Targeting Individuals for Microfinance Programs:
The Impact of Educational Attainment
SIS, American University
4400 Massachusetts Avenue NW, Washington DC 20016
Microfinance loan repayment levels are quite low; discouraging banks and other microcredit
institutions from expanding their lending scope to reach the majority of the rural poor. In this
study, I examine candidate client characteristics that affect loan repayment. Mainly, I study the
relationship between loan repayment and educational attainment, but also other borrowing
history variables affecting repayment as well. In particular, data is used from the Matlab
[Bangladesh] Health and Socioeconomic Survey (MHSS) of 1996, to examine the microfinance
borrowing history of 773 individuals from across the country. My findings suggest that greater
levels of educational attainment improve loan repayment rates. I also find that first time
borrowers are less likely to repay their loans. Microenterprises are a vital source of income for
those of the rural poor in emerging economies. It is imperative that microcredit institutions
uphold their mission of issuing microcredit to these clients.
I am grateful to Andy Schaalman. Andy provided helpful feedback for this empirical analysis.
Historically commercial efforts to provide credit and financial services to poor, rural
communities of developing countries have generally failed. To compensate, many governments
developed credit programs to fill this void. These programs yielded mixed results, incentivizing
international organizations, such as USAID and the United Nations, to design and deploy their
own microfinancing programs. These programs too have produced varied results in terms of
program reach and credit access. Of those individuals who did gain access to credit through
these programs, results have indicated there is room for improvement in terms of loan
repayment. Are there measurable characteristics that should be considered for targeting
microfinancing candidates that would ensure greater loan repayment rates, thus further
incentivizing commercial entities and international organizations to extend the reach of their
In developing countries, does a microfinance candidate’s education level impact loan repayment?
Ha: Microfinance client candidates with higher educational attainment are more likely to
repay their loans.
For this statistical analysis the independent variable is educational attainment. The dependent
variable is household borrowing history of interest-free loans, loan repayment. Other
independent variables that will be measured are: loan amount, months allowed for payback,
number of borrowing transactions, and purpose of loan. These independent variables are
indicators of the interest-free loans granted and may be associated with loan repayment. The
following variables are nominal: loan amount, months allowed for payback, and number of
borrowing transactions. The following variables are ordinal: educational attainment, purpose of
loan, loan repayment. All data is gathered from the Matlab [Bangladesh] Health and
Socioeconomic Survey (MHSS) of 1996.
Microfinance is thought of to be as the most important recent tool for reducing poverty in
developing economies, as credit is important to the incomes of people in these countries.
Microenterprises created from this credit is a critical source of income for the poor.
2. Literature Review
Over the last couple of decades alleviating poverty in developing economies has become a focus
of many international organizations and a focus for governments, domestically and
internationally. For example, the United Nations has implemented several microfinancing
programs, previously declaring 2005 as the “Year of Microcredit”. Many individuals of these
countries have witnessed a surge in available credit enabling them to support themselves and
their families through creating their own microenterprises. However, the faith in microfinance
does not appear to be based on solid evidence, as interviews with the poor of select countries and
non comprehensive program evaluations have indicated. There is much evidence to suggest that
while microfinance programs may help some, they are not reaching and providing credit to the
vast majority of the rural poor.
Overall, little unbiased research exists to evaluate the impact of these programs, and even further
fewer attempts to examine the relationship between client candidate characteristics and loan
repayment. The literature review for this research includes an overall analysis of empirical
evidence on microcredit, two country (India and Thailand) empirical studies on credit access and
benefits of microcredit, and an empirical assessment of clients of microfinance programs in
Micro-credit is claimed to be the best solution to most of the problems facing the rural poor in
developing countries. However, the implementation of any formal program is plagued by three
challenges: exact targeting of clients of those in need of the program; distinguishing creditworthy
borrowers; and enforcing loan repayment (Chavan and Ramakumar, 2002). Understanding and
working to improve these problems is vital for the survival of microfinance programs. Chavan
and Ramakumar found that while the implemented programs have led to positive income
increases, they have only been marginal adjustments. Perhaps if these errors are address, these
programs could be more successful.
