the impact of microcredit on poverty and women's
-
Upload
khangminh22 -
Category
Documents
-
view
0 -
download
0
Transcript of the impact of microcredit on poverty and women's
THE IMPACT OF MICROCREDIT ON POVERTY AND WOMEN’S
EMPOWERMENT:
A CASE STUDY OF BANGLADESH
By
SAYMA RAHMAN
A Thesis Submitted in Fulfilment of the Requirements
For the Award of the Degree
of
Doctor of Philosophy
School of Economics and Finance
College of Law and Business
University of Western Sydney
Sydney Australia
March 2007
ii
THE IMPACT OF MICROCREDIT ON POVERTY AND WOMEN’S
EMPOWERMENT: A CASE STUDY OF BANGLADESH
Abstract
The microcredit program in Bangladesh is a unique innovation of credit delivery
designed to enhance the income generating activities of the poor. Its uniqueness is
reflected in its collateral-free group-based lending strategy. The program extends
small loans to poor people, mainly women, for self-employment activities thus
allowing clients to achieve a better quality of life. This program is regarded as a very
exciting anti-poverty tool for the poorest, especially for women.
This study investigates the impact of microcredit on economic indicators as well as
consumption behaviour of the borrowers. It further analyses the impact of
microcredit on women’s empowerment. Primary data has been collected from the
borrowers of two major microcredit institutions in Bangladesh. Alongside the
borrowers, data have also been collected from non-borrowers of the same village to
compare the impact between borrowers and control group. The empirical work has
used sophisticated econometric techniques. Five different econometric methods -
OLS, 2SLS, Probit, Tobit and SURE estimators - have been applied to the sample
data of this study.
The most important finding indicates that microcredit programs are effective in
increasing borrowers’ income, assets and consumption but it is more pronounced
towards high income borrowers than low income borrowers. It further finds that
microcredit programs are empowering for women.
iii
STATEMENT OF AUTHENTICATION
I, Sayma Rahman, declare that this thesis has not been submitted, either in whole or
in part, for a degree at this university or any other academic institution. I also certify
that the work presented in this thesis is, to the best of my knowledge and belief, my
own work and original except as acknowledged in the text.
Sayma Rahman
------------------------------
Signature of Candidate
iv
DEDICATION
TO MY MOTHER, SISTERS AND BROTHERS
FOR YOUR FAITH IN ME
TO AWAB
FOR ENCOURAGING AND SUPPORTING ME
WITH MY LOVE…
v
ACKNOWLEDGEMENTS
Thank you, to the following people who assisted me and made it possible for me to
complete this thesis by offering their support, encouragement and professional
consultation in different ways:
First of all, to Professor P N (Raja) Junankar, my principal supervisor, thank you so
much for all the valuable advice and encouragement that I have received from you
for your professional support and for entrusting me with this work. Thank you for all
the confidence and generosity you showed me during my PhD study, especially in
the last few months.
To my co-supervisor, Dr Girijasankar Mallik, thank you for your supervision and
discussions, and for your interest and comments throughout this study. I would like
to express my gratitude to you for your time and encouragement that you have
provided me throughout this study. Thank you for helping me with data entry and
compilation.
I also express my sincere gratitude to Professor Anis Chowdhury for providing me
with updates on the issue of my research whenever he found any. My appreciation
goes to Dr Sudhir Lodh for his comments at different stages of my thesis. Special
thanks and appreciation go to Muhammad Arifur Rahman for providing me
comments and support. Many thanks to all my friends and colleagues for their
friendly assistance and encouragement in many matters during my PhD study. I also
wish to thank Angela Damis for proof reading my thesis and all the academic and
vi
administrative staff in the School of Economics and Finance at the University of
Western Sydney for necessary support and help.
Thank you all; without your support, this study would have not been possible.
Sayma Rahman
vii
TABLE OF CONTENTS
-----------------------------------------------------------------------------------------------------
Abstract ii
Statement of Authentication iii
Dedication iv
Acknowledgements v
List of Tables and Figures xi
Chapter One: Scope and Framework of the Study………………………….…....1
1.1 Introduction………………………….…….…………………............1
1.2 How Far the Grameen Bank Differs from BRAC……………………3
1.3 Research Problem and Question……….….………..…………….......4
1.4 Why is the Problem Worthy of Research?...........................................6
1.5 Data and the Econometric Approach…………………………………8
1.6 Thesis Outline…………………………………………...……………9
Chapter Two: Survey of Literature: Rural Credit Market and the Bangladesh
Economy …………………………………………………………………………...12
2.1 Relevant Theories on Rural Credit Market (Part A)………………..12
2.1.1 The Rural Credit Market……………………………………14
2.1.2 The Monopolistic Nature of the Rural Credit Market………15
2.1.3 Interlinked Factor Markets………………………………….19
2.1.4 Potential Risk………………………………………………..22
2.1.5 Significance of Credit Policy in the Rural Areas…………...23
2.1.6 Conclusion of Part A………………………………………….24
2.2 The Bangladesh Economy: Pre-Microcredit (Part B)……..……......25
2.2.1 Inception of Microcredit Institutions……………………….27
2.2.2 Background of the Grameen Bank……………...………….28
2.2.3 Growth and Expansion of the Grameen Bank.......................29
2.2.4 Savings Mobilisation by the Grameen Bank……………….32
2.2.5 Background of BRAC………….…………………………..33
2.2.6 Growth and Development of BRAC………….……………35
2.2.7 Savings Mobilisation by BRAC………...…….……...…….37
viii
2.3 The Bangladesh Economy and Recent Development………………39
2.3.1 Conclusion of Part B………………………………………..42
2.4 Introduction to Review of Literature (Part C)………………………43
2.4.1 Impact of Microcredit on Poverty ………………………….45
2.4.2 Women’s Empowerment……………………………………54
2.4.3 Household Consumption …………………………………...58
2.4.4 Conclusion of Part C………………………………………….61
Chapter Three: Assessment of Income and Asset Accumulation of Microcredit
Borrowers…………….………………………………………….………....63
3.1 Introduction…………………………………………………………63
3.2 Previous Studies on Impact Assessment…………………………....65
3.2.1 Data …………………………………………………….…..…66
3.2.2 The Selectivity Problem……………………………………....67
3.2.3 Research Questions…………………………………….……..69
3.2.4 Hypotheses……………………………………………………69
3.3 Model Specification…………………………………………………70
3.3.1 Endogeneity of Credit Program……………………………..72
3.3.2 Methodology………………………………………………...72
3.4 Specification of the Instruments and Variables……………………..73
3.4.1 Description of the Variables…………………………...……...74
3.4. 2 Specification of the Variables…………………………..……75
3.5 Estimation Results and Discussion………………………………….78
3.5.1 Descriptive Statistics of the Variables………………… ……78
3.5.2 Impact of Microcredit on Household Outcomes…………...…80
3.5.3 Impact of Microcredit based on Different Income Levels……85
3.5.4 Estimating Censored Regression Problem …………………...89
3.5.5 Censored Problem on different Income levels………………..93
3.6 Conclusion…………………………………………………………..95
ix
Chapter Four: Assessment of Consumption of Microcredit Borrowers
Compared to Non-borrowers………………………………………………….…97
4.1 Introduction……………………………………………………………..97
4.2 Literature Review on Expenditure Elasticity………………………….100
4.3 Introduction to Relevant Theories…………………………..…………104
4.3.1 Theory of Mixed Demand…………………………………...105
4.3.2 An Almost Ideal Demand System…………………………...107
4.3.3 Specification of the AIDS model……………………………108
4.4 Model Specification……………………………………………………112
4.5 Empirical Results……………………………………………………....118
4.5.1 Percentage and Mean Consumption ……………………….118
4.5.2 Preliminary Findings of the Descriptive Table…………...…121
4.5.3 Estimation Results and Discussion…………………………..125
4.5.4 Test of Significant Difference……………………………….134
4.5.5 Testing for Uncorrelated Error Terms…………………….....138
4.6 Conclusion……………………………………………………………..143
Chapter Five: Factors Affecting Women’s Empowerment in Microcredit
Programs...………………………………………………………..148
5.1 Introduction………………...………………………………………….148
5.2 Defining Women’s Empowerment……………………………….……154
5.3 Towards Development of Empowerment Index……………….………156
5.4 Calculation of Empowerment Index…………………………………...159
5.4.1 Economic Security Index……………………………………159
5.4.2 Purchase Decision Index…………………………………….160
5.4.3 Control over Asset Index…………………………………….161
5.4.4 Mobility Index……………………………………………….161
5.4.5 Awareness Index…………………………………………….162
5.4.6 Empowerment Index……………………………………..….163
5.5 Summary Table of the Indices…………………………………………164
5.6 What Causes Women to be Empowered?..............................................170
5.6.1 Research Questions...………………………………… ……171
5.6.2 Hypotheses…………………………………………………..171
5.6.3 Specification of the Variables…………………………….....172
x
5.7 Model Specification…………………….………………..……………172
5.7.1 Model of the Study………………………………………….173
5.8 Estimation Results and Discussion……………………………..……..173
5.8.1 Factors Affecting Empowerment…................................... ....174
5.8.2 Factors Affecting Empowerment for both Groups………..…176
5.8.3 Factors Affecting Empowerment: Different Groups…...…....179
5.8.4 Do Microcredit Programs Empower Women?.........…..…….181
5.9 Conclusion…………………………………………………………......185
Chapter Six: Conclusion and Future Research….……….…………………….188
6.1 Introduction…………………………………………….……………...188
6.2 Summary of the Study………………………………….……………...189
6.3 Key Findings of the Study……………………………………………..192
6.4 Policy Implications………………………………………..…………...200
6.5 Study Limitations………………………………………………..…….201
6.6 Avenues for Future Research………………………………….…..…..202
6.7 Concluding Notes……………………………………………………...203
APPENDICES……………………………………………………………………205
7.1 Appendix A: Global Poverty Figure…………………………….…….205
7.2 Appendix B: The Sixteen Decisions of the Grameen Bank……...……206
7.3 Appendix C: Household Survey Questionnaire………………….……208
7.4 Appendix D: Map of Bangladesh……………………………………..225
7.5 Appendix E: List of the Village……………………………………….226
7.6 Appendix F: Descriptive Tables of Consumption Patterns…………....227
7.7 Appendix G: Experiences from Field Trip…………………………….236
7.8 Appendix H: Abstract of the Paper Published in the FIBR Conference
Proceedings……………………………………………………...…249
7.9 Appendix I: Abstract of the Paper accepted for publication in a
forthcoming issue in the Journal of Development Areas……….…250
7.10 Appendix J: Abstract of the Paper published in the Conference
Proceedings……………………………………………………….251
REFERENCES…………………………………………………………………..252
xi
List of Tables and Figures
List of Tables:
Table 2.1 Number of Schools, Students and Graduates from BRAC 36
Table 2.2 Job Creation by BRAC 36
Table 2.3 Comparing BRAC with the Grameen Bank 38
Table 2.4 Major Socio-Economic Conditions: 1975-2005 40
Table 2.5 Sectoral GDP Growth Rates: 1979/80-1999/00 41
Table 3.1 Correlation Matrix of the Variables 77
Table 3.2 Descriptive Statistics of Dependent and Independent Variables 79
Table 3.3 2SLS Estimation of Amount of Borrowing on
Household Outcome: Log of Total Expenditure 81
Table 3.4 2SLS Estimation of Amount of Borrowing on
Household Outcome: Log of Assets 82
Table 3.5 2SLS Estimation of Amount of Borrowing on
Different Income Level: Log of Total Expenditure 86
Table 3.6 2SLS Estimation of Amount of Borrowing on
Different Income Level: Log of Assets 87
Table 3.7 Tobit Estimation of Amount of Credit Using Censored Data 90
Table 3.8 OLS Estimation of Household Outcomes Using Estimated Value
Of Amount of Credit from Tobit Estimation 92
Table 3.9 OLS Estimation of Household Outcomes Based on Different Income
Level Borrowers: Expenditure (Income) 93
Table 3.10 OLS Estimation of Household Outcomes Based on Different Income
Level Borrowers: Assets 94
Table 4.1 Average Consumption of Borrowers and Non-Borrowers
Of three Districts 120
Table 4.1a T-test Results of Mean Expenditure and Budget Share of Different
items Consumed by Borrowers and Non-Borrowers 121
Table 4.2 OLS Estimation of the AIDS Model 127
Table 4.3 Testing for Differential Slope Coefficient 136
Table 4.4 Seemingly Unrelated Regression Estimates 140
xii
Table 4.5 Testing for Different Slope Coefficients Using SURE 141
Table 4.5a Estimation of Income Elasticity for Borrowers and non-borrowers141
Table 4.6 Comparing and Contrasting OLS and SURE 142
Table 5.1 Defining Women’s Empowerment 151
Table 5.2 Borrowers’ and Non-Borrowers all indices according to Districts 165
Table 5.3 Empowerment Index of Borrowers and Non-Borrowers
According to Districts 166
Table 5.4 Probit Model: Factors Affecting Empowerment Index
Borrowers and Non-Borrowers Separately 175
Table 5.5 Probit Model: Factors Affecting Empowerment Index
Pooling Full Data Set 178
Table 5.6 Probit Model: Factors Affecting Empowerment Index
Based on Different Data Set 181
Table 5.7 OLS Estimation of the Equations 183
Table 5.8 Probit Estimation: Using the Estimated Value of the
Amount of Credit 184
List of Figures:
Figure 2.1 Monopoly Market 17
Figure 2.2 Number of Branches 31
Figure 2.3 Number of Borrowers 31
Figure 4.1 Consumption Tree . 99
1
Chapter One
SCOPE AND FRAMEWORK OF THE STUDY
1.1 Introduction
This research explores the impact of microcredit organisations in the Bangladesh
economy. It is one of the first comprehensive studies that analyses the impact of
microcredit in a broader spectrum. This is an original study not only in terms of use
of a new data set but also in terms of use of econometric techniques1 that have not
been used by other researchers in this area. The academic discourse in this research
is quite relevant and goes a long way to contributing to the existing literature. This
study touches on wide-ranging theoretical issues related to this field and highlights
the developments in the literature on the subject area. Microcredit is not only a very
topical issue the founder of the Grameen Bank has just been awarded the 2006 Nobel
Peace Prize but it has also been a topic of interest to researchers since its inception in
early 1970s. In this research we have analysed the impact of microcredit programs
on the borrowers of two large microcredit institutions in Bangladesh − the Grameen
Bank and the Bangladesh Rural Advancement Committee (BRAC).
The Grameen Bank is the largest moneylending institution in the world, measured in
terms of numbers of borrowers. In 2005 alone, Grameen Bank’s borrower count rose
from 4 to 5 million. Microcredit programs have created a revolution not only in
Bangladesh but also through out the world for their novelty. Microcredit is a unique
1 This is the first study that uses An Almost Ideal Demand System (AIDS) model to analyse the consumption behaviour of the microcredit borrowers, Tobit estimations to analyse the impact of microcredit on household outcomes and Two Stage Least Square estimation to analyse the impact of microcredit on women’s empowerment.
2
innovation in credit delivery techniques that enhances income generating activities.
Its uniqueness is reflected in its collateral-free group-based lending strategies, high
repayment rates and also a special focus on women. A microcredit program2 extends
small loans to poor people, mainly women, for self-employment activities thus
allowing clients an opportunity to achieve a better quality of life. It is the most
sensational anti-poverty tool for the poor people, especially women (Microcredit
Summit 1997, World Bank). For these reasons, microcredit programs in Bangladesh
have drawn the attention of academics, researchers, international agencies and policy
makers throughout the world. The Grameen Bank the largest microcredit institution
and BRAC the largest non-governmental organisation (NGO) – have been the
pioneers of microcredit in Bangladesh for almost three decades.
There are more than a thousand microcredit institutions providing financial and
social development services in Bangladesh (Khalily, Imam and Khan, 2000 p. 105).
The Grameen Bank and BRAC are the major contributors to this credit market.
About 94 % of the borrowers in this market are female. Bangladesh’s microcredit
program is widely known as a ‘lending program to the poor without any collateral’.
The main focus of this endeavour is to eliminate poverty and empower women in
rural country areas in Bangladesh (Khandker, 2003).
2 According to Khalily, Imam and Khan (2000), “program” is synonymously used with “institution.” Program encompasses those microcredit or micro-finance institutions that are providing micro-financial services.
3
1.2 How Far the Grameen Bank Differs from BRAC
In order to achieve the main objective of delivering credit the Grameen Bank and
BRAC both provide various support programs including social services3 and social
development programs along with financial discipline. As part of such initiatives the
Grameen Bank sets out guidelines and codes of conduct for the borrowers promoting
social and financial discipline (see Appendix B). To develop social awareness among
borrowers, BRAC provides skills training, adult literacy and primary health care etc.
BRAC’s social development approach is more extensive (integrated program)
compared to the Grameen Bank’s minimalist (credit only) approach.
One of the central concerns of both the Grameen Bank and BRAC is to support the
poor - making productive use of their credit and income. It should be mentioned that
although there are differences between the Grameen Bank and BRAC in their
approaches to loan provision, both institutions adopt a credit based poverty
alleviation approach in rural Bangladesh.
The credit delivery model developed by the Grameen Bank does not require tangible
collateral. The “group liability” may be interpreted as intangible collateral in the
sense that, in case of default, a member loses its opportunity for future loans. The
model developed by the Grameen Bank has been replicated in many parts of the
world (Khandker et al., 1995, p. 2). Malaysia, Guinea, Malawi, Colombia, the
Philippines and Nigeria are some less developed countries (LDCs) that have
replicated the Grameen model. In developed countries, the Women’s Self-
3 These include providing free education, awareness development for using proper sanitation, pure drinking water, plantation and birth control measures.
4
employment Project of Chicago, Illinois, the Good Faith Fund in Pine Bluff
Arkansas, in the United States and the Native Self-employment Loan Fund of
Toronto, Ontario in Canada are noteworthy (Wahid, 1993).
1.3 Research Problem and Questions
The founder of the Grameen Bank, Professor Muhammad Yunus, won the Nobel
Peace Prize in 2006 for waging a war against poverty with a revolutionary
microcredit system. He was praised by the Nobel Committee as “a leader who has
managed to translate visions into practical action for the benefit of millions of
people, not only in Bangladesh, but also in many other countries”. This signifies the
importance of a critical examination of the microcredit program.
Although there have been quite a number of studies on various aspects of
microcredit4, none of those studies has comprehensively analysed whether programs
are successful in bringing about an improvement in the quality of life of the
borrowers. On the other hand, the impact assessment studies often show
contradictory results. This study, therefore, analyses the impact of microcredit on
various household outcomes. It examines the consumption patterns of borrowers
compared to non-borrowers and assesses the impact of microcredit on women’s
empowerment.
4 Stiglitz (1993) and Varian (1990) in their theoretical paper have used the principal-agent framework to explain group-based lending. Hossain’s study (1988) is one of the authoritative studies on the impact of credit programs. Gonzalez-Vega and Chavez (1993) and Yaron (1992) have examined the subsidy issue in the context of financial viability of credit programs.
5
This study raises a number of critical questions that need to be addressed by
economists.
Question One: Do microcredit programs in Bangladesh improve various household
outcomes such as income and asset accumulation?
Question Two: Do microcredit programs in Bangladesh provide improved
consumption for borrowers compared to that of the non-borrowers?
Question Three: Do microcredit programs in Bangladesh empower women?
In the existing literature there are two general hypotheses about the impact of
microcredit programs.
First Hypothesis: Microcredit programs are successful in improving the standard of
living of the rural people5 and improving their consumption behaviour.
Second Hypothesis: Microcredit programs are successful for empowering women.
This study answers these questions and examines the hypotheses by using
econometric techniques such as the Ordinary Least Square estimation (OLS), Two-
stage Least Square estimation (2SLS), Tobit, Probit and Seemingly Unrelated
Regression Estimates (SURE). It compares the performance of borrowers with a
control group, non-borrowers to find out if borrowers are better off compared to non-
borrowers.
5 Khandker et al. (1998) showed credit programs have a positive impact on income, production and employment particularly in non-farm sector. Mustafa et al. (1996) showed that not all money borrowed is invested but some portions are used for consumption. Hashemi et al. (1996) and Zaman, (1998) showed that credit programs are empowering for women.
6
The existing literature does not provide a comprehensive study, that is, one covering
all the goals that the credit programs claim to achieve. This study incorporates the
major issues such as impact assessment and conducts a thorough analysis of the
impact of credit on consumption and on women’s empowerment. It uses a new
approach to measure consumption and integrates some new explanatory variables in
the general model specification. In particular this study does not look at the impact
of microcredit on poverty reduction in Bangladesh overall.
1.4 Why is the Problem Worthy of Research?
The success stories of microcredit are fading nowadays as the programs are drawing
increasing amount of attention as well as inevitable criticism. The evidence from the
literature shows that the impact of microcredit on poverty in Bangladesh is
contradictory. There is research to suggest that the access to credit has the potential
to reduce poverty significantly (Hossain, 1988; Khandker, 1998a; Wahid, 1994); on
the other hand, there are studies, that argue that microcredit has a minimal impact on
poverty reduction (Morduch, 1999 and 2000; Weiss and Montgomery, 2005). In
spite of their recent fame, microcredit programs are facing criticism, on the grounds
that they charge very high rates of interest and the poor have not benefited from the
programs. As micro finance transactions are very small in volume, they are unlikely
to have a sustained aggregate impact on poverty reduction. Also, there is concern
that funding for microcredit programs could have been used for other much needed
programs such as health, water, and education. Credit programs may enable poor
people to improve their conditions, but they do not eliminate the need for other basic
social and infrastructure services. Some studies even claim that the benefits from
microcredit are so small that borrowers often borrow from other informal sources to
7
repay loans (Matin, 1999). The very purpose of the micro-finance movement as a
lending instrument for the poor is thus questioned.
In spite of rigorous involvement of world development organisations to reduce
poverty, world poverty has been increasing over the last decade6. This implies that in
many countries the existing policies may not be effective in reducing poverty.
Professor Yunus’ Nobel Peace award reinforces the fact that the world’s
development planners are considering microcredit as one of the key anti-poverty
tools for poorest people. World development organisations such as the World Bank
are planning to replicate the model widely to eradicate poverty from the world.
Since, on the one hand, the impact of microcredit programs on poverty eradication is
still a controversial issue and on the other, world development organisations are
considering microcredit as a key solution to eradicate world’s poverty, it is
extremely important to conduct a thorough analysis of the impact of microcredit
programs in a LDC such as Bangladesh. It is also important to find an answer to the
controversy as well as to find out the extent of the poverty reduction effect of
microcredit programs.
Even though there is a broad consensus about credit objectives, program outreach
and high repayment rates, it is not clear how far credit programs benefit the poor and
whether the poor are utilising the loans in a proper way. It is necessary to investigate
how far microcredit programs are successful in (1) increasing income and asset
accumulation (alleviating poverty) via increasing the standard of living of the poor,
6 Even though the Global Poverty figures between 1990 to 1999 shows an overall reduction in poverty in the world, but in some part such as Europe and Central Asia, Latin America and the Caribbean, Sub-Saharan Africa and in China poverty has increased. Please refer to Appendix A.
8
(2) increasing the consumption of borrowers, and (3) empowering rural women to
whom they are delivering loans. It is imperative to investigate the programs with
first-hand information collected from the beneficiaries to find out the actual benefit
received by the microcredit clients, the extent of clients’ awareness and the extent of
poverty reduction as a result of the intervention of the microcredit programs.
This study analyses the impact of microcredit in terms of poverty reduction and
women’s empowerment using primary data. A detailed questionnaire7 with a series
of questions has been administered to 571 usable respondents out of which 387 are
borrowers and 184 are non-borrowers. Appropriate econometric methods are used to
analyse the responses. Some existing models that have been used by other
researchers are also used after modifying the model specification and/or the
estimation methods.
1.5 Data and the Econometric Approach
Primary data have advantages over secondary data as there is less possibility of
biased estimation caused by omitted variables. On the other hand, the World Bank
and other international agencies have already used secondary data extensively.
Therefore, we have collected primary data during July-August 2004 for the purpose
of our analysis. A well-organised questionnaire is prepared for data collection. Our
study draws the sample from three major districts in Bangladesh, viz., Gazipur,
Dinajpur and Chokoria. The districts have been chosen on the basis of different agro-
7 Please refer to Appendix C
9
climatic and economic conditions. From each district, five villages8 have been
selected in a cluster. Forty households has been selected using multistage stratified
random sampling from each of these five villages. The samples of the households’
have been randomly selected without replacement from the list of households
available from the program’s local office of each village. The non-borrowers data
has been collected from the same cohort to make a proper comparison. Finally the
total sample size eventually was approximately 600.
These cross-section data are estimated firstly, using the basic methods of Ordinary
Least Squares (OLS), secondly, the instrumental variable method Two Stage Least
Square (2SLS), thirdly, censored Tobit estimation, binary Probit estimation and
Seemingly Unrelated Regression Estimates (SURE) to estimate multiple equation
models.
1.6 Thesis Outline
This study provides a comprehensive body of theoretical and empirical work on
microcredit. It outlines theoretical considerations of the monopolistic nature of the
rural credit market on the basis of which microcredit programs have been developed.
The study presents an extensive analysis of the theory of rural credit as well as an
empirical examination. It is organised in six chapters which are as follows:
8 The map of Bangladesh showing the areas from which data has been collected and the list of the villages are provided in Appendix D and E respectively.
10
Chapter Two outlines the relevant theories on the rural credit market. It then
discusses the inception of microcredit institutions in Bangladesh, provides an overall
picture of the Bangladesh economy before and after the introduction of the
microcredit movement. Finally the chapter provides review of a wide range of
literature on microcredit.
Chapter Three provides an empirical analysis of the impact of microcredit on
various household aspects such as income and asset accumulation. This chapter
closely follows the model of Pitt and Khandker (1996) but extends the analysis by
introducing new variables. It examines the impact of microcredit on various
household outcomes using Two-Stage Least Squares estimation. It further analyses
the impact on household outcomes for different income level borrowers. Finally,
after pooling the data for both borrowers and non-borrowers, Tobit estimation is
undertaken for the censored data.
Chapter Four presents an empirical analysis of the consumption patterns of
borrowers compared to non-borrowers using the AIDS model. Ordinary Least
Squares (OLS) estimation as well as Seemingly Unrelated Regression Estimate
(SURE) is used for the model and results are compared. We allow for differences
between borrowers and non-borrowers in their consumption patterns in this chapter.
Chapter Five provides an empirical analysis of women’s empowerment. Firstly, the
chapter considers alternative definitions of empowerment that has appeared in the
literature. Secondly, the chapter defines empowerment from a new perspective and
develops an empowerment index based on this definition. Thirdly, it finds out the
11
difference between borrowers and non-borrowers in terms of various empowerment
correlates. Fourthly, it studies the factors that affect women’s empowerment. Finally,
the chapter shows whether credit programs are effective in empowering women.
Finally, Chapter Six summarises the results of the study together with policy
implications and provides some suggestions for further research.
12
Chapter Two
SURVEY OF LITERATURE: RURAL CREDIT MARKET AND
THE BANGLADESH ECONOMY
This chapter is organised in three parts. Part A outlines relevant theories on rural
credit market. Part B discusses the historical perspective leading to the inception of
microcredit institutions in Bangladesh. It also examines the scope of the activities
undertaken by these credit institutions. Further, it makes an informal attempt to
understand the ways in which microcredit programs may have contributed towards
the development of the Bangladesh economy. The chapter further examines the
broad trends in socio-economic developments that have taken place over recent
decades in Bangladesh and endeavours to relate these to the activities of the
microcredit institutions. Finally, Part C provides a review of literature on
microcredit.
Part A
2.1 Relevant Theories on the Rural Credit Market
Prior to the introduction of microcredit institutions, the debate in development
economics literature centred on the availability of rural financing in developing
countries (Reserve Bank of India, 1977). Several previous studies on rural financing
show the existence of large interest rate differentials between rural and urban regions
in developing countries (Wharton, 1962; Bailey, 1964; Rahman, 1979). However,
these studies suggest that such interest rate differentials may be of existence for a
temporary period. Basu (1997) suggests that arbitrage across regions on such
13
financing may take place because it is possible to borrow from low interest regions
and then lend the sum to high interest regions (p. 268). Basu further states that
according to standard comparative static analysis this process will continue until
market forces eliminate interest rate differentials between urban and rural regions.
Saleem (1987) argues that the existence of high interest rate differentials can only be
explained by the segmentation of regional funds through risk or monopolistic
barriers.
As well, the existence of such relatively high interest rates in rural regions may be
caused by an inherent and known higher risk default factor to recover the principal
from the borrowers (Tun Wai, 1958; Raj, 1979). While this is an attractive
proposition, research evidence suggests that it may only be a partial explanation of
rural-urban interest rate differentials (Bottomley, 1975). Another explanation for
such differentials may be caused by high transaction costs in channelling finance
from urban to rural areas. These high transaction costs ultimately act as a barrier for
formal commercial banking institutions entering the rural funds market. One of main
foci of rural financing is that the size of rural loans is generally smaller than urban
loans. Rural loans are more disparate requiring relatively higher administrative and
overhead costs when undertaken by established commercial banks (Lipton, 1976). In
contrast, local lenders have a cost advantage with local knowledge leading to lower
information search costs in providing loans. Bottomley (1964) argues that the
administrative costs of collecting a large number of local small loans are less for
local lenders specialising in specific rural regions. Thus, cost differentials based on
this asymmetric information leads to high entry barriers and local monopolies in
rural fund markets.
14
It should be noted that in the past (that is, prior to the introduction of microcredit
institutions in Bangladesh) small informal lenders (mostly local merchants, large
landowners and moneylenders) dominated rural funds markets in many developing
countries (Bhaduri, 1977). In general, these informal lenders find it profitable to
invest their surplus capital (funds) in local small loans that have higher returns,
instead of depositing the surplus with formal commercial banks. According to
standard market theory (as has been stated earlier), such high entry barriers leading
to local monopolies will result in an under-allocation of funds to rural regions.
2.1.1 The Rural Credit Market
Prior to the introduction of microcredit in Bangladesh, credit was not widely
available in the rural economy. Informal lenders (as in most developing countries),
were the sole providers of rural credit (Adams and Fitchett, 1992; Ghate, 1992).
These informal lenders (who are commonly known as “moneylenders”) had an
information advantage about their clients, which helped them to monitor their
activities in an easier and cheaper way. These moneylenders used to charge high
interest rates. For example, Hossain (1988 p. 23) argues that in the case of
Bangladesh interest rate charged by these moneylenders were about 10% per month.
These moneylenders basically provide loans only to the people who worked for them
or over whom they have some sort of control. The nature of operation of these
moneylenders may be described in terms of a monopolist or duopolistic market. The
following section provides the relevant theoretical explanation of the monopolist
market.
15
2.1.2 The Monopolistic Nature of the Rural Credit Market
It is a common practice in rural areas that moneylenders lend to people over whom
they have some control. For example, a merchant lends only to those who buy/sell
regularly from their shop or business. The rural credit market is defined as
“fragmented”9 (Basu, 1997). This credit market is further characterised by a whole
array of interest rates differentials. The theoretical relevance of this fragmented rural
economy is demonstrated here in terms of a monopolistic market.
The personalised relationship between the borrower and the lender determines most
often the terms and conditions of rural lending. In general, it is observed that each
lender lends to the person in whom they have greater personal confidence. That is,
according to Bottomley (1964), the monopoly power of the rural moneylender is
dependent on the intimate knowledge of the borrower’s circumstances.
The monopolistic rural credit market as analysed by Basu (1997) is discussed below.
Following Basu’s (1997) model the demand function is derived, assuming the
borrower can get loans only from one moneylender and the borrower is a price taker,
as follows:
)1.2(0)(),( ' <= iLiLL
where L is the amount of credit and i the interest rate.
The inverse function of Equation (2.1), according to Basu, is just another way of
looking at the same relation, as follows:
9 A fragmented rural credit market means that each moneylender with a potential client becomes a small “Credit Island” and the credit market effectively becomes broken up into small fragments (Basu 1997).
16
)2.2(0)(),( ' <= LiLii
This relationship is shown in Figure 2.1, where the line AD shows the demand
function and AM represents the marginal revenue curve. The personalised relation
ensuring there is no default, the moneylender chooses L amount of loan with i
interest to maximise his/her interest earnings.
Now assuming that the lender has the option of investing money elsewhere and
earning an interest r on the lending, the lender’s objective function, according to
Basu, is derived, as follows:
LrLLiL
−)(max}{
The first-order condition of this problem is shown as follows:
)3.2()()( ' LLiLir +=
The left-hand side of this equation represents the marginal cost of giving loans and
the right-hand side is the marginal revenue. Hence, Equation 2.3 depicts the standard
equilibrium in a monopolistic market where MC=MR. This relationship is shown in
Figure 2.1, as having equilibrium level of amount of loan and interest rate which is
indicated by *L and *i respectively.
17
A
i
*i
r C E
O *L M F D L
B
Figure 2.1 Monopoly Market
According to Basu, “the rural financial market is fragmented into little monopolistic
islands … the rural economy itself will be characterised by a whole array of interest
rates” (p. 270). Basu also states that this model is not without limitations. Firstly,
Basu argues, “The analysis presumes that the borrower always has enough money to
repay a loan”. In reality, given the widespread poverty of rural borrowers, this is
unrealistic (p. 271). The second criticism, according to Basu, is that “monopoly
analysis is general” (p. 271).
In Figure 2.1, it is shown that the urban interest rate r is lower than i* (rural interest
rate) when the moneylender has access to the urban market. Here it is presumed that
the borrower always has enough money to repay the loan. In reality, however,
18
according to Basu, the borrowers often face a liquidity crisis and, therefore,
sometimes use alternative mode of repayments including personal labour; mortgage
land and personal belongings.
According to Basu (1997, p. 271), the demand function of the borrower (Equation
2.2) is treated as “primitive” It is derived assuming that the borrower takes loans
only in order to invest. Let R denote the total earnings from investment and L denote
the amount of loan; the function is derived, as follows (p. 272):
0)0(,0)(,0)(),( "' =<>= RLRLRLRR
Basu (1997) advocates that in the case of a producer who employs workers up to the
point where wage equals marginal product of labour, given i, the borrower chooses L
so that 1+i equals marginal earnings from L. Thus, according to Basu,
“The marginal earnings curve represents the demand curve for loans and the area
under this curve is equal to the total earnings of the borrower” (p. 272). Therefore,
the demand curve for loans put forward by Basu is as follows:
)(1 ' LRi =+
This is the equation of the curve AD in Figure 2.1.
If the moneylender-monopolist considers lending a certain amount *OL , according to
Basu the maximum amount the borrower is willing to pay for this is *OABL . Basu
shows that, according to Figure 2.1, the lender has the option of earning an interest
of r elsewhere; the lender’s net income will be ABCr.
19
If the lender adopts this strategy of extracting the maximum the borrower is willing
to pay, according to Basu the lender does best by giving loans equal to OF, for a net
income of AEr. Since AEr> *BCri Basu argues that, “by this strategy the monopolist
earns a larger profit than he would by behaving like a textbook monopolist” (p. 272).
This finding of Basu (1997) is consistent with that of Bardhan (1976 and 977) where
he uses a model that shows the equilibrium percentage of revenue is higher, other
things remaining the same for the monopolistic landlord.
2.1.3 Interlinked Factor Markets
The rural credit market is not only dominated by moneylenders who are
monopolistic in nature as discussed above. The landlord-tenant relationship is also
an important characteristic of the rural market. The relationship described in the
literature is master-serf in nature. A widely noted theoretical paper on the interlinked
factor markets in an agrarian economy is by Bhaduri (1973) where he shows that the
landlord who is a provider of consumption loans to a tenant may have no incentive to
adopt yield increasing innovations, if the landlord’s interest income from their loans
to the tenant goes down (because the tenant will borrow less as they shares the
increase in yield) sufficiently to offset his share of the increased yield. Bhaduri’s
(1973) four prominent features of the agricultural economy make it easy to
understand the theory on rural credit market. Bhaduri portrays the agricultural
market as “semi-feudal” because the existing relations of production have more in
common with classical feudalism of the master-serf type than with industrial
capitalism. The features are as follows:
20
(1) Sharecropping: Sharecropping is a common form of land tenure used in the
agricultural sector. According to Bhaduri (1973) “The landlord leases out his land
for at least one full production cycle and the net harvest10 is shared between tenant
and the landlord on some legally stipulated basis” (p. 121). He further mentioned
that this tenancy system may have enormous complications as the tenant may have
some land of his own or working on fixed capital or the entire amount is supplied by
the landlord.
(2) Perpetual Indebtedness: “The tenant is always heavily indebted. A substantial
portion of the tenant’s legal share of the harvest is taken away immediately after the
harvest as repayment of past debt with interest, thus reducing his actual availability
balance of the harvest well below his legal share of the harvest” (Bhaduri 1973, p.
122). According to Bhaduri (1973) this does not leave enough food for the tenant to
survive from this harvest to the next harvest. The only way of survival is to borrow
for consumption. This actually perpetuates the indebtedness of the tenant.
(3) Landowner as the Lender of Consumption-loans: “This perpetual indebtedness of
the tenant is combined with another important factor which lends the whole system
the definite character of semi-feudalism the lender of the consumption-loan is also
typically the tenant’s landlord” (Bhaduri, 1973 p. 122). Thus the tenant leases their
land from the same man to whom they are perpetually indebted and this reduces
them virtually to the state of a traditional serf. The tenant is more or less tied to their
landlord. According to Bhaduri (1973), this semi-feudal landlord exploits the tenant
both through their traditional property rights in land and through usury and both
10 Net harvest as defined by Bhaduri (1973) is the gross harvest minus seed required for the next harvest.
21
these modes of exploitation are important features of the agricultural market in
LDCs.
(4) Inaccessibility to the Market: “The semi-feudal economic relationship between
the tenant and his landlord works with full severity when the rate of interest on
consumption-loans is extraordinarily high. The tenant is usually not credit-worthy in
any commercial banking sense because he has no asset to borrow against. His only
lender is usually his landlord, who lends against the future harvest and the tenant has
to borrow on terms which the latter dictates” (Bhaduri, 1973, p. 122).
These four types of landlord-tenant relationship appeared in the theory of
agricultural economy in different ways. Bardhan (1980) shows that a monopolistic
landlord does land rationing at a fixed rental share. Bardhan (1980) argues that “the
tenant is dependent on the landlord for credit to finance a given subsistence
consumption in the lean season11 and he pays back the loan along with interest at the
end of the harvest” (p. 89). According to Bardhan (1980), “since the interest rates are
high, it matters a great deal as to how long one has to wait until income comes. The
tenant knows that as a wage labourer he can get some wage income immediately in
lean season, whereas as a tenant he has to wait until the crop is harvested in peak
season” (p. 89).
In the model developed by Bardhan (1976 and 1977) the landlord has three sources
of income, rental income from leased-out land, income from self-operated land net of
wages paid in lean and peak seasons, and interest income. They maximise their total
11 In agricultural production there are two seasons lean and peak. Lean is the preparation season and peak is the harvest season.
22
income with respect to their two decision variables, the amount of land to lease out
and the amount of labour to use on their self-operated farm. According to Bardhan
1980, “the tenant considers his income from cultivating the leased-in area and from
wage income net of interest payment on lean-season consumption credit and
maximises it with respect to his only decision variable, labour-intensity per acre on
his leased-in area.” Bardhan (1980) shows in his comparative static parametric
variations model that the equilibrium percentage of area under tenancy will be higher
for the monopolistic landlord.
2.1.4 Potential Risk
The Interlinked rural credit market characterised by Basu (1997) is when the
landlord agrees to take on a tenant at a fixed rent and also agrees to provide the
tenant with credit at a certain interest rate. According to Basu (1997) there is some
risk associated with a loan on part of the landlord. There is always a probability that
certain portion of loans will not be repaid. According to Basu (1997), “if a debtor is
the lender’s tenant or has some historical ties with the lender, it is very unlikely that
the debtor will get away with defaulting. If, on the other hand, a debtor has no
dealings, nor any prior ties, with the lender, it is likely that the debtor will not repay,
there being hardly any legal machinery in backward regions to enforce repayment”.
It is not unrealistic that for every moneylender among all potential borrowers there is
a set of people from whom the lender can recover money and there is another set
over whom the landlord has no control. Therefore, according to Basu (1997) if a
landlord does not discriminate between potential borrowers, he will attract all
“lemons” and eventually become bankrupt. The landlord in general provides loans to
23
those over whom they have some control and where there is no ‘potential risk’ of
default.
2.1.5 Significance of Credit Policy in the Rural Areas
The monopolistic nature of the rural credit market has become a concern for policy
makers in understanding the nature and depth of the market. In the case of LDCs,
credit policy is a great worry for policy makers because the moneylenders are the
only credit providers and they charge exorbitant interest rates.
For over a century official policy has been directed at replacing traditional
moneylenders with formal institutions such as cooperatives, rural development bank,
and commercial bank branches. However, these institutions may not cater for the
needs of rural borrowers due to the lack of securities, which is a pre-requisite for
formal borrowing. The government of LDC and the relevant global institutions such
as the World Bank and IMF are always concerned to improve the condition of poor
people in the rural areas. Since there is always an unsatiated demand for rural
finance in rural areas, the governments of LDCs take various initiatives. The major
concern for rural people is how to survive. If they (rural people) could obtain
affordable long-term credit for investment as well as for consumption, then they
could improve their living conditions.
Since 1976 in Bangladesh the microcredit institutions such as the Grameen Bank
have made a remarkable achievement in rural lending. They have set their objective
to reduce poverty by mobilising resources and targeting credit and non-lending
services to poor rural households for the purpose of setting up new enterprises
24
(Yunus, 1984; Hossain, 1984; Getubig, 1992; and Khandker, 1996). Much has been
written about this new class of “banks for the poor”, and there have been enthusiastic
replications in a number of countries including the United States (Ravallion and
Wodon, 2000). Several nations are now trying to emulate the Grameen Bank model.
2.1.6 Conclusion of Part A
Earlier theoretical discussion shows that the landlords can prevent a tenant from
renting other land or from working for wages. Landlords can include these
restrictions in the contract or they can enforce them. The landlord also has an extra
control instrument by means of contract terms that limit the tenant to their wage
rate12 (reservation utility) (Binswanger and Rosenzweig, 1984, p 17).
Some interesting observations may be derived about the microcredit programs of
Bangladesh based on the above mentioned theoretical discussion of interlinked rural
factor markets. The microcredit movement that has been initiated in Bangladesh is a
breakthrough in the history of rural financing in developing countries. Its operating
style is in many ways different from the traditional moneylenders, who lend to rural
borrowers, as well as landlords, who make tenants work for them. It has already been
mentioned earlier that the traditional moneylenders are monopolistic in nature. Thus,
microcredit programs do not reveal such characteristics. Unlike traditional
moneylenders that make their tenants work for them, the borrowers of microcredit
are allowed to choose the activities in which they are willing to invest. The credit
delivery mechanism is very different from traditional lending systems. This does not
12 Bardhan and Srinivasan (1971) developed a similar model where the landlord is not allowed to control tenancy size.
25
allow a master-serf relationship as portrayed by Bhaduri between the loan providers
(microcredit institutions) and their borrowers. It is exciting to see that the programs
are now trying to change the traditional rural lending system and trying to make rural
financing available and affordable for everyone. One very interesting observation
about the microcredit programs are that they do not distinguish between clients’
“potential risk”, rather they provide loans to everyone. It has removed the
moneylenders to some extent and is trying to save the rural economy from very high
interest rates. The popularity of the programs is demonstrated by their outreach and
high recovery rate. The features that attracted most attention are their use of “peer
monitoring” which has substituted the need for collateral. Although lending is
provided to individuals, borrowers are formed into groups of four or five. According
to the Grameen Bank model, each of these groups is held jointly accountable for the
behaviour of its members. If one member is in default to repay the credit then the
likelihood of obtaining credit by the other members becomes remote. That is, one
member’s non-repayment can cause all other members to be denied future credit.
This collective microcredit model creates peer pressure and also encourages the
development of mutual help within groups.
