LEARNING CURVE SPILLOVERS AND TRANSACTIONS COST

IN THE MICROFINANCE INDUSTRY OF THE PHILIPPINES

In partial fulfillment of the requirements for Economic Research

Jovi C. Dacanay

Department of Economics

School of Social Sciences

Ateneo de Manila University

October 22, 2014

 Acknowledgments

I would like to thank the following who have guided and inspired me in the course of

completing this research:

Professor Victor Venida for his patience and guidance during our Economic Research

classes and in providing a most helpful critique of our drafts,

Professor Noel De Guzman for his valuable comments before, during and after the

presentation,

My family, friends and colleagues from the University of Asia and the Pacific for helping,

inspiring and motivating me to finish this research.

2
3
 ABSTRACT

Microfinance institutions (MFIs) in the Philippines have gained a reputation for

operating as a for-profit institution reaching the poor through micro-lending. The

problem or issue which the study addresses is to determine how MFIs in the Philippines

are able to attain operational self-sufficiency, the established indicator for financial

viability among microfinance institutions, in spite of high transactions cost. The

phenomenon may be verified by the following research question: does the behavior of

operational and transactions costs among group and individual microfinance lenders

manifest experience or learning curve spillovers and a U-shaped supply curve? The

study has two objectives. First, using appropriate measures of financial and social

performance, the study shall empirically verify the phenomenon of experience or learning

curve spillovers among MFIs. Second is to estimate the supply curve for loans to the

unbanked poor and verify that it is U-shaped.

The results of the pooled least squares with cross-section random effects

regression estimation show that both NGOs and rural banks are attaining both objectives

of operational self-sufficiency and moderate to good social performance, through the

spillover effects of learning, that is, fast learning for rural banks and moderate learning

for NGOs. Operational and transactions costs are high but decreasing for both rural

banks and NGOs. Older, more mature NGOs and rural banks are able to set transactions

cost at the prescribed level of 11%-25%. Such costs ensure that the MFIs operate in

order to both financial performance and outreach.

Keywords: Transactions cost, operational self-sufficiency, social performance,

experience curve, spillovers

4
 TABLE OF CONTENTS

Acknowledgments ……………………………………………………………….………..2

Abstract ………………………………………………………………………….………..4

Chapter I. Introduction ………………………………………………………….………10

A. Statement of the Research Problem…...……………………………………..…..10

B. Significance of the Research Question…...……………………………….….….11

C. Scope and Limitations………..…………………………………………...….….12

D. Methodology ………………..………………………………………….….…….13

E. Definition of Terms ………………………..……………………….…..………14

Chapter II. Operational Self-Sufficiency, Cost per Dollar of Loan and Transactions

Cost among Microfinance Insitutions……………………………………....…...16

A. Historical Development ……………………..…………………………….……16

B. Policy Issues ………………………………..…………………………….……19

C. Theoretical Issues…………………………………..……………………..….…23

D. Empirical Issues…….…………………………………...……………….…...….27

Chapter III. Transactions Cost for Microfinance Institutions…………………….…..…22

A. Hypothesis…………………...…………………………………………….…..…27

B. Theoretical Framework and Methodology.……………………………………...29

C. Methodology………………………………………………………….…...……..42

D. Data Sources……………………………………………………...…..………….52

Chapter IV. Analysis of Results……………………………………….……..………….53

A. Descriptive Analysis of MFIs…………………………………….....….………..53

5
B. Learning Curve Spillover and U-Shaped Supply Curve Analysis.…...…….…...59

Chapter V. Summary, Conclusion and Recommendation ................................................75

Bibliography ......................................................................................................................76

Appendices……………………………………………………………………………….80

6
 LIST OF TABLES

Table 4.1 Transactions Cost of Rural Banks and NGOs from 2000 to 2011 by Age ......53

Table 4.2 Transactions Cost of Rural Banks and NGOs (2000 to 2011) by Age and

Operational Self-Sufficiency (OSS) ......................................................................54

Table 4.3 Relevant Information on Chosen Rural Banks and NGOs (2000-2011) ..........55

Table 4.4 Transaction Cost per Borrower of Rural Banks and NGOs (2000 to 2011) by

Age Operational Self-Sufficiency (OSS) and Profitability (Return on Assets

(ROS) and Return on Equity (ROE)) .....................................................................56

Table 4.5 Selected Indicators for Operational Efficiency and Social Performance….....57

Table 4.6 Variables Used for the Learning Curve Spillover and U-Shaped Supply Curve

Regressions………………………………………………..…………….……….59

Table 4.7 Experience or Learning Curve Spillover Regression………………………...62

Table 4.8 Experience or Learning Curve Spillover Regression Asset Specificity……..63

Table 4.9 U-Shaped Supply Curve Regression for Rural Banks and NGOs………..….68

Table 4.10 Experience or Learning Curve Spillover, Transactions Cost and a U-Shaped

Supply Curve for Rural Banks and NGOs……………………….…………..….73

7
 LIST OF FIGURES

Figure 3.1 Why Poor Consumers Did Not Receive Loans ................................................30

Figure 3.2 Why Money Lenders Lend at High Interest Rates ...........................................32

Figure 3.3 Why MFIs are Able to Lower Interest Rates....................................................36

Figure 3.4 Learning or Experience Curve of Micro-Lending ............................................37

Figure 3.5 Profit Margin, Operational Self-Sufficiency and Return on Assets for Rural

Banks and NGOs …………………….……………………..……………...…… 48

Figure 4.1 Experience Curve for Rural Banks and NGOs .................................................61

Figure 4.2 Experience Curve for Rural Banks and NGOs using Asset Specificity ……..63

Figure 4.3 Estimated U-Shaped Supply Curve for Rural Banks and NGOs……….……66

Figure 4.4 Transactions Cost and Experience or Learning Curve Spillovers ....................66

Figure 4.5 Operational and Social Outreach Efficiency among Rural Banks and NGOs .70

8
 LIST OF APPENDICES

Table A.1. Descriptive Statistics (2003-2011)…...........................................................81

Table B.1. Market Share of Selected Rural Banks (2009-2011)………………………..82

Table B.2. Market Share of Selected NGOs (2009-2011)………………………………84

Table C.1 Experience Curve Regression on Cost per Dollar of Loan………………….85

Table C.2 Experience Curve Regression on Personnel Cost per Dollar of Loan………86

Table C.3 U-Shaped Supply Curve Regression on Cost per Dollar of Loan…………..87

Table C.4 U-Shaped Supply Curve Regression on

Personnel Cost per Dollar of Loan………………………………………….88

9
 CHAPTER I

INTRODUCTION

The Philippines maintains its standing as No. 1 in the world in microfinance

regulatory environment. For four years in a row (2009-2012), the Economist Intelligence

Unit’s global survey has ranked the Philippines as number one in the world in terms of

policy and regulatory framework for microfinance. The Philippines is also consistently

ranked at the top ten for having a good microfinance business environment. The survey

noted that the Philippines recorded material gains in transparency in pricing given the

BSP’s issuance of improved rules on transparency and disclosure. However, the

microfinance industry in the Philippines suffers from the incidence of over-borrowing

and the high cost of transactions.

A. Statement of the Research Problem

The study investigates the relationship between transactions cost and operational

self-sufficiency in the microfinance industry of the Philippines. When micro-lenders

target the entrepreneurial poor, monitoring regularity of payments of borrowers may be

costly. This leads to the main problem of the paper: In spite of the presence of

transactions cost, can the microfinance industry achieve operational efficiency? The

study has two objectives. First, using appropriate measures of financial and social

performance, the study shall empirically verify a downward sloping experience curve.

Second is to estimate the supply curve for loans to the unbanked poor and verify that it is

U-shaped. Both objectives shall lead the study to a verification of decreasing operations

cost and transactions cost, thereby allowing MFIs to operate more efficiently.

10
B. Significance of the Research Question

The Philippines has a favorable regulatory environment which allows rural banks

to operate in a financially viable manner in spite of allotting a certain percent of credit

made accessible to the unbanked, that is, micro entrepreneurs. Reason is that the

regulatory environment allows rural banks to financially include them. Why? There has

been a significant increase in the number of active borrowers in the Philippines since

1994. Microfinance activities have been existing previous to the 1997 establishment the

regulatory body called National Credit Council. Through microfinance, the unbanked,

that is, micro entrepreneurs who do not have access to the lending activities and practices

of commercial banks, had access to loans. Now, financial services to the poor are

provided by NGOs, cooperative and rural banks. (Economic Intelligence Unit (EIU),

2013; Rosenberg, 2013)

Previous regulation by the Central Bank of the Philippines had directed credit,

also called subsidized credit which did not work. Reason: micro entrepreneurs have to

be trained and assisted in order to finance and operate a business viably. There was a

need for human capital development (Ashta, 2009).

In spite of the Philippines scoring high internationally in terms of creating a

policy environment that enhances financial inclusion and effective financial access to the

unbanked, it scores low on financial education and monitoring. Payments system,

transparency in lending and data collection for existing NGOs can be improved so as to

closely monitor and evaluate the financial and social outreach performance of these

MFIs. The Philippines scored low on a supporting institutional framework which would

incentivize MFIs to achieve accounting transparency, adherence to international

11
standards of accounting, dispute settlements for unfair and time-consuming loan

processes, an integrated credit bureau and policy and practice of financial transactions

through agents (EIU 2013). Even if the microfinance industry has allowed e-commerce,

with the cooperation of telecommunications services such as Smart Communications and

Globe Telecom, to be made available for the payment of loans, loan officers are crucial in

order to monitor and ensure the regular payment of loans. All these seem to denote high

transactions cost.

Such high transactions costs may hinder the further expansion of financial access

to the unbanked, as there is still 2.9 million families which are unserved by microfinance

institutions. (Microfinance Council of the Philippines, 2006). There is a large, about

67%, unmet demand for micro loans to the entrepreneurial poor, henceforth denoted as

the unbanked.

C. Scope and Limitations

The study focuses on the occurrence of transactions cost as human asset

specificity to explain the reason for high operating costs among MFIs. However, such

types of costs also imply experience and learning. The theoretical formulation of

experience, learning and transactions costs proper to the microfinance industry is non-

existent to date.

Due to a lack of access to actual data per firm over the amount of loans and

interest charged to each borrower, the study focused on annual data per MFI made

accessible in the Microfinance Information eXchange Portal (MIX). Borrowing and

payment behavior of creditors would be important in order to detect the occurrence of

mission drift in an MFI. Indicators used in the regressions are based on the accepted

12
accounting variables which are accessible through the MIX website. As a consequence,

relevant variables in the study are interest rates charged to micro-borrowers, human asset

specificity and social performance. The indicators used in the global industry are

operating expense ratio (also known as operating cost per dollar of loan), personnel

expense ratio and the social efficiency index. The computations used by the industry

shall be mentioned in Chapter 3.

The social performance indicators used in the study are number of borrowers,

depth of outreach (Average loan size over GDP per capita also termed as national loan

size), and the social efficiency index (Operating Expense Ratio x cost per borrower)

Outreach is only one aspect of social performance for MFIs.

The phenomenon of multiple loans across various MFIs cannot be observed from

the data used, as this phenomenon would involve detailed MFI data. Anecdotal

information claim that some MFIs apply for a loan to other MFIs so as to repay their

loans, leading to over-indebtedness. Previous to 2011, MFIs were allowed to lend only

up to US$3,500. However, a BSP circular issued in 2011 has allowed rural banks to lend

up to US$7,000.

D. Methodology

Using appropriate financial data provided by the Microfinance Information

Exchange Portal, the likelihood of a U-shaped supply curve for loans to the unbanked

poor, due to the high unit cost of transactions, shall be discussed. Then, an analysis of

the factors explaining transactions cost, through an estimation of the appropriate/proper

measure of operational self-sufficiency, an evaluation of selected social performance

measures for the microfinance industry, and, indicators for cumulated output shall be

13
done. Operational self-sufficiency is the measurement used to denote financial viability

among MFIs, whereas, the social performance index and the number of women

borrowers and outreach are the variables used to indicate social performance. The

regression estimation procedure, using unbalanced panel data, with cross-section random

effects shall be performed in order to evaluate whether or not the chosen MFIs for the

study are either: (a) decreasing their transactions cost but at a constant or level of

operational self-sufficiency, or, (b) improving their operational self-sufficiency and

lowering their transactions cost at the same time.

E. Definition of Terms

The following terms are used based on the stated definitions. The

conceptualization and definition of these terms follow Armendáriz and Szafarz (2011)

and Copestake (2007).

Operating expenses include the costs of implementing the loan activities—

personnel compensation, supplies, travel, depreciation of fixed assets, etc. Operating

expenses consume the majority of the income of most microlenders’ loan portfolios, so

this component is the largest determinant of the rate the borrowers end up paying

Financial viability refers to the operational and financial sustainability of MFIs.

Operational and financial sustainability refers to the achievement of cost effective

strategies in order to reach the unbanked and entrepreneurial poor. Financial outcomes

would indicate a lowering of portfolio at risk (90 days), constant or sustained returns to

assets and cost of operations over total assets.

In a broad perspective, the social performance of MFIs is associated with their

social objective, which is extending credit by targeting the poor and the excluded, the

14
adaptation of services and products to target customers, and improvement in the equity

capital of beneficiaries through the creation and strengthening of community relations.

Specifically, the social performance of MFIs refers to their mission: the alleviation of

poverty by extending credit to the unbanked poor, while at the same time operating cost

effectively, that is, achieving operational and financial sustainability or viability.

Financial inclusion, as an objective of MFIs, refers to their capacity to extend

credit to the unbanked entrepreneurial poor in a financially viable way. It is used

synonymously with the social and financial performance of MFIs.

Transaction costs in credit markets therefore are indirect financial costs generated

by various processes, including the cost of searching and collecting relevant information.

They are indirect costs caused by frictions in the flow of credit funds, preventing credit

markets from reaching efficient market equilibrium. Consequently transaction costs of

lending consist of the costs of administering credit, coordination costs and the costs of the

risk of default. It is further highlighted that administrative costs are those which are

directly attributable to the processing, delivering and administering of loans while

coordination costs are those resources a financial institution dedicates to ensuring that

clients adhere to terms stipulated in loan contracts.

15
 CHAPTER II

OPERATIONAL SELF-SUFFICIENCY, COST PER DOLLAR OF LOAN AND

TRANSACTIONS COST AMONG MICROFINANCE INSTITUTIONS

This chapter contains a review of related literature grouped into those which

discuss historical development, policy issues, theoretical and empirical issues.

A. Historical Development

Past government initiatives for poverty alleviation in the 60s to 80s were focused

on direct credit and guarantee programs, which provided massive credit subsidies to bring

down the cost of borrowing for target sectors. These programs resulted in very limited

effectivity and outreach, and at a great cost to the government’s budget. These programs

were met with massive repayment problems, capture of funds by large-scale borrowers,

neglect of deposit mobilization and huge fiscal costs for the government. (Jimenez and

Roman 2011).