Further, minimizing the costs of default is “crucial for the sustainability of any credit provision
programme or institution” (Chavan and Ramakumar, 2002 p. 959). Some studies have
concluded default rates from these programs are between 50 and 75 percent. Evidence has also
suggested default rates increase with the size of the loan and with return borrowers (Chavan and
Ramakumar, 2002). In essence, curbing the default rate is important for incentivizing
microcredit institutions and borrowers to develop better reaching programs and to participate in
these programs. If high default rates persist, the additional costs will be shifted onto borrowers,
increasing their debt and thus defeated the objective of the program (Chavan and Ramakumar,
Basu and Srivastava similarly find that despite its perceived growing global importance, the
practice of microfinance has been relatively poor at reaching the majority of rural poor and
producing a positive impact over the past decade (2005). These discouraging results stem from
banks’ concerns regarding the safety of their loans. Banks, financial services entities, want to
ensure their loans are safe and “will not take on the time-consuming task of helping groups
manage the bookkeeping of their internal savings and loan accounts” (Basu and Srivastava, 2005
p. 1754). Microfinance programs in emerging economies will not be successful if the right
clients are not targeted, nor the clients are not educated to understand how these programs work.
Basu and Srivastava’s analysis indicates linkage between banks and targeting the poor has
resulted in “mission drift” as pressures to produce good financial results have led to the serving
of richer clients with larger loans (2005).
This drift is a possible explanation for the shallow reach and small impact microfinance
programs have made thus far. Further, Basu and Srivatava analyze what types of communities
are associated with poverty and what candidate client characteristics may indicate why these
programs are not reaching those in need (2005). Their results suggest the percentage of illiterate
households is likely to be correlated with poverty; additionally, higher rates of illiteracy are
associated with fewer “self help groups”/ microcredit institutions in that community (Basu and
To reinforce the argument that these programs are not reaching the poor as much as they are the
relatively wealthy, Coleman provides empirical evidence (2006). In his study, weighted t-tests
indicate that program participant households are significantly wealthier than those of
nonparticipants (Coleman, 2006). Further, Coleman interviewed poor rural villagers who stated
they would be interested in joining a microfinance program, but they do not believe they are
qualified to join as these programs are perceived to be for the rich. Coleman’s regressions yield
these villagers understand the concept of loan repayment and show understanding that
creditworthiness is a significant determinant. In interviews with five other villages, Coleman
found that none of the villages identified the rural bank program as one that targeted the poor
(2006). Poor targeting is a significant problem for these programs.
When examining an earlier assessment of microfinance clients in Uganda, the empirical evidence
is somewhat different. The assessment concludes that these programs positively and
significantly had a positive impact on clients (Barnes, Morris, & Gaile, 1999). Further, this
research assessing client characteristics and determines clients tended to have higher educational
attainment than non-clients; this characteristic was found statistically significant (Barnes, et. al,
1999). In fact Barnes, et. al. find that these clients actually invested in further education after
theses programs. Fifty-nine percent of clients cited their microenterprises as the main source of
funds for expenditures on education, averaging $137 a month (Barnes, et.al, 1999).
This literature supports to objective of this research; examining candidate client characteristics is
important for predicting success of microfinance programs. This literature also indicates
educational attainment may be a significant predictor for determining loan repayment.
3. The Model
Given empirical findings, greater levels of education attainment should be associated with
interest-free loan repayment. From the survey, the significant indicators with the largest amount
of data available are: education attainment, loan amount, months allowed for payback, number of
borrowing transactions, and purpose of loan.
Loan Repayment = 0 + 1Edu +2 Amount +3Months + 4NBorrow + 5Purpose
where, Edu = education indicator (primary, secondary, or unattached secondary)
Amount = loan amount
Months = months allowed for payback
NBorrow = number of borrowing transactions
Purpose = purpose of loan
It is expected that higher levels of educational attainment will result in a negative correlation
with loan repayment; therefore, the more educated an individual is the less likely the loan will
not be repaid. Microcredit institutions tend to have a greater reach into more educated
communities because these individuals are able to understand the purpose of the loan and the
advantages to repayment.
Data collected for this study is at the nominal and ordinal levels of measurement. The data is
analyzed from the Matlab [Bangladesh] Health and Socioeconomic Survey (MHSS) 1996
survey. Only data from Bangladesh is represented in this study.
The dependent variable, loan repayment, was calculated from the survey’s Household Economy,
Borrowing History of Interest-Free Loans; respondents answered “yes” or “no”. The main
independent variable, educational attainment, was calculated from the survey’s Household roster,
Household Members. Education levels were measured in primary, secondary, or unattached
secondary. Loan amount, months allowed for payback, number of borrowing transactions, and
purpose of loan were all measured within the Borrowing History data. Purpose of loan was by
23 different responses. According to the data, the majority of loans were used for business
supplies; the average responses were fish cultivation, rickshaw, and boat.