Part B
2.2 The Bangladesh Economy: Pre-Microcredit
Bangladesh became an independent nation in 1971 after a brutal civil war with
Pakistan that lasted about nine months. The government of Bangladesh has faced the
difficult task of reconstructing and rehabilitating the fragile economy since then.
Bangladesh’s economy was forced through a very vulnerable period followed by
26
several natural calamities including the devastating flood of 1974. The War and
natural calamities combined took a heavy toll on mainly the agriculture-based
economy of Bangladesh. Consequently, the production of rice, which is the staple
food, faced a major setback (Hossain, 1988). Soon after independence the economy
had to rely quite heavily on imported food. For example, Bangladesh’s imported
food was worth13 Tk14 500, Tk 870 and Tk 1,310 million at constant prices in the
years 1973-74, 1974-75 and 1975-76 respectively.
In the face of a growing population, the agricultural sector of the economy could not
expand sufficiently. This sector also failed to create more jobs and absorb additional
people joining the labour force every year due to the high population growth (Wahid,
1993). In contrast, the manufacturing sector had an even a bleaker image with poor
structure as well as small size during the 1970s. The major manufacturing industries
were nationalised by the government soon after liberation but they proved to be
liabilities rather than assets due to widespread mismanagement and corruption.
Overall, the economy was characterised by low economic growth, high population
growth and inequality in distribution of resources during 1970s. Khandker (1998a)
identified landlessness15 in rural areas as a major contributor to poverty in the
country at that time. Along with landlessness, lack of human and physical capital
among the poorer population in the rural areas prevented the economy from
productive investment. Coleman (1999) rightly argues that the poor often find
13 Bangladesh Bank Economic Trends, vol. XV (5), (1990), p. 18. 14 Tk. stands for taka, the currency of Bangladesh. 15 In Bangladesh, households with less than 0.5 acre are considered functionally landless, since the amount of land they own cannot be a significant source of income (Hossain, 1988, p.15).
27
themselves in a vicious circle: they are producing at a subsistence level, which in
turn makes it harder for them either to invest in productive resources or to gain
access to formal credit market.
2.2.1 Inception of Microcredit Institutions
In the post-independence period, state owned commercial banks and specialised
agricultural banks such as the Bangladesh Krishi Bank (BKB) could not cater to the
needs of the rural people in the country for several reasons. Firstly, these banks
required collateral for providing loans, which rural poor people found difficult to
arrange. Secondly, the loan sanctioning procedures and other related formalities were
too cumbersome for mainly less educated rural people. Thirdly, these financial
institutions preferred handling large loans rather than small loans. Finally, another
problem was the high transaction costs for the banks that include cost of obtaining
information about the rural borrowers to access their creditworthiness, monitoring
the use of loans and ensuring repayments.
For the reasons outlined above the commercial and other formal banks could not
operate successfully in the rural financial market of Bangladesh. Due to non
availability of credit from the formal sources the rural people had to rely heavily on
informal money lenders who used to charge exorbitant interest rates. All these
factors taken together effectively opened up an opportunity for a new type of
financial institution to come into existence which is called the “microcredit
28
institution”16 (most recently these microcredit institutions are also often called
microfinance institutions).
Nobel laureate Professor Muhammad Yunus, the founder of the largest microcredit
institution in Bangladesh – Grameen Bank was convinced that the poor people were
more capable of taking care of themselves than what was and currently is portrayed
in traditional literature (Yunus, 1999, p. 69). He observed that poor villagers with
virtually no assets worked hard at a variety of crafts and service jobs and in the
absence of any major shocks could somehow manage to survive (Yunus 1999, p.
71). Unfortunately, they could not accumulate any capital because firstly, their
income was quite low and secondly, out of this income, after deducting for
sustenance, whatever insignificant surplus they gathered was going to the local
moneylenders from whom they had to borrow (a small amount of ) capital. Thus, if
the poor villagers were to succeed, according to Yunus, (1999) “credit” was the key.
2.2.2 Background of the Grameen Bank
The Grameen Bank of Bangladesh is one of the largest microcredit institutions in
Bangladesh. In 1976, Professor Yunus tried an experimental research project in a
village near Chittagong University. In his experimental project, he started
lending the equivalent of $26 to a group of 42 workers. Borrowers invested this
16 Microcredit is defined as a credit provided to “poor” free of collateral through institutionalised
mechanism (the only collateral is the “peer” collateral). This credit is made available as and when needed, at the doorstep of the client (Bajwa, 2001). On the other hand, microfinance is inclusive of savings and other services. As a matter of fact, this concept has been used and applied in variety of contexts and used interchangeably with microfinancing - such examples include Badan Kredit Kecamatan (BKK) Indonesia, Agriculture Cooperation, and Banks Rakyat Indo Unit Desa.
29
small amount in craft making and repaid the loan on time. From this small
experiment, he realised lending to the poor had a future.
In March 1978, the project developed an appropriate mechanism for delivering
credit to the poor. In June 1979, the Grameen Bank project was launched in a
wider area with assistance from the Bangladesh Bank (Central Bank of
Bangladesh). Within a year 24 branches were set up. It ensured 98 % recovery of
loans by the due date. With assistance from ‘International Funds for Agricultural
Development (IFAD)’ the project expanded its activities to other districts besides
Dhaka. In 1982-83, 50 branches were set up in five districts. A government
ordinance in September 1983 transformed the project into a “bank”, a specialised
financial institution for the rural poor.
2.2.3 Growth and Expansion of the Grameen Bank
The Grameen Bank has reversed conventional banking practice by removing the
need for collateral and created a banking system based on mutual trust,
accountability, participation and creativity (a group-base lending strategy). The bank
focused its market mainly on the rural areas where population density is very high.
Prior to the introduction of microcredit institutions, a large number of rural people
could not get access to formal credit. Microcredit institutions have taken the risk of
lending to mainly rural people who had been neglected in the past. In the past these
borrowers were classified as high risk borrowers. Most interestingly, through the
introduction of microcredit these so-called “high-risk” borrowers were allowed to
30
borrow collateral-free loans17 and the recovery rate of these loans is considered to be
very high, that is about 95%18 (Yunus 1999, p. 114).
The Grameen Bank is not only the largest rural finance institution in Bangladesh; it
is also the largest moneylending institution in the world in terms of number of
clients. In 2005, the bank’s borrower count rose from 4 to 5 million. 94% of
borrowers are women. With 2,345 branches, the Grameen Bank provides services in
75,359 villages, covering 90% of the total number of villages in Bangladesh. At
present the bank disburses general, seasonal, housing and other types of loan under
the categories of Processing and Manufacturing, Agriculture and Forestry, Livestock
and Fisheries, Service, Trading, and Peddling and Shopkeeping. Since its inception,
the Grameen Bank has sanctioned 290.03 billion taka19 (The Grameen Bank, Annual
Report, 2006).
Figures 2.2 and Figure 2.3 shows the expansion of Grameen Bank membership and
branches over 12 years. It is observed from the figure that the membership has
increased by approximately 150% and the number of branches have grown by more
than 50% in 12 years.
17 Although the Grameen Bank does not require any tangible collateral in giving loans, but the “group liability” may be interpreted as intangible collateral in the sense that, in case of default, a member loses its opportunity for future loans. 18 In 1999, the recovery rate was 95 %which has gone up to 98 % in 2006. 19 One taka is $US 0.017 (approximately).
31
Figure 2.2 Number of Branches
Number of Branches
0
500
1000
1500
2000
2500
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Years
Branches
Source: The Grameen Bank, Annual Report (2006).
Figure 2.3 Number of Borrowers
Number of Borrowers
0
100000
200000
300000
400000
500000
600000
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Years
Borrowers
Source: The Grameen Bank, Annual Report (2006).
32
In addition to its financial services, the Grameen Bank implements diverse programs
to promote social development. It encourages members to open nursery schools at its
centres, distribute seeds and seedlings to encourage gardening and plantation, and it
helps members and their families to improve their health and nutrition. The bank has
opened 14,804 schools by 2003. The total number of children enrolled in these
schools grew from 71,467 in 1985 to 396,289 by 1994 which is a remarkable
increase of 454.5% from 1985 to 1994. Children’s enrolment at schools is still
increasing at a reasonable rate with the expansion of the banks branches (The
Grameen Bank Annual Report, 2003).
2.2.4 Savings Mobilisation by the Grameen Bank
The Grameen Bank mobilises savings by requiring members to make deposits of
different types. The objectives in doing so are to overcome market imperfections and
promote financial security for members/borrowers. These savings are alternative
sources of credit for borrowers and help to gradually minimise the bank’s
dependence on outside borrowing.
The Grameen Bank’s members are required to contribute one taka weekly to the
group fund as savings. These contributions are held as individual savings that are
refundable when members drop out. Individual members are also required to
contribute 5% of the principal amount borrowed to the group fund. Unlike their
individual savings, group members cannot reclaim this group fund contribution, but
they are allowed to borrow from this fund with the approval of other members.
33
There are two other types of loan, from the “emergency fund” and “welfare fund”.
The emergency fund offers protection against debt or liability when a member dies,
or when theft, loss and damage to property (including livestock and crops) of
members/borrowers occurs. The children’s welfare fund was designed to provide
education for members’ children in schools managed and run by the Grameen Bank
members, and to support children’s involvement in small-scale income earning-
projects.
Although the Grameen Bank initiated microcredit programs, subsequently several
other institutions have started providing microcredit in Bangladesh. At this stage, the
major microcredit institutions in Bangladesh include the Grameen Bank, Bangladesh
Rural Advancement Committee (BRAC), which is the biggest non-government
organisation (NGO),20 Rural Development Project-12 (RD-12), Bangladesh Rural
Development Board (BRDB), Association of Social Advancement (ASA) and
Proshika Manobik Unnayan Kendra (PROSHIKA).
Since we have confined our analysis in the study to the Grameen Bank and BRAC,
the following section discusses BRAC.
2.2.5 Background of BRAC
The Bangladesh Rural Advancement Committee (BRAC), the largest non-
governmental organisation in Bangladesh, was formed in 1972 as a relief
20 Non-government organisations provide help to those who fail to accumulate sufficient resources to survive. In its most ambitious form, an NGO is a private, not for profit institution dedicated to influencing the working structure of government and ensuring the greater welfare of all citizens (Mustafa et al. 2000)
34
organisation for the post-war period. The founder of BRAC, Mr Fazle Hasan Abed
realised that pervasive poverty could not be addressed with short-term relief
measures. Thus in 1973, BRAC shifted its focus from relief to long-term community
development. It started its operation with the objective of improving the economic
and social status of the rural poor. Since then BRAC has operated as an NGO, but its
journey has not always been very smooth. It faced many challenges from different
corners of society including the rural communities and the village elites who
controlled many of the social and economic opportunities of the poor, but these
could not stop the activities of BRAC.
BRAC realised that rural people lack the skills and opportunities to derive benefit
from their own labour. The organisation presently delivers both social services and
credit to its members through its two major programs, the rural development
program (RDP) and the rural credit program (RCP). The RDP was started in 1986,
the RCP in 1990. Both programs have evolved over time through the process of
learning by doing with the objective of poverty reduction and empowering women.
Women in the rural areas of Bangladesh play a vital and unacknowledged role in
production both within and outside the household. Women are often the sole
providers for their households due to the death of a spouse, divorce, desertion or the
migration of men looking for paid labour. Recognising these factors and the low
status of women in Bangladeshi society, BRAC targeted women in its programs and
made women’s empowerment a pillar of its programming. The RDP addresses the
socio-economic development of underprivileged rural women through access to
35
credit, capacity development, savings mobilisation, institution building and
awareness raising.
2.2.6 Growth and Development of BRAC
BRAC’s social development programs are aimed partly at alleviating poverty and
partly at supporting government organisations. BRAC also emphasises the role of
women in development by mobilising more women than men. It has helped improve
women’s social status by increasing their economic role in the family. BRAC
intervenes in education, health care, legal activities, vulnerable group development,
and skill development. Its non-formal primary education program and its oral
dehydration therapy are widely known.
BRAC has developed two primary school models the non-formal primary education
program and the primary education program for older children. In addition, BRAC
assists the government in expanding primary education throughout the country.
BRAC’s primary education models aim to reduce mass illiteracy, ensure women’s
education and involve communities in their own socio-economic development.
The non-primary education program, begun in 1985, is a three-year program for 8 to
10 year-old boys and girls. These children, who have never attended school, are
enrolled in the fifth grade of the formal primary school system after graduating from
this program. The primary education for older children program, begun in 1988 is a
three-year program for 11 to 16 year-olds who have never attended school. Books
and other materials are provided free for the students.
36
Table 2.1 Number of BRAC Schools, Students and Graduates
Bangladesh Rural Advancement Committee (BRAC) (as at June 2006
Number of Primary Schools 3,200
Number of Students 1 million
Number of Graduates 3.49 million
Number of Pre-Primary School 20,168
Number of Students 0.54 million
Source: Bangladesh Rural Advancement Committee, Annual Report (2006).
Table 2.2 Job Creation21
by BRAC
Job Creation by BRAC (as at June 2006)
Poultry 1,708,145
Livestock 570,266
Agriculture 853,390
Forestry 79,062
Fisheries 2,772,330
Sericulture 25,549
Horticulture 184,031
Handicraft Product 15,223
Small Enterprise 136,159
Small Trade 2,635,212
TOTAL 63,53,482
Source: Bangladesh Rural Advancement Committee, Annual Report (2006).
21 Job creation here does not refer to traditional textbook definition. Rather it refers to the number of people who have invested in different sectors.
37
Table 2.1 shows, recent figures for the number of students who have graduated from
BRAC schools since BRAC’s inception. It also shows the number of schools and
pre-primary schools established by BRAC and the number of students enrolled.
Table 2.2 shows job creation by BRAC at the sectoral level since BRAC’s inception
until June 2006.
2.2.7 Savings Mobilisation by BRAC
Savings mobilisation is an integral part of BRAC’s lending process. In addition to
depositing two taka per week as individual savings, borrowers must pay 10% of the
principal loan amount. Out of this, 5% is transformed into individual savings, 1% is
kept for a group insurance fund, and 4% is given to a group fund. The basic objective
of the group insurance scheme is to provide financial support, a maximum of Tk
5,000 to the family of the deceased member so that the family is not displaced and
does not suffer financial hardship. BRAC pays 9% interest on both individual
savings and group funds. This rate is higher than that offered by commercial banks
(BRAC Annual Report, 2004).
Table 2.3 provides some basic statistics in respect of the Grameen Bank and BRAC.
Though we do not directly compare these two institutions in this study, as is evident
from the table both the institutions are quite comparable (similar) in terms of many
indicators.
38
Table 2.3 Comparing BRAC with the Grameen Bank
Items BRAC Grameen Bank
Number of Borrowers 4.51million 6.61million
Number of Branches 1172 2226
Villages Covered 65,000 71,371
Loan Disbursement Tk. 185,580 million
(US$ 3385million)
Tk. 290.03 billion
(US$ 5.72 billion)
Repayment Rate 98.85% 98.85%
Interest Rate Charged 15% 20%
Source: 2006 Annual Reports of the Grameen Bank and BRAC.
39
2.3 The Bangladesh Economy and Recent Developments
Even though the Bangladesh economy went through a very difficult time in the
period following independence, it has recovered substantially in the last two decades
and has shown its potential for becoming an example of success. The UNDP Human
Development Report (2000) referred Bangladesh as “a case that stood out as a
potential instructive story of development”. Philip Browning, the former editor of the
Far Eastern Economic Review, wrote in a very recent article in the International
Herald Tribune on 7th May 2005:
“Bangladesh is a paradox. It lacks good governance and is beset by natural calamities, corruption and self-destructive political infighting. Yet its gross national product persistently maintains a growth rate of 6%, well above average for developing countries. It has overtaken India on several social indicators. Its aid dependence has fallen from 6% to 1.8% of gross domestic product…...Of course this is still a desperately poor, overpopulated country, where 50% of children are underweight. But India is no better on that score. Bangladesh has made much more progress than its neighbours over the past 10 years, becoming self-sufficient in food. Social progress has been even more marked. Educational standards may be poor, but primary school education is even more striking. There are now more girls than boys at secondary level. Gender equality also seems reflected in the fall in fertility rate, which has halved from 6.0 to 3.0 in two decades the steepest fall almost anywhere other than China. It is now below India’s and far below Pakistan’s. The lowest birth rate is linked to a steep fall in child mortality, and to the enhanced economic role of women as small-scale village entrepreneurs and as garment workers. Economic advancement has been underpinned by the individual initiative of hundreds and thousands of Bangladeshis working overseas. Their remittances exceed the net earnings of the garment industry, amounting to more than $3.5 billion a year, mostly from Britain, the United States and Middle East…..”
Despite many obstacles some man-made and some natural Bangladesh experienced
moderately accelerated growth in the 1990s, compared to the previous decades. Its
economic growth rate has increased from 2% in 1975 to 6% in 2005. Bangladesh has
become successful in reducing population growth which has decreased from 3% to
1.48% in last 30 years. The life expectancy of both males and females has increased
40
and the infant mortality rate has fallen dramatically. The country has demonstrated
striking progress in primary and secondary school enrolment for both males and
females. Bangladesh’s poverty has fallen steeply in last three decades. The
accelerated economic and social development of the country is summarised in Table
2.4.
According to the Human Development Index (HDI) developed by UNDP,
Bangladesh has made significant progress. Its HDI increased from 0.345 in 1975 to
0.520 in 2005. The HDI ranking shows that Bangladesh has done better than what
was expected given its GDP level.
Table 2.4 Major Socio-Economic Conditions in Bangladesh 1974-2005
Indicator 1974-75 1984-85 1994-95 2005
Real GDP Growth (%) 2.00 3.72 5.21 6.04 Annual Population Growth Rate 3.00 2.07 1.95 1.48 Infant Mortality Rate 140 112 94 58 (below I year, per '000 live births Life Expectancy at Birth (years) Female 47 56 56.7 59 Male 49 55 56.5 58 Adult Literacy (%) 20 33 37 59.2 Primary School Enrolment (% of school age population) Male 50 63 77 96.5 Female 40 43.8 70 94.5 Child Malnutrition 78 70 68 55 Population below poverty Line 70 55 47 44.7
Sources: Asian Development Bank, Bangladesh Bureau of Statistics,
Ministry of Finance, and United Nations Development Program
41
Sources of Growth
Bangladesh has seen an emergence of progressive entrepreneurs, and good
macroeconomic management has kept inflation in single digits (6% in 2004
according to The World Factbook). Recent reforms in fiscal management,
governance, state-owned enterprises, banking, telecommunications and energy have
also shown encouraging results. Foreign direct investment flows have supported
infrastructure, energy and export-oriented manufacturing.
Bangladesh’s accelerated economic growth has occurred in all broad economic,
agricultural, industry, and services sectors. Table 2.5 shows sectoral GDP growth
rate from the early 1980s to 2000. The agricultural sector accelerated from 2.68% to
3.22% industrial output from 5.70% to 6.96% and service sector GDP from 3.83% to
4.48% over this period.
Table 2.5 Sectoral GDP Growth Rates: 1979/80 – 1999/00
Five-yearly average Decadal average Sector
1980/81-1984/85
1985/86-1988/89
1990/91-1994/95
1995/96-1999/00
1980/81-1999/00
1990/91-1999/00
Agriculture 2.68 2.40 1.55 4.89 2.54 3.22
Crop production 2.69 2.69 -0.43 3.86 2.69 1.72
Fisheries 3.06 1.64 7.86 8.56 2.35 8.21
Others 2.40 2.21 2.53 3.30 2.31 2.92
Industry 5.70 5.80 7.47 6.44 5.75 6.96
Manufacturing 4.69 5.27 8.20 5.59 4.98 6.90
Large and Medium 4.44 5.43 8.41 5.49 4.94 6.95
Small scale 5.41 4.89 7.69 5.87 5.15 6.78
Construction 6.44 5.59 6.27 8.80 6.02 7.54
Others 11.50 10.68 6.43 4.90 11.09 5.67
Services 3.83 3.58 4.14 4.81 3.71 4.48
TOTAL GDP 3.72 3.74 4.15 5.23 3.73 4.69
(Annual average; in constant 1995/96 producer prices)
Source: Osmani et al. (2003)
42
2. 3.1 Conclusion of Part B
We can make an important observation from our discussion in this chapter. The
socio-economic indicators that have shown notable achievements of Bangladesh
over the past two decades are all related, directly or indirectly, to some areas in
which microcredit institutions have had an operational involvement. We have seen
that the two major micro finance institutions chosen for this study are involved not
only in delivering credit but also in providing some important social services such as
mass literacy program, population control, proper sanitation etc. Their credit delivery
service is directly associated with entrepreneurial skills development and job
creation which has direct implications for economic growth. In addition, the social
services provided by these institutions are expected to underpin the overall
development process indirectly. Therefore, it would not be an exaggeration to call
these institutions “development partners” of the government of Bangladesh. With
this backdrop, the relevance of an objective analysis of the contributions of
microcredit programs in the development process of Bangladesh cannot be ignored.
Despite considerable improvement in Bangladesh’s socio-economic scenario over
the past decades, an all-out effort of both government and non-government
organisations is still extremely important to help the country meet several daunting
challenges lying ahead. For example, Bangladesh’s poverty rate remains very high,
with nearly half of its 146 million people living below the poverty line. Bangladesh
still has the highest incidence of poverty in South Asia and the third highest number
of poor people living in a single country after India and China (Bangladesh Country
Brief, World Bank, September 2004). The challenges are magnified by a population
density of more than 800 people per square kilometre one of the highest in the world.
43
These challenges further stress the importance of this study. By providing a thorough
analysis of the performance of these institutions in relation to their major objectives,
this study will uncover the areas in which these institutions should pay careful
attention so that they play a meaningful role in tackling future challenges as active
development partners of the government.
Part C
2.4 Introduction to Review of Literature
The literature on poverty research in Bangladesh is wide-ranging. A significant body
of research on various aspects of poverty in Bangladesh and erstwhile East Pakistan
can be found in economics and development studies literature. Before the emergence
of an independent Bangladesh in 1971, poverty literature about this region appears to
have mainly focused on inequality in income, consumption and wealth between the
then East and West Pakistan. Emphasis was generally placed on economic disparity
between the two regions while such disparity used to be generally cited as the
principal reason causing inequalities in income and wealth of the two regions.
The post-independence literature on poverty in Bangladesh can be generally
characterised as positive and descriptive studies (Rahman and Hossain, 1996;
Osmani, 1978). Attempts have been made to measure and analyse poverty mostly
from a positivist perspective. Within the broad definition of poverty, research has
been undertaken to examine cross-section and time series patterns of poverty
(Rahman and Islam, 1987, 1988). A major part of poverty research is occupied with
the agricultural sector mainly focusing rural poverty (Quesem, 1991; Nahar, 2004;
Osmani, 1989). Urban poverty has drawn the attention of poverty researchers mainly
44
since the 80s when the number of slums and slum dwellers in Dhaka city (the capital
city of Bangladesh) started to rise almost exponentially. Until such time, poverty
used to be generally seen as a rural phenomenon. However, research on urban
poverty ushered a new era of poverty research by revealing the nature, extent,
dimensions and possible consequences of urban poverty. Most of these researches
find a link between rural-urban migration and rise in urban poverty (Rahman, 1978).
While rural poverty, caused by landlessness, unemployment (due mainly to rise in
labour supply in the agricultural sector) and flood and river erosion, is seen as one of
the main causes of rural-urban migration, it is found to be causing urban poverty as
well (Muqtada, 1981).
Another important area of poverty research attempts to determine whether poverty
has a gender bias. In other words, this category of poverty literature looks to examine
whether poor women are poorer than poor men. Some of these researches have found
an association between gender and poverty. Occupational distribution of poverty has
been the focus of some researchers. Notable work on this issue mainly concentrates
on poverty among garment workers (Khan, 1985).
A major part of the poverty research is dedicated to measuring the success or
effectiveness of various poverty alleviation measures undertaken by government and
non-government agencies. Such research includes measuring the success of food for
work programs, agricultural subsidies, agricultural credit and microcredit and other
interventionist programs undertaken by the government and NGOs on different
occasions (Rahman and Razzaque, 2000). Initiatives of several NGOs, particularly
those of the Grameen Bank, BRAC, Proshika Manobik Unnaya Kendra
45
(PROSHIKA), Association of Social Advancement (ASA) and others have been the
focus of a number of studies. Keeping pace with the growth in coverage and volume
of microcredit, the body of such research has also grown to cover various aspects of
microcredit including its role in poverty alleviation, empowerment of the poor
especially poor women, and comparative evaluation of government and NGO
microcredit programs.
Three major dimensions: are under consideration in this study: impact of credit on
household outcomes, impact of credit on women’s empowerment and impact of
credit on household consumption. The literature on these three broad topics is
studied in this chapter. Since not enough literature is found on the impact of
microcredit on consumption patterns a review of literature on consumption (overall
not specific to microcredit) patterns is provided.
2.4.1 The impact of Microcredit on Poverty (Household Outcomes) in
Bangladesh
Microcredit has appeared in the literature as a popular source of financing that
provides small loans in rural and remote regions of developing countries (Bornstein,
1998; Johnson and Rogaly, 1997; Zeller and Sharma, 2000). As far as developing
countries are concerned, Bangladesh may be considered as the pioneer, having
started this financial innovation that provides loans to the poor especially to women
engaged in self-employment projects that allow them to generate income and in
many cases, begin to build wealth and eliminate poverty (Hulme and Mosley, 1996;
Yunus, 1983; World Bank, 1994).
46
The World Bank (Microcredit Summit, 1997) classified the microcredit program in
Bangladesh as one of the most effective anti-poverty tools for the poorest people.
The program extends small loans to unemployed poor people who are not
creditworthy. These individuals lack collateral, and stable employment and therefore
cannot meet even the most minimum qualifications to gain access to formal credit.
The microcredit (or known as microfinance) program, provides collateral-free small
loans to extremely impoverished people (mostly women) for income-generating
activities thereby reducing poverty (Hossain, 1988; Rahman, 1995; Khandker,
1998a).
Since loans are provided in a group, according to Stiglitz (1993), Varian (1990), and
Ghatak (1999) and Besley and Coate (1995) the members in a group are well placed
to umpire the creditworthiness and scrutinise the actions of their peers, as a
consequence mitigating the problems of adverse selection and moral hazard. Group
lending also gives incentives to members to avoid excessively risky projects
(Stiglitz, 1993). It also provides insurance to other members in the event their
projects fail (Coleman, 1999). Mutual trust among group members created by their
long association with each other provides strong inducement to self-monitoring
which reduces the monitoring cost down to zero.
Many studies have attempted to measure the impact of microcredit on poverty,
income, employment, contraceptive use and fertility (Hossain, 1988; Hulme and
Mosley, 1996; Hashemi and Schuler, 1996). One of the limitations of these studies is
that they fail to address whether improvement in the quality of life is due to program
participation or not.
47
Existing literature on the impact of microcredit on poverty in Bangladesh provides a
controversial picture. Apart from the studies that suggest access to credit has the
potential to significantly reduce poverty (as mentioned above) others argue that
microcredit has a minimal impact on poverty reduction (Morduch, 1999 and 2000;
Weiss and Montgomery, 2005). However, one issue is beyond controversy: everyone
agrees that the “vulnerability22” of the poor has been reduced due to microcredit
programs.
On the issue of vulnerability it is worth mentioning, the study conducted by
Montgomery, Bhattacharya and Hulme (1996), which found an improvement in
household income, is higher for third time borrowers compared to first time
borrowers. There is also a study by Mustafa et al. (1996) that showed the older
members’ asset valuation is 112% higher compared to those of the newer members.
The study further showed that the average weekly expenditure of the household is
higher for the older members than for the newer ones.
Another interesting finding of the study is that it shows 80% of the credit is invested
and the rest is used for consumption. Money is fungible23 and often the cash obtained
from the microfinance institutions24 is used to meet immediate consumption needs.
Several empirical studies support that the view that credit market involvements
improve both consumption and production of the poor via smoothing consumption
and reducing constraints in production (Feder et al., 1988 and Foster, 1995). Even
22 The term “vulnerability” appeared in the literature as the crisis-coping mechanisms of the poor especially when there are natural disasters (Hashemi et al. 1996; Montgomery et al. 1996; Morduch, 2000; Husain, 1998). 23 The dictionary meaning of fungible is something that is exchangeable or substitutable. 24 Synonymously used for microcredit programs.
48
though it is evident from the literature that not all money borrowed is invested by the
households, a portion of it is used for consumption. Therefore, it may be assumed
that microcredit may benefit households in terms of income as well as consumption.
However, there are costs associated with joining the program. The explicit cost is the
interest payment and the implicit cost is the opportunity cost of attending meetings
etc. Rural women in Bangladesh are preoccupied with household work and
producing non-market products which makes the opportunity cost of wage
employment higher for women. This issue is addressed by Khandker (1998b) where
he shows that households would benefit from withdrawing labour from the wage
market if funds are made available to buy the minimum capital needed to initiate
home-based marketable products.
It is observed that the credit program makes a borrower switch from wage
employment to self-employment in farm or non-farm sectors. This process may
increase self-employment, but wage employment will reduce. However, Khandker et
al. (1998) show that wage employment may have declined, but the increases in self-
employment are large enough to offset a reduction in wage employment at the
village level. It has also been suggested in the literature that rural wages have
increased (as there is a reduction in labour supply) at the expense of wage
employment due to the microcredit program. The study further shows that the
Grameen Bank’s credit program has induced about 13.5% increase in the rural wage
and that is due to a reduction in wage employment. In another study, Rahman and
Khandker (1994) show that male and female employment in the Grameen Bank
villages are 14 and 39% higher respectively than those in non-program areas.
49
It has been argued in the literature that, the marginal gain from micro borrowing to
participants may be large, but the accrued total benefits from microfinance in
reducing poverty are likely to be small, as microfinance transactions are often too
small in volume to have a sustained aggregate impact on poverty reduction. Yaron et
al. (1998) have argued that absence of appropriate methodology prevents to find out
the welfare impact or the poverty level in the presence of the microcredit program.
It has been argued that the microcredit program may not be the only way to alleviate
poverty. There may be other non-credit ways such as “food for work” and “targeted
wage employment schemes” to eradicate poverty. Also the grants or donations
received by the government may have alternative uses. Even if clients benefit from
program participation, it may hurt others in the society. It is therefore important to
know the externality effect of the program. It may happen that the benefit accruing to
the poor are generated at the cost of others in the society, so that the program
intervention is not Pareto efficient.25 The World Bank (1994) says Pareto efficiency
is likely to be achieved in areas of economic growth that permits new development
of activities benefiting the participants without hurting anyone else. On the other
hand, Khandker (1998b) argues; that since subsidised funds have alternative uses
through which the poor can benefit, program evaluation of microfinance institutions
must be made based on the cost-effectiveness of alternative anti-poverty programs.
Another issue suggested in the literature is state intervention in alleviating poverty
(Besley, 1997). However, Besley recognises that, there are sections of the poor that
the government does not reach; therefore asking for government intervention
25 Program participation will be Pareto efficient when it benefits at least one person while leaving everyone else at least as well off as before (World Bank 1994).
50
becomes unrealistic in practice. In Bangladesh, the formal financial system has
limited success in targeting the poor with the notable exception of Bangladesh Rural
Development Board’s (BRDB) RD-12 project (Khandker, 1998a).
Researchers have used “program outreach” as a measure of microcredit program
evaluation (Yaron, McDonald and Piperk, 1997; Chaves and Gonzalez-Vega, 1996).
Poverty outreach is important for the governments, policymakers and the donors in
order to know the extent to which microfinance institutions serve the poor in
disadvantaged locations. It may happen that branches of microfinance institutions are
established in locations with better access to transport and communication
infrastructure. Research has showed that microcredit services are geared more
towards the poor who reside in relatively well-developed areas rather than the poor
in more remote and less developed regions. This may not be consistent with the
client density of existing branches that shows more clients in less favourable and
more distressed locations of the rural economy. However, microcredit programs may
be effective in reaching target clients and may generate benefits that are marginal
that do not result in growth of the economy.
Along with outreach there is another issue that has concerned the researchers. It is
well known that all microcredit organisations in Bangladesh are dependent on grants,
soft loans as well as subsidies (Adams and Von Pischke, 1992). Questions such as
“How much subsidy those institutions enjoy”? “Are they able to survive in the long
run if subsidies are withdrawn and if so, how?” (Yaron, 1992b; Yaron, 1994;
McNamara, 1998; Rahman, 1999; Buckland, 1998)’. “Who benefits from
51
microcredit program and program outreach?” (Yaron, 1992a; Sharma and Zeller,
1999)’are posed in the literature.
In the case of microfinance, fixed and transaction costs associated with branch
establishment in rural areas make credit delivery programs expensive ventures to
administer. A large proportion of the literature on microfinance addresses the
financial viability of the targeted credit programs. The literature has covered issues
such as cost efficiency of microcredit programs, where a primary concern is to see
the amount of cost to deliver the services.
In a comparative study among the major small-scale credit programs in Bangladesh
that provide productive credit and other services to the poor such as those provided
by the Grameen Bank, BRAC, and RD-12 project of BRDB, Khandker et al. (1998)
using housing survey data collected by the World Bank and Bangladesh Institute of
Development Studies (BIDS) in 1991-92, attempted to quantify the village-level
impact. Their econometric analysis shows that these programs have positive impacts
on income, production and employment particularly in the non-farm sector.
In a separate study, Pitt and Khandker (1998) evaluated the effect of same group-
based credit programs (again the Grameen Bank, BRAC, and BRDB’s RD-12
programs) on a variety of household behaviours and on the intra-household
distribution of resources. The study estimated the impact of participation, by gender,
in each of the three group-based credit programs on women and men’s labour
supply, boys’ and girls’ schooling, expenditure, and assets using the “Weighted
Exogenous Sampling Maximum Likelihood-Limited Information Maximum
52
Likelihood-Fixed Effects” (WESML-LIML-FE) approach. They found that credit is
a significant determinant of many outcomes such as household expenditure, non-land
assets held by women, male and female labour supply and boys’ and girls’
schooling. Furthermore, credit provided to women is more likely to influence these
behaviours than credit provided to men. The impact of credit on these six outcomes
provided to women is found significant at the 5% level. Annual household
consumption expenditure, the most comprehensive measure available of program
impact, increased 18 taka for every 100 additional taka borrowed by women from
these credit programs, compared with 11 taka for men. These findings suggest that
credit is not perfectly fungible within a household.
An extension of Pitt and Khandker’s (1998) study is conducted by Khandker (2003)
where he estimated the long-run impacts of microfinance on household consumption
and poverty in Bangladesh, based on panel data. The household survey data was
collected in 1992-93 and 1998-99. Pitt and Khandker (1998) showed that the
endogeneity of both microcredit program placement and program participation is a
serious issue and findings could be misleading if this endogeneity is not taken into
consideration while estimation. The method used by Pitt and Khandker (1998) is
based on cross-section data but they employed a quasi-experimental survey design to
resolve the problems of endogeneity associated with non-random program placement
and self-selected program participation. In the study, Khandker used panel data
analysis, which helped measure the program effects on long-term household or
individual welfare.
53
Morduch (1998) using the same data as used by Pitt and Khandker (1998), (BIDS-
World Bank data, 1992-93) but employing a different technique (difference-in-
difference method), finds that program effects are either non-existent or are very
small. Morduch found no evidence of an increase in consumption (and therefore
reduction in poverty) using the same data. Therefore, the contradictory results in
findings provide scope for further study in the area of impact of credit in poverty
reduction.
It appears from the literature that the impact assessment on microcredit provides a
contradictory result. Therefore it is justifiable to assess the impact of microcredit on
various household indicators. It also appears from the literature that many
researchers have used ‘savings’ as an indicator that may reduce poverty. There may
be other indicators that may also contribute to reduce poverty. As far as poverty
alleviation of the borrowers are concerned it is necessary to analyse how these credit
programs may influence in bringing higher income and assets for the borrowers.
Therefore, in this study we have considered “assets” and “income” as poverty
reduction indicators instead of “savings” alone. Some researchers have analysed the
impact of credit on savings using instrumental variable or Two-Stage Least Square
(2SLS) estimation. However, using Tobit estimation along with 2SLS is also a
possible alternative to determine the impact of credit on income and assets using
non-borrowers data as the censored observation. This is a unique idea that has not
been found so far in the literature. Analysing the impact of microcredit on economic
indicators and comparing it as between borrowers and non-borrowers (where the
non-borrowers are the “control group”) is another interesting dimension of research.
54
2.4.2 Women’s Empowerment
The term “women’s empowerment” is an important topic appearing throughout the
literature, but the number of studies in this context is limited. Women’s
empowerment is a subjective issue and is defined by researchers based on their
perception on the issue. In terms of an appropriate methodology, it is a difficult issue
to measure. Some researchers have used a comparison sample as a methodology but
again it is sometimes impossible to find two villages identical in all attributes
making empowerment measurement even harder.
“Women’s empowerment” has been defined by different authors in different ways26.
Rao and Kelleher (1995) define women’s empowerment as the “capacity of women
to become economically self-reliant with control over decisions affecting their lives
and freedom from violence.” Holcombe (1995) defines “empowerment” as “the
sharing of control, and the entitlement and ability to participate in influencing
decisions regarding the allocation of resources”. Hashemi and Schuler (1996) have
argued that a woman’s empowerment is reflected in her relative physical mobility,
economic security, ability to buy things on her own, freedom from domination by the
family, political and legal awareness and participation in public protests and political
campaigning. Sen (1997) has stated, “empowerment is about changes in favour of
those who previously exercised little control over their lives”.
The literature reveals that the impact of credit programs on women’s empowerment
contradicts to each other. Goetz and Sengupta (1996) have found 18% of women
26 Women’s empowerment as defined by this study is discussed in detail in Chapter Five. Chapter Five also provides a table comparing definitions of the term as defined by different researchers.
55
have full control over the loans while 39% of the loans are controlled by the men
even though loans are issued in a woman’s name. According to Goetz and Sengupta
(1996) the credit program is not empowering. They have measured women’s
empowerment in terms of women’s managerial control over loan use. They have
found that the degree of control over loan use increases steadily for a while but
decreases as membership length increases. The findings of this study may not be
dependable because the sample of the study is very small.27
Hashemi et al. (1996) on the other hand, have found that the Grameen Bank and
BRAC have had significant positive effects on women’s empowerment. They argue
that minimalist28 credit programs empower women. They collected data through
participant observation and informal interviews. They chose six villages where both
the Grameen Bank and BRAC operate. They used a combination of sample survey
data and case study approach to argue that the success of the Grameen Bank in
empowering women is due to its strong central focus on credit, and its skilful use of
rules and rituals to make the loan program function.
Hashemi et al. (1996) created an “empowerment index” composed of eight indicators
The indicators are mobility, economic security, ability to make small purchases,
ability to make larger purchases, involvement in major decisions, relative freedom
from domination by the family, political and legal awareness, participation in public
protests and political campaigning. They arbitrarily chose a cut-off point of five and
27 Goetz and Sengupta (1996) have used a sample of 253 female borrowers covering four rural credit providers in Bangladesh. 28 There are group of economist who believe that credit programs would still be helping the poor fight poverty by giving credit to any poor person who is able to repay a loan without dictating to that person how the loan should be used (Mustafa et al., 2000).
56
said if someone scores five points out of eight categories the person is empowered
and not empowered otherwise. To avoid selection bias they used respondents’
demographic and socio-economic characteristics such as age, education, relative
wealth, religion, geographic division and number of surviving children, which may
create heterogeneity problems. The unobserved heterogeneity problem is not
considered in the study. This could have been minimised by dividing the sample into
some strata according to age, education and wealth. The problem of having a small
sample size (after stratifying) could be resolved by using dummies.
Mizan (1993) has used a similar approach as Hashemi et al. (1996). He has used an
index called the Household Decision Making (HHDM) Scale, which is composed of
questions on matters such as decisions of food purchase, education and marriage of
children, expenses on medical for self and husband, the woman’s business earnings,
purchase and sale of land, hiring of outside labour, purchase of agricultural inputs,
the provision of financial support to husband’s family and purchase of clothes for
self and other household members. Mizan has found that the number of years of
borrowing from the Grameen Bank has a positive and significant effect on the
HHDM score. Since only HHDM is used as the indicator of empowerment, this may
be considered as a limitation of the study.
In a recent paper Ashraf et al. (2006) have defined women’s empowerment as
women’s decision making power. In the study they have used randomised controlled
trials and examined whether access to product leads to an increase in women’s
decision-making power within the household. Using data from the Philippines they
have found a positive impact, of access to product and women’s decision making
57
power. The study has suggested that this leads to a shift towards female-oriented
durables goods purchased in the household.
To assess women’s empowerment, Zaman (1998) first divided the questions into
three sections “women’s ownership and control over assets”, “women’s general and
legal knowledge” and “knowledge on fertility and mobility of the women”.
In terms of ownership and control over resources a list of common household assets
is presented and respondents are asked whether they own the items themselves, and
if so whether they could sell them of their own accord, whether they could keep the
proceeds from sale and whether the latter actually ever happened. The legal and
political knowledge section has focused on women’s awareness regarding dowry,
marital age, divorce and “union parishad” chairman’s (local elected representative)
name. The “fertility” section has probed into issues such as whether the woman has
decided to have a child (in conjunction with her husband) or whether it is due
entirely to her husband’s, or even her mother-in-law’s will. The mobility section lists
a number of sites in the locality such as the marketplace and questions whether the
respondent has visited these places in the last four months and if so whether they
went alone or not.
All of these “empowerment correlates” are binary variables29 with the value one for
“yes” and zero for “no”. Zaman (1998) has decided not to construct “empowerment
indices” of any sort due to the problem of assigning subjective weights to different
responses.
29 The responses are transformed into binary variables where necessary; for instance “general knowledge” example discussed above, the “incorrect” and “don’t know” responses are merged into one category.
58
It is clear from the above literature that women’s empowerment has drawn attention
of the researchers, and it is always an interesting issue. It is thus necessary to explore
the idea further and redefine women’s empowerment from a different perspective. It
is also important to find out the factors that affect women’s empowerment. Further,
it is necessary to find out the difference between borrowers and non-borrowers in
terms of women’s empowerment and finally to see if credit programs make
borrowers more empowered or not. To do so, we need to measure “women’s
empowerment”. Since it is a subjective issue this may be measured through index by
using binary variables, which may also eliminate the problem of assigning subjective
weights.
The next section considers the literature on consumption patterns and the
methodology used to measure consumption behaviour as studied in the context of
developed and developing countries.
2.4.3 Household Consumption Behaviour
Studies of consumption behaviour have been pioneered by Stone (1953), and are
carried out by estimation of systems of demand equations explicitly derived from
consumer theory. The availability of household survey data from developed
countries explains the huge empirical literature that has attempted to estimate
systems of demand equations for different consumption categories. The estimation of
the consumption-income relationship or Engel relationship from cross-section data
has been paid considerable attention in the literature on developed countries.
59
Blundell and Ray (1984) have studied the Engel curve analysis on demand system,
while Giles and Hampton (1985) have studied the same on household expenditure
using New Zealand data. Beneito (2003) has estimated income elasticity using Engel
curve analysis in relation to the Spanish economy using Seemingly Unrelated
Regressions Estimator (SURE) and Three-Stage Least Squares Estimates (3SLS).
Sawtelle (1993), on the other hand, has estimated two linear Engel functions for
household total expenditure and 15 aggregate classification of consumer durable,
non-durable and service expenditures using United State’s cross-section data. Each
expenditure equation is estimated using OLS and the 95 %level of significance is
applied in testing all regression results for the models. Using data from the United
States, Lee and Brown (1986) examine food expenditure on household data.
Apart from Engel model, Deaton and Muellbauer (1978) have introduced An Almost
Ideal Demand System (AIDS) model, which has been subsequently adopted by Ray
(1980). Moschini and Vissa (1993) have used the alternative of using direct
approximation to mixed demands with an application to meat demand in Canada.
The number of studies on consumption in the context of developing countries is not
overwhelming. In this regard it is worth mentioning the study conducted by
Weiskoff (1971), on demand elasticity in the developing economy. Ray (1980)
added family size to the AIDS model and applied it to Indian budget data to estimate
expenditure, price, and size elasticity. The usual mean-variance assumptions are
made about the stochastic error term, and F-tests are carried out to investigate the
hypotheses. Ray has estimated single equations using OLS estimates in the absence
60
of cross-equation or non-linear restrictions. In another study, Meenakshi and Ray
(2002) combine the expenditure and demographic information contained in the unit
records of nearly 70,000 households to analyse rural poverty in India. In the presence
of non-linearity and cross-equation restrictions of the parameters appearing in each
equation the study estimated the model as a system of equations using non-linear
Full Information Maximum Likelihood (FIML) estimation using SHAZAM 8.0
computer package.
Dey (2000) has analysed the demand for fish in Bangladesh. The adding-up
conditions imply a singular variance-covariance matrix for the disturbances if all the
“n” demand relations of the equations are estimated jointly. The study has employed
the Iterative Nonlinear Seemingly Unrelated Regression (ITSR) method of
SYSNLIN procedure of SAS (SAS1984) to estimate the share equations.
Hendriks and Lyne (2003) have used panel data on two villages of Africa. They have
used OLS estimates on absolute budget share (ABS) as well as marginal budget
share (MBS). The ABS is the consumption on each item divided by the total
expenditure while MBS is the changes in consumption on each item divided by the
changes in total expenditure.
Ferdous (1997 and 1999) uses the AIDS model on household consumption using
secondary data from the Bangladesh Bureau of Statistics. She has estimated each
single equation using OLS estimates after applying the Breusch-Pagan test for
heteroscedasticity.
61
There appears to be no literature found on the impact of microcredit on consumption
using the AIDS model. Since it is a new idea, it needs to be explored. It is thus
important to find out the impact of microcredit on household consumption using the
AIDS model. It is also important to find out the difference between borrowers and
non-borrowers in terms of consumption.
2.4.4 Conclusion of Part C
Part A introduces the relevant theories on the rural credit market; Part B describes
Bangladesh economy before and after microcredit movement and Part C provides a
review of the wide range of literature on microcredit research. It is not feasible here
to look at all the broad areas carefully and identify every gap from the literature for
research. We have thus confined our study to three areas such as: the impact of credit
on household outcomes, the impact of credit on women’s empowerment and the
impact of credit on household consumption. The most appropriate way to conduct
such particular type of research is through primary data.
At this stage we have developed models30 focusing on these three issues. We have
also developed a detailed questionnaire composed of all the questions necessary to
find out the relevant answers to the problems. Apart from the basic questions of the
respondents the questionnaire is composed of three sections. These three sections
consist of specific questions designed to conduct research on the three issues of
interest (that is the impact of credit on income and assets; the consumption patterns
of both borrowers and non-borrowers; and the extent of women’s empowerment). A
30 Models are developed and selected to analyse the basic problems identified by the study. Models corresponding to each problem are provided in relevant chapters.
62
sample questionnaire used for collecting primary data is provided in Appendix C and
the data collection procedure is described in Chapter Three.
63
Chapter Three
ASSESSMENT OF INCOME AND ASSET ACCUMULATION OF
MICROCREDIT BORROWERS31
3.1 Introduction
Microcredit programs that were initially introduced in Bangladesh have become
popular worldwide. This initiative to provide small finance has not only been
adopted by developing countries (for example, Bangladesh, Malaysia, Guinea,
Malawi, Colombia, the Philippines and Nigeria), but it also has been introduced by
developed nations (for example, United States and Canada). The microcredit
programs in Bangladesh have become successful mainly because of their high
recovery rate, group-based lending strategy and focus on women.
As already stated these programs provide small loans to rural people especially to the
poor with the aim of eradicating poverty. In particular, one of the main objectives of
these programs, that of “credit to the poor” is intended to stimulate the rural
economy and boost the rural financial market (Rahman and Khandker, 1994, p. 50).
However, the impact of microcredit has so far been found contradictory. Several
studies have found that the credit programs have a positive impact in eradicating
poverty (Hossain, 1988; Khandker, 1998b; Yaron, 1994) while others have found the
impact to be negative (Morduch, 2000: Weiss and Montgomery, 2005). To
substantiate the truth of the controversy it is important to assess the impact of these
31 A research paper based on the findings of this chapter has been presented and published in the conference proceedings (refereed) for the Fourth International Business Research Conference. The abstract of the paper is provided in Appendix H.
64
credit programs on household income and asset accumulation of the microcredit
borrowers in Bangladesh. In so doing, it is necessary, firstly, to identify factors that
are essential measures/indicators of poverty and, secondly, to find out whether these
factors are ultimately affected by these credit programs as they aim to eradicate
poverty.
Khandker (2000) considers household savings as an indicator and found that this
factor has an influence on eradicating poverty. Khandker argues that credit
programs32 do stimulate savings as microcredit borrowers are compelled to save
money every week, which they are entitled to withdraw at the end of their
membership. In addition, Khandker (2000) has found that credit programs have a
positive impact in generating not only voluntary savings (mandatory under credit
programs) but also additional savings among the borrowers. Apart from savings, it
can be argued that there are other factors that may contribute to eradicate such
poverty. For example, income and accumulation of assets of the household may be
considered as additional determining factors. It is likely that with the introduction of
microcredit programs, borrowers will have better income as well as more assets. It is
therefore necessary to analyse how these credit programs can influence higher
income and assets for borrowers. In so doing, Two-Stage Least Square and Tobit
estimation are used to determine the impact of microcredit on borrowers’ income and
assets.
This chapter is organised as follows. Section 3.2 begins previous research on the
impact of microcredit on poverty, elaborates the sources of data, and develops
32 By credit programs we are referring only to microcredit programs.
65
hypotheses. Section 3.3 specifies the model for the analysis. Section 3.4
demonstrates the necessary instruments and the variables for the analysis. Section
3.5 analyses the results and findings. A conclusion is drawn is the final section.
3.2. Previous Studies on Impact Assessment
Khandker et al. (1998) have used data from the three most important microcredit
programs in Bangladesh namely, the Grameen Bank, Bangladesh Rural
Advancement Committee (BRAC) and RD-12 project run by the Bangladesh Rural
Development Board (BRDB). Khandker et al. have attempted to quantify the village-
level impacts of these programs using OLS estimation. The econometric analysis
shows that these programs have positive impacts on income, production and
employment particularly in the rural non-farm sector.
Pitt and Khandker (1998) in a study have estimated the impact of participation by
gender in each of the three group-based credit programs in Bangladesh on women
and men’s labour supply, boys and girls schooling, and expenditure and assets using
Weighted Exogenous Sampling Maximum Likelihood-Limited Information
Maximum Likelihood-Fixed Effects (WESML-LIML-FE). They have used housing
survey data collected in 1991-92 by the World Bank and Bangladesh Institute of
Development Studies (BIDS). Their findings suggest that credit is a significant
determinant of many outcomes such as household expenditure, non-land assets held
by women, male and female labour supply and boys’ and girls’ schooling.
Furthermore, credit provided to women is more likely to influence these behaviours
than credit provided to men.
66
Morduch (1998), using the same data set, has obtained a very different result. Instead
of using regression discontinuity design to estimate marginal impacts (as used by Pitt
and Khandker), Morduch has used average impacts. He has made an assumption that
there is no spill over from program clients to non-participants. With no spill over, the
average impact per participant is calculated by dividing the impact per eligible
household by the proportion of eligible households that participate. Morduch used
difference-in-difference estimates and indicates that program participants do not
benefit in terms of greater consumption levels, but they participate because they
benefit from risk reduction.
Khandker (2003) has estimated the long-run impacts of microfinance on household
consumption and poverty in Bangladesh. Montgomery et al. (1996) have used
retrospective questioning to calculate changes in household income. They have
found that improvements in household income are greater for third-time borrowers
compared to the first-time borrowers. Mustafa et al. (1996) have found that older
members’ gross household asset values are greater than those of newer members.
3. 2.1 Data
In order to determine the impact of microcredit on income and asset accumulation,
data are collected from primary sources. As well, a detailed questionnaire (see
Appendix C) was prepared. Appropriate econometric methods are used to analyse
the collected data and test the hypotheses. Modification of existing models used by
prior researchers by adding new variables and dummies to estimate the equation is
carried out for the development of hypotheses and analysis.
67
The sample is drawn from the three major districts of Bangladesh. These are
Gazipur, Dinajpur and Chokoria where both the Grameen Bank and BRAC operate.
The districts are chosen on the basis of different agro-climatic and socio-economic
conditions. From each district five villages33 are chosen at random. The borrowers
are selected in a cluster from each village. Non-borrowers are selected from the same
village sharing the same socio-economic and cultural background as the borrowers.
Data of non-borrowers are collected from the same cohort to provide a control group
for comparison. The sample of borrowers and non-borrowers are randomly selected
without replacement from the list of households available from the programs’ local
office in each village. From all three districts 387 borrowers and 184 non-borrowers
were interviewed through a questionnaire.
3.2.2 The Selectivity Problem
One frequently observed source of bias in cross-sectional data is that the sample may
not be randomly drawn from population. This may cause selection bias. This bias
can be corrected through Heckman’s two-step correction procedure. The problem
with the Heckman procedure is to identify suitable instruments. The first equation is
used to construct a selectivity term called the “Mills ratio” which is added to the
second stage outcome equation.
The term “selectivity problem” appeared in the literature in different ways. Khandker
(1996) addressed the “selectivity problem” using an econometric technique that had
as its basis the assumption that households with more than 0.5 acres of land are not
33 Location of the districts is shown in the map of Bangladesh in Appendix D, list of the villages is provided in Appendix E.
68
included in microcredit programs. Khandker argues due to non-availability of
suitable identifying instruments Heckman’s corrections cannot be used. His view has
been criticised as “half an acre restriction” has its limitations given that a sizeable
proportion of credit programs in Bangladesh include members who do not fulfil this
land criterion (Zaman, 1998; Mustafa et al., 1996; Montgomery et al., 1996).
Reilly (1990) mentions the possibility of obtaining identification by exploiting the
fact that the Mills ratio term is a non-linear function of the exogenous variables used
in the first stage equation. Hence, all the variables in the first stage equation can
enter the second stage, along with the Mills ratio term, in order to identify the
selectivity effect. This “identification on functional form” procedure is normally
used to test for the sensitivity of the estimates from the Heckman procedure to the
particular identification variable used. This identification is obtained “technically”
instead of being “theoretically-based”.
According to theory as stated by Hsiao (2003), when the truncation is based on some
variables and these variables are used as dependent variables in the estimation
process, such estimation will often create what is commonly referred to as selection
bias.
In this study we have, therefore, estimated the models carefully for dependent
variables that do not cause selection biases. We have used household “assets”, and
“incomes”, in Chapter Three, “consumption” in Chapter Four and “the
empowerment index” in Chapter Five as our dependent variables.
69
There is another type of selectivity problem which is called the self-selection
problem. “Self-selection” is a term used to indicate any situation in which
individuals select themselves into a group. We know that the decision to participate
in a microcredit program is self-selective. This type of self-selection problem may be
corrected through Heckman’s two stage correction procedure. The problem with the
Heckman procedure is to identify suitable instruments. Since no suitable instruments
have been identified which would permit the use of techniques such as the Heckman
procedure to correct this self-selection bias, we did not look at the causal impact of
program participation in this study.
3.2.3. Research Questions
(1) Does the microcredit program increase various household outcomes such as:
income and asset accumulation of borrowers?
(2) Are there any differences in income and asset accumulation between different
income level borrowers?
3.2.4. Hypotheses
First Hypothesis: Microcredit programs improve the living standards of rural
households via better household outcomes.
Second Hypothesis: High income level borrowers are better off in term of
income and asset accumulation.
70
3.3 Model Specification
The primary concern of this chapter is to estimate the impact of microcredit on
household outcomes such as income and asset accumulation in the context of
Bangladesh’s major microcredit institutions, namely, the Grameen Bank and
Bangladesh Rural Advancement Committee (BRAC).
To analyse the impact of microcredit using data from borrowers of these two
institutions, we have adapted the model used by Pitt and Khandker (1996).
Pitt and Khandker (1996) considered the credit programs C ij which is the amount of
borrowing of i-th borrower in j-th village depends on some household characteristics,
some village specific characteristics and some other variables. Thus the model used
by Pitt and Khandker (1996) is as follows:
)1.3(C
ijijijcijcij ZVXC επγβα ++++=
where X ij is the vector of exogenous household characteristics (for example, some
demographic factors of the household); ijV is the vector of village characteristics
(community infrastructure); Z ij is also a vector of a set of household or village
characteristics which are different from the Xs and Vs in that they affect C ij . cβ ,
cγ and π are unknown parameters, and c
ijε is a random error composed of three
components. The random error is composed of the following:
71
)2.3(c
ijjjc
ij e++= μηε
where jη and jμ are an unobserved household-specific effect and village-specific
effects respectively, and c
ije is a non-systematic error uncorrelated with the other
error components or the regressors.
According to Pitt and Khandker (1996), the household outcome ijY (we have used
“income” and “assets” whereas Pitt and Khandker used “savings” alone) depends on
the amount of C ij as well, which may be explained as:
)3.3(y
ijijjiyijyij CVXY εδγβα ++++=
where ,yβ yγ and δ are unknown parameters and y
ijε is the error term.
Equation 3.1 shows that amount of borrowing depends on some demographic,
village-specific and some other variables. On the other hand, Equation 3.3 shows
that the dependent variable (we used income and assets separately) depends on the
same demographic, village-specific factors and also amount of borrowings. Amount
of borrowings may be denoted here as a jointly dependent variable (as it appears as a
dependent variable in Equation 3.1 while as an independent variable in Equation 3.3)
making Equations 3.1 and 3.3 a simultaneous equations model. In the language of
simultaneous equations models this type of jointly dependent variables is called an
endogenous variable.
72
3.3.1 Endogeneity of Credit Programs
In the above simultaneous equations model the “amount of borrowing” is an
endogenous variable. The following reasons are provided by Pitt and Khandker
(1996) for “amount of borrowing” to be treated as an endogenous variable.
1. Non-random placement of credit programs
Credit programs are not randomly allocated across the villages of Bangladesh; rather
they are often placed in poorer and more-flood-prone areas.
2. Village attributes may affect both credit demand and household outcomes ( ijY )
Village infrastructural attributes may not be well measured in the data and are likely
to affect both demand for credit and the household outcomes.
3. Household attributes affect both credit demand and household outcomes
Since the households are heterogeneous, household attributes are likely to affect both
credit program and household outcomes.
3.3.2 Methodology
Models containing simultaneous equations are estimated using “Instrumental
Variables” in general. Maximum Likelihood Estimates could also be a possible
alternative, which provides an efficient result. However, if the equations are
73
members of simultaneous equations, OLS estimates will provide biased estimates. In
this chapter we have used Two-Stage Least Squares (2SLS) to estimate the above-
mentioned simultaneous equation model.
Along with 2SLS we have also used Tobit estimation using data from both
borrowers and non-borrowers. Tobit estimation is appropriate for a sample in which
information on the regressand is available only for some observations (known as a
“censored sample”) and not for the rest. Data of “amount of credit” is available for
borrowers only and obviously not for non-borrowers. Therefore, “amount of credit”
will be zero for a group of people (non-borrowers) and contain values for another
group (borrowers). In such cases of censored data the most appropriate estimation
method is Tobit estimation. Tobit estimation is often called a “Limited Dependent
Variable Regression model” because of the restriction put on the values taken by the
regressand (Gujarati, 1995, p. 572).
3.4 Specification of the Instruments and the Variables
In the above-mentioned model, the exogenous regressors Z ij are the instrumental
variables. General norms of choosing instruments are to use variables, which are
uncorrelated with the error term but correlated with the explanatory variables
(independent variables). We have used number of earners in the household and types
of houses as instruments.
74
We have adapted the above-mentioned simultaneous equations model for the
purpose of analysis. Vectors of the dependent and independent variables, as well as
the instrumental variables used in the model, are specified below.
3.4.1 Description of the Variables
We have considered that “the amount of borrowing” depends on demographic
variables such as the “age of the female (borrower)” and the “age of the male
(husband of the borrower)” in the household. It may also depend on the “education
of the female” as well as the “education of the male” in the household. Education of
the male/female is measured as years of schooling. Unlike Pitt and Khandker (1996),
we have considered in this study the number of adult males/females as a percentage
of family size (MAPFS/FAPFS) instead of the number of adult male/female
(ADMALE/ADFEM). The variables MAPFS and FAPFS are measured as the
number of adult male/female divided by the total family size times one hundred.
Since family size is likely to vary, we have considered MAPFS/FAPFS as a better
measure compared to ADMALE/ADFEM. We have also considered in this chapter
“gender of the household head” as a determining factor of amount of borrowing. In a
traditional Bangladeshi society, it is likely that families are male headed. There may
be situations where the female is a widow or abandoned by the husband, or the
husband has migrated to another place for making a living in such circumstances we
may expect a family to be female-headed.
In terms of the village specific variables, we have considered that the amount of
borrowing depends on infrastructural facilities of the village such as electricity, and
paved roads. Dummy variables are used to identify these variables.
75
The instrumental variables used in the study are the “number of earners” and “types
of houses”. The variable “types of houses” is expressed in terms of a dummy, where
the dummy is one for brick and/or tin (roofs and wall) houses and zero for non-brick
and non-tin houses.
For 2SLS estimation we have used borrowers’ data only. In this study we have used
“total expenditure34” (proxy for income) as one of the dependent variables. Not all
the surveyed households used in this study are on a fixed salary. We have therefore,
decided to consider total monthly expenditure as a more appropriate measure than
self-reported monthly income. People are more likely to give accurate answers for
expenditures than for their incomes. They are likely to become sensitive about
declaring incomes as it may not have been reported for taxation purposes. We have
also used “assets of the households” as another dependent variable. We have
preferred in this study to use assets instead of savings as savings are self-reported
and poorly measured. On the other hand, “assets” are physical assets measured in
market value. Therefore, assets of the households such as furniture, television, radio
and other household items other than land and houses are considered for estimation.
3.4.2 Specification of the Variables
Xij = vector of demographic characteristics are:
• Number of males as percentage of family size (MAPFS)
• Number of females as a percentage of family size (FAPFS)
• Age of the female-borrower (AFEM)
34 Total expenditure is measured from the total food and non-food items consumed by the households in a month. Detailed description of all the food and non-food items as well as their average consumption is provided in Chapter Four.
76
• Age of the male-husband of the borrower (AM)
• Education of the female-borrower (EFEM)
• Education of the male-husband of the borrower (EM)
• Gender of the household head (GHH = 1 for male, and zero otherwise).
Vij = vector of village level characteristics (community infrastructure) are:
• Village has electricity (V1 = 1, if there is electricity and zero otherwise)
• Village has paved roads (V2 = 1, if there are paved roads and zero
otherwise).
ijY = dependent variables are the sets of outcomes such as:
• Total expenditure of the households (TE)
• Assets of the households excluding land (ASE).
ijZ = the instruments are the sets of variable other than X’s and V’s such as:.
• Number of earners in the household (NOE) • Types of houses (TYH = 1 for household with tin/brick roofs and/or walls,
and zero otherwise).
C ij = the endogenous variable;
• The amount of credit (CRE).
77
Table 3.1 shows the correlation between all the variables. In our estimation we have
used “log of credit”, “log of total expenditure” and “log of household assets” instead
of “amount of credit”, “total expenditure” and “household assets” respectively. The
table shows the correlation between variables as it has been used for estimation.
Table 3.1 Correlation Matrix of the Variables
AF EM
EF EM
AM EM GHH NOE LN CR
LN ASE
TYH MA PFS
FA PFS
LN TE
AFEM
1
EFEM
-0.14 1
AM
0.97 -0.15 1
EM
0..67 0.95 -0.04
1
GHH
-0.23 0.14 -0..25
0..21 1
NOE
0..30 -0.12 0.29 -0.12 -0.10 1
LNCRE
0..27 -0.07 0.28 0.02 -0.12 0.24 1
LNASE
0..34 0.13 0.32 0.2 -0.01 0.20 0.43 1
TYH
0.06 0.05 0.03 0.12 0.05 0.04 0.07 0.15 1
MAPFS
0.31 -0.03 0.31 0.07 -0.05 0.21 0.17 0.24 0.00 1
APFS
0.03 0.12 0.03 0.07 -0.07 -0.03 0.09 0.09 -0.09 0.02 1
LNTE
0.543 0.124 0.52 0.21 0.04 0.20 0.21 0.57 0.12 0.24 0.003 1
From Table 3.1 it is observed that most of the variables are not highly correlated
with each other except for the “age/education of the female” and the “age/education
of the male”. Age of the male is highly correlated with the age of his wife. Similarly,
education of the male is also highly correlated with his wife’s education. To avoid
multicollinearity we therefore used either male age/education or female
age/education in an equation while estimating models.
78
3.5 Estimation Results and Discussion
The descriptive statistics of the dependent variables and the independent variables
are provided in Table 3.2. Tables 3.3 and Table 3.4 show the estimation results of the
impact of microcredit on household outcomes on total expenditure (income) and
assets respectively. Due to multicollinearity we do not consider age/education of
male and female together in the same equation. We have divided borrowers into
three equal groups (33.33 %) based on their income level. The groups are identified
as high, middle and low income level borrowers. Tables 3.5 and 3.6 provide
estimation results of different income level borrowers. Finally, data have been
pooled from the full set (borrowers and non-borrowers) and Tobit estimation is used
for the censored data. The STATA 8.0 statistical package is used to estimate the data
and White corrected standard errors are used to allow for heteroscedasticity. The
Hansen-Sargan test for over identifying restrictions is allowed for exogeneity of the
instruments.
3.5.1 Descriptive Statistics of Dependent and Independent Variables
Table 3.2 provides the summary statistics of the variables. Average age of females in
the study is 37 years with maximum age of 75 years and a minimum of 16 years.
Average education of the female is 4 years of schooling with the highest education to
the level of a bachelor degree (14 years of schooling). Average age of males is 43
years with a maximum of 80 and a minimum of 20 years.
79
Table 3.2 Descriptive Statistics of Dependent and Independent Variables
Variables Mean Standard Deviation
Minimum Maximum
Age of the female (Years)
36.88 10.28 16 75
Education of the female (Years of schooling)
3.82 3.28 0 14
Age of the male (Years)
43.02 10.43 20 80
Education of the male (Years of schooling)
4.60 3.90 0 16
Gender of the household head (%)
0.91 0.27 0 1
Number of males as a percentage of family size (%)
29.48 16.77 - -
Number of females as a percentage of family size (%)
27.18 13.78 - -
Number of earners in the household
1.73 0.70 1 5
Family size
5.4 1.95 2 15
Amount of borrowing (Taka)
8017.51 10268 4000 100,000
Total Expenditure(Taka)
6743.89 4289.02 1300 27000
Assets (Taka)
7126 68931 504 1,625,000
Type of houses
0.52 0.49 0 1
Average education of the male is approximately 5 years of schooling with the
maximum of a master’s degree (16 years of schooling). 91% of the households are
male-headed the rest by a female. Average number of adult males as a percentage of
family size is 29% while for females it is 27%. Average number of earners in a
80
family is 1.73 and the average family size is 5.4. Average amount of borrowing is
8017 taka35 and 51 paisa36 with the highest of 100,000 taka. Average monthly total
expenditure of the household is found to be 6,743 taka with the highest 27,000 taka.
Average accumulated assets are 7126 taka with a maximum of 1,625,000 taka and a
minimum of 504 taka. 52% of the houses are found to have tin or brick roofs and/or
walls and the rest are built with other local housing materials such as straw, mud etc.
3.5.2 Impact of Microcredit on Household Outcomes: Income and Asset
In this study we have estimated simultaneous equations model to determine the
impact of borrowing on household outcomes such as total expenditure (income) and
assets using 2SLS estimation. Equations 3.1 and 3.3 of Section 3.3 are estimated
separately for total expenditure and assets as the dependent variables. Variables such
as amount of credit, total expenditure and assets are log transformed. Table 3.3 and
3.4 show the estimated results for total expenditure and assets respectively. Table 3.1
shows high correlation between age and education of male and female. Two separate
equations for male and female (age and education) are estimated. Results of these
two equations are shown as Equation 1 and 2 (columns 1 and 2) in Table 3.3 and 3.4.
35 Taka is the currency of Bangladesh. 36 Paisa is the lowest denominator of currency.
81
Table 3.3 2SLS Estimation of Amount of Borrowing on Household Outcome:
Log of Total Expenditure/Income
Explanatory Variables Log of
Total Expenditure (Income)
Equation 1 Equation 2
Constant 4.35***
(2.58)
3.72***
(2.12)
Log of Amount of Borrowing 0.95***
(3.74)
1.94***
(3.66)
Age of the Female 0.56**
(2.07)
-
Age-squared (female) -0.28***
(-2.58)
-
Education of the Female 0.38*
(1.82)
-
Age of the Male - 0.55**
(2.06)
Age-squared (male) - -0.46***
(-2.78)
Education of the Male - 0.22**
(2.28)
Gender of the Household Head 0.31***
(3.87)
0.29***
(3.41)
Number of Males as a
Percentage of Family Size
-0.001
(-0.73)
-0.001
(-0.71)
Village has Electricity 0.01
(0.12)
0.04
(0.39)
Village has Paved Roads -0.03
(-0.62)
-0.008
(-0.16)
2R
0.25 0.14
F-statistic (8,379)
Prob. > F
23.33
0.0000
22.14
0.0000
Number of Observations 387 387
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values using White corrected standard errors.
82
Table 3.4 2SLS Estimation of Amount of Borrowing on Household Outcome:
Log of Assets
Explanatory Variables Log of Assets
Equation 1 Equation 2
Constant -3.72**
(-2.12)
-7.37*
(-1.68)
Log of Amount of
Borrowing
1.41***
(2.72)
1.51***
(2.74)
Age of the Female 0.01**
(2.05)
-
Education of the Female 0.02**
(2.28)
-
Age of the Male -
0.56**
(2.07)
Education of the Male -
0.04**
(2.67)
Gender of the Household
Head
0.41*
(1.72)
0.48*
(1.69)
Number of Males as a
Percentage of Family Size
0.003
(0.52)
0.003
(0.54)
Village has Electricity 0.56**
(2.07)
0.48
(1.69)
Village has Paved Roads -0.12
(-0.90)
-0.07
(-0.52)
2R 0.16
0.17
F-statistic (7, 379)
Prob. >F
25.52
0.0000
22.29
0.0000
Number of Observations 387 387
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values using White corrected standard errors.
83
Table 3.3 shows that coefficient of total expenditure (income) is significant and
positive. This implies incomes of the households are affected positively by the
amount of credit. As amount of credit increases expenditure (income) of the
households also increase. Age of the female as well as male is found significant and
positive. That tells us that as age of both male and female increases, income of the
household also increases.
To see the effect of age we have further estimated the equation using age-squared.
We have found negative and significant coefficient of age-squared for both male and
female. This implies that as age increases income of the household increases but
after certain level it starts dropping. This is a realistic finding as in real life people’s
income increases as they grow older but falls after a certain level when they stop
working.
Table 3.3 also shows a positive coefficient for education of both male and female.
That means education has a positive role in providing higher income for the
households. The coefficient of education for female and male is found significant at
5% and 10% level respectively. In our sample most of the households are male-
headed. Our estimation results from Table 3.3 suggest that the male-headed
households have a larger expenditure (income) for the family.
Table 3.4 shows the 2SLS estimation results of the simultaneous equations model as
discussed in Section 3.3 of this chapter using “assets” as one of the dependent
variables. We have used log transformed assets and credit to estimate our model. As
mentioned earlier, to avoid multicollinearity we have estimated two separate
84
equations for male/female age and education. These are shown separately in columns
1 and 2 of Table 3.4 as Equations 1 and 2 respectively.
Table 3.4 shows that the microcredit program affects household assets significantly
and positively. The positive coefficient of assets implies that as credit increases,
assets of the households also increase. This could be interpreted as one of the
achievements of microcredit programs. Table 3.4 also shows a positive and
significant coefficient for age of the male and female. That means as age of the male
increases, household assets also increase. We also could conclude that education
contributes to better quality of life through enhancing household assets for both male
and female. Education enlightens everyone and plays an important role in bringing
better assets for the households. Table 3.4 also shows that gender of the household
head is positive and significant at the 10% level. The positive and significant
coefficient of electricity implies that having electricity in villages may facilitate
household assets.
Summarising Tables 3.3 and 3.4 we may conclude that microcredit programs are
facilitating households to have better income as well as assets. We therefore accept
the null hypothesis that the microcredit programs improve household outcomes. Our
results suggest that the microcredit programs are successful in improving the quality
of life of borrowers through accumulation of assets and income. These results are
consistent with the findings of Pitt and Khandker (1996).
85
3.5.3 Impact of Microcredit on Household Outcomes Based on Different Income
Level Borrowers
Since we have found a positive impact of microcredit on household outcomes, now
we want to see how this impact varies among different income level borrowers. We
have therefore divided the borrowers into three equal groups according to their
income levels. Each group contains 33.33% borrowers and are classified as high,
middle and low income groups. As we know 94% of microcredit borrowers are
female at the same time there is a high correlation between age and education of
male and female. We have therefore, decided to focus only on female age and
education in our following estimations.
Tables 3.5 and 3.6 show the 2SLS estimation of the effects of microcredit programs
on household outcomes (total expenditure and assets) based on different income
level borrowers. The results of total expenditure and assets are shown in Table 3.5
and 3.6 respectively.
The findings of Tables 3.5 and 3.6 are very interesting. We have found positive and
significant coefficient of income for high income level borrowers. For middle
income level borrowers’ income is positive but significant at the 10% level. The
table also shows that the age of the female is positive and significant across all
income levels.
86
Table 3.5 2SLS Estimation of Amount of Borrowing on Different Income Level
Borrowers: Log of Total Expenditure
Explanatory
Variables
High Income
Level Borrowers
Middle Income
Level Borrowers
Low Income
Level Borrowers
Constant 5.00*** (3.69)
6.69***
(3.99)
5.31***
(2.94)
Log of Amount of Borrowing
0.455*** (2.86)
0.12* (1.68)
0.23
(1.17)
Age of the Female 0.01**
(2.56)
0.13***
(3.11)
0.3***
(3.21)
Education of the
Female
0.024*
(1.90)
0.01*
(1.86)
0.01
(0.08)
Gender of the
Household Head
0.33**
(2.23)
0.04
(0.50)
0.32**
(2.54)
Male as Percentage
of Family Size
-0.004
(-1.17)
-0.001
(-0.55)
-0.001
(-0.39)
Village has
Electricity
0.001
(0.01)
0.07
(1.24)
0.59***
(3.71)
Village has Paved
Roads
-0.11
(-0.88)
-0.06
(-0.82)
-0.02
(0.27)
2R .25 .29 .27
F-statistic (7, 121) Prob. > F
23.22
0.0000
24.31
0.0000
23.21
0.0000
Number of observations
129 129 129
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values using White corrected standard errors.
87
Table 3.6 2SLS Estimation of Amount of Borrowing on Different Income Level
Borrowers: Log of Assets
Explanatory
Variables
High Income
Level Borrowers
Middle Income
Level Borrowers
Low Income
Level Borrowers
Constant 6.60*** (4.28)
5.31***
(3.60)
4.35***
(2.58)
Log of Amount of Borrowing
1.69*** (2.85)
0.12* (1.68)
1.99
(1.30)
Age of the Female 0.18*** (3.17)
0.01**
(2.40)
0.02***
(2.40))
Education of the
Female
0.003***
(2.58)
0.19
(1.21)
0.001
(1.09)
Gender of the
Household Head
-0.16
(-1.42)
-0.65
(-1.09)
-0.18
(-1.07)
Male as Percentage
of Family Size
0.008***
(2.89)
0.18**
(2.58)
0.33**
(2.23)
Village has
Electricity
0.03**
(2.00)
0.18
(1.07)
0.008
(0.81)
Village has Paved
Roads
-0.20
(-1.49)
-0.001
(0.05)
-0.12
(-0.72)
2R .21 .26 .31
F-statistic (7, 121) Prob. > F
20.68
0.0000
23.99
0.0000
21.19
0.0000
Number of observations
129 129 129
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values using White corrected standard errors.
88
Table 3.5 further shows that as age increases income of the households increase for
all income level borrowers. Education of the female is positive and significant at the
10% level for high and middle income level borrowers. Male-headed households
provide better income for high and low income level households. From Table 3.5 we
have found that having electricity in the village increases income for the low income
level borrowers.
Table 3.6 shows the results of the impact of microcredit on household assets on
different income level borrowers. It is interesting to see from the significant and
positive coefficient that as amount of borrowing increases, assets increase only for
the higher income level borrowers. The coefficient is positive and significant at 10%
level for the middle income level borrowers. Age of the female is positive and
significant across all income level borrowers. This implies that as age of the female
increases, households possess more assets. However as education of the female
increases household assets increase for higher income level borrowers only.
Households with more adult males provide better assets across all income groups.
Having electricity in the village, increases assets only for the higher income level
borrowers.
Summarising the results of Tables 3.5 and 3.6, our findings suggest that the high-
income group borrowers are better off compared to middle and low income group
borrowers. We therefore accept the null hypothesis that the high income group is
better off compared to low and middle income group borrowers. The tables further
show that, as the amount of credit increases, income and assets of high income group
borrowers increase. This is interesting to see as education of the female increases
89
household assets increases only for the higher income level borrowers. Electricity in
the village expedites income for the low income level borrowers only.
3.5.4 Estimating Censored Regression Problem on Household Outcomes
So far in our 2SLS estimation we used only the borrowers’ data to determine the
impact of microcredit on household outcomes. Now we want to pool the whole data
set including borrowers and non-borrowers to determine the impact of microcredit
on household outcomes. We have thus used Tobit estimation to estimate the
censored data. Amount of credit contains some values for the borrowers and zeros
for the non-borrowers. Tobit estimation is normally used for this type of censored
data. Tobit estimation provides the best estimation for model containing censored
observation. Zero values of the dependent variable correspond to censored
observations while positive values correspond to actual transactions. James Tobin
proposes the Tobit model37 for analysing the censored regression problem. To fit in
the same simultaneous equations model used earlier in this chapter we used Tobit
estimation on Equation 3.1 of Section 3.3 containing censored data for amount of
borrowers. Thus predicted rather than actual borrowing is used in Equation 3.3 to
estimate the impact of borrowing on various household outcomes using OLS
estimation. In the following estimation we have used variables without logs such as
“amount of credit”, “assets” and “income”. Table 3.7 shows the results of Tobit
estimation and Table 3.8 shows OLS estimation of household outcomes using the
estimated value of amount of borrowing obtained from Tobit estimation.
37“Estimation of Relationships for Limited Dependent Variables”, Econometrica, January 1958.
90
Table 3.7 Tobit Estimation of Amount of Credit Using Censored Data
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values.
In our previous estimations we overruled the possibility that the village that has
electricity may also have paved roads. The coefficient of the dummy variable for
“paved roads” was not found significant in our previous estimations. We have,
therefore, used only “electricity” in the following estimations as one of the
community infrastructure factors.
In our Tobit estimation as shown in Table 3.7 we see that out of 571 observations
there are 183 left-censored observations, 1 right-censored observation and 386
uncensored observations. That leaves us with 570 observations.
Constant 2.47* (1.77)
Age of the Female 0.09*** (3.95)
Education of the Female 0.44*** (2.48)
Gender of the Household Head -.23 (-1.42)
Number of Earners in the Family 1.13*** (3.16)
Types of Houses 0.13 (0.28)
Village has electricity 1.73*** (2.28)
Number of Males as a Percentage of Family Size
0.04*** (2.86)
Number of Observations 570
Pseudo 2R 0.045
LR χ 2 (7)
Prob.> χ 2
76.63 0.0000
91
Table 3.7 shows the Tobit estimation results of the amount of credit (not log
transformed) using censored data. From the table we see that the credit programs are
positively and significantly affected by the age of the female. The microcredit
programs are also affected positively and significantly by the education of the
female. It is good to see that females’ education helps them to manage credit
properly. It is also interesting to see a positive and significant coefficient of “the
number of earners” in the family. That means if a family has more earning members
that helps them to have quick repayment as well as obtain more credit. The positive
and significant coefficient of the variable “electricity” suggests that having
electricity in the village facilitates borrowing by the households. Number of adult
males in the family also facilitates credit programs. This may be concluded from the
positive and significant coefficient of the variable (MAPFS). Tobit estimation results
show that the credit program is also affected by infrastructural facilities such as a
village having electricity.
Table 3.8 shows OLS estimation results of household outcomes: expenditure
(income) and assets using the estimated values of amount of credit obtained from
Tobit estimation. Using censored data provides us slightly better results than the
2SLS estimation. This could be due to having more observations in censored
estimation.
From Table 3.8 we see that the impact of microcredit on household income is
positive and significant. This implies that the income of the household increases with
respect to amount of credit. This is one of the achievements of the microcredit
programs. Table 3.8 further shows age of the female is significant and positive. This
92
implies that as age of the female increases, income of the household increases. In
real life we see people’s income increase with age. Education of the female in the
household is positive and significant. It implies that as the female becomes more
educated, income of the household increases. This is also another realistic finding.
Table 3.8 further shows having electricity in the village enhances household income.
From Table 3.8 we see that the microcredit program enhances household assets. It
also shows that age and education of the female in the household as well as having
better infrastructural facilities in the village facilitates household assets.
Table 3.8 OLS Estimation of Household Outcomes Using Estimated Value of
Amount of Credit from Tobit Estimation
Explanatory variables Income Assets
Constant 6.86*** (9.51)
7.28*** (9.54)
Estimated Value of Amount of Credit
0.72*** (2.40)
0.69*** (2.69)
Age of the Female 0.49*** (3.22)
0.54*** (2.96)
Education of the Female 0.02*** (2.88)
0.06*** (2.82)
Gender of the Household Head
-0.86 (-0.16)
-0.088 (-0.97)
Village has Electricity 0.72** (2.39)
0.14** (2.40)
2R 0.13 0.15
F ( 5, 563) Prob. > F
19.31 0.0000
17.31 0.0000
Number of observations
570 570
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values.
93
3.5.5 Estimating Censored Regression Problem of Different Income Level
Households
To be consistent with our previous 2SLS analysis we have decided to determine the
impact of credit based on different income levels of borrowers using the censored
data in Tobit estimation. We have thus divided the total sample into three equal
groups containing 33.33% in each group. The groups are classified as high, middle
and low income group. We have used the estimated value of the amount of credit
obtained in Tobit estimation and estimated Equation 3.3 using OLS estimation for all
three groups.
Table 3.9 OLS Estimation of Household Outcomes Based on Different Income
Level Borrowers: Total Expenditure (Income)
Explanatory
variables
High Income
Group
Middle Income
Group
Low Income
Group
Constant 3.12*** (3.58)
3.09*** (3.98)
3.59*** (4.99)
Estimated Value of Amount of Credit
0.10*** (2.67)
0.59** (2.21)
0.08 (1.10)
Age of the Female 0.05*** (2.89)
0.42** (2.25)
0.52** (2.26)
Education of the Female
0.57*** (2.77)
0.65*** (2.51)
0.61** (2.09)
Gender of the Household Head
-0.33 (-0.58)
-0.54 (-0.89)
-0.65 (-0.61)
Village has electricity
2.47*** (2.58)
1.48** (2.26)
2.43** (2.25)
2R 0.15 0.16 0.13
F ( 5, 184) Prob. > F
6.80 0.0000
5.52 0.0000
5.43 0.0000
Number of observations
190 190 190
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values.
94
Table 3.10: OLS Estimation of Household Outcomes Based on Different Income
Level Borrowers: Assets
Explanatory
variables
High Income
Group
Middle Income
Group
Low Income
Group
Constant 2.15*** (2.78)
2.09*** (3.05)
3.56*** (2.58)
Estimated Value of Amount of Credit
0.015*** (2.98)
0.07** (2.40)
0.03 (1.55)
Age of the Female 0.006*** (2.22)
0.01** (2.32)
0.15** (2.24)
Education of the Female
0.02*** (2.71)
0.06*** (2.55)
0.05** (2.22)
Gender of Household Head
-0.55 (-0.51)
-0.45 (-0.75)
-0.02 (-0.63)
Village has electricity
0.81*** (2.22)
0.71*** (2.61)
0.21 (1.09)
2R 0.14 0.13 0.17
F ( 5, 184) Prob. > F
5.66 0.0000
6.54 0.0000
5.51 0.0000
Number of observations
190 190 190
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and figures in the parentheses show the t-values.
Tables 3.9 and 3.10 show the OLS estimation of household outcomes based on
different income level borrowers using the estimated value of amount of credit
obtained from Tobit estimation. Interestingly the results as shown in Table 3.9 and
3.10 reveal that the impact of microcredit are more pronounced for high income and
middle income group households compared to low income group households. The
results from Table 3.9 and 3.10 are consistent with the results obtained from 2SLS
estimation. They show that age and education of female affect income as well as
assets positively and significantly. The tables further show that having electricity in
the village expedites household outcomes such as income and assets.
95
3.6 Conclusion
In this chapter we have analysed the impact of microcredit on household outcomes
such as income and assets. We have modified the simultaneous equations model
suggested by Pitt and Khandker (1996) by assessing the impact of microcredit
separately on household expenditure (income) and assets. We have further
differentiated the impact of credit on different income level borrowers. In addition
we have also used some refined and log transformed variables to estimate the model
compared to previous studies.
From the estimation it may be concluded that the microcredit programs are effective
in generating higher income and assets for borrowers. This is a very important
finding as it confirms the claims made by the microcredit providers. We may
conclude from this finding that microcredit is providing better quality of life for its
borrowers. We know from experience that people’s income increases as they grow
older. Our estimation results suggest that age of the female as well as male has a
significant and positive impact on income and assets. As someone grows older their
income also grows. We also wanted to see what happens to someone’s income as the
person grows even older. To do so, we have added “age-squared” in the equation and
estimated the model. The negative and significant co-efficient tells us someone’s
income increases as the person grows older but income starts to decline after a
certain level when they stop working. Our results also suggest that education of the
female as well as male are an important factor in affecting income and assets
positively. Education helps people to be more enlightened. According to this finding,
education enhances borrowers’ income and assets. We have also found from this
96
analysis that infrastructural facilities in the village facilitates borrowers’ income and
assets.
This study also suggests that as the number of earners increases in a household,
amount of borrowing also increases. This could be due to the fact that having more
earning members increases the ability to repay loans quickly as well as obtain more
loans. Male-headed households are able to enhance household income and assets. It
is interesting to see that even though microcredit facilitates household outcomes the
impact varies between different income level borrowers. Our results suggest that
microcredit programs help bring better outcomes for high income group borrowers
compared to medium and low income group borrowers.
In general it may be concluded that microcredit has a pronounced positive effect on
income and assets of borrowers. However, microcredit seems to be more effective
for the high income group borrowers compared to middle and low income group
borrowers. Microcredit is more effective for educated borrowers compared to less
educated borrowers. Infrastructural facilities facilitate the household outcomes. It is
also found that families having more earning members enjoy better household
outcomes. Overall this study suggests that even though microcredit is an attractive
tool for producing better outcomes in terms of income and assets, it is more effective
for relatively wealthier borrowers compared to non-wealthy borrowers. It would be a
good research question to pursue in the future, to uncover in the future why this is
the case. At the same time, our results suggest that there should be some adjustment
to the existing microcredit programs to achieve the intended outcome, that is, to
serve the purpose of those who are at the bottom of the society.
97
Chapter Four
ASSESSMENT OF CONSUMPTION OF MICROCREDIT
BORROWERS COMPARED TO NON-BORROWERS38
4.1 Introduction
This chapter investigates the consumption behaviour of borrowers in the microcredit
program and compares it with that of non-borrowers of the same village. To capture
the essence of the consumption behaviour of the borrowers, various food and non-
food items, which are commonly used in the rural households, are considered as an
indicator of their consumption pattern. Total expenditure is used as a proxy for
income as it is simply the sum of the expenditure items. It is common to use total
expenditure in place of income as an explanatory variable in the expenditure
relationship because of the conceptual and measurement errors associated with
observations on income. With total expenditure as a proxy for income, one can use
income as an instrumental variable to obtain consistent estimates. Household
expenditure is used in this chapter as a proxy for income as suggested by Alderman
(1993), Devereux (1993), Puetz, (1993) and Hazell and Röell (1983).
Over a period of time microcredit provides collateral free loans to rural people, and it
is argued that there would be a significant difference in the expenditure patterns of
borrowers compared to non-borrowers. The underlying hypothesis in this regard is:
38 Part of this chapter has been accepted for publication in a forthcoming issue of the Journal of Developing Areas. Please refer to Appendix I for the abstract of the paper.
98
“Are the borrowers in the microcredit program better off in terms of consumption
behaviour than the non-borrowers”?
We use budget share of the items consumed by borrowers and non-borrowers and the
items are classified as food and non-food. Budget share is defined as: when a dollar
of a consumer’s budget is selected randomly, the possibility of that money spent on
the ith commodity is yqpw iii /= where ip is the price and iq is the quantity
bought of that commodity and y is the income or total expenditure. When the budget
share is used, the value shares like probabilities should be non-negative and add up
to one. For the purpose of calculation, budget share is calculated for each item
consumed by the households. The basic items that are commonly consumed by the
households are selected to find out the consumption pattern. Figure 4.1 illustrates the
items consumed by the households as a consumption tree.
99
Fig 4.1 Consumption Tree of Food and Non-food Items Consumed by
Borrowers and Non-borrowers
Food
Total Expenditure
Food
Non-Food
Cereal
Lentils Cigarette
s and
betel leaf
Protein Vege-
tables
Sugar
Spices Cooking
Oil
Milk and
Eggs
Fuel
Clothing Education
Health
Travel
Entertain
-ment
Electricity
100
Food items are classified as follows: Cereals (rice, wheat and products of similar
type); lentils (pulses of various kinds); vegetables (includes fruit); meat and fish
(includes, chicken, beef, mutton and fish of all types); milk and eggs and cigarettes,
betel-leaf and betel nuts.
Non-food items are considered as follows: Fuel (includes wood, kerosene and
similar items used for cooking); electricity, cloth, education (includes tutor’s fee and
books and stationery etc); and health (includes medicines, doctor’s bills and hospital
bills).
The rest of this chapter is structured as follows: Section 4.2 provides a literature
review. Section 4.3 discusses the relevant theories. Section 4.4 specifies the
econometric model. Interpretations of the results are provided in Section 4.5. A
conclusion is drawn in the final section.
4. 2 Literature Review on Expenditure Elasticity
The method of Ordinary Least Squares (OLS) is a classical approach to the problem
of estimating the coefficients of a regression equation. However, if the equation is a
member of simultaneous equations, OLS estimates will provide biased estimates
(Summers, 1959). In estimating household expenditure elasticities, total expenditure
is quite often used as a proxy for income while budget share of the item consumed is
used as the dependent variable. In Engel curve analysis, income or total expenditure
is used as the independent variable in OLS analysis. According to Liviatan (1961),
neither of these variables is a satisfactory index of the true economic position of the
family. This results in biased estimates of the income elasticities of the various
101
consumption categories. The bias can, however, be eliminated in large samples by
using both income and total expenditures in the estimation procedure. This can be
accomplished by applying the method of “instrumental variables” to Engel curve
analysis, with recorded income serving as the instrumental variable.
On the other hand, Prais and Houthakker (1955) have estimated budget shares using
OLS estimates. They have stated, “So long as the item of expenditure is a small
proportion of the budget it is not to be expected that serious biases will result….”
According to Summers (1959), if income data are available, the parameters of each
equation could be estimated directly by Least Squares applied to each equation
individually. For brevity, this procedure will be called Single Equation Least
Squares.
Massell (1969) has expressed the expenditure relationship as:
ijkjk
kijijijiiij uDRNXX +++++= ∑=4
1
*
3
*
2
*
10
* γββββ (4.1)
where ijX = expenditure by household j on item i, ∑=i
ijj XX total expenditure
by household j, N = household size, R = the ratio of subsistence to total expenditure,
1 4,.....D D are district dummy variables, iju is the disturbance, and an asterisk denotes
a logarithm. The 'sβ and 'sγ in the equation are the parameters to be estimated. In
addition, the model has used double log relationship and has found to provide a good
fit to satisfy the assumption of homoscedastic residuals.
102
After using additional instruments as “cash income” and “farm size”, the equation
becomes:
)2.4(..... 4714
*
3
*
2
*
10
*
jjjjjjj vDDNTYX +∂++∂+∂+∂+∂+∂=
)3.4(....... 4714
*
3
*
2
*
10 jjjjjjj wDDNTYR +++++++= ξξξξξξ
where Y = cash income, T = land size of the j-th individual, v and w are the
disturbances, an asterisk denotes a logarithm. Since the above two equations are a set
of simultaneous equations, Massell (1969) has used Two-Stage Least Squares
(2SLS). The evidence of heteroscedasticity or non-linearity has been tested for each
individual consumable item in his paper.
As has been established by the theory of consumption, the demand functions that are
derived from the maximisation problem of the consumer make the quantity
demanded of a good or service dependent on the income level of the individual and
on prices. To be consistent with the microeconomic theory (that is that the consumer
is a utility maximiser) it is required that the restrictions such as (1) homogeneity of
degree zero in prices (that is the consumers have no money illusion, or that they react
to relative, not nominal prices); (2) symmetry or that the effect of a change in the
price of good i on the demand for good j is same as the effect of a change in the price
of good j on the demand for good i for constant income, and (3) a restriction for
adding up or that all the budget shares add up to one. The demand functions should
meet the above-mentioned adding-up restriction, the homogeneity and symmetry
conditions. However, when prices are constant, one obtains the specification
corresponding to the Engel curve, where the demand of goods is dependent only on
103
the income of the individual. In such a case, the only condition to be met is the
adding-up restriction since all the others are derived from the consideration of prices
(Beneito, 2003).
There are quite a number of properties, which a “good” Engel curve should have.
These properties stem from considerations on economic, statistical and practical
questions in demand analysis, but unfortunately there is no general agreement to the
above properties in the literature. Therefore, there is no a priori reason to trust in and
use one certain algebraic formulation of Engel curve only. It is well known that
income elasticities for the same commodity and estimated for the same set of
household budget data can differ widely when they are derived from different
functional types of Engel curves (Dax, 1987).
Ray (1980) has used an extension of an Almost Ideal Demand System (AIDS)
model, which was introduced by Deaton and Muellbauer (1978). Ray has used
restrictions such as adding up, homogeneity, symmetry and concavity in the model.
He has used family size and applied AIDS model to Indian budget data to estimate
expenditure, prices and size elasticities. The usual mean-variance assumptions are
made about the stochastic error term, and F-tests are carried out to investigate the
hypotheses. Ray has estimated single equations using OLS estimates in the absence
of cross-equation or non-linear restrictions, while Beneito (2003) has estimated
income elasticities using Engel curve analysis using Seemingly Unrelated
Regressions Estimator (SURE) and Three Stages Least Square Estimates (3SLS).
104
Based on the above-mentioned literature, in this chapter we have firstly estimated
single equations using heteroscedasticity corrected OLS estimation and then have
used iterative Seemingly Unrelated Regression Estimates (SURE) to analyse the
model containing multiple equations.
4.3 Introduction to Relevant Theories
Empirical studies in applied demand analysis using system of demand equations
typically rely on one of two assumptions: either prices are assumed to be
predetermined or quantities are assumed to be predetermined. These assumptions
lead to quantity-dependent or direct demand systems, such as the (direct) translog
(Christensen, Jorgenson and Lau, 1975), the AIDS (Deaton and Muellbauer, 1980),
or the (direct) differential or Rotterdam model (Theil, 1975). Direct demands are the
usual representation of preferences for the individual consumer, who is typically
taken as making optimal consumption decisions for given prices and income. Its use
at the aggregate level is equivalent to assuming that supplies are perfectly elastic and
that demands adjust to clear the market. The alternative of assuming that quantities
are predetermined and that prices adjust to clear the market leads to price-dependent
or inverse demand systems, such as the (inverse) translog (Christensen, Jorgenson
and Lau, 1975) or (inverse) Rotterdam model (Theil, 1975; Barten and Bettendorf,
1989).
In addition to the two polar cases of direct and inverse demand functions, another
class of models allows one to sidestep estimation of both demand and supply
functions in a simultaneous equations framework. This class is the case of “mixed
105
demand” function (Chavas, 1984) where prices of some goods are predetermined
such that the respective quantities demanded adjust to clear the market. However, for
the remaining set of goods quantities supplied are predetermined and prices must
adjust to clear the market.
The following section will discuss the theory of mixed demands and AIDS model
The mixed demand systems needs the knowledge of both direct and indirect utility
functions to characterise the demand properties whereas the Translog and AIDS
model cannot be used to specify a mixed demand system because the flexible
functional forms do not have a close form dual representation.
4.3.1 Theory of Mixed Demands
The mixed demand theory as described by Moschini and Vissa (1993)
assumes mA xxxx ,....,( 21= ) the vector of commodities chosen optimally and
),....,( 21 nmmB xxxx ++= the vector or commodities in fixed quantity whose price is
optimally determined. According to Moschini and Vissa if ip denote the nominal
price of good i, and y is total expenditure on Ax and Bx (income, for short), then
ypv ii /≡ is the corresponding normalised price, and Av and Bv are the vectors of
normalised prices of the two subsets of goods. Mixed demands are then derived from
the constrained optimisation problem (Samuelson, 1965; Chavas, 1984):
BA
BBAABABA
vx
xvxvvvVxxU )4.4(}1.|),(),(max{ =+−
106
where U (.,.) and V (.,.) are direct and indirect utility functions, quasi-concave and
quasi-convex in their respective arguments. The solution to Equation (4.4) gives
Marshallian mixed demands )1,,( BAA xvx and )1,,( BAA xvv . Evidently, at the
optimum,
[ ] [ ] ),1,,()1,,(),1,,( , BABABABBAA xvVxvvvVxxvxU ≡= where V is the mixed utility
function.
The mixed demand functions )1,,( BAA xvx and )1,,( BAB xvv as stated by Moschini
and Vissa (1993) satisfies the typical restrictions of consumer theory. Firstly, they
satisfy the adding-up condition 1.. =+ BBBA xvxv implied by the budget constraint.
Secondly, the homogeneity condition implies that )1,,( bAA xvx is homogeneous of
degree zero in nominal prices and income, that is )1,,( BAA xvx = ),,( yxpx BAA .
Similarly, the optimal nominal prices of group B are homogeneous of degree one in
),,( ypA implying that )1,,( BAB xvv is homogeneous of degree zero in ),( ypA .
According to Moschini and Vissa (1993), ),,((.). yxppvyp BABBB ≡= follows that
the mixed utility function is homogeneous of degree zero in Ap and y, which
implies, ).,,()1,,( yxpVxvV BABA =
As stated by Moschini and Vissa (1993), the symmetry restrictions can be illustrated
in terms of the compensated mixed demand functions. According to Chavas (1984),
the compensated mixed demands are the same as the compensated demands under
rationing, although mixed demands should be carefully distinguished from rationed
demands (in the case of the latter, some markets do not clear). Moschini and Vissa
107
(1993) further stated that the compensated rationed demand can be characterised in
terms of the restricted cost function defined by Gorman, (1976) and Deaton, (1981).
The authors (Moschini and Vissa, 1993) mentioned that a possible reason for the
scarcity of applications of mixed demand systems is that the knowledge of both
direct and indirect utility functions required that characterises the demand properties.
4.3.2 An Almost Ideal Demand System
Stone (1953) has estimated a system of demand equations derived explicitly from
consumer theory. There has been a continuing search for alternative specifications
and functional forms. Many models have been proposed, but perhaps the most
important one is the linear expenditure system, the Rotterdam model (Theil, 1965
and 1976) and Translog model (Christensen, Jorgenson and Lau 1975). Another type
of model has been introduced by Deaton and Muellbauer; the AIDS model which is
comparable to Rotterdam and translog models but has considerable advantages over
both. The advantages of AIDS model are as follows:
• It gives an arbitrary first-order approximation to any demand system.
• It satisfies the axioms of choice exactly.
• It aggregates perfectly over consumers without invoking parallel linear Engel
curves.
• It has a functional form which is consistent with known household-budget
data.
• It is simple to estimate, largely avoiding the need for non-linear estimation.
108
• It can be used to test the restrictions of homogeneity and symmetry through
linear restrictions on fixed parameters.
Although many of these desirable properties are possessed by one or other of the
Rotterdam or translog models, neither possesses all of them simultaneously.
4.3.3 Specification of the AIDS model
In most of the literature on systems of demand equations, the starting point has been
the specification of a function, which is general enough to act as a second-order
approximation to any arbitrary direct or indirect utility function or, more rarely, a
cost function. Alternatively, it is possible to use a first-order approximation to the
demand functions themselves as in the Rotterdam model (Theil, 1965 and 1976).
Deaton and Muellbauer (1980) followed these approaches in terms of generality but
started not from some arbitrary preference ordering, but from a specific class of
preferences, which by the theorems of Muellbauer (1975 and 1976) permit exact
aggregation over consumers: the representation of market demands as if they were
the outcome of decisions by a rational representative consumer. These preferences,
known as PIGLOG39 class, are represented via the cost or expenditure function,
which defines the minimum expenditure necessary to attain a specific utility level at
given prices. The underlying assumption is to maximise utility subject to a budget
39 The cost function underlying both the Engel curves and the Tornqvist type of indexes has been constructed based on the PIGLOG functional form (Deaton and Muellbauer, 1980). The PIGLOG model was developed to treat aggregate consumer behaviour as if it were the outcome of a single maximising consumer. This problem of how to treat aggregate consumer behaviour as if it was the outcome of a single maximising consumer exists because it is neither necessary nor desirable for a macroeconomic relationship to perfectly mimic its microeconomic foundation. Therefore, the market demand functions that are being estimated may not have the desirable properties of micro-demand functions. This is known in economics as the aggregation problem.
109
constraints. Deaton and Muellbauer denoted this function c(u,p) for utility u and
price vector p, and define the PIGLOG class by
{ } { } )5.5()(log)(log)1(),(log pbupaupuc +−=
With some exceptions, u lies between 0 (subsistence) and 1 (bliss) so that the
positive linearly homogeneous functions a(p) and b(p) can be regarded as the costs
of subsistence and bliss, respectively.
As mentioned by Deaton and Muellbauer (1980), the specific functional forms are
taken for log a(p) and log b(p). For the resulting cost function to be a flexible
functional form, it must possess enough parameters so that at any single point its
derivatives 2222 /,/,/,/,/ ucandpucppcucpc ijii ∂∂∂∂∂∂∂∂∂∂∂∂ can be set
equal to those of an arbitrary cost function as shown by Deaton and Muellbauer
(1980). Therefore, according to Deaton and Muellbauer the cost function takes the
following form:
)6.4(loglog2
1log)(log *
0 jkk j
kjkk
k pppaapa ∑∑∑ ++= γ
)7.4()(log)(log ∏+=k
ikppapb ββ
From the above-mentioned equation the AIDS cost function may be written as:
)8.4(loglog2
1log),(log 0
*
0 ∏∑∑∑ +++=k
kkjk
k jkjk
kk puppppuc ββγαα
110
where *,, ijii and γβα are parameters. As mentioned by Deaton and Muellbauer it
can easily be checked that c(u, p) is linearly homogeneous in p (as it must be a valid
representation of preferences) provided that:
.0,1 ==== ∑∑∑∑ j jk ijj iji i βγγα
Deaton and Muellbauer showed that it is also straightforward to check that Equation
4.8 has enough parameters for it to be a flexible functional form. Since utility is
ordinal, Deaton and Muellbauer (1980) chosen a normalisation such that, at a point,
.0/log 22 =∂∂ uc According to Deaton and Muellbauer the choice of the functions
a(p) and b(p) in Equations 4.6 and 4.7 is governed partly by the need for a flexible
functional form. However, as mentioned by Deaton and Muellbauer, the main
justification is that this particular choice leads to a system of demand functions with
the desirable properties, which are demonstrated below.
Deaton and Muellbauer (1980) further mentioned that the demand functions can be
derived directly from Equation 4.8. It is a functional property of the cost function
(Shephard, 1953 and 1970) that its price derivatives are the quantities
demanded: ii qppuc =∂∂ /),( . Multiplying both sides by ),(/ pucpi gives:
)9.4(),(log
),(logi
ii
i
wpuc
qp
p
puc ==∂∂
where iw is the budget share of good i. Hence, logarithmic differentiation of
Equation 4.8 gives the budget shares as a function of prices and utility:
111
)10.4(log 0∏∑ ++= kki
jjijii pupw βββγα
where:
)11.4()(2
1jiijij γγγ +=
Again, as mentioned by Deaton and Muellbauer (1980), for a utility-maximising
consumer total expenditure x is equal to c(u,p) and this equality can be inverted to
give u as a function of p and x, the indirect utility function. According to Deaton and
Muellbauer if this is done for Equation 4.8 and substituted the result into Equation
4.10 the budget shares is obtained as a function of p and x. These are the AIDS
demand function in budget share form:
{ } )12.4(/loglog pxpw ijj
ijii βγα ++= ∑
where P is price index defined by
)13.4(loglog2
1loglog 0 j
j kkkjk
kk pppP ∑∑∑ ++= γαα
The restrictions on the parameters of Equation 4.8 plus Equation 4.11 imply
restrictions on the parameters of the AIDS Equation 4.12. These are taken in three
sets:
112
)14.4(001111
=== ∑∑∑ −==
n
ii
n
iij
n
ii βγα
)15.4(0=∑j
ijγ
)16.4(jiij γγ =
According to Deaton and Muellbauer (1980), provided Equations 4.14, 4.15 and 4.16
hold, Equation 4.12 represents a system of demand functions that add up to one.
However, as stated by Deaton and Muellbauer the total expenditure ),1( =∑ iw is
homogeneous of degree zero in prices which satisfies Slutsky symmetry conditions.
Deaton and Muellbauer have further showed that in the absence of changes in
relative prices and “real” expenditure (x/P) the budget shares are constant and this is
the natural starting point for predictions using the model. Nevertheless, the changes
in relative prices work through the terms ijγ ; each ijγ represents 10 2 times the effect
on the ith budget share of a 1 % increase in the jth price with (x/P) held constant
(Deaton and Muellbauer 1980). Yet the changes in real expenditure operate through
the iβ coefficient; these add to zero and are positive for luxuries and negative for
necessities.
4.4 Model Specification
The AIDS model has been used by Ray (1980). It is originated from Engel curve
analysis, and it is the time series counterpart of Engel function as suggested by Leser
(1976) is as follows:
113
)17.4(log CXB iii ++= βα
where iB and X are the budget shares of the i-th item and total household
expenditure respectively and iα and iβ are the parameters. The model allows
negative coefficients for necessary items and positive for luxury and inferior goods.
AIDS as obtained from the PIGLOG cost function corresponding to the PIGL (Price
Independent Generalised Linearity) after the choice of appropriate functional form is
given by:
nji
CPP
XB jijiii
,.....2,1,
)18.4(loglog
=++⎟⎠
⎞⎜⎝⎛+= ∑γβα
where ,i iα β and ijγ are the parameters and P is an overall price index defined in
terms of individual prices by:
)19.4(loglog2
1loglog 0 j
i jiijii PPPP ∑∑∑ +∂+∂= γ
These functions should meet the well-known adding-up restrictions and the
homogeneity, symmetry and non-negativity conditions. However, when prices are
constant (as in this case40) one obtains the specification corresponding to the Engel
40 In this study cross-section data is collected from three districts of Bangladesh over a period of time. There is insignificant difference in prices found between districts. Therefore, prices are assumed to be constant over the period.
114
curve, where the demand for goods is dependent only on the income of the
individual. In such a case, the only condition to be met is the adding-up restrictions
since all the others are derived from the consideration of prices (Beneito, 2003). The
adding-up restrictions are:
)20.4(001111
=== ∑∑∑ −==
n
ii
n
iij
n
ii βγα
The price formulation Equation 4.18 makes Equation 4.19 a non-linear system of
equations. To avoid non-linear estimation Deaton and Muellbauer (1978) used the
Stone (1954) index as a convenient approximation:
)21.4(logloglog 0
*
ii
i PPp ∑+=≅ βα
Using the utility basis of the PIGLOG model as cost function at the reference year
(p=1) for an individual the linear equation model will take the following form:
)22.4(log.)log(log 00 jj
ijii
iii PPXB ∑∑ +−−+= γβαβα
Ray (1980) incorporated the family size in the above model by using the Barten
(1964) type household utility function and deflating total household expenditure by
family size following Houthakker (1957), Weiskoff (1971) and Sener (1977). The
model takes this form:
115
)23.4(log.)/log( *
0 fPPxB ijj
ijii φγβα +++= ∑
where f, x =(X/f), ijγ and φ denote family size, per capita household expenditure and
the effect of prices and the effect of family size respectively.
As there are no price data (as mentioned above) we have assumed that all the
households face the same prices. This assumption follows Ferdous (1997). So
Equation 4.23 becomes:
*
'
'
loglog
,
)24.4(log
log
PPc
cwhere
fxBor
cfxB
iji
ij
ii
iiii
iiii
βγαα
φβαφβα
−=+=
++=+++=
∑
However, the dependent variable may not only be influenced by variables that can be
quantified on some well-defined scale but also by variables that are basically
qualitative in nature. Such qualitative variables usually indicate the presence or
absence of a “quality” or an attribute. One method of quantifying these attributes is
by constructing dummy variables that take on values of one or zero, where one
indicates the presence and zero indicates the absence of the attribute.
Since the objective of this chapter is to see if there is a significant difference in the
consumption behaviour of the borrowers and non-borrowers of microcredit, a
116
dummy variable called BD (borrower dummy41) is therefore introduced in the model,
where BD = 1 for the borrowers and zero otherwise. However, district dummies are
also included to see if there is any significant difference in consumption between
districts. As the sample data have been collected from three different districts of
Bangladesh, two district dummies are introduced, where DD 1 = 1, for Gazipur and 0
otherwise, and DD 2 = 1 for Dinajpur and zero otherwise. After adding the dummy
variables the model becomes:
12.......2,1
)25.4(log 21
=∀++++++=
i
iiiiiiii DDDDBDfxB εχπγδβα
In the above equation:
BD = 1 for borrowers
= 0 otherwise (that is, non-borrowers)
DD 1 = 1 for Gazipur
= 0 otherwise (that is, Dinajpur and Chokoria)
DD 2 = 1 for Dinajpur
= 0 otherwise (that is, Gazipur and Chokoria)
Mean of non-borrowers:
)0()0|( iii BDBE γα +==
41 We know that the decision to participate in a microcredit program is self-selective. This type of
self-selection problem may be corrected through Heckman’s two stage correction procedure. The problem with the Heckman procedure is to identify suitable instruments. Since no suitable instruments have been identified which would permit the use of Heckman procedure to correct self-selection bias, we did not look at the causal impact of program participation in this study.
117
iα=
Mean of borrowers:
)1()1|( iii BDBE γα +==
ii γα +=
From these regressions it is seen that the intercept term iα gives the mean
consumption of the non-borrowers and ii γα + gives the mean consumption of
borrowers. The coefficient iγ tells the degree of difference in consumption of
borrowers from non-borrowers.
A test of the null hypothesis that the consumption by borrowers is the same as non-
borrowers (that is, 0=iγ ) can easily be made by running regression of Equation 4.25
in the usual Ordinary Least Squares (OLS) manner and finding out whether or not
the basis of the t-statistics and the computed coefficient is statistically significant. In
a similar way, a null hypothesis can be developed for the district dummies to tell
how far Gazipur and Dinajpur are different from the third district.
Equation 4.25 has been estimated using heteroscedasticity corrected OLS estimates
applying the method suggested by White corrected standard errors. To see the
direction as well as the magnitude of changes in consumption between different
income groups, the samples are divided into three groups according to their income
per capita. Samples are classified according to their income per capita instead of
incomes only because it is found that there are wide variations in incomes in the
sample.
118
4.5 Empirical Results
This section will provide empirical results with the interpretation of the estimates. A
brief discussion of the summary statistics is provided here to give an idea of average
consumption of the items in each district.
4.5.1 Percentage and Mean Consumption of the Items by Districts
Before estimating the above mentioned Equation 4.25 it is worthwhile to look at the
average consumption of various food and non-food items in all three districts for
both the borrowers and non-borrowers. Table 4.1 shows the average consumption of
food and non-food items for both borrowers and non-borrowers in all three districts.
The percentage consumption of each item is shown in parentheses. Percentage
consumption of each item is calculated by dividing the mean consumption by the
total expenditure multiplied by hundred.
According to Table 4.1, the percentage consumption of staple food such as cereals,
pulses and vegetables, of non-borrowers of all three districts are higher than
borrowers. On the other hand, percentage consumption of protein items such as meat
and fish and milk and eggs are higher for borrowers in all districts. Percentage
consumption of the total non-food items is higher for borrowers and percentage
consumption of the total food items is higher for non-borrowers in Gazipur and
Dinajpur district but not in Chokoria district. Cigarette consumption of non-
borrowers is also higher in all three districts. Two conclusions may be derived from
119
the above discussion. Firstly, borrowers spend more of their budget on protein than
on cereals and secondly, non-borrowers spend more on food than on non-food items.
While looking at the non-food data, Table 4.1 shows fuel consumption is higher for
borrowers in Gazipur and Dinajpur districts; electricity consumption is higher for
non-borrowers in all districts. Again the clothing consumption is higher for the
borrowers of Gazipur and Dinajpur but not in Chokoria district.
Expenditure on education is higher for the borrowers in all three districts but
expenditure on health does not reveal a pattern. Finally by looking at the total non-
food items, it may be concluded that the spending on non-food items is higher for the
borrowers in Gazipur and Dinajpur districts and cumulatively in all districts for the
borrowers.
Table 4.1a shows the t-test results of different items consumed by borrowers and
non-borrowers. T-test assesses whether the means of treatment group (borrowers) are
statistically different from the control group (non-borrowers). Our t-test results show
that there are significant differences at 1% level in all food items except for
cigarettes between borrowers and non-borrowers. There are significant differences
between borrowers and non-borrowers in terms of all non-food items as well but for
health, education and electricity the difference is significant at 5% level only.
120
Table 4.1 Average Consumption of Borrowers and Non-Borrowers of Three
Districts
(Figures are in taka)
Gazipur Dinajpur Chokoria Total
B NB B NB B NB B NB
Cereal
835.80 (13.68)
743.02 (16.07)
874.78 (19.91)
919.65 (22.99)
725.59 (15.11)
716.12 (16.93)
812.05 (16.23)
792.93 (18.66)
Pulses 196.74 (3.22)
195.55 (4.23)
191.78 (4.37)
199.53 (4.98)
197.99 (4.12)
190.16 (4.49)
95.50 (3.90)
195.08 (4.57)
Vegetables 559.13 (9.15)
509.77 (11.02)
520.84 (11.85)
503.01 (12.57)
450.00 (9.37)
390.29 (9.23)
510 (10.12)
467.69 (10.94)
Meat & Fish
1317.34 (21.56)
732.35 (15.84)
1003.43 (22.84)
848.65 (21.21)
1166.85 (24.30)
837.04 (19.80)
1162.54 (22.9)
806.01 (18.95)
Milk & Eggs
293.81 (4.81)
198.72 (4.29)
174.54 (3.97)
152.47 (3.81)
214.7 (4.47)
209.29 (4.95)
227.8 (4.42)
186.82 (4.35)
Cigarette 245.40 (4.02)
239.15 (5.17)
130.91 (2.98)
138.25 (3.45)
197.42 (4.11)
173.58 (4.10)
191.24 (3.70)
183.66 (4.24)
Cooking Oil
202.65 (3.31)
182.96 (3.95)
165.07 (3.75)
172.81 (4.32)
168.39 (3.50)
148.09 (3.50)
178.0 (3.52)
167.95 (3.92)
Spices 228.19 (3.73)
205.45 (4.44)
169.44 (3.85)
153.33 (3.83)
160.88 (3.35)
139.25 (3.29)
186.17 (3.64)
166.01 (3.85)
Sugar 119.49 (1.95)
91.22 (1.97)
49.71 (1.13)
43.39 (1.08)
53.86 (1.12)
42.74 (1.01)
74.35 (1.4)
69.11 (1.35)
Total Food 3998.56 (65.45)
3098.21 (67.02)
3280.53 (74.69)
3131.12 (78.28)
3335.61 (69.48)
2846.69 (67.33)
3538.23 (69.87)
3025.34 (70.87)
Fuel 175.96 (2.88)
131.54 (2.85)
134.23 (3.05)
111.80 (2.79)
168.61 (3.51)
179.61 (4.24)
159.6 (3.15)
140.98 (3.29)
Electricity 194.74 (3.19)
167.26 (3.61)
63.34 (1.44)
62.69 (1.56)
131.00 (2.72)
117.53 (2.78)
129.69 (2.45)
115.82 (2.65)
Clothing 225.64 (3.69)
129.81 (2.80)
177.40 (4.03)
149.52 (3.78)
181.65 (3.78)
175.67 (4.15)
194.90 (3.83)
151.66 (3.57)
Education 250.74 (4.10)
97.55 (2.11)
92.92 (2.11)
55.55 (1.38)
141.53 (2.95)
99.70 (2.35)
161.73 (3.05)
84.266 (1.95)
Health 247.31 (4.05)
187.79 (4.06)
153.71 (3.50)
131.58 (3.29)
205.62 (4.28)
186.04 (4.40)
202.21 (3.94)
168.47 (3.92)
Travel 574.80 (9.40)
485.47 (10.5)
323.48 (7.36)
232.53 (5.81)
373.84 (7.78)
364.17 (8.61)
424.04 (8.18)
360.72 (8.3)
Entertain
ment
9.90 (0.16)
5.56 (0.12)
2.72 (0.05)
13.49 (0.33)
0 0 12.62 (.07)
19.05 (0.15)
Other 431.53 (7.06)
319.15 (6.9)
164.24 (3.73)
111.26 (2.78)
262.76 (5.47)
258.2 (6.1)
286.17 (5.42)
229.53 (5.26)
Total Non-
food 2110.64 (34.55)
1524.25 (32.98
1111.57 (25.31)
868.47 (21.72)
1465.03 (30.52)
1380.95 (32.67)
1562.41 (30.12)
1257.89 (29.12)
TOTAL
6109.2 (100)
4622.46 (100)
4392.1 (100)
3999.59 (100)
4800.64 (100)
4227.64 (100)
5100.64 (100)
4283.23 (100)
Note: B and NB stands for borrowers and non-borrowers respectively.
The figures in parentheses show the percentage of consumption in terms of total expenditure.
121
Table 4.1a T-test Results of Mean Expenditure and Budget Share of Different
items Consumed by Borrowers and Non-Borrowers
Cereal 3.85***
(0.00)
Fuel 1.16***
(0.00)
Pulses 4.29***
(0.00)
Electricity 1.61**
(0.05)
Vegetable 2.97***
(0.00)
Clothing 5.63***
(0.00)
Meat & Fish 11.17***
(0.00)
Education 2.10**
(0.02)
Milk & Eggs 4.67***
(0.00)
Health 1.67**
(0.04)
Cigarette 0.35
(0.3 6)
Travel 2.75***
(0.00)
Oil 4.65***
(0.00)
Others 2.28***
(0.01)
Spices 2.53***
(0.00)
Total Non-Food 2.29***
(0.01)
Sugar 8.30***
(0.00)
Total Food 3.39***
(0.00)
(***) significant at 1%level (**) significant at 5%level (*) significant at 10%level and the p-values are shown in the parentheses.
4.5.2 Preliminary Findings from the Descriptive Table
The descriptive table is provided in Appendix F which gives the mean, median,
maximum, minimum, standard deviation, normality, skewness and kurtosis of all the
items consumed by the borrowers and non-borrowers and also their family size,
family income, total expenditure and total production of all three districts. Tables are
122
provided separately for the borrowers and non-borrowers and also for both groups as
a whole.
Average cereal consumption is highest in Dinajpur than for the two other districts.
The standard deviation of cereal consumption in Dinajpur is also the highest. In
Gazipur and Chokoria the average consumption of cereal among borrowers are
higher than non-borrowers. The average pulses consumption is almost same in all
three districts. Vegetable consumption is lowest in Chokoria than in the two other
districts.
Average protein consumption is observed as higher for borrowers than non-
borrowers, compared to the average cereal consumption. The standard deviations for
protein items for borrowers in all three districts are higher than the standard
deviation of cereal. It may be concluded here that borrowers are better off in terms of
protein consumption than non-borrowers in all three districts. When comparing all
the districts, the households (both borrowers and non-borrowers) of Gazipur
appeared to have highest average consumption of protein than the other two districts.
Since Gazipur is close to the capital city Dhaka, people may have more opportunities
to make better earnings to have better consumption. It may also be due to better
transportation around the capital that creates a better availability of food.
By looking at the mean value of total food and comparing with total non-food, it
appears that the mean consumption of total food always exceeds the mean of total
non-food in all districts for both borrowers and non-borrowers. In all three districts it
is observed that the average food and non-food consumption is higher for borrowers
123
than non-borrowers. Among the districts, Gazipur is better off in terms of
consumption for both food and non-food items.
From Table 4.1 it is clear that the average income is higher for borrowers in all three
districts. While comparing the districts it appears that the average income is higher
for both borrowers and non-borrowers in Gazipur district. In terms of income,
Chokoria lies in the second position followed by Dinajpur. Even though the average
income of both borrowers and non-borrowers of Gazipur are higher than other two
districts, the dispersion between the incomes of the two groups is also higher in that
district which can be revealed by the higher standard deviation.
In terms of average family size Gazipur has the lowest family size followed by
Chokoria and then Dinajpur. However, the ranking of average total production is
Gazipur (highest), Dinajpur and then Chokoria.
The people of Gazipur spend more on education than the people of the other two
districts. The ranking from high to low is Gazipur, Chokoria and Dinajpur. Again
among borrowers and non-borrowers of all districts it appears that borrowers spend
more on education than non-borrowers in all districts. The same conclusion may be
derived for the consumption on health and clothing in all three districts. It is
surprising to notice that even consumption on cigarettes, betel leaf and entertainment
are also found higher in Gazipur than in the other two districts. However, overall
consumption on entertainment is very poor in all three districts. (Entertainment is
considered here as going to the cinemas and going to theatres).
124
It is remarkable that the total production of non-borrowers of Dinajpur and Chokoria
are higher than borrowers. In contrast, the total production of borrowers in Gazipur
is higher than non-borrowers. However, as with consumption total production is also
highest in Gazipur compared to the other districts.
The result of the Jarque-Bera test shows that the items consumed by both the groups
are not normally distributed in all districts. The concept of skewness and kurtosis of
a probability density function (PDF) tell us something about the shape of the PDF.
Skewness is a measure of asymmetry and kurtosis is a measure of tallness or flatness
of a PDF. In most of the cases items consumed are positively skewed, some are low
in value but very few were found negatively skewed among both borrowers and non-
borrowers. On the other hand, the value of kurtosis is found to be over three in most
of the cases.
It is interesting to see that, in all three districts among both borrowers and non-
borrowers there are some people who do not spend any money on items such as
education, health, travel and entertainment as well as milk and eggs, sugar and
cigarettes. This null expenditure on items may be explained separately. Since
primary education is free in rural Bangladesh some people may not spend money on
education (that is tutorial fee). Some people may be too poor to afford to spend on
items such as health, travel and entertainment. The same logic may apply for some of
the food items as well. The non-expenditure may also be possible due to the fact that,
on that particular month when the survey was conducted, some people may not have
spent any money on that item (the problem of purchase infrequency). While
estimating the model the null expenditure has been taken care of by deleting the
125
entire observation from the estimation procedure for each item. We have not
estimated the equations for travel and entertainment as the sample size become too
small after omitting the null expenditure.
4.5.3 Estimation Results and Discussion
The AIDS model as specified in Equation 4.25 has been estimated by OLS in the
absence of cross-equation and non-linear restrictions. In the absence of price data42,
we assume that all the households experience the same price. However, when prices
are constant, the demand of goods is dependent only on the income of the individual.
In that case, the only condition to be met is the adding-up restriction since all the
other restrictions are derived from consideration of prices (Beneito, 2003). It is worth
mentioning that in this chapter we have considered adding-up restriction by
estimating (n-1) number of equations.
The heteroscedasticity corrected OLS estimates are then obtained using the STATA
8.0 statistical package by applying White corrected Standard errors which is
particularly suitable for cross-section data of family budgets. Twelve items that are
commonly consumed in rural areas are used for the purpose of estimation.
The households are divided into three groups according to their per capita incomes to
see the sign as well as the magnitude of difference in consumption between different
groups. As income increases, it is expected that the consumption of staple food item
(necessary good) will decrease. By dividing the samples into different income groups
42 Since we use cross-section data over a period of time it is assumed that the price of the items has not changed during that period among the districts.
126
we wanted to see the magnitude of changes in consumption between groups. The
reason for grouping the sample according to per capita income instead of income is
due to the fact that there is wide variation in income among the households. Samples
are divided into three equal groups each containing 33.33% of total size and termed
as low, medium and high income group.
Table 4.2 shows OLS estimates of the AIDS model as represented by Equation 4.25.
White corrected standard errors are used to correct for heteroscedasticity.
The t-statistics are shown in the parentheses.
127
Table 4.2 OLS Estimation of the AIDS Model for All Districts.
Constant Log of Total
Expenditure
per capita
(LNTEPC)
Family
Size
Borrower
Dummy
District
Dummy
1
District
Dummy
2
2R No.
of
Obs
Cereals Low income per capita
1.25*** (15.93)
-0.16*** (-13.90)
0.0002 (0.11)
-0.001 (-0.26)
0.04*** (4.00)
0.06*** (8.33)
0.34 190
Medium income per capita
1.08*** (15.47)
-0.13*** (-13.07)
-0.004***
(-2.81)
0.002 (0.49)
0.03*** (6.13)
0.04*** (9.01)
0.35 190
High income per capita
0.80*** (19.21)
-0.09*** (-16.05)
-0.001 (-1.35)
-0.013** (-2.31)
0.02*** (4.49)
0.02*** (4.24)
0.31 191
Pulses Low income per capita
0.21*** (6.06)
-0.02*** (-4.43)
-0.0007 (-1.31)
-0.001 (-0.60)
0.002 (0.54)
0.008*** (2.93)
0.17 189
Medium income per capita
0.15*** (4.00)
-0.02*** (-2.67)
-0.0009 (-1.38)
-0.002 (-0.73)
-0.007** (-2.49)
-0.0002 (-0.11)
0.10 183
High income per capita
0.16*** (9.27)
-0.02*** (-6.53)
-0.001***
(-2.67)
-0.008*** (-3.10)
-0.0002 (-0.12)
0.003 (1.22)
0.29 195
Vegetables Low income per capita
0.59*** (8.98)
-0.07*** (-6.53)
-0.006***
(-5.12)
-0.001 (-0.29)
0.02 (0.10)
0.03*** (6.83)
0.37 190
Medium income per capita
0.42*** (6.51)
-0.04*** (-4.37)
-0.007***
(-6.32)
0.0008 (0.25)
0.01*** (2.61)
0.03*** (7.18)
0.36 190
High income per capita
0.49*** (18.71)
-0.05*** (-14.92)
-0.005***
(-7.80)
0.004 (1.46)
0.02*** (5.29)
0.02*** (4.09)
0.33 191
Sugar Low income per capita
-0.07*** (-2.92)
0.01*** (3.63)
0.002*** (4.80)
0.009*** (4.51)
-0.009***
(-2.69)
-0.01*** (-4.91)
0.32 184
Medium income per capita
-0.05 (-1.10)
0.01 (1.61)
0.002*** (3.53)
0.01*** (3.66)
-0.003 (-0.96)
-0.01*** (-3.67)
0.20 184
High income per capita
0.05* (1.75)
-0.002 (-0.54)
0.0003 (0.49)
0.01*** (3.96)
-0.004 (-1.54)
-0.01*** (-3.70)
0.133 183
Protein (Meat and Fish) Low income per capita
-0.29*** (-3.63)
0.05*** (4.55)
0.006*** (4.48)
0.03*** (6.06)
-0.009 (-1.09)
-0.008 (-1.50)
0.32 189
Medium income per capita
-0.39*** (-3.85)
0.06*** (4.40)
0.02*** (6.05)
0.03*** (5.63)
0.00007 (0.01)
0.005 (0.95)
0.31 191
High income per capita
-0.17** (-2.31)
0.04*** (3.61)
0.01*** (4.68)
0.05*** (4.77)
0.009 (1.29)
-0.004 (-0.42)
0.34 190
128
Table4.2 (cont’d) Constant Log of Total
Expenditure
Per Capita
(LNTEPC)
Family
Size
Borrower
Dummy
District
Dummy
1
District
Dummy
2
2R No.
of
Obs
Protein (Milk and Eggs) Low income per capita
-0.03 (1.05)
0.01*** (2.84)
-0.0001 (-0.31)
-0.002 (-0.92)
-0.02*** (-4.98)
-0.009***
(-3.28)
0.15 189
Medium income per capita
-0.009 (-0.23)
0.006 (0.99)
0.003*** (3.90)
0.0007 (0.30)
-0.007 (-1.99)
-0.006** (-2.06)
0.14 183
High income per capita
-0.07** (-2.04)
0.02*** (3.24)
0.002*** (3.25)
0.007 (0.30)
-0.007** (-1.99)
-0.006** (-2.06)
0.14 183
Cigarettes Low income per capita
0.07 (1.64)
0.0002 (0.04)
-0.003*** (-4.71)
-0.008** (-2.42)
-0.002 (-0.38)
0.011*** (3.46)
0.19 166
Medium income per capita
0.03 (0.69)
0.002 (0.24)
0.0009 (0.98)
0.0008 (0.31)
0.006** (2.16)
0.0009 (0.23)
0.02 166
High income per capita
0.16*** (4.58)
-0.01*** (-3.03)
-0.002** (-2.21)
0.006 (1.49)
0.004 (1.32)
-0.008** (-2.47)
0.11 165
Total Food Low income per capita
1.73*** (16.91)
-0.16*** (-10.08)
-0.004** (-2.13)
0.02** (2.08)
0.009 (0.75)
0.08*** (8.03)
0.30 190
Medium income per capita
1.59*** (9.42)
-0.14*** (-5.87)
0.005 (1.57)
0.03*** (3.00)
0.02** (2.20)
0.06*** (5.21)
0.31 190
High income per capita
1.65*** (11.60)
-0.14*** (-7.46)
0.003 (0.91)
0.05*** (3.15)
0.05*** (3.74)
0.01 (0.64)
0.29 191
Fuel Low income per capita
0.18*** (3.36)
-0.02** (-2.07)
-0.004*** (-4.21)
0.001 (0.28)
-0.03*** (-4.81)
-0.02 (-4.60)
0.24 190
Medium income per capita
0.16*** (3.93)
-0.02*** (-2.72)
-0.002** (-2.42)
-0.006*** (-2.72)
-0.01*** (-4.42)
-0.008***
(-3.06)
0.19 190
High income per capita
0.10*** (3.25)
-0.008* (-1.90)
-0.002*** (-2.96)
0.003 (1.15)
-0.01*** (-4.57)
-0.009***
(-2.98)
0.21 191
Electricity Low income per capita
0.25*** (8.97)
-0.03*** (-6.90)
-0.003*** (-6.22)
-0.004* (-1.71)
-0.001 (-0.30)
-0.008***
(-3.35)
0.35 134
Medium income per capita
0.17*** (4.15)
-0.02*** (-2.77)
-0.003*** (-4.36)
0.0005 (0.26)
-0.003 (-1.26)
-0.003 (-0.84)
0.23 134
High income per capita
0.13*** (5.38)
-0.01*** (-3.59)
-0.002*** (-5.43)
-0.004** (-2.07)
-0.0002 (-0.14)
-0.002 (-0.97)
0.23 133
129
Table 4.2 (cont’d) Constant Log of Total
Expenditure
Per Capita
(LNTEPC)
Family
Size
Borrower
Dummy
District
Dummy
1
District
Dummy
2
2R No.
of
Obs
Education Low income per capita
0.07 (1.33)
0.001 (0.14)
-0.004***
(-3.69)
-0.005 (-1.24)
-0.02** (-2.22)
-0.01 (-3.31)
0.18 94
Medium income per capita
0.02 (0.34)
0.006 (0.67)
-0.003***
(-2.83)
-0.003 (-0.66)
-0.005 (-0.99)
0.0002 (0.06)
0.06 94
High income per capita
-0.15** (-2.26)
0.02*** (2.98)
-0.001 (-1.65)
0.003 (0.44)
0.005 (0.90)
0.01* (1.84)
0.14 94
Cloth Low income per capita
-0.01 (-0.52)
0.006 (1.66)
0.0002 (0.58)
0.002 (1.59)
-0.004 (-1.44)
0.004 92.35)
0.07 190
Medium income per capita
0.06*** (2.84)
-0.005 (-1.46)
0.0004 (0.71)
0.003** (2.30)
-0.005*** (-3.12)
0.004 (2.17)
0.20 190
High income per capita
0.05** (2.16)
-0.001 (-0.43)
-0.0001 (-0.29)
0.004 (1.71)
-0.002 (-1.35)
0.008** (2.16)
0.14 191
Health Low income per capita
-0.02 (-0.74)
0.01** (2.22)
0.0003 (0.67)
-0.003 (-1.27)
0.005 (0.66)
-0.006** (-2.31)
0.11 185
Medium income per capita
0.0009 (0.02)
0.006 (0.70)
-0.0007 (-0.82)
0.003 (0.71)
-0.004 (-0.78)
-0.007** (-2.59)
0.02 185
High income per capita
0.09** (2.47)
-0.003 (-0.79)
-0.003***
(-3.83)
-0.002 (-0.48)
-0.005 (-1.31)
-0.002 (-0.51)
0.08 186
Total non-Food Low income per capita
-0.73*** (-7.16)
0.16*** (10.08)
0.004** (2.13)
-0.02** (-2.08)
-0.009 (-0.75)
-0.08*** (-8.03)
0.44 190
Medium income per capita
-0.59*** (-3.51)
0.14*** (5.87)
-0.005 (-1.57)
-0.03*** (-3.00)
-0.03** (-2.20)
-0.05*** (-5.21)
0.31 190
High income per capita
-0.65*** (-4.58)
0.14*** (7.46)
-0.003 (-0.91)
-0.05*** (-3.15)
-0.05*** (-3.74)
-0.01 (-0.64)
0.29 191
(***) significant at 1%level (**) significant at 5%level (*) significant at 10%level and the t-statistics are shown in the parentheses.
130
The R 2 values suggest that the model fits better for items such as cereals, vegetables,
total food and total non-food. Low R 2 is a common observation for the consumer
demand behaviour while analysing survey data. The reason may partly be the
absence of some factors in the model, which are unobservable and difficult to
include. A negative value of expenditure coefficient for any item indicates a
necessity and a positive value indicates a non-necessity item.
From Table 4.2 cereals consumption expenditure is found significant and it is a
necessity item for all income groups, as the sign of the coefficient is found negative.
This implies that consumption of cereal is more responsive to changes in income for
all income groups. A small change in income will cause bigger change in cereal
consumption for lower income group people than middle and higher income group
people.
Family size is found statistically significant for middle income group. Borrower
dummy is found statistically significant for higher income group, which implies that
the higher income group borrowers are better off in terms of cereal consumption. We
therefore reject the null hypothesis that borrowers’ consumption is same as for non-
borrowers. District dummy is also found significant which implies that there are
variations among districts in cereal consumption; we therefore reject the null
hypothesis that the consumption is same across districts.
Pulses and vegetables are necessity for all income groups. The coefficient of pulses
and vegetables are also found significant for all income groups. Sugar is found to be
a necessity only for the higher income groups. Family size is found significant for
pulses consumption only for the higher income group, while, family size is found
131
significant for vegetable consumption for all income groups. In case of sugar
consumption, family size is significant only for low and middle income group
people. As family size increases sugar consumption will also increase for low and
medium income group people. Borrower dummy is found significant for pulses
consumption for the higher income group. This implies that the higher income group
borrowers are better off in terms of pulses consumption. Borrower dummy is found
significant for all income groups in case of sugar consumption. Significant district
dummy shows that there is a variation in consumption of vegetables, pulses and
sugar between districts.
Protein consumption is divided into two categories. Beef, mutton, chicken and fish
of all types are estimated separately than milk and eggs. For all income groups’
protein is found to be a significant and non-necessary item. Protein consumption is
found responsive to changes in income for all income groups. For the higher income
group bigger changes in income will cause smaller changes in protein consumption.
This implies the higher income group is more saturated in protein consumption. In
terms of protein consumption family size is found significant and positive. This
implies that as family size increases, protein consumption will also increase. District
dummy is found significant in case of milk and eggs consumption across different
income levels.
Borrower dummy43 is found significant for all income groups in case of meat and
fish consumption. This implies that there is a difference in consumption of meat and
fish between borrowers and non-borrowers. Protein (meat and fish) is considered as
43 The coefficient of the variable (borrower dummy) may be little higher in the absence of Heckman’s two-stage procedure.
132
an expensive food item compared to other food items in rural Bangladesh. Our
results suggest that the microcredit borrowers are better off in terms of protein
consumption. From these results we may conclude that the microcredit programs are
doing well in providing better consumption for borrowers. We have emphasised this
point since we found significant borrower dummy for all income groups. If borrower
dummy was significant, only for the higher income group borrowers, a question may
arise regarding whether only higher income group people are microcredit borrowers.
Cigarettes, betel leaf and betel nuts are found significant and necessary for higher
income group people. Demand for total food is found significant in all income
groups and more responsive to changes in income. For the lower-income group it
decreases at a higher rate than the higher-income group. Significant district dummy
shows there is a difference in total food consumption across districts.
Borrower dummy is found significant for all income groups for total food
consumption. This implies that there is a difference in total food consumption
between borrowers and non-borrowers. This further implies that borrowers are better
off in terms of total food consumption. This result reinforces our previous suggestion
that the microcredit programs are successful in bringing better consumption for
borrowers.
Fuel and electricity are found significant and necessity items. Family size is
significant for both fuel and electricity consumption. District dummies are found
significant for fuel consumption but not for electricity. In case of fuel consumption
borrower dummy is significant only for the middle income group people while the
133
same dummy is significant for low and high income group people in case of
electricity consumption.
Education is found non-necessary but significant for higher income group people.
This may be due to free primary education in rural Bangladesh. Clothing is found
necessary for middle and higher income group people, but not for the poor people.
On the other hand expenditure on health is found to be a necessary item only for the
higher income group but not necessary for the lower and middle income group.
Total non-food items are found non-necessary for all income group people. Borrower
dummy is found significant only for the total non-food item but not for education,
clothing or health (that compose total non-food). This implies there is no difference
between borrowers and non-borrowers in terms of consumption on education,
clothing and health. The significant coefficient of total non-food item for borrowers
may be due to more spending on electricity by borrowers.
On the other hand, district dummy is found significant for items such as clothing,
education at the 10% and 5% levels respectively, and health at the 5% level. The
district dummy is found significant for total non-food item implies that there is
difference in consumption of these items among districts.
Budget share on food items especially meat, fish and eggs are found significant
which is consistent with Deaton’s result based on British data and Ray’s result based
on Indian data on rural areas. The absence of a significant coefficient for clothing,
contrasts with Deaton and Ray both.
134
To examine the robustness of these results an alternative model has been estimated
here where the samples are not divided into groups but income dummies are used
instead.
The model is as follows:
12.......2,1
)26.4(loglog 2121
=∀++++++++=
i
iiiiiiii DDDDDDBDfxB εχπγδβα
where D 1 = 1 for the low income group and zero otherwise,
D 2 = 1 for the middle income group and zero otherwise.
Similar results are obtained from this model so we are not reporting the results here.
4.5.4 Test for Significant Difference between Borrowers and Non-borrowers
The model that has been estimated in the previous section assumes that the
qualitative variables (dummy variables) affect the intercept but not the slope
coefficient of the various subgroup regressions. But what if the slopes are also
different? If the slopes are in fact different, testing for differences in the intercepts
may be of little practical significance. Therefore, it is necessary to develop a general
methodology to find out whether two regressions are different, where the difference
may be in the intercepts or the slopes or both.
In so doing, the following equation is estimated:
12.......2,1
)27.4(*log 21
=∀+++++++=
i
iiiiiiiii DDDDBDXBDfXB εχπλγδβα
135
To see the implications of model and assuming that 0)( =iuE , we obtain the
following from Equation 4.27:
XXBDBE i 11),0|( βα +==
XXBDBE i )()()1|( 1111 λβγα +++==
These are respectively the mean consumption functions for the regression model.
The intercept coefficient as well as the slope coefficient is different as the dummy
takes the value of zero and one. It is worth mentioning that the introduction of the
dummy variable BD in the multiplicative form (D multiplied by X) enables us to
differentiate between the slope coefficient of borrowers and non-borrowers, just as
the introduction of the dummy variable in the additive form enables us to distinguish
between the intercepts of the two groups.
136
Table 4.3: Testing for Differential Slope Coefficient
(***) significant at 1%level, (**) significant at 5%level, (*) significant at 10%level. The t-statistics are shown in the parentheses and probabilities are shown in the square brackets.
Table 4.3 shows the OLS estimation using the differential slope coefficient. It has
also been tested whether the coefficient of the borrower dummy and the
multiplicative term are jointly zero through coefficient tests of Wald coefficient
restrictions.
Constant
Log of Total
Expenditure
Per capita
(LNTEPC)
Family
Size
Borrower
Dummy
(BD)
LNTEPC
*BD
District
Dummy
1
District
Dummy
2
Test of
Restriction
Vector
0== ii λγ
(F-statistics)
2R
Cereal 1.10*** (21.38)
-0.13*** (-17.59)
-0.02*** (-2.85)
-0.08 (-1..62)
0.01* (1.65)
0.03*** (7.80)
0.05*** (11.12)
1.41 [0.24]
0.30
Pulses 0.16*** (7.38)
-0.02*** (-4.97)
-0.004** (-2.15)
0.03 (1.25)
-0.004 (-1.39)
-0.001 (-0.88)
0.004*** (2.72)
3.98 [0.02]
0.38
Vegetables 0.60*** (16.38)
-0.06*** (-12.91)
-0.035*** (-9.35)
-0.03 (-0.85)
0.004 (0.91)
0.02*** (6.26)
0.03*** (10..38)
0.89 [0.41]
0.3 4
Sugar -0.08*** (-5.18)
0.02*** (6.78)
0.01*** (4.70)
0.03* (1.86)
-0.003 (-1.21)
-0.005*** (-2.85)
-0.01*** (-8.01)
32.48*** [0.00]
0.33
Protein (meat and
fish)
-0.48*** (-9.33)
0.07*** (10.44)
0.06*** (9.89)
-0.01 (-0.29)
0.007 (1.00)
0.001 (0.35)
-0.0002 (-0.07)
55.053*** [0.00]
0.39
Protein (milk and
eggs)
-0.03 (-1.52)
0.01*** (3.67)
0.009*** (3.43)
-0.08*** (-3.81)
0.01*** (4.01)
-0.007*** (-3.59)
-0.007*** (-3.75)
11.23*** [0.00]
0.26
Cigarettes 0.11*** (3.77)
-0.006 (-1.58)
-0.009*** (-3.03)
-0.05* (-1.86)
0.008* (1.87)
0.006*** (2.66)
0.003* (1.68)
1.76 [0.17]
0.05
Total food 1.53*** (16.64)
-0.13*** (-9.56)
-0.002 (-0.23)
-0.18** (-1.98)
0.03** (2.39)
0.04*** (4.57)
0.05*** (8.95)
22.60*** (0.00)
0.45
Fuel 0.18*** (7.18)
-0.01*** (-4.62)
-0.01*** (-6.60)
-0.03* (-1.76)
0.005* (1.82)
-0.01*** (-8.12)
-0.01*** (-5.92)
1.94 [0.14]
0.23
Electricity 0.21*** (9.99)
-0.02*** (-7.58)
-0.02*** (-8.65)
-0.02 (-1.16)
0.003 (1.17)
-0.002* (-1.81)
-0.005*** (-3.09)
0.69 [0.50]
0.38
Education 0.09** (2.37)
-0.004 (-0.72)
-0.01*** (-3.49)
-0.11*** (-2.74)
0.02*** (2.67)
-0.003 (-0.94)
-0.001 (-0.50)
4.22** [0.02]
0.08
Cloth -0.001 (-0.08)
0.005*** (2.61)
-0.0002 (-0.17)
-0.01 (-0.78)
0.002 (1.10)
-0.005*** (-4.33)
0.004*** (4.36)
11.04*** [0.00]
0.15
Health 0.05** (2.16)
-0.0001 (-0.02)
-0.007*** (-2.65)
0.003 (0.14)
-0.0009 (-0.22)
-0.004* (-1.88)
-0.007*** (-4.44)
0.79 [0.45]
0.04
Total
non-food
-0.53*** (-5.80)
0.13*** (9.56)
0.002 (0.23)
0.18** (1.98)
-0.03** (-2.39)
-0.04*** (-4.57)
-0.05*** (-8.95)
22.60*** [0.00]
0.35
137
From Table 4.3 it is observed that the intercept coefficient for all food items except
for milk and eggs and all non-food items except for clothing are statistically
significant. By looking at the slope coefficient it is evident that items such as cereals,
protein (both types) fuel, education, total food and total non-food are statistically
significant. The income elasticity of items such as cereals, vegetables, sugar, fuel,
electricity, total food and non-food items for non-borrowers are found steeper than
borrowers, which implies small changes in income will cause a bigger change in
consumption for non-borrowers. This inelastic function implies that cereals,
vegetables, sugar, fuel, electricity, total food and non-food items are necessary items
for non-borrowers.
On the contrary, it is observed that items such as pulses, meat, fish, eggs, milk and
clothes reveals a more inelastic (steeper) function for borrowers. Therefore, it can be
concluded that for borrowers’ items such as pulses, meat, fish, eggs, milk and clothes
are necessary items. It is worth mentioning here that these are relatively expensive
items compared to other food and non-food items in rural Bangladesh. According to
our findings, once borrowers find such items (pulses, meat, fish, eggs, milk and
clothes) as necessary, it can be concluded that microcredit programs must be doing
well in providing better consumption for borrowers. Thus the findings from Table
4.3 are consistent with the results of Table 4.2.
From Table 4.3 it is observed that the intercepts as well as the slopes of the demand
functions are different for both food and non-food items. Therefore, the result
strongly indicates that there are significant difference between borrowers and non-
borrowers in terms of consumption of the items. For testing the values of the
138
coefficient whether jointly zero or not we develop the null hypothesis 0== ii λγ .
The coefficient test result shows statistically significant coefficient for items such as
sugar, protein, education, clothing, total food and total non-food. We therefore reject
the null hypothesis that the coefficients are jointly zero.
4.5.5 Testing for Uncorrelated Error Terms44
The model we have used in this chapter contains multiple equations. Since we used
budget share as the dependent variable, it would be unrealistic to expect that the
equation errors would be uncorrelated. A set of equations that has contemporaneous
cross- equation error correlation is called Seemingly Unrelated Regression Estimates
(SURE) system. The SURE model of Zellner, (1962) consists of m regression
equations, each of which satisfies the assumptions of the standard regression model.
It may appear that each single equation used in the model is unrelated but the
equations could be correlated through the errors terms since budget share is
considered in the model which contains multiple equations. We therefore, have
estimated Equation 4.25 again using Iterated Seemingly Unrelated Regression
estimates using the STATA 8.0 statistical package.
We have estimated 12 equations earlier covering food and non-food items. Some of
the equations are added together to reduce the number of equations to make it
suitable for the SURE estimation in the STATA statistical package. Twelve
44 A paper based on the results of this section has been published in the Conference Proceedings
(refereed), November 7, 2005, ISBN 0-7326-2283-2, Department of Management, Monash University Australia. Please refer to Appendix J for the abstract of the paper.
139
equations are reduced to seven by adding some items together (such as cereals with
pulses, protein of both types together, fuel with electricity, health with education,
and finally fruits with vegetables). Now, six equations are estimated instead of seven
to consider adding-up restriction. Table 4.4 provides the results from iterated SURE
analysis. The figures in parentheses show the Z-statistics at 95 % confidence
interval.
To be consistent with our OLS estimation, we also have added the multiplicative
dummy variable in the model and estimated Equation 4.27 using iterated SURE. The
results are provided in Table 4.5. The results of the SURE estimation are consistent
with our OLS estimation. Table 4.5 shows steeper income elasticities for items such
as protein and clothing for the borrowers. That implies that a large change in income
will bring smaller changes in consumption of protein and clothing for the borrowers.
This result also reinforces the conclusion that borrowers are better off in terms of
consumption of expensive items compared to non-borrowers.
We have calculated the income elasticity for the items consumed by households. The
results are derived from Table 4.5 and produced in Table 4.5a. We have found
income elasticity greater than one for items such as protein, clothing and health and
education. These results are justifiable as rural Bangladeshi people are quite poor. To
find out the difference in income elasticity between borrowers and non-borrowers it
is observed that, there is no difference between the groups in terms of consumption
of all items except for protein.
140
Finally, Table 4.6 provides a comparison between the results from OLS (Table 4.2)
with SURE (Table 4.4).
Table 4.4 Seemingly Unrelated Regression Estimates
Cereals
and Pulses
Protein Fruits and
Vegetables
Fuel and
Electricity
Clothing Health and
Education
Constant
1.18***
(44.89)
-0.57***
(-18.53)
0.55***
(32.63)
0.19***
(13.97)
-0.014*
(-1.92)
-0.07***
(-3.18)
Log of Total Expenditure Per Capita (LNTEPC)
-0.14***
(-38.44)
0.10***
(23.33)
-0.06***
(-26.00)
-0.02***
(-7.70)
0.007***
(6.75)
0.02***
(6.41)
Family Size -0.001*
(-1.83)
0.01***
(10.79)
-0.006***
(-10.06)
-0.003***
(-6.26)
0.00001
(0.06)
0.0006
(0.81)
Borrower dummy (BD)
-0.003
(-0.88)
0.04***
(9.94)
0.0002
(0.10)
-0.003
(-1.46)
0.004***
(4.34)
-0.002
(-0.85)
District
Dummy 1
0.03***
(7.01)
-0.007
(-1.29)
0.02***
(6.67)
-0.006***
(-2.83)
-0.005***
(-4.19)
-0.009**
(-2.43)
District
Dummy 2
0.05***
(12.7)
-0.007
(-1.51)
0.03***
(11.65)
-0.02***
(-11.77)
0.004***
(4.31)
-0.013***
(-3.79)
No of
Observations
571 571 571 571 571 571
R 2 0.31 0.30 0.34 0.27 0.15 0.11
Χ 2 2424.53 885.16 1058.14 210.30 106.97 73.17
(***) significant at 1% level, (**) significant at 5% level, (*) significant at 10% level and the Z-statistics are shown in the parentheses.
141
Table 4.5: Testing for Differential Slope Coefficient Using SURE Estimation
Cereals
and Pulses
Protein Fruits and
Vegetables
Fuel and
Electricity
Clothing Health and
Education
Constant
1.21*** (25.28)
-0.44*** (-7.99)
0.56*** (18.67)
0.19*** (7.79)
-0.001 (-0.12)
-0.04 (-0.97)
Log of Total Expenditure Per Capita (LNTEPC)
-0.15*** (21.02)
0.08*** (10.16)
-0.06*** (-14.55)
-0.01*** (-4.21)
0.005*** (2.63)
0.01*** (2.61)
Family Size -0.002* (-1.94)
0.011*** (10.15)
-0.006*** (-10.06)
-0.003*** (-6.18)
-0.00 (-0.15)
0.00 (0.62)
Borrower Dummy (BD)
-0.041 (-0.84)
-0.11* (-1.93)
-0.02 (-0.91)
-0.006 (-0.26)
-0.01 (-0.79)
-0.04 (-0.99)
District
Dummy 1
0.03*** (7.02)
-0.006 (-1.28)
0.01*** (11.56)
-0.006*** (-2.83)
-0.005*** (-4.18)
-0.009** (-2.42)
District
Dummy 2
0.05*** (12.62)
-0.008* (-1.71)
0.03*** (11.56)
-0.02*** (-11.75)
0.004*** (4.23)
-0.01*** (-3.85)
LNTEPC*BD 0.005 (0.77)
0.02*** (2.68)
0.004 (0.92)
0.0006 (0.16)
0.002 (1.12)
0.006 (0.93)
No of
Observations
571 571 571 571 571 571
R2
0.30 0.31 0.35 0.26 0.26 0.11
Χ 2 2427.67 903.41 1060.54 210.33 210.33
74.15
(***) significant at 1% level, (**) significant at 5%level, (*) significant at 10% level and the Z-statistics are shown in the parentheses.
Table 4.5a: Estimation of Income Elasticity for Borrowers and Non-borrowers
Cereals
and
Pulses
Protein Fruits and
Vegetables
Fuel and
Electricity
Clothing Health
and
Education
Income
Elasticities 0.85 1.08 0.94 0.99 1.005 1.01
Results are derived from Table 4.5.
142
Table 4.6: Compare and Contrast Between OLS and SURE Results
Ordinary Least Square
Estimation
Seemingly Unrelated
Regression Estimation
Cereal -As income increases consumption of cereal decreases, which makes cereal a necessary item. - - Borrower dummy is found significant for the high income group borrowers. This implies there is difference in cereal consumption between the borrowers and non-borrowers in the high income group. -Both the district dummies are found significant, which implies there is difference in consumption of cereal between districts.
-As income increases consumption of cereal decreases, which makes cereal a necessary item. - We found negative co-efficient for the family size. - Borrower dummy is not found significant, which implies there is no difference between the borrowers and non-borrowers in terms of cereal consumption. -Both the district dummies are found significant, which implies there is difference in consumption of cereal between districts.
Protein - -Co-efficient of family size is found positive and significant. -Borrower dummy is found significant for all income groups especially in case of meat and fish. -District dummy is found significant for milk and eggs.
-The co-efficient of protein is non-negative and significant which makes protein a non-necessary item. -Co-efficient of family size is found positive and significant. -Borrower dummy is found significant. -
Fruits and
Vegetables
-Vegetable is found necessary item -Family size is found significant. -There is no difference between vegetable consumption of borrowers and non-borrowers. -District dummies are found significant.
-Vegetable is found necessary item. -Family size is found significant. -There is no difference between vegetable consumption of borrowers and non-borrowers. -District dummies are found significant.
Fuel and -Fuel and electricity is found necessary item. -There is no difference
-Fuel and electricity is found necessary item. -There is no difference
143
electricity between borrowers and non-borrowers in terms of fuel consumption. -District dummies are found significant.
between borrowers and non-borrowers in terms of fuel consumption. -District dummies are found significant.
Clothing -Clothing is found necessary for middle and high-income group people. - -District dummy is only found significant high income group.
-Clothing is found non- necessary item. -Borrower dummy is found significant. -District dummy is found significant.
Health and
Education
-Health and education is a non-necessary item. - -Borrower dummy found not significant. -District dummies are found significant.
-Health and education is a non-necessary item. -As family size increases expenditure on health and education will also increase. -Borrower dummy found not significant. -District dummies are found significant.
4.6 Conclusion
This chapter has investigated the consumption behaviour of borrowers in the
microcredit programs compared to non-borrowers of the same category using
primary data collected through a structured questionnaire circulated in three major
districts of Bangladesh. In order to investigate the impact of per capita monthly
expenditure on budget share of different items consumed by borrowers and non-
borrowers AIDS (An almost Ideal demand System) model has been considered.
We have considered a model that consists of a borrower dummy as well as district
dummies to determine the difference between borrowers and non-borrowers and as
well as between districts. Twelve consumable items commonly used in rural
Bangladesh are used for the purpose of estimation. The samples are divided into
144
three equal groups according to their income per capita to determine the difference in
consumption between different income groups. A test of significant difference
between borrowers and non-borrowers is considered along with the coefficient test.
Finally to test for the uncorrelated error terms, Seemingly Unrelated Regression
Estimation (SURE) has also been used using the STATA 8.0 statistical package.
From the analysis it is observed that cereals, pulses, vegetables, fuel and electricity
are found to be necessity items for all income groups. Sugar, cigarettes, clothing and
health are found to be a necessity only for the high income group people but not for
the low or middle income group people. Meat, fish, milk and eggs are found
significant for all income groups which is consistent with Deaton’s result based on
British data, and Ray’s result based on Indian data. Total food is found statistically
significant for the borrowers and the coefficient becomes smaller with respect to the
higher income level.
It may be derived from the analysis that there is a difference in consumption between
different income group people. Consumption of cereals is more responsive to
changes in income for all income groups, but for the lower income group cereals
consumption falls at a faster rate with respect to income than for the middle and
higher income group. A small change in income will cause bigger change in cereals
consumption for the lower income group people than the middle and higher income
group people. Protein consumption is found responsive to changes in income for all
income groups. For the higher income group bigger changes in income will cause
smaller changes in protein consumption. This implies that the higher income group is
more saturated in protein consumption. The statistically significant coefficient of
145
total food for borrowers (which is also smaller with respect to higher income group)
implies that as income increases, the proportion of expenditure on total food
becomes smaller for the higher income group compared to the lower income group.
Similar results are obtained from total non-food items.
Borrower dummy is found significant in case of cereals, pulses and electricity for the
high income group. From this it may be concluded that borrowers are better off in
terms of food and non-food consumption and higher income borrowers are better off
in some items than the low income borrowers. Borrower dummy is also found
significant for sugar, protein, total food and total non-food for all income groups.
This result suggests that borrowers of all income levels are better off in terms of
consumption of such (relatively expensive) items. Our results further suggest that
microcredit programs are making it possible for borrowers to consume
comparatively expensive food and non-food items and have a better quality of life.
In most of the food items except for meat and fish district dummies are found
significant. Significant coefficients are found in case of all non-food items except for
education, health and clothing. This suggests that there is difference in consumption
in terms of most of the items between different districts.
From the results of the differential slope coefficient test, it is found from the inelastic
demand curve that items such as cereals, vegetables, sugar, fuel, electricity, total
food and non-food are necessary for non-borrowers. On the other hand, for
borrowers, items such as pulses, meat, fish, eggs, milk and cloth are found necessary,
which is demonstrated from the inelastic demand curve of the items.
146
From the result of the joint coefficient test, it is evident that there is a difference in
consumption between borrowers and non-borrowers. Borrowers as a whole are better
off in terms of consumption of all food items and some non-food items such as fuel
and electricity. Interestingly, higher income group borrowers are found statistically
significant and better off in terms of consumption of few food items and total non-
food items.
The results of the SURE estimation are not different from the results of the OLS
estimation. We found that there is no difference between borrowers and non-
borrowers in terms of cereals (staple food) consumption. In terms of protein
consumption borrowers are better off than the non-borrowers. Borrowers consider
protein as a necessary item whereas non-borrowers consider protein as a non-
necessary item. This is a significant difference between borrowers and non-
borrowers.
In terms of vegetables, fuel and electricity consumption there is no difference
between borrowers and non-borrowers while in terms of clothing consumption it is
found that borrowers are better off. In terms of expenditure on health and education,
we found no difference between borrowers and non-borrowers but there is difference
between districts in terms of expenditure on education and health.
In summary we may conclude that borrowers are better off in terms of consumption
of most of the food and non-food items compared to non-borrowers. Borrowers of all
income levels are better off in consumption of expensive food items. These results
suggest that microcredit programs are successful in bringing better consumption for
147
borrowers. The programs are able to make borrowers afford relatively expensive
food and non-food items. Our results further suggest that microcredit programs are
doing well enough to produce better quality of life for borrowers by providing better
consumption. Since borrowers of all income levels are demonstrating better
consumption on most of the expensive food items, show the positive impact of
microcredit programs on consumption. We can not conclude that the higher income
borrowers are better off as we have found significant coefficient of two items
(cereals and pulses) only for that group. Apart from credit, we do not have enough
information to identify other factors that may have impacted better consumption on
borrowers.
148
Chapter Five
FACTORS AFFECTING WOMEN’S EMPOWERMENT IN
MICROCREDIT PROGRAMS
5.1 Introduction
According to a study conducted by the United Nations (2004), more than one billion
people around the world live below the poverty line45. It is argued that women and
children of developing countries represent most of these people. This study further
argues that women’s empowerment is a critical factor in eradicating women’s
poverty. In Bangladesh, non-government organisations (NGOs) and microcredit
institutions have been contributing to break the cycle of poverty. Abed (2000) argues
that women’s economic emancipation is pivotal for alleviating poverty. In so doing,
many microcredit institutions in Bangladesh have launched various credit programs
in providing credit mainly to rural women.
The fundamental principles of the microcredit institutions and NGOs in Bangladesh
are based on poverty alleviation and women’s empowerment. These institutions put
forward a view that if a poor woman is given an opportunity she would build up her
assets i.e. financial security (Abed, 2000). These institutions have also realised that
destitute women would perhaps adopt any new opportunity quicker than men (Write,
2000). There could be many reasons for such behaviour: firstly, women may not
waste money on gambling, drinking alcohol, cinemas or cigarettes like men.
45 The poverty line is the level of income below which one cannot afford to purchase all the resources one requires to live. People who have an income below the poverty line have by definition no discretionary disposable income (Sen, 1976).
149
Secondly, women give more emphasis on the well being of their children. That is
money managed by a woman’s hand in a household will bring more benefit to the
family as a whole than money spent by men. Thirdly, women may be better savers
and good at repaying loans on time. Men, on the other hand, have different priorities,
which do not give the family the top position (Yunus, 1999). In relation to
Bangladesh’s traditional social structure, of which gender discrimination is a
fundamental characteristic, Yunus, (1999 p. 88) argues:
A poor woman in our society is totally insecure; she is insecure in her
husband’s house because he can throw her out any time he wishes… She
cannot read and write, and generally she has never been allowed out of her
house to earn money, even if she has wanted to. She is insecure in her in-
law’s house, for the same reason as she was in her parents’ house: they are
just waiting to get her out so they will have one less mouth to feed.
As well, it cannot be assumed that Bangladeshi women in the formal sector are
liberated from discrimination, as it may be apparent when one examines the
formal banking sector. In relation to Bangladesh, where there is a power
distance46 between men and women, Yunus (1999 p. 87) argues that the formal
procedures of obtaining loans are more difficult for women compared to men, as
the bank require approval from the male spouse and/or male guardian.
46 Power distance is a cultural index derived by sociologist Geert Hofstede, (2001). It measures how much a culture has respect for authority. The Arabic-speaking nations, Latin America, Russia and nearly all of Asia are high in power distance. Most of Europe, Canada, Australia and Israel are low in power distance. Japan and Mediterranean Europe fall in the middle. In a high power distance culture, its acceptable for a supervisor to display his authority, while in a low power distance culture, supervisors are expected to treat employees respectfully.
150
There are general views that less educated, rural women are likely to be loyal
and faithful. This is a reason why the microcredit institutions in Bangladesh have
targeted women to provide loans. Of course, the ultimate objective for such a
strategy is to alleviate women’s poverty in order to empower them. In
Bangladesh, 94% of microcredit borrowers are women and the recovery rate of
loans is 98%. Since 1985, the Grameen Bank has been channelling credit to
women. It is also evidenced from the literature that household consumption
increases by 18% when lending to women compared to 11% when lending to
men (Khandker, 1998b).
There exist different views on the impact of microcredit on women’s empowerment.
Although some researchers are sceptical about the positive effects of a credit
program towards such an empowerment, most of the extant studies put forward a
view that microcredit contributes towards women’s empowerment (Amin et al.,
1994; Naved, 1994; Hashemi et al., 1996; White, 1992). However, Goetz et al.
(1996) argue that there is an inverse relationship between the loan amount and
control (used as proxy for empowerment). They have found that control on loans
diminishes beyond a threshold level of membership. Montgomery et al. (1996) have
also expressed some reservations about the empowering effect of credit programs.
They argue that microcredit reinforces the existing gender discrimination and
inequalities and contributes little to alter the social status of women.
This chapter attempts to examine the factors that affect women’s empowerment, the
difference between borrowers and non-borrowers in terms of empowerment and also
151
the effects of credit programs on women’s empowerment in Bangladesh. It is
therefore necessary to elaborate further on how such empowerment can be defined.
Table 5.1 followed by Section 5.2 provides an examination of the literature on how
various researchers have defined the term in carrying out research in Bangladesh.
Based on this literature on women’s empowerment as outlined here, it will then be
necessary to construct the concept of women’s empowerment to be used in the
study. The following sections are devoted to this end.
Table 5.1: Defining Women’s Empowerment
Authors
(& Dates)
Definition of Women’s
Empowerment (Extract)
Research
Framework Used
Programs
Under
Study
Results
Hashemi et al. (1996)
Women’s empowerment is reflected in her relative physical mobility, economic security, ability to buy things on her own, freedom from domination by the family, political and legal awareness and participation in public protests and political campaigning.
Empirical study based on ethnographic data collected through participant observation and informal interviews.
Grameen Bank and BRAC47
Credit program is empowering and it reduces incidence of violence against women due to women’s involvement in the program.
Zaman (1998, 2001)
Ownership and control over assets, general and legal knowledge and knowledge about fertility and mobility of women.
Empirical study based on primary data collected from borrowers.
BRAC Credit program is empowering
Banu et al. (2001)
The capacity of women to reduce their socio-economic vulnerability and their dependency on their husbands or other male counterparts, in terms of their ability to become involved in income-
Study used Chen and Mahmud’s (1995) conceptual framework as well as a scoring method based on study findings. Both quantitative
BRAC Credit program is empowering – husbands learnt how to value wives, as they are the
47 BRAC stands for Bangladesh Rural Advancement Committee, largest non-government organisation (NGO) of Bangladesh.
152
generating activities and freely spend the income thus generated; to accumulate assets over which they can have right of sale and profit; increase their contribution to household expenditure and thereby acquire a greater role in household decision-making and finally, increase their self confidence and awareness of social issues.
and qualitative data were used.
avenues to bring loans into the household for capital investment.
Amin and Pebley (1994)
Woman’s household decision-making power, her control over household resources, physical mobility outside the home and autonomy and attitudes and aspirations.
Qualitative information was gathered through semi-structured interviews. Multiple logistic regressions were used to estimate the impact of group membership and program effect.
BRAC Credit program is empowering - involvement in micro-credit program reduces their chances of abandonment by the husband. Husbands consider themselves as worthy as they can bring credit to the family.
Mustafa et al. (1996)
Opportunities to generate income and control over their income.
Field survey. BRAC Credit program is empowering.
Ackerly (1995)
Used knowledge of accounting for loan activity as a proxy of empowerment.
Different organisation’s effectiveness in achieving the goal of empowerment was compared through the variable called ‘knowledge of accounting.’
Save the Children (NGO), Grameen Bank and BRAC
Credit program is empowering.
Mizan (1993)
Household decision-making index (HHDM) used as a proxy for empowerment: decision of food purchase, education and marriage of children, medical expenses, earnings and business, purchase of agricultural inputs, financial support to husband’s family, etc.
Uni-dimensional variable was used to identify decision-making power of borrowers and non-borrowers.
Grameen Bank
Number of years of borrowing has positive and significant effect on HHDM score.
Goetz and Sen Gupta
Women’s continued high demand for loan and their
Qualitative investigation of
Safe the Children’s
An inverse relationship
153
(1996) manifestly high propensity to repay is taken as a proxy indicator for control and empowerment.
loan histories. A sample 253 female borrowers from four rural credit providers in Bangladesh was used.
Fund, RD 12 (government funded NGO), Grameen Bank and BRAC
between loan amount and control exits. Women’s high repayment rate and sustained high demand for loans is assumed to reflect effective loan investment strategies by women. The paper has raised the issue of women’s loss of direct control over the loan.
Montgomery et al. (1996)
Control over asset is considered as a proxy for empowerment.
Secondary source and small field survey of 67 borrowers.
BRAC This study has reservations about the empowering effect of microcredit.
Kabeer, N. (2001)
Conceptualised women’s empowerment in an understanding of the relationships of dependence, interdependence and autonomous way.
Qualitative methodology.
The Small Enterprise Development Project (SEDP)
Access to loans enhanced women’s role in minor decisions only.
From Table 5.1 it is evident that there is no coherent definition and/or measurement
index of empowerment. It is apparent from the table that the definitions of
empowerment include: control over assets (Goetz and Sen Gupta, 1996;
Montgomery et al., 1996); women’s relative physical mobility, economic security,
freedom from domination, political and legal awareness and participation in public
protests and political campaigns (Hashemi et al., 1996; Banu et al.; 2001; Zaman,
1998 and 2001); control over household resources (Amin and Pabley, 1994);
154
opportunity to generate income and control over income (Mustafa et al., 1996);
handling of loan applications (Ackerly, 1995); and participation in household
decision-making (Kabeer, 2001; Mizan, 1993).
5.2 Defining Women’s Empowerment
United Nations Population Information Network (POPIN)48 has defined women’s
empowerment based on five components, which are as follows:
• women's sense of self-worth;
• their right to have access to opportunities and resources;
• their right to have the power to control their own lives, both within and
outside the home;
• their right to have and to determine choices;
• their ability to influence the direction of social changes to create a better
social and economic order, nationally and internationally.
Rao and Kelleher (1995) define women’s empowerment as the “capacity of women
to become economically self-reliant with control over decisions affecting their lives
and freedom from violence”. Holcombe (1995) defines empowerment as “the
sharing of control, and the entitlement and ability to participate in influencing
decisions regarding the allocation of resources”. Hashemi and Schuler (1996) have
argued that women’s empowerment is reflected in her relative physical mobility,
economic security, ability to buy things on her own, freedom from domination by the
family, political and legal awareness and participation in public protests and political
48 See http://www.un.org/popin/
155
campaigning. Sen (1997) has stated that, “Empowerment is about changes in favour
of those who previously exercised little control over their lives”.
From these definitions we may visualise an empowered woman. An empowered
woman is confident in her ability, she is capable of leading her life independently,
she is socially as well as economically independent, she is opinionated, enlightened
and has freedom from all sorts of domination and finally she is someone who is
capable of standing for her own rights. Now we may formulate a new definition of
women’s empowerment based on the above definitions:
Women’s empowerment comprises women’s education and
knowledge to enhance her understanding about her surroundings, her
ability to control her life, freedom from domination by not depending
on anyone else’s income, her ability to participate in decision-making
process, her capability to make independent decisions and finally her
independence in terms of mobility.
These subjective attributes are difficult to measure. We have therefore used some
proxies to measure different attributes of women’s empowerment. From the
definition we have identified the key features that represent women’s empowerment.
In the definition of women’s empowerment the key elements are: women’s
economic security, her ability to participate in decision-making processes, her
control over assets, her mobility and finally her knowledge and awareness about her
surroundings. In this study we used proxies for each type of attribute to define
women’s empowerment as well as to formulate an empowerment index.
156
Based on the above definition, proxies are used to define women’s empowerment. In
this study we have used proxies such as “Economic Security Index” (ESI), “Purchase
Decision Index” (PDI), “Control over Assets Index” (COAI), “Mobility Index” (MI)
and finally “Awareness Index” (AI) to develop an “Empowerment Index”. Detailed
derivations of each individual index are discussed below.
5.3 Towards Development of a Women’s Empowerment Index
An attempt has been made in this study to extend the definition of empowerment
from a different perspective. In this study empowerment is defined as women’s
holding of her own assets (not family assets), her ability to participate in the buying
and selling process, her control over the sales proceeds, her independence in terms of
mobility (that is, her ability to visit a place unaccompanied by anyone), and her
awareness around general knowledge questions.
Definitions developed in previous studies have not considered women’s individual
ownerships on various assets such as land, house, domestic productive assets (e.g.
cattle, ducks and chickens), savings, jewellery, poultry farm, business enterprises
(e.g. shops). It is important to consider these individual ownership factors while
measuring empowerment proxies. In addition, it can be argued that the
empowerment factor may be caused by women’s participation in the decision-
making process of buying and selling of her assets. We have therefore asked the
question, “Whether the woman participates in the decision-making process of buying
or selling of her assets only if she possesses any assets in her own name”. That
157
means in the case where the answer of the ownership question is negative the
following decision-making question is considered irrelevant.
Like other studies, we also have considered “control over assets” as an empowering
factor. We further have considered that the woman has control over the asset if she is
allowed to keep the proceeds from the sale of her asset. It should be noted that
control over the proceeds of the disposed assets is also an important variable for
women’s empowerment. That is, the sale proceeds are handled by women without
any intervention by the spouse or male guardian.
It is also assumed that women’s “mobility” is an important factor in calculating
women’s empowerment. In other words, whether a woman is able to visit
(occasionally or often) local markets, urban areas (sub-district town), and her
parent’s place; and whether the woman can visit these places by herself or
accompanied by someone (e.g. sons, daughters, friends, husband, neighbours).
A further factor considered in this study is women’s general knowledge. This general
knowledge is measured based on the following questions: (1) whether the concerned
woman knows the name of the local council (Union Parishad) chairman; (2) whether
the woman knows the name of the Prime Minister of the country; (3) whether the
woman knows the minimum marital age of a bride; (4) whether the woman is aware
of divorce procedures; (5) whether the woman has any reservations about other
women working outside (or going out of the house); (6) whether the woman uses
contraceptives and participates in the decision of birth control with her husband; (7)
whether the woman is aware of the AIDS disease; (8) whether the woman knows
158
anyone who has taken or given a dowry in her village; and (9) whether the woman
has received training from the Grameen Bank and BRAC. Although these general
knowledge attributes are not exhaustive, they are taken to measure women’s general
knowledge.
While collecting data it was observed that there was no one in the village who had
not given a dowry for their daughter’s wedding or received a dowry in their son’s
wedding. Therefore, in calculating the awareness index we have not considered
question (8). Similarly, since all the borrowers from both the Grameen Bank and
BRAC have training from their respective organisations, again question (9) is deleted
while calculating the index. Therefore, the remaining seven questions directly
relevant to general knowledge are considered to calculate the awareness index.
The questions asked to both borrowers and non-borrowers are binary in nature.
Therefore there are only two possible answers to all these questions, either
“affirmative” or “negative”.
We have developed a few indices for each category that contribute towards the
development of the Empowerment Index (EI). We have assigned a cut off point49 for
every index to calculation the EI. The following section provides a detailed
description of calculation of the EI.
49 The cut of points of each category are described in the following pages of this chapter.
159
5.4 Calculation of the Empowerment Index
The above-mentioned categories are the Economic Security Index (ESI), Purchase
Decision Index (PDI), Control over Asset Index (COAI), Mobility Index (MI), and
Awareness Index (AI). Calculations of these small indices are described below.
5.4.1 Economic Security Index (ESI)
To find out the ESI of a woman (either a borrower or non-borrower) we ask
questions such as whether she possesses any assets in her own name such as land,
house, domestic productive assets (e.g. cattle, duck, chicken), savings50, jewellery, a
poultry farm or a business enterprise, (e.g. shop). Assets are categorised into four
groups. The first category of asset consists of land and house (valuable assets), the
second category consist of cows, goats, ducks and chickens (productive assets), the
third one consists of cash, savings and jewellery and the final one consists of poultry
or shops in her own name. Jewellery is not taken into the calculation of the index as
it is assumed that females are most likely to have some sort of jewellery, such as
bangles, chains, rings or at least a nose pin. Therefore, we do not take into
consideration that having jewellery is empowering. To find out the ESI, women
owning any of the above-mentioned assets on her own from any of the above
categories scored one point and zero otherwise. Therefore, out of four categories if
someone scored two she is considered empowered in terms of ESI. Therefore, the cut
off point is considered two in this case. That means anyone scoring two or more will
be scored one in the second step to calculate the EI, zero otherwise.
50 Savings is defined here as money saved in the banks or cash saved at home. These figures are self-reported.
160
5.4.2 Purchase Decision Index (PDI)
Some studies have defined “empowerment” as participating in decision-making
processes (Mizan, 1993). Other studies such as Mahmood (2003) and Kabeer (1998)
have considered women’s joint decision-making with her husband as empowering.
We do not consider joint decision-making as empowering. However, if someone
owns any of the above-mentioned assets in her own name, only then it is relevant to
ask if she participates in the decision-making process of buying or selling of her
assets. If she does not possess any of the assets, it is assumed irrelevant to ask such a
question. If she takes her own decision she is considered empowered in terms of
PDI.
One may possess three or four assets categorised above but participate in the
purchase or selling decision of one or two only. The reason for considering PDI in
such a way is to see whether she really has the ability to participate in the decision-
making process (buy/sell) of her asset. Out of the four groups of assets, we have
considered the cut off point to be two. So if she scores two or more, she is
considered empowered in terms of the PDI. That means anyone scoring two or more
will be scored one in the second step to calculate the EI, zero otherwise.
If she possesses no assets in her own name we have assumed it is irrelevant to ask if
she participates in the decision-making process of the item. In that case she scored
zero in PDI as well.
161
5.4.3 Control over Asset Index (COAI)
To calculate COAI, questions are asked such as “Are you allowed to keep the money
from the sale of your asset or do you spend from your own pocket to buy the asset?”
In the ESI if there is a positive answer (she owns some assets in her own name) only
then will she answer this section. Otherwise, it is irrelevant to ask such a question.
She scores one for a positive answer and zero otherwise.
5.4.4 Mobility Index (MI)
To find out the mobility of a female we ask questions such as “Do you visit places
(such as local markets, sub-district town, parent’s place, court, BRAC, and/or
Grameen Bank), generally, occasionally, on your own or with someone?” Out of the
above-mentioned places visiting court, BRAC, and the Grameen Bank is not
considered in MI calculation. A very poor response has been found for visiting court,
so it is ignored in the MI calculation. Further, it is observed that every member has
to attend meetings with their respective organisations; therefore, the other question is
not also considered.
Visiting a place may depend on necessity. In that case it does not matter if someone
visits a place frequently or rarely. For that reason we consider visiting a place both
generally and occasionally is empowering as long as she can visit the place on her
own. It is worth mentioning here that visiting a place accompanied by someone is
not considered in this study.
162
To calculate each mobility indicators such as local markets, sub-district town and
parent’s place we have considered four dummies such as D 1 , D 2 , D 3 , and D 4 where
D 1 is the woman visits a place generally and on her own. D 2 = she visits a place
occasionally and on her own. D 3 = she visits a place generally but accompanied by
someone. D 4 = she visits a place occasionally but with someone. Even though four
dummies are derived but for the purpose of calculating MI index we only consider
first two dummies out of four. That is, if a woman visits a place by herself, not
accompanied by anyone, generally or occasionally she is considered empowered.
Therefore, we have not considered D 3 , or D 4 , where she is visiting a place
accompanied by someone. The cut off point decided here is two. If a woman scores
two or more out of three places (such as local markets, sub-district town and parent’s
place) she is considered empowered in terms of mobility index and is scored one in
the second step to calculate the EI, or zero otherwise.
5.4.5 Awareness Index (AI)
To find out the AI, seven general knowledge questions are asked to the female
(either borrowers or non-borrowers). The questions are (1) Do you know the name of
the Union Parishad’s chairman? (2) Do you know the name of the Prime Minister of
the country? (3) Do you know the legal minimum marital age of a bride? (4) Are you
aware of the divorce procedures? (5) Do you have any reservation about female
working outside the house? (6) Do you use contraceptives and participate in the
decision of birth control with your husband? (7) Are you aware of the disease AIDS?
There were two other questions in the questionnaire which are not considered while
determining the AI. The questions are on “dowry” and on training from the Grameen
163
Bank and BRAC office. For both the questions the response found100%, so these
two questions are ignored from the calculation of awareness index. We therefore,
have considered seven questions out of nine questions to calculate the AI. These
seven questions are directly relevant to general knowledge. The cut off point
estimated here is five. If someone scored five or more out of seven questions, she is
considered empowered and scored one in the second step of calculating EI and zero
otherwise.
5.4.6 Empowerment Index (EI)
The latent variable empowerment (E) is measured through an index called
Empowerment Index (EI). To calculate the EI, all indices such as ESI, PDI, COAI,
MI and AI are added. The cut off point decided here is three. If someone scores three
or more out of five indices is considered empowered and scored one, zero otherwise.
This may be symbolically written as:
Empowerment (E) = (ESI + PDI + COAI + MI + AI) ≥ 3
Empowerment Index (EI) = 1 if E ≥ 3
= 0 otherwise.
In this study we have calculated the EI for both borrowers’ and non-borrowers’,
investigated the factors affecting women’s empowerment, identified any difference
in terms of empowerment between borrowers and non-borrowers and, finally,
examined whether a credit program has any empowering effect on women.
164
5.5 Summary Table of the Indices
Tables 5.2 and 5.3 provide the mean and percentages of all the indices used to
calculate the EI. Table 5.2 shows the average and percentage of all ownership items,
items related to participation in decision-making process, control over assets and
mobility indicators of all borrowers and non-borrowers of all three districts.
Out of the 125 borrowers in Gazipur it is found that only 31 borrowers own land in
their own name. This comprises 25% of the borrowers of Gazipur owning land or
houses in their own name51. The borrowers of Dinajpur own the second highest
percentage of houses or land. Out of total borrowers only 11% of those sampled
possess land or houses in their own name and only 0.5% of non-borrowers possess
their own land.
In terms of owning productive assets it is observed that the non-borrowers of
Gazipur are the highest in percentage ownership. In case of Dinajpur and Chokoria,
borrowers own more productive assets than non-borrowers. The difference between
borrowers and non-borrowers in terms of owning productive assets are quite low in
Gazipur and Dinajpur but quite high in Chokoria.
51 The reason for borrowers having land or houses in their own name could be due to the new rule introduced by the Grameen Bank that, if a person has to take a loan for housing or purchasing a piece of land, his wife’s name has to be included in the property. In general almost all loans are granted towards female.
165
Table 5.2 Borrowers’ and Non-Borrowers’ Ownership, Purchase Decision,
Control, Mobility and Awareness Indices according to Districts
Gazipur Dinajpur Chokoria Total
Indices Classification
of the indices
Bor.
Non-
bor.
Bor.
Non-
bor.
Bor.
Non-
bor.
Bor.
Non-
bor
House/Land 31 (25)
1 (2)
9 (7)
0 (0)
3 (2)
0 (0)
43 (11)
1 (0.5)
Productive Asset
76 (61)
37 (69)
80 (61)
36 (57)
71 (55)
22 (33)
227 (58)
95 (52)
Savings 78 (62)
22 (41)
71 (54)
36 (57)
79 (61)
33 (49)
228 (59)
91 (49)
Economic
Security
Index
(ESI)
Poultry/Shop 2 (2)
0 (0)
5 (4)
0 (0)
0 (0)
0 (0)
7 (2)
0 (0)
House/Land 26 (21)
0 (0)
9 (7)
0 (0)
3 (2)
0 (0)
38 (10)
0 (0)
Productive Asset
43 (34)
18 (33)
26 (20)
15 (24)
31 (24)
11 (16)
100 (26)
44 (24)
Savings 39 (31)
9 (17)
25 (19)
16 (25)
32 (25)
16 (24)
96 (25)
41 (22)
Purchase
Decision
Index
(PDI)
Poultry/Shop 2 (2)
0 (0)
5 (4)
0 (0)
0 (0)
0 (0)
7 (2)
0 (0)
Control
over Asset
Index
(COAI)
75 (60)
21 (39)
38 (29)
17 (27)
47 (36)
23 (34)
160 (27)
61 (33)
Generally on Her Own
25 (20)
4 (7.5)
18 (14)
6.3 (10)
17 (16.8)
5 (7.4)
60 (15)
15 (8)
Occasionally on Her Own
24 (19)
10 (18)
40 (30)
22 (35)
41 (39.4)
17 (25.3)
105 (27)
49 (27)
Generally Accompanied by Others
21 (17)
8 (15)
35 (26)
15 (24)
37 (35.9)
22 (32.3)
93 (24)
45 (24)
Mobility
Index (MI)
Occasionally Accompanied by Others
38 (30)
15 (28)
23 (17)
14 (22)
24 (23.3)
17 (25)
85 (22)
46 (25)
Awareness
Index (AI)
Affirmative Responses from the Questions
84 (68)
34 (63)
81 (61)
40 (63)
81 (78)
43 (64)
246 (64)
117 (64)
Listens to
News
(Radio/TV
)
94 (75)
34 (63)
75 (57)
39 (62)
68 (52)
37 (55)
237 (61)
110 (59)
Number of
Observatio
ns
125 54 132 63 130 67 387 184
(Figures in the parentheses show the percentage)
166
Table 5.3 Empowerment Index of Borrowers’ and Non-Borrowers’ According
to Districts
Districts Gazipur Dinajpur Chokoria Indexes Bor. Non-
bor. Bor. Non-
bor. Bor. Non-
bor. Economic
Security
Index (ECI)
68 (54.4)
17 (31.5)
58 (44)
23 (37)
41 (32)
9 (13)
Purchase
Decision
Index (PDI)
63 (50.4)
24 (44.4)
52 (39)
25 (40)
49 (38)
26 (39)
Control over
Asset Index
(COAI
75 (60)
21 (38.9)
38 (29)
17 (27)
47 (36)
23 (34)
Mobility
Index (MI) 43
(34.4) 13
(24) 54
(41) 31
(49) 60
(46) 23
(34)
Awareness
Index
(AI)
84.5 (68)
33.8 (63)
81 (61)
40 (63)
81 (78)
42.7 (63.8)
Empowerme
nt Index
(EI)
81 (64.8)
28 (51.6)
73 (55)
39 (62)
78 (60)
39 (58)
(Figures in the parentheses show the percentage)
In Gazipur and Chokoria borrowers have more savings than in Dinajpur. In Dinajpur
the savings of borrowers and non-borrowers are not too far apart. It is observed that
only two out of 125 borrowers possess a business or poultry in their own name. No
non-borrowers are found with any business in their own name. When we summarise
the ownership indicators it is observed that the borrowers as a whole are better off
compared to non-borrowers. In terms of districts borrowers of Gazipur are better off
compared to two other districts.
In Gazipur, borrowers and non-borrowers participate more in the decision-making of
their property. Again the borrowers of Gazipur are better managers52 of their savings
by themselves. This savings management ability is also found highest in Gazipur in
52 By the term managing savings we considered that the female can save her money by her own will and spend from the savings if necessary without consulting her male counterpart.
167
percentage terms compared to the two other districts. In Dinajpur non-borrowers
manage more of their savings compared to the borrowers and in Chokoria savings
management by both the groups are pretty much the same.
Sixty per cent borrowers of Gazipur are likely to keep by themselves the proceeds of
the sale of their property (this is how “control over assets” is measured). Both
borrowers and non-borrowers of Gazipur have more control over their assets
compared to the two other districts. Since Gazipur is close to Dhaka, there may be a
capital city influence for women having more control over assets. In Dinajpur and
Chokoria borrowers have more control over assets than non-borrowers.
In terms of women’s mobility it is observed that borrowers of all three districts visit
places on their own compared to non-borrowers. Among borrowers, females of
Gazipur are more independent in terms of mobility than the other two districts. The
mobility ranking from high to low is Gazipur, Chokoria then Dinajpur. Visiting
places either generally or occasionally depends on someone’s need. Therefore we do
not distinguish women’s mobility based on frequency of visiting. It is observed from
Table 5.2 that most women (both borrowers and non-borrowers) are more
comfortable visiting places accompanied by someone. This may be a cultural norm.
Finally from Table 5.2 it is observed that from the sample over 50% women of all
categories are politically conscious and listen to radio/television news regularly. This
information is gathered but not used in the analysis.
168
Table 5.3 shows the summary of Table 5.2 in terms of various indices. It shows the
calculation of EI derived from all other indices in terms of districts. From the table it
is observed that borrowers of all districts are more empowered in terms of ESI
compared to non-borrowers.
Table 5.3 shows that among all districts, borrowers of Gazipur are more empowered
in terms of ESI. The economic security ranking (high to low) is Gazipur, Dinajpur
and then Chokoria for borrowers while for non-borrowers it is Dinajpur, Gazipur and
then Chokoria.
In terms of PDI, both borrowers and non-borrowers of Gazipur take a larger part in
the decision-making process than the two other districts. This may be due to better
facilities available in Gazipur (near the capital city, Dhaka) compared to the two
other districts. This made both borrowers and non-borrowers more empowered in
terms of decision-making in Gazipur. However, in Dinajpur and Chokoria the
difference between borrowers and non-borrowers is very small in terms of PDI and it
seems that non-borrowers are slightly better off compared to borrowers in these two
districts.
It is observed from Table 5.3 that borrowers of all districts have more control over
assets compared to non-borrowers. It is further observed that the difference in terms
of control over assets between borrowers and non-borrowers is quite low in Dinajpur
and Chokoria district while it is quite high in Gazipur district.
169
As mentioned earlier, the analysis considered women’s mobility only if they are
capable of travelling independently to places. Women dependent on someone for
visiting places are not considered empowering in the study. From Table 5.3 it is
observed that non-borrowers of Dinajpur are the most mobile of all. We could not
find any convincing explanation for such a result. It is interesting to see that the
women (both borrowers and non-borrowers) of Dinajpur and Chokoria are more
mobile compared to Gazipur. While comparing borrowers with non-borrowers, it is
observed that borrowers of Gazipur and Chokoria are more mobile compared to non-
borrowers. In Dinajpur non-borrowers are more mobile than borrowers. It is further
observed that the difference between borrowers and non-borrowers in terms of
mobility is higher in Chokoria than two other districts.
It is observed from Table 5.3 that, over 60% of women (both borrowers and non-
borrowers) is politically aware and has better general knowledge. The borrowers of
Chokoria are the most empowered in terms of the AI. Table 5.3 shows that
borrowers of Gazipur and Chokoria are more empowered in terms of the AI
compared to non-borrowers of the same district. It is further observed that non-
borrowers of Dinajpur have better general knowledge compared to borrowers of the
same district.
Finally, Table 5.3 provides the EI. It is observed from the table that, over 64% of
borrowers and 52% of non-borrowers in Gazipur are empowered. The table further
shows, non-borrowers of Dinajpur are more empowered than borrowers in the same
district. This finding may be explained through a casual observation that Dinajpur is
culturally more advanced compared to two other districts. In Chokoria, 60% of
170
borrowers and 58% of non-borrowers are empowered. The ranking of empowerment
from high to low for borrowers is Gazipur, Chokoria and Dinajpur while for non-
borrowers it is Dinajpur, Chokoria and Gazipur.
5.6 What Causes Women to be Empowered?
Once the EI is calculated it is worth exploring the effect of various factors that may
cause women to be empowered. Before identifying the factors we have considered
some questions. The questions are: 1) Does age of the female matter? 2) Does her
education matter? 3) Does the spouse’s age and education make any difference in
terms of empowerment? 4) Does income of the household have any effect? 5) Can
asset accumulation make a woman empowered? 6) Does women’s empowerment
depend on the locality where she resides?
After careful consideration of the aforementioned questions, we have identified
factors that may affect women’s empowerment. The factors are: age of the female,
age of the husband, education of the female, education of the husband, income of the
household, assets of the household and the location where she resides.
According to theory, it is expected that younger women are more empowered; it is
also expected that a young man would encourage his wife to be more empowered.
We may therefore expect a negative coefficient of age of both male and female when
regressed against the binary variable empowerment index. Similarly, theory tells us
that education has a positive impact on woman’s empowerment. We may expect a
positive coefficient for education of both male and female. We may also expect a
171
positive coefficient of income and assets. In terms of location we expect a positive
coefficient of DD 1 (Gazipur) as Gazipur is close to the capital city Dhaka.
Based on the factors as discussed above we have developed research questions and
hypotheses. Research questions, hypotheses and descriptions of the variables are
provided below:
5.6.1 Research Questions
(1) Is a younger woman more empowered than an older woman?
(2) Are there any differences between the borrowers and non-borrowers in
terms of empowerment?
5.6.2 Hypotheses
First Hypothesis: Microcredit borrowers are more empowered than the non
borrowers.
Second Hypothesis: Microcredit programs are empowering for the women.
172
5.6.3 Specification of the Variables53
The specification of variables is as follows:
Demographic Factors (DF)
• Age of the female (borrower/non-borrower) (AFEM)
• Age of the male (husband of the borrower/non-borrower) (AM)
• Education of the female (borrower/non-borrower) (EFEM)
• Education of the male (husband of the borrower/non-borrower) (EM).
Economic Factors (EF)
• Income of the households (YH)
• Assets of the household (ASE). .
Dummy Variables (DV) • Borrower dummy (BD=1 for borrowers and 0 otherwise)
• District dummy (DD 1 =1 for Gazipur and 0 otherwise)
• District dummy (DD 2 =1 for Dinajpur and 0 otherwise).
5.7 Model Specification
Recapping the whole story, we may now develop a model from above-mentioned
variables and dummies. The latent variable “empowerment” is measured by an index
that is binary in nature. This index takes the value of one and zero only. Therefore
53 Summary statistics of the variables are provided in Chapter Four in Table 4.2.
173
the probability of someone being empowered as denoted by P(EI) is shown by the
following model.
5.7.1 Model of the Study
)1.5(
)(
21 iiiii
iiiiiii
DDDDBDASE
YHEMEFEMAMAFEMEIP
εϕθφπμλγδβα
++++++++++=
where ϕθφπμλγδβ ,,,,,,,, are the parameters of the variables age of the female,
age of the male(husband), education of the female, education of the male (husband),
household income (total expenditure is taken as a proxy for income), household
assets54, borrower dummy and both district dummies respectively of the i-th female.
And iε is the error term.
5.8 Estimation Results and Discussion
From the correlation matrix provided in Table 3.1 of Chapter Three, we see that
some variables such as age of the male, age of the female, education of the male and
education of the female are highly correlated. We have, therefore, estimated the
model using male age/education and female age/education in separate equations.
Equation 5.1 is estimated using Probit55 estimation. Since the EI is binary in nature
Probit estimation is used to regress the derived index on various factors. It is
assumed that the estimation error term of the Probit model is normally distributed.
54 Assets are physical assets such as furniture, radio, television and other household items except land and houses valued at market price. 55 An alternative to logistic regression analysis is Probit analysis. The term "Probit' was introduced in
the 1930s by Chester Bliss and stands for probability unit. Probit estimation uses the cumulative
normal probability distribution (Gujarati 1992 p. 356).
174
The STATA 8.0 statistical package is used to estimate the model. White corrected
standard errors56 are used to allow for heteroscedasticity.
5.8.1 Factors Affecting Empowerment - for Borrowers and Non-borrowers
Separately
To find out the factors affecting empowerment, we have divided firstly, the total
sample according to two groups - borrowers57 and non-borrowers to see if these two
groups are different in terms of empowerment. We also want to determine if there is
any difference between women of different districts in terms of empowerment. We
have therefore considered district dummy in the equation. Along with that we have
used the quadratic term of income to see if the quadratic term fits better for the
model. To be consistent with the previous chapters we have used total monthly
expenditure as a proxy for income. After adding the quadratic term and district
dummies the model is as follows:
)2.5()()( 21
2
iiiiiiiii DDDDYHYHEFEMAFEMEIP εϕθνμγβα +++++++=
56 White (1980) has derived a heteroscedasticity consistent covariance matrix estimator that provides correct estimates of the coefficient covariance in the presence of heteroscedasticity of unknown form (Gujarati 1992, p. 449). 57 We know that the decision to participate in a microcredit program is self-selective. This type of
self-selection problem may be corrected through Heckman’s two stage correction procedure. The problem with the Heckman procedure is to identify suitable instruments. Since no suitable instruments have been identified which would permit the use of Heckman procedure to correct this self-selection bias, we did not look at the causal impact of program participation in this study.
175
We have estimated Equation 5.2 separately for borrowers and non-borrowers. Table
5.4 shows the estimation results of the two groups separately. White corrected
standard errors are used to allow for heteroscedasticity and goodness of fit is
measured by the Wald Chi-square test. The output from the Logistic Regression
model provides the pseudo R – square (which is a measure of improvement to fit in a
model due to the independent variables) and is also reported in the table.
Table 5.4: Probit model: Factors Affecting Empowerment Index (Borrowers
and Non-Borrowers)
Borrowers Non-borrowers
Constant 1.46*** (3.82)
1.89*** (2.87)
Age of the Female -0.03*** (-4.81)
-0.03 (-3.20)
Education of the Female 0.07*** (3.47)
0.02 (0.59)
Household Income -0.00 (-0.07)
-0.00 (-0.83)
Income Squared 0.00 (0.43)
0.00 (0.79)
District Dummy 1 0.11 (0.67)
-0.07 (-0.08)
District Dummy 2 -0.18 (-1.06)
-0.02 (-0.08)
No. of Observation 387
184
Wald χ 2 47.74
19.24
Pseudo R2
0.11
0.09
(***) significant at 1% level, (**) significant at 5% level and (*) significant at 10% level. Figures in the parentheses show the z-values.
From Table 5.4 it is observed that the age of the borrowers is significant and
negative. It is a significant factor in determining women’s empowerment but it
affects empowerment negatively. The sign of the coefficient for both borrowers and
176
non-borrowers is found negative. In other words, the estimation results suggest that
younger women are more empowered. Further, education of the borrowers is found
significant and positive. This implies that education of the borrower affects
empowerment in a positive way. This is a reflection that the education plays an
important role in making borrowers empowered.
Since the quadratic term is not found significant we dropped the variable “income-
squared” in our following estimation. We used log transformed “income” and
“assets” variable instead to see if the model fits better. Now instead of separating the
sample into two groups (borrowers and non-borrowers) we would like to determine
the affect of factors for the total sample as a whole. We have therefore considered
borrower dummy as well as district dummy to determine if there is any difference
between borrowers and non-borrowers and the districts respectively. District dummy
is calculated as DD 1 = 1 for Gazipur and 0 otherwise, and DD 2 = 1 for Dinajpur and
zero otherwise. Table 5.5 shows the estimation results of Equation 5.1 for the total
data set using log transformed variables for income and assets.
5.8.2 Factors Affecting Empowerment: Total Data Set Borrowers and Non-
borrowers
We pooled the whole data set to estimate Equation 5.1. Table 5.5 shows the
estimated results of Equation 5.1 considering all the variables. Due to the
multicollinearity problem we do not consider age/education of male and female
together in the same equation.
177
The first column of Table 5.5 shows the estimation result of Equation 5.1
considering all the variables. The results are shown as Equation 1 (column 1 in Table
5.5). The estimation result as shown in Equation 1 shows that age of the female is
significant and it affects empowerment negatively. Education of the female is
significant and affects empowerment positively.
Equation 1 of Table 5.5 further shows that Gazipur is significantly different from the
two other districts. Since Gazipur is close to Dhaka the capital city, it may be
perceived that there may be some capital city influence in Gazipur that makes
women more empowered in that district. It may be that the people of Gazipur have
better access to all facilities that may bring a higher standard of living as well as
empowering the female in that district.
From the estimation results of Equation 1 in Table 5.5 we see that the borrower
dummy is not significant. We therefore have decided to drop this variable and
estimate the model keeping age/education of male and female in separate equations.
Equations 2 and 3 are estimated using female and male age/education separately
after dropping borrower dummy. The results of Equations 2 and 3 are provided in the
second and third columns of Table 5.5 respectively.
Table 5.5 further shows that the coefficient of the age of the male is significant and
negative. This is an interesting finding. This may be interpreted as a younger male
(male counterpart could be husband or father) encouraging the female to be
empowered. The table also shows that the male partner’s education affects women’s
empowerment significantly and positively. There could be a different interpretation
178
of the sign of the coefficient of age/education of the male counterpart. This could
also be due to the correlation between female age/education with male
age/education.
Table 5.5: Probit model: Factors Affecting Empowerment Index: (Pooling Full
Data Set)
Equation 1 Equation 2 Equation 3
Constant -0.182 (-0.17)
-0.18 (-0.17)
0.03 (0.03)
Age of the Female -0.02*** (-3.88)
-0.03*** (-3.90)
-
Education of the
Female
0.04** (2.29)
0.04** (2.30)
-
Age of the male -
- -0.03*** (-4.09)
Education of the
Male
- - 0.05*** (2.73)
Log of Income 0.03 (0.28)
0.04 (0.29)
0.03 (0.19)
Log of Assets 0.07 (1.32)
0.08 (1.32)
0.08 (1.43)
Borrower Dummy 0.001 (0.01)
- -
District Dummy 1 0.59*** (2.90)
0.50*** (2.90)
0.49*** (2.88)
District Dummy 2 0.21 (1.30)
0.21 (1.31)
0.23 (1.44)
No. of
Observation
379
379 379
Wald χ 2 32.82
32.82 37.26
Pseudo R2
0.06
0.07 0.07
(***) significant at 1% level, (**) significant at 5% level and (*) significant at 10% level. Figures in the parentheses show the z-values.
179
5.8.3 Factors Affecting Empowerment: Considering Different Income Level
Households
From Tables 5.4 and Table 5.5 we have observed that the coefficient of the variables
“income” or “log of income” are not found significant. We have overlooked so far
the possibility that the variables income and assets could be correlated. We have
therefore dropped the income variable from Equation 5.1 and estimated the model
considering log transformed assets alone as an economic variable. We have used
age/education for male and female in separate equations. We have divided the
households into three equal groups based on their income level to determine if their
income level has any impact towards women’s empowerment. We used two income
dummies such as 1DY and 2DY . 1DY = 1 for low income group and zero otherwise.
2DY = 1 for middle income group and zero otherwise. After adding the dummies the
model is as follows:
)3.5()( 21 iiiiiiii DYDYLnASEEFEMAFEMEIP εϕθλγβα ++++++=
We have estimated Equation 5.3 for the whole group (borrowers and non-
borrowers). Equations 3 and 4 (columns 1 and 2 of Table 5.6) show the estimation
results of Equation 5.3 for female and male respectively. From Table 5.6 we found
that the age of female is significant. Education of the female is significant at the 10%
level. The coefficient of age of the male is found negative and significant. This
implies that a younger male also encourages a woman’s empowerment. The
coefficient of education for both male and female is found positive. This reinforces
180
the importance of education for both male and female in bringing women’s
empowerment.
Table 5.6 further shows that after dropping the income variable (as one of the
economic factors) we found the coefficient of assets is positive and significant at the
5% and 10% levels. Families with more assets are definitely privileged which may
contribute towards better facilities to empower the female in the family.
Furthermore, it is observed that the middle income group household is significantly
different (at the 5% and 10% level) from low and high income group household.
There is no plausible explanation for the negative coefficient of this income dummy.
There may be some data discrepancies (one of the limitations of primary data) which
prevent a reasonable explanation here.
181
Table 5.6: Probit Model: Factors Affecting Empowerment Index: Based on
Different Income Level Households
Equation 3 Equation 4
Constant 0.28 (0.55)
0.38 (0.72)
Age of the Female -0.03*** (-3.94)
-
Education of the Female 0.04* (1.75)
-
Age of the Male -
-0.03*** (-4.07)
Education of the Male -
0.04** (2.33)
Log of Assets 0.11* (1.94)
0.11** (1.97)
Low Income Dummy 0.02 (0.11)
0.04 (0.21)
Middle Income Dummy -0.33** (-2.01)
-0.32* (-1.94)
No. of Observations 387 387
Wald χ 2 26.94
34.63
Pseudo R2
0.06
0.07
(***) significant at 1% level, (**) significant at 5% level and (*) significant at 10%level. Figures in the parentheses show the z-values.
5.8.4 Do Microcredit Programs Empower Women?
In this section we want to determine the effect of microcredit program on women’s
empowerment. We have therefore considered only the borrowers’ sample (as non-
borrowers do not have credit). We have used the simultaneous equations model as
presented in Chapter Three with some modification for the purpose of estimation.
182
As in Chapter Three we first assume “credit” depends on “Demographic Factors”
(DF), “Economic Factors” (EF) and unknown factors different from DF and EF.
These unknown factors are the “instrumental variables” used in the model. The
model takes the following form:
)4.5(C
iiicicici ZDDEFDFC επδγβα +++++=
where iC is the amount of credit of i-th individual, iDF is the vector of some
demographic variables (age, education, husband’s age, husband’s education) of i th
borrower, iEF is the vector of some economic variables (household assets,
household income) of i-th household, iDD is the district dummy of i-th district and Z
represents the instrumental variable, which is different from DF, EF and DD. In this
equation we have used “types of houses” and “number of earners” as instruments.
πγβ ,, are the parameters and icε is the error term. In our estimation we used log
transformed variables for amount of credit as well as for the economic variables.
Women’s empowerment may also depend on credit. We have therefore; formulate
the following equation which shows that the EI may also depend on the same
demographic factors, economic factors and also on “credit”. Therefore, the equation
takes the following form:
)5.5()(E
iiiEiEi CEFDFEIP εδγβα ++++=
where, P( iEI )is the probability of i-th borrower to be empowered. δγβ ,, are the
parameters and iEε is the error term. To solve these two simultaneous equations
183
models, we first estimate Equation 5.4 using OLS estimation. We then estimate
Equation 5.5 using the estimated value of amount of credit using Probit estimation.
The results of OLS estimation of Equation 5.4 is provided in Table 5.7. White
corrected standard errors are used to allow for heteroscedasticity.
Table 5.7: OLS Estimation of Equation 5.4
Constant 5.92*** (12.68)
Age of the Female 0.017*** (5.28)
Education of the Female -0.002 (-0.24)
Log of Income 0.26*** (4.75)
District Dummy 1 0.46*** (5.51)
District Dummy 2 -0.08 (-1.19)
Types of Houses 0.07 (1.23)
Number of Earners 0.06 (1.44)
Number of Observations 387
R-Squared 0.32
(***) significant at 1%level, (**) significant at 5% level and (*) significant at 10% t level. Figures in the parentheses show the t-values.
From Table 5.7 we see that credit programs are significantly and positively affected
by the age of borrowers. It also is affected positively by income. The district dummy
for Gazipur shows that Gazipur is significantly different from the two other districts
in terms of program intervention.
184
Table 5.8: Probit Estimation: Using the Estimated Value of Amount of Credit
Constant -2.69* (-1.56)
Age of the Female -0.04*** (-4.34)
Education of the Female 0.07*** (3.63)
Estimated Value of Log of Credit 0.55* (1.94)
Log of Income -0.11 (-0.68)
Gender of the Household Head 0.36 (1.54)
Number of Observations 387
Wald χ 2
28.50
Pseudo R2
0.06
(***) significant at 1% level, (**) significant at 5% level and (*) significant at 10%. (Figures in the parentheses show the z-values)
Estimated value of amount of credit obtained from Table 5.7 is used to estimate
Equation 5.5. The results are shown in Table 5.8. We see from Table 5.8 that
woman’s age affect woman’s empowerment significantly. We also see that woman’s
education is significant and positive for woman’s empowerment. Finally we also see
that the coefficient of amount of credit is significant at the 10% level. We may
therefore conclude that microcredit programs affect women’s empowerment
positively. As the amount of credit increases borrowers become more empowered.
This finding reinforces the claim made by the microcredit providers that programs
help in making women empowered.
185
5.9 Conclusion
In this chapter first of all we have introduced “women’s empowerment” as defined
by different researchers. A comparison of the definitions provided by different
researchers is presented in a tabular form. Based on the formal definitions we have
derived our own definition from a different perspective. We then used proxies to
measure women’s empowerment (a subjective term) to quantify it. Detail derivation
of all proxies (termed as indices) are provided in this chapter.
In this chapter we have tried to find out if there is any significant difference between
borrowers and non-borrowers in terms of empowerment. We therefore divided the
sample into two groups and have estimated the equation separately for the two
groups. In this chapter we have also determined factors that may have effect on
women’s empowerment. Finally we have tried to find out whether microcredit
programs have any affect on women’s empowerment. We have used modified
version of the model used by Pitt and Khandker (1996) as presented in Chapter
Three and estimated the simultaneous equations model using 2SLS estimation.
It is observed from the analysis that in terms of ownership of houses, land and
businesses, borrowers of Gazipur are better off compared to non-borrowers as well
as of other districts. It further observed that the borrowers as a whole are better off in
terms of Economic Security Index (ESI) in all districts. However, by looking at the
individual figures it appears that non-borrowers of Dinajpur and Gazipur have more
savings and productive assets respectively compared to the borrowers. There is no
plausible explanation for such a result. In terms of decision-making the overall
186
picture shows that borrowers are better off. In terms of mobility it appears that
female of all districts are more comfortable going out with someone instead of by
themselves. This could be a cultural norm. From careful observation it appears that
borrowers of Chokoria are most mobile. In Dinajpur it appears that non-borrowers
are more mobile compared to borrowers.
The overall picture shows that in terms of awareness there is no difference between
borrowers and non-borrowers. However, the individual depiction shows that
borrowers of Gazipur and Chokoria are more empowered in terms of general
knowledge compared to non-borrowers but not in Dinajpur.
In terms of the Empowerment Index (EI) it is observed that borrowers of Gazipur are
the most empowered compared to non-borrowers as well as other districts. When
comparing borrowers with non-borrowers of all districts, it appears that the
borrowers of Gazipur and Chokoria are more empowered than non-borrowers
whereas in Dinajpur non-borrowers are more empowered than borrowers.
In this study we have tried to find out the factors that may affect women’s
empowerment. Factors are divided in terms of demographic and economic criteria.
Some dummy variables (borrower, district, income) are considered to estimate the
model. It is observed from the analysis that age of the female affects empowerment
negatively. That implies younger women are more empowered. We also have found
that education of the female affects empowerment positively which shows the
importance of education in bringing about women’s empowerment. The most
interesting finding is the age and education of the male partner. It appears that young
187
and educated male encourages women to be more empowered. We further have
observed that assets are positively related to women’s empowerment.
In this study we have found that Gazipur is significantly different from two other
districts in terms of empowerment and also middle income group borrowers are
different from low and high income groups.
We have separated borrowers and non-borrowers to find out if there is any
significant difference between the groups. We also have estimated the full data set
using the borrower dummy. The estimation result (insignificant borrower dummy)
shows that there is no difference between the two groups in terms of empowerment.
This could be due to demonstration effect that non-borrowers are equally empowered
as borrowers. We therefore reject the null hypothesis that borrowers are better off in
terms of empowerment than non-borrowers.
Finally, we have tried to find out the impact of microcredit on women’s
empowerment. 2SLS estimation is used in a simultaneous equations model. It is
observed from the analysis that microcredit places an important role in bringing
women’s empowerment. The positive coefficient shows that as credit increases
woman becomes more empowered. This reinforces the claim made by the
microcredit providers. This result is consistent with Hashemi et al. (1996), Zaman
(1998), Amin and Pabley (1994) and Ackerly (1995). We may conclude that even
though microcredit programs are empowering we cannot say with confidence that
this is only due to microcredit program intervention.
188
Chapter Six
CONCLUSION AND FUTURE RESEARCH
6.1 Introduction
This is an impact assessment study on microcredit organisations in Bangladesh. This
study does not examine the overall poverty level in Bangladesh. Instead it examines
the impact of microcredit programs on household outcomes, income (expenditure)
consumption and women’s empowerment. This is an original and comprehensive
study. The uniqueness of the study is reflected not only in its use of primary source
data, but also in the extent of research performed in the study. Different issues are
addressed in separate chapters. Individual attention has been given to each issue.
Each analytical chapter has addressed new topics and ideas in order to add a new
dimension to the existing research.
The study provides a comprehensive theoretical consideration of rural credit markets
and relates the theory to the performance of microcredit programs in Bangladesh. It
also shows how these programs have overcome the shortcomings of rural financial
markets. The study further establishes the link between the program’s involvement
and the recent socio-economic development of Bangladesh.
The study is conducted using primary data collected from borrowers of two major
microcredit institutions of Bangladesh - the Grameen Bank and Bangladesh Rural
Advancement Committee (BRAC).
189
The impact of microcredit on various household outcomes has been chosen for
study, since such a wide-ranging study has not been conducted previously. The
existing impact assessment studies often provide contradictory results. Such an
intensive study on consumption using microcredit data has not been found in the
prevailing literature. Therefore the consumption behaviour of borrowers compared to
non-borrowers has been considered for analysis in this study. Women’s
empowerment is an important topic. This study defines women’s empowerment in a
different perspective and measures the latent variable “empowerment” through an
index as well as assessing the impact of microcredit on women’s empowerment.
This concluding chapter summarises the study in Section 6.2. Main findings and the
contribution of this study to the existing literature are provided in Section 6.3. Policy
implications are suggested in Section 6.4. Study limitations and suggestions for
further research are provided in Section 6.5 and 6.6 respectively. Concluding notes
are provided in Section 6.7.
6.2 Summary of the Study
The introductory chapter has outlined the framework of this study. It has pointed out
the general hypotheses about the links between microcredit and poverty. Given the
importance of microcredit in the economic emancipation of rural people, the
question arises in the study: “Do microcredit programs in Bangladesh improve
various household outcomes such as income and assets?” To examine the study’s
hypotheses and to answer the question, a comprehensive range of theoretical and
empirical models has been considered.
190
Chapter Two is of fundamental importance. It has presented a broad theoretical
consideration of how the rural financial market developed and worked especially
prior to the microcredit period. Microcredit institutions have reversed the system to
some extent and made rural financing available and affordable for everyone instead
of having a master subordinate relationship in the rural economy. The program has
eliminated potential risk which is a remarkable contribution in rural financial
markets.
Chapter Two also provides an overview of the Bangladesh economy mainly before
the inception of microcredit institutions. It has further discussed the journey of the
two microcredit institutions under study in the past 30 years. It has also shown the
growth and development of these institutions and highlighted their involvement in
social development. The chapter has established a link between the contribution of
these institutions and the recent socio-economic development of Bangladesh.
The last part of Chapter Two has reviewed a wide range of literature on the impact of
microcredit on poverty, women’s empowerment and consumption behaviour. Since
no literature has been found on the impact of microcredit on consumption, an overall
literature on consumption in the context of developed as well as developing countries
have been presented in Chapter Two.
Chapters Three, Four and Five are the core chapters of the study. The three chapters
have been designed in three different ways. They differ in terms of use of
econometric techniques as well as use of models. In Chapter Three we have assessed
the impact of microcredit on various household outcomes such as income (total
191
expenditure is used as a proxy for income) and assets. The chapter has adapted a
modified version of the model used by Pitt and Khandker (1996). In the
simultaneous equations model we have divided the independent variables in terms of
demographic as well as community infrastructure.
We used Two-Stage Least Squares estimation (2SLS) to determine the impact of
credit on the log transformed variables (total expenditure and asset accumulation).
The borrowers are divided into three categories according to their income levels. We
have estimated the impact of credit on different income level borrowers. We have
used Tobit estimation using non-borrowers as the censored group.
Chapter Four has presented a comparative analysis of microcredit borrowers with
non-borrowers in terms of consumption. We have considered consumable food and
non-food items that are commonly used by the rural people. The chapter has
introduced the relevant theories such as mixed demand theory and an Almost Ideal
Demand System (AIDS) and then it has specified the model that has been used for
analysis.
In Chapter Four we have divided the households into three groups based on their
income level. In this chapter we have estimated each equation in the model using
OLS to see the difference in consumption between different income level
households. We have then tested for significant differences between borrowers and
non-borrowers. To test for robustness we have introduced an interactive variable in
the model to see the difference between the two groups in terms of the intercept as
192
well as the coefficient terms. Joint coefficient tests have also been undertaken in this
chapter.
The model that has been estimated in Chapter Four contains multiple equations; OLS
estimation estimates each equation in the model one at a time. Seemingly Unrelated
Regression (SURE) analysis estimates all the equations in the model as a set
allowing for covariance of error terms. Therefore, we have estimated the multiple
equations model using SURE in this study.
Chapter Five is on women’s empowerment. Women’s empowerment has been
defined from a different perspective and an index is developed to measure the latent
variable “empowerment”. In the development of an empowerment index we have
discussed five empowerment correlates that have been used to define empowerment
in this study. These small correlates are used to develop the EI. Since EI is a binary
variable Probit estimation is used to analyse the model. Finally in this chapter we
have used the same simultaneous equations model as used in Chapter Three to find
out the impact of credit on women’s empowerment.
6.3 Key Findings of the Study
The 2SLS estimation in Chapter Three shows that the amount of credit has a positive
impact in bringing higher income and assets for borrowers. This is a very important
finding as it confirms the claims made by the microcredit providers. This is an
indication that microcredit programs are helping borrowers to achieve better quality
of life through higher income and assets. Age of both male and female is important
193
in bringing higher income and assets. Our results suggest that income and assets
accumulation increases as people gets older. Education plays an important role to
enhance borrowers’ income and assets. Number of adult males as a percentage of
family size is also a significant factor that brings higher income and assets. The
analysis further shows that male-headed households benefit from microcredit
programs. Infrastructural facilities such as having electricity are important factors for
credit programs to be effective in producing household outcomes. However, when
these borrowers are divided into different income levels it gives an interesting
picture. The findings of this study show that even though microcredit is improving
overall household income and assets it is more effective for higher income level
borrowers. As the amount of credit increases, income of only the high income level
borrowers increases. It may be concluded from these findings that microcredit
programs are effective only for the high income group borrowers in producing better
household outcomes. Tobit estimation is used considering non-borrowers as the
censored data. The estimation results are no different than the 2SLS estimation.
To estimate the consumption of different income group borrowers and non-
borrowers we have divided the households into three groups in Chapter Four. The
AIDS model is initially estimated using OLS. From the estimation cereal is found to
be a necessary item for all income group households. Small changes in income will
cause bigger changes in cereal consumption for low income group households. That
means cereal is income elastic for low income group households. It is found that
higher income group borrowers are better off in terms of cereal consumption. The
estimation result shows that there is difference in consumption between districts in
terms of cereal consumption.
194
Lentils (pulses) and vegetables are found to be necessary items for all income group
households. It is also found that the high income level borrowers are better off in
terms of pulses consumption. The test result suggests that there is a variation in
consumption of vegetables, pulses and sugar consumption between districts.
Protein (beef, mutton, chicken, fish, milk and eggs) are found to be non-necessary
items for all income group households. Borrower dummy is found significant for all
income groups in case of meat and fish consumption. This implies that there is a
difference in consumption of meat and fish between borrowers and non-borrowers.
Our results suggest that microcredit borrowers are better off in terms of protein
consumption. From these results we may conclude that the microcredit programs
improve consumption for borrowers.
Cigarette, betel leaf and betel nuts are found significant and necessary for high
income group households. Demand for total food for the low income group
household decreases at a higher rate compared to the high income group when there
is a decrease in income. Borrower dummy is found significant for all income groups
for total food consumption. This implies that there is a difference in total food
consumption between borrowers and non-borrowers. This further implies that
borrowers are better off in terms of total food consumption. This result strengthens
our previous suggestion that the microcredit programs are successful in improving
consumption for borrowers.
Education expenses are found non-necessary but significant for the high income
group. However, expenditure on health is found necessary only for high income
195
group households. Total non-food items are found non-necessary for all income
levels. Significant borrower dummy implies that there is difference between
borrowers and non-borrowers in terms of consumption of total non-food items.
The test of robustness indicates that the interactive variable is significant for cereal,
protein, cigarettes, total food, education, fuel and total non-food. It is observed that
items such as pulses, meat, fish, eggs, milk and cloth are more inelastic (steeper)
function for the borrowers. Therefore, it can be concluded that for borrowers’ pulses,
meat, fish, eggs, milk and cloth are essential items.
It is observed from the test result that for both groups, the intercepts as well as the
slopes are different for both food and non-food items. Therefore, the result strongly
suggests that there are significant differences between borrowers and non-borrowers
in terms of consumption of items. The coefficient test result shows statistically
significant coefficients for items such as sugar, protein, education, clothing, total
food and total non-food. Similar results are obtained from SURE estimation.
Chapter Five estimates the EI using Probit model. From the summary table it is
observed that out of 125 borrowers in Gazipur district only 31 borrowers own land in
their own name. The borrowers of Dinajpur own the second highest percentage of
houses or land. Out of total borrowers only 11% of the sample possesses land or
houses in their own name and only 0.5% of non-borrowers possess their own land.
The productive assets are classified as cattle, ducks and chickens. It is observed that
in terms of productive assets the non-borrowers of Gazipur are highest in percentage
196
ownership. However, in Dinajpur and Chokoria districts borrowers own more
productive assets compared to non-borrowers. It is interesting to see that the
differences in ownership in productive assets are quite low between borrowers and
non-borrowers in all districts. This may be due to the fact that these are common
items in rural Bangladesh.
In Gazipur and Chokoria borrowers have more savings than in Dinajpur. Our study
further shows that only two borrowers of Gazipur and five borrowers of Dinajpur
possess business or poultry in their own name. None of the non-borrowers are found
to have any such businesses in their own names. When summarising the Economic
Security Index (ESI) it is found that the borrowers as a whole are better off in terms
of ownership of assets and in particular the borrowers of Gazipur are better off
compared to two other districts.
In Gazipur district borrowers and non-borrowers both participate more in the
decision-making process. Both the borrowers and non-borrowers of Gazipur have
more control over assets compared to two other districts. Since Gazipur is close to
Dhaka, there may be a capital city influence in women having more control over
assets. In Dinajpur and Chokoria borrowers have more control over assets than non-
borrowers.
In terms of women’s mobility it is observed that the borrowers of all three districts
are more comfortable to visit places on their own compared to the non-borrowers.
Among the borrowers, the females of Gazipur are more independent in terms of
197
mobility, than the female of the other two districts. The mobility ranking from
highest to lowest is Gazipur, Chokoria and then Dinajpur.
The above findings are summarised as indices and used to develop the EI. The
economic security index (ESI) ranking from highest to lowest is Gazipur, Dinajpur
and Chokoria. However, for non-borrowers it is Dinajpur, Gazipur and Chokoria. In
terms of the Purchase Decision Index (PDI), both borrowers and non-borrowers of
Gazipur take a greater part in decision-making process compared to the two other
districts. However, in Dinajpur and Chokoria the difference between borrowers and
non-borrowers is less in terms of PDI, and it appears that the non-borrowers are
slightly better off compared to the borrowers in these two districts.
The analysis further shows that borrowers of all districts have more control over
assets compared to non-borrowers. It is observed that the difference in terms of
control over assets between borrowers and non-borrowers is low in Dinajpur and
Chokoria district while it is quite high in Gazipur district.
It is interesting to see that the women (both borrowers and non-borrowers) of
Dinajpur and Chokoria are more mobile compared to Gazipur. However, comparing
borrowers with non-borrowers, it is observed that borrowers of Gazipur and
Chokoria are more mobile compared to non-borrowers. In Dinajpur, non-borrowers
are more mobile than borrowers. It is further observed that the difference between
borrowers and non-borrowers in terms of mobility is higher in Chokoria than two
other districts.
198
The analysis further shows that over 60% of women (both borrowers and non-
borrowers) are politically aware and have better general knowledge. Borrowers of
Chokoria are the most empowered in terms of the Awareness Index (AI). The table
shows that borrowers of Gazipur and Chokoria are more empowered in terms of AI
compared to non-borrowers of the same district. It is further observed that non-
borrowers of Dinajpur have better general knowledge compared to borrowers of the
same district.
In estimating factors affecting the EI, the test result shows that age of the female is
significant and facilitates women’s empowerment. It further shows that younger
women are more empowered compared to older women. Education of the female is
also a significant factor in facilitating women’s empowerment. It is interesting to see
that the age and education of the male is also significant in facilitating women’s
empowerment.
Assets are other factors that enhance women’s empowerment. However, the
estimation result further suggests that there is no difference between borrowers and
non-borrowers in terms of empowerment. This could be due to demonstration effects
that non-borrowers are equally empowered as borrowers. It is also found that
Gazipur is significantly different from the two other districts. Finally, the 2SLS
estimation result shows that microcredit programs are significant and positive in
bringing women’s empowerment. We may therefore conclude from the results of this
study that microcredit facilitates women’s empowerment.
199
By summarising the results of all the aforementioned analytical chapters we may
conclude that microcredit programs are helping to create higher income and asset
acquisition for borrowers as a whole. From our results it may be concluded that the
microcredit programs provide better quality of life for its borrowers. Our findings
confirm the claims made by microcredit providers. However, when we segregate the
borrowers into groups it appears that microcredit programs are not as effective for
the low and middle income group borrowers as they are for high income group
borrowers.
In terms of consumption, it is observed that borrowers of all income levels are better
off in terms of consumption of relatively expensive food items compared to non-
borrowers. While comparing borrowers of different income levels our results suggest
that high income level borrowers are better off in terms of consumption of few items.
We may conclude that microcredit borrowers are better off in terms of consumption
of most items compared to non-borrowers and high income borrowers are slightly
better off compared to low and middle income level borrowers.
It is found that borrowers as a whole are better off in terms of holding assets
(measured by ESI) compared to non-borrowers. In terms of PDI the results are
mixed. It is found that borrowers of Gazipur and Chokoria are more mobile
compared to non-borrowers while borrowers of Gazipur and Chokoria are more
empowered compared to non-borrowers. However, non-borrowers of Dinajpur are
more empowered compare to borrowers. This finding may be explained through a
casual observation that the district of Dinajpur is culturally more advanced compared
to the two other districts.
200
Our results further suggest that age, education of both female and male as well as
assets of the household facilitates women’s empowerment. It is interesting to find
more than 50% empowered borrowers and non-borrowers in all districts. Finally, it
is found that microcredit programs are positively and significantly facilitating
women’s empowerment. Therefore, we may say that microcredit programs are
successful in empowering its borrowers. We further suggest that microcredit may
also be successful in making non-borrowers (those who are not members of credit
programs) empowered through a demonstration effect.
6.4 Policy Implications
The stated goals of the microcredit institutions are generally consistent with the
development objectives of the government of Bangladesh. Our findings suggest that
credit programs administered by two of these leading institutions are effective in
increasing borrowers’ assets and income. This study also suggests that their
programs are facilitating the empowerment of women, who constitute almost half of
the potential labour force of the country. Therefore, the government should take note
of these findings and try to integrate microcredit institutions like the Grameen Bank
and BRAC to make a coordinated effort to overcome major hindrances to the
economic development of Bangladesh.
One incidental finding of this study deserves special attention. Though the overall
evidence of this study tends to suggest that microcredit is an attractive tool to
produce better outcomes in terms of income, assets and consumption we have also
found that such favourable outcomes often accrue to the relatively wealthy
201
borrowers compared to the non-wealthy borrowers. We interpret this finding as an
apparent weakness of the present programs. Therefore, some adjustment to existing
microcredit programs should be made to achieve the intended outcome of such
programs, which is to assist those at the bottom of the society.
6.5 Study Limitations
There are some limitations in this study. The theory does not provide a unique model
to guide the empirical research for examining the impact of microcredit on various
household outcomes. In this respect, economists and researchers have applied
different econometric methodologies and techniques to establish the possible link
between microcredit program and household outcomes.
More precisely, there is no comprehensive theoretical framework in terms of rural
financial market and/or microcredit programs that are required for microcredit and
its economic impact on household outcomes. In addition, many of the empirical
studies suffer from model misspecification and are also sensitive to data coverage.
This study is based on survey data with limited observations. Since this study is
based on primary data, we could not estimate the impact over time and there may be
some data discrepancies. Income, asset accumulation and consumption expenditure
are self-reported and measured only over a relatively short period in the absence of
time series or panel data.
202
However, given these limitations our study has used sophisticated econometric
techniques on consistent cross-section data collected from three different districts of
Bangladesh from both borrowers and non-borrowers of both the Grameen Bank and
BRAC. Appropriate econometric analyses along with related diagnostics are
provided so as to ensure the rigor of economic interpretations of the results.
6.6 Avenues for Future Research
Microcredit is an attractive tool to produce better outcomes in terms of income and
assets but some of our results suggest that it is more effective for relatively wealthy
borrowers compared to non-wealthy borrowers. It would be a good research question
to pursue in future to uncover why this is the case.
From this discussion it may be concluded that borrowers of microcredit are better off
in terms of consumption patterns compared to non-borrowers. But this does not
necessarily tell us that borrowers are better off only due to microcredit intervention.
It may be the fact that the well-off sections of the society are the borrowers of
microcredit program. It would be a good research question to pursue in the future
through panel data estimation to determine whether this is the case.
203
6.7 Concluding Notes
Microcredit programs in Bangladesh represent a breakthrough for rural financial
markets. Two particularly attention-grabbing features of these programs are their
high recovery rates and the heavy emphasis that they place on women. The unique
credit delivery methods have mitigated the problems of moral hazard and adverse
selection. The founder of the Grameen Bank Professor Yunus, has won the 2006
Nobel Peace Prize for waging war against poverty. This event has strengthened the
importance of microcredit. On paper these eye-catching features may sound too good
to be true, so a comprehensive investigation conducted at this stage would prove to
be invaluable in uncovering the truth. Therefore an examination was crucial from
borrowers’ point of view in understanding the effectiveness of the program.
This study examines the impact of microcredit programs on its borrowers through
the collection and compilation of primary source data. Every attempt has been made
to cover all the economic aspects affecting borrowers in order to make this analytical
study as thorough and comprehensive as possible. Attempts have also been made to
substantiate the claims made by microcredit providers with real world evidence.
An interesting finding of this study is that microcredit has been successful in creating
a better quality of life for borrowers, through enhancing their income, consumption
and assets. Our findings strongly suggest that borrowers are better off in terms of
consumption of most of the food and non-food items. Most importantly this study
shows that microcredit programs are successful in empowering women. It is also
clear from the findings that the intensive credit intervention in the rural economy is
204
creating a demonstration effect among everyone including non-borrowers. There is
no doubt from the results that microcredit has continued to have a revolutionary
effect on the rural economy of Bangladesh. This study has already highlighted how
the programs have spurred the continual growth and development of the people
living in rural areas. The continued success of microcredit in Bangladesh has created
an exemplary template for the rest of the world to adopt.
205
Appendices
7.1 Appendix A: Global Poverty Figures
Global Poverty in 1990 and 1999*
Number of people living on less than $1 per day (millions)
Region 1990 1999
East Asia and Pacific 486 279
Excluding China 110 57
Europe and Central Asia 6 24
Latin America and the Caribbean
48 57
Middle East and North Africa
5 6
South Asia 506 488
Sub-Saharan Africa 241 315
Total 1,292 1,169
Excluding China 917 945
Source: Global Economic Prospects 2003, World Bank.
* The poverty figures have not been updated since 1999.
206
7.2 Appendix B: The Sixteen Decisions of the Grameen Bank
1. The four principles of Grameen Bank-discipline, unity, courage, and hard
work-we shall follow and advance in all walks of our lives.
2. We shall bring prosperity to our families.
3. We shall not live in dilapidated houses. We shall repair our houses and
work towards constructing new houses as soon as possible.
4. We shall grow vegetables all the year round. We shall eat plenty of them
and sell the surplus.
5. During the planting seasons, we shall plant as many seedlings as possible.
6. We shall plan to keep our families small. We shall minimize our
expenditures. We shall look after our health.
7. We shall educate our children and ensure that they can earn enough to
pay for their education.
8. We shall always keep our children and the environment clean.
9. We shall build and use pit latrines.
207
10. We shall drink tube well water. If it is not available, we shall boil water
or use alum.
11. We shall not take any dowry in our son’s weddings; neither shall we give
any dowry in our daughter’s weddings. We shall not practice child
marriage.
12. We shall not inflict any injustice on anyone; neither shall we allow
anyone to do so.
13. For higher income, we shall collectively undertake bigger investments.
14. We shall always be ready to help each other. If anyone is in difficulty,
we shall all help.
15. If we come to know of any breach of discipline in any centre, we shall all
go there and help restore discipline.
16. We shall introduce physical exercise in all our centres. We shall take part
in all social activities collectively.
208
7.3 Appendix C: Household Survey Questionnaire
Household Survey Questionnaire
(This survey questionnaire will only be used for academic purposes)
Identification:
Village Name:--------------------------------------- ---- Bari Name:------------------------------------------ ---- DSS Household (Family) Number:-------------- ------ BRAC eligible Household: Yes NO BRAC VO Member Household: Yes NO Name of Respondent: ------------------------------------------ Name of Household Head: ------------------------------------ Gender of Household Head: ---------------------Male Female
Date of Interview: DD MM YY Interviewed by: Name ------------------------------------------ Signature----------------------------------------
209
Household Survey Questionnaire
Village specific questions
This section will be filled up by any village representatives
Questions
1) Village has Grameen Bank Yes No
2) Village has BRAC Yes No
3) Village has Commercial Bank Yes No
4) Village has electricity Yes No
5) Village has paved road Yes No
6) Village has development program
Yes No
7) How far is the thana headquarter?
Within 1-5 kms Within 6-10 kms More than 10 kms
8) How far is the nearest town? Within 1-5 kms Within 6-10 kms More than 10 kms
9) How far is the commercial bank?
Within 1-5 kms Within 6-10 kms More than 10 kms
10) Access to Primary School. In the village Within 1-5 kms Within 6-10 kms More than 10 kms
11) Access to Secondary/Higher Secondary School.
In the village Within 1-5 kms Within 6-10 kms More than 10 Kms
12) Access to Degree college. In the village Within 1-5 kms Within 6-10 kms More than 10 Kms
210
13) Access to water Tube Well Well River Pond/Lake
14)
How far is the medical centre?
In the village Within 1-5 kms Within 6-10 kms More than 10 Kms
15) Is doctor available? In the village Within 1-5 kms Within 6-10 kms More than 10 Kms
16)
What is the average female wage per month?
17) What is the average male wage?
18) Prices of the items of HSQ258
19) Percentage of Hindu population
20) Proportion of village households owing land of 50 decimals or less
58 Household Survey Questionnaire 2
211
Household Survey Questionnaire
Section 1.1
Individual specific questions
a) Marital Status Single Married Sec (b) Widow Separated Divorced b) If married, do you live in a Single family? Extended family? Sec (c) c) If you live in an extended family, how many generations are living together? Section 1.2
Household specific questions
Relationship
with the
borrower
Sex Age Education
Year of
Schooling
Source of
Income59
Income
Per
month
Land
ownership.
(1) (2) (3) (4) (5) (6) (7)
Borrower 1
Others 2
3
4
5
6
7
8
9
59 income from farming, wage labourer etc.
212
Members if not residing with the household
Sex Age Education Source
of
Income60
Income
Per
month
Land
ownership.
(1) (2) (3) (4) (5) (6) (7)
Father of
the
household
head
Mother of
the
household
head
Father of
the
spouse
Mother of
the
spouse
Brother
of the
household
head
Sister of
the
household
head
Brother
of the
spouse
Sister of
the
spouse
60 wage labourer, farming etc.
213
Section 1. 3 Loan Specification
a) Do you have any loans at present Yes No b) If yes, please answer the following questions Source Use of loans
Cash Kind Land Agri Business Con. Wed. Other
Type Value
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Loan1
Loan2
Loan3
Loan4
Friend/relatives =1, Moneylender =2, BRAC =3, GB = 4, Commercial Bank =5, Other =6. Con. Stands for Consumption and Wed. stands for daughter’s wedding
Section 1.4
Borrower specific information (BRAC/GB)
a) How many times did you borrow money from BRAC/GB b) Please specify the following
No.of
loans
Date
started
Date
ended
(contract)
Date
ended
(actual)
Instalment
(Amount) No. of
Instalments PurposeAmount
(Total)
(1) (2) (3) (4) (5) (6) (7) (8)
1
2
3
4
Sec 1.5
214
Section 1.5
Household current savings
Type Amount
(1) (2)
Cash in hand
BRAC
Grameen Bank
Other Banks
Other
Section 1.6
Income from different sources of the family members (Last month)
Income
Generating
Activities
Name
of the
family
membe
rs
No. of
hours
worked
Income from sales Expenditure
Cas
h
In kind Cash In kind
Goods Value
Taka
Goods Value
Taka
(1) (2) (3) (4) (5) (6) (7) (8)
Wage
Agriculture
Langgol chalano
Ropon
Nirano
Kata
Processing
Other
Daily labour other than Agriculture
Net making
Katha salai
Madur tairi
Bamboo and cane work
Egg selling
Milk selling
Chuckles Selling
Veg. garden
215
Mudir dokan
Biri making
Brick breaking
Jogan dai
Poultry
Nursing
Food for work
Overseas
Section 1.7
Assets
Items Land size/
number
Value of the
Items
(1) (2) (3)
1) Type of houses Brick
Tin
Mud
Other
2) Chair
3) Table
4) Cycle
5) Bed
6) Mosquito net
7) Radio
8) TV
9) Fishing net
10) Rickshaw
11) Tube well
12) Sanitary toilet
13) Duck
14) Chicken
15) Cow
16) Goat
17) Mudi shop
18) Mobile phone
19) Other assets: List
216
Household survey Questionnaire 2 (For both male and female borrowers and non-borrowers)
Section 2.1
Household consumption expenditure (Food)
Items Consumption Own
production Gift
Kg
(Monthly)
Price61
Cons
(1) (2) (3) (4) (5) (6)
Cereal
Coarse rice
Fine rice
Flour/Atta
Corn
Other
Pulses and
oil seeds
Mushuri dal
Mung dal
Kashari dal
Kalai
Boot dal
Mator dal
Other
Fruits and
Veg.
Green leaf
Green veg.
Soya bin oil
Mustard oil
Potato
Chilly
Onion
Garlic
Fruits
Others
Spices
Turmeric
Cumin
Coriander
Other
61 The prices of different items may also be collected (if not found from the respondent) from village representatives.
217
Protein
Big fish
Items Consumption
Own
production Gift
Kg
(Monthly)
Price62
Cons
(1) (2) (3) (4) (5) (6)
Small fish
Chicken
Beef
Mutton
Eggs
Cow’s milk
Powder milk
Sugar/Molasses
Others
Tea
Cigarette/Biri
62 The prices of different items may also be collected (if not found from the respondent) from village representatives.
218
Section 2.2
Household consumption expenditure (Non-food)
Items Unit Own
Production
Gift
(1) (2) (3) (4) (5) (6)
Fuel Monthly
Kerosene
Wood
Gas
Electricity
Other
Clothes Yearly
Women’s wear
Men’s wear
Kid’s wear
Education Yearly
Books
School fees
Tutor’s fees
Other
Health Yearly
Medicine
Doctor
Travel to doctor /Hospital
Other
Other
expenditure
Yearly
Utensils
Bedding
Furniture
House repair
Cycle
TV
Radio
Other
219
Items
Unit Own
Production
Gift
(1) (2) (3) (4) (5) (6)
Travel Monthly
Bus
Train
Rickshaw
Entertainment
Cinema
Theatres
Circus
Other
220
Household survey questionnaire 3
(For female borrowers and non-borrowers only)
Section 3.1
General Questions
(This section will only be filled if HSQ1 has not been filled) Identification:
Village Name: -------------------------------------- ----- Bari Name: ------------------------------------------ ---- DSS Household (Family) Number: ------------- ------- BRAC eligible Household: Yes NO BRAC VO Member Household: Yes NO Name of Respondent: ------------------------------------------ Age of the Respondent: ----------------------------------------- Name of Household Head: ------------------------------------ Gender of Household Head: ---------------------Male Female
Date of Interview: DD MM YY Interviewed by: Name ------------------------------------------ Signature----------------------------------------
221
Section 3.2
Ownership and Purchase decision
Items Do you own these
items
Does your husband consult with you in buying or selling those items
Are you allowed to keep the money from the sales/or did you buy from your own money
(1) (2) (3) (4)
Yes No Yes No Yes No
Generally Occasionally Generally Occasionally
House/land
Chicken/Duck
Cow/Goat
Big Plants
Veg. garden
Jewellery
Savings
Boat
Fishing Net
Cycle
Rickshaw
Weaving machine
Sewing machine
Langal
Deep tube well
Shop
Pond
Fish farm
Poultry farm
Repairing house
N/A N/A
Cloth for you N/A N/A
Cloth for the kids
N/A N/A
Household items
N/A N/A
Other items N/A N/A
222
Section 3.3
Mobility
In last four months
did you visit these
places
Yes No If yes
Generally Occasionally On
your
own
Accompanied
by others
(1) (2) (3) (4) (5) (6)
Local Market
Sub divisional town
Father’s place
BRAC office
ICDDR,B office
Court
Health centre
Govt. Office
Other Places
Section 3.4
Family issues
Items Between you and your
Husband
Between you and your in-
laws
(1) (2) (3)
Yes No Yes No
Your money has been taken away from you against your will
Your land/Jewellery/Cow has been taken away from you against your will
You have been forbidden to visit your father’s place
You have been physically abused
223
Section 3.5
General knowledge and use of contraceptives
Items Yes No
(1) (2) (3)
Do you know the name of the local Union Parishad Chairman? (Interviewer must know all the answers before hand)
Do you know the name of the Prime Minister of the Country?
Do you know anybody who has taken or given dowry in their son/daughter’s wedding in your village?
Do you know that dowry is illegal?
Do you know the minimum marriage age for a girl?
Do you know the procedure of divorce?
Women are working outside house now; do you think it is good?
Do you use any contraceptives
Do you take the decision of using it in consultation with your husband?
224
Section 3.6
Some additional questions
Items
(1) (2) (3)
What is the ideal family size in your view?
If you were given another chance how many kids would you have?
What should be the time gap between two kids in your opinion?
What is the number of your surviving children
What is the age of your youngest child?
How old are you?
How old were you when you got married?
How many years of schooling did your parents do?
Do you watch NEWS from TV/Radio or both?
Are you aware of AIDS? Yes No
In your assessment are you in good health
Yes No
In your assessment what is the general condition of your children
Good Not good
How many times have you visited your parent’s house in last year
If you do not visit quite often, is it because of
1. Due to security reason
2. You are unable to go on your own
3. You do not have any one to go with
4. Your husband does not allow you to go
Did you receive any training from BRAC/GB?
Yes No
226
7.5 Appendix E: List of the Villages
The following list shows villages from which data have been collected.
Bangladesh Gazipur 1. Bagia 2. Meghlal 3. Kholapara 4. Borabo 5. Ambagh Dinajpur 1. Sheikh hati 2. Ramjiban pur 3. Paschim Shibrampur 4. Bangshi pur 5. Kamalpur Chokoria 1. Bharamuhuri 2. Binamara 3. Nishpan khali 4. Lakharchar 5. Kahariaghona
227
7.6 Appendix F: Descriptive Tables of consumption patterns Table 1
Descriptive table (borrowers of Gazipur district)
Number of observations 125
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 835.881 755 2200 450 288.23 237.08 (0.00)
1.98 8.42
Pulses 197.12 192 650 0 101.51 42.21 (0.00)
1.00 5.04
Vegetables 561.16 552.5 1130 134 186.22 0.83 (0.65)
0.18 3.13
Meat & Fish 1385.65 1080 5500 165 1028.01 29.64 (0.00)
1.07 3.99
Milk & Eggs 444.23 287.5 1750 18 336.67 16.50 (0.00)
0.87 3.33
Cigarette 363.88 310 1200 0 234.21 191.89 0.00()
0.41 2.94
Cooking Oil 203.07 198 450 36 91.17 6.69 (0.033)
0.55 2.75
Spices 228.34 220 450 58 96.11 6.72 (0.039)
0.54 2.68
Sugar 299.05 280 700 0 170.07 4.58 (0.020)
0.33 2.35
Total Food 4518.41 4001 12745 1308 2275.92 14.21 (0.00)
0.81 3.12
Fuel 176.12 144.5 890 11 110.19 963.61 (0.00)
2.51 15.58
Electricity 194.81 200 800 60 114.17 756.80 (0.00)
2.86 13.54
Clothing 236.43 210 560 50 142.92 11.37 (0.0033)
0.58 2.10
Education 252.32 100 1500 0 363.03 117.11 (0.00)
1.89 5.83
Health 247.46 210 1000 0 207.03 81.86 (0.00)
1.50 5.55
Travel 572.02 400 3100 0 647.64 429.32 (0.00)
2.71 10.23
Entertainment 10 0 500 0 58.88 10893.8 (0.00)
6.44 46.87
Other 427.48 295 2500 0 533.87 280.89 (0.00)
2.37 8.51
Total non-
food 2116.60 1622 9030 231 1835.52 275.58
(0.00) 2.34 8.52
Total expenditure
6635.01 5484 18912 1539 3803.14 37.02 (0.00)
1.21 4..07
Total Production
880.90 765 3200 0 742.04 10.64 (0.006)
0.71 3.06
Family size 4.61 4 10 2 1.32 87.31 (0.00)
1.20 6.29
Family Income
8877.81 6750 27000 2000 5579.32 43.40 (0.00)
1.29 4.13
The figures in the parentheses represent probability.
228
Table 2
Descriptive table (non-borrowers of Gazipur district)
Number of observations 54
Mean Median Max. Min. St. dv. Norm. Skew
.
Kurt.
Cereal 745.28 695 2500 450 291.65 1345.37 (0.00)
4.26 26.41
Pulses 195.53 194 420 0 79.10 1.38 (0.50)
0.35 3.38
Vegetables 507.82 543 950 185 138.28 2.64 (0.30)
0.21 4.02
Meat & Fish 624.71 550 2600 100 469.15 97.18 (0.00)
2.07 8.25
Milk & Eggs 193.17 187.5 555 30 102.40 26.29 (0.88)
1.20 5.51
Cigarette 265.48 250 600 0 131.67 0.18 (0.63)
0.13 3.12
Cooking Oil 182.81 184 380 56 68.58 3.78 (0.15)
0.63 3.35
Spices 206.03 205 350 90 57.63 0.023 (0.98)
0.04 2.94
Sugar 137.07 115 350 0 83.43 2.75
(0.25)
0.50 2.48
Total Food 3057.23 3078.5 7956 1085 1168.32 63.58 (0.00)
1.59 7.3
Fuel 131.40 125 282 22 49.13 5.59 (0.06)
0.50 4.21
Electricity 166.63 192.5 250 50 45.10 5.50 (0.04)
-0.76 2.56
Clothing 123.75 102.5 310 20 71.47 14.13 (0.02)
1.24 3.57
Education 95.57 30 400 0 118.26 8.46 (0.014)
0.97 2.74
Health 189.28 130 1000 20 163.06 303.67 (0.00)
2.81 13.41
Travel 488.07 450 1200 100 256.00 4.09 (0.13)
0.68 3.04
Entertainment 5.55 0 200 0 30.19 2575.62 (0)
5.66 34.88
Other 325.28 300 700 0 182.60 0.69 (0.70)
0.21 2.64
Total non-
food
1525.67 1445 2608 430 484.80 0.42 (0.78)
0.18 2.74
Total
expenditure
4582.90 4386 9230 1515 1407.26 4.89 (0.06)
0.58 3.93
Total
Production
632.98 600 1660 0 471.38 2.40 (0.27)
0.31 2.15
Family size 4.38 4 9 2 1.23 25.99 (0.00)
1.11 5.65
Family
Income
4450.81 4000 11667 1400 1755.04 56.50 (0.00)
1.57 6.89
The figures in the parentheses represent probability.
Table 3
229
Descriptive table (Both borrowers and non-borrowers of Gazipur district)
Number of observations 179
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 809.41 750 2500 450 291.35 902.23 (0.00)
2.57 12.75
Pulses 196.65 192 650 0 95.09 56.22 (0.00)
0.92 5.05
Vegetables 545.58 550 1130 134 174.90 3.62 (0.15)
0.28 3.40
Meat & Fish 1163.35 750 5500 100 964.45 83.45 (0.00)
1.38 4.87
Milk & Eggs 370.88 242 1750 0 310.12 63.40 (0.00)
1.29 4.37
Cigarette 335.14 300 1200 0 213.88 12.89 (000)
0.61 3.47
Cooking Oil 196.94 196 450 36 85.54 12.15 (0.00)
0.63 3.03
Spices 221.83 210 450 58 86.87 12.47 (0.001)
0.64 3.17
Sugar 251.73 220 700 0 167.00 13.84 (0.000)
0.66 2.69
Total Food 4091.55 3318 12745 1085 2120.22
44.41 (0.00
1.12 3.96
Fuel 163.02 135 890 11 98.56 2048.11 (0.00)
2.76 18.66
Electricity 186.58 200 800 50 99.78 1855.43 (0.00)
3.20 17.46
Clothing 203.51 150 560 20 136.15 25.03 (0)
0.90 2.70
Education 206.53 100 1500 0 319.78 349.55 (0.00)
2.30 8.08
Health 230.47 200 1000 0 196.56 198.29 (0.00)
1.77 6.75
Travel 547.5 400 3100 0 562.64 1048.01 (0.00)
3.03 13.21
Entertainment 8.65 0 500 0 51.87 22281.22 (0)
6.94 55.86
Other 397.62 300 2500 0 461.58 763.22 (0.00)
2.75 11.51
Total non-
food 1943.97 1512.5 9030 231 1587.1
28 832.01 (0.00)
2.84 11.93
Total expenditure
6035.52 5103.5 18912 1515 3415.24
119.71 (0.00)
1.57 5.49
Total Production
808.47 705 3200 0 682.39 22.46 (0.00001)
0.83 3.52
Family size 4.54 4 10 2 1.29 116.03 (0.00)
1.18 6.16
Family Income
7592.85 5875 27000 1400 5188.71
116.03 (0.00)
1.67 6.16
The figures in the parentheses represent probability.
230
Table 4
Descriptive table (borrowers of Dinajpur district)
Number of observations 132
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 874.78 797.5 2450 350 344.35 231.91 (0.00)
1.96 8.16
Pulses 191.78 177 540 0 82.30 35.62 (0.00)
0.84 4.91
Vegetables 520.84 500 1200 200 168.95 37.53 (0.00)
0.97 4.74
Meat & Fish 719.68 450 3000 120 644.73 40.29 (0.00)
1.48 4.52
Milk & Eggs 227.15 150 950 0 220.95 107.71 (0.00)
1.80 5.55
Cigarette 185.66 200 680 0 171.88 16.67 (0.05)
0.85 3.28
Cooking Oil 165.07 144 550 48 76.34 116.42 (0.00)
1.30 6.78
Spices 169.43 165 560 50 75.13 155.66 (0.00)
1.44 7.46
Sugar 147.18 99.5 650 0 153.40 87.19 (0.00)
1.62 5.2
Total Food 3201.61 2733 9510 1070 1687.62 78.24 (0.00)
1.54 5.217
Fuel 143.19 120 480 11 88.55 57.67 (0.00)
1.28 4.97
Electricity 68.03 0 320 0 89.69 23.86 (0.00)
1.04 2.96
Clothing 182.11 120 750 20 148.29 134.50 (0.00)
1.84 6.28
Education 92.91 0 1250 0 203.88 1160.30 (0.00)
3.37 15.86
Health 153.71 120 550 0 112.14 66.07 (0.00)
1.38 5.08
Travel 323.48 200 2000 0 426.93 272.95 (0.00)
2.21 8.47
Entertainment 2.27 0 200 0 19.40 41462.64 (0.00)
9.08 87.90
Other 164.20 100 2000 0 274.09 2247.43 (0.00)
3.78 21.74
Total non-
food
1129.93 876.5 6290 64 1086.26 374.64 (0.00)
2.34 9.78
Total
expenditure
4331.54 3631 15460 1415 2672.75 161.47 (0.00)
1.87 6.92
Total
Production
217.42 95.5 1060 0 276.34 50.42 (0.00)
1.40 4.14
Family size 5.8 5 12 3 1.92 14.58 (0.00)
0.80 3.25
Family
Income
5257.62 4500 20000 1500 3089.17 178.12 (0.00)
1.81 7.39
The figures in the parentheses represent the probability.
231
Table 5
Descriptive table (non-borrowers of Dinajpur district)
Number of observations 63
Mean Median Max. Min. St. dv. Norm. Ske. Kurt.
Cereal 926.09 780 2300 450 454.42 44.04 (0.00)
1.72 5.27
Pulses 201.33 188 660 44 109.35 82.26 (0.00)
1.68 7.53
Vegetables 505.48 477.5 1250 200 168.70 87.98 (0.00)
1.63 7.83
Meat & Fish 461.61 240 3000 65 557.22 198.36 (0.00)
2.57 10.09
Milk & Eggs 166.17 118 1100 15 160.71 837.21 (0.02)
3.45 19.62
Cigarette 153.70 150 650 0 143.86 15.62 (0.20)
1.02 4.36
Cooking Oil 173.27 167.5 450 50 72.87 23.34 (.00)
1.03 5.17
Spices 154.19 145 330 50 62.52 5.38 (0.05)
0.72 3.03
Sugar 101.37 36 650 0 132.93 143.11 (0.00)
2.37 8.73
Total Food 2843.25 2356.5 9780 1159 1556.78 122.23 (0.00)
2.10 8.43
Fuel 114.06 104 380 12 72.96 32.41 (0.00)
1.31 5.37
Electricity 63.87 0 280 0 83.43 10.10 (0.01)
0.98 2.82
Clothing 133.82 100 550 20 111.36 121.58 (0.00)
2.22 8.22
Education 56.45 0 500 0 95.43 153.67 (0.00)
2.26 9.24
Health 132.5 100 500 0 106.35 51.02 (0.00)
1.64 5.99
Travel 229.83 200 3000 0 406.91 2946.01 (0.00)
5.20 35.35
Entertainment 13.49 0 600 0 81.42 5134.88 (0.00)
6.47 45.29
Other 109.83 0 1000 0 180.89 225.00 (0.00)
2.43 10.96
Total non-
food
854.09 592 6780 184 937.11 1674.09 (0.00)
4.28 26.97
Total
expenditure
3697.35 3002.5 16560 11412
2413.12 429.00 (0.00)
2.88 14.52
Total
Production
410.69 300 2928 0 508.58 223.61 (0.00)
2.34 11.02
Family size 6.66 6 15 3 2.78 19.19 (0.00)
1.26 4.03
Family
Income
3862.14 3100 20333 1000 2814.65 874.98 (0.00)
3.44 19.90
The figures in the parentheses represent the probability.
232
Table 6
Descriptive table (both borrowers and non-borrowers of Dinajpur district)
Number of observations 195
Mean Median Max. Min. St. dv. Norm. Skew Kurt.
Cereal 889.27 785 2450 350 382.46 259.39 (0.00)
1.93 7.12
Pulses 194.29 178 660 0 91.70 197.94 (0.00)
1.35 7.13
Vegetables 515.08 500 1250 200 168.57 102.90 (0.00)
1.18 5.65
Meat & Fish 634.78 350 3000 65 627.65 147.76 (0.00)
1.72 5.50
Milk & Eggs 207.85 135 1100 0 204.70 326.23 (0.00)
2.15 7.93
Cigarette 175.83 180 680 0 163.42 30.98 (0.01)
0.92 3.60
Cooking Oil 167.57 155 550 48 74.98 138.81 (0.00)
1.22 6.33
Spices 164.23 150 560 50 71.54 186.41 (0.00)
1.34 6.96
Sugar 132.05 72 650 0 148.16 181.44 (0.00)
1.81 6.02
Total Food 3080.99 2595 9780 1070 1649.37 165.34 (0.00)
1.69 5.96
Fuel 133.77 113 480 11 84.97 100.32 (0.00)
1.33 5.27
Electricity 67.13 0 320 0 87.52 33.65 (0.00)
1.02 2.92
Clothing 166.33 110 750 20 138.84 262.53 (0.00)
1.99 7.05
Education 80.84 0 1250 0 176.79 2824.26 (0.00)
3.74 20.07
Health 146.56 100 550 0 110.32 113.69 (0.00)
1.46 5.32
Travel 294.10 200 3000 0 421.68 1534.24 (0.00)
3.05 15.31
Entertainment 5.89 0 600 0 48.99 108135 (0.00)
10.20 116.54
Other 147.10 0 2000 0 248.34 3891.4 (0.00)
3.81 23.50
Total non-
food
1033.03 822 6490 94 986.72 1159.08 (0.00)
2.79 13.55
Total
expenditure
4122.74 3450 16560 1412 2597.98 404.91 (0.00)
2.13 8.62
Total
Production
278.89 120 2928 0 375.55 1429.55 (0.00)
2.64 15.16
Family size 6.07 6 15 3 2.26 81.42 (0.00)
1.27 4.88
Family
Income
4806.77 4000 20333 1000 3066.58 495.42 (0.00)
2.14 9.52
The figures in the parentheses represent the probability.
233
Table 7
Descriptive table (borrowers of Chokoria district)
Number of observations 130
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 724.17 685 1375 310 213.34 12.18 (0.002)
0.68 3.57
Pulses 196.97 200 410 45 80.27 4.49 (0.099)
0.44 2.83
Vegetables 448.71 450 1200 150 149.40 96.62 (0.00)
1.06 6.62
Meat & Fish 875.37 595 2675 0 706.50 27.79 (0.000)
1.12 3.12
Milk & Eggs 296.70 200 975 0 230.42 32.09 (0.00)
1.18 3.49
Cigarette 269.42 200 850 10 184.52 23.79 (0.00)
1.00 3.57
Cooking Oil 167.51 170 360 50 55.34 2.88 (0.17)
0.28 3.55
Spices 160.19 150 350 50 57.37 7.42 (0.021)
0.55 3.35
Sugar 247.97 210 680 16 169.33 16.72 (5.213)
0.87 2.91
Total Food 3387.05 2998 7889 935 1304.63 19.00 (0.000075)
0.87 3.65
Fuel 168.61 156 360 20 71.17 2.21 (0.32)
0.20 2.50
Electricity 131 150 500 0 105.86 2.66 (0.26)
0.35 3.03
Clothing 181.65 165 500 20 92.26 19.28 (0.000)
0.90 3.54
Education 141.53 120 600 0 149.82 20.43 (0.000)
0.95 3.33
Health 205.61 200 600 0 128.26 25.23 (0.000)
1.02 3.67
Travel 373.84 300 2000 0 333.59 293.66 (0.00)
2.06 9.09
Entertainment 0 0 0 0 0 - - -
Other 262.76 200 1500 0 295.63 189.83 (0.00)
1.85 7.61
Total non-
food
1465.03 1297.5 5120 175 927.72 88.35 (0.00)
1.48 5.74
Total
expenditure
4800.7 4324 11666 1175 2053.77 20.88 (0.000)
0.91 3.68
Total
Production
156.32 0 1590 0 284.25 303.91 (0.00)
2.27 8.94
Family size 5.64 6 11 2 1.49 3.69 (0.15)
0.16 3.76
Family
Income
6150.19 5500 17000 1300 2862.91 51.75 (0.00)
1.19 4.96
The figures in the parentheses represent the probability
234
Table 8
Descriptive table (non-borrowers of Chokoria district)
Number of observations 67
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 718.78 667.5 1565 425 219.27 36.00 (0.00)
1.37 5.39
Pulses 192.06 180 350 72 83.17 5.61 (0.06)
0.36 1.76
Vegetables 391.96 365 650 200 106.38 2.29 (0.28)
0.38 2.50
Meat & Fish 485.68 382.5 1750 125 332.85 97.56 (0.000016)
2.00 7.39
Milk & Eggs 209.29 200 500 0 87.43 25.99 (0.00002)
0.81 5.57
Cigarette 201.89 200 650 15 110.76 33.19 (0.66)
1.06 5.74
Cooking Oil 149.42 135 260 90 47.33 4.95 (0.08)
0.54 2.05
Spices 140.30 130 240 80 46.93 5.09 (0.077)
0.37 1.86
Sugar 141.65 94 550 12 120.62 21.13 (0.00)
1.27 4.09
Total Food 2637.95 2356 5935 1379 969.88 19.97 (0.000046)
1.61 5.49
Fuel 181.19 177.5 320 32 73.47 1.84 (0.3587)
0.02 2.18
Electricity 117.53 150 250 0 89.48 6.29 (0.0431)
-0.23 1.56
Clothing 130.84 105 320 50 63.27 18.67 (0.2759)
1.24 3.79
Education 101.21 0 400 0 129.31 8.68 (0.0109)
0.75 2.06
Health 188.25 200 500 50 78.85 31.07 (0.000001)
1.18 5.37
Travel 368.18 300 1500 0 250.02 108.28 (0.00)
1.73 8.22
Entertainment 0 0 0 0 0 - - -
Other 262.12 200 1000 0 211.88 9.86 (0.00721)
0.80 3.98
Total non-
food
1348.93 1276.5 4040 395 619.52 61.42 (0.00)
1.38 6.83
Total
expenditure
3986.89 3658 9975 1911 1455.60 45.82 (0.00011)
1.51 6.3
Total
Production
289.51 0 1480 0 393.32 16.02 (0.000537)
1.19 3.38
Family size 5.86 5 11 2 2.30 4.43 (0.1134)
0.41 2.03
Family
Income
4498.23 4100 10000 2000 1612.33 11.16 (0.00513)
0.88 3.95
The figures in the parentheses represent the probability.
235
Table 9
Descriptive table (both borrowers and non-borrowers of Chokoria district)
Number of observations 197
Mean Median Max. Min. St. dv. Norm. Skew. Kurt.
Cereal 722.37 685 1585 310 215.00 40.11 (0.00)
0.92
4.21
Pulses 195.32 200 410 45 81.07 8.07 (0.01)
0.41 2.44
Vegetables 429.70 430 1200 150 138.94 155.45 (0.00)
1.09 6.76
Meat & Fish 744.81 550 2675 0 633.88 90.48 (0.00)
1.50 4.37
Milk & Eggs 269.72 200 975 0 203.31 101.57 (0.00)
1.51 4.77
Cigarette 246.80 200 850 10 166.35 67.17 (0.00)
1.22 4.49
Cooking Oil 161.45 150 360 50 53.77 5.32 (0.06)
0.38 3.22
Spices 153.52 150 350 50 55.02 12.04 (0.002429)
0.58 3.29
Sugar 212.35 180 680 12 162.40 37.05 (0.00)
1.04 3.39
Total Food 3136.08 2659 7675 922 1459.78 76.69 (0.00)
1.21 3.69
Fuel 196.16 165 650 20 113.82 155.55 (0.128)
1.53 6.07
Electricity 126.42 150 500 0 100.38 2.017 (0.3645)
0.24 2.88
Clothing 163.22 120 600 20 101.89 128.88 (0.000001)
1.50 5.57
Education 127.30 100 600 0 144.10 29.99 (0.00)
0.94 3.24
Health 198.95 200 600 0 114.40 61.26 (0.00)
1.16 4.42
Travel 370.55 300 2000 0 307.17 492.97 (0.00)
2.06 9.55
Entertainment 0 0 0 0 0 - - -
Other 261.21 200 1500 0 269.52 287.01 (0.00)
1.74 7.77
Total non-
food 1443.86 1275 5450 175 903.62 225.96
(0.00) 1.71 6.67
Total expenditure
4579.94 3934 12135 1115 1856.07 51.81 (0.00)
1.33 4.48
Total Production
209.26 0 1590 0 335.65 162.94 (0.00)
1.76 5.71
Family size 5.72 6 11 2 1.80 5.71 (0.057)
0.41 3.06
Family Income
5575.67 5000 17000 1300 2632.89 133.20 (0.00)
1.40 5.88
The figures in the parentheses represent the probability. Max., Min., Stdy., Norm., Skew., and Kurt., are respectively the abbreviation for maximum, minimum, standard deviation, normality, skewness and Kurtosis used in here.
236
7.7 Appendix G: Experiences from Field Trip
My experience from fieldwork
Sampling Procedure
Throughout the course of our fieldwork study, my team and I interviewed six
hundred and ninety-five borrowers and non-borrowers from all over Bangladesh.
The study drew upon the population from four major districts of Bangladesh. These
were based on different agro climatic and economic conditions. The districts were
Gazipur, Dinajpur, Khulna and Chittagong. A cluster of villages from each district
was selected. The household samples were selected using multistage stratified
random sampling from each village in each district. The data collection took place in
July 2004. During this period, nearly two-thirds of the country was under water.
Before leaving for Bangladesh, I had anticipated that I would encounter problems
associated with the monsoon season. However, I did not expect to experience
flooding during my stay. Nevertheless, this was the only time that I could manage to
go for fieldwork between semesters.
My research team consisted of three girls and two boys from Jahangirnagar
University, who assisted me in conducting the fieldwork. We visited several villages
in order to conduct our research. These included Gazipur, located near Dhaka,
Dinajpur, which covers the Northern area, Chokoria, situated near the coast and
Khulna, which lies near south-west of Bangladesh (appendix 1). Interestingly
enough, we only found seven male borrowers in our total surveyed sample as the
vast majority were female.
237
2. My experience from fieldwork
I underwent several exciting experiences as I visited the numerous villages of
Bangladesh. The villagers assumed that we represented the “Sarker63” (Bangladesh
Government). They were convinced that we were collecting information to pass on
to the Government, because we kept asking questions about their assets, incomes and
expenditures.
We repeatedly mentioned that the information that we were gathering would be used
strictly for academic purposes. However, the villagers were not at all convinced by
our remarks. They believed that our research would somehow reach the Government,
who would in turn take care of their problems. In the Chokoria district, one village
representative took me to visit a few of the poorest families and suggested that I
videotape their poor living conditions in order to show the Government. They
expected that I would return with grants, some time in the future. In almost all cases,
the villagers inquired about how they would be reimbursed for their time and service
to us.
The experience was very different in Gazipur. The majority of locals harboured
suspicions regarding our work and were thus unwilling to talk to us. They believed
that we were income tax people, because we wanted information related to their
financial situation.
63 “Sarker” is the Bengali term for the Government.
238
In Gazipur, we were easily able to determine which people were borrowers, because
their clothes, furniture and home accessories provided us with a clear indication of
their present financial situation. We observed that the borrowers were relatively
wealthy and vice versa. All the villages in the Gazipur district were equipped with
electricity, proper toilets, as well as a large range of electrical equipment. These
included colour televisions, stereos, mobile phones and even personal computers.
Gazipur appeared to be the richest place that we had surveyed in terms of health and
wealth of the residents. We noted that the infant mortality rate was much lower in
Gazipur than the other areas that we visited.
Since Gazipur is situated so close to Dhaka, we observed a certain ‘capital-city
influence’ reflected in the attitudes of the local villagers We experienced several
difficulties while trying to get people to talk about personal aspects of their lives,
such as their incomes and expenses. Getting an exclusive interview with borrowers
proved to be an impossible task for us. This was primarily due to husbands of female
borrowers forbidding their wives to disclose private matters to ‘outsiders’.
The district of Dinajpur appeared to be significantly poorer than Gazipur. Villages in
Dinajpur did not have electricity or sanitary toilets. Borrowers were mostly factory
workers, daily labourers and blacksmiths. The crop fields in the villages did not
belong to any of the borrowers. We found the people of Dinajpur to be very friendly,
approachable and down to earth. They took us inside the house offered us a variety
of food items including cow’s milk, rice crisps, jackfruit and other home-grown
fruits. They possessed very few material items, but appeared to be very content with
what they had. The local people were very excited, because they were being filmed.
239
One particular elderly woman even went to the trouble of putting on her best blouse
in order to be captured on film.
One of our most interesting memories of Dinajpur was being able to witness girls
riding bicycles. This scene was unique to the area and would rarely be found
elsewhere in the country. The intervention of Non-Government Organisations
(NGOs) such as BRAC64, led to the introduction of bicycles for the female field-
workers. Even more unusual was our witnessing of girls that were wearing “hijab”65,
riding bikes in the same area. The significance of such seemingly simple scenarios
may not be obvious to everyone, so one must keep the conservative nature and
culture of the locals in mind to fully grasp their importance.
Among the many villages that we visited in the Chokoria district, we came across a
certain village resided by Muslim majority. It was a beautiful village surrounded by
coconut trees. We found a number of NGOs working together at this particular
village. Every household had borrowers from the NGOs. Sometimes, they borrowed
from more than one NGO at the same time.
In the same village, two of our male co-workers found it exceedingly difficult to
interview the female borrowers, because the villagers were very strict Muslims.
According to Islam, females are not allowed to see non-family members without
“parda”66. The female borrowers spoke to the male interviewers from behind doors
64 BRAC stands for Bangladesh Rural Advancement Committee 65 ‘Hijab’ is the head cover used by Muslim women. 66 “Parda” is used to define restrictions in Islam that separate females from males other than their relatives.
240
or fences. The replies to the questions that they were asked were conveyed back to
the interviewers via their husbands.
Some female borrowers told us not to take their photographs, as they did not seek
permission from their husbands. We spoke to one of the schoolteachers working for
BRAC in this village. He mentioned that they always encounter these type of
problems when foreign donors come to visit this village, because the borrowers do
not talk to strangers.
We also visited some villages resided by Hindu and Buddhist majority in the
Chokoria district. The villagers took us to some of their temples and requested that
we videotape the experience. Throughout our journey through the various villages of
the Chokoria district, we witnessed people using water from ponds that was
extremely dirty. They said that they possessed just one tube-well, which was arsenic
free and usable for drinking water.
The Buddhist village that we visited was called “Borua para”. The main occupation
of the villagers was to catch crabs that would ultimately be sold in the international
market. Loans from the Grameen Bank enabled these people to establish such
business.
2.1 Problems in collection data
The time that we had selected for data collection was not at all congenial for
fieldwork. We walked mile after mile through rain, mud and even through flood-
241
water. We realised that carrying out such intensive fieldwork requires an enormous
amount of dedication and perseverance. When we interviewed people we tried to
avoid interrupting their daily routines. Thus, we spoke to them while they cooked,
milked their cows or washed their clothes.
We encountered a number of problems while we were collecting data. Firstly we
wanted to investigate typical household expenses in a certain amount of detail. This
proved to be exceedingly hard for them to explain, as the villagers did not live by a
fixed budget and did not keep track of their daily expenses. Some people felt
uncomfortable when we asked questions about their monthly expenditure. Others
were unable to reply because they could not remember their previous month’s
expenditure. Certain items such as milk, eggs, chicken and fish were produced at
home. This made it difficult for them to estimate their average consumption. We
therefore, had to cross-question them in order to find out the actual amount. We
faced similar problems with the use of cooking wood, as villagers often gathered dry
leaves and wood from neighbours. We also had difficulties with peoples’ incomes.
Not all of them were wage earners, so they were unable to provide details about their
incomes. In many cases the figure that we received would be an over or under-
estimation of the actual figure. We tried calculating the figures from other questions.
In our questionnaires, we included a number of items, which we expected that the
villagers would consume, but in reality, we observed that they consume only a small
variety of common items.
In almost all the cases, the female villagers were unable to provide their husband’s
age or even their own age. We tried finding that from their eldest child’s age and the
242
age at which they were married. Surprisingly, they were also unable to provide us
with general information about their village such as the approximate distance from
one place to another.
Apart from the previously stated difficulties, we also encountered problems relating
to effective communication. We found it difficult to understand various regional
dialects that were specific to various parts of Bangladesh. Although the entire
country speaks a single language (Bengali), these dialects are significantly different
from each other.
My inexperience relating to village life is now very apparent to me and I feel that
this is reflected in the questionnaire, which I prepared prior to my arrival. To record
the type of the houses, I listed four categories: brick, tin, mud and straw houses in
the questionnaire. In reality, we found that most houses were made up of different
combinations of products. Notable features included straw roofs, mud walls and mud
floors. I now realise that I could have asked questions about the type of roof, walls
and floor separately. This in turn, would have greatly assisted me in gathering
information from the locals.
3. Beyond credit program
Do micro credit programs really help the poor? This question cannot be answered in
a single sentence. The impact that micro credit programs have had on the rural
economy has produced mixed results. I spoke to many female borrowers who were
satisfied with the program. Most of them had been borrowers for around ten to
243
fifteen years. It made them extremely happy to be given the opportunity to borrow
money in order to gather assets. They were glad to have tin roofs, tube wells and
land, which would not be possible without the financial assistance received from the
micro credit program. They were also content with the fact that they were now able
to posses a number of cows, goats and chicken. In addition, they stated that they
would not be able to save up to own properties unless they were offered these types
of loans.
We observed that the relatively wealthy families were the ones who were benefited
by the program. Similarly, the long-term borrowers also appeared to be in better
financial condition.
An interesting observation that we made was that although NGOs lend money to
women, in most cases this money is used and managed by the male members of the
family. Generally, the female borrowers borrowed money while their husbands
invested in businesses. The ratio of women whose money was managed by someone
other than themselves to those that handled their own finances was roughly 9:1. As a
result, one may ask why NGOs still continue to lend money to women. This question
needs to be investigated further.
However, there were still some cases where women had successfully borrowed and
managed money. Two such examples included a woman that I met in Gazipur and
one that we met in Dinajpur. Both these women had invested their money wisely and
were thus able to accumulate assets on their own. Such experiences proved to be
244
very empowering for women as they gained confidence in their abilities as
individuals.
The micro credit program has been proven to be highly successful in terms of
recovery. We did not find a single person that had become a defaulter at any time,
because the Grameen Bank takes every measure to ensure that nobody misses his or
her weekly instalment. The field workers who collect the weekly instalments do not
leave until the person pays the money that is owed. They often spend the entire day
focusing on a single debtor. According to borrowers, the debt collectors make sure to
count every single paisa67 and do not express any leniencies or make exceptions for
anyone. Field workers are accountable for the recovery, so at times they may behave
harshly or act rude towards borrowers. In some cases, there were even reports about
fighting breaking out among the field workers and borrowers. It has also been
reported that if the borrowers were unable to pay the money, field workers even took
the tin off the roofs. Poorer borrowers often sold their cows or their jewellery in
order to repay the money. Sometimes other members of the group lent money to the
borrowers in order to help them out. It was also reported that the borrowers
frequently borrowed money from moneylenders so that they could pay off their loan.
Even more interesting was the fact that these moneylenders were also borrowers
from the Grameen Bank, who lend money to people in need. They do not invest the
money borrowed from the Grameen Bank but use it for lending to others. The
interest rate that they charge varies according to peoples’ needs. They charge higher
interest rates when the necessity of the situation is too extreme. Judging from this
67 Paisa is the lowest denomination of Bangladeshi currency.
245
example, one could question whether the Grameen Bank patronises moneylenders
instead of removing them.
One of the new schemes offered by the Grameen Bank is to lend money for buying
mobile phones that will be used commercially. The borrower places the phone in a
shop nearby and people are allowed to use it in case of necessity. Although this
scheme may sound appealing several borrowers in Gazipur expressed their
dissatisfaction with the program. According to the contract, the Grameen Bank lends
25,000 Taka68 to the borrowers in order to buy a mobile phone. However, for the
first three months they are not allowed to keep any profit earned. All revenue earned
from the phone would have to be returned to the bank. They are allowed to keep
profits, three months after paying the weekly instalment. The lady that we spoke to
did not have any problems with this program, because she had other income from her
family to support her. However, she expressed concern about others who sought to
support themselves from such investments. She felt that poor people could not
possibly get through the three months if banks did not allow them to keep any of the
profit.
3.1 Savings and lending mechanism
Grameen Bank and the BRAC borrowers save a certain amount of their money apart
from their weekly instalment. These savings are kept in the bank. Recently the
Grameen Bank has introduced the Deposit Pension Scheme (DPS), which matures in
five to ten years. BRAC also has some long-term savings schemes. The BRAC
borrowers reported that BRAC does not allow them to withdraw their savings even
68 Taka is the currency of Bangladesh. One US Dollar is equivalent to approximately sixty Taka.
246
after they mature. On the other hand, most Grameen Bank borrowers do not
experience such troubles. Many borrowers are now withdrawing membership from
BRAC because of this. There are several NGOs such as ASA, PROSHIKA,
KARITAS, ANSAR and VDP working within the same villages. People are
therefore moving from BRAC to other NGOs. These NGOs are trying to overcome
the problems that the old NGOs have had in the past. Villagers reported that the new
NGOs are much more relaxed about making repayments. They allow a few days to
repay the money and are not as strict as the Grameen Bank and BRAC.
Both the Grameen Bank and BRAC allow people to take out an additional loan when
their first loan is almost repaid. The Grameen Bank waives all debt during the event
of a borrower’s death. As mentioned earlier, in most cases, the husbands of female
borrowers generally manage their money. Thus, attempts are being made to
introduce a similar debt-waiver in the event of a husband’s death.
Interestingly, many borrowers reported that the field workers were very persuasive in
their attempts to convince them to take out loans. However their attitudes changed
when the borrowers had trouble making repayments. It has also been reported that
the Grameen Bank field workers always try to convince people to borrow money and
thus remain indebted at all times. Once a loan has been repaid, they relentlessly
attempt to convince the borrower to take out yet another loan. The field workers
informed my co-workers that the administrative and overhead costs are the same
whether people are borrowing money or not. They therefore prefer to provide more
loans so that the bank can make more money.
247
3.2 Physical abuse
Rural Bangladesh is not free from physical abuse towards women. Our
questionnaires included a question, which asked whether the female interviewee had
ever been physically abused by her husband, in-laws or other relatives. We were
surprised to find out how many women answered “yes” to the question. In some
cases they even told us how they are being abused. Once, while we were visiting the
village of “Shibrampur” in the Dinajpur district, my co-workers and I noticed that we
were being followed by a very beautiful young lady with watery eyes. She told us
that her husband physically abused her all the time. She could not give us a reason to
explain why she was being abused. Furthermore, she mentioned that whenever her
brother came over, her husband told him to take her back. Being a traditional Hindu,
she was not permitted to get divorced as it was against her religion. She therefore,
has to stay with him forever and accept her fate.
3.3 Dowry
Dowry is also another interesting issue. The Grameen Bank has “sixteen decisions69”
that each borrower must recite at every meeting. One of these decisions was that “we
will not give or take dowry in our daughter’s/son’s marriage”. Unfortunately the
dowry custom is still widely common in rural Bangladesh. We asked borrowers
whether they had heard of anyone in the local community who had either given or
taken dowry. The usual reply that we received was that the wedding could not have
even taken place without a dowry. A typical situation is that a man with no education
69 Please see Appendix B.
248
and no job asks for a dowry of at least fifty to a hundred thousand Taka from the
poor girl’s father. This figure increases with the groom’s educational qualifications
and wealth. To marry off their daughters, parents must sometime sell the last of their
assets or borrow from the Grameen Bank.
In Dinajpur, we met a borrower who was at the time working as a daily labourer.
She never worked outside her own house before. She used the borrowed money to
pay the dowry for her daughter’s wedding. Her new son in-law demanded more
money, so she had to take out a new loan in order to pay him off. The borrower had
to start working outside the house so that she could pay off the debt. She mentioned
that she and her husband could barely afford a single meal per day after paying off
the weekly instalment and they still had several years left to pay off the debt.
In every village, we saw borrowers working hard to pay off the debts. Many villagers
borrowed money for investment but in the end this money was used for
consumption. In most of these cases, the money was spent on dowry. As a result they
must do extra work in order to gather money to pay off their debts. Being trapped in
debt has had been very detrimental to the health and well-being of the borrowers.
Mental anxiety and stress have led to villagers reporting that they were unable to
sleep at night. One village representative reported that the general health of the
women has deteriorated since they became borrowers. The same representative
stated that, like so many others, he wished that the dowry was removed from society.
249
7.8 Appendix H: Paper Published in the Fourth International Business and
Research Conference Proceedings
Microcredit Programs and Economic Indicators: Are the Higher Income
Borrowers Better Off? Evidence from Bangladesh
Sayma Rahman
School of Economics &Finance University of Western Sydney
Locked Bag 1797, Penrith South DC NSW 1797, Australia Tel +61 2 9831 8020 Fax +61 2 4620 3787
Email: [email protected]
Abstract
The microcredit program in Bangladesh provides small loans to rural people especially to the poor with the purpose of eradicating poverty. This study investigates the impact of microcredit on economic indicators of the borrowers and compares if the impact is the same across borrowers having different income levels. Household savings, assets and income are considered as causal factors that may contribute towards eradicating poverty. To estimate the impact of microcredit on such indicators we have used simultaneous equations model. Primary data has been collected from the Grameen Bank and the Bangladesh Rural Advancement Committee (BRAC) borrowers of some selected villages from three major districts in Bangladesh. The Two-Stage Least Squares (2SLS) estimation results show that the microcredit programs is effective in generating higher income, assets and savings for the borrowers in general. However, that impact is not found to be uniform across income levels of the borrowers and higher income borrowers seem to be better off compared to the middle and lower income borrowers. This study further shows that the age and education of the household head and his/her partner in the family are significant in bringing about better household impact.
I would like to thank Professor P. N. (Raja) Junankar and Dr. Girijasankar Mallik for their comments on earlier drafts of the paper. Any errors are my responsibility.
250
7.9 Appendix I: Paper accepted for publication in the Journal of Developing Areas in a forthcoming issue
Is there a significant difference between the consumption behaviour of the
borrowers of microcredit programs compared to the non-borrowers? A
Bangladesh experience
Sayma Rahman*
School of Economics &Finance University of Western Sydney
Locked Bag 1797, Penrith South DC NSW 1797, Australia Tel +61 2 9831 8020 Fax +61 2 4620 3787
Email: [email protected]
Abstract
This paper investigates the consumption behaviour of the borrowers of two
major microcredit institutions in Bangladesh and compares that with the non-borrowers. Primary data has been collected from the borrowers of the Grameen Bank and Bangladesh Rural Advancement Committee (BRAC) operating in three major districts of Bangladesh. Along with the borrowers, data has also been collected from the non-borrowers of the same village to compare the differences in consumption patterns between the two groups. This study analyses the impact of per capita monthly expenditure and other household characteristics on the budget share of eleven items (food and non-food) consumed by the borrowers and non-borrowers. Results from the estimation on linear and quadratic model suggest that borrowers of microcredit programs are better off in terms of consumption than the non-borrowers.
JEL Classification: C81, D19
Keywords: Microfinance, elasticity
* I would like to thank Professor P. N. (Raja) Junankar and Dr. Girijasankar Mallik for their comments on earlier drafts of the paper. Any errors are my responsibility.
251
7.10 Appendix J: Paper published in the Conference Proceedings, Monash
University
Comparing the Consumption Pattern of Micro-credit Borrowers and Non-
borrowers: Evidence from Bangladesh
Sayma Rahman*
School of Economics & Finance University of Western Sydney
Locked Bag 1797 Penrith South
DC NSW 1797 Australia
Tel +61 2 9831 8020 Fax +61 2 4620 3787
Email:[email protected]
Abstract
This paper investigates the consumption behaviour of the borrowers of two
major micro-credit institutions in Bangladesh and compares this with that of the non-borrowers. Primary data has been collected from the Grameen Bank and the Bangladesh Rural Advancement Committee (BRAC) borrowers of some selected villages from three major districts in Bangladesh. Alongside the borrowers, data has also been collected from the non-borrowers of the same village to compare the differences in consumption patterns between the two groups. In investigating the impact of per capita monthly expenditure and other household characteristics on the budget share of the items consumed by the borrowers and non-borrowers, the study relies on the AIDS (An Almost Ideal Demand System) framework. The study analyses the monthly budget share of twelve household consumables (food and non-food items) and finds that the borrowers of micro-credit are better off in terms of consumption than the non-borrowers.
JEL Classification: C81, D19
Key words: Microcredit, consumption pattern, budget share, AIDS model.
* I would like to thank Professor P. N. (Raja) Junankar and Dr. Girijasankar Mallik for their comments on earlier drafts of the paper. Any errors are my responsibility.
252
REFERENCES
Abed, F. H. 2000, Microfinance NGO’s in Bangladesh Growth, Impact and Challenges, paper presented at the Asian Regional Conference, Dhaka, Bangladesh (November). Ackerly, B.A. 1995, Testing the Tools of Development: Credit Programs, Loan involvement and Women’s Empowerment, IDS Bulletin 26: (3), 56-68. Adams, D.W. and D. A. Fitchett, 1992, Informal Finance in Low-income Countries. Boulder, Colo, West View Press. Adams, D. W. and J. D. Von Pischke, 1992, Micro Enterprise Credit Programs: Déjà Vu. World Development 20, 1463-70. Alderman, H. 1993, Obtaining Useful Data on Household Incomes from Surveys, in Von Braun, J & Puetz, D. (eds.), Data Needs for Developing Countries: New Directions for Household Surveys, International Food and Policy Research Institute, Washington, DC, 193-201. Amin, S. and A. R. Pebley, 1994, Gender Inequality within Households: The Impact of a Women’s Development Program in 36 Bangladeshi Villages, The Bangladesh Development Studies XXII (2&3). Ashraf, N., D. Karlan, and W. Yin, 2006, Female Empowerment: Impact of a Commitment Savings Product in the Philippines, Yale University Economic Growth Centre Discussion Paper No. 949. Asian Development Bank, www.asiandevelopment Bailey, F. G. 1964, Capital, Saving and Credit in Highland Orissa (India), Chicago: The Aldine Press. Bajwa, R. 2001, Talk Given to the Participants of 75th National Management Course, held at Pakistan Administrative Staff College, Lahore. Bangladesh Bank, 1990: Economic Trends, XV: (5):17-18 Bangladesh Country Brief, The World Bank, September, 2004, www.worldbank.org Bangladesh Bureau of Statistics, http://www.bbs.gov.bd/ Banu, D., F. Farashuddin, A. Hossain and S. Akhter, 2001, Empowering Women in Rural Bangladesh: Impact of Bangladesh Rural Advancement Committee’s (BRAC’s) Program. Bardhan, P. K. 1976, Variatins in Extent and Forms of Agricultureal Tenancy: Analysis of Indian Data across Regions and over Time, Economic and Political Weekly 11, 1505-11 and 1541-46.
253
Bardhan, P. K. 1980, Interlocking Factor Markets and Agrarian Development: A Review of Issues, Oxford Economic Papers, 32: (1), 82-98. Bardhan, P. K. 1977, Variation in the Form of Tenancy in a Peasant Economy, Journal of Development Economies, 4, 105-18. Bardhan, P. K and T. N. Srinivasan, 1971, Crop-sharing tenancy in agriculture: A Theoretical and Empirical Analysis, American Economic Review 61, 48-64. Barten, A. P. 1964, Family Consumption, Prices and Expenditure Patterns, in P.E. Hart, G. Mills and J. Whitaker (eds.). London, Butterworth. Barten, A. P. and. L. J. Bettendorf, 1989, Price Formation of Fish: an Application of an Inverse Demand System, European Economic Review, 33: 1509-1525. Basu, K. 1997, Analytical Development Economics: The Less Developed Economy, Massachusetts; London, The MIT Press. Beneito, P. 2003, A Complete System of Engel Curves in the Spanish Economy, Applied Economics 35, 803-816. Besley, T. and. S. Coate, 1995, Group Lending Repayment Incentives and Social Collateral, Journal of Development Economics, 46, 1-18. Besley, T. 1997, Political Economy of Alleviating Poverty: Theory and Institutions, World Bank Research Observer. Washington, D.C. The World Bank: 117-134. Bhaduri, A. 1977, On the Formation of Usurious Interest Rates in Backward Agriculture, Cambridge Journal of Economics 1, 341-52. Bhaduri, A. 1973, A Study in Agricultural Backwardness under Semi-Feudalism, Economic Journal, 83, 120-37. Binswanger, H. P., and M. R. Rosenzweiz, 1984, Contractual Arrangements, Employment and Wages in Rural Labour Markets in Asia, Yale University Press. Blundell, R. and R. Ray, 1984, Testing for linear Engel curve and additively separable preferences using a new flexible demand system, The Economic Journal 94, 800-11.
Bornstein, D. 1998, The Microcredit Movement is Revolutionising International Development. www.civnet.org/journal/issue6/erpdborn.htm. Bottomley, A. 1964, Monopoly Profit as a Determinant of Interest Rates in Underdeveloped Rural Areas, Oxford Economic Papers, 16, 431-437. Bottomley, A. 1975, Interest Rate Determination in Underdeveloped Rural Areas, American Journal of Agricultural Economics, 57, 279-291. Browning, P. 2005, The Puzzle of Bangladesh, International Herald Tribune, Far
254
Eastern Economic Review, 7th May. BRAC Annual Reports (2004 and 2006), www.brac.net/ Buckland, J. 1998, Social Capital and Sustainability of NGO Intermediated Development Projects in Bangladesh, Community Development Journal, 33, 236-248. Chavas, J. P. 1984, The Theory of Mixed Demand Functions, European Economic Review 24, 321-344. Chaves, R. A. and C. Gonzalez-Vega, 1996, The Design of Successful Rural Finance Intermediaries: Evidence from Indonesia, World Development, 24, 65-78. Christensen, L. R., D. W. Jorgenson and L. J. Lau, 1975, Transcendental Logarithmic Utility Functions, American Economic Review 65, 367-383. Coleman, B.E. 1999, The Impact of Group Lending in Northeast Thailand, Journal of Development Economics 60, 105-141. Dax, P. 1987, Estimation of Income Elasticities from Cross-Section Data, Applied Economics 19, 1471-1482. Deaton, A. and. J. Muellbauer, 1978, An Almost Ideal Demand System, University of Bristol. Deaton, A., and. J. Muellbauer, 1980, An Almost Ideal Demand System. American Economic Review 12, 152-163. Deaton, A. 1981, Theoretical and Empirical Approaches to Consumer Demand under Rationing, in A Deaton (ed.) Essays in the Theory and Measurement of Consumer Behaviour in Honour of Sir Richard Stone, Cambridge, Cambridge University Press,. Devereux, S. 1993, Collection of Production and Income Data: A Commentary Analysis, International Food and Policy Research Institute, Washington, DC. Dey, M. M. 2000, Analysis of Demand for Fish in Bangladesh, Aquaculture Economic and Management 4: (12), 65-83. Feder, G., T. Onchan, Y. Chalamwong and C. Hongladarom, 1988, Land Policies and Farm Productivity in Thailand, Baltimore, Johns Hopkins University Press. Ferdous, R. 1997, Consumer Demand Behaviour in Rural Bangladesh, Dhaka University Journal of Science, 45(1), 109-119. Ferdous, R. 1999, Investigating Gender inequality in Our Family Budgets, Dhaka University Journal of Science, 47(1), 109-117.
255
Foster, A. D. 1995, Prices, Credit Markets and Child Growth in Low Income Rural Areas, The Journal of the Royal Economic Society, 105, 551-70. Getubig, I. P. Jr., 1992, The Role of Credit in Poverty Alleviation: The Asian Experience, Economic Development Institution, The World Bank, Washington, DC. Ghate, P. 1992, Informal Finance: Some Findings from Asia, London: Oxford University Press. Ghatak, M. 1999, Group lending, local information and peer selection, Journal of Development Economics, 60, 27-50.
Giles D. E. A. and P. Hampton, 1985, An Engel Curve Analysis of Household Expenditure in New Zealand, The Economic Record, 61, 450-62. Goetz, A. and R. Sengupta, 1996, Who Takes the Credit? Gender, Power, and Control over Loan Use in Rural Credit Programs in Bangladesh, World Development 24(1), 45-63. Gonzalez-Vega and R.A. Chaves, 1993, Indonesian Rural Financial Markets (Mimeo). Gorman, W. M. 1976, Tricks with Utility Functions, Cambridge: Cambridge University Press. Gujarati, D. N. 1995, Basic Econometrics, McGraw- Hill, New York. Gujarati, D. N. 1992, Essentials of Econometrics, McGraw Hall, New York. Hashemi, S. M., S. R. Schuler and A.P. Riley, 1996, Rural Credit Programs and Women's Empowerment in Bangladesh, World Development 24(4), 635-653. Hazell, P. and. A. Röell, 1983, Rural Growth Linkages: Household Expenditure Patterns in Malaysia and Nigeria, International Food and Policy Research Institute. Hendriks, S. L., and. M. C. Lyne, 2003, Expenditure Patterns and Elasticities of Rural Households Sampled in two Communal Areas of Kwa Zulu-Natal, Development Southern Africa 20(1), 105-127. Hofstede, G. H. 2001, Culture’s consequences: Comparing values, behaviours, institutions and Organisations across nations, Thousand Oaks, Calif, Sage. http://www.un.org/popin/ Holcombe, S. 1995, Managing to empower: the Grameen Bank's experience of poverty alleviation NJ Zed Books, 36-37. Hossain, M. 1984, Credit for the Rural Poor: The Grameen Bank in Bangladesh, Bangladesh Institute of Development Studies, Dhaka.
256
Hossain, M. 1988, Credit for Alleviation of Rural Poverty: The Grameen Bank in Bangladesh, IFPRI Research Report 65, International Food Policy Research Institute, Washington, D.C. Houthakker, S. H. 1957, An International Comparison of Household Expenditure Patterns Commemorating the Centenary of Engel's law, Econometrica, 25, 532-551. Hsioa, C. 2003, Analysis of Panel Data, Cambridge University Press, U.K, 09-10. Hulme, D. and P. Mosley, 1996, Finance against poverty, London, Routledge Int. Publishing Company. Hussain, A. M. 1998, Poverty Alleviation and Empowerment: The Second Impact Assessment Study of BRAC’s, Rural Development Program, BRAC, Dhaka. Johnson, S., and B. Rogaly 1997, Microfinance and Poverty Reduction, London, Oxford; OXFAM. Kabeer, N. 1998, Money Can’t Buy Me Love, Re-Evaluating Gender, Credit and Empowerment in Rural Bangladesh, Discussion Paper 363 IDS, Sussex. Kabeer, N. 2001, Conflicts over Credit: Re-Evaluating the Empowerment Potential of Loans to Women in Rural Bangladesh, World Development, 29, 63-64. Khalily, B.A., M. O. Imam, and S. A. Khan, 2000, Efficiency and Sustainability of Formal and Quasi-formal Microfinance Programs – An Analysis of Grameen Bank and ASA, The Bangladesh Development Studies, XXVI (2&3). Khan, M. M. 1985, Labour Absorption and Unemployment in Rural Bangladesh, Bangladesh Development Studies, XIII, 3 & 4, 67-88. Khandker, S. R. 1998a, Fighting Poverty with Microcredit Oxford University Press. Washington, D.C. Khandker, S. R., B. Khalily and Z. Khan, 1995, Grameen Bank, Performance and Sustainability, World Bank Discussion Papers 306. Khandker, S. R., 1998b, Micro-credit Program evaluation: A critical review, IDS Bulletin 29(4), 11-20. Khandker, S. R. 1996, Grameen Bank: Impact, Costs and Program Sustainability, Asian Development Review, 14, 97-130. Khandker, S. R. 2000, Savings, Informal Borrowing, and Microfinance, Bangladesh Development Studies XXVI: (2 &3), 49-78. Khandker, S. R. 2003, Micro-finance and Poverty: Evidence Using Panel Data from Bangladesh, World Bank Policy Research Working Paper 2945.
257
Khandker, S. R., H. A. Samad, et al., 1998, Income and Employment Effects of Micro-credit Programs; Village-level Evidence from Bangladesh, Journal of Development Studies 35(2), 96-124.
Lee,J. Y. and M. Brown, 1986, Food Expenditures at Home and Away Home in the United States: A Switching Regression Analysis, The Review of Economics and Statistics, 68, 142-47. Leser, C. E. V.1976, Income, Household Size and Price Changes, Oxford Bulletin of Economics and Statistics 38, 1-10. Lipton, M. 1976, Agricultural Finance and Rural Credit in Poor Countries, World Development, 4. Liviatan, N. 1961, Errors in Variables and Engel Curve Analysis, Econometrica, 29(3). Massell, B. F. 1969, Consistent Estimation of Expenditure Elasticities from Cross-Section data on Households Producing Partly for Subsistence, Review of Economics and Statistics 5: (12), 136-142. Matin, I. 1999, Rapid Credit Deepening and A Few Concerns: A Study of Grameen Bank in Madhupur, Unpublished Manuscript. McNamara, N. and S. Morse, 1998, Donors and Sustainability in Provision of Financial Services in Nigeria, IDS Bulletin, 29, 91-101. Meenakshi, J. V., and. R. Ray, 2002, Impact of Household Size and Family composition on Poverty in Rural India, Journal of Policy Modelling 24, 539-559. Mizan A. N. 1993, Women’s Decision Making Power in Rural Bangladesh: A Case study of Grameen, in Abu Wahid (ed) The Grameen Bank: The Poverty Relief in Bangladesh, West View press, Dhaka, Bangladesh 97-126. Montgomery, R., D. Bhattacharya and D. Hulme, 1996, Credit for the Poor in Bangladesh: The BRAC Rural Development Program and the Government Thana Resource Development and Employment Program' in Hulme, D. and Mosley, P. Finance against Poverty, 1 and 2, Routledge, London. Morduch, J. 2000, The Microfinance Schism, World Development, 28: (4), 617-29. Morduch, J. 1998, Does Microfinance Really Help the Poor? New Evidence from Flagship Programs in Bangladesh, Harvard-Institute for Development, Department of Economics and HIID, Harvard University. Morduch, J. 1999, The Microfinance Promise, Journal of Economic Literature, 37: (4), 1569-1614. Moschini, G. and. A. Vissa, 1993, Flexible Specification of Mixed Demand System, American Journal of Agricultural Economics 75, 1-9.
258
Muellbauer, J. 1975, Aggregate, Income, Distribution and Consumer Demand, Review of Economic Studies 62, 525-543. Muellbauer, J. 1976, Community Preferences and the Representative Consumer, Econometrica 44, 979-999. Mustafa, K., Z. Ahmed and T. Azid, 2000, NGOs, Micro-finance and poverty alleviation: Experience of the rural poor in Pakistan, The Pakistan Development Review. 39: (4), 771-792. Mustafa, S., I. Ara, D. Banu, A. Hossain, A. Kabir, M. Mohsin, A. Yusuf and S. Jahan, 1996, Beacon of Hope: An Impact Assessment Study of BRAC’s Rural Development Program, Dhaka, BRAC. Muqtada, M. 1981, Poverty and Famines in Bangladesh, Bangladesh Development Studies, IX (1), 1-34. Nahar, N. 2004, Concepts and Different Poverty Measures in Rural Bangladesh, AIUB Journal of Business and Economics (AJBE) 1: (5), 57-72. Naved, R. 1994, Empowerment of Women: Listening to the Voices of Women, in the Bangladesh Development Studies, ‘Special Issue on Women, Development and Change, XXII, (2 &3), edited by S. Amin, BIDS, Dhaka. Osmani, S. R. 1989, Limits to the Alleviation of Poverty through Non-Farm Credit, Bangladesh Development Studies XVII: (4) 1-20. Osmani, S. R. 1978, On the Normative Measurement of inequality, The Bangladesh Development Studies 6: (4), 417-442. Osmani, S.R., W. Mahmud, B. Sen, H. Dagdeviren, and A. Seth, 2003, Pro-Poor Policies in Bangladesh, Asia and the Pacific Regional Workshop on Macroeconomics of Poverty Reduction, UNDP. Pitt, M. and S.R. Khandker, 1996, Household and Intra-household Impacts of the Grameen Bank and Similar Targeted Credit Programs in Bangladesh, World Bank Discussion Papers 320, Washington, DC. Pitt, M. and S. Khandker, 1998, The Impact of Group-Based Credit Programs on Poor Households in Bangladesh: Does Gender of the participant Matter? Journal of Political Economy 106, 958-996. Prais, J. S. and. H. S. Houthakker, 1955, The Analysis of Family Budget, Cambridge University Press. Puetz, D. 1993, Improving Data Quality in Household Surveys. Washington, DC: International Food and Policy Research Institute. Quasem, M. A. 1991, Limits to the Alleviation of Poverty through Non-Farm Credit: A Comment, Bangladesh Development Studies XIX: (3) 129- 134.
259
Raj, K. N. 1979, Keynesian Economics and Agrarian Economies, Delhi, Allied Press. Rahman, A. 1999, Microcredit Initiatives for Equitable and Sustainable Development: Who Pays? World Development 27, 67-82. Rahman, A. and R. Islam, 1987, An Empirical Account of Hired Labour Market in Rural Bangladesh, The Bangladesh Development Studies XV: (1) 129-142. Rahman, A. and R. Islam, 1988, Labour use in Bangladesh – An Empirical Analysis, The Bangladesh Development Studies XVI: (4) 1-40. Rahman, H. 1995, Mora Kartik: Seasonal Deficits and the Vulnerability of the Rural Poor, in H. Rahman, and M. Hossain (eds.), Rethinking rural poverty Bangladesh, Dhaka, University Press Limited (UPL). Rahman, H. Z. and M. Hossain, 1996, Rethinking Rural Poverty: Bangladesh as a case study, University Press Limited, Dhaka, Bangladesh. Rahman, R. I. 1978, Measurement of Rural Unemployment: A Dis-aggregative Approach, Bangladesh Development Studies 6: (1) 101-110. Rahman, R. I. and S. R. Khandker, 1994, Role of Targeted Credit Programs in Promoting Employment and Productivity of the Poor in Bangladesh, The Bangladesh Development Studies, XXII (2&3), 50-92. Rahman, A. 1979, Usury Capital and Credit Relations in Bangladesh Agriculture: Some Implications for Capital Formation and Capitalist Growth, Bangladesh Development Studies, 7, 1-46. Rahman, A. and A. Razzaque, 2000, On Reaching the Hardcore Poor: Some Evidence on Social Exclusion in NGO Programs, The Bangladesh Development Studies, XXVI, (1), 1-36. Rao, A., and D. Kelleher, 1995, Engendering Organisation Change: The BRAC Case in Getting Institution Right for Women Development, IDS Bulletin 29 (3). Ravallion, M. and Q. Wodon, 2000, Banking on the Poor? Branch Location and Non-farm Rural Development in Bangladesh, Review of Development Economics 4(2), 121-139. Ray, R. 1980, Analysis of a Time Series of Household Expenditure Surveys for India, Review of Economics and Statistics, 595-602. Reilly, B. 1990, Occupational Endogeneity and Gender Wage Differentials for Young Workers: An Empirical Analysis Using Irish Data, The Economic and Social Review, 21 (3), 311-328. Reserve Bank of India, 1977, Indebtedness of Rural Households and the Availability of International Finance, All-Indian Debt and Investment survey 1971-2, Bombay.
260
Saleem, S. T. 1987, On the Determination of Interest Rates in Rural Credit Markets: A Case Study from Sudan, Cambridge Journal of Economics 11, 65-72. Samuelson, P. A. 1965, Using Full Duality to Show that Simultaneously Additive Direct and Indirect Utilities Implies Unitary Price Elasticity of Demand. Econometrica 33, 781-796. Sawtelle, B. A. 1993, Income Elasticities of Household Expenditure: A US Cross-Section Perspective, Applied Economics 25, 635-644. Sen, G. 1997, Empowerment as an Approach to Poverty, Human Development Report, Background Paper (New York: The UNDP) 96-97. Sen, A. K. 1976, Poverty: an Ordinal Approach to Measure, Econometrica, 44 (2), 219-231 Sener, T.1977, An International Comparison of Demand Elasticities: Empirical Analysis of Consumption Patterns, Studies in Development 15, 124-159. Shephard, R. W. 1970, Theory of Cost and Production Functions, Princeton, N.J. Shephard, R. W. 1953, Cost and Production Functions, Princeton, N. J. Squillace, S. 2004, The Palli Karma-Sahayak Foundation as a Regulator of Microfinance Institutions in Bangladesh, Master of Arts in Law and Diplomacy Thesis, The Fletcher School, Tufts University. Stiglitz, K. 1993, Peer Monitoring and Credit Markets: Oxford University Press. Stone, J. R. N. 1953, The Measurement of Consumers ' Expenditure and Behaviour in the United Kingdom, 1920-1938, (Vol. 1): Cambridge University Press. Stone, J. R. N. 1954, Linear Expenditure Systems and Demand Analysis: An Application to the Pattern of British Demand, Economic Journal 64, 511-527. Summers, R. 1959, A Note on Least Squares Bias in Household Expenditure Analysis, Econometrica 27(1), 121-126. The Grameen Bank Annual Reports (2003 and 2006), http://www.grameen-info.org/ The Microcredit Summit, 1997, The Microcredit Summit: Declaration and plan of action. Washington, D.C. The World Fact Book, http://www.bartleby.com/151. Theil, H. 1965, The Information Approach to Demand Analysis, Econometrica 33, 67-87. Theil, H. 1975-1976, The Theory and Measurement of Consumer Demand (Vol. I and II,), Amsterdam, North-Holland.
261
Tobin, J. 1958, Estimation of Relationship for Limited Dependent Variables, Econometrica, 26(1). Tun Wai, U. 1958, Interest Rates outside the Organised Money Markets of Underdeveloped Countries. IM F. Staff Paper (6). United Nations, 2004, Women and Poverty http://topics.developmentgateway.org/gender. United Nations, Development Program, www.undp.org/ UNDP, Human Development Report, 2000, http://hdr.undp.org/report Varian, H. R. 1990, Monitoring Agents with Other Agents, Journal of Institutional and Theoretical Economics 146, 153-174. Wahid, A. N.W. 1993, The Grameen Bank Poverty Relief in Bangladesh, West View Press. Wahid, A. N. M. 1994, The Grameen Bank and Poverty Alleviation in Bangladesh: Theory, Evidence and Limitations, The American Journal of Economics and Sociology, 53(1). Weiskoff, R. 1971, Demand Elasticities for a Developing Economy, Cambridge: Harvard University Press. Weiss, J. and H. Montgomery, 2005, Great Expectations: Microfinance and Poverty Reduction in Asia and Latin America, Oxford Development Studies, 33(3/4):391-416. Wharton, C. 1962, Marketing, Merchandising and Money Lending: A note on Middlemen Monopsony in Malaya, Malaysian Economic Review, 7, 24-44. White S.C. 1992, Arguing with the Crocodile: Gender and Class in Bangladesh, Zed Books, London. Write, A. N. Graham, 2000, Micro-Finance Systems- Designing Quality Financial Services for the Poor, The University Press Limited, Dhaka, Bangladesh. World Bank, 1994, The World Bank's Strategy for Reducing Poverty and Hunger, Environmentally Sustainable Development Studies, Monograph series 4. Yaron, J. 1994, What Makes Rural Finance Institute Successful? The World Bank Research Observer. 9: (1), 49-70 Yaron, J. 1992a, Assessing Development Finance Institutions: A Public Interest Analysis, The World Bank Discussion Paper, Washington, D. C. Yaron, J. 1992b, Successful Rural Finance Institutions. Washington, D.C., World Bank Discussion Papers 150. Monograph series 4.
262
Yaron, J., B. McDonald and G. Piprek 1997, Rural Finance: Issues, Design and Best Practices, Environmentally Sustainable Development Studies and Monographs Series, No.14, The World Bank, Washington D. C. Yaron, J. et al. 1998, Promoting Efficient Rural Financial Intermediation, The World Bank Research Observer 13(2), 147-70. Yunus, M. 1983, Group -based Savings and Credit for the Rural Poor: Grameen Bank in Bangladesh. Group-based Savings and Credit for the Rural Poor, Papers and Proceedings of a workshop, Bogra (Bangladesh), Geneva: ILO. Yunus, M. 1984, Group-Based Savings and Credit for the Rural Poor: The Grameen Bank in Bangladesh, in Group-Based Savings and Credit for the Rural Poor, International Labour Office, Geneva. Yunus, M. 1999, Banker to the Poor, The Autobiography of Muhammad Yunus, Founder of the Grameen Bank. Aurum press limited. London. Zaman, H. 1998, Who Benefits and to What Extent? An evaluation of BRAC's Microcredit Program University of Sussex, D.Phil Zaman, H. 2001, Assessing the Poverty and Vulnerability Impact of Microcredit in Bangladesh: A case study of BRAC, World Bank.
Zeller, M. and M. Sharma, 2000, Many Borrow, More Save, and All Insure: Implications for Food and Microfinance Policy, Food Policy 25, 143-167. Zeller, M. and M. Sharma, 1999, Placement and Outreach of Group-Based Credit Organisation: The Cases of ASA, BRAC, and PROSHIKA in Bangladesh, World Development, 12: 2123-36. Zellner, A. 1962, An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests of Aggregation Bias, Journal of the American Statistical Association, 57, 500-509.
.