodity groups6 from
Economy
2005-06 onwards (base 1999-2000 = 100).
h) Index numbers of Imports - QI and UVI for 9 commodity groups7 from
2005-06 onwards (base 1999-2000 = 100) Similar type of data on imports.
i) India’s overall Balance of Payment (in US $ and Indian Rupees) under: A
current account, B capital account, C. errors and omissions, D. overall
balance and E. Monetary movements from 2008-09 onwards.
j) Invisibles by category of transactions (in US $ and Indian Rupees) from
1994-95 onwards.
k) Exchange rate of Indian Rupee vis-à-vis US $, UK £, DM/€ and Japanese
¥ from 1975 onwards (calendar year – annual average, financial year –
annual average and end year rates).
l) Real Effective Exchange Rate (REER) and Net Effective Exchange Rate
(NEER) of Indian Rupee 36-country bilateral weights) with base 2004-05
from calendar year 2005 onwards.
m) Indices of REER and NEER.
n) External assistance in the form of loans and grants from 1985-86 onwards
(in US $ and Indian Rupees), authorization, utilization and debt service
payments.
o) NRI deposits outstanding (in US $ and Indian Rupees) from 1996 onwards
in different types of accounts like NR (E)RA, FCNR(A), FCNR(B),
NR(NR)RD, NRO, FC (B&O)D, FC(O)N and total. It has increased from
US $17.446 billion in 1996 to US $103.844 in 2014.
p) Foreign investment inflows (in US $ and Indian Rupees) namely, gross
investments, repatriation/ disinvestment, FDI by India, net foreign direct
investment, net portfolio investment from 2000-01 onwards.
q) Foreign exchange reserves (in US$ and Indian Rupees) from 1956-57 onwards.
r) India’s external debt (in US $ and Indian Rupees) from 1996 onwards8,
which, inter alia, providesconcessional debt as percentage of total debt,
short term debt as % to total debt, debt stock-GDP ratio (%) and debt
service ratio (%).
6
The commodity groups are: I Food and live animals (6 sub-groups), II. Beverages and
Tobacco, III. Crude materials, inedible, except fuel (4 sub-groups), IV. Mineral fuels,
lubricants and related materials (3 sub-groups), V. Animal and vegetable oil, fats and
waxes, VI. Chemicals and related products (6 sub-groups), VII. Manufactured goods
classified chiefly by material (5 sub-groups), VIII. Machinery and transport equipment (9
sub-groups) and IX. Misc. manufactured articles (4 sub-groups). In a few cases, indices for
specific commodities under a sub-group are also given.
7
The commodity groups are: I Food and live animals (5 sub-groups), II. Beverages and
Tobacco (2 sub-groups), III. Crude materials, inedible, except fuel (7 sub-groups), IV.
Mineral fuels, lubricants and related materials (3 sub-groups), V. Animal and vegetable oil,
fats and waxes, VI. Chemicals and related products (8 sub-groups), VII. Manufactured
goods classified chiefly by material (7 sub-groups), VIII. Machinery and transport
equipment (9 sub-groups) and IX. Misc. manufactured articles (3 sub-groups). In a few
cases, indices for specific commodities under a sub-group are also given.
8
The debts are classified into I. Multilateral, II. Bilateral, III. International Monetary Fund,
IV. Trade credit, V. Commercial borrowing, VI. NRI & FC (B&O) deposits, VII. Rupee
debt, VIII. Total long-term debt, IX. Short-term debt and X. Gross total debt. Most of the
groups have sub-groups A. Government borrowing, and B. Non-Government borrowing,
68 further disaggregated into concessional and non-concessional borrowing.
Some of these data are from the DGCI&S and the rest are from the Balance of Trade and Finance
Payments (BoP) data of the RBI. The BoP data reported by RBI show the
value of merchandise imports on the debit side and that of exports on the credit
side. It also shows trade balance – a trade deficit or a trade surplus – depending
upon whether the difference ‘export - imports’ is negative or positive. These
are all shown in the balance payment format as part of current account, which
also shows another entity ‘invisibles’. However, there is a divergence in trade
deficit/surplus in merchandise trade shown by DGCI&S data and that shown
by RBI’s BoP data. This discrepancy between the two sources also affects
data on current account deficit (CAD) or surplus (CDS), since current account
deficit/surplus is the total of trade deficit/surplus and net invisibles (inflow of
invisibles net of outflow in the category ‘invisibles’9) For example, the
Economic Advisory Council to the Prime Minister noted, in its Report on
Balance of Payments (BoP) submitted to the Prime Minister recently, that the
divergence between the two sources of data on trade was growing10. It noted
that trade deficit is projected for the year 2005-06 at 5.2 per cent of GDP on the
basis of trade data from DGCI&S, while it is 7.7 per cent according to RBI’s
BoP data – a difference of the size of 2.5 per cent of GDP. CAD based on trade
data of DGCI&S is only 0.3 per cent of GDP while it is 2.9 per cent of GDP if
the estimate of CAD is based on trade data from BoP. The reasons for the
divergence in the data between the two sources are, as noted by the Council:
• DGCI&S tracks physical imports and exports while BoP data tracks
payment transactions relating to merchandise trade;
• DGCI&S data do not capture Government imports, which are exempted

from customs duty. Defence imports fall into this category; and
• DGCI&S data do not capture imports that do not cross customs boundary
(for example, oil rigs and some aircrafts) while they are still paid for and
get captured in BoP data.
23.2.2 Services Trade

Besides export and import of merchandise, a number of services, like
transportation services, travel services, software, Information technology-
Enabled Services (ITES), business services and professional services are
exported and imported. These are captured by “non-factor services” included in
the entry “Invisibles” in the Tables on India’s Overall BoP and on Key
Components of India’s BoP in the RBI Handbook. The handbook has also a
table (Table 140 in RBI’s Handbook of 2013-14) that gives the distribution of
‘invisibles’ by transactions – credit, debit and net, under the Current Account,
giving separately for a) Services: Travel, Transportation, Insurance, G.n.i.e,
Miscellaneous (software services, business services, financial services and
communication services); b) Transfers: Official, Private; c) Income: investment
income, compensation of employees.

1) What kind of Trade Data is compiled by DGCI&S?
………………………………………………………………………………
………………………………………………………………………………
9
See under the section on Finance later in this Unit.
10
News report in the Hindu dated February, 23, 2006 Chennai edition – Business page. 69
Data Base of Indian 2) What is the existing base year for constructing unit value Index and
Economy
quantum index by DGCI&S?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) What are the reasons for divergence between two sources of trade data?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) List the various transactions covered in the ‘net invisibles’.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
23.3 FINANCE
The finance sector consists of public finances, the central bank or the Reserve
Bank of India, the scheduled banks, urban and rural cooperative banks and
related institutions. The financial market consists of the stock exchanges
dealing with scripts like shares, bonds and other debt instrument, the primary
and secondary markets, the foreign exchange market, the treasury bills market
and the securities market where financial institutions, mutual funds, foreign
institutional investors, market intermediaries, the market regulator the
Securities Exchange Board of India (SEBI), the banking sector and the RBI all
play important roles. It also has the insurance (life and general) and pension
funds as well as their respective regulators, i.e., the Insurance Regulatory and
Development Authority (IRDA) and the Pension Funds Regulatory and
Development Authority (PFRDA). It has the holding companies, which invest
in various subsidiaries and controls its operations through its stocks11. There is
also the unorganized sector made up of financial operators like individual
money-lenders and pawn shops, insurance agents, unregistered stock brokers/
11
One example of a Holding Company is Tata Sons, whose subsidiaries include TISCO, TCS,
70 Tata Motors, Tata Housing, etc.
sub-brokers, etc. In the System of National Accounts, 2008 (SNA 2008)12, the Trade and Finance
financial sector has been sub-divided into nine sub-sectors, depending on its
type of business.
23.3.1 Public Finances

What would we like to know about public finances? We would like to know
how they are managed. What are the sources of such finances and how and on
what are they spent? Does the Government restrict its expenditure within its
means or does it spend beyond the resources available to it? Does it, in the
process, borrow heavily to finance its expenditure? The Budget documents of
the Central and State Governments, the pre-Budget Economic Survey of the
ministry of finance and the Reserve Bank of India (RBI) Handbook of
Statistics on the Indian Economy 2005 and the Monthly Bulletin of the RBI
provide a variety of data on public finances. The National Accounts Statistics
of the CSO, MoSPI also provides the output, GVA, gross savings, etc. of the
financial sector and its various sub-sectors. EPWRF website will also help to
access data in a time series format on public finances.The Economic Survey,
for instance, gives an overall summary of the budgetary transactions of the
Central and State governments and Union Territory Administrators. This
includes the internal and extra-budgetary resources of the public sector
undertakings for their plans. It indicates the total outlay, the current revenues,
the gap between the two, the manner in which the gap is financed by net
internal and external capital receipts and finally, the overall budgetary deficit.
It gives the break-up of the outlay into developmental and non-developmental
outlays and the components of the latter, the components of current revenues-
tax revenue and non-tax revenue – and sub-components of these and the
components of internal and external capital receipts. The RBI Handbook
2013-14 presents the following time series data in respect of public finances.
a) Central Government Finances
i) Key deficit indicators13 of the Central Government – gross fiscal

deficit, gross primary deficit, net primary deficit, revenue deficit,
primary revenue deficit, drawdown of cash balances, net RBI credit
from 1975-76 onwards;
ii) Major components of Central Govt. Receipts – tax revenue (direct and
indirect taxes and their components) non-tax revenue (one of its
important components is interest receipts) and capital receipts;
12
Discussed in Unit 20 in the discussion on National Accounts Statistics.
13
GFD = total expenditure including loans (net of recovery) – revenue receipts (including
external grants and non-debt capital receipts); NFD = GFD – net lending of the Central
Govt; GPD = GFD – interest payment; NPD = net interest payments; RD = revenue
receipts – revenue expenditure; PRD = RD – interest payment; BD (Conventional Deficit)
= all expenditure (revenue and capital) – all receipts (revenue and capital); with the
discontinuation of ad hoc treasury bills and 91-day treasury bills, the concept of
conventional budget deficit has lost its relevance since 1/4/97. The figures shown against
BD from 1997-98, therefore, represent draw down of cash balances from RBI; Net RBI
Credit to Govt. = the sum of variations in the RBI’s holdings of (i) Central Govt. dated
securities, (ii) treasury bills, (iii) Rupee coins, and (iv) loans and advances from RBI to the
Central Govt. since 1/4/97 adjusted for changes in the Central Govt.’s cash balances with
the RBI in the case of the Centre. 71
Data Base of Indian iii) Major14 heads of capital receipts of the Central Government – market
Economy
borrowings, small savings, provident funds15, special deposits, recoveries
of loans, disinvestments receipts16, external loans (net);
iv) Major heads of Central Govt. expenditure – revenue expenditure and
its important components (defence, interest payments and subsidies),
capital expenditure including loans and advances and defence
expenditure; also the breakup of expenditure into developmental and
non-developmental heads and the shares of economic services and
social services in developmental expenditure;
v) Centre’s gross fiscal deficit and its financing – GFD receipts, GFD
expenditure, Gross Fiscal Deficit, financing of GFD through external
financing and internal financing (market borrowings, other borrowings,
draw down of cash balances);
vi) Gross capital formation from budgetary resources of the Central

Government – fixed assets, work stores, increase in stocks of foodgrains
& fertilisers, gross financial assistance for capital formation to State
Governments, Non-Departmental Commercial Undertakings (NDCUs)17
and others;
vii) Public sector Plan outlay, its sectoral profile and the manner in which
it is financed – from sources like own resources, domestic market
borrowings and net capital inflow from abroad;
viii) Financing of Public Sector Plan – balance from current revenue,

contribution of public enterprises, borrowings (including long and
medium term borrowings), small savings, deficit financing, net capital
inflow from abroad, central assistance and others.
b) Finances of the State Governments
i) Key deficit indicators of the State Governments – fiscal deficit, gross
primary deficit, revenue deficit, primary revenue deficit, overall deficit,
net RBI credit to States (annual variation) from 1975-76 onwards;
ii) Pattern of receipts of the State Government – total Revenue receipts:
tax receipts (like sales tax and State excise duties), the share of Central
taxes like income tax and union excise duties; and non-tax receipts like
interest receipts, grants from the Centre, total capital receipts;
iii) Pattern of major capital receipts of the State Government – loans from
centre, recovery of loans and advances, market loans, State provident
fund (net), special securities issued to NSSF;
iv) Expenditure pattern – revenue and capital expenditure; capital outlay,
loans and advances by State Government, developmental expenditure
and the shares of economic services and social services in it and non-
developmental expenditure and the shares of interest payments,
administrative services and pension and miscellaneous general
services;
14
Adjusted for changes in classification effected in 1974-75 and 1987-88.
15
Only Govt. provident funds, as the Public Provident Fund (PPF) is part of small savings
since 1998-99.
16
These are not to be treated as budgetary receipts as these are to be credited to a separate
fund.
17
72 NDCU means the Central and State Public Sector Units
v) States’ Gross Fiscal Deficit and its financing – loans from Central Trade and Finance
Government, market borrowings, special securities issued to NSSF and
others,
vi) Outstanding liabilities of the State Governments.
c) Combined Finances of Central and State Governments
i) Combined deficit;
ii) Receipts and Disbursements;
iii) Direct and Indirect tax revenues;
iv) Developmental and non-developmental expenditure;
v) Market borrowings;
vi) Interest rates on dated securities (Range and weighted averages) of
Central and State Government;
vii) Outstanding liabilities;
viii) Ownership of Central and State Govt Securities – 11 categories like
RBI, scheduled commercial banks, cooperative banks, primary
dealers, insurance companies, financial institutions, mutual funds,
provident funds, corporates, foreign institutional investors and others;
d) Transactions with the Rest of the World
We have enumerated the kind of data available on Central and State
Government finances. But these relate mainly to domestic finances and
transactions. What about our transactions with the rest of the world? We
have looked at one of these, namely, trade in merchandise and services in
the section on trade. There are a number of other areas in which India
interacts with the rest of the world. Foreign exchange flows into the country
as a result of exports from India, external assistance/aid/loans/borrowings,
returns from Indian investments abroad, remittance and deposits from Non-
Resident Indians (NRI) and foreign investment (foreign direct investment –
FDI – and portfolio investment) in India. Foreign exchange reserves are
used up for purposes like financing imports, retiring foreign debts and
investment abroad. What is the net result of these transactions on the
foreign exchange reserves? What are the trends in these flows and their
components? What is the size of the current account imbalance relative to
GDP and its composition? If it is a deficit is it investment-driven? What is
the size of foreign exchange reserves relative to macro-aggregates like
GDP, the size of imports and the size of the short-term external debt18?
18
One can measure the size of the forex reserves as equivalent to so many weeks or months of
imports. Alternatively, one can relate it to financial stability. The Guidotti-Greenspan Rule
sets a standard for this. According to this rule, reserves should equal one year’s short-term
debt. Yet another way of examining the size of the reserves is to look at it is terms of its
opportunity cost. The reserves earn a zero real return measured in domestic terms. If on the
other hand, these are invested either domestically in infrastructure or in a fully diversified
long-term way in global markets, substantial incremental benefits would accrue to the
domestic economy. [see the Hindu, April 3, 2006, page 17 “Intriguing Pattern of Global
Capital Flows” excepts from the L.K. Jha Memorial Lecture delivered recently
(March 24, 2006) in Mumbai by Prof. Lawrence Summers. President of Harvard
University and former Secretary of the Treasury in the Clinton Administration, at the
invitation of the RBI.] The full text of the lecture may also be published in the RBI
Monthly Bulletin. The lecture discusses the implications of the rising forex reserves of the
developing countries and the opportunities and challenges that these present. 73
Data Base of Indian The RBI Handbook 2013-14 gives time series data (which are often in
Economy
US$ as well as in Indian Rupees) on a number of these parameters:
i) India’s overall BoP showing current account and capital account and key
components of these like trade balance, invisible19, types of foreign
investment, net external assistance, net commercial borrowing, rupee debt
service and net NRI deposits. It also shows monetary movements in terms
of increase/decrease in forex/reserves, IMF and SDR allocation,
ii) Exchange rate of Indian Rupee vis-à-vis US $, UK £, DM/€ and Japanese
¥ from 1975 onwards (calendar year – annual average, financial year –
annual average and end year rates),
iii) Real Effective Exchange Rate (REER) and Nominal Effective Exchange
Rate (NEER) of Indian Rupee based on 36 country bilateral weights with
base 2004-05 from calendar year 2005 onwards,
iv) Indices of REER and NEER,
v) External assistance in the form of loans and grants from 1985-86 onwards
(in US $ and Indian Rupees), authorization, utilization and debt service
payments,
vi) NRI deposits outstanding (in US $ and Indian Rupees) from 1996
onwards in different types of accounts like NR (E)RA, FCNR(A),
FCNR(B), NR(NR)RD, NRO, FC (B&O)D, FC(O)N and total. It has
increased from US $17.446 billion in 1996 to US $103.844 in 2014,
vii) Foreign investment inflows (in US $ and Indian Rupees) namely, gross
investments, repatriation/ disinvestment, FDI by India, net foreign direct
investment, net portfolio investment from 2000-01 onwards,
viii) Foreign exchange reserves (in US $ and Indian Rupees) from 1956-57
onwards,
ix) India’s external debt (in US $ and Indian Rupees) from 1996 onwards20,
which, inter alia, provides concessional debt as percentage of total debt,
short term debt as % to total debt, debt stock-GDP ratio (%) and debt
service ratio (%).
FDI has assumed importance in the context of the country’s efforts to meet the
increasing need for investment capital for its expanding economy. Data on FDI
are important. These are available from August, 1991 onwards. FDI data
collected by RBI and the Department of Industrial Policy & Promotion
(DIPP) in the Ministry of Commerce & Industry before 2000 related only to
equity capital. The coverage of the data has since then has been expanded so
that these are in accordance with the best international practices. FDI data
collected by RBI now cover (a) equity capital, (b) reinvested earnings (retained
earnings of FDI companies); and (c) other capital (inter-corporate debt
transaction between related entities). FDI data from 2000-01 onwards are,
therefore, not comparable with data of earlier years. The Department of
19
The component transactions that make up ‘invisible’ have been enumerated in the section
on Trade – Services.
20
The debts are classified into I. Multilateral, II. Bilateral, III. International Monetary Fund,
IV. Trade credit, V. Commercial borrowing, VI. NRI & FC (B&O) deposits, VII. Rupee
debt, VIII. Total long-term debt, IX. Short-term debt and X. Gross total debt. Most of the
groups have sub-groups A. Government borrowing, and B. Non-Government borrowing,
74 further disaggregated into concessional and non-concessional borrowing.
Industrial Policy and Promotion of the Ministry of Commerce and Industries Trade and Finance
publishes monthly data on FDI inflow in India21. It presents time series data on
the break-up of FDI flows by
1) Cumulative FDI inflows (equity inflows + reinvested earnings + other

capital) from 2000 – 2014,
2) Cumulative amount of FDI equity inflows (excluding amount remitted
through RBI’s + NRI schemes),
3) FDI inflows during the financial year till that month,
4) FDI equity inflows (month-wise) during that financial year and that
calendar year,
5) Share of top investing countries FDI equity inflows,
6) Sectors attracting highest FDI equity inflows, showing 10 different areas,
like service sector, construction development, telecommunication,
computer software and hardware, drugs and pharmaceuticals, automobile
industry, chemicals (other than fertilisers), power, metallurgical industries,
hotel and tourism,
7) Statement of RBI’s regional offices (with State covered) received FDI
equity inflows,
8) Country-wise FDI equity inflows from April 2000 onwards (from 140
countries, FIIs, NRI),
9) Sector-wise FDI equity inflows from April 2000 onwards (showing 63
industrial sectors).
The RBI handbook 2013-14 and the SEBI Handbook of Statistics of Indian
Securities Market22 present time series on FII. The SEBI handbook, gives
i) Foreign Investment Inflows,

ii) Trends in FII to portfolio investment,
iii) Trends in resource mobilization by mutual funds,
iv) International Securities Market (market capitalization, no. of listed
companies, value of shares traded, no. of trading days,
v) International Fixed Income Market (no. of bonds listed, value. of bonds
listed, value of bonds traded),
vi) International Derivatives market (stock options, stock futures, stock index
options, stock index futures).
FDI data can also be accessed in the RBI website www.rbi.org.in.
21
You can visit the site
http://dipp.gov.in/English/Publications/FDI_Statistics/FDI_Statistics.aspx#MainContent to
get the monthly FDI inflow.
22
http://www.sebi.gov.in/sebiweb/home/list/4/32/0/0/Handbook%20of%20Statistics 75
Economy
1) Name the documents that provide different kinds of data on public finances.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) What is the difference between Gross Fiscal Deficit (GFD) and Net Fiscal
Deficit (NFD)?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Identify the transactions other than trade in merchandise that India has
with the rest of world.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
23.3.2 Currency, Coinage, Money and Banking

Economic transactions need a medium of exchange. We have come a long way
from the days of barter and come to the use of money and equivalent financial
instruments as the medium of exchange. Banks function as important financial
intermediaries not only in this process but also in matters of resource
mobilization and the deployment of such resources. The central bank of the
country (the RBI in India) regulates the functioning of banking system. In
addition, it issues currency notes and takes steps to regulate the money supply
in the economy in order to achieve the objectives of ensuring adequate credit to
development activities and at the same time to maintain stability in prices. We
should, therefore, be interested in data on money supply or the stock of money
and its structure and the factors that bring about changes in these, the kind of
aggregate that need monitoring, the transactions in the banking system in
pursuance of the nation’s development objectives, the flow of credit to
different activities, indicators of the health and efficiency of banks which are
the custodians of the savings of the public. We should also be interested in data
on prices as price level affects the purchasing power of money and indices of
76
prices appropriate for the purpose/group in question – consumer prices for Trade and Finance
producers and different groups of consumers.
Most of these of data are compiled by the RBI on the basis of its records and
those of NABARD and returns that it receives from banks and can be found in
RBI Bulletins and RBI handbook. The wholesale Price Index (WPI) is
compiled by the Economic Advisor’s Officer in the Ministry of Industry, the
Consumer Price Index for Industrial Workers (CPI – IW) and CPI for
Agricultural Labour (CPI - AL) by the Labour Bureau, Shimla, Ministry of
Labour and CPI for rural and urban (CPI – Rural, CPI-Urban and CPI
Rural+Urban) by the CSO. All these are published by the agencies concerned
and also presented in the RBI and CSO publications mentioned above. The
Consumer Price Index released on 12th of each month by the CSO also
provides retail prices of selected commodities/services, separately for the rural
and urban areas in India. Two other reports of the Reserve Bank of India
published every year – the Report on Currency and Finance and the Report
on Trends in Banking provide a wealth of information of use to analyse. The
RBI also maintains an online database in searchable format, where one can
select the parameters for which one need to use the data. You can also right
click on the selected page, copy the data necessary for your analysis and paste
it on a blank excel workbook to carry out your own analysis 23. The EPWRF
website also provides data on Banking, Money and Finance.
The RBI Handbook 2013-14 presents time series data on
i) Liabilities and assets of the RBI, 1980 onwards. The liabilities are deposits
from Central Government, State Governments, Scheduled Commercial
Banks, Scheduled State Co-operative Banks, Non-Scheduled Co-operative
Banks, other Banks and others. The assets are notes and coins, balances
held abroad, loans and advances to Central and State Govts. banks and
other agencies, bills purchased and discounted, investments and other
assets;
ii) Components of money stock, namely, Reserve Money (M0) made up of

currency in circulation, other deposits with RBI and bankers’ deposits with
RBI, currency with the public (= currency in circulation – cash with bank),
Narrow Money (M1) consisting of currency with the public, other deposits
with RBI and demand deposits and Broad Money (M3) comprising Narrow
Money and time deposits;
iii) Sources of money stock consisting of net RBI credit to Central Government,
net RBI credit to State Government, other banks’ investments in Government
securities, RBI credit to commercial sector, Other banks’ credit to
commercial sector, net foreign exchange assets of the RBI, net foreign
exchange assets of other banks, Government’s currency liabilities to
public, net non-monetary liabilities of RBI, net non-monetary liabilities of
other banks, RBI’s gross claims on banks;
iv) Average monetary aggregates, like currency with the public, demand
deposits, time deposits, other deposits with the RBI, reserve money,
narrow money, broad money, net bank credit to Government, bank credit
to governmental sector, net foreign exchange assets of the banking sector,
23
See the link http://dbie.rbi.org.in/DBIE/dbie.rbi?site=home 77
Data Base of Indian Government’s currency liabilities to the public, banking sector’s net non-
Economy
monetary liabilities;
v) Major monetary policy rates and reserve requirements: bank rate, LAF
(REPO, reverse REPO, and MSF) rates, CRR and SLR;
vi) Monthly data series on the detailed composition of each of the components
of the money stock (the “C” components) and the composition of
individual sources of change in the money stock; also defining new
monetary aggregates24 NM2 and NM3, these being respectively equal to
“M1 + short-term time deposits” and “[NM2+ long-term deposits +
call/term funding from Financial Institutions (FIs)] = [domestic credit +
Government’s currency liability to the public + net forex assets of the
banking sector – capital account – other items (net)];
vii) Monthly data series on Liquidity Aggregates25, namely, L1= NM3 + Postal
deposits26; L2 = L1 + liabilities27 of FIs; L3 = L2 + public deposits with
NBFCs28 (L3 is compiled on a quarterly basis);
viii) Monthly average price of gold and silver in domestic (Mumbai) and
foreign markets;
ix) Selected Aggregates of Scheduled Commercial Banks (SCBs) like
outstanding demand and time deposits, investment in Govt. and other
securities, bank credit (food and non-food), cash in hand and balance with
the RBI.
x) Deployment of non-food credit to priority sector and its sub-sectors (like
agriculture, small scale industries), industry and its groups, whole sale
trade and export credit; and short and long term direct and indirect
institutional credit to agriculture and allied activities and to farmers by size
of holdings;
xi) Consolidated balance sheets of SCBs; Gross and net Non-Performing
Assets (NPAs) of SCBs by bank groups29; Distribution of SCBs and
different sub-group of SCBs by Capital to Risk-weighted Assets Ratio
(CRAR);30 and
xii) Important banking indicators of Regional Rural Banks (RRBs), State
Cooperative Banks, Primary Agricultural Coop. Societies (PACS), State
Coop. Agricultural and Rural Development Banks and Primary Coop.
(A&RD) Banks;
24
See footnote 19.
25
The methodology for compiling liquidity aggregates is available in the “New Monetary and
Liquidity Aggregates” in the RBI Bulletin of November, 2000. The acronyms NM2and NM3
are used to distinguish the new monetary aggregates from the existing monetary aggregates.
26
Post office SB deposits + PO time deposits + PO recurring deposits + other deposits + PO
Cumulative Time Deposits.
27
Term money borrowings + CDs + term deposits.
28
Non-Banking Financial Companies. Estimates of public deposits are generated on the basis
of a sample study of more than 1000 NBFCs with public deposits of Rs. 20 crores or more.
29
1) public sector banks, 2) old private sector banks, 3) new private sector banks, and 4)
foreign banks in India.
30
The detailed instructions issued by the RBI on CAR (Capital Adequacy Ratio) or CRAR
can be seen at http://rbi.org.in/scripts/BS_ViewMasCirculardetails.aspx?Id=8133&Mode=0.
SCBs had to comply with a minimum CRAR of 8 per cent up to the end of March, 1999 and
9 per cent from the end of March, 2000. The sub-group 1 in the footnote 25 is split into two
78 sub-groups – SBI group and nationalized banks.
Trade and Finance
23.3.2 Financial Markets
What would we like to know about the financial market and its functioning?
We should like to know about the ways in which financial resources can be
accessed and at what cost. What are the prevailing interest rates payable for
funds to meet short-term or long-term requirements? How do new ventures
access the large amount of resources that are need for the new ventures? How
do term lending institutions access funds required for their operations? What
are the sources of funds?
The RBI, which regulates banking operations and the operations of the NBFCs
and FIs and the Securities Exchange Board of India (SEBI), which regulates
the capital market, and the Department of Company Affairs that administers
the Companies Act, are the major sources of data on financial markets. The
RBI Handbook of Statistics of the India Economy – 2013-1431 and the
Handbook of statistics on the Indian Securities Market - 201432 published
by SEBI contain comprehensive data on the financial market. The two together
provide annual time series data on several aspects of the financial market:
i) The structure of interest rates – call/notice money rates, commercial bank

rate, lending rates of banks, prime lending rates (PLR) of term lending
institutions like IDBI, dividend and yields of the units of UTI, annual
gross redemption yields of Govt. Securities and average annual price and
yield rate of Central Govt. securities (SGL transactions);
ii) Financial assistance sanctioned and disbursed and financing of project

cost of companies by FIs, loans sanctioned by HDFC and NABARD, and
refinancing operations of National Housing Bank (NHB);
iii) Aggregate deposits of NBFCs and Non-Banking Non-Financial Companies

(NBNFCs)33 ;
iv) Taxable and tax-free bonds issued by public sector undertakings – both
public issue of bonds and privately placed bonds;
v) Resource mobilization in the Private Placement Market – financial and

non-financial institutions in the public and private sector;
vi) Net resources mobilized by mutual funds (MFs) – MFs sponsored by

banks, financial institutions (FIs), UTI and the private sector;
vii) New capital issues (number and amount mobilized) by non-govt. public
Ltd. Companies;
31
Also available on the RBI website http://www.rbi.org.in . In fact, the entire Handbook is
available on the website. The web version can be downloaded from the website. For some
of the financial market data longer time series are provided in the CD ROM and web
versions. The CD ROM now incorporates intelligent search features that allows for searches
across tables, enabling users to select one or more of any data series for any selected time
period from any table in the handbook in a user-friendly manner. The data can be
downloaded in the form of a user-defined spreadsheet table that can be read by most
standard econometric software.
32
Copies can be had from Research Deptt., SEBI, World Trade Centre I, 29th floor, Cuffe
Parade, Mumbai-400 005. This is also available on the website of SEBI –
www.sebi.gov.inand can be downloaded from the website.
33
After the new regulatory framework for NBFCs came into force in 1998, the NBFCs and
Residuary NBCs. 79
Data Base of Indian viii) Absorption of private capital issues – the no. of issuing companies, the
Economy
number of shares and amount subscribed by promoters etc., and Govt.,
FLs etc., and the number of shares and amount subscribed by public,
other than underwriters and other groups;
ix) Investments of LIC by sector and instrument and of UTI by instrument;
x) Assets and liabilities of institutions like IDBI, NABARD, EXIM Bank,

NHB and SIDBI;
xi) Annual averages share price indices – BSE SENSEX (base 1978-79
=100), BSE National (base 1983-84 = 100) and RBI Index (base 1980-81
= 100) and Market Capitalisation34;
xii) Market intermediaries like stock exchanges (cash and derivatives

market), brokers, corporate brokers, sub-brokers, custodians, FIIs,
depositories, merchant bankers, bankers to issues and underwriters
registered with SEBI; and registered brokers by stock exchanges and by
ownership categories – proprietary, partnership and corporate; (SEBI);
xiii) Long-term capital raised during 1957-90 (pre-reform period) through

shares, debt and loans; (SEBI);
xiv) Annual and monthly data series on resources raised by the corporate
sector through (i) equity issues and (ii) debt issues (public issues and
private placement) and the share of private placement in total debt and
total resource mobilization and the share of debt in total resource
mobilization; (SEBI);
xv) Pattern of funding for non-govt, non-financial public limited companies –

I. Internal sources [(i) reserves and surplus, and (ii) depreciation], and II.
External sources [(paid up capital through new issues and premium),
borrowings (debentures, from loans and from FIs) and trade dues and
other current liabilities]; (SEBI);
xvi) Annual and monthly series on resources mobilized, instrument wise, from
the primary market – number and amount mobilized by category of issue
(public issue and rights issue); by type of issues [by listed companies and
initial Public Offerings (IPO)]; by equities at par and equities at a
premium; cumulative convertible preference shares (CCPS); bonds; and
others; (SEBI);
xvii) Annual and monthly series of data on capital raised by (i) industrial
(economic activity) classification (banks/FIs), industries like electronics,
and engineering, entertainment, finance etc.; (ii) size of capital raised;
(iii) sector (public and private); and (iv) region (north, east, south and
west); (SEBI);
xviii) Annual and monthly data series on the number and quantum of Euro
Issues; (SEBI);
34
The compilation of the RBI Index was discontinued from 1999-00. The compilation of
market capitalization – all India was discontinued by BSE since 1999-00. The SEBI
Handbook provides data on market capitalization – all India from 1999-00 as given in the
80 publication of the National Stock Exchange (NSE).
xix) Annual and monthly data series on transactions of MFs on the Stock Trade and Finance
Exchanges – gross purchases and sales and net purchase/sales in (a)
equity, and (b) debt; (SEBI);
xx) Trends on trading on stock exchanges – the number of shares traded and
the number and value of shares delivered; (SEBI);
xxi) Indicators of liquidity – Market capitalization – GDP ratio (BSE), market

capitalization – GDP ratio (NSE), turnover ratio – BSE, traded value ratio
– BSE, traded value ratio – NSE; (SEBI);
xxii) Trends in foreign investment flows – direct and portfolio investment;35

(SEBI);
xxiii) Annual and monthly series on trends in FII investment – gross purchases
and sales and net investment; (SEBI);
xxiv) Comparative evaluation of Indices through Price to Earnings Ratio and

Price to Book Ratio (these are monthly averages of closing values) for
BSE SENSEX, BSE 100 Index, S&P CNX NIFTY, CNK NIFTY Junior;
(SEBI); and
xxv) Survey of investor households – a joint effort of SEBI and the National
Council of Applied Economic Research – result giving an estimate of
investor and non-investor households by type of investment, household
by type of instruments invested in and so on.

1) Which document does contain the methodology for compiling liquidity
aggregates?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) List the kind of time series data available in RBI Handbook 2013-14.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
35
Figures from 1995-96 include acquisition of shares of Indian companies by non-residents
under S/6 of FEMA, 1999. Those from 2000-01 have been revised with expanded coverage
to approach international best practices and are, therefore, not comparable with earlier data. 81
Data Base of Indian 3) Identify 3 different sub-sectors of financial market and name a few major
Economy
sources of data for each of these sub-sectors of financial markets.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
23.4 LET US SUM UP

We have, in this Unit, surveyed the data available in the area of trade and
finance. We noted that data on the volume of merchandise trade by
commodities and countries, direction of trade, trade balance, quantity and unit
value index numbers of imports and exports and indices depicting different
measures of the terms of trade are available from the DGCI&S and are also
presented as a time series in the RBI Handbook 2013-14, besides the
Handbook of Statistics of the SEBI. Data on trade in services and on
merchandise are available as part of BoP data of the RBI. There is a divergence
between merchandise trade deficit/surplus data provided by DGCI&S and that
shown by BoP data of RBI and the reasons for such a discrepancy are related to
the manner in which the two sources collect the basic data.
Data are available in the RBI Handbook (long time series) on public finances
such as receipts and expenditure of the Central and the State Governments, the
manner in which these Governments mobilize resources through taxes and non-
tax revenue borrowings, the patterns of their expenditure – developmental
outlays, debt servicing, economic and social services and so on – and the
various types of (fiscal) deficits that they run and the manner in which these are
financed. Data on balance of payments, inflow of foreign investment, foreign
aid/borrowings, assistance and the size of foreign debt are also available in the
RBI Handbook. So are the data on the money stock, the different monetary
aggregates like M0, M1 and M3, and the new monetary aggregates NM2 and
NM3 and “C” components of the money stock, the sources of change in the
money stock and their “S” components and liquidity aggregates L1, L2 and L3.
The role of the banks and financial institutions in the mobilization of savings
and the deployment of credit to different economic activities, their functional
and operational efficiency in terms of indicators like NPA and CRAR can be
examined well with the time series data available in the RBI Handbook and
such an effort can be supplemented with information on profitability provided
by the RBI’s Report on Trends in Banking. Similarly, the functioning of the
financial market in terms of the structure of interest rates in different markets,
resources mobilized through different modes like capital issues, private
placement, equities with and without premium, preference shares and CCPS,
total resources raised from the primary market, trends in trading in stock
exchanges, share price indices and market capitalization, indicators of liquidity
like traded value ratio and market capitalization to GDP ratio and measures of
comparative evaluation of indices (BSE SENSEX, BSE 100 Index, etc.,) like
Price to Earnings Ratio and Price to Book Ratio, can be analysed adequately
82
with the data available on these aspects in the RBI Handbook and the Trade and Finance
Handbook of Statistics, published by SEBI. The RBI Handbook, especially
with the special features incorporated in its website version, stand out as an
invaluable source of data on finance and also trade.
23.5 EXERCISES
1) How can you analyse trend in foreign trade? Discuss the role of unit value
index and quantum index in this regard.
2) Discuss the kind of time series data on money and banking compiled by
RBI. How this data can be useful for research?
3) List the types of data on financial market compiled by SEBI. To what

extent is it adequate to analyse the financial markets?

Ministry of Finance, : Economic Survey – 2005-06 and earlier years,
Govt. of India Ministry of Finance, Govt. of India, New Delhi.
Budget Documents, Ministry of Finance, Govt.
of India, New Delhi.
Reserve Bank of : Report on Currency and Finance – 2005, RBI,
India Mumbai.
Report on Trends in Banking – 2005, RBI,
Mumbai.
World Trade : International Trade Statistics – 2005, WTO,
Organization Geneva.
Also accessible at http://www.wto.org/english/res_e/statis_e.htm

PROGRESS EXERCISES
2) 1998-99.

1) The budget documents of the central and the state governments, the pre-
budget economic survey, RBI Indian Economy 2005, Monthly Bulletin of
the RBI.
2) The excess of total expenditure including loans over revenue receipts – net
lending of the central government.
3) See Sub-section 23.3.1 (sub head and ‘d’) 83
Economy
1) RBI Bulletin, Nov. 2000 under the title ‘New Monetary and Liquidity
aggregates’.
3) The RBI Handbook of Statistics of Indian Economy 2005 and the
Handbook of Statistics on Indian Securities Market 2005 provide
comprehensive data on the financial market.
84
Social Sector
UNIT 24 SOCIAL SECTOR
Structure
24.0 Objectives
24.1 Introduction
24.2 Employment, Unemployment and Labour Force
24.2.1 Magnitude of Employment and Unemployment
24.2.2 Quality/Adequacy of Employment
24.2.3 Labour Welfare
24.3 Education
24.3.1 Educational Infrastructure
24.3.2 Infrastructure Utilization and Access to Educational Opportunities
24.4 Health
24.4.1 Health Infrastructure
24.4.2 Public Health, Morbidity and Mortality Indicators
24.4.3 National Family Health Survey (NFHS)
24.5 Shelter and Amenities
24.6 Social Consequences of Development
24.7 Environment
24.8 Quality of Life
24.9 Let Us Sum Up
24.0 OBJECTIVES
• know the sources of data on various aspects of employment and unemployment,
labour welfare, education, health, shelter, safe drinking water, etc., which
determine the quality of life of the people;
• Identify the various organization/agencies involved in the compilation of
data on different aspects of social sector;
• describe the different concepts used in collection of data on social sector;
and
• explain the kind of data/information available in the various publications
containing data on social sector.
24.1 INTRODUCTION
Social Sector consists of education, health, employment, shelter, sanitation,
other housing amenities, environment, and adverse consequences of
development and levels of living or quality of life in general. Investments in
85
Data Base of Indian this sector pay rich dividends in terms of rising productivity, distributed
Economy
growth, reduction in social and economic inequalities and levels of poverty
although after a relatively longer time span than in the case of investment in
physical sectors. Let us look at the kind of data available in this sector.
24.2 EMPLOYMENT, UNEMPLOYMENT AND

LABOUR FORCE
Employment is the means to participate in the development process. Creation
of employment opportunities is an important instrument for tackling poverty
and to empower people, especially women. Regarding data on employment,
firstly, we should know how many are employed and how many are ready to
work but are unable to gain access to employment opportunities. How do
women fare in these matters? Or, for that matter, what is the experience of men
and women belonging to different social/religious/disadvantaged groups? Are
children employed in any economic activity that is not only hazardous to their
health but which also adversely impacts on our dream of a golden future for
them through efforts to ensure their mental and physical well-being? What is
the quality of employment opportunities available to the work force? What are
the conditions in which people work? Let us first look at data on the magnitude
of employment and unemployment.
24.2.1 Magnitude of Employment and Unemployment