In total, the Household survey collected data from 4,364 households, capturing 2,687 residential
compounds. The Borrowing Data represented 773 observations. The data is large enough to
make sufficient references to examine a case of microfinance programs and candidate clients.
5. Estimates and Empirical Findings
The purpose of this study is to determine if higher educational attainment levels increased loan
repayment of microfinance programs emerging economies. This study uses the MHSS 1996
survey to examine microfinance loan repayment in Bangladesh. The MHSS 1996 provides both
the dependent variable and the independent variables (educational attainment, loan amount,
months allowed for payback, number of borrowing transactions, and purpose of loan) from the
Borrowing History data. The null hypothesis states that microfinance clients will all educational
attainment levels will repay their loans. The research hypothesis states that microfinance clients
with higher levels of educational attainment levels will be more likely to repay their loans.
First descriptive statistics were run to determine central tendencies, range, and variability of the
variables. The descriptive statistics give greater detail into the types of microfinance loans that
were issued, as well as to the clients. The majority of those individuals had not repaid their loan
(Figure 1) and attained a primary education level (Figure 2). Loan amounts for this survey
ranged from 50-200,000 takas and the average amount was 5,180.93 takas (Table 1). The
average months allowed for payback was 50 month and the average number of borrowing
transactions was close to 1 (Table 1). Individuals cited the purpose of the loan, which showed
the majority of the loans were used for microenterprise investments. The average responses
were: to purchase a boat, to purchase a rickshaw, or for fish cultivation.
Bivariate correlations were run to determine if there were relationships between any of the
variables (Table 2). Only one significant relationship was found and that was between the two
main variables: loan repayment and educational attainment. This further supports the research
hypothesis, indicating that there is an association between these two variables, which may cause a
change in the other.
Lastly, a probit regression analysis of the dependent variable (loan repayment) was performed to
determine the Best Linear Unbiased Element Model (BLUE) (Table 3). This Model will be used
to determine unbiased estimations of the relationships between the variables; thus, further
determining a decision to accept or reject the null hypothesis. As seen in Table 3, all Models have
a sufficient sample size (773) to rely up on the regression. In all six models, loan repayment and
educational attainment are both significant at the 5% level. The intercept coefficients remain
somewhat stable around the same value throughout all models.
While Model 1 has both significant variables, adjusted R is very low; therefore, this Model is not
the best estimator of the data. Models 2 and 3, both incorporate the significant variables, but also
incorporate highly non-significant models, also making them not the best predictor.
Models 4 and Models 5 both show number of borrowing transactions is significant at the 10%
level, in addition to the intercept and educational attainment. However, Model 4’s adjusted R is
relatively low. This indicates Model 5 is the best Model to estimate the relationships between
these variables; however, because previous empirical findings suggest the number of borrowing
transactions is significant, another model should be created. Model 6 was created to determine the
significant between the three significant variables: loan repayment, educational attainment, and
number of borrowing transactions.
While Model 6 indicates the intercept is still significant, educational attainment becomes less
significant (now at the 10% level) and number of borrowing transactions becomes non-significant.
The intercept’s coefficient also greatly increased. Also, adjusted R is very low. These changes
may indicate some incomplete data; thus, altering the true data. Therefore, Model 5 is the BLUE.
6. Conclusions and Policy Implications
The results from Model 5 explain there is a negative relationship between loan repayment and
educational attainment; greater levels of educational attainment indicate lower levels of unpaid
loans. This Model also shows a positive relationship between loan repayment and number of
borrowing transactions; the less borrowing transactions indicate lower loan repayment rates.
These results allow us the reject the null hypothesis and accept the research hypothesis.
Further, the results indicate that these relationships can be relied upon 6.6% of the time; this is
relatively low, but it is the second highest adjusted R values of the probit regression. This low
reliability reduces the profoundness of the research hypothesis. While the hypothesis is
significant, the reliability of the data is low.
However, this data presents the opportunity for future research. A more recent, but similar survey
and statistical analysis should be conducted in other emerging economies. It would be interesting
to explore the relationship between loan repayment and number of borrowing transactions, as other
data has indicated the more a client borrows, the less likely they are to repay; while this data, only
seems to have focused on first time borrowers. It would also be interesting to investigate whether
educational attainment is correlated with number of borrowing transactions.
There a two main policy implications from this research. First, microfinance programs
implemented in emerging economies should first target candidate clients with higher levels of
educational attainment in the rural poor areas. Targeting these clients, will ensure greater loan
safety for banks/microcredit institutions; thus, greater levels of loan repayment. These higher
levels of repaid loans will encourage these institutions to expand the scope of their lending
practices, expanding beyond the wealthier individuals and reaching the true potential beneficiaries
of these programs: the rural poor.
Second, the aid in the process of implementing successful microfinance programs, microcredit
institutions should develop educational programs for their clients and candidate clients. An
educational program that provides greater understanding of the borrowing process and reinforces
the importance of loan repayment, will allow for a more educated clientele. A more educated
clientele will yield higher loan repayment rates, incentivizing these institutions to expand their
lending scope; thus, further reaching a larger majority of the rural poor.
Barnes, C., Morris, G., & Gaile, G. (1999). As Assessment of Clients of Microfinance Programs
in Uganda. International Journal of Economic Development. 1; 1; pp. 80-121.
Basu, P. & Srivastava, P. (2005). Microfinance and Rural Credit Access for the Poor in India.
Economic and Political Weekly. 40; 17; pp. 1747+1749-1756
Chavan, P. & Ramakumar, R. (2002). Micro-Credit and Rural Poverty: An Analysis of
Empirical Evidence. Economic and Political Weekly. 37; 10; pp.955-965.
Coleman, B.E. (2006). Microfinance in Northeast Thailand: Who Benefits and How Much?
World Development. 34;9; pp. 1612-1638.
Rahman, O., Menken, J., Foster, A., & Gertler, P. (1996). Matlab [Bangladesh] Health and
Socioeconomic Survey (MHSS). ICPSR02705-v5. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2001. doi:10.3886.
Appendix 1. Tables
Table 1. Descriptive Statistics
Variable Mean Standard Deviation Coefficient of Observations
Loan Repayment 0.150 0.357 2.856 782
Educational 1.430 0.559 0.391 774
Loan Amount 5,180.926 12,502.56 2.413 783
Months Allowed for 50.024 47.919 0.958 783
Number of 1.309 0.743 0.567 783
Purpose of Loan 17.156 12.280 0.716 783
Data source: Data are taken from the MHSS, 1996.
Table 2. Correlation Table
Variable Loan Educational Loan Amount Months Purpose of Number of
Repayment Attainment Allowed for Loan Borrowing
Educational 0.091 1.0000
Loan Amount -0.007 -0.004 1.000
782 774 783
Months -0.030 -0.025 0.029 1.000
Allowed for (0.410) (0.482) (0.416)
Payback 782 774 783 783
Purpose of 0.008 -0.024 0.015 -0.018 1.000
Loan (0.816) (0.503) (0.684) (0.609)
782 774 783 783 783
Number of -0.058 0.012 -0.020 -0.032 0.021 1.00
Borrowing (0.104) (0.738) (0.572) (0.376)\ (0.559)
Transactions 782 774 783 783 783 783
Data source: Data are taken from from the MHSS, 1996.
Estimates significant at the 5% level are in boldface, p-values in parentheses.
Table 3 . Probit Regression Analysis
Dependent Variable: Loan Repayment
Model 1 Model 2 Model 3 Model 4 *Model 5 Model 6
Intercept -1.389 -1.385 -1.336 -1.137 -1.151 -1.194
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Educational 0.240 0.240 0.239 0.239 0.240 0.240
Attainment (0.012) (0.012) (0.013) (0.013) (0.013) (0.012)
Loan Amount -0.786 -0.626 -0.835 -0.824
(1,000,000) (0.844) (0.876) (0.835) (0.837)
Months Allowed -0.000 -0.001 -0.001
For Payback (0.392) (0.372) (0.375)
Number of -0.155 -0.155 -0.154
Borrowing (0.082) (0.081) (0.086)
Purpose of Loan (0.853)
N 773 773 773 773 773 773
R2 0.013 0.044 0.073 0.035 0.066 0.015
Estimates significant at the 5% level are in boldface, p-values in parentheses. Estimates significant at the 10%
level are in boldface and italics. *Denotes best Model.
Appendix 2. Figures
Figure 1. Descriptive Statistics
Dependent Variable: Loan Repayment
Source: MHSS 1996
*0 = Not Repaid; 1 = Repaid
Figure 2. Descriptive Statistