01 Small-Area Population Estimations and Projections

Demographic Techniques:
Small-Area Estimations
and Projections
KOSTAT-DANE Collaborative Project on Strengthening Statistical
Capacity of Colombia
APPI (Asia-Pacific Population Institute)
Expert Team
jkh96@cnu.ac.kr
A. Small-area
Demography
Introduction
• Estimates and projections for small areas are used extensively
in the public and private sectors
• Because of population size and data availability issues, small-
area methods areas face methodological challenges not
commonly encountered at larger geographical scales
• In this lecture,
 define “smallness” for estimation and projection purposes;
 describe the types and sources of data used;
 discuss the methods for estimating and projecting small-area
populations;
 assesse recent methodological developments, in particular the impact
of GIS and spatial techniques.
2/16/2018 Small area estimations and projections 3

Estimate, Projection, and Forecast
• Distinction of the terminology

 Estimate: information about the past or current population, not based
on a census or population register
 Projection/Forecast: information about the future population
• The terms projection/forecast are used without any distinctions,
but should be differentiated according to the expected likelihood
of their outcomes:
 Projection: a numerical outcome of a particular set of assumptions
regarding the future population
 Forecast: a projection believed most likely to provide an accurate
prediction of the future population

Estimate, Projection, and Forecast-
Continued
• In addition to the time dimension, a key distinction between
estimates and projections involves the type of data used.
• Since projections refer to the size of the population at a future
point in time, they cannot be based on actual data comprising
the components of population change.
• In this regard, projections must be based on the extension of
either current or expected population trends into the future

How can “smallness” be defined?
• No universally accepted definition exists for what constitutes a

small area for population estimation and projection purposes.
However, it commonly involves administrative units below the
national level.
• Examples: states, counties, cities, municipalities, townships, wards,
local government areas, postal areas, school districts, census tracts,
and census blocks
• Alternatively, a small area can refer to any subpopulation or
domain for which direct estimates of adequate precision cannot
be produced. Subpopulations or domains are subsets of larger
populations defined by criteria other than geographic affiliation
such as age, sex, income, educational level, health status, etc.
• Smallness can thus be understood in terms of size or in terms
of data availability.

Who produce subnational, or small-area
estimates and projections?
• A variety of government agencies, research institutes, and private
businesses produce subnational population projections.
• In the US, the Census Bureau makes projections for states; most
state governments (or their designees) make projections for counties
in their states; and many local and regional governments make
projections for cities, census tracts, block groups, and other small
areas.
• Private businesses make (or compile from other sources) projections
for states, counties, subcounty areas, and a variety of customized
geographic areas and demographic subgroups.
• Subnational projections have become increasingly common over the
last few decades, especially for small areas. Similar trends have
occurred in other countries, including Australia, Canada, India, Israel,
New Zealand, and virtually all countries in Europe.
•

Challenges Facing Small-Area Demography
• The boundaries of small areas often change over time, thus making
time series analyses difficult.
• Many types of data, in particular those covering more detailed
population characteristics, are not tabulated for smaller areas,
necessitating the use of proxy variables
• Because of a paucity of data, there are often no discernible past
patterns of change that can serve as a basis for estimation or
projection, which may require the application of model rates based on
areas for which data are available but which may not be directly
comparable.
• Even when data are available for small areas, they may be less
reliable because of smaller sample sizes and greater sampling
variability.
• Location-specific factors such as institutional populations, seasonal
populations, facility closings, or changes in zoning have a greater
impact on population changes in small area.

South Korea: Administrative Division

Colombia: Administrative Division

USA: Administrative Division

China:
Administra
tive
Division
2/16/2018
Small area estimations and projections 12

Note: Official statistics
• Official statistics are statistics published by government

agencies or other public bodies such as international
organizations as a public good
• They provide quantitative or qualitative information on all major
areas of citizens' lives, such as population living conditions,
health, education, environment, and development
• Of course, governmental agencies at all levels, including
municipal, county, and state administrations, may generate
and disseminate official statistics
• Official statistics should be objective and easily accessible and
produced on a continuing basis so that measurement of
change is possible.

Fundamen
tal
Principles
of National
Official
Statistics

Two Categories of Data for Small-Area
Demography
• Two data categories are distinguished: direct data and indirect
(symptomatic) data.
 Direct data: censuses, administrative records, surveys, etc.
 Indirect data: infant vaccination school enrollment, tax returns, vital
statistics, employment statistics, voter registration, electrical hook-ups,
housing counts, etc.
• There is no clear distinction between the two categories of data;
the same data can be direct for one type of estimate and
indirect for another.
• For example, birth and death counts are direct data when used
to estimate natural change in a population but indirect data
when used in the census-ratio method to estimate total
population.
• The usefulness of indirect data depends on the extent to which
factors other than population size and distribution influence
them.
Censuses as the Primary Source of
Small-Area Demography
• Essential features of a population census: universality,
individuality, periodicity, and simultaneity (UN Principles and
Recommendation for Population and Housing Censuses, 3rd
Revision)
• In general, Censuses provide accurate and comprehensive
data, regular repetition, fine geographic detail, and relative
ease of data access; potential downsides include cost,
enumeration difficulties, infrequent updates, falling response
rates, differential under-enumeration, and concerns about
disclosure control.
• While censuses are still generally considered the “gold
standard,” in many countries there has been a movement
towards greater utilization and application of administrative
data sources and sample surveys to provide more timely data.

2015
Population
and Housing
Censuses in
Korea
2/16/2018
Small area estimations and projections 17

Administrative Records as the Source for
Small-Area Demography
• Administrative records, which are collected by national,
departmental, and local government agencies for registration,
licensing, and program administration purposes, can provide
ongoing information on a variety of demographic events and
characteristics.
• The most important administrative records for small area
estimates and projections are vital statistics (especially birth
and death data), which tend to be widely available and
generally quite accurate, at least in the developed countries,
and which form crucial inputs in many estimation and projection
models.
• Other sources of administrative records come from a variety of
fields and cover data related to the economy, education, health,
social services, safety and security, community resources and
participation, housing, and the environment.

Note: History of Vital Statistics System in
Korea
• 1948~55: Produced vital statistics via questionnaires different from the form of
register by the National Bureau of Statistics
• 1962: Statistics Law and Vital Statistics regulation (Ordinance by the Economic
Planning Board)
• 1970: Integrated topics for Family Registration and Vital Statistics System
• 1980: Began issuing annual report for Vital Statistics (containing data from 1970
to 1979)
• 1982: Began issuing Annual Report for Cause of Death Statistics
• 1997: Used local input method with the development of Vital Statistics Report
System
• 2004: Converted the report system of Vital Statistics into the Web-based report
system
• 2007: Enacted the regulation of Vital Statistics
• 2008: The Family Register System changed to the family relations registration
system; converted to C/S-based vital events registration input system

Population Register
• Population registers are a special case of administrative records:

• Universal registers: ones which include the entire population
• Partial registers: ones which are established for specific administrative
purposes and cover those directly affected by the particular program (e.g.,
infant vaccination, school enrollment, military conscription, voter
registration, etc.)
• A universal population register has the potential to deliver timely, cost-
effective, and precise population estimates even for small
geographical units, but a number of challenges –ranging from
confidentiality and privacy, up-front costs, to accuracy and technical
issues – must be overcome for the system to become established.
• To be most useful, population registers must be accurate and up-to-
date, allow for linking from one source to another via a personal
identification number, and be embedded in a supportive legislative
framework

National Population Register of India

Survey as the Source for Small-Area
Demography
• Surveys have their central features which are the use of a fixed
design, the collection of data in standardized form from
individuals or organizations such as schools or businesses, and
the selection of representative samples from known populations.
• Surveys are often used to collect data on variables not covered
in a census or by administrative records. Surveys come from
governmental/official statistics, academic/social research, and
commercial/advertising/market research.

Survey…-Continued
• Small sample sizes often limit the utility of survey estimates for
small areas. This can be the case even for surveys designed
specifically to provide accurate and timely demographic, social,
and economic data on an ongoing basis for large and small
areas, such as the American Community Survey in the United
States.
• To combat the challenges posed by small sizes, small area
estimation techniques have been developed that apply indirect
estimators which “borrow strength” by using values of the
variable of interest from related areas and/or time periods, thus
increasing “effective” sample sizes.

Colombia:
National
Household
Survey
Small area estimations and projections 2/16/2018 24

Colombia:
Demographi
c and Health
Survey (DHS)

American
Community
Survey
(ACS) in the
US

B. Demographic
Estimation
Types of Population Estimation
• Based on time reference and derivation methods, population

estimates can be divided into:
 intercensal estimates, which relate to a date between two censuses
and take the results of these censuses into account,
 postcensal estimates, which relate to a date following a census that
take that census into account, but not later censuses.
• Intercensal estimates can be regarded as interpolations, and
postcensal estimates as extrapolations. Estimates can also be
made for dates prior to census taking (precensal), which are of
interest to historical demographers in particular.
• The estimation methods are primarily used for making
postcensal estimates, which are the most common type of
population estimation. Intercensal and precensal estimates can
be made with some of these methods but usually require a
different approach.

Types of Population Estimation-
Continued
• Population estimates can further be divided whether they are
made for a legally resident de jure population or for a physically
present de facto population.
• Most population estimates follow a de jure definition, and the
methods described below are primarily applicable to them.
• De facto population estimates, which are useful for estimating
daytime, visitor, seasonal, homeless, and disaster-impacted
populations, for the most part utilize different sets of techniques

Importance of Data Sources for
Population Estimation
• The type and quality of data available are crucial determining
factors when choosing a method for population estimation.
• Estimation accuracy, and how different methods account for
uncertainty, are important factors to consider.
• Additional criteria include the provision of necessary detail, face
validity, plausibility, costs of production, timeliness, and ease of
application and explanation

Extrapolation techniques
• Extrapolation techniques, which range from simple linear

change to complex ARIMA models, rely solely on the pattern of
past population changes to estimate postcensal population, and
assume that trends in the postcensal period will be similar to
historical trends.
• Simple extrapolation models are most useful for postcensal
estimates close to the last census, when resources are limited,
and for very small areas or demographic subgroups; complex
models allow the construction of probabilistic intervals around
the estimates but are more difficult to implement and not
necessarily more accurate.

Ratio extrapolation methods
• Ratio extrapolation methods express the population of a

subgroup as a proportion of a larger population.
• Share-of-growth,
• Shift-share,
• Constant-share,
• Like simple extrapolation techniques, they have small data
requirements and are easy to apply, but share the general
shortcomings of extrapolation methods in that they do not
account for differences in demographic characteristics or the
components of growth, and they ignore potentially relevant
information related to postcensal population changes

Symptomatic Estimation techniques
• Symptomatic estimation techniques include

 the housing unit methods
 the censal-ratio methods.

Housing unit method
• The housing unit method relies on the assumption that nearly
everyone in a population lives in some type of housing structure.
• The housing unit method calculates the population of an area as
equal to the number of occupied housing units (households) times the
average number of persons per household plus the number of
persons living in group quarters (e.g. prisons, college dormitories,
military barracks, nursing homes).
• The number of households can be estimated using measures of
construction activity such as building permits or certificates of
occupancy, using utility data such as a residential electric or telephone
customers, from property tax records, and from aerial photographs;
• The average number of persons per household can be taken from the
most recent census, extrapolated as a trend from the two most recent
censuses, or estimated using post-censal data in combination with
data from the last census;
• The group quarters population can be obtained either directly from the
facilities or from a past census.

Censal-ratio
method
using
housing
permits

Censal-ratio method
• The censal-ratio method, introduced as the “vital rates
method”, is related to the ratio extrapolation methods, but is
based on ratios of symptomatic data to total population
rather than proportions of national, state, or regional totals.
• The technique involves computing the ratio of symptomatic
data to total population at the time of the last census,
extrapolating the ratio to the estimate date, and dividing the
estimated ratio into the value from the symptomatic series for
the estimate date.
• Symptomatic data that can be used in the censal-ratio
method include birth and death statistics, school enrollment
data, tax returns, number of electric, gas, or water meter
accounts, number of building permits issued, bank receipts,
motor vehicle registrations, and voter registration rolls.

Censal Ratio
method
using
symptomatic
data

Regression techniques
• Regression techniques derive population estimates by means of
symptomatic indicators of population change. The ratio-correlation
method is the most widely-used regression technique for population
estimation.
• Ratio-correlation method involves relating changes in several
symptomatic indicators to population changes – expressed in the form
of ratios to totals for geographic areas – by a multiple regression
equation. The symptomatic variables used are similar to those for the
censal-ratio method.
• Regression techniques for population estimation have a firm
foundation in statistical inference, which allows for the construction of
meaningful measures of uncertainty.
• Regression techniques rely for their accuracy on the validity of the
assumption that the relationship between the independent and
dependent variables observed in the past will persist in the post-
censal period; they also require judgment with regard to the reliability
and consistency of coverage of the symptomatic indicators.

Estimates of population for the counties in
Colorado, US: 2000-2010 using regression
techniques

Component methods
• Component methods are based on the demographic balancing

equation
𝑃2 = P1 + B − D + IN − OM
= 𝑃1 + 𝐵 − 𝐷 + 𝑁𝑀
• All component methods generally employ birth and death data

but vary in how migration is estimated.

Component methods-Continued
• The component method-II and the tax returns/administrative

records method developed by the US Census Bureau are
widely used .
• The component method-II employs changes in school
enrollments to estimate migration, whereas the tax
returns/administrative records method uses address changes
on federal income tax returns.
• If migration data of sufficient quality are available, the cohort-
component method can also be used

Component methods-Continued
• Component method divides the population at the starting date
into age-sex groups (cohorts) and accounts separately for the
fertility, mortality, and migration each cohort experiences until
the estimate date.
• Although useful when estimates by age and sex are required,
the cohort-component method is data- and computationally
intensive and more commonly applied for population projections
than for estimates.
• Component methods are attractive, because they specifically
account for the three components of population change, births,
deaths, and migration.
• Limitations include that they assume the continuation of
historical patterns in symptomatic data in measuring migration,
that they can be resource-intense and require data which may
not be available for all small areas, and that they may not be
appropriate for areas with substantial “special” populations.

Sample-based methods
• Sample based methods of population estimation, which include
synthetic methods, structure preserving estimation (SPREE),
the ranked set samples method (RSS), and Bayesian methods,
are more commonly employed by statisticians than by
demographers.
• Other methods that are sometimes used for creating small area
estimates include structural models, economic-demographic
models, dual system estimation, microsimulation models, neural
networks, the grouped answer method, social network analysis,
and various methods related to spatial demography.
• A local census, which provides a direct count of small area
populations independent of any other data source, can also be
used, though it is expensive and its accuracy dependent on the
response rate achieved.

Composite methods
• The composite method, developed by Bogue and Duncan

(1959), is a “portfolio” of separate estimation methods that are
tailored to particular segments of the population such as age
groups.
• (censal-ratio method using births for the population aged 0–4)
+(component method using school enrollment data for the population
aged 5–17) + the censal-ratio method using births and deaths for the
population aged 18–64) + (Medicare data for the population aged 65
and older)
• The estimates for each age group are then summed up to yield
an estimate of total population.
• Many combinations of methods and data are available, though
there exists
• little empirical guidance in terms of which to choose

Combining Estimates
• Another approach is to combine estimates made with two or more
different estimation techniques.
• In its most simple way, this involves averaging or weighting the
estimates directly; alternatively, one can join estimates from
sample surveys and estimates based on demographic analysis as
independent variables in a regression model.
• There is evidence suggesting that combining estimates made with
different methods, data, and assumptions improves accuracy and
reduces biases.

Recent Trends
• Recently, research has been devoted to small area population
estimation methods that utilize remote sensing and GIS
technologies. Dasymetric mapping, once confined to visualizing
population distribution, is also increasingly being used for
estimating the population of small areas.
• These methods are attractive since they are not tied to the
administrative boundaries associated with census or survey
data and allow population estimates to be made for user
defined areas.
• They also show potential for areas where census data and
administrative records are unavailable, unreliable, or out of date,
as is the case in many countries in the developing world.
• Yet, there currently exists a disconnect between academic
research and the actual adoption and use of remotely sensed
technologies for estimation purposes, with lack of expertise,
technical, application, and financial issues preventing more
widespread adoption by practitioners.
GIS Software to
Remodel Population
Data using
Dasymetric Methods
Method of thematic mapping,

which uses areal symbols to
spatially classify volumetric
data.
47 2/16/2018 Small area estimations and projections

Flowchart of
Population
Estimation
Using Remote
Sensing and
GIS
Technology
Section C: Demographic
Projection
Methods for Population Projection
• Many of the methods used for producing population estimates

can also be applied when preparing population projections and
the sources of data and factors to be considered when
choosing among methods are similar as well.
• However, there are important differences since estimates are
concerned with giving an accurate population count for a past
or present date, whereas projections require that assumptions
about future population changes be made. Projections involve
more uncertainty, and projection methods vary widely in how
they attempt to model it.
• As was the case with estimates, there are various ways to
classify projection methods. For small areas, four types of
projection models recommended are: trend extrapolation,
cohort-component, structural, and microsimulation models.

Trend extrapolation
• Trend extrapolation models determine future population values
solely by their historical values; they range in complexity from
simple methods such as linear, geometric, and exponential
growth to complex polynomial, logistic curve fitting, and ARIMA
time series models.
• Ratio methods – in which the population or population change
of a small area is expressed as a proportion of population or
population change of a larger area in which the smaller area is
located – is another type of extrapolation that is often used.
• Recent research developments involve the application of ratio
methods in combination with GIS techniques to produce small
area projections based on grids.

Domestic Population Forecasting:
Logistic Curve Method

ARIMA model
• Given a time series of data Xt , the ARMA model is a tool for
understanding and, perhaps, predicting future values in this
series.
• The model consists of two parts, an autoregressive (AR) part
and a moving average (MA) part.
• The AR part involves regressing the variable on its own lagged (i.e.,
past) values.
• The MA part involves modeling the error term as a linear combination
of error terms occurring contemporaneously and at various times in the
past.
• The model is usually referred to as the ARMA(p,q) model where
p is the order of the autoregressive part and q is the order of
the moving average part

Pakistan: Projected population using
different ARIMA models
FORECASTING THE POPULATION OF PAKISTAN USING ARIMA MODELS

Muhammad Zakria* and Faqir Muhammad
Department of Mathematics and Statistics, Allama Iqbal Open University, Islamabad
*Corresponding author’s e-mail: zakriauaf@yahoo.com
Trend extrapolation-Continued
• Trend extrapolation models have a number of shortcomings:
• do not account for differences in demographic composition or for
differences in the components of growth,
• provide little or no information on the projected demographic
characteristics of the population,
• have no theoretical content and generally cannot be related to theories
of population change.
• On the other hand, data requirements, ease of application,
timeliness, and low cost of simple trend and ratio models make
them an attractive option for small areas where data availability
and reliability issues may prevent the application of more
complex models.
• Furthermore, simple trend models often provide fairly accurate
projections for short and even long horizons, and the empirical
evidence suggests that more complex and sophisticated
models do not offer more accurate projections, at least for total
population.

Cohort-component models
• Cohort-component models are the most commonly used

population projection method, taking separately into
consideration the three components of demographic change:
births, deaths, and migration.
• Most models subdivide the population into age and sex groups,
and each demographic subgroup is projected separately;
further subdivisions by race, ethnicity, or other demographic
characteristics are possible.
• While assumptions about the future need to be made for all
three components of population change in a cohort-component
model, migration tends to be the most significant, most volatile,
and hardest to predict for small areas.

An Example
of Cohort
Component
Method for
Simulating
Population
Growth
Cohort-component models-Continued
• Approaches for projecting migration include methods which use

only base period migration data, methods which require some
additional qualitative or quantitative information, and methods
based on quantitative projections of independent variables.
• Cohort-component models vary further in whether they apply
gross or net migration data. Gross migration models range from
complex migration pool and multi-regional models that are
primarily useful for larger areas to simpler bi-regional models
that are easier to apply for small areas.
• Net migration models can follow a top-down or bottom-up
approach; a simplified version proposed by Hamilton and Perry
that treats mortality and migration as a single unit is also
sometimes used

Cohort-component models-Continued
• The cohort-component method is a very popular projection

method because it can incorporate many different data sources,
assumptions, and application techniques.
• It is well suited to provide projections of demographic
characteristics in addition to projections of total population.
• However, the method is also very data intensive and relatively
expensive to apply, and the potential lack of complete and
reliable data can be a serious obstacle when projecting the
population for small areas.

Structural models
• There are two major types of structural models:

• economic-demographic models
• urban system models
• Structural models form the third major group of projection
methods, together with trend extrapolation and cohort-
component projection models.
• In a structural model, population change is related to changes
in one or more explanatory variables, and the focus is generally
on modeling migration.

Economic-demographic models
• Examples of the former include econometric models, models
that balance labor supply and demand, models based on
population/employment ratios, and regional economic models.
• Economic-demographic models generally focus on economic
variables to project migration, though some include amenities
as well.
• Most models are complex, resource intensive, expensive to
develop and apply, and primarily used for relatively large areas
such as counties, metropolitan areas, states, and nations.

Urban models
• For smaller areas, urban system models are more commonly
applied; these typically include land use and transportation
characteristics in addition to economic variables and rely
heavily on GIS techniques.
• Like most economic-demographic models, urban systems
models tend to be very resource intensive and often require a
substantial degree of technical expertise.

Additional approaches-Housing unit
method
• The housing unit method, though more commonly applied to
prepare population estimates, also shows potential for small
area population projections.
• The method takes into account local housing supply and, either
explicitly or implicitly, residential land availability, and can be
used to constrain cohort-component projections.
• This approach promises to be particularly useful for growing
urban areas and areas earmarked for residential development,
though projecting its various components is challenging.

Additional approaches-Microsimulation
models
• Microsimulation models differ from traditional demographic
modeling as follows
• they use a sample rather than the total population;
• they work on the level of the individual or the household rather than
with grouped data;
• they rely on repeated random experiments to derive a projection.
• By modeling individual behavior, spatial microsimulation
models avoid aggregation bias, are internally consistent,
provide very detailed projection outputs, and allow for
scenario and what-if analyses.
• On the other hand, the microsimulation models tend to be
extremely complex, require extensive data and staff
resources, and there have been few attempts to validate the
assumptions and outcomes of these models

Additional approach-Spatial diffusion and
spatial-temporal models
• Originating in the geographic concept of spatial diffusion – the
spread of a particular phenomenon over space and time –
spatial diffusion and spatial-temporal projection models
incorporate population spillovers and legacy effects arising from
neighboring growth rates and characteristics into the model.
• While these methods address a major shortcoming of most
projection models used for small areas – which treat each unit
of geography as an independent, stand-alone entity rather than
as an entity surrounded by other areas with which they interact
– they have not been found to outperform forecasts made with
simple extrapolation methods

Synthesis of approaches: Combining
projections
• Combining projections using simple averages, weighted
averages, trimmed means, and composite approaches, is
another option.
• Combined projections are ideally derived from methods that
differ substantially from one another and that draw from
different sources of information; combining has been found to
improve forecast accuracy under many.

Synthesis of approaches-Continued
• in addition to projections of total population and by demographic
characteristics, many types of planning, budgeting, and analysis
also require projections of households, school enrollment,
employment, health, poverty, or other population-related
characteristics.
• While structural models are often applied for these purposes, the
participation-ratio and cohort-progression methods are worthwhile
to consider. In the former, socioeconomic characteristics are
related to demographic characteristics through the use of ratios; in
the latter, projections are developed by “surviving” people with
particular socioeconomic characteristics.

Addendum: Probabilistic Projection
Methods
• Among the most important advances in recent years has been
the development of probabilistic projection methods.
• Probabilistic methods acknowledge that forecasting the
population entails a significant amount of uncertainty, especially
for longer time periods, for places with small or rapidly changing
populations, for certain age groups, and for the demographic
components of population change.
• Conventional deterministic projections provide a single
population number, or high and low variants that do not attach
probability to the high-low ranges

Probabilistic Projection Methods-
Continued
• Probabilistic methods, in contrast, assess and
communicate uncertainty by providing forecasts that
come with probability distributions or that are fully
probabilistic, i.e. generated by probabilistic population
renewal.
• Predictive intervals are commonly made with three
distinct approaches:
 Time-series analyses;
 Expert-based probabilistic projections
 Ex post analysis, i.e. the use of past forecast errors.

Probabilistic Projection Methods-
Continued
• Most of the research on probabilistic projection methods has

focused on large geographic areas such as nations, states, and
regions.
• Yet a probabilistic approach is arguably even more important for
small areas, given the greater uncertainty of future population
changes at the subnational level.
• Articles read at the international workshops and conferences
discuss some of the challenges of providing probabilistic
projections for small areas and offer directions for further
methodological development.

Small area: from administrative division
to user-defined area (?)
• Alho, Juha M. (2001). “with the development of the geographic
information system (GIS), small area estimation and forecasting
will no longer be limited by administrative boundaries.”
• Arguing that administrative boundaries are frequently not well
suited to many types of social research, and that domains of
interest may intersect several areas or be too small for their
effect to be discernible in small area data, the introduction of
GIS techniques was viewed as a major step forward for small
area estimation and forecasting.
• Other experts also predicted a bright future for GIS in small
area analysis, but argued that while GIS techniques may be
useful when combined with remote sensing to develop small-
area population estimates, for projections the main value of GIS
would likely be in developing databases rather than in
constructing projections per se.
Small-area estimates and projections,
Continued
• Recently, research has focused on population estimation using
remote sensing and GIS techniques. Not only have these
techniques created new and alternative means of acquiring
population estimates for small areas previously estimated with
traditional methods, they also allow estimates to be made for
user defined areas that were hard to come by in the past.
• On the other hand, the literature on projections made using GIS
techniques remains sparse. Projections require assumptions
about the future, and while GIS and spatial techniques can add
to our understanding of past changes and present population
distributions, they do not make forecasting them easier.
• It seems likely that for the foreseeable future GIS and spatial
techniques will remain most useful as tools for distributing
populations projected with established demographic models to
areas smaller than, or different from, those they were originally
made for, rather than providing an alternative projection
approach.
Continued
• Most small area estimates and projections are still made for
administrative units, and a major shift towards user defined
areas is unlikely to happen soon.
• Most of the required input data are still collected for
administrative units, and methods that allow for customized
bottom-up aggregations – such as microsimulation models – or
methods that transcend administrative boundaries – such as
remote sensing and GIS techniques – have not yet replaced
estimates and projections made with established demographic
models.

Continued
• Administrative units can be inconvenient to work with, but they
often represent meaningful entities for which knowledge about
past and present population changes exists.
• While small area population estimates and projections will
always contain an element of uncertainty, this professional
knowledge, when combined with a careful selection of data and
methods, still provides the best foundation for satisfactory
estimation and forecasting outcomes.


01 Small-Area Population Estimations and Projections

Uploaded by

Copyright:

Available Formats

01 Small-Area Population Estimations and Projections

Uploaded by

Document Information

Original Description:

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

01 Small-Area Population Estimations and Projections

Uploaded by

Copyright:

Available Formats

Demographic Techniques:

2/16/2018 Small area estimations and projections 3

• Distinction of the terminology

2/16/2018 Small area estimations and projections 4

2/16/2018 Small area estimations and projections 5

• No universally accepted definition exists for what constitutes a

2/16/2018 Small area estimations and projections 6

2/16/2018 Small area estimations and projections 7

2/16/2018 Small area estimations and projections 8

2/16/2018 Small area estimations and projections 9

2/16/2018 Small area estimations and projections 10

2/16/2018 Small area estimations and projections 11

Small area estimations and projections 12

• Official statistics are statistics published by government

2/16/2018 Small area estimations and projections 13

2/16/2018 Small area estimations and projections 14

2/16/2018 Small area estimations and projections 16

Small area estimations and projections 17

2/16/2018 Small area estimations and projections 18

2/16/2018 Small area estimations and projections 19

• Population registers are a special case of administrative records:

2/16/2018 Small area estimations and projections 20

2/16/2018 Small area estimations and projections 21

2/16/2018 Small area estimations and projections 22

2/16/2018 Small area estimations and projections 23

Small area estimations and projections 2/16/2018 24

Small area estimations and projections 2/16/2018 25

Small area estimations and projections 2/16/2018 26

• Based on time reference and derivation methods, population

2/16/2018 Small area estimations and projections 28

2/16/2018 Small area estimations and projections 29

2/16/2018 Small area estimations and projections 30

• Extrapolation techniques, which range from simple linear

2/16/2018 Small area estimations and projections 31

• Ratio extrapolation methods express the population of a

2/16/2018 Small area estimations and projections 32

• Symptomatic estimation techniques include

2/16/2018 Small area estimations and projections 33

2/16/2018 Small area estimations and projections 34

Small area estimations and projections 2/16/2018 35

2/16/2018 Small area estimations and projections 36

Small area estimations and projections 2/16/2018 37

2/16/2018 Small area estimations and projections 38

2/16/2018 Small area estimations and projections 39

• Component methods are based on the demographic balancing

• All component methods generally employ birth and death data

2/16/2018 Small area estimations and projections 40

• The component method-II and the tax returns/administrative

2/16/2018 Small area estimations and projections 41

2/16/2018 Small area estimations and projections 42

2/16/2018 Small area estimations and projections 43

• The composite method, developed by Bogue and Duncan

2/16/2018 Small area estimations and projections 44

2/16/2018 Small area estimations and projections 45

Method of thematic mapping,

47 2/16/2018 Small area estimations and projections

• Many of the methods used for producing population estimates

2/16/2018 Small area estimations and projections 50

2/16/2018 Small area estimations and projections 51