Data on employment in selected sectors are available from several sources.
Data on Employment in the organized sector of the economy is available from
the Employment Market Information (EMI) programme of the Directorate
General of Employment & Training (DGE&T), in the Ministry of Labour.
This is based on statutory returns submitted1 to the employment exchange of
the area every quarter by non-agricultural establishments in the private sector
employing at least 25 persons and all public sector establishments (excluding
defense forces and Indian missions posted abroad) irrespective of their size.
Smaller establishments, that is, those employing 10 to 24 persons submit such
returns on a voluntary basis. Data on employment levels in the organized sector
of the economy (as defined above) is thus available every quarter at district
levels through Quarterly Employment Reviews released by the DGE&T and
the Directorates of Employment (DEs) in the State Governments and Union
Territory Administrations. The Annual Employment Reviews released by the
DGE&T and the DEs provide more detailed data on employment at the three
levels of NIC code. The returns also provide data on the number of vacancies
occurring in the establishment during the quarter and the number filled. This is
the only source of employment data on the organized sector of the economy
that is available at quarterly intervals. This is, however, subject to several
limitations. First of all, it excludes the entire agricultural sector, self employed
persons and part-time workers. In addition, adjustments are to be made to take
care of non-response in the submission of the returns. Incompleteness of the
employers’ register is also an impediment. Establishments in the organized
sector of the economy also submit statutory biennial returns providing the
occupational distribution of their employees and the educational profile of
those employed in selected occupations. These returns provide the basis for the
preparation of biennial reports on the occupational pattern of employment and
1
These statutory returns are submitted under the Employment Exchanges (Compulsory
86 Notification of Vacancies) Act, 1959 and the Rules of 1960 made there under.
the educational profile of selected occupations in the public and private sectors. Social Sector
However, non-response in the submission of these returns has over the years
deteriorated so much that the data on the occupational/education pattern has
now lost its utility.
The DGE&T and the DEs in the State Governments and Union Territory
administrations also provide data on the number of job seekers by age, sex and
education qualifications, and the socially and physically challenged and their
distribution by sex and education on the live register of employment
exchanges, the number of vacancies notified to the employment exchanges
under the Act referred to in the preceding paragraph and the number of job
seekers placed in jobs by the employment exchange. It would not at all be fair
to measure the efficiency of the employment exchanges by comparing the
number of placements with the number of vacancies notified to the exchanges
or with the number on the live register for several reasons. First, vacancies
being filled through public service commissions, other commissions, recruitment
boards and the like and those filled through competitive examinations need not be
notified to employment exchanges. Second, public sector undertakings need to
fill from the Employment Exchanges only vacancies carrying salaries below a
certain level specified by Government. Finally, establishments in the private
sector need to only notify their vacancies to the employment exchanges. They
are not obliged to confine their choice of candidates for recruitment to the
nominees of the exchanges and are free to fill their vacancies through the open
market. Sample surveys of those on the live register at any point of time have
shown that a sizeable proportion of the job seekers are already employed, some
are unemployed and some have registered at more than one exchange. At the
same time, surveys of employment and unemployment like the NSSO
quinquennial surveys (see below) have shown that not all the unemployed are
registered in the employment exchanges. Thus, the size of the live register can
overestimate unemployment and, from another angle, can also underestimate
unemployment. These factors may not also cancel each other and it is,
therefore, difficult to consider the size of the live register as a reasonably
accurate estimate of the level of unemployment. Nevertheless, it does represent
the extent of pressure in the job market, especially for government and public
sector jobs.
The second source is the Economic Census. This has been conducted in 1977,
1980, 1990, 1998, 2003 and 2013 covering all economic enterprises in the country
except those engaged in crop production and plantation and provides data on
employment in these enterprises 2 . These thus include also the unorganized
sector outside crop production, animal husbandry and plantations. The third
source is the Annual Survey of Industries (ASI), which gives us estimates of
factory sector employment, i.e., all factories registered under the Factories Act,
1948 and the Bidi and Cigar Workers Condition of Employment Act, 19663.
The ASI presents estimates of levels of employment by industries at the three-
digit level of NIC 2008 codes along with industry level aggregates like
investment, output, inputs, value added and so on4 so that one can derive useful
technical ratios to facilitate analysis of the role of the factors of production in
2
The first report on the Economic Census 2013 is available on the website of the Ministry
of Statistics and Programme Implementation (MoSPI) www.mospi.gov.in .
3
See the Sub-section 21.3.3 titled “Factory Sector – Annual Survey of Industries” in Unit 21
on Agricultural and Industrial Data.
4
See the list of principal characteristics given in footnote 22 in the section referred to in the
preceding footnote. Sub-section 21.3.3 referred to therein also gives an idea of the kind of
data on employment available from ASI. 87
Data Base of Indian industry. The quality of ASI data is thus tied to the completeness of the frame
Economy
of factories, which, in turn depends on the quality of the enforcement of the
Factories Act, 1948 and the Bidi and Cigar Workers (Conditions of
Employment) Act, 1966. Data on employment in Railways is available from
the Railway Board, Ministry of Railways, those on employment in the
Banking Sector from the Reserve Bank of India (RBI), those on employment
in posts and telecommunications respectively by the Department of Posts and
the Ministry of Telecommunications and those on employment in insurance
by the Ministry of Finance. The reports on the Census of Central
Government Employees conducted by the DGE&T every year and the
Census of Central Government Employees conducted by the respective State
Governments provide a time series data of employment in Government sector.
The publications5 of the Labour Bureau (Ministry of Labour), Shimla present

estimates of average daily number of workers in Government and local fund
factories and other factories – male and female adults (those who have
completed 18 years of age), male and female adolescents (those who are aged
between 15 to 18) and boys and girls (those who are not yet 15 but not below
14) – and employment in factories by industry groups. These also provide
estimates of average daily employment in mines and quarries for different
minerals. The first is based on the statutory returns received from factories
through the State Governments under the Factories Act, 1948 and the second
on similar returns received by the Director General of Mines Safety (DGMS)
under the Mines Act, 1952 as amended in 1983. The third set of data that they
provide is employment in shops, commercial establishments and restaurants,
theatres, etc., based on the returns submitted to the State Governments under
the Shops & Establishments Act and Weekly Holidays Act, 1942. All suffer
from inadequacy of response in the matter of submitting statutory returns.
Further, the last one suffers from an added limitation arising from the fact that
these Acts are in force only in certain urban areas, towns and cantonments. Of
late, similar returns are being collected by some of the Gram Panchayats, who
work under the Department of Rural Development of the respective State
Governments. Lastly, the publications give data on employment in tea, coffee
and rubber plantations on the basis of data received from the Tea, Coffee and
Rubber Boards and the Ministry of Agriculture. It must be mentioned here that
employment in these specific sectors will also be covered by the EMI
programme to the extent establishments in these sectors are covered by the
Act governing the EMI programme and the extent to which these
establishments submit the relevant statutory returns.
Comprehensive data on employment and unemployment covering the

entire country at regular intervals are available from two sources, namely,
the Population Census conducted by the Office of the Registrar General and
Census Commissioner, India every ten years and the quinquennial sample
surveys on employment and unemployment conducted by the National
Sample Survey Office (NSSO). The former is based on a complete
enumeration of the population. Workers in the Census are enumerated as main
workers and marginal workers. The Census 2011 presents data, at the
village/urban block level (for slums in all statutory towns also, which have
been treated as separate enumeration blocks in Census 2011), on the number
employed (male and female) or the size of the male and female workforce
5
Indian Labour Yearbooks (the latest is for 2009 and 2010) and Annual Employment
Unemployment Survey (the 4th report of this pertaining to 2012-13 is the latest) Labour
88 Bureau (Ministry of Labour and Employment), Shimla.
(main workers and marginal workers) and its distribution by four economic Social Sector
activity categories – (i) cultivators, (ii) agricultural labourers, (iii) household
industries and (iv) other activities. The rest of the population, namely, those
who are neither main workers nor marginal workers, are categorized as non-
workers. All other data on employment and unemployment are available at the
district level upwards for rural and urban areas and the urban data by urban
agglomerations and size classes of cities and towns. In Census 2011, till
December 2014, the following data has been published:
a) The Primary Census Abstract (PCA), published for each village and urban
ward, provide, among others, the number of female and male main
workers, marginal workers and non-workers for overall, SC and ST for
each of the four-fold classification of economic activity, i.e., cultivators,
agricultural labourers, household industry and other workers;
b) The number of female and male main workers, marginal workers and non-
workers in each single year of age, classified further by four-fold
classification of economic activity and social group disaggregated at India,
State and district level.
The kinds of data on employment and unemployment available from the

Census 2001, and to be published in Census 2011 for rural (R) and urban (U)
and (R+U) is shown below:6
a) The distributions of the male and female main workers by economic

activity (the first digit level of the NIC code) – ten categories of economic
activities7 and classified further by age groups and educational level (for
the all India table) and by educational level8 for the table on India, States
and Union Territories;
b) Similar distributions for male and female marginal workers;
c) Main workers, marginal workers, non workers, those marginal workers

that are available for or are seeking work (unemployed) and those non-
workers that are available for or are seeking work (unemployed), by sex,
social group and educational level;
6
Census tables released so far are available on CDs for users on payment of the prices
indicated in the Census website http://www.census.nic.in. Most of the India and State/UT
level tables can be viewed in / downloaded from the census website. Highlights of and
introductions to tables, concepts and definitions and an advance release calendar are also in
the website. Tables are also available at State/Union Territory/District City level on CDs.
7
In Census 2001, the NIC 1998 was used for dissemination of the results. In NIC ’98, the first
level had the following categories: a) Agriculture, Hunting & Forestry, b) Fishing, c) Mining
& Quarrying, d) Manufacturing, e) Electricity, Gas and Water Supply, f) Construction,
g) Wholesale and Retail Trade, h) Hotels and Restaurants, i) Transport, Storage and
Communications, j) Financial Intermediation, k) Real Estates and Business Activities,
l) Public Administration and Defence, and Compulsory Social Security, m) Education,
n) Health & Social Work, o) Other Community, Social and Personal Service Activities,
p) Private Households with Employed Persons, q) Extra-Territorial Organizations and
Bodies. In Census 2001 tables, Categories A& B were grouped together, J & K were grouped
together and L to Q were grouped together.
8
1)Illiterate, 2) literate but below metric, 3) metric/secondary but below graduate, 4) technical
diploma or certificate not equal to degree, 5) graduate and above other than technical degree,
and 6) technical degree or diploma equal to degree or post graduate degree and a further sub-
classification of item 6 into engineering and technology, medicine, agriculture and others. 89
Data Base of Indian d) Similar distributions for unemployed SC population, ST populations and
Economy
further classification of the distribution for unemployed general population
by religious communities;
e) Distributions of (i) main workers, (ii) marginal workers, and (iii) non-
workers by main activity, educational levels, age and sex; and
f) Similar distributions for SC population, ST population, further classification of

the distribution in item “e” above for the general population by religious
communities;
g) The number of main workers, marginal workers and non-workers among

the disabled by type of disability, age and sex; and
h) The male and female workforce by detailed economic activities at the

three or four digit levels, distribution by industrial categories (according to
National Industrial Classification – 1998 (NIC 1998) and a separate
table by occupations (as per the National Classification of Occupations –
1986 (NCO 1968) and educational profile and;
i) The number of marginal and non-workers who are seeking/available for

work;
j) The major non-economic activity of the non-workers, namely, student,

pensioner, household duties, etc. by age and sex. This category also provide
information on vulnerable sections of the society like beggars, etc.
The NSSO quinquennial surveys on employment and unemployment

constitute the other major source of data on employment and unemployment.
The last survey for which results have been published is the 68th Round survey
conducted during July, 2011 and June, 2012. The earlier surveys related to
October, 1972 to September, 1973, July, 1977 to June, 1978, January – December,
1983, July 1987 to June, 1988 July, 1993 to June, 1994, July 1999 – June 2000 and
July 2004- June 2005. These measure employment/unemployment/labour force
status on the basis of the Usual Status and Current Status and arrived at four
measures of employment/unemployment/labour force, namely, the Usual
Principal Status (UPS), the Usual Principal and Subsidiary Status (UPSS), the
Current Weekly Status (CWS) and the Current Daily Status (CDS) employment/
unemployment/labour force status. These concepts and the measures have been
discussed in the Unit on “Employment and Unemployment: Policy
Implications” in Block 1 of MEC – 005 on Indian Economic Policy in your
first year course. These surveys provide the per thousand distribution of
variables by characteristics along with estimated number of persons and/or the
sample number of persons for each column and row characteristic. These can
be used to estimate new proportions – like for example the worker participation
rate or the unemployment rate among the poor (those below a certain income
class) and the non-poor – and also estimates of the magnitude of employment
or unemployment using the estimated sample proportions and the relevant
population projection for the midpoint of the survey period. The surveys 9
9
Some of the reports relating to employment and unemployment on the 68th round are Report
No. 554 on “Employment and Unemployment Situation in India” and Report No. 557
on “Informal Sector and Conditions of Employment in India”. All these can be
downloaded from the MoSPI website: www.mospi.gov.in by becoming a registered user,
90 which is free of charge.
provide the following type of data10 for Rural (R), Urban (U) and (R+U) at Social Sector
national and State levels11:
a) Distribution of male and female UPS/UPSS/CWS/CDS workforce by

employment status – self-employed, regular wage/salaried employment
and casual labour – and by age group, educational level and primary,
secondary and tertiary economic activity12;
b) Distribution of UPS/UPSS male and female workforce by economic
activity up to the three-digit code of NIC; and similar distribution by
occupation up to the three-digit code of NCO 1968;
c) Distribution of male and female UPS/UPSS/CWS/CDS workforce by
employment status and by primary, secondary and tertiary economic
activity and by MPCE classes;
d) Distribution of UPS workers in each employment status by usual
subsidiary economic activities;
e) Distribution of UPSS employed by place of work (same village/town,
another village/town) and by distance of the place of work from the place
of residence;
f) Proportion of persons who are more or less regularly employed and the
distribution of persons who are not regularly employed by duration for
which they are available for work; (underemployment)
g) Distribution of UPS employed by “CWS employed”, “CWS unemployed”,
“out of the labour force by the CWS criterion”, showing underemployment
among the UPS employed;
h) Similarly distribution of the CWS “employed”, “unemployed,” and “out of
the labour force” on each half-day of the reference week giving the
underemployment among the CWS employed;
i) Distribution person-day of employment by (i) principal 13 economic
activity of the household, (ii) household type14, and (iii) household land
cultivated class, [distribution at (iii) only for rural areas];
j) Distribution of households and female-headed households by the number
of male and female adults (aged 15+) employed under the UPS criterion –
households with no adult UPS employed, number with one male adult
UPS employed, number with one female adult employed, number with one
male adult and one female adult employed and so on;
10
The unit record data, that is the basic data collected in the surveyed households, are
available in CDs from the Computer Centre of the MoSPI on signing an undertaking and
making a payment. The details can be seen at http://mospi.nic.in/Mospi_New/upload/
nsso/ratelist_UnitData.pdf.
11
Technically, data should be available by 70 and odd NSS regions being made up of one or
more contiguous districts. As the unit record data are available on floppies from NSSO
on payment, one can make any kind of tabulation for any region or regions. However, one
should be checking the sample size while preparing such estimates, as that affects the
reliability of an estimate.
12
In some cases, only by agricultural sector and the non-agricultural sector.
13
The economic activity that accounted for the maximum of the income of the household in
the 365 days preceding the date of the survey.
14
The nature and type of work from which a household derives its major income is an
important indicator of the activity pattern of its members. The classification of households
by household type is based on the major economic activity of the household during the 365
days preceding the date of the survey. 91
Data Base of Indian k) UPS/UPSS/CWS/CDS unemployment rates (can be easily estimated from
Economy
the tables on distribution of 1000 persons by various characteristics, as
indicated in the main paragraph) for all groups together and for different
groups like gender, age group, educational level, MPCE classes,
household type, household land cultivated class and principal household
economic activity, etc.;
l) Average daily salary/wage earnings of regular wage/salaried employees
aged 15-59 by sex, sector of work and education; a similar distribution by
occupational groups and education; and
m) Average daily wage/earnings per day received by casual labour by sex, age
group, type of operation, and sub-round15.
Data on employment are thus available from several sources. Estimates of
employment derived from different sources for the same sector or sub-sector
will differ. While this may be frustrating to a lay person, it opens up an exciting
and challenging opportunity to an analyst to unravel the factors responsible for
such variations and to find ways and means of tackling such factors. Such
factors could be differences in the concepts, definitions and mode of collection
of data (sample survey, census or administrative/statutory reporting systems)
used by the different sources. There may be other factors causing the
divergence in estimates. Some steps could be taken to remove at least a few of
these factors. Data collecting agencies should adopt the same concepts and
definitions for enquiries on the same or similar subjects, as far as possible. The
use of trained personnel in data collection does have a favourable impact on the
quality of data and should receive consideration. Estimates based on sample
surveys should accompanied by estimates of the standard error to which the
estimates from the survey are subject16.
24.2.2 Quality/Adequacy of Employment
NSSO survey data (items above) can be used to analyse the quality and
adequacy of employment. Besides underemployment among the employed, other
aspects of quality can also be looked at. One is the proportion of the employed by
employment status, especially the proportion of the work force in ‘casual
employment’. The second is a comparison of the average daily earnings of male and
female, regular and casual workers in different sectors and operations with the
prevailing statutory minimum wage and the poverty line. The third is a look at the
pattern of employment status of the workers belonging to poor households. These
will throw light on the quality of employment enjoyed by the workforce, quality in
terms of the intermittent nature of work, tenure of employment, employment
security and low wages and above all, the prevailing gender differences in these
aspects of quality. Gender differences in access to employment in different
occupations and sectors of economic activity would be clear from a perusal of the
tables listed in the preceding section. Even aspects of the question of the
invisibility of the female worker, leading to underestimation of the female
workforce can be examined17.
15
One year long round of the Survey has four sub-rounds to take note of the seasonal effect
on the variables under investigation.
16
Reports on NSSO surveys do provide information on the concepts, definitions, coverage,
sampling design and estimation procedures adopted and the standard errors of estimates
presented in the reports.
17
NSSO Report No. 554 referred to in an earlier footnote. See also “Gender and
Employment in India”, T.S. Papola & Alakh N. Sharma (Ed), published by Indian
Society of Labour Economics and Institute of Economics Growth in association with
92 Vikas Publishing House Ltd, New Delhi. (1999).
The Labour Bureau, Shimla under the Ministry of Labour also publishes18 data Social Sector
on wage levels in the organized sector and on the welfare of labour. Data on
total earnings of factory workers are collected through the statutory returns
under the Payment of Wages Act, 1936 from establishments defined as
factories under the Factories Act, 1948. The Act was applicable to employees
with earnings19 up to Rs. 200. This limit has been increased from time to time
to 1600 in 1982. These returns are, however, not received from all factories.
The Bureau also conducts Occupational Wage Surveys in selected industries at
regular intervals (called first round, second round and so on) and publishes
reports20 on these surveys) that provide the distribution of workers by levels of
earnings and by occupations for each industry. These reports facilitate analysis
of variation in levels of earnings within occupations, across gender within
occupations, across regions and industries and to judge the adequacy of such
employment opportunities in the light of statutory minimum wages and the
poverty line. We have seen in the Unit on Agricultural and Industrial Data that
the ASI gives information on the total wages, total emoluments and employee
compensation along with mandays of workers and mandays of employees in
different industries in the factory sector. These thus help in deriving the
average wage per worker-manday and the average emoluments per employee-
manday and the average employee compensation per employee-manday.
As for the unregistered sector and the unorganized part of other sectors, the
DE, NDE and OAE Surveys 21 and other Establishment Surveys of the
CSO in different sectors (till the early 1990s) and the unorganized sector
surveys of the NSSO 22 provide data on average annual earnings for men,
women and children in such establishments. Regular reports like “Wage
Rates in Rural India” (the latest published relates to 2004-05) and the Report
on the Working of the Minimum Wages Act, 1948 (the latest is for 2003), of
Labour Bureau enable an analysis of rural/unorganized sector wage levels vis-
à-vis statutory minimum wages and poverty line.
The prevalence of child labour, which depicts an exploitative dimension of the

economic system, can be seen from the Census and NSSO tables on the
distribution of the work force by age groups and activity, the former down to
the district level and the latter up to the State level (technically down to NSS
regions due to availability of the unit record data on floppies/CDs). Such data
may not fully reflect the ground level realities because surveys may not be able
to extract information on employment of children, for various reasons. Further,
such data are also dated. These can help only in drawing attention to the areas
where child labour is prevalent so that the authorities concerned can initiate
further action. It is usually the Non-Government Organisations (NGOs) active
in the field of the rights of children that succeed in locating establishments
employing children and press the Government to take further action.
18
Yearbooks of Labour Statistics, (various years), Ministry of Labour, Shimla.
19
Earnings include basic wages, dearness allowance, money value of concessions, annual or
prepaid bonus, and arrears.
20
Reports on Occupational Wage Surveys (Successive Rounds – the latest publications relate
to the sixth round), Ministry of Labour, Labour Bureau, Shimla.
21
Directory Establishments, Non Directory Establishments and Own Account Enterprises. The
first two employ at least one hired worker and a total of 6 or more persons and 1 to 5 persons
respectively. In Own Account Enterprises, no hired worker works on a regular basis. These
surveys are now being conducted by the NSSO. The of NSS Report 549, for the year 2010-
11 is the latest report on unorganized non agricultural enterprises (excluding construction),
which provide data on DE, NDE and OAE.
22
Besides the reports listed in Sub-section 21.3.4 of unit 21. 93
Data Base of Indian
Economy
24.2.3 Labour Welfare
The Labour Bureau publishes23 data on several aspects of labour welfare – data
on industrial injuries, injuries in mines, compensation to workers for injuries
and death, industrial disputes, health insurance, provident fund and trade
unions of employers and workers. Statistics on industrial injuries are collected
through statutory returns under the Factories Act, 1948 that provides that
industrial accidents due to which the affected persons are prevented from
attending to work for at least 48 hours should be reported to the inspector of
factories. These depend on returns, and the sizeable non-response in the
submission of these mars the quality of this source of data on incidence of fatal
and non-fatal injuries in factories in mines under the Mines Act, 1952, as
amended in 1983. Initially, data on the number of serious accidents and the
accident rate were classified as ‘fatal’ and ‘serious’. Subsequently, from 1984,
accidents and the accident rate (accidents per 100 employees) were classified
as ‘fatal’ ‘spot serious’24 and ‘reportable serious’25. Complete statistics are not
available for the same reasons mentioned earlier. Statistics on ‘compensated
injuries’ and the amount of compensation paid, both classified as (resulting in)
‘death’ and (leading to) ‘disability’ – ‘permanent’ and ‘temporary’, are
collected under the Workmen’s Compensation (WC) Act, 1923 on the basis of
annual returns received from the State Governments, Posts & Telegraph
Departments and the Railway Board for different Zonal Railways under the
Act. Compensation is payable to workers employed in ‘scheduled
employments’26 for injuries due to accidents resulting in death or disablement
for more than three days, provided that it is not caused through the fault of the
worker himself. The number of injuries reported in the returns does not reflect
the total number of injuries that occur, as all the injuries are not compensated
for. Further, many of the establishments covered by the Act fail to submit
returns and, therefore, the information received by the State Governments
about the compensated injuries and the amount of compensation paid is
incomplete. Compensation for injuries in establishments covered by the
Employees State Insurance (ESI) Act, 1948 are paid under the ESI Act and not
under WC Act. These limitations affect the trends and comparability over time
of compensated accidents.
Statistics on the number of (a) industrial disputes27, (b) workers involved in the
disputes, (c) mandays lost and (d) causes of disputes – (i) wages and
allowances, (ii) bonus, (iii) personnel matters, (iv) retrenchment, (v) leave and
hours of work/shift working, (vi) indiscipline and violence, (vii) others, and
(viii) cause not known. Completeness of these data would depend on the extent
to which outside authorities are involved in the resolution of the disputes. How
are the workers and the employers, the two pillars of industrial progress,
organized to fight for their rights? The Trade Unions Act, 1926 regulates the
formation and functioning of such organizations. The annual returns under the
23
Yearbooks of Labour Statistics referred to in an earlier footnote.
24
This is one where one or more persons have received serious bodily injury. This means any
injury which involves, or in all probability will involve, (a) the permanent loss of any part
or section of the body or use of any part or section of the body or (b) the permanent loss of,
or injury to, sight or hearing or (c) any permanent incapacity or (d) the fracture of any bone
or one or more joints or bones of any phalanges of hand or foot.
25
This is one where one or more persons have received “reportable injury”. This means any
injury other than a serious bodily injury, which involves, or in all probability will involve,
the enforced absence of the injured person from work for a period of 72 hours or more.
26
Employments specified in the Schedule appended to the Act under reference.
27
94 Disputes resulting in work-stoppages involving 10 or more workers.
Act received from the State Governments relate only to those organizations that Social Sector
have been registered with the State Governments under the Act. And it is not
obligatory on these organizations to secure registration under the Act. The
response rate even among the registered ones is less than 50 per cent. The data
provided in the Yearbooks of Labour Statistics relates to the number of (i)
workers’ organizations on the register (ii) those submitting returns, (iii)
percentage response, (iv) membership at the end of the year, (v) income
including balance carried over from the previous year, (vi) expenditure, and
(vii) balance of funds at the close of the year and similar data for employers’
organizations. Information on the membership of the employers’ and workers’
organizations by NIC 1987 is available in the Yearbooks. Statistics relating to
the working of various welfare funds like the Coal, Mica etc., Mines Welfare
Funds and the ESI Corporation and the Employees Provident Fund
Organisation (EPFO) – number of beneficiaries/members, etc., - are also
provided by Yearbooks and the annual reports of the ESIC and EPFO28.
As for the unorganized sector, the Labour Bureau’s reports on their ongoing
programme of surveys on (i) the working and living conditions of Scheduled
Caste/Tribe workers, (ii) living conditions of unorganized workers, and (iii)
contract labour provide information on the qualitative aspects of employment
in terms of variables like wage levels, working and living conditions, work
place safety and amenities at the work place of these classes of unorganized
workers in rural and urban areas.

1) Name the sources that provide data on employment.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Explain the kind of data on employment and unemployment available
from the population census 2001.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) State the different measures of employment/unemployment/labour force
used by NSSO in different quinquennial surveys.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
28
See also the Section on Health in this Unit. 95
Data Base of Indian 4) Identify the different aspects of labour welfare on which data are compiled
Economy
by the Labour Bureau.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
24.3 EDUCATION
Education is an important instrument for empowering people, especially
women. Education nurtures and develops their innate talents and capabilities
and enables them to contribute effectively to the development process and reap
its benefits. It is also an effective instrument for reducing social and economic
inequalities. We have built up over the years a vast educational system in an
attempt to provide education for all, to ensure that the skill and expertise needs
of a developing economy are met satisfactorily and at the same time, to
monitor the functioning of the educational system as an effective instrument for
tackling inequalities. We would, therefore, like to look at data on different
aspects of the educational and training system such as its size and structure, the
physical inputs available to it for its effective functioning, its geographical
spread, the type of skills and expertise it seek to generate, the access of
different sections of society and areas of the country to it and the progress
made towards the goals like ‘total literacy’, ‘universalisation of secondary
education’, ‘education for all’ and ‘removal of inequalities’.
The United Nations, in the year 2000, has put achieving of universal primary
education as one of the eight Millennium Development Goals, to be achieved
by 2015. Education plays an important role in shaping the post 2015
development agenda as well. The Government of India has enacted the Right to
Education Act in August 2009. Therefore, the importance of research on the
progress of the nation towards achieving these goals and its effects on the other
social and economic goals is increasing by day.
The Department of School Education and the Department of Higher Education

of the Ministry of Human Resources Development (MHRD), the National
University of Educational Planning and Administration (NUEPA), National
Council for Educational Research and Training (NCERT) and the University
Grants Commission (UGC) collect and publish educational statistics, conduct
research studies and surveys in the area of education. The International
Standard Classification of Education (ISCED) is used worldwide to compile
and compare cross-country statistics of education. In 2014, for the first time,
the MHRD has developed the Indian Standard Classification of Education
(InSCED), as a part of collection, dissemination and presentation of statistical
data on education. Educational activities have been first classified into 16
broad levels from A through O and X. Levels A to E pertain to School
Education, levels F to L pertains to Higher Education, levels M, N and O
pertain to Certificate Courses, In-service training and Adult Education. The
detailed structure can be accessed at http://mhrd.gov.in/sites/upload_files/
mhrd/files/statistics/InSCED2014_0.pdf.
96
The annual publication “Educational Statistics at a Glance” provide a host of Social Sector
information on school education, like number of educational institutions, level
and Stage-wise enrollment in school and higher education, teacher-student
ratio, drop-out rates, examination results, public expenditure on education, etc.
The MHRD also publishes a compendium on Universities in India, providing
University-wise data on student enrollment, number of teachers, staff quarters,
hostels, etc. The Annual Report of the MHRD also provides information on
school and higher education in India. All these reports can be accessed free of
charge at the website www.mhrd.gov.in in the sections documents and reports
and the section statistics.
The NUEPA has developed a District Information System of Education

(DISE, can be accessed at www.dise.in ) to record data from the schools and
junior colleges in all districts of India. Analytical reports based on this data,
like school and facility related indicators, enrollment based indicators, teacher
related indicators and EDI and analytical tables can be accessed at
http://www.dise.in/AR.htm. These reports are available from the period 200-02
onwards, although the time series varies among topics. Provisional results of
the “Report of the eighth AIES”, (http://www.aises.nic.in/surveyoutputs)
published by the NCERT, for the year 2009, in spite of being a bit dated,
provides important data on school education statistics. While the data in the
publication of MHRD are based on returns from the State Governments and
Union Territory Administrations, the last one is based on a field survey. The
University Grants Commission and MHRD provide data on university level
courses in professional and technical courses. The MHRD has set up a
National Technical Manpower Information System (NTMIS) –
(seehttp://www.iamrindia.gov.in/_aboutNTMIS.htm) with lead centre at the
Institute of Applied Manpower Research (IAMR), New Delhi and 21 nodal
centres located at different States. It provides data on technical manpower –
intake in and output from institutions and the utilization patterns of such output
in detail through Tracer Studies 29 and other studies. These are disseminated
through reports released from time to time by IAMR. The Employment-
Unemployment surveys of the NSSO and Education surveys conducted in
the 52nd (1995-96), 64th (2007-08) and 71st (2014) Round of NSSO and the
social and cultural tables of the population Census (ORGI) also provide
useful data on literacy and educational composition of the population or stocks
of educated manpower of different levels of education and their utilization
patterns – snapshots, at specific points of time, of the impact of the efforts in
the area of education.
The kind of data available in these publications and sources are indicated below:
24.3.1 Educational Infrastructure

a) Levels of literacy and progress of efforts under literacy campaigns; (MHRD)
b) The number of institutions – colleges (including universities, deemed
universities and institutions of national importance) and all types of
general, technical and professional schools; (MHRD & UGC)
29
Tracer Studies are, as the name suggests, trace or follow up specific cohorts or batches of
students passing out of institutions to find out their present activity status. This helps in
evaluating the trends over time in (a) the capacity of the economy to absorb the specific
category of manpower in employment, (b) the waiting time for employment for the category
of manpower, and (c) market value, in terms of salary levels in first employment, of the
manpower category. These surveys are also called Cohort Studies. 97
Data Base of Indian c) The number of universities, deemed universities and institutions of
Economy
national importance, the distribution of number of colleges by categories
like those teaching arts, science & commerce, oriental learning and
different categories of professional subjects like law, agriculture,
engineering & technology and medicine and the distribution of schools in
the area of general education (and stages like pre-primary and primary
etc.,) and different categories such as those dealing with vocational,
professional, special education etc. (MHRD & UGC);
d) Teacher (male and female and those belonging to SC/SCcommunities), the
proportion trained among the, terms and conditions of their employment
and attrition of the stock of teachers by cause. (AIES);
e) Patterns of management of colleges including professional colleges and
deemed universities, recognition by appropriate bodies, availability of
teachers and facilities like laboratory and equipment etc. (UGC);
f) Patterns of management (Government, local bodies, private management,
religious and linguistic minority trusts/organizations and so on), medium
of instruction, type of school buildings (pucca, kutcha, thatched hut etc.,),
crowding (number of rooms in the building and the number used for
instructional purposes), availability and adequacy of facilities like drinking
water, labatories and urinals (and whether these are available separately
for girls). (Sixth AIES);
g) Teachers in professional and technical institutions and the number of
technical teacher training institutions. (MHRD & UGC);
h) The number of Industrial Training Institutions (ITIs) and Advanced
Training Institutes, and the training capacity for apprenticeship training in
industry under the Apprenticeship Act, 1961 for those passing out of ITIs
and the vocational stream of schools. (MoLE and MHRD);
i) The number of institutions for training instructors for it is. (MoLE)30;
j) The number of Vocational Rehabilitation Centres (VRCs) for the
Physically Handicapped31 set up by the DGE&T all over the country for
giving adjustment training and placement in suitable employment through
the Special Employment Exchanges for this group of people. (DGE&T32);
k) Expenditure on education by programmes like the Sarva Siksha Abhiyan,
the Total Literacy Campaign and Adult Education; direct and indirect
expenditure on recognized institutions of education. (MHRD).
24.3.2 Infrastructure Utilisation and Access to Educational

Opportunities
a) Literacy rates – overall and age-specific, among SC/ST and adults;
(Census – Social & Cultural Tables; NSSO & MHRD);
b) The number of students and the number of female students in the
institutions specified in item ‘b’ above;
30
Annual Reports of the Ministry of Labour, the part relating to DGE&T.
31
The phrase “physically challenged” instead of the phrase “physically handicapped” is being
brought into use nowadays in public discussions on the subject.
32
98 See the preceding footnote.
c) The number of students and the number of girl students by courses and Social Sector
stages of education in recognized institutions – from the nursery class to
the high/higher secondary level and in schools for vocational &
professional education and special education and similar data for rural
areas. (MHRD);
d) The number of male and female workers and marginal workers in the age
group 5-9 and 10-14 – working children; (Census and NSSO); a
comparison of this with the population in these age groups and data on
school enrolment would lead to an assessment of the number of children
who are neither in school nor in the workforce;
e) Enrolment and output in institutions specified in items ‘h’, ‘I’ and ‘j’
above;
f) Enrolment (and enrolment of female students) in university level general
education courses by stages (degree, post graduate degree,
diploma/certificate, research) and university level professional and
technical education courses by faculty. (MHRD & UGC);
g) Similar information of the kind mentioned above for Scheduled Castes and
Tribes (SC/ST) and females belonging to such sections of society. (MHRD
& UGC);
h) Teacher-Pupil ratios at different levels of education. (MHRD);
i) Intake and output of graduates and post graduates in different disciplines
and faculties. (MHRD & UGC);
j) Coverage of population in appropriate age groups by different stages of
education. (MHRD);
k) Dropout rates at different stages of education. (MHRD);
l) Distribution of population attending educational institution by age, sex and
type of educational institution for general and SC/ST population;
m) Access of the general and SC/ST population to (or availability of) facilities
for different levels of education in rural habitations and urban settlements
belonging to different population slabs, in terms of distance of the
habitation from the facility; similar information regarding non-formal
education (NFE) centers. (Sixth AIES);
n) Availability of special institutions suited to different types of disability of
children in different villages and towns and the enrolment of disabled
children in such institutions. (Sixth AIES);
o) Utilization patterns of different categories of professional and technical
manpower in general and trends in waiting time for employment,
utilization patterns and bargaining power for employees’ compensation
(salary etc.,) through tracer studies. (IAMR);
p) The disabled among main workers, marginal workers and non-workers by
type of disability, age and sex. (Census);
q) Stocks of educated manpower and also selected categories of technical
manpower for the general population, SC/ST and by religious communities;
99
Data Base of Indian r) The distribution of population by different levels of educational attainment
Economy
(including illiterates, literates without any educational attainment) for
population groups like SC/ST/Backward Classes (BC)/general households
and those belonging to different religions, household types like (a) those
with different sizes of land holdings; (b) households self-employed in
agriculture, households self-employed outside agriculture, agricultural
labour households, other labour households and households depending
largely on regular wage/salaried employment outside agriculture; and (c)
households belonging to different monthly per capita consumption
expenditure (MPCE) classes. (NSSO and Census)33.
One should not forget to mention in this context the National Human
Development Reports (NHDR) prepared by the Planning Commission (now
rechristened as NitiAyog) and the HDRs prepared by different State
Governments as very useful sources for the manner in which available data on
education, health, and other areas have been utilized to compute human
development indicators and also as important sources for basic data on these
sectors.
1) Name the organizations/institutions that collect and publish the data on

different aspects of education.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) What do you understand by the term ‘Tracer Studies’? How are these
useful?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
3) Prepare a chart stating the kind of educational data available in the various
documents published by MHRD and UGC.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
33
See also the sub-section (b) on education in the Section on ‘Level of Inequality on Non-
income Aspects of Life in Unit 4 on Poverty & Inequality: Policy Implications in Block 1 of
100 MEC – 005 on Indian Economic Policy in your first year course.
Social Sector
24.4 HEALTH
One of the important dimensions of quality of life is health. A healthy
individual can contribute effectively to production of goods and services.
Investment in health is, therefore, an essential instrument of raising the quality
of life of people and the productivity of the labour force. What is the health
status of the population? What are the challenges to the health of the population
and how are these being tackled? What kind of data is available about these
aspects of the population, the health infrastructure and the efforts being made
to deal with problems of health? What is the impact of these on the health
situation, especially of women and children? The World Health Organisation
(WHO), India website (http://www.who.int/countries/ind/en/ ) provides country
statistics on health profile, nutrition, mortality and burden of disease and risk
factors like those of alcohol and tobacco. The Central Bureau of Health
Intelligence of the Ministry of Health and Family Welfare publishes the
“National Health Profile (NHP) of India” every year. The publications
“Sample Registration System (SRS): Statistical Reports”, “SRS
Compendium of India’s Fertility & Mortality Indicators, 1971-2001”,
Mortality Statistical and Cause of Death”, SRS Bulletin and the Social &
Cultural Tables (C Series Tables) of Census 2001 of the Office of the
Registrar General of India (RGI), Ministry of Home Affairs and the “Report
on the National Family Health Survey (NFHS-4) – 2014-15” of the Ministry
of Health and Family Welfare contain a large amount of information on these
aspects of health, at the State level. For district level information, you can see
the reports of the Annual Health Survey (AHS) of the ORGI and the District
Level Health Survey (DLHS) of the MoH&FW. The Government of India
launched the National Rural Health Mission in 2005, which has now been
expanded to cover urban areas and called the National Health Mission. From
2005, the MoH&FW publishes, as an offshoot of the NHM, the Rural Health
Statistics (RHS) report each year.
24.4.1 Health Infrastructure

The RHS of the MoH&FW provide the following types of data:
i) Demographic indicators – state-wise number of villages, rural and urban

population, population growth rates and density, estimates of crude birth
rates, crude death rates and infant mortality rates;
ii) Rural health infrastructure – number of health sub-centres in each 5-year
plan period, number of Primary Health Centres, number of community
Health Centres, number of sub-centres, PHCs and CHCs functioning at the
end of the year, number of sub-divisional hospitals, district hospitals and
mobile medical teams, shortfall in health infrastructure, building position
of sub-centres, PHCs and CHCs;
iii) Health manpower in rural areas – Health workers (female)/ANM at sub
centre, PHCs, Health workers (male) at sub centre, number of sub-centres/
PHCs without ANMs/Health worker (Male), health assistants/LHVs at
PHCs, doctors at PHCs, number of PHCs without doctors/ lab technicians/
pharmacists, number of PHCs with AYUSH facility, surgeons at CHCs,
obstrreticians and gynaecologists at CHCs, physicians at CHCs,
paediatricians at CHCs, general dutty medical officers at CHCs,
radiographers at CHCs, pharmacists at CHCs and PHCs, lab technicians at
CHCs, nursing staff at PHCs and CHCs, etc.; 101
Data Base of Indian iv) Facilities available at sub centre – number of sub centres functioning, with
Economy
ANM quarter, ANMs living in sub centre quarters, number of sub centres
functioning as per IPHS norms, number of sub centres with regular water
supply, with electricity, with all-weather motorable approach road, number
of PHCs functioning, with labour room, with operation theatre, with at
least 4 beds, regular water supply, with electricity, with all-weather
motorable approach road, with telephone, with computer, with referral
transport, registered RKS, functioning as per IPHS (Indian Public Health
Standard) norms;
v) Facilities available at CHCs - number of CHCs functioning, with all four
specialists, with computer/ statistical assistant for MIS/ Accountant, with
functional laboratory, with functional OT, with functional labour room,
with at least 30 beds, with functional X-ray machine, with quarters for
specialist doctors, with referral transport, registered RKS, functioning as
per IPHS norms, number of CHCs having a regular supply of allopathic
drugs for common ailments, with AYUSH drugs for common ailments;
vi) Training of medical and paramedical personnel –ANM/ HW (F) training
schools funded by Government of India, established by Government of
India, health and family welfare training centres, MPW (M) training
centres;
vii) Rural health care – some parameters of achievement – classification of
States/UTs by average rural population covered by a sub centre, by a PHC,
by a CHC, average radial distance covered by PHCs, ratio of LHV/ health
assistant training schools to LHV/ health assistant, etc.;
viii) Expenditure on health and family welfare – overall and on individual
programmes like malaria control, filarial control and national leprosy
eradication programme.
24.4.2 Public Health, Morbidity and Mortality Indicators

i) Progress of the programme for vaccination of children and pregnant
women;
ii) Time series on the number of notified/reported cases of and deaths due to
diseases like cholera, small pox, acute diarrhoeal diseases, malaria,
kalaazar, Japanese encephalitis and meningitis;
iii) The number of cases detected, treated and discharged in respect of
diseases like leprosy and tuberculosis;
iv) The number of patients treated, discharged and deaths due to (a) different
types of cancer in specialized cancer hospitals, (b) different types of
mental diseases, and (c) communicable diseases like diphtheria,
poliomyelitis, tetanus (neonatal and others), hepatitis and rabies;
v) Progress in the National Aids Control Programme and other National
Control/Eradication programmes;
vi) Utilization by beneficiaries of facilities provided by CGHS, ESI scheme,
Control/Eradication programmes;
vii) Incidence of morbidity and mortality by causes in zonal Railway
hospitals;
102
viii) Incidence of morbidity by diseases in the ESI scheme; Social Sector
ix) Causes of death statistics: distribution of deaths due to (a) selected

diseases by age and sex, (b) specific causes under the group of diseases
peculiar to infancy, and (c) causes related to childbirth, (d) causes related
to childbirth and pregnancy (maternal mortality) and (e) accidents due to
different types of natural and other causes (RGI);
x) Medical certification of cause of death – distribution of such deaths by
age, sex and major cause groups (18 groups of causes) and the extent of
coverage of such certification to total deaths in each of the 18 groups
(RGI);
xi) Birth Rates (BRs), Death Rates (DRs) and natural growth rates and State-
wise BRs and DRs (from SRS, RGI);
xii) Mortality Indicators like the Crude Death Rate (CDR), Infant Mortality
Rate (IMR)34, Nano-Natal Rate (NNR)35, Post-Natal Rate (PNR)36, Still
Birth Rate (SBR)37 , Age-Specific Death Rates (ASDRs) and Maternal
Mortality Rates (MMR) (from SRS, RGI);
xiii) Fertility data – 2001 (F series Tables) (census 2001);
xiv) Disabled population in India by type of disability, sex and age in
general/SC/ST population and a further classification by marital status for
the general population (from Census 2001); and
xv) Expectation of Life at Birth (e0) 38, expectation of life at ages 10, 20, 30,
40, 50 and 60 for males, females and persons (from RGI).
24.4.3 National Family Health Survey (NFHS)

What is the impact of the efforts made to expand the size and reach and also
enhance the quality of the health services on the health status of the family,
especially of women and children? The first National Family Health Survey
(NFHS-1) conducted in 1992-93 succeeded in answering this question and also
in building up an important demographic and health database in India. This
success paved the way for the conduct of the second National Family
Health Survey in 1998-99 (NFHS-2), third NFHS in 2005-06 and NFHS-4
in 2014-15, to strengthen this database further and facilitate implementation
and monitoring of population and health programmes in the country. NFHS – 4
is covering, for the first time, all the 35 States and UTs as per Census 2011.
The survey is based on a representative sample of 5,60,000 households, up
from about 1,90,000 households of NFHS-3. The survey provides State-level
estimates of demographic and health parameters and also data on various socio-
economic and programmatic factors that are crucial for bringing about desired
changes in India’s demographic and health situation.NFHS-1 and NFHS-2
were funded by the United States Agency for International Development, with
supplemental funding from UNICEF. NFHS-3 funding was provided by the
United States Agency for International Development, the Department for
34
Relates to the number of children who die before their first birthday.
35
Relates to the number of children who die in the first month of their life.
36
Relates to children who die after the first month of life but before their first birthday.
37
Relates to children who are born dead.
38
The number of years that a person born today will, on the average live. Similary, ei., (i=10,
20, 30, 40, 50, 60 ….) is the number of years that a person aged i years today will, on an
average live. 103
Data Base of Indian International Development (United Kingdom), the Bill and Melinda Gates
Economy
Foundation, UNICEF, the United Nations Population Fund, and the
Government of India. Assistance for the HIV component of the NFHS-3 survey
was provided by the National AIDS Control Organisation and the National
AIDS Research Institute39. The field data collection in the NFHS is conducted
by different research organisations. A total of 18 such organisations collected
data in NFHS-3 during December 2005 to August 2006. It has been planned to
conduct the NFHS-4 through 11 reputed field agencies like the AMS, DRS,
EHI, GFK, GIM, IIHMR, NIELSON, SPVM, RDI, SPYM and VIMARSSH.
NFHS-3 provides
a) urban and rural estimates for most States, (b) separate estimates for 8 cities,
and (c) estimates for the slum areas. Besides a national report, reports are
prepared for the States. The NFHS reports can be downloaded free from the
site http://www.rchiips.org/nfhs/sub_report.shtml. The kind of data on health
and nutrition status of women and children and estimates of parameters related
to these presented by the NFHS, are indicated in brief below:
i) Educational level of the household population, school attendance of boys

and girls and reasons for not attending school;
ii) Distribution of housing by availability of basic amenities40, ownership of
agricultural land, house and livestock and durable goods; standard of
living indicators based on these and habits like drinking, tobacco and
smoking;
iii) Distance of households from the nearest health facility and distribution of
rural residents living in villages that have selected facilities and services;
iv) Age at first marriage of (women) respondents, their exposure to mass
media, their employment status and aspects of their empowerment or lack
of it, including domestic violence;
v) Current fertility, variation in current fertility by various factors, fertility
trends, pattern of outcome of pregnancy, median number of children ever
born and living to ever married women, patterns of birth order and birth
intervals, median age of women at the first birth and the last birth of
child, patterns of fertility preference and sex (of the child) preference;
vi) Knowledge of use and time of first use of contraception, reasons for
discontinuation of, or for not using contraception;
vii) Estimates of age-specific death rates, crude death rates, infant and child
mortality41;
viii) Morbidity by selected diseases and its variation over States;
ix) Vaccination of children, vitamin A supplementation for children,
prevalence of acute respiratory infection, fever and diarrhoea, treatment
of diarroea and awareness of treatment like ORS packets;
39
See the website http://www.rchiips.org/nfhs/nfhs3.shtml which provides details of all the 4
NFHS surveys conducted so far.
40
Electricity, source of drinking water, time taken to get water, method of purifying water,
sanitation facility, fuel for cooking, type of house (pucca, semi-pucca and kutcha) and
persons per room.
41
Child mortality relates to children who die between their first and 4th birthday. Under-five
104 mortality relates to children who die between their 4th and 5th birthdays.
x) Knowledge about AIDS and ways of avoiding it; Social Sector
xi) Food consumption of women, nutritional status of women, anaemia

among them, iodisation of salt consumed in households and body mass
index42 - an indicator of the heath status of women;
xii) Median duration of breastfeeding – overall and in different States, type of
food consumed by children, nutritional status of children, anaemia among
them and indicators of acute and chronic malnutrition among children –
weight for age index, height for age index and weight for height index43;
xiii) Health problems of pregnancy, antenatal care, assistance during delivery,
place of delivery, post-partum care, and care and treatment of reproductive
health problems; and
xiv) Couple protection rate44.
It is necessary to make a reference here again to the NHDR 2001 of the
Planning Commission and the HDRs of the State Governments for their
importance in the field of health, education and other sectors.
1) Indicate the publications which contain the data on different aspects of

health.
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
2) What type of database is provided by National Family Health Survey?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
42
“Body mass index (BMI) = [weight in kgms.] / [height in metres)(height in metres)]. See
Unit 4 on Poverty & Inequality – Policy Implications” (section on human development
indicators) in Block 1 of MEC – 005 on Indian Economic Policy in your first year course.
43
The National Institute of Nutrition, Hyderabad has recommended that the nutritional status
of the international reference population recommended by the WHO could be used as the
standard for India. The Weight for Age Index is a composite measure that takes into
account both chronic and acute malnutrition. Children who are more than 2 standard
deviations (s.d.) below the median value of the index for the reference population are
“underweight”. The Height for Age Index measures linear growth retardation. Children
who are more than 2 s.d. below the median value of the index for the reference population
are considered short for their age, or are “stunted”. This is chronic undernutrition. The
Weight for Height Index examines body mass in relation to body height. Children who are
more than 2 s.d. below the median value of the index for the reference population are
considered too thin or “wasted”. This is acute undernutrition. (NFHS – II). See the
preceding footnote on BMI also.
44
The percentage of currently married women aged 15 to 49 years using family planning
methods at the time of the survey. 105
Data Base of Indian ……………………………………………………………………………….
Economy
3) What do you mean by the term ‘weight for age index’?
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
……………………………………………………………………………….
24.5 SHELTER AND AMENITIES

Another important dimension of the quality of life is shelter and access to
amenities like safe drinking water, toilet, adequate lighting and safe fuel for
cooking. We have already discussed this in sub-section (d) under the section on
“Levels of Inequality in Non-income Aspects of Life” in Unit 4 on “Poverty
and Inequality: Policy Implications” in Block 1 of MEC – 005 on “Indian
Economic Policy” in your first year course. The population Census, from 1991
onwards, have collected data on housing and amenities of the population
during the house-listing and housing census phase, i.e., in the years 1990, 2000
and 2010. From Census 2001 onwards, results on parameters related to this
aspect is disseminated for the urban slums as well. While Census 2001 results
on urban slums relate to the towns with 20,000 or more population, all statutory
towns were considered for identifying the urban slums and disseminating
results for them. The results based on Census 2011 on housing stock, amenities
and assets, at district level can be downloaded from http://www.censusindia.
gov.in/2011census/hlo/HLO_Tables.html. This site alsso provides such tables
for the urban slums. While the Population Census provides data at a much
disaggregated level, a higher quantum of information is provided by the
surveys conducted by the NSSO in their surveys on housing condition and
condition of urban slums. The NSS survey reports based on surveys conducted
in the 49th round (1993), 58th round (2002) and 69th round (2012)provide results
at State/UT level on these parameters. The 69th round of NSS survey has
published three reports – (i) “Key indicators of drinking water, sanitation,
hygiene and housing condition in India”, (ii) “Drinking water, sanitation,
hygiene and housing condition in India” and (iii) “Housing condition and
amenities in India” – contain the latest data on housing conditions including
those of dwelling in slums. The NFHS also gives some data on housing and
amenities. These sources of data provide the following kind of information
(data presented for rural/urban/urban slum/other urban areas in the case of
items a to f):
a) Distribution of households by type of structure –pucca, semi – pucca and

kutchha;
b) Access to toilet facility – SC/ST/other households – within the premises or
not; type of latrine;
c) Access to safe drinking water – SC/ST/other households, major source of
drinking water;
d) Use of electricity for lighting – SC/ST/other households;
e) SC/ST/other households with no access to electricity, safe drinking water
106 and toilet;
f) SC/ST/other households with access to electricity, safe drinking water and Social Sector
toilet;
g) Percentage of villages connected by roads45;
h) Type of slum – notified or non-notified -, availability of infrastructure like
sewerage system, drainage system, means of garbage disposal and pucca
road(s) within and approaching the slum, how has the slum been there, and
whether the land on which the slum is located is owned by Local Bodies,
State Government, etc.;
i) Duration of residence of the household in the slum, reasons for coming to
the slum, the place from which the household came, whether the
household had any document of identification and whether it had at any
time tried to leave the slum;
j) Possession of durable goods and rent being paid for the premises; and
k) Workforce participation rates, female workforce participation rates,
workers and non-workers and workers by broad categories of economic
activity, including household, industry; and demographic features of the
slum population like sex ratio – overall and for children in the age group
0-6, etc.
24.6 SOCIAL CONSEQUENCES OF

DEVELOPMENT
The development experience of the last few decades show that the cost of
development is not shared equally by all sections of the society. Often the
burden of the cost falls almost entirely on the poor and the voiceless. The best
example of inequitable sharing of the burden of the cost of development
programmes is provided by the current controversy over the Sardar Sarovar
Project (SSP) over the river Narmada and the struggle of the Narmada Bachao
Andolan (NBA) on behalf of the people displaced by the construction of the
dam and the consequent submerging of their villages and their lands or are
likely to be displaced with each installment of rise in the height of the dam.
The other recent example that invited wide attention is the acquisition of the
lands of tribals in Orissa for the industries in Kalinganagar in the State. The
problem of people ousted from the land submerged by the water filling the
reservoirs created by the dams or, for that matter, people whose lands are
acquired by the State for purposes of development (building factories, dams
etc.,), called “oustees”, has been coming up ever since the first development
projects were taken up in 1951. The State promised jobs to the oustees in the
industries that came up in their lands in addition to compensation for the land
acquired or alternative land to prevent loss of their livelihood. The question
that has continued to agitate the minds of most development analysts is about
the extent to which the rehabilitation of the displaced or the “oustees” has been
successfully completed. What data are there on this vital question?
It may be mentioned that “oustees” were one of the priority categories for
registration as jobseekers in the employment exchanges and placement in
employment. The DGE&T, the organization concerned with employment
exchange at the national level, is one source where some information on (i)
45
Roads are all categories of roads surfaced or unsurfaced, district road, highways and rural
roads. (Basic Road Statistics of India, Ministry of Surface Transport, New Delhi.) 107
Data Base of Indian how many of them were registered as job seekers under these categories, (ii)
Economy
how many of these were actually placed in a job and (iii) how many languished
on the “live registers” of employment exchanges for a long time before passing
into oblivion, because if he or she did not renew her registration at specified
intervals he or she would have ceased to be on the “live” register of job
seekers. Past data with the DGE&T in the Ministry of Labour might show the
proportion of those who just ceased to be on the register. This source of data,
however, would have represented a very small part of whatever rehabilitation
was done, because the kind of vacancies that can be filled through employment
exchanges is restricted to jobs in Government not filled through competitive
examinations, public service commission and other commissions/recruitment
boards and jobs up to a specified salary level in public sector undertakings. The
private sector is not obliged to fill its vacancies through the employment
exchanges. Other than this source, Government – the Central or the State
Governments – do not seem to have any data on the number of families which
were displaced and the number rehabilitated. Data maintained by the State
Governments concerned with the implementation of rehabilitation under, for
example, the Sardar Sarovar Project, do not seem to reflect the actual situation
on the ground, as has been shown by the Report of the Group of Ministers
which toured some of the affected areas in Madhya Pradesh in April, 200646.
It was in 1994 that a study supported by the Indian Council of Social

Science Research (ICSSR) attempted estimates of the extent of displacement
of families between 1951 and 1990 due to projects like mines, dams, industries
and wild life sanctuaries. It also estimated the proportion of those displaced in
the Fifties and the Sixties who were resettled till 1980. It pointed out that there
was practically no improvement in the Eighties. A study on development-
induced displacement in West Bengal, between 1947 and 2000, conducted by
a team led by Walter Fernandes has made estimates of (a) the number of
people adversely affected by projects, (b) the number physically displaced, (c)
the number resettled by the projects, (d) the rest that had been left to fend for
itself following displacement, and (e) the proportion of those in (d) who were
dalits and tribals. Yet another study published in Economic and Political
Weekly by Manipadma Jena quotes official figures (of the Government of
Orissa) of families displaced due to development projects between 1950 and
1993 and of the extent of land that had to be acquired by the Government for
the purpose. Jena’s study also presents data on (a) the number of villages
affected by the Hirakud Dam that was constructed between 1948 and 1957, (b)
the number of families and people whose livelihood was disrupted due to
displacement, (c) the number of families (all belonging to Scheduled
Castes/Tribes) which were displaced forcibly by police, and (d) the number of
families resettled in rehabilitation camps47.
This aspect of development projects will cast greater responsibilities on the

respective project authorities in these matters and closer monitoring of the
implementation of resettlement plans. These will necessarily require the setting
46
The Hindu, Chennai edition of Monday, the 17th April, 2006 carried the full text of ‘the
Group of Ministers’ (GoM’s) Report on the OP-ED page.
47
The studies referred to in this paragraph and their contents are taken from the article
“Creating Dispensable Citizens” by Usha Ramanathan, in the Chennai edition of The
Hindu of Friday April 14, 2006.UshaRamanathan is an Honorary Fellow of the Centre for
the Study of Developing Societies, New Delhi. The estimates made by the studies are given
in the article. These have not been indicated here as the purpose of this Unit is only to
indicate the very few sources of data that are available on this important social aspect of
108 development and not to discuss the problem.
up of a reliable and transparent system of collection, compilation and analysis Social Sector
of statistics relating to displacement of people due to the project, their
demographic, social, cultural and economic profiles, their resettlement in
alternative livelihood, number absorbed in employment in the industrial
establishment(s) for which the lands of the oustees were acquired, details
relating to the levels of living of the oustees in the area of resettlement in terms
of parameters like incomes, access to basic needs of life like water, shelter,
education for their children and health and medicare at least until a reasonable
number of years after resettlement, the time taken and the expenditure incurred
for resettlement and so on.
24.7 ENVIRONMENT
The process of development adversely affects the environment and through it
the quality of life of society. For instance the excessive use of fertilizers and
pesticides rob the soil of its nutrients. Letting sewers and drainage and industrial
effluents without prior treatment into rivers and water bodies pollute these water
bodies causing destruction of aquatic life and endanger the health of people using
such polluted water. The recent outcry in Tirupur, near Coimbatore, a place well
known for garment exports, against untreated effluents from garment factories
being let into the river used for drinking purposes is a case in point. The exhaust
fumes containing Carbon Monoxide (CO) and lead (Pb) particles let in to the
air we breathe by vehicles using petrol or diesel is an example of air pollution.
The best example of industrial pollution through insufficient safety measures is
the Bhopal gas disaster where lethal gases leaking from a factory’s storage
cylinder killed many people immediately and maimed many others for life. The
forest cover of the country is continuously getting reduced due to
indiscriminate felling of trees leading to reduction in rainfall and changes in
rainfall pattern, besides climatic changes. The destruction of mangroves along
seacoasts for housing/tourism development often leads to soil erosion along the
coast by the sea. The adverse effects of current models of development on
environment and the realization of the need to take note of the cost to
development represented by such effects have now led to the development of
environmental economics as a new discipline in economics.
The system of national accounts currently in use throughout the world, suffers
from extreme narrowness. Vast quantities of information relevant for economic
evaluation do not appear in them. Some don't because the appropriate data are
hard, even impossible, to collect; but others don't because until recently the
theory and practice of economic evaluation didn't ask for them. The demand for
green national accounts has arisen because of a growing recognition that
contemporary national accounts are an unsatisfactory basis for economic
evaluation. The qualifier “green” signals that we should be especially
concerned about the absence of information on society's use of the natural
environment. As an extension of national accounting, the development of green
accounting framework and related researches are therefore being undertaken
around the world. The CSO, for the first time in India, have published the
Report “Green National Accounts in India” in the year 2013. You can
download a copy of this report from the Social Statistics Division of the CSO.
The Central and State Pollution Control Boards and the Ministry of
Environment and Forests (MOEF) evolve and, monitor implementation of
policies to protect the environment. Statistics on environment are collected
through this process by the agencies mentioned above and the CSO. The
109
Data Base of Indian annual reports of the MOEF 48 and the Compendium on Environment
Economy
Statistics, India 2003 published by the CSO from time to time are excellent
sources of data on environment. The latter especially is very comprehensive
and includes a very informative write up. The Compendium (and the annual
report of MOEF) can be accessed in the respective website of the two
organizations. The type of data on environment available from these
publications are mentioned below by way of illustration:
a) Ambient Air Quality Status [concentration of Sulphur di-oxide, Nitogen

di-oxide and Solid Particulate Matter49 (SPM) in air] in major cities of
India;
b) Percentage of petrol-driven two-wheelers, three wheelers and four-
wheelers meeting CO emission standards; and
c) Water quality of Yamuna river (in the Delhi stretch) in respect of selective
physio-chemical parameters between April, 1998 and March, 1999 –
dissolved oxygen (milligrams/litre), Biological Oxygen Demand (BOD)
(mg/l), faecal coliforms 50 (number/100ml), total coliforms (number/100ml)
and ammonical nitrogen (mg/l).
24.8 QUALITY OF LIFE

We have already looked at several of the factors determining the quality of life
of the people – education, health, employment, shelter and amenities and
environment. We have also referred to the plight of displaced people uprooted
from their normal life by development projects. One other factor, an important
one, is the level of income or consumption. We have already looked at the data
available on this aspect of life in the Unit on “Poverty and Inequality: Policy
Implications” in Block 1 of MEC-005 on “Indian Economic Policy” – A
compulsory course in first year. The relevant data are available from successive
quiquennial surveys of the NSSO on consumer expenditure, namely, those on
levels of consumption of different MPCE classes. That Unit also discusses
dimensions of poverty and inequalities in income (consumption) and non-
income aspects of life and about HDIs measuring shortfalls in human
development in the population, Gender Development Indices (GDI) measuring
gender discrimination, BMIs evaluating the health status of women and the
measures, Weight for Age Index Height for Age Index and Weight for Height
Index gauging the nutritional status of children. All these measures are also
available from these sources, namely, NSSO, the Planning commission and the
National Human Development Report 2001 and those of the State
Governments for judging the quality of life of the Scheduled Castes and Tribes.
The reports (prepared every year) of the Commission for Scheduled Castes and
Tribes provide (review) data on:
i) the progress in education of these sections of society;

ii) Progress in the utilization of the reservation quotas in different services of
the Governments and the public sector undertakings and the efficacy of
measures taken by Government in enabling members of the weaker
sections of society in gaining access to these opportunities;
48
The latest report of 2013-14 can be accessed at http://www.moef.nic.in/sites/default/files/
annual_report/AR-2013-14-Eng.pdf.
49
SPM consists of metallic oxide silicon, calcium and other deleterious metals.
50
The most common contamination in water is from disease-bearing human wastes which is
110 usually detected by measuring faecal coliform levels.
iii) enforcement of laws like the Untouchability Act; and Social Sector
iv) difficulties of these sections of society – ill-treatment, practice of

untouchability, discrimination and lack of access to financial and technical
assistance to entrepreneurs belonging to these classes.
The Ministry of Social Justice and Empowerment, the Ministry of Tribal

Affairs, the National Commission of Backward Classes (NCBC), the State
Backward Class Commissions, the Minorities Commission, the National
Commission on Women and the State Commissions on Women also conduct
surveys, bring together and review a large amount of data on the progress of or
the lack of progress in curbing exploitation of, and violence against these
sections of society, raising the shares in education and employment of these
communities and groups and in empowering and bringing these sections into
the mainstream of life. These are comprehensive sources of data on the efforts
being made and need to be made to empower these sections of society and
elevate their status in society. The Commission for the Aged (Senior Citizens)
and the Commission for Children are sources of data on these sections gathered
at one place from various primary sources51 and analysed to highlight issues
that need attention. The Ministry of Social Justice and Empowerment (Annual
Report of the Ministry), the nodal agency that is entrusted with the
responsibility for the development and empowerment of the physically and
mentally challenged, the weaker sections of society and women and the
development of children and the welfare of the aged, constitute another source
of data on the status of these vulnerable groups in society.
1) What kind of data/information is provided by NSSO on housing conditions

in India?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
2) Do you think that the data on registrants with the employment exchanges
provide adequate information about the persons displayed by development
projects? Give reasons in support of your answer.
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
51
For instance, for the aged, important sources of data are (a) the NSSO Report No. 446 (52nd
Round) “The Aged in India – A Socio-economic Profile: 1995-96”. The Census 2001 is a mine
of information on the Aged as it presents a number of social, economic and cultural
characteristics by age groups and gender, one such group being 60+. Data on women and
children and the disabled also flow from the Census and other sources of data for these grops
have already been referred to elsewhere in this unit. 111
Data Base of Indian 3) Name the documents, which contain the statistics on environment.
Economy
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
4) Which department/ministry has been assigned the responsibility to provide
information data on development and empowerment of weaker sections of
the society?
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
………………………………………………………………………………
24.9 LET US SUM UP

We have tried to enumerate the various sources of data in the social sector and
looked at the kind of data available in different sub-sectors of the social sector.
We have seen that there are a number of sources of data on employment
covering parts of the economy. There are two sources that are comprehensive
and cover the entire economy. These enable an analysis of trends in
employment and unemployment over the years, the industrial, occupational and
educational composition of the employed and their age profile, the educational
profile and the age structure of the unemployed, the quality and adequacy of
employment in terms of wages, tenure of employment and employment
security of workers. Similar analysis is possible for several subsections of the
economy and society that are relevant to the country’s objectives of economic
and social policies. The database on welfare of labour is somewhat incomplete
in some areas, calling for stricter monitoring of the submission of statutory
returns and the implementation of existing legislation. Enormous amount of
data on educational infrastructure, the utilization of such facilities and the
output of the educational system are available and periodic comprehensive
surveys add to the richness of the database. However, there is a need to go into
such discrepancies as may be there between data collected by the reporting
system in the educational administration and the data thrown up by the surveys,
to initiate corrective action to tone up the large database built up over the years.
Another important need is to reduce the time lag in the availability of data for
policy makers as well as researchers. Data on health are being built up from
more than one source, namely, the Registrar General of India on aspects like
vital rates such as birth and death rates, IMR, MMR and morbidity rates and
life expectancy, while the CBHI collects and brings together data from the
different wings of the Ministry of Health and the State Health Departments and
the different health professional councils. Health and Medicare and health
situation data are comprehensive in terms of item coverage but time lags need
to be reduced so that data relating a particular time point is complete in terms
of coverage of all States and Union Territories. Data from NFHS is a valuable
112
addition to the database on health, especially from the point of view of health Social Sector
and nutrition status of women and children. The database on shelter and
amenities has been updated recently by the Census 2011 and the NSSO survey
of the 69th round (July – December, 2012) and also includes specific data on
urban slums. A database on environment – air pollution, soil degradation and
noise pollution – is getting built up steadily. Data on levels of living, both in
terms of income and non-income aspects of life and for the general population
as well as for the weaker sections of society can be derived from existing data
and used as a basis for planning programmes to reduce poverty and inequalities.
We have also noted a grey area in the data systems we have considered. There
is no data worth the name on people displaced from their livelihoods by
development projects and the number among these who have been resettled in
alternative livelihood. There is a growing realization, at least among some
sections of development analysts and thinkers that the burden of the costs of
development is invariably borne mostly by the poor and the voiceless. There is
an urgent need to build up a transparent database on people displaced by
development projects and their resettlement and the basic amenities made
available to them.
Ministry of Environment and Forest : State of Environment Report, India

(2009)
Ministry of Statistics &Programme : Green National Accounts in India – A
Implementation (2013) Framework
Ministry of Labour and Employment : Indian Labour Yearbook 2009 and 2010
(2010)

PROGRESS EXERCISES
1) i) Employment Market Information (EMI) Programme of Directorate

General of Employment and Training (DGE&T), Ministry of
Labour
ii) State Directorates of Employments (DES)
iii) Economic Census
iv) Annual Survey of Industries
v) Population Census
vi) Quininquennial Sample Surveys on employment and unemployment
conducted by NSSO.
3) Usual Principal Status (UPS), Usual Principal and Subsidiary Status (UPSS),
Current Weekly Status (CWS) and the Current Daily Status (CDS).
113
Economy
1) The department of School Education and higher education of Ministry of
Human Resource Development, NCAER, UGC, Institute of Applied
Manpower Research etc.
2) See footnote no. 29
3) Do yourself

1) See Section 24.4
2) State Level estimates of demographic and health parameters and the data
on various socio-economic factors crucial for bringing changes in
demographic and health situations.
3) Composite measure that takes into account both chronic and acute
malnutrition.

1) See Section 24.5
2) See Section 24.6
3) The annual Reports of Ministry of Environment and Forest and the
compendium of environment statistics.
4) Ministry of Social Justice and Empowerment (annual reports of the
Ministry).
114

MEC - 109 Research Methods in Economics PDF

Uploaded by

Copyright:

Available Formats

MEC - 109 Research Methods in Economics PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MEC - 109 Research Methods in Economics PDF

Uploaded by

Copyright:

Available Formats

MEC-109

After completion of Master’s degree in Economics, many of you may intend to

Block 1: This block deals with the theoretical perspectives/foundations guiding

1.2 RESEARCH METHODOLOGY AND ITS

Research Methods, comprises of research techniques and tools. Research

Thus, methodology, methods and tools/techniques are three distinct elements of

1.3 THEORETICAL PERSPECTIVES

Realism Constructivism Pragmatism

• Constructivism begins by assuming that everyone has unique experience

• Pragmatism considers realism and constructivism as two alternate ways to

The characteristics of positivism and post positivism as theoretical and

1.4 APPROACHES TO SOCIAL ENQUIRY

Mixed methods research by combining quantitative methods and qualitative

Quantitative approach can be further sub-classified into: inferential approach;

In Inferential approach, database is established through survey method and

Qualitative approach deals with the subjective assessment of attitudes, opinions

Table 1.1: Typical Characteristics of Qualitative and Quantitative Approach

1.5 RESEARCH STRATEGIES

Each strategy has different starting points (Lewis-Beck, 2004):

Check Your Progress 1

1.6 RESEARCH PROCESS

Defining Research Problem

Collection of Data: Source, tools, and methods

Fig. 1.1: Flowchart of Research Process

a) Defining the research problem: Selecting and properly defining the

1) rationale behind the research problem;

4) developing the ideas through discussions; and

5) re-phrasing the research problems identified in (i) above into a working

The different steps to be followed while defining the research problem,

• statement of the problem first in a general way to be later sharpened

• understanding the nature of the problem, and

• surveying the available literature.

• Technical terms and words or phrases used in the research problem

• Basic assumptions or postulates relating to the research problem need

• A clear and unambiguous statement of the investigation should be

• The sources of data and its limitations must be explicitly mentioned.

c) Formulation of Hypothesis: Specification of working hypothesis (or

d) Research Design: Research design is the logical conceptual structure within

e) Collection of Data: Collection of data is an essential part of the research

f) Analysis of Data: Analysis involves steps like categorization, coding,

Processing of Data Analysis of Data

Editing Descriptive and Inferential analysis/

Bivariate analysis Multi – variate analysis

1.7 HYPOTHESIS: ITS TYPES AND SOURCES

1.8 THE NATURE, SOURCES AND TYPES OF DATA

1) Time Series Data: It is a set of observations on the values that a variable

1.9 MEASUREMENT SCALES OF VARIABLES

Check Your Progress 2

1.10 LET US SUM UP

1.11 KEY WORDS

1.12 SOME USEFUL BOOKS

1.13 ANSWERS OR HINTS TO CHECK YOUR

Check Your Progress 2

KNOWLEDGE: POSITIVISM AND

History of Science is a study of the development of scientific ideas and institutions

It is obvious that answers to questions such as these are of importance in figuring

2.2 POSITIVIST PHILOSOPHY OF SCIENCE

2.2.1 Central Tenets of Positivist Philosophy

Where L1……………………………………..Ln is a set of laws, I2………………………………In is a