Learning from these experiences, the government policies of the last decade have

shaped reforms to develop a market oriented financial and credit policy environment that

promotes and supports private institutions to broaden and deepen their services; while

government will instead focus not on the actual provision of credit but on creating the

enabling policy environment. It is during this period that the National Strategy for

Microfinance was created imbibing the following principles: 1) Greater role of the private

microfinance institutions in the provision of financial services, 2) Existence of an

enabling policy environment that facilitates the increased participation of the private

sector in microfinance, 3) Adherence to market oriented financial and credit policies, 4)

16
Non-participation of government line agencies in the implementation of credit and

guarantee programs.4 Several laws were enacted in support of this strategy namely;

Social Reform and Poverty Alleviation Act, Agriculture and Fisheries Modernization

Act, Barangay Micro-Business Enterprises Act, Executive Order 138 and the General

Banking Act of 2000. With the National Strategy for Microfinance and the subsequent

policy issuances and laws that were passed, the microfinance market has been driven by

the private sector with the government only providing the enabling policy and regulatory

environment. The strategy also laid the groundwork for the establishment of a regulatory

framework for microfinance as well as the uniform set of performance standards for all

types of microfinance institutions. (Jimenez and Roman, 2011)

Given this backdrop, there are now three types of institutions that provide

microfinance services: the nongovernmental organizations (NGOs), cooperatives and

rural banks. The success factor in developing a range of microfinance institutions is that

the policy environment encourages the development of these different types of

institutions recognizing the strength of each type in delivering microfinance services. The

current players include eight microfinance oriented banks and 187 thrift and rural banks

with some level of microfinance operations, approximately 300 NGOs where around 30

will have sizeable and significant microfinance portfolios, and 50 cooperatives engaged

in pure microfinance activities. Together, these institutions are providing microfinance

services to approximately 1.3 to 1.5 million families. (Jimenez and Roman, 2011). This

number amounts to close to a third of the entrepreneurial poor. (Also see Microfinance

Report for RP, 2011)

17
 Internationally, women, usually, are the clientele, low level of credit available to

the poor in spite of the possibility of high returns (Ashta, 2009; Armendáriz and Szafarz,

2011). The mission of the MFI is to alleviate poverty by extending credit to the

unbanked poor. There are two factors which prompts an MFI to increase its average

loan size over time, thereby lowering depth of outreach. First, progressive lending,

which, in the microfinance jargon, pertains to the idea that existing clients can reach up to

higher credit ceilings after observing a “clean” repayment record at the end of each credit

cycle. Second, cross-subsidization, which entails reaching out to unbanked wealthier

clients in order to finance a larger number of poor clients whose average loan size is

relatively small. These two explanations are in line with the MFI social objective.

(Armendáriz and Szafarz, 2011)

B. Policy Issues

The Bangko Sentral ng Pilipinas (BSP) remains at the forefront of establishing a

policy and regulatory environment conducive to financial inclusion. This stems from the

recognition that financial inclusion is a worthy policy objective that could and should be

pursued alongside the promotion of stability and efficiency in the financial system. It is

also believed that financial inclusion is a key component of inclusive growth. The BSP is

actively implementing policy and program initiatives to realize the Philippine

government’s vision for the financial sector: “an inclusive financial system which

provides for the evolving needs of a diverse public” (Philippine Development Plan 2011-

2016). The financial inclusion initiatives of the BSP are focused on the major areas of:

1) Policy, Regulation and Supervision, 2) Financial Inclusion Data, 3) Financial

Education and Consumer Protection, and 4) Financial Inclusion Advocacy.

18
 In spite of the absence of a regulatory body for NGOs, the BSP regulates NGOs

through circulars and ordinances. The BSP aims at promoting transparency and good

governance among MFIs through the Issuance of Rules Regarding the Relationship

Between Banks and their Related Microfinance Non‐Governmental Organizations

(Circular 725, 16 June 2011). This issuance recognizes the possible synergy between a

bank with microfinance operations and a related microfinance NGO/Foundation. While

this has become a successful model for some, the issuance aims to ensure that the banks

with related microfinance NGOs are able to safely and viably coexist by mitigating

possible operational, governance and reputational risks. The salient features of the

issuance includes a) requiring clear contractual agreements between the two entities, b)

prohibiting bank personnel from holding any concurrent, full time positions that may

cause them to be involved in the daily operations of related NGOs/foundations and c)

issuing general principle sand standards that will govern the business relationships

between banks and their related NGOs/foundations. (BSP Financial Inclusion Initiatives

Year-End Report December 2011).

Traditionally, the central bank defined microfinance loans to be below P150,000

(US$3,500). In December 2011, BSP issued Circular 744 that allows banks to offer the

option of “Microfinance Plus” loans of up to P300,000 (US$7,000). This is intended to

lessen the occurrence of over-indebtedness among MFIs as small borrowers seem to

resort to borrowing from other MFIs so as to pay the outstanding balance from a loan.

There is still, however, a need for an integrated credit bureau. Under the Republic

Act 9510 (also known as the Credit Information Act (CISA)), signed into law in 2008, all

regulated entities would be required to submit positive and negative information to a

19
credit bureau under the Securities and Exchange Commission (SEC). A crucial

component of the legislated is the creation of the Central Credit Information Corporation

(CCIC) to receive and consolidate the credit data and to act as a central registry of credit

information which will provide access to standardized information in order to overcome

and replace the existing system of fragmented credit bureaus. Although the law’s

implementing rules and regulation were approved in May 2009, the establishment of the

Central Credit Information Corporation (CICC) has not yet resulted in the operation of a

functioning, active credit bureau. Under the CISA implementing rules and regulations,

the proposed corporation will have a seed capital of Php 75 million. But so far, CCIC

only received Php 17.5 million. The future central depositary of credit history is also still

looking for an office, still forming the organization chart, and outlining plans for the rest

of the year.

Meanwhile, the private sector is pursuing its own initiative. The Banker’s

Association of the Philippines (BAP) has its own credit bureau, which the Rural Bankers

Association of the Philippines, the association of rural banks, is also using. BAP’s credit

information database contains approximately 3.8 million accounts and information

pertaining to unpaid loans, loans under litigation, bounced checks and credit card debt.

To encourage rural banks to subscribe to the database BAP-CB lowered the fee for each

inquiry from Php 11 (US$ 0.25) to Php 5.6 (US$ 0.13) and removed the upfront

subscription fee. As of January 2011, twenty rural banks have joined the bureau. Access

to credit information can help rural banks identify whether loan applicants already have

outstanding loans and avoid the problem of over-indebtedness, which can be harmful to

20
the borrowers, the banks and the sector as a whole. BAP’s credit bureau is almost

exclusively for non-microfinance clients, however.

As a result, the system of fragmented credit bureaus remains intact. For example,

cooperatives based in Mindanao have formed their own credit bureau called CCBOL.

Perhaps the most promising initiative was taken by the leading MFIs. In early 2013, the

seven largest microfinance providers (Taytay sa Kauswagan Inc. (TSKI), OK Bank,

CARD Bank, CARD NGO, Negros Women, Ahon sa Hirap and ASA Philippines), which

together serve about 70 percent of the estimated one million micro-borrowers in the

country, signed a memorandum of agreement on the creation of the credit bureau called

the “Microfinance Data Sharing System (MiDAS)”. Initially, MiDAS is meant to focus

on negative information, i.e. delinquent borrowers, with the view later on of establishing

and implementing programs aimed at client rehabilitation. The business requirements of

MiDAS are unique to its users, the microfinance institutions, with a special feature that

allows for Barangay (town or village) level search for delinquent borrowers. It is the

intention of the founders to expand the coverage of the credit bureau to other MFIs as

well in the future. (EIU 2013).

New rules issued by the BSP and effective July 1, 2012 outlaw the use of flat

interest rate calculation methods for regulated institutions. Unregulated NGO-MFIs and

cooperatives are encouraged to follow suit but the BSP lacks the authority to require them

to do so. This makes the calculation of an effective interest rate difficult. Through the

flat balance calculation method, the interest rate is applied to the initial loan amount

throughout the entire loan term. Through this method the borrower pays interest on the

full loan amount even though the amount they have over the loan term is less and less as

21
they repay the loan. Interest rates calculated using the flat balance appear much cheaper

than declining balance rates, but are in fact nearly twice as expensive. For example, an

annual interest rate of 15% charged on a flat balance results in almost the same amount in

interest payments as an annual interest rate of 30% charged on a declining balance. This

can make comparison between the prices of loans difficult, posing a serious obstacle to

MFIs in terms of their ability to make informed price-setting decisions and to clients in

terms of comparing the prices of the loan products available to them. Through the

declining, or reducing, balance interest rate calculation method, the lender charges

interest on the loan balance that the borrower has not yet repaid. This amount declines

over time as the borrower repays the loan, so that interest is only charged on money that

the borrower is in possession of.

C. Theoretical Issues

When one observes the behavior of commercial lenders, there seems to be a

causal link as to why there is low level of credit supplied to the poor in spite of high

demand. There is a high demand for credit among the entrepreneurial poor, but these

borrowers have high transactions cost, and commercial lenders would incur low profit.

The transactions cost come about as more effort is needed among commercial lenders,

that is, monitoring costs to ensure frequent and constant payment of loans has to be done.

As a result, there is a discontinuous marginal revenue curve for borrowers among the

micro-borrowers or the entrepreneurial poor vis-à-vis borrowers from small and medium

enterprises. This phenomenon partly explains the existence for a U-shaped supply curve,

a downward sloping steep supply curve for low levels of borrowings, after which it

22
reaches a minimum point, then the supply curve slopes upwards for higher levels of

borrowings (Ashta, 2009).

The microfinance industry has addressed or mitigated the occurrence of high

transactions cost either by individual lending (rural banks) or group lending but with

individual liability (NGOs). NGOs tend to have high information cost, thus, they seek

other sources of funding, i.e. donors or investors (Ashta,2009). With transactions cost

operationalized as governance cost, one can explain the institutional organization of

MFIs, i.e. vertically integrated firm has to be the structure due to asset specific

transactions. (Williamson 1989). Skilled loan officers invest time and effort in order to

understand the difficulties faced by micro-borrowers in achieving their weekly repayment

of loans. This investment of time and effort may be viewed as a sunk cost for both NGOs

and rural banks. Through time, there is an observed decrease in operational costs per

borrower, reflecting the possibility of a downward sloping experience curve for the

industry (Rosenberg, et al, 2013).

Transaction cost economics adopts a contractual approach to the study of

economic organization. Questions such as the following are germane: Why are there so

many forms of organization? What main purpose is served by alternative modes of

economic organization and best informs the study of these matters? Striking differences

among labor markets, capital markets, intermediate product markets, corporate

governance, regulation, and family organization notwithstanding, is it the case that a

common theory of contract informs all? What core features- in human, technology, and

process respects- does such a common theory of contract rely on? These queries go to the

heart of the transaction cost economics research agenda. (Williamson, 1985)

23
 As compared with other approaches to the study of economic organization,

transaction cost economics (1) is more micro-analytic, this means that the behavior of

agents within a firm explain governance costs which are usually not covered within the

scope of the governance costs of the market. (2) is more self-conscious about its

behavioral assumptions, (3) introduces and develops the economic importance of asset

specificity, (4) relies more on comparative institutional analysis, (5) regards the business

firm as a governance structure rather than a production function, (6) places greater weight

on the ex post institutions of contract, with special emphasis on private ordering (as

compared with court ordering), and (7) works out of a combined law, economics and

organization perspective. The basic transaction cost economics strategy for deriving

refutable implications is this: assign transactions (which differ in their attributes) to

governance structures (the adaptive capacities and associated costs of which differ) in a

discriminating (mainly transaction cost economizing) way. (Williamson, 1985)

However, the crux, or what gives a rationale for a transactions cost perspective is

that the stage or process of transactions can be divided into several phases. Previous

theories on economic organizations assume that no friction arises in the technological

process of production. For a service industry such as banking, once a borrower has been

properly screened for investment viability by the loan officers of the commercial bank,

the fulfillment of the requirements and therefore the terms of agreement of the contract

are understood by the borrower who holds oneself liable to the conditions for borrowing.

Frictions in the smooth flow of contractual arrangement in transactions give way

to transactions cost. The economic counterpart of friction is transaction cost: for that

subset of transactions where it is important to elicit cooperation, do the parties to the

24
exchange operate harmoniously, or are there frequent misunderstandings and conflicts

that lead to delays, breakdowns, and other malfunctions? Transaction cost analysis entails

an examination of the comparative costs of planning, adapting, and monitoring task

completion under alternative governance structures. (Williamson, 1985)

An application of the theory on governance costs on the microfinance industry

shall now be discussed. The discussion centers on the reasons why credit does not reach

the poor. First, it has been observed that there is a lack of available complementary

inputs when lending to micro-entrepreneurs. For example, there is a lack of financial and

accounting education, there is insufficient financial capital to invest and acquire human

capital, land or entrepreneurial ability in order to improve the productivity of land or any

foregoing business activity. Second, investment is not based on the marginal productivity

of capital or the returns on capital, but on the risk-adjusted returns. The poor may

promise higher returns but they also represent much higher risk (Akerlof 1970; Stiglitz

and Weiss 1981). An adverse selection problem usually happens. These problems have

been reduced by monitoring but such activity lead to high cost due to the small scale of

the loan made available to many borrowers. This aspect characterizes the asset

specificity of negotiations for loans and the monitoring of loan payments. Third, is the

transactions cost of loans. Partly, these are the costs of monitoring and of creating legal

liens on whatever collateral the poor may have to offer. But besides these, the time the

banker spent in helping an illiterate man fill out an application form (to have his

particulars available), the time spent to process the loan and to take back cash in small

installments and keep records, all have to be costed (Shankar 2007).

25
D. Empirical Issues

Empirical findings based on various data sources show that there is a difference in

performance when more years are covered per MFI. This may imply that the number of

years of existence of an MFI is directly proportional to operational self-sufficiency

(financial revenue over the sum of financial expense, impairment loss and operating

expense), and, the capacity to sustain outreach targets with commercial funding and

internally sourced equity (Rosenberg, et al, 2013). Rosenberg, et al (2013) and

Armendáriz, et al (2011) studied empirical outcomes first before any theoretical

explanation is given; i.e. social performance, financial performance and the current

phenomenon of mission drift.

Shankar (2007) focused on the empirical assessment of microfinance performance

in India, using indicators of transactions cost, especially with group lending, making use

of Rosenberg’s (2002) computation of information cost. Shankar (2007) stresses the

importance of commercially operated MFIs, which allows the recognition of the role of

transactions cost when examining the interest rates charged by MFIs. The results

highlight the mistakes committed by the government of India in handling the MFIs in

Andhra Pradesh. Microfinance activity was governed by state sponsored loans crowding-

out private sector initiative and the needed financial discipline which arises through

commercial or market-led operations in microfinance. Rural officials use microcredit as

a way to increase patronage within the region of Andhra Pradesh. The government of

India decided to stop microfinance activity in this state leading to client drop-out, access

to funding froze, many MFIs in the area defaulted. (EIU (2013)).

26
 CHAPTER III

TRANSACTIONS COST FOR MICROFINANCE INSTITUTIONS

This section defines the research hypothesis, presents the theoretical framework

and empirical model to be used.

A. Hypotheses

The research hypothesis to be tested in the study is: higher experience or learning

curve spillover effects lessens operating and transactions cost.

For the microfinance industry, regulation by the Bangko Sentral ng Pilipinas

(BSP) plays an important role in monitoring operating costs and operational self-

sufficiency. But only rural banks are regulated by the BSP. Thus, the study shall

proceed with an analysis for rural banks and another analysis for NGOs.

Operating costs for rural banks may manifest higher experience or learning curve

spillover effects and lesser transactions cost than NGOs. This phenomenon may be due

to their longer commercial lending experience, their strict observance of industry

standards for operational self-sufficiency, and the larger average loan portfolios,

comparison to MFIs, handled by loan officers.

Operating costs for NGOs may manifest lower experience or learning curve

spillover effects and may have higher transactions cost than rural banks. This

phenomenon may be due to their shorter commercial lending experience, the non-

regulated nature of their operations, and, smaller average loan portfolios in comparison to

that handled by loan officers in rural banks.

27
B. Theoretical and Conceptual Framework

The theoretical approach to the study of the efficiency and productivity of MFIs

focuses on the manner in which transactions between the micro-borrower and the loan

officer are done. For group-lending MFIs, loans per borrower are radically of a smaller

size than that of commercial banks, henceforth referred to as the traditional banking

sector, with loans averaging to at most a tenth of loans given by the traditional banking

sector to their borrowers. Also, borrowers of MFIs may operate their business in remote

areas, far from the nearest MFI. As a consequence, loan officers of MFIs usually have to

travel to these areas weekly so as to collect loan payments from a group of borrowers,

usually numbering to 30-40 in the Philippines. Some loan officers of group lending

borrowers collect more often than once a week. This process involves time and effort for

loan officers as problems of repayment of loans have to be immediately resolved per

collection period, i.e. 52 collection periods annually, so as to ensure that borrowers repay

their loans.

Transactions of rural banks involved in MFI operations, henceforth referred to as

simply rural banks, do not have to go to the area where their individual borrowers operate

their business. The borrowers go to the most accessible rural bank in order to turn-in

their weekly repayment. Given this background the theoretical approach most

appropriate to the study of the efficiency and productivity of MFIs in the Philippines

would be transactions cost economics.

These specific characteristics of microcredit can now be applied to the standard

theory of average and marginal revenues. With asymmetric information explanation,

marginal costs are much higher for the poorer people than for richer people. The

28
economies of scale effect of the transactions cost explanation remains, as well as the

ultimate increase in risks with over-indebtedness, and the supply curve becomes

downward sloping as loan size increases for richer people. The final curve is U-shaped as

in the standard literature because after some point over-borrowing by rich people

increases their risks. (Ashta, 2009).

All the above reasons imply that even if there are no usury laws in a country,

lending to the poor is not possible for the commercial sector, as illustrated in Figure 3,

which puts together the Marginal Revenue curve (MR) and supply curve for loans (based

on the marginal cost curve). Although the simplistic model that is proposed suffers from

many limitations as enumerated above, it is good for explaining the fundamentals.

Although other authors may have said the same thing, the representation is new and may

therefore lead to additional insights. (Ashta, 2009)

Figure 3.1 shows that the first portion of the MR curve does not have any

equilibrium with the supply curve. As a result, the organized sector did not lend to the

poor. The associated average revenue (AR) curve is consequently of little relevance to

any monopolistic banker who decides to enter the market. The final equilibrium is at low

rates of interest (ro) but is available only to wealthy borrowers. Since this end of the

market is more likely to be subject to (perfect) competition, interest rates are not going to

be much higher than marginal cost, but we can associate a demand curve if for

institutional reasons there is monopoly power. In this case, the demand curve would start

at the same level as this portion of the discontinuous supply curve and (in the simple

linear case) would slope downward at half the speed (the ARrich curve would take twice

29
as long as the MRrich curve to reach the X axis). Interest rates would then be higher at rm.

(Ashta, 2009)

The MR of capital curve is discontinuous at a point owing to effect of human

capital available only to not so poor borrowers. The supply curve is U shaped owing to

asymmetric information and transaction costs. As a result, poor people do not receive

loans but wealthier people do, at a fairly low interest rate, ro if there is competition and rm

if there is not. (Ashta, 2009)

Why do money lenders lend at high interest rates? Figure 3.1 illustrates this

graphically, using the discontinuous marginal revenue curve and U-shaped supply curve

analysis developed above. The lower asymmetric information costs of moneylenders

lower the supply curve of credit facing poorer borrowers. As a result, they also get loans,

30
but at higher interest rates of rMLm, much higher than the rates charged by the organized

sector of ro to richer borrowers. Since moneylenders are monopolistic, they charge rates

on the associated Average Revenue curve. The supply curve of moneylenders cannot fall

lower because their costs are high for various reasons: no access to cheap deposits, little

access to debt from the organized sector (they may get personal loans by pledging

personal assets), making them reliant only on their own equity capital. This equity capital

may also suffer from seasonal variations in demand, indicating that the moneylender may

have to recover higher interest rates in the busy season. (Ashta, 2009)

Moneylenders have lower transaction costs and lower information asymmetry. As

a result, they push down the supply curve for poorer borrowers to the dashed line shown

in the diagram. They charge high interest rates rML, much higher than the interest rates ro

charged by the competitive organized sector to the wealthier borrowers. (Ashta, 2009)

Why do banks refuse to enter the market for microcredit? If moneylending is

profitable, why don’t new entrants like banks increase competition and drive down

interest rates and profits? At the very least, they should drive these down to rMLc. The

suggested answers are several and are similar to reasons why banks don't come in.

(Ashta, 2009)

The first is based on transaction cost or barriers to entry. A new entrant has high

start-up costs: if he is not from the locality, he does not know the customers and does not

have the case histories. These are developed by experience. Also, it is difficult, time

consuming and expensive to market and monitor in isolated villages, driving up costs

especially for small transaction sizes. As a result, a new entrant faces higher risks and

higher costs than the local established moneylender. (Ashta, 2009) (See Figure 3.2)

31
 Also, the existence of more than one moneylender (competition as opposed to

monopoly) breaks down the business model of the established moneylender. The

borrower could shift from one moneylender to another. Greater competition will increase

default rates to the first institution, drive down interest rates and reduce the total number

of customers. As a result, moneylending would become a loss-making proposition,

driving out competition. Therefore, increasing the number of moneylenders is not a

solution. (Ashta, 2009)

Finally, more important, the market equilibrium in this segment would remain at

OD in Figure 3.2 and interest rates would not fall below the MR curve. Thus,

moneylending and other business would be competing only for lowering profits to the

32
segment as a whole. All this means that the market of DC for loans in this segment is still

not being served and, in the face of excess labor available to people in the same segment,

productive capacity is underutilized and economic development is suboptimal. (Ashta,

2009)

How can microcredit overcome information asymmetry and other barriers? One

method to overcome problems linked with information asymmetry used by microcredit

organizations is group lending. Although there are many different forms of group lending

exercised by microfinance institutions (solidarity groups, village banking, branch

banking, etc.), the essential method consists in lending to individuals in a group, and

using group pressure to ensure repayment by individuals. This group pressure may come,

for example, if other members of the group would get loans only if the first borrowers

pay back. As a result of this, people would not like to form groups with others who are

unlikely to pay back. Thus, the adverse selection problem is avoided without the bank

getting information. (Ashta, 2009)

The second tool for overcoming asymmetric problems is the provision of

incentives: to threaten to stop lending, on the individual level, is incremental loans or

progressive lending. Thus, if the borrower’s project can be divided into a series of

projects, the MFI lends a small amount first for one project and the next loan is given

only if the first one is repaid. Thus, the borrower is assured of funding for his project, if

he can overcome the moral hazard issues of being capable of managing the project and be

willing to repay. Another method used by MFIs is to collect repayments in public. Thus,

an agent passes at a fixed time once a week and all the borrowers are present and repay

him in front of everybody. This reduces collection costs as well as creates a social

33
pressure to repay on time. In some MFIs, there are also medals given for a series of

successful repayments, acting as further reinforcements both for the individual receiving

the medal as well as the others watching him receive it. (Ashta, 2009)

The third tool mentioned above was frequent repayments. The fourth tool is non-

traditional collateral. A fifth method used by MFIs is to focus on women. All of the

above methods do not necessarily require information being transferred from the

individual borrower to the bank, but they overcome the problems of asymmetric

information. However, a sixth method is used to improve the information available to the

MFI. This includes contacting neighbors to find out information about a potential

borrower. Some MFIs also encourage cross reporting where borrowers are encouraged to

be whistle blowers, in the interest of the larger group, if they think that some borrower is

not going to repay. (Ashta, 2009)

The final result of all these factors is illustrated in Figure 3.3. For graphic

simplicity the average revenue curve of the moneylender is taken off but we can see he is

charging the higher monopolistic rate. Since no market is yet saturated, MFIs may also be

in the position of monopolies. As a result they would charge interest rates somewhere

between rMFIc and rMFIm, depending on whether their mission is purely social and purely

for-profits. (Ashta, 2009)

These developments have led to money flowing from the MFIs to the poor to the

extent of DD': with moneylenders, the market obtained OD, with MFIs, it obtains OD'.

If the supply curve only shifts downward in this range and not outward to the

right, it would also affect the equilibrium solution for richer markets. As shown in Figure

3.3, the market of the poor now extends from OA earlier to OA' because these people

34
now demand more loans since their productivity has gone up. The next segment therefore

shifts to the right and the MR curve of the rich shifts to MR'RICH. (Ashta, 2009)

Microfinance institutions have lower cost of capital and they overcome

information asymmetry. As a result, they push down the supply curve for poorer

borrowers further to the dotted line shown in the diagram. At the same time, group

monitoring and involvement may actually boost the performance on projects and push the

MR curve of the poor to MR’. The MFIs charge interest rates rMFIc, or rMFIm, depending

on whether there is competition or monopoly in the local market, in either case much

higher than the interest rates ro charged by the organized sector to the wealthier

borrowers, but much lower than the interest rates rML charged by money-lender. (Ashta,

2009)

The perfect competition equilibrium interest rate to the rich rises from ro to r'o.

Some rich borrowers borrow less. Correspondingly, the unserved people in the poor

market is D'C' since only those willing to pay interest rates above r'o will be able to use

the capital with their available labor, given the new human capital thanks to MFIs, other

Non-Government Organizations and governmental dispositions. Although international

flows may not affect the national markets of developing countries immediately, they

would affect the relative availability of credit in developed countries and the raising of

interest rates may occur. (Ashta, 2009)

35
 Transactions cost economics can then be applied to the microfinance industry.

The occurrence of a lowering of interest rates applied to borrowers with a satisfactory

payment history would be possible only with the specialized effort applied by loan

officers who have invested time and effort to provide information and education to

micro-borrowers who may not have acquired the education necessary to carry-on and

manage financial and business negotiations. The type of transactions cost proper to

microfinance are relationship-specific assets: site specificity, physical asset specificity,

dedicated specificity and human asset specificity. (Williamson, 1989 and Besanko, et al,

2010). Understanding the nature of behavioral characteristics in the monitoring of loan

payments is not a question of understanding the statistical risks involved in the

36
transactions, which often require a large number of instances, i.e. renegotiations. The

bilateral business agreements are specific to the circumstances of the micro-borrower,

and would require specific talents and communication capabilities on the part of the loan

officers. These human-asset specific relationships enable the MFI to incur agency costs.

The study focuses only on the asset specific causes of transactions cost, and specifically,

on those which refer to human assets such as skills and capabilities which allow a longer

term duration of negotiations between the micro-borrower and the loan officer.

As the scale of the transactions increase, the firm’s demand for borrowings

increase, and a vertically integrated firm can better exploit economies of scale and scope

in production. Vertical integration is more likely to be the preferred mode of organizing

the transaction for any given level of asset specificity. This process involves a learning

phase for each firm involved in microfinance or micro-credit lending. (See Figure 3.4)

37
 The experience of a firm at any given age may be measured in a number of ways

including, inter alia, the age of the firm, the cumulative prior output of the firm, which for

microfinance would mean average loan balance per borrower, and, number of active

borrowers, the average tenure of its employees, or the average length of related work

experience of its employees. The most popular implementation assumes that the current

unit cost of a firm of age v, c(v), is a decreasing function of its cumulative prior output,
𝜈
𝛾(𝜈) = ∫0 𝑥(𝑠)𝑑𝑠, where 𝑥(𝑠) is the firm’s output given its market share at age v. In

most research, most especially in empirical and macroeconomic applications, a power

rule, a functional form for cost, c(v), using constant elasticity to scale, of the following

form:

𝑐(𝜈) = 𝑐(0)𝛾(𝜈)−𝛽 = 𝑚 + 𝑐𝑜 𝑒 −𝑦̅ (3.1)

is assumed. (Arrow 1962, Thompson, 2010). For equation (1), c refers to cost, m refers

to fixed costs, 𝑦̅ is the cumulated industry output at the end of the period and  refers to

industry spillovers, i.e. organizational structures so as to improve repayment and other

relevant experiences in dealing with micro-borrowers is made known to the entire

industry. Note that the equation used to estimate the experience curve uses firm figures

but the result will refer to the total industry as market shares per firm in the MFI industry

are not significantly different from each other, i.e. the Hirshman-Herfindahl index is <

0.15, indicating that market shares per firm in terms of number of borrowers and gross

loan portfolio have a close to a perfectly competitive market structure. (Ghemawat and

Spence, 1985). The computation of the Hirshman-Herfindahl index can be seen in

Appendix B.1-B.2. This specification of the learning curve assumes that with a close to

38
competitive market structure, cost structures are not significantly different and spillovers

occur. Proprietary experience per firm due to the existence of patents is almost nil.

The differences in the cost curves will depend on the spillover parameter, , the

mean of the assumed exponential function, which is also interpreted as a rate or speed of

cost decrease. Lower levels of  would indicate a faster learning time since the intercept

at time = 0 would also be higher. Thus, from Figure 3.4, | 3 | < | 2 | < | 1 |. Note that

the spillover rate is the reciprocal of the coefficient of the variables indicating cumulative

output, when a regression estimation is done on unit cost as dependent variable and a

cumulative output indicator as regressor. Ghemawat and Spence (1985) would refer to

experience or learning curve spillovers in order to incorporate the seminal concept of

learning-by-doing from Arrow (1962).

Scale economies may be present for the commercial banking sector. However,

such may not be the case for banks dealing with micro-entrepreneurs. The decrease in

the cost of transactions within a specific level of borrowings among micro-entrepreneurs

may not necessarily denote scale economies but may just be an indication that more

experience in transacting with the unbanked poor enables microfinance institutions to

acquire a higher level of learning, thereby, lessening the costs of monitoring. This

phenomenon may be summarized by a downward sloping experience curve, which

flattens rightwards. Passive learning, a phenomenon which gives rise to learning from

experience, refers to the conventional economic characterization of organizational

learning by doing as an incidental and costless byproduct of a firm’s production activities

(Thompson, 2010). A firm that increases productivity through passive learning will be

said to move along an experience curve (Thompson, 2010).

39
 The experience curve, also synonymously termed as the learning curve, is the

curve that relates unit costs to accumulated volume (Spence, 1981, Ghemawat and

Spence, 1985). The learning curve is believed to characterize the costs in some industrial

markets, and can be an important determinant of competitive behavior in those markets

(Spence, 1981). When there is a learning curve, the short-run output decision is a type of

investment decision. It affects the cumulated output, a stock, and through it, future costs

and market position (Spence, 1981). When additions to output lower future costs, it is

appropriate for the firm to go beyond the short-run profit maximizing level of output.

Thus, on an optimal output path for a firm, marginal short-run profits, as a function of

output are negative. In fact an optimal path is one in which, at every time, t, marginal

short-run profits equal the present value of the total cost reduction over all subsequent

periods of an additional unit of output at t. When applied to the microfinance industry,

profits from operations, specifically, operational self-sufficiency (OSS), converges to a

level that is competitive and sustainable, i.e. greater than 100%, averaging to 107% for

NGOs and 118% for rural banks. Younger rural banks and NGOs usually experience

higher levels of operational self-sufficiency but eventually, the OSS converges to 107% -

118%. When such levels of operational self-sufficiency are being observed, then the

bank is able to achieve financial sustainability in its operations, that is, operational along

with social performance efficiency are both achieved. Transactions cost may still

increase as loan officers adjust to the demands of micro-borrowers as they demand higher

loans. But at this stage, more competitive levels of interest rates may be applied to

micro-borrowers, thus, allowing the supply curve for credit to start increasing.

40
 The contribution of Ghemawat and Spence (1985) to the learning curve literature

is when they derive a cost function with constant elasticity to scale production, thereby

allowing an industry-wide trend for learning curves. When firms have constant elasticity

to scale production functions, then, market dominance does not exist, the possibility of

proprietary knowledge does not occur and information diffusion happens frequently. The

decreasing unit cost trend with cumulated output represents the industry-wide marginal

cost curve. This observation was also mentioned in Randon et al (2006) when applying

non-strategic learning curve spillovers to the semiconductor industry. They agree with

Spence (1985) that as long as the industry structure is close to a competitive market

model, spillovers occur even in the event of high research and development expenditures

and technological advancements. The frequency of information diffusion due to the

absence of market dominance leads to two effects: the efficiency effect and the

disincentive effect. Spillovers, according to Ghemawat and Spence (1985) causes costs

to decline, the efficiency effect, but they also tilt the marginal cost toward the current unit

cost and hence makes output choices less aggressive. This phenomenon may explain

why firms in the microfinance industry do not aggressively offer different types of loan

products to their borrowers, even if unit costs have declined due to extensive lending

experience. However, because of the competitive nature of the industry, unit costs

eventually reach an asymptotic point, where unit costs level off.

The manner in which NGOs and rural banks deal with micro-creditors, i.e.

techniques to monitor micro-entrepreneurs, is not proprietary knowledge. In fact, the

manners in which group and individual lending is done is made known to those in the

microfinance industry. Thus, experience and learning curve spillovers occur among

41
personnel of rural banks and NGOs thereby lowering entry barriers into the industry, and

these experience and learning curve spillovers result to cost declines in operational and

transactions cost.

The occurrence of an experience curve or a learning curve spillover, that is, a

decrease in operational and transactions cost in the case of the microfinance industry,

with cumulated output, differs from economies of scale. Economies of scale refer to the

ability to perform activities at a lower unit cost when those activities are performed on a

larger scale at a given point in time. The behavior of the experience or learning curve

spillover refers to reductions in unit costs due to cumulated experience over time.

Economies of scale may be substantial even when learning or experience is minimal. This

is likely to be the case in mature, capital-intensive production processes. Likewise,

experience may be substantial even when economies of scale are minimal, as in such

complex labor-intensive activities as the production of handmade watches. (Besanko and

Braeutigam, 2014). In industries wherein asset specificity occurs due to the diversity of

transactions handled by loan officers, initially high costs may happen in transactions and

negotiations carried out even by skilled personnel. Through time, learning occurs on the

part of bank personnel monitoring loan payments as well as micro-creditors, resulting to

experience or learning curve spillovers.

C. Methodology

To be able to test the stated hypotheses, the empirical methodology shall proceed

with two steps, which also pertain to the objectives of study. The study therefore

attempts to empirically verify (1) a decreasing cost over cumulative output indicated by

age, also called a learning or experience curve spillover, and (2) a U-shaped supply

42
function, for the microfinance industry. Pooled least squares regression with cross-

section random effects shall be performed for (1) and (2).

The coefficient of age upon establishment and average loan balance per borrower

shall be the indicator for experience or learning curve spillovers. The greater the spill-

over effects, the greater the capacity of the MFI to decrease transactions cost, assuming

that the data covers a time period from zero output up to a planning period. The

reciprocal of the coefficient of the cumulative output shall be the indicator for the

spillover rate as the regression model to be performed will not be a logarithmic function.

Experience in negotiations and transactions with micro-borrowers communicated within

each MFI and to the whole MFI industry is likely to result to a trend towards lowering

transactions cost.

Due to the unique nature of MFI operations, wherein, the human capital

specificity of each transaction with borrowers is also a function of the average loan per

borrower, a regression shall be performed for both rural banks and NGOs with human

capital specificity as dependent variable. Human capital specificity, as indicated by

personnel expense per average loan portfolio per borrower, accounts for more than 50%

of operating expenses. Thus, the study would verify if MFIs undergo a learning curve as

regards transactions with micro-borrowers. The data only extends to aggregates per firm

and not by loan officer or employee per firm. Thus, the study shall use average loan per

borrower and age upon establishment as the indicator for cumulative output, the

operating expense ratio or cost per dollar of loan as the indicator for unit cost, and,

personnel expense ratio or the personnel cost per dollar of loan as the indicator for human

asset specificity, or the transactions cost indicator.

43
 The functional form used for the learning or experience curve assumes spillovers,

that is, rivals learn from each other’s experience, i.e. organizational structures are made

known to all in the industry, due to the presence of a regulatory body monitoring the

financial viability of the players. If learning is neither too slow nor too rapid, no

significant entry barriers arise.

Using the operating expense ratio as the cost indicator of MFI operations and

dependent variable, a regression using operational self-sufficiency, average loan balance

per borrower, number of active borrowers, capitalization, size, actual age (upon

establishment), square of actual age, and, outreach as explanatory variables, can be done,

answering objective 1. The regression model used is adapted from Cull, Demirgüç-Kunt,

Morduch (2007) who used the same independent variables to explain financial

performance. In their empirical model, labor and capital costs were included as

regressors. For the study, the operating expense ratio was been used, indicator for unit

cost, as a dependent variable and cumulated output would be indicated by age upon

establishment, and, average loan balance per borrower for objective 1. The estimated

regression for equations (3.2) and (3.3) shall be plotted with age upon establishment and

operating expense ratio (unit cost indicator) and personnel expense ratio (asset specificity

indicator).

The reciprocal of the coefficient of average loan balance per borrower and age

upon establishment shall be the indicators for the experience or learning curve spillover

rate. Through time, it has been observed, microfinance firms increase their average loan

balance per borrower as they are more capable, with greater experience, to monitor and

manage increasing amounts of credit made available to borrowers. To achieve an

44
asymptotic curve, the square of age shall be included in the regression equation. The

experience curve can be estimated using the following equation, with the operating

expense ratio used as the indicator for unit cost:

Operating Cost per dollar of Loanijt (Operating Expense Ratio) = αij + β1 OSSijt +

β2 Average Loan Balance per Borrowerijt + β3 Number of Active Borrowersijt +

β4 Capital Costijt + β5 Gross Loan Portfolio per Assetijt + β6 Age upon

Establishmentijt + β7 Age upon Establishment Squaredijt β8 Scaleijt +

β9 Outreachijt + β10 Number of Women Borrowersijt + β10 Social Efficiency

Indexijt + εit (3.2)

Note that performance indicators have been used as regressors in the empirical

model. This specification was used due to the need for MFIs to achieve the regulated or

standard levels for operational self-sufficiency, which should be greater than 100%. This

level of operational self-sufficiency has to be reached even while achieving its social

outreach motives. The meaning of the above-mentioned explanatory variables and the

expected signs from the regression are indicated in Table 4.7.

The downward sloping learning or experience curve happens as a result of a

decreasing operating cost per dollar of loan across time, i.e. more years of experience by

the micro-borrower in transacting with the MFIs and its personnel. As mentioned by

Shankar (2007), transactions costs do seem to decrease with time for the Philippines. In

fact, there is much hope for interest rates charged to micro-borrowers due to an

expectation that as micro-lenders acquire more experience they learn to lend more

45
efficiently (Rosenberg, et al, 2013). Thus, what is expected to be observed for operating

costs must also be observed for transactions cost, the main cost driver for MFIs. This

observation will be verified through the following regression, with the personnel expense

ratio used as the indicator for transactions cost:

Personnel Cost per dollar of Loanijt (Personnel Expense Ratio) = αij + β1 OSSijt +

β2 Average Loan Balance per Borrowerijt + β3 Number of Active Borrowersijt +

β4 Capital Costijt + β5 Gross Loan Portfolio per Assetijt + β6 Age upon

Establishmentijt + β7 Age upon Establishment Squaredijt β8 Scaleijt +

β9 Outreachijt + β10 Number of Women Borrowersijt + β10 Social Efficiency

Indexijt + εit (3.3)

Neo-classical economic theory on competitive market structures points out that

cost improvements happen as firms (or the whole industry in a given market) acquire

more experience. Eventually, efficiency lessons are learned, and the experience curve

flattens out. At this point efficiency improves slowly even in the absence of

technological breakthroughs (Rosenberg, et al, 2013). In addition to the learning curve,

there is hope that the pressure of competition will force lenders to find more efficient

delivery systems (Rosenberg, et al, 2013). There seems to be, therefore, a global

evidence for a flattening operating cost curve for the microfinance industry, as micro-

lenders such as MFIs, gain more lending experience to micro-borrowers with time or age.

The study shall use high quality data reported in the Microfinance Information

eXchange Portal (MIX). Profitability patterns can be observed across the two main

46
institutional types, also denoted as lending types in the study, identified in the MIX are as

follows:

1. Individual-based Rural Bank lenders: institutions that use standard bilateral

lending contracts between a lender and a single borrower. Liability for repaying the loan

rests with the individual borrower only, although in some cases another individual might

serve as a guarantor;

2. NGO group lenders: institutions that employ contracts based on a group with

either joint or individual liability implemented with solidarity groups (in the spirit of

contracts used initially at the Grameen Bank in Bangladesh and at BancoSol in Bolivia).

Loans are made to individuals within the group, which has between 30 to 40 members

depending on the institution and location.

In a study by Morduch et al (2007), it has been observed that village banks have

the highest portfolio yields (indicator of financial self-sufficiency) and expense ratios

(indicator of operating cost self-sufficiency), but at the same time has the lowest return

on assets (indicator of profitability). The individual based lenders are observed to have

the most financial and cost efficient mode of operations and are also the most profitable.

For RP, the MIX data can only specify the type of market, i.e. poor vis-à-vis

better-off clients. Rural banks usually lend individually. Grameen type rural banks loan

to groups, but liability is rendered to each individual member of the group. NGOs lend to

groups but liability for loans can either be individual or by group. But accounting for

liability cannot be observed from the MIX data. From Figure 3.5, it can be observed that

rural banks have financial and cost efficient mode of operations and are also the most

profitable.

47
 Figure 3.5
Profit Margin, Operational Self-Sufficiency and Return on Assets for
Rural Banks and NGOs

120% 118% 107%

100%

80%

60%

40%

20% 13%
2% -2% 2% Profit Margin
0% OSS
Rural Banks NGOs
ROA
-20%

Source: MIX Market Data

This observation shows that the analysis of the operational efficiency by lending

type i, referring to either individual-based lenders such as rural banks, or, group lenders

or NGOs, across j firms and t years, from 2003 to 2011, would be affected by the level of

operating and transactions cost, and vice versa. When costs per loan transaction are

placed as dependent variable, the equation may be interpreted as a supply for loans

function, obtaining the regression procedure to account for Objective 2 of the study. The

basic regression model to be used, with the personnel expense ratio used an the indicator

for human asset specificity, is as follows:

48
Transactions Costijt (Personnel Expense Ratio) = αij + β1 OSSijt + β2 Average Loan

Balance per Borrowerijt + β3 Average Loan Balance per Borrower Squaredijt +

β4 Number of Active Borrowersijt + β5 Capital Costijt + β6 Gross Loan Portfolio

per Assetijt + β7 Age upon Establishmentijt + β8 Scaleijt + β9 Outreachijt + β10

Number of Women Borrowersijt + β11 Social Efficiency Indexijt + εit (3.4)

Due to the large percentage share of personnel expense per dollar of loan to

operating expense per dollar of loan, the same set of explanatory variables in equation

(3.5) shall also be done with operating expense per dollar of loan, the indicator for unit

cost for the supply curve equation, as dependent variable:

Operating Costsijt (Operating Cost per Dollar of loan) = αij + β1 OSSijt + β2 Average

Loan Balance per Borrowerijt + β3 Average Loan Balance per Borrower

Squaredijt + β4 Number of Active Borrowersijt + β5 Capital Costijt + β6 Gross

Loan Portfolio per Assetijt + β7 Age upon establishmentijt + β8 Scaleijt +

β9 Outreachijt + β10 Number of Women Borrowersijt + β11 Social Efficiency

Indexijt + εit (3.5)

The dependent variable is transactions cost, indicated by personnel expense per

average loan balance per borrower, the indicator used for human asset specificity.

Transactions cost also signifies the personnel cost per dollar of loan. It shall be regressed

on OSS (operational self-sufficiency by MFI institution, i.e. rural bank or NGO), average

loan balance per borrower, the square of the average loan balance per borrower (to

49
capture either an asymptotic behavior or a U-shaped curve, number of active borrowers).

MFI History would refer to age and scale of the MFI, and, orientation would refer to the

target market or outreach and type of outreach such as the number of women borrowers.

The capital structure indicator shall not use the debt-to-equity ratio but shall use the gross

loan portfolio over assets, as the debt-to-equity ratio for MFIs would need more

information on funding sources. Note that the expected signs for equations (3.4) and

(3.5) are the same as the expected signs for equations (3.2) and (3.3), and the explanatory

variables are the same except for an additional variable which the square of the average

loan balance per borrower which serves as the indicator for the behavior of costs.

If the supply curve is expected to be U-shaped, then the curve reaches a minimum

level of costs then starts to increase as the amount of loans by micro-entrepreneurs

increases. More loan officers and personnel have to be employed in order to process and

assess the viability of bigger loan amounts, as well as assess the credit history and

capacity to pay of the borrower. When the supply curve of MFIs starts to slope upwards,

then one can say that the behavior of the loan officers and personnel would compare and

follow the usual behavior of personnel employed in commercial banks. Note that

equations (3.4) and (3.5) incorporate the experience or learning curve spillover indicators.

These regressors aim to capture the existence of a supply curve in the microfinance

industry that is primarily denoting a downward-sloping experience or learning curve.

When the learning curve spillovers are large enough as to influence players in the

industry to achieve better monitoring of clients, then an industry-wide decrease in

operating and transactions cost is expected to occur, prolonged, verified by an asymptotic

marginal cost curve that flattens at a minimum point. It may start increasing as

50
cumulated output increases, that is, when micro-borrowers are more capable of handling

and managing larger loans for their micro-businesses.

MFIs listed in the Microfinance Information Exchange Portal (MIX), opt to be

transparent with their data, and thus, their financial and operational viability. This

attempt would imply that they observe the rules on the financial and operational self-

sufficiency indicators imposed on all MFIs. However, the two main types of lenders

studied: rural banks (mostly individual lenders) and NGOs (combination of individual

and group lenders), would competitively behave differently, i.e. outreach vis-à-vis

profitability. Thus, the NGOs will have to be analyzed separately from the rural banks.

Some types of lenders try to attract the better-off borrowers (high profitability ratios)

and/or sustainable borrowers (low operational costs). This may be true for rural banks.

Other lenders, such as NGOs, target poorer borrowers.

D. Data Sources

Data for the study was obtained from the Microfinance Information eXchange

Portal (MIX), which accounts for close to 60% of the total number of active borrowers

around the world. MIX is the premier source, open access, incorporated since 2002, for

relevant microfinance performance data and analysis. MIX provides performance

information on microfinance institutions (MFIs), funders, networks and service providers

dedicated to serving the financial sector needs for low-income clients. MIX fulfills its

mission through a variety of platforms.

The MIX Market (http://www.mixmarket.org) provides instant access to financial

and social performance information covering approximately 2,000 MFIs around the

51
world. MIX is a non-profit organization headquartered in Washington, DC with regional

offices in Azerbaijan, India, Morocco, and Peru.

The current study uses the MIX portal for financial and social performance data

for 104 MFIs, i.e. rural banks, NGOs and credit cooperatives. A total of only 50 MFIs:

18 Rural Banks and 32 NGOs, with historical data from 1998 to 2011, are included in

the dataset. Not all MFIs have a complete set of financial and social performance

indicator for all the years specified.

52
 CHAPTER IV

ANALYSIS OF RESULTS

The results of the study shall be discussed based on the objectives proposed.

A. Descriptive Analysis of the Operational Efficiency and Social Performance for

MFIs

From the above-stated objective 1, that is, an evaluation of MFI performance

based on financial viability and outreach, show the following results. A comparison of

operational self-sufficiency, and outreach was done with rural banks and NGOs based on

the MIX Portal database from 2000-2011 with a total of 50 rural banks and 32 NGO. It

can be seen that more mature MFIs are able to achieve lower levels of personnel expense

per dollar of loan, i.e. less than US$ 26.00. Due to the strict regulatory nature of the

commercial banking system, most rural banks are able to control their transactions to less

than US$ 26.00.

Table 4.1
Transactions Cost (Personnel Expense per US$100 of Loan) of Rural Banks and
NGOs from 2000 to 2011 by Age
Annual Age
Transactions Cost Young Mature Total Remarks on
(Personnel Expense Outreach
per US$100 of Rural NGOs Rural NGOs Rural NGOs
Loan) Banks Banks Banks
Low: Less than $26 0 4 46 10 46 14 Narrow to
Moderate: $26-$50 1 2 2 10 3 12 moderate

Moderate to
High: $51-$100 0 0 0 3 0 3 broad with
several network
affiliations
Total 1 6 48 23 49 29
Source: MIX Market Data (http://www.mixmarket.org/)
Note: Commercial sources of funds is the usual source of funding for all MFIs.

53
 Not only are rural banks able to control their transactions cost, they are also able

to strictly observe low risk and financially sustainable operations. NGOs, most of whom

are not regulated, operate to sustain profits, decrease risk but increase outreach. Most

rural banks are mature, whose client size mostly have a medium to large asset size. Also,

rural banks have a market clientele or target market ranging from micro firms (better-off

micro firms) to small and high end businesses. NGOs vary from young and new to

mature firms. Client outreach is composed of small to large firms (by asset size). But

NGOs primarily target low end (or poor) businesses as clients, indicated by a narrow to

moderate level of outreach (See Tables 4.2 and 4.3).

Table 4.2
Transactions Cost (Personnel Expense per US$100 of Loan) of Rural Banks and NGOs
(2000 to 2011) by Age and Operational Self-Sufficiency (OSS)
Age and Operational Self-Sufficiency (OSS)
Transactions Young Mature Total
Cost Rural Remarks

Rural Banks
(Personnel Banks NGOs Rural Banks NGOs
Expense per on

NGOs
Outreach
OSS > 1

OSS > 1

OSS > 1

OSS > 1
OSS < 1

OSS < 1

OSS < 1

OSS < 1

US$100 of
Total

Total

Total

Total
Loan)

Low: Less 0 4 4 2 44 46 2 8 10 46 14
than $26 Narrow to
Moderate: 1 1 2 2 2 2 2 8 10 3 12 moderate
$26-$50
Moderate
to broad
High: 0 0 0 3 3 0 3 with
$51-$100 several
network
affiliations
Total 0 1 1 0 6 6 4 44 48 7 16 23 49 29
Source: MIX Market Data (http://www.mixmarket.org/)
Note: The lowest level of operational self-sufficiency achieved by a bank from 2000 to 2011 is
0.48, and, 0.18 for NGOs for all samples. Commercial sources of funds is the usual source of
funding for all MFIs.

Rural banks are primarily focused on profitability as seen in the average profit

margin, OSS and ROA’s of the 50 listed rural banks in the MIX Portal. Cost per

54
borrower is close to double that of NGOs but NGOs have more than double the number

of active borrowers. Loans extended by rural banks are more than double that of NGOs

(See Table 4.4 for the specific MFIs).

Table 4.3.
Relevant Information on Chosen Rural Banks and NGOs (2000-2011)

Number of Active Borrowers

Operational Self-Sufficiency

Average Loan Balance per

Average Loan Balance per

Borrower per capita Gross

Total Operating Expenses

Cost per Borrower (US$)
Return on Assets (in %)

over Total Loans (in %)

National Income (in %)
Profit Margin (in %)

Portfolio at Risk
(90 days) (in %)

Borrower (US$)
Selected (in %)
Indicators by
Lending Type

Rural Banks
Mean 13 118 2 7.5 102 17,436 668 46 22
Median 12 114 2 6.0 79 9,683 454 33 19
Maximum 49 195 16 44.0 528 267,282 7,139 521 92
Minimum -108 48 -16 0.0 20 208 46 4 2
Number of 276 279 229 181 194 256 256 255 229
Observations
NGOs
Mean -2 107 52,380 6.8 44 118 8 45
Median 8 108 3% 18,603 3.0 38 93 6 41
Maximum 46 185 23 606,488 50.0 165 1,414 69 124
Minimum -463 18 -96 837 0.0 11 30 3 20
Number of 214 214 188 153 183 226 226 223 189
Observations
Source: MIX Market Data (http://www.mixmarket.org/)

Very little improvements in operational efficiency can be observed from 2000-

2005 and 2006 to 2012. One can observe an increase in the average loan balance per

borrower as the well an increase in the number of active borrowers for both rural banks

and NGOs. (See Table 4.5).

55
 Table 4.4
Transaction Cost per Borrower of Rural Banks and NGOs (2000 to 2011) by Age, Operational Self-Sufficiency (OSS) and Profitability (Return on Assets (ROS) and Return on Equity (ROE))
Annual Age
Transaction Young Mature Total
Rural Banks NGOs Rural Banks NGOs Remarks
Cost Rural
NGOs
(in US$) OSS < 1 OSS > 1 Total OSS < 1 OSS > 1 Total OSS < 1 OSS > 1 Total OSS < 1 OSS > 1 Total Banks

Bangko Santiago de Libon, Bangko

Kabayan, Bangko Luzon, Bangko Mabuhay,
BCB Cantilan Bank, CBMO, FICO, First
Macro Bank, GM Bank, Mallig Plains RB, VEF, ASHI,
New RB Victorias, PR Bank, RB Bagac, RB ECLOF-
FCBFI, RB
Low: Less Cainta, RB Camalig, RB Cotabato, RB Philippines,
0 0 0 0 JVOFI, 3 Datu 35 CMEDFI, Serviamus 8 35 11 Funding mainly via
than $26 Dipolog, RB Digos, RB Guinobatan, RB MEDF,
HSPFI Paglas commercial sources,
Labason, RB Lebak, RB Liloy, RB Loon, PMDF,
RB Mabitac, RB Makiling, RB Oroquieta, PALFSI outreach is narrow to
RB Pagbilao, RB San Enrique, RB Solano, moderate
RB Sto. Tomas, RB Talisayan, RB Tangub
City, Valiant RB

56
Kasagana-Ka, Kazama
Moderate: 1st Valley Bank, FAIR Bank, PBC, RB
0 0 0 0 0 0 0 5 CEVI Grameen, KCCDFI, 5 5 5
$26-$50 Montevista, Progressive Bank
MILAMDEC

High: RB San ABS-CBN, ARDCI, Funding via

0 1 0 DSPI 1 OMB K Bank, RB Katipunan 3 0 5 4 6 commercial sources
$51-$100 Jacinto ASKI, DSPI, RSPI
and grants, outreach is
ASA moderate to broad,
Very High: OK CARD-NGO, KMBI,
0 0 0 PAGASA Philippines, 3 Card Bank, Green Bank 3 0 5 3 8 several network
> $100 Bank NWTF, TSKI, TSPI
Life Bank affiliations
High information cost,
possible via funding
sources, especially for
Total 0 1 1 1 6 7 2 43 45 7 16 23 47 30
NGOs. Target varies
from poor borrowers
to SMEs.
Source: MIX Market Data
Note: The lowest level of operational self-sufficiency achieved by a bank from 2000 to 2011 is 0.48, and, 0.18 for NGOs for all samples. Rural Banks and NGOs in green have a positive ROS and ROE, those in red have a
negative ROS and ROE averaged from 2000 to 2011.
 Table 4.5
Selected Indicators for Operational Efficiency and Social Performance
Average Operational Social Loan
Loan Self- Efficiency Personnel Officer
Balance per Sufficiency Index2 Productivity Productivity
Borrower
Descriptive

2000-2005

2006-2012

2000-2005

2006-2012

2000-2005

2006-2012

2000-2005

2006-2012

2000-2005

2006-2012
Statistics

Mean 490 769 1.19 1.16 39 27 103 104 72 86

Banks

Median 375 495 1.16 1.13 23 19 96 98 60 50

Rural

Maximum 2461 7139 1.95 1.91 219 163 195 1,040 490 871
Minimum 46 62 0.66 0.48 9 2 41 11 6 1
Mean 76 129 1.04 1.09 139 97 133 182 137 322
NGOs

Median 70 109 1.06 1.08 134 95 132 70 132 126

Maximum 214 470 1.58 1.85 229 229 198 727 284 1,884
Minimum 30 44 0.18 0.38 47 21 44 11 46 8
East Asia 11-14 110%-120% 200%-280%
Average1

Eastern Europe 26-45 60%-90% 155%-190%

World

Latin America > 1.0 27-37 110%-140% 300%-310%

World <10-47 High productivity, excellent
asset quality
Source: MIX Market Data (http://www.mixmarket.org/), MicroRate (2014)
Notes: 1World Average uses 2009 to 2012 data
2
Excellent < 30, Good 30-50, Moderate 50-100, Poor > 100

Slight improvements in operational efficiency through the operational self-

sufficiency indicator, can be observed with NGOs. However, their social efficiency

index falls within poor levels, with a slight improvement to moderate level in 2006 to

2012. One can compare this with rural banks, whose social efficiency index improved in

2006-2012. The social efficiency index is computed as the cost per dollar of loan (or

operating expense ratio) over cost per borrower and allows one to compare MFIs with

different credit methodologies. NGOs seem to face several factors which make the cost

of per dollar of loan more expensive. For example, MFIs in the Philippines provide

complementary services such as health or training which will increase operating expenses

57
but are not directly related to the cost of providing the loan. For NGOs, however, these

complementary services allow them to fulfill their outreach mission.

B. Learning Curve Spillover and U-Shaped Supply Curve Analysis

All the regression results made use of a pooled least squares with cross-section

random effects model. The estimation obtained a strong correlation between fixed and

random effects, as the heterogeneity among rural banks and NGOs takes into account the

varying intercepts, the indicator of fixed costs, of the regressions obtained per firm. The

random effects model gave the correct signs for the explanatory variables, thereby giving

sound theoretical results. The Durbin-Watson statistic obtained in all the regressions was

less than 1.4, indicating problems of serial correlation. Problems of endogeneity among

the explanatory variables are expected in the regression, as the other explanatory

variables were meant to explain operational self-sufficiency, which was used as a

regressor in the estimation procedure.

In spite of all the limitations of the regression model used, the study focuses on

the level of significance that would be obtained from the coefficients of the cumulated

output indicators. The levels of significance obtained from the regression with random

effects and the regression without random effects did not differ significantly.

The regression on the learning curve and the U-shaped supply curve shall use the

following variables with their corresponding significance and expected signs. (See Table

4.6). Specific indicators shall be observed in the regressions: (a) For the experience or

learning curve spillover regression, the presence of spillovers among MFIs shall be

captured through a learning curve that levels-off as the amount of cumulated output over

time increases. These indicators shall be age and the square of age upon establishment,

58
 and average loan balance per borrower; (b) For a U-shaped supply curve, also known as

the U-shaped marginal cost curve, the presence of a downward-sloping experience or

learning curve is expected either to be asymptotic to the cumulated output indicator or

that it reaches a minimum then increases as the supply of loans increases. The

coefficients of the experience or learning curve spillovers are expected to be significant

and negative.

Table 4.6
Variables Used for the Learning Curve and U-Shaped Supply Curve Regressions
Relevance and Computation and Commercial Bank or
Variable Used Expected Sign for all Regressions Traditional Banking Equivalent
 This ratio provides the best indicator  Operating Expense over Average Loan
of the overall efficiency of a lending Portfolio.
institution. Also referred to as the  Goal is lower than 35% for urban MFIs
Efficiency Ratio, measuring the but leading MFIs have 10% or lower.
Operational institutional cost of delivering loan  Similar to the Efficiency Ratio or
Expense Ratio services compared to the average Cost/Income ratio used by the
also called loan size of its portfolio. traditional banking sector to determine
Cost per  Therefore, a general rule is the how efficiently the bank uses its assets
Dollar of lower the Operating Expense Ratio, and liabilities within internal operations
Loan. the higher the efficiency. related to the loan portfolio. This ratio
 This variable is used as the measures the amount of non-interest
dependent variable for the expenses (operating expenses,
regression in the learning curve, excluding provisions of loan losses)
indicator for cost per unit. needed to support operating revenues.
 Used as an indicator for asset
specificity, or, human asset  Compensation or salaries over
Personnel specificity. Average Loan Portfolio
Expense Ratio  Used as the dependent variable for  No equivalent with the traditional
the regression on transaction cost banking sector
as asset specificity.
 Financial Revenue over (Financial
 Indicator for Financial Self- Expense + Impairment Loss +
Sufficiency, and, thus Operating Expense)
sustainability.
 Financial revenues are revenues from
 Used as an explanatory variable for the loan portfolio and from other
Operational the learning curve and transactions financial assets and are broken out
Self- cost regression. separately and by type of income
Sufficiency  Expected to have a negative sign (interest, fee).
 MFIs operate with the goal of  Equal to or greater than 100%
achieving sustainability. Thus, it  No equivalent with the traditional
acts as an explanatory variable for banking sector
indicating efficiency of operations

Average Loan  Gross loan portfolio over number of active borrowers.

 Used as an indicator for supply of loans, also indicating cumulated output over
Portfolio time and is expected to have a negative sign for all regressions
59
Square of  Indicator for a U-shaped supply curve and is therefore expected to have a
Average Loan positive sign for all regressions
Portfolio
 The number of individuals or entities who currently have an outstanding loan
balance with the MFI or are primarily responsible for repaying any portion of
Number of the Gross Loan Portfolio. Individuals who have multiple loans with an MFI are
Active counted as a single borrower.
Borrowers  Used as an explanatory variable for the regression on the learning curve and U-
shaped supply curve.
 Expected to have a positive sign
 Computed as total liability over total
equity as in the traditional banking
 The simplest and best-known sector
measure of capital adequacy  Reveals the extent to which the bank
because it measures the over-all
funds operations with debt rather than
leverage of the institution. Hard to equity, allowing banks to monitor
put a standard level as MFIs have solvency and analyze their capital
various sources of local and structure.
Debt to international funding  Varies considerably depending on the
Equity Ratio  It is of particular interest to lenders type of institution. NGOs typically
because it indicates how much of a have lower Debt/Equity (1:1 to 3:1)
safety cushion (in the form of levels than regulated MFIs, such as
equity) there is in the institution to rural banks, which even have lower
absorb losses. levels than commercial banks. The
 Expected to have either a positive only way to strengthen an NGO’s
or negative sign equity is by reinvesting profits or
through grants and donations.
 Indicator for all outstanding principals due for all outstanding client loans per
Gross Loan dollar of assets, includes all outstanding client loans (current, delinquent and
Portfolio over renegotiated, except those written-off) and includes interest receivables.
Assets  Expected to have either a positive or negative sign
 Age upon establishment as a rural bank or as an MFI.
 Indicator for accumulated output over time
Age upon
Establishment  Used as explanatory variable for the learning curve regression and U-shaped
supply curve regression
 Expected to have a negative sign
Square of Age  Indicator for a learning curve that levels-off with a high level of accumulated
upon output, due to the phenomenon of spill-overs.
Establishment  Expected to have a positive sign
 Large Number of borrowers > 30,000
 Medium Number of borrowers 10,000 to 30,000
Outreach  Small Number of borrowers < 10,000
 Categorical variable: 1 refers to small, 2 refers to medium, 3 refers to large
 Expected to have either a positive or negative sign
 Large: Africa, Asia, ECA, MENA: >8 million; LAC: >15 million;
 Medium: Africa, Asia, ECA, MENA: 2 million–8 million; LAC: 4 million–15
million;
Scale  Small: Africa, Asia, ECA, MENA: <2 million; LAC: <4 million
 Categorical variable: 1 refers to small, 2 refers to medium, 3 refers to large
 Expected to have either a positive or negative sign
Percent of  Number of women over total number of active borrowers. An indicator of
Women outreach as some NGOs focus on servicing women.
Borrowers  Used as an explanatory variable and may have a positive or negative sign

60
  A proxy for how efficiently the
institution is providing loans while
neutralizing the effects of average
loan size on efficiency (both
operating expense ratio and cost per
borrower are each heavily  Operating Expense Ratio over Cost per
influenced by the loan size). Borrower (Operating Expenses over
Social  The Social Efficiency Index allows Number of Active Borrowers). This is
Efficiency for a more direct comparison of an index.
Index different types of MFIs with  No equivalent with the traditional
different credit methodologies. banking sector
 Used as an explanatory variable for
the regression on the learning curve
and transactions cost.
 Expected to have either a positive or
negative sign.
Source: MicroRate (2014), Rosenberg, et al (2013), MIX Market Data
(http://www.mixmarket.org/)

From Figure 4.1, it can be observed that rural banks do have higher experience or

learning curve spillover rates than NGOs, when we use age upon establishment as the

indicator for spillover effects.

61
 Table 4.7
Experience or Learning Curve Spillover Regression
(For Rural Banks and NGOs)
Method: Pooled Least Squares with Cross-Section Random Effects
Operating Expense per Average
Explanatory Variables Expected Loan Portfolio
Sign
Rural Banks NGOs
C + 69.52 *** 122.69 ***
6.41 12.58
Operational Self-Sufficiency - -11.58 *** -22.70 ***
(OSS) 2.81 4.48
Average Loan Balance per - -0.002 ** -0.04 **
Borrower 0.001 0.02
Number of Active Borrowers + 8.81E-05 *** 1.74E-05 *
1.72E-05 1.16E-05
Debt-to-Equity Ratio + or - -0.90 *** 0.03 na
0.29 0.12
Gross Loan Portfolio over + or - -30.91 *** -56.29 ***
Total Assets 4.79 5.89
Age upon Establishment - -0.40 ** -0.71 *
0.22 0.62
Age upon Establishment + 0.005 * 0.03 *
Squared 0.004 0.02
Scale (1 small, 2 medium, 3 -3.56 *** 1.57 na
large) + or -
1.03 1.40
Outreach (1 small, 2 medium, + or - 1.08 na -0.95 na
3 large) 0.96 1.78
Women Borrowers as a + or - 3.59 ** -7.72 na
Percent of Total Borrowers 1.77 6.26
0.07 *** 0.01 na
Social Efficiency Index + or -
0.02 0.03
Adjusted R-squared 0.53 0.39
F-statistic 15.45 10.71
Durbin-Watson Statistic 1.35 0.82
Cross-Sections Included 45 29
Total Unbalanced Panel Observations 140 165
Years Covered 2001-2011 2000-2011
Source: Author’s Estimates, MIX Market Data (http://www.mixmarket.org/)
Note: Italicized numbers refer to the standard deviation. Regression uses White
diagonal standard errors and covariance (degrees of freedom corrected). All p-
values of the F-statistic are very significant or p less than 0.01. p-values used
are: *** p less than 0.01, ** p less than 0.05 and greater than 0.01, * p less than
0.15 and greater than 0.05. Swamy and Arora estimator of component variances.

62
 From Table 4.7, we would observe a lower coefficient level, in absolute terms, for

age upon establishment for rural banks than for NGOs, as a lower coefficient for age

would refer to a higher experience or learning curve spillover rate, the reciprocal of the

coefficient indicates the spillover rate. The square of age is positive and moderately

significant for the regression for NGOs and rural banks, stating that the curve flattens as

age increases. The slower learning process for NGOs may be due to the high social

efficiency index as NGOs face high levels of operational expenses when training loan

officers and personnel, following-up and educating micro-borrowers.

When the dependent variable is changed into the asset specificity indicator, which

is the personnel expense ratio, it can be observed that the learning curve flattens out with

age, but the regression result no longer holds age and the square of age as a significant

explanatory variable for rural banks, only for NGOs. (See Figure 4.2)

63
 Table 4.8
Experience or Learning Curve Spillover Regression and Asset Specificity
(For Rural Banks and NGOs)
Method: Pooled Least Squares with Cross-Section Random Effects
Personnel Expense per Average
Explanatory Variables Expected Loan Portfolio
Sign
Rural Banks NGOs
C + 25.93 *** 59.15 ***
3.47 8.09
Operational Self-Sufficiency - -6.46 *** -17.47 ***
(OSS) 1.69 2.71
Average Loan Balance per - -0.0005 * -0.03 **
Borrower 0.0004 0.02
Number of Active Borrowers + 4.66E-05 *** -3.01E-06 na
6.71E-06 6.27E-06
Debt-to-Equity Ratio + or - -0.31 ** -0.04 na
0.14 0.06
Gross Loan Portfolio over + or - -7.48 *** -15.78 ***
Total Assets 2.29 3.93
Age upon Establishment - 0.10 na -0.55 *
0.11 0.43
Age upon Establishment 0.001 na 0.03 ***
Squared +
0.002 0.01
Scale (1 small, 2 medium, 3 -0.86 * 0.48 na
large) + or -
0.51 0.90
Outreach (1 small, 2 medium, + or - -0.65 * 0.59 na
3 large) 0.42 1.11
Women Borrowers as a + or - 1.12 * -2.90 na
Percent of Total Borrowers 0.77 3.32
0.08 *** 0.04 *
Social Efficiency Index + or -
0.02 0.02
Adjusted R-squared 0.58 0.33
F-statistic 13.51 7.07
Durbin-Watson Statistic 1.25 0.91
Cross-Sections Included 43 28
Total Unbalanced Panel Observations 102 136
Years Covered 2003-2010 2003-2011
Source: Author’s Estimates, MIX Market Data (http://www.mixmarket.org/)
Note: Italicized numbers refer to the standard deviation. Regression uses White
diagonal standard errors and covariance (degrees of freedom corrected). All p-
values of the F-statistic are very significant or p less than 0.01. p-values used
are: *** p less than 0.01, ** p less than 0.05 and greater than 0.01, * p less than
0.15 and greater than 0.05. Swamy and Arora estimator of component
variances.

64
 The average loan balance per borrower for rural banks and NGOs continues to be

negative and moderately significant with the regression on transactions cost. We can

conclude that the experience or learning curve spillover phenomenon among MFIs is

explained by increasing average loan balance per borrower and age for NGOs. For rural

banks, age of the establishment no longer explains transactions cost (See Table 4.8).

Once the operational expense ratio has reached a low level of US$20, asset specificity no

longer plays a significant role as micro-borrowers increase their average loan size. The

goal of continuing efficient operations along with social performance requires a high

level of capability among loan officers and personnel when dealing with higher average

loan portfolios among micro-entrepreneurs. NGOs, on the other hand, are already on

their way to achieving higher spillover effects. Both operating costs and transactions cost

are not yet achieving an asymptotic level, but they are decreasing.

The succeeding regressions shall now combine the phenomenon of experience or

learning curve spillovers with the presence of a U-shaped supply curve. This time, the

other indicator for accumulated output, average loan balance per borrower, shall be used.

This is also the indicator for output for the U-shaped supply curve. (See Figure 4.3)

When the personnel expense ratio, the indicator used for asset specificity is

graphed using average loan balance per borrower on the abscissa, a steep downward

sloping experience curve can be observed not only for NGOs but also for rural banks.

While the evidence of a flattening experience can be observed with rural banks, the

experience curve for NGOs seem to be reaching a low personnel cost per dollar of loan.

The average loan balance per borrower, however, has not increased and thus a flattening

experience curve cannot be observed. (See Figure 4.4)

65
66
 From the regression results of Table 4.9, one can see the consistently negative

sign of operational self-sufficiency and outreach for both NGOs and rural banks. The

extent of outreach through the social efficiency index and the high level of operational

self-sufficiency all seem to go hand-in-hand with a steeply downward sloping operating

and personnel cost vis-à-vis age upon establishment and the average loan balance per

borrower. Only rural banks seem to manifest a slightly increasing marginal cost curve,

denoting the well-behaved portion of the supply curve for an individual firm.

Thus, in response to objective 2, the results from Table 4.9 seem to indicate that

only rural banks are starting to manifest a well-behaved supply curve as they are no

longer operating at the steep portion of the marginal cost curve (See Figures 4.3 to 4.4).

Operating costs are decreasing as the average amount of loan portfolio increase. The

learning curve spillover effects among rural banks is already manifesting a marginal cost

curve that has reached a minimum level, and is on its way towards an upward sloping

marginal cost curve. On the other hand NGOs, most of whom have a narrow extent of

outreach, i.e. only dealing with small borrowers, and, with smaller average loan

portfolios per borrower than rural banks, do not reflect an upward sloping marginal cost

curve. In fact, most NGOs are not yet operating at the minimum level of operating costs,

unlike that of rural banks. Average loan balance per borrower hardly reach US$500 (See

Figure 4.4). Transactions costs for NGOs, however, are approaching the minimum level

reached by rural banks but over-all operating cost, that is, cost per dollar of loan, are still

far from the minimum level of operating costs per dollar of loan achieved by rural banks.

International regulatory bodies for MFIs put the standard to be within the range of

US$10-US$26 per dollar of loan.

67
 Table 4.9
U-Shaped Supply Curve Regression for Rural Banks and NGOs
(With Transactions Cost and Experience or Learning Curve Spillovers)
Method: Pooled Least Squares with Cross-Section Random Effects
Personnel Expense per Operating Expense per
Explanatory Expected Average Loan Portfolio Average Loan Portfolio
Variables Sign
Rural Banks NGOs Rural Banks NGOs
C + 26.10 *** 49.94 *** 68.23 *** 113.69 ***
3.32 10.96 5.86 15.22
Operational Self- - -6.29 *** -17.24 *** -10.80 *** -22.04 ***
Sufficiency (OSS) 1.69 2.90 2.79 4.61
Average Loan -0.002 * -0.01 na -0.006 *** -0.03 na
Balance per -
Borrower 0.001 0.07 0.002 0.09
Average Loan 2.7E-07 * -3.4E-05 na 9.3E-07 ** -2.3E-05 na
Balance per +
Borrower Squared 2.2E-07 0.0001 3.7E-07 0.0002
Number of Active + 4.7E-05 ** 1.3E-06 na 8.0E-05 *** 2.3E-05 **
Borrowers 6.7E-06 6.3E-06 1.7E-05 1.1E-05
Debt-to-Equity + or - -0.32 ** -0.03 na -1.00 *** -0.02 na
Ratio 0.14 0.06 0.28 0.12
Gross Loan -7.78 *** -15.02 *** -32.70 *** -56.01 ***
Portfolio over + or -
Total Assets 2.26 4.11 4.72 5.95
Age upon - -0.04 * 0.42 ** -0.11 * 0.21 na
Establishment 0.04 0.22 0.07 0.29
Scale (1 small, -0.73 na 0.39 na -2.93 *** 1.16 na
2 medium, 3 large) + or -
0.52 0.93 1.04 1.39
Outreach (1 small, + or - -0.71 * -0.20 na 0.68 na -1.01 na
2 medium, 3 large) 0.43 1.11 1.00 1.79
Number of Women + or - 1.06 na -2.50 na 3.94 ** -7.26 na
Borrowers 0.77 3.49 1.75 6.32
Social Efficiency 0.06 *** 0.05 * 0.06 *** 0.02 na
Index + or -
0.02 0.03 0.02 0.04
Adjusted R-squared 0.58 0.30 0.55 0.39
F-statistic 13.75 6.30 16.58 10.62
Durbin-Watson Statistic 1.19 0.88 1.34 0.84
Cross-Sections Included 43 28 45 29
Total Unbalanced Panel 102 136 140 165
Observations
Years Covered 2003-2010 2003-2011 2001-2011 2000-2011
Source: Author’s Estimates, MIX Market Data (http://www.mixmarket.org/)
Note: Italicized numbers refer to the standard deviation. Regression uses White diagonal
standard errors and covariance (degrees of freedom corrected). All p-values of the F-statistic are
very significant or p less than 0.01. p-values used are: *** p less than 0.01, ** p less than 0.05
and greater than 0.01, * p less than 0.15 and greater than 0.05. Swamy and Arora estimator of
component variances.

68
 The results of the regressions may indicate that NGOs are devoting resources on

operations which will enable them to extend loans to micro-borrowers. Efficient

operations seem to be compromised by the need to monitor, educate and follow-up

repayments done by micro-entrepreneurs. This phenomenon may explain why, in spite of

a levelling-off of the cost per dollar of loan as well as transactions cost for rural banks,

their social efficiency index is positively related to cost. This observation may be verified

by the comparison of operational self-sufficiency and the social efficiency index. This

result may suggest that NGOs have to re-think their organization and operations so as to

achieve better levels of operational self-sufficiency and social performance efficiency, as

their attempt to increase the outreach of their social projects might be comprised with less

efficient operations and therefore sacrificing financial sustainability.

The extent of capitalization of the MFI may be indicated by the debt-to-equity

ratio and the gross loan portfolio over assets. In principle, MFIs should not lend beyond

their level of assets. Thus, more efficient operations may be indicated by a negative

relationship of this indicator with costs. Both MFIs and rural banks seem to reflect this

negative relationship. However, the structure of debt would be indicated by the debt-to-

equity ratio. Efficiency would again be reflected by a negative relationship of this

indicator with costs. However, this behavior can only be observed with rural banks as the

debt-to-equity ratio has a positive relationship with costs for NGOs. This regression

result may verify that NGOs have an extensive source of grants that help cover their

outreach activities. However, such expenses do not seem to allow a healthy level of

operational self-sufficiency and social efficiency index for NGOs. (See Figure 4.5)

69
 It can be observed that a social efficiency index which is greater than 100 is likely

to have an operational self-sufficiency that is lower than 1.0. This trend is most observed

with NGOs. The social efficiency index for some rural banks, though, may fall within

the moderate range but would report an operational self-sufficiency level which is lower

than 1.0. Most of these rural banks loan more than 10%, the regulated level by the

Bangko Sentral ng Pilipinas, of their credit line to MFIs.

The research hypotheses that were tested in the study led to the following results:

1. Hypothesis for rural banks. Operating costs for rural banks have manifested

higher experience or learning curve spillovers and lesser transactions cost than NGOs.

This phenomenon is observed to be explained by their longer commercial lending

experience, and, their strict observance of industry standards for operational self-

70
sufficiency. With efficiency gains, rural banks are operating at the low marginal cost

levels in the supply curve attributed to the individual lending mode of transactions done

among rural banks involved with MFI operations which follow the traditional or market-

oriented manner of transacting with clients. Lower transactions cost among rural banks

enable them to achieve high operational efficiency. Empirically, the presence of lower

transactions cost denoting high learning or spillover effects have enabled rural banks to

operate efficiently thereby allowing them to perform their outreach activities with more

productivity and lesser cost. The satisfactory level of their social efficiency index

indicates that financial and operational expenses are not compromised with extensive

social outreach. All these indicators seem to precede the eventual occurrence of an

upward sloping marginal cost curve, and thereby, allowing rural banks serving MFI

clients to initiate operating at the well-behaved portion of the supply curve.

2. Hypothesis for NGOs. Operating costs for NGOs have manifested lower

experience or learning curve spillovers and have higher transactions cost than rural

banks.

This phenomenon may be due to their shorter commercial lending experience,

and, the non-regulated nature of their operations. Transactions cost among NGOs are

high and are less likely to attain industry standards for operational self-sufficiency. With

lesser efficiency gains, some NGOs may be experiencing disincentive effects, that is,

being unable to increase their average loans credited to micro-borrowers, and, NGOs

may not yet be operating at low marginal cost levels in the supply curve. This can be

explained by the group lending with individual liability lending mode of transactions

done among NGOs involved with MFI operations which do not follow the traditional or

71
market-oriented manner of transacting with clients. Loan officers and personnel have to

achieve a certain level or manner of negotiating capabilities to effectively deal with

micro-borrowers. While still at the high level of transactions and operating costs, NGOs

into microfinance are not yet in the capacity to operate at the well-behaved portion of the

marginal cost curve, also known as the upward sloping supply curve. More efficient

operations will have to be achieved while they attempt to accomplish their social

outreach. Their social efficiency index seems to denote that financial resources are

sacrificing operations thereby disallowing NGOs to operate more efficiently.

In terms of objectives of the study, the results are denoted by Table 4.10. The

downward trend of cost per dollar of loan and transactions cost through time, indicated

by age, establishes the presence of an experience or learning curve spillover for rural

banks and NGOs. (See Figures 4.1 and 4.2) This is the result when the proxy variable in

the scatterplot for cumulated output is time, that is, age upon establishment, per firm

included in the data. When the proxy variable for cumulated output in the scatter plot is

average loan balance per borrower, one can observe a very steep downward sloping

marginal cost curve for NGOs and a U-shaped marginal cost curve for rural banks when

the unit cost used is either cost per dollar of loan or transactions cost. (See Figures 4.3

and 4.4). This result establishes the need for a downward sloping with a levelling-off

trend for cost per dollar of loan and transactions cost, before the marginal cost curve

starts to increase. This phenomenon is observed only among rural banks, and denotes a

U-shaped supply curve.

Efficient levels of operational self-sufficiency seems to coincide with the

downward trend of transactions cost and cost per dollar of loan for both rural banks and

72
NGOs. However, social outreach efficiency, indicated by a low social efficiency index,

as of the moment, does not seem to coincide with the trend observed with costs as the

social efficiency index has a positive relationship with costs.

Table 4.10
Experience or Learning Curve Spillovers, Transactions Cost and a U-Shaped
Supply Curve for Rural Banks and NGOs
Experience or Learning Experience or Learning Curve Spillovers,
MFIs Curves Transactions Cost and a U-Shaped Supply Curve
(Objective 1) (Objective 2)
1. Lower transactions cost, lower cost per dollar of
loan, leading to a U-shaped supply curve
Higher experience or
Rural
learning curve spillovers
Banks 2. Excellent level of the social efficiency index but
translate to faster learning
has a positive relationship to operating and
transactions costs
1. Higher transactions cost and operating
expenses. NGOs still operate at the steep portion
Lower experience or
of the marginal cost curve
learning curve spillovers
NGOs
translate to slower to
2. Poor level of the social efficiency index but has
moderate learning
a positive relationship to operating and
transactions cost
Source: Author’s Regression Results

73
 CHAPTER V

SUMMARY, CONCLUSION AND RECOMMENDATION

The results of the study can be summarized by Table 5.1. The combination of

low operating and transactions costs, due to a fast learning environment due to high

experience or learning curve spillovers, all seem to be positively related with a

satisfactory or low social efficiency index. This combination of indicators allows the

MFI to reach a minimum level of marginal costs, while at the same time allowing MFIs

to operate efficiently and achieve an appropriate level of outreach. Higher levels of loans

transacted by micro-entrepreneurs entail higher personnel and operational expenses in

order to properly monitor larger loan portfolios. Only rural banks manifest this trend.

On the other hand, NGOs have to improve learning from the operations of other

NGOs, thereby eventually achieving a higher level of spill-over effects. As of the

moment, with a wide variety of micro-borrowers with very small loans, operational and

transactions costs are still high, although showing a downward or decreasing trend

through time.

The results lead to the following conclusion: high experience or learning curve

spillovers in the microfinance industry allow lenders to learn from the experience of other

lenders negotiating with micro-entrepreneurs. Learning from the experience of other

MFIs allows the entire industry to reach a low level of operating and transactions costs.

This level of costs, however, enables the MFI to appropriately reach its social outreach

mission while at the same time achieve operational self-sufficiency, the indicator for

financial viability used in the study. This phenomenon, though, is being observed only

74
among rural banks. NGOs are still operating at the steep portion of marginal costs but

operating and transactions costs are showing a decreasing trend through time.

It is recommended that the observed relationships between the experience curve,

operating costs, transactions costs and social efficiency be verified among a sample of

MFIs so as to tract those costs which allow transactions and negotiations among micro-

borrowers to be costly. Focusing the study on a few MFIs, who would represent a

substantial proportion of active borrowers in the Philippines, would enable researchers to

determine those costs, specifically costs related to human asset specificity, which explain

why NGOs continue to operate at the steep portion of the marginal cost curve. With this

intent, there seems to be a rationale behind the existence of a U-shaped supply curve for

the microfinance industry of the Philippines, and for the microfinance industry, in

general.

Verification of the existence of a U-shaped supply curve, through the presence of

significant experience or learning curve spillovers before reaching the minimum level of

marginal costs, may lead to the necessary conditions by which micro-lenders are able to

reach an efficient level of costs and risks. This efficient level of operations, which

combine financial viability and social outreach, may be the starting point through which

the microfinance industry, namely rural banks and NGOs, be included in the financial

sector.

75
 BIBLIOGRAPHY

Ahlin, Christian, Jocelyn Lin, Michael Maio 2011 Where does microfinance flourish?
Microfinance institution performance in macroeconomic context. Journal of
Development Economics 95, 105-120.

Akerlof, George A. 1970 The Market For ‘Lemons’: Quality Uncertainty and the Market
Mechanism. Quarterly Journal of Economics 84, 3: 488-500.

Armendáriz, Beatriz and Arianne Szafarz 2011 On Mission Drift in Microfinance

Institutions, in The Handbook of Microfinance” Beatriz Armendáriz and Marc
Labie, editors. MA, USA:World Scientific Publishing, pp. 341-366

Arrow, Kenneth J. 1962 The economic implications of learning by doing. Review of

Economic Studies 29 (3), 155–173.

Ashta, Arvind 2009 Microcredit Capital Flows and Interest Rates: An Alternative
Explanation Journal of Economic Issues Vol. XLIII No. 3 September, pp. 661-
683 (DOI 10.2753/JEI0021-3624430305. Accessed: September 7, 2013)

Balkenhol, Bernd and Marek Hudon 2011 Efficiency in The Handbook of

Microfinance, Beatriz Armendáriz and Marc Labie, editors. MA, USA: World
Scientific Publishing, pp. 383-396

Bangko Sentral ng Pilipinas 2011 Financial Inclusion Initiatives Year-End Report

Bazinzi N, Mangeni P, Nakabuye Z, Brendah A, Agasha E. 2013 Transaction Costs and

Outreach of Microfinance Institutions in Uganda. Issues in Business Management
and Economics 1(6):125-132 (http://www.journalissues.org/journals-
home.php?id=2, Accessed: September 7, 2013)

Besanko, David, David Dranove, Mark Shanley, Scott Schaefer 2013 Economics of
Strategy. Sixth Edition. New Jersey, USA: John Wiley and Sons. pp. 119-120,
139-141

Boston Consulting Group. 1972. Perspectives on Experience. Boston Consulting Group,

Boston, MA

Copestake, James 2007 Mainstreaming Microfinance: Social Performance Management

or Mission Drift? World Development Vol. 35, No. 10. pp. 1721-1738
(DOI:10.1016/j.worlddev.2007.06.004. Accessed: February 27, 2012)

76
Cull, Robert, Asli Demirgüç-Kunt and Jonathan Morduch 2007 Financial Performance
and Outreach: a Global Analysis of Leading Microbanks Economic Journal 117,
F107-F133, February.

Dacanay, Jovi C. 2012 Can Outreach and Sustainability Co-Exist in the Microfinance
Industry of the Philippines? An NGO and Rural Banks Performance Analysis
Working Paper. Submitted to the Faculty of the Department of Economics for
Workshop in Economics 2. School of Social Sciences, Ateneo de Manila
University, March.

Economic Intelligence Unit (EIU) 2013 Global Microscope on the Microfinance Business
Environment. London, UK: Economic Intelligence Unit

Ghemawat, Pankaj and A. Michael Spence 1985 Learning Curve Spillovers and Market
Performance The Quarterly Journal of Economics Vol. 100 Supplement, pp.
839-852 (http://www.jstor.org/stable/1882926 Accessed: 28/03/2014)

Jimenez, Eduardo C., Pia Bernadette Roman 2011 Case Study on Philippines Electronic
Banking: Delivering Microfinance Services to the Poor in the Philippines”
Working Paper. Bangko Sentral ng Pilipinas

Karlan, Dean S. and Jonathan Morduch 2009 Access to Finance” Chapter 2. Handbook of
Development Economics. Dani Rodrik and Mark Rosenzweig (eds). North
Holland: Elsevier Science, Volume 5.

Karlan, Dean S. and Jonathan Zinman 2010 Expanding Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts Review of Financial
Studies 23(1): 433-464

McIntosh, Craig and Bruce Wydick 2005 Competition and Microfinance Journal of
Development Economics 78, 271-298.

Microfinance Council of the Philippines 2006 Philippine Country Profile on

Microfinance

Microfinance Information Exchange Portal 2011 Microfinance Report for the

Philippines. (http://www.mixmarket.org/) (Accessed: February 15, 2013)

MicroRate 2014 The Technical Guide: Performance and Social Indicators for
Microfinance Institutions. www.ada-microfinance.org; www.microfact.org.
(Accessed: April 15, 2014)

Randon, Emanuela, Ahmad Naimzada 2006 Dynamics of the non linear learning curve
with spillovers in a differentiated oligopoly: effects on industry structure” Journal
of Evolutionary Economics 17: 95–106 (DOI 10.1007/s00191-006-0032-6,
Accessed October 20, 2014)

77
Riordan, Michael H., Oliver E. Williamson 1985 Asset Specificity and Economic
Organization International Journal of Industrial Organization 3: 365-378.

Rosenberg, Richard 2002 Microcredit Interest Rates. Consultative Group to Assist the
Poor (CGAP): Washington DC

Rosenberg, Richard, Scott Gaul, William Ford, Olga Tomilova 2013 Microcredit Interest
Rates and Their Determinants 2004-2011 Access to Finance Forum. Report by
the Consultative Group to Assist the Poor (CGAP) and its Partners. No 7 June

Saito KA, Villanueva, DP 1981 Transaction Costs of Credit to the Small-Scale Sector in
the Philippines Economic Development and Cultural Change 29(3):631-640.

Shankar, Svita 2007 Transaction Costs in Group Microcredit in India Management

Decision, Volume 45 No. 8 pp. 1331-1342

Spence, A. Michael 1981 The Learning Curve and Competition The Bell Journal of
Economics Vol. 12 No. 1 (Spring) pp. 49-70 (http://www.jstor.org/stable/3003508
Accessed: 23/03/2014)

Stiglitz, Joseph E. and Andrew Weiss 1981 Credit Rationing in Markets with Imperfect
Information American Economic Review 71, 3: 393-410.

Tchamanbé Djiné, Louise 2012 Location and Performance of Microfinance Institutions

Asian Research Publishing Network (ARPN) Journal of Science and Technology
Vol. 2, No. 3 April (http://www.ejournalofscience.org)

Tchamanbé Djiné L. 2009 Les faillites bancaires en Afrique subsaharienne. Paris:

Harmattan

Thompson, Peter 2010 Learning by Doing Chapter 10. In Handbook of the Economics of
Innovation Volume 1. Edited by Bronwyn H. Hall and Nathan Rosenberg. North
Holland: Elsevier Science, pp. 429-476.

Williamson, Oliver 1989 Transaction Cost Economics Chapter 3. In Handbook of

Industrial Organization Volume 1. Edited by R. Schmalensee and R. D. Willig.
North Holland: Elsevier Science, pp. 150-159.

Williamson, Oliver 1975 Markets and Hierarchies: Analysis and Antitrust Implications.
New York, USA: Free Press.

78
APPENDICES

79
 Table A.1
Descriptive Statistics (2003 to 2011)

Average Percentage
Selected Gross
Loan Scale Outreach of Women
Descriptive Transactions Operational Number of Loan Portfolio Age
Balance Real (1-Small; (1-Narrow; Borrowers
Statistics with Cost Self- Active Portfolio at Risk (0-Young;
per Yield 2-Medium; 2-Moderate; to Total
common (in US$) Sufficiency Borrowers over (90 days) 1-Mature)
Borrower 3-Large) 3-Broad) Active
samples Assets
(US$) Borrowers

Rural Banks
Mean 48.78 1.16 703.03 24,122 0.25 0.65 0.06 0.93 2.13 1.73 57%
Median 20.48 1.13 472.00 12,007 0.23 0.69 0.05 1 2 2 60%
Maximum 747.31 1.76 5,731.00 311,380 0.67 0.89 0.44 1 3 3 100%
Minimum 3.20 0.48 46.00 208 (0.02) 0.32 0 0 1 1 2%
Std. Dev. 97.66 0.16 745.78 43,471 0.12 0.14 0.07 0.25 0.68 0.73 29%

80
Skewness 5.23 0.23 3.90 4.74 1.19 (0.74) 2.24 (3.42) (0.16) 0.47 (0.20)
Kurtosis 32.87 7.59 24.20 27.97 5.02 2.71 11.69 12.70 2.19 2.01 1.83
Observations 117 117 117 117 117 117 117 117 117 117 117
NGOs
Mean 151.78 1.09 114.52 73,105 0.46 0.71 0.06 0.72 1.82 2.27 93%
Median 77.13 1.09 96.00 25,321 0.45 0.70 0.03 1 2 2 99%
Maximum 1,153.11 1.74 470.00 606,488 0.75 1.19 0.48 1 3 3 110%
Minimum 16.00 0.38 44.00 1,157 0.13 0.42 (0.09) 0 1 1 1%
Std. Dev. 180.98 0.20 68.55 100,502 0.12 0.15 0.09 0.45 0.78 0.76 12%
Skewness 2.68 (0.58) 2.96 2.51 0.17 0.73 2.69 (1.00) 0.33 (0.50) (3.64)
Kurtosis 11.89 6.01 13.91 10.64 2.53 3.58 10.72 2.00 1.72 1.88 24.19
Observations 141 141 141 141 141 141 141 141 141 141 141
Source: MIX Data 2013
Notes: If MFIs intend to fund growth via commercial funds, then the transactions cost should be from US$ 10 to US$ 25; OSS has to be greater
than 1.0 and Portfolio at risk (90 days) should be less than 0.05. Transactions cost is calculated as administrative and personnel expenses
over average loan portfolio per borrower.
 Table B.1
Market Share of Selected Rural Banks (2009-2011)
Market Share (in %)
Gross Number
No. Rural Banks into Microfinance Loan of Active Gross Number
Lending Portfolio Borrowers Loan of Active
(in US$) Portfolio Borrowers
1 1st Valley Bank 69,078,307 64,719 13.6 7.1
2 Banco Santiago de Libon 3,397,136 11,407 0.7 1.3
3 Bangko Kabayan 22,397,278 11,394 4.4 1.3
4 Bangko Luzon 7,918,003 9,610 1.6 1.1
5 Bangko Mabuhay 9,191,583 6,715 1.8 0.7
6 Bukidnon Cooperative Bank 5,509,136 1.1
(BCB)
7 Cantilan Bank 16,974,467 21,305 3.3 2.3
8 Center for Agricultural and Rural 38,251,373 240,615 7.5 26.4
Development (CARD Bank)
9 Cooperative Bank of Misamis 11,431,706 24,029 2.2 2.6
Oriental (CBMO)
10 First Agro-Industrial Rural Bank 8,782,969 20,539 1.7 2.3
(FAIR Bank)
11 First Isabela Cooperative Bank 25,260,159 25,072 5.0 2.8
(FICO)
12 First Macro Bank 11,477,170 7,618 2.3 0.8
13 GM Bank 28,798,526 31,932 5.6 3.5
14 Rural Green Bank of Caraga 36,534,854 60,005 7.2 6.6
(Green Bank)
15 Kausawagan Bank (K Bank) 5,410,927 33,815 1.1 3.7
16 Mallig Plains RB 7,454,387 23,133 1.5 2.5
17 New RB of Victorias 1,987,736 2,529 0.4 0.3
18 OMB 5,927,615 31,162 1.2 3.4
19 People's Bank of Caraga (PBC) 20,134,877 55,808 3.9 6.1
20 PR Bank 48,639,582 9.5
21 Progressive Bank 6,902,829 28,210 1.4 3.1
22 RB Bagac 1,041,571 622 0.2 0.1
23 RB Camalig 7,924,049 11,137 1.6 1.2
24 RB Cotabato 432,914 0.1
25 RB Digos 7,815,288 7,921 1.5 0.9
26 RB Dipolog 12,968,218 2.5
27 RB Guinobatan 3,527,309 12,711 0.7 1.4
28 RB Katipunan 15,551,204 36,324 3.1 4.0
29 RB Labason 2,123,021 1,625 0.4 0.2
30 RB Lebak 3,716,046 4,875 0.7 0.5
31 RB Liloy 1,677,402 0.3

81
32 RB Loon 722,513 508 0.1 0.1
33 RB Mabitac 5,273,418 13,037 1.0 1.4
34 RB Makiling 3,650,288 3,005 0.7 0.3
35 RB Montevista 5,667,814 22,399 1.1 2.5
36 RB Oroquieta 4,684,648 4,708 0.9 0.5
37 RB Pagbilao 817,624 705 0.2 0.1
38 RB San Enrique 1,607,953 3,456 0.3 0.4
39 RB Siargao 2,348,886 7,096 0.5 0.8
40 RB Solano 2,859,520 5,146 0.6 0.6
41 RB Talisayan 4,759,958 18,743 0.9 2.1
42 Opportunity
(OK Bank)
Kausawagan Bank 6,155,402 31,162 1.2 3.4
43 Valiant RB 22,992,262 15,567 4.5 1.7
Average/Hirschman-Herfindahl Index 11,855,301 23,957 0.0561 0.0964
Source: MIX Website

82
 Table B.2
Market Share of Selected NGOs (2009-2011)
Market Share (in %)
Gross Loan Number
No. NGOs into Microfinance Lending Portfolio of Active Gross Number
(in US$) Borrowers Loan of Active
Portfolio Borrowers
1 ABS-CBN 3,001,218 1.2
2 ARDCI 4,074,446 22,123 1.6 0.9
3 ASA Philippines 24,415,521 298,970 9.8 12.5
4 Ahon sa Hirap, Inc (ASHI) 4,515,103 21,838 1.8 0.9
5 Alalay sa Kaunlaran, Inc. (ASKI) 13,900,604 53,919 5.6 2.2
6 Center for Agricultural and Rural
Development (CARD NGO) 56,508,101 580,999 22.7 24.2
7 Community
(CEVI)
Economic Ventures, Inc. 4,469,872 33,832 1.8 1.4
8 Cebu Micro-Enterprise Development
Foundation, Inc. (CMEDFI) 1,442,999 10,905 0.6 0.5
9 Daan sa Pagunlad, Inc. (DSPI) 3,263,197 54,926 1.3 2.3
10 Ecumenical
(ECLOF - PHL)
Church Loan Fund 2,364,657 5,806 0.9 0.2
11 FCBFI 2,009,112 15,626 0.8 0.7
12 Hagdanan sa Pag-uswag Foundation,
Inc. (HSPFI) 1,693,025 17,814 0.7 0.7
13 Kasagana-Ka
Inc.
Development Foundation, 1,751,571 17,236 0.7 0.7
14 Kazama Grameen, Inc. 3,073,497 27,536 1.2 1.1
15 Kasanyangan-Mindanao
Inc. (KCCDFI)
Foundation, 2,833,290 24,434 1.1 1.0
16 Kabalikat para sa Maunlad na Buhay,
Inc. (KMBI) 13,757,522 222,070 5.5 9.3
17 Life Bank Foundation, Inc. 19,829,145 236,917 7.9 9.9
18 Mentors Philippines 1,282,394 11,268 0.5 0.5
19 MILAMDEC Foundation, Inc. 2,414,600 30,661 1.0 1.3
20 Negros Women for Tomorrow
Foundation, Inc. (NWTF) 10,987,599 93,637 4.4 3.9
21 Pagasa 8,679,634 99,317 3.5 4.1
22 People's Alternative Livelihood
Foundation of Sorsogon, Inc. (PALFSI) 2,027,240 13,834 0.8 0.6
23 RSPI 2,506,870 30,795 1.0 1.3
24 Serviamus 1,057,999 10,452 0.4 0.4
25 Taytay sa Kauswagan, Inc. (TSKI) 22,898,570 183,455 9.2 7.7
26 Tulay sa Pag-unlad, Inc. (TSPI) 32,609,652 278,018 13.1 11.6
27 VEF 2,101,619 1,695 0.8 0.1
Average/Hirschman-Herfindahl Index 9,239,595 92,234 0.1041 0.1173
Source: MIX Website

83
 Table C.1
Experience Curve Regression on Cost per Dollar of Loan
Dependent Variable: OPEXPLOAN
Method: Panel EGLS (Cross-section random effects)
Date: 10/17/14 Time: 16:46
Sample (adjusted): 2001 2011
Periods included: 11
Cross-sections included: 45
Total panel (unbalanced) observations: 140
Swamy and Arora estimator of component variances

Variable Coefficient Std. Error t-Statistic Prob.

C 69.52489 6.409659 10.84689 0.0000

OSS -11.57717 2.806924 -4.124504 0.0001
ALBPERBOR -0.001802 0.000921 -1.955402 0.0527
NOACTIVEBORR 8.81E-05 1.72E-05 5.114304 0.0000
DEBTEQUITYRATIO -0.895406 0.287716 -3.112113 0.0023
GLPASSETS -30.90733 4.793453 -6.447821 0.0000
ACTUALAGE -0.403047 0.219714 -1.834419 0.0689
ACTUALAGE^2 0.004898 0.003657 1.339396 0.1828
SCALE -3.556822 1.034076 -3.439613 0.0008
OUTREACH 1.083155 0.961189 1.126891 0.2619
WOMENBORR 3.591124 1.773621 2.024741 0.0450
SOCEFFINDEX 0.067404 0.019906 3.386151 0.0009

Effects Specification
S.D. Rho

Cross-section random 5.190107 0.6498

Idiosyncratic random 3.810356 0.3502

Weighted Statistics

R-squared 0.570339 Mean dependent var 8.023466

Adjusted R-squared 0.533415 S.D. dependent var 6.282887
S.E. of regression 4.202551 Sum squared resid 2260.664
F-statistic 15.44627 Durbin-Watson stat 1.351327
Prob(F-statistic) 0.000000

Unweighted Statistics

R-squared 0.731749 Mean dependent var 22.90631

Sum squared resid 6312.152 Durbin-Watson stat 0.534819

84
 Table C.2
Experience Curve Regression on Personnel Cost per Dollar of Loan
Dependent Variable: PERSONEXPLOAN
Method: Panel EGLS (Cross-section random effects)
Date: 10/17/14 Time: 16:47
Sample (adjusted): 2003 2010
Periods included: 8
Cross-sections included: 43
Total panel (unbalanced) observations: 102
Swamy and Arora estimator of component variances

Variable Coefficient Std. Error t-Statistic Prob.

C 25.93240 3.470638 7.471939 0.0000

OSS -6.457636 1.691225 -3.818318 0.0002
ALBPERBOR -0.000548 0.000402 -1.363177 0.1762
NOACTIVEBORR 4.66E-05 6.71E-06 6.942789 0.0000
DEBTEQUITYRATIO -0.307186 0.141670 -2.168321 0.0328
GLPASSETS -7.483822 2.292931 -3.263868 0.0016
ACTUALAGE -0.101536 0.114545 -0.886428 0.3777
ACTUALAGE^2 0.000975 0.001838 0.530524 0.5971
SCALE -0.862356 0.513587 -1.679084 0.0966
OUTREACH -0.647553 0.424644 -1.524930 0.1308
WOMENBORR 1.120842 0.768798 1.457914 0.1483
SOCEFFINDEX 0.079024 0.019435 4.066050 0.0001

Effects Specification
S.D. Rho

Cross-section random 2.587678 0.8069

Idiosyncratic random 1.265904 0.1931

Weighted Statistics

R-squared 0.622813 Mean dependent var 2.769111

Adjusted R-squared 0.576713 S.D. dependent var 2.224399
S.E. of regression 1.399094 Sum squared resid 176.1717
F-statistic 13.50989 Durbin-Watson stat 1.249467
Prob(F-statistic) 0.000000

Unweighted Statistics

R-squared 0.631175 Mean dependent var 9.813529

Sum squared resid 930.3512 Durbin-Watson stat 0.235696

85
 Table C.3
U-Shaped Supply Curve Regression on Cost per Dollar of Loan
Dependent Variable: OPEXPLOAN
Method: Panel EGLS (Cross-section random effects)
Date: 10/17/14 Time: 16:48
Sample (adjusted): 2001 2011
Periods included: 11
Cross-sections included: 45
Total panel (unbalanced) observations: 140
Swamy and Arora estimator of component variances

Variable Coefficient Std. Error t-Statistic Prob.

C 68.22808 5.855925 11.65112 0.0000

OSS -10.80088 2.794467 -3.865095 0.0002
ALBPERBOR -0.006290 0.002275 -2.764963 0.0065
ALBPERBOR^2 9.31E-07 4.14E-07 2.247778 0.0263
NOACTIVEBORR 8.60E-05 1.72E-05 5.003892 0.0000
DEBTEQUITYRATIO -0.999595 0.279644 -3.574521 0.0005
GLPASSETS -32.69619 4.715825 -6.933292 0.0000
ACTUALAGE -0.106115 0.067341 -1.575799 0.1175
SCALE -2.928632 1.041321 -2.812421 0.0057
OUTREACH 0.680822 0.969660 0.702125 0.4839
WOMENBORR 3.941804 1.754660 2.246478 0.0264
SOCEFFINDEX 0.061478 0.020653 2.976739 0.0035

Effects Specification
S.D. Rho

Cross-section random 4.741461 0.6029

Idiosyncratic random 3.848184 0.3971

Weighted Statistics

R-squared 0.587588 Mean dependent var 8.722539

Adjusted R-squared 0.552147 S.D. dependent var 6.520431
S.E. of regression 4.283784 Sum squared resid 2348.903
F-statistic 16.57904 Durbin-Watson stat 1.341756
Prob(F-statistic) 0.000000

Unweighted Statistics

R-squared 0.739874 Mean dependent var 22.90631

Sum squared resid 6120.952 Durbin-Watson stat 0.548509

86
 Table C.4
U-Shaped Supply Curve Regression on
Personnel Cost per Dollar of Loan
Dependent Variable: PERSONEXPLOAN
Method: Panel EGLS (Cross-section random effects)
Date: 10/17/14 Time: 16:49
Sample (adjusted): 2003 2010
Periods included: 8
Cross-sections included: 43
Total panel (unbalanced) observations: 102
Swamy and Arora estimator of component variances

Variable Coefficient Std. Error t-Statistic Prob.

C 26.10448 3.316345 7.871461 0.0000

OSS -6.291076 1.690990 -3.720351 0.0003
ALBPERBOR -0.001984 0.001271 -1.560691 0.1221
ALBPERBOR^2 2.66E-07 2.19E-07 1.216618 0.2269
NOACTIVEBORR 4.71E-05 6.70E-06 7.038634 0.0000
DEBTEQUITYRATIO -0.318624 0.141350 -2.254149 0.0266
GLPASSETS -7.784014 2.264328 -3.437670 0.0009
ACTUALAGE -0.040065 0.035605 -1.125252 0.2635
SCALE -0.730180 0.520832 -1.401949 0.1644
OUTREACH -0.706997 0.427059 -1.655500 0.1013
WOMENBORR 1.063388 0.766034 1.388172 0.1685
SOCEFFINDEX 0.064824 0.022642 2.862972 0.0052

Effects Specification
S.D. Rho

Cross-section random 2.587760 0.8078

Idiosyncratic random 1.262253 0.1922

Weighted Statistics

R-squared 0.626973 Mean dependent var 2.761836

Adjusted R-squared 0.581381 S.D. dependent var 2.221757
S.E. of regression 1.389767 Sum squared resid 173.8307
F-statistic 13.75175 Durbin-Watson stat 1.195439
Prob(F-statistic) 0.000000

Unweighted Statistics

R-squared 0.633882 Mean dependent var 9.813529

Sum squared resid 923.5229 Durbin-Watson stat 0.224419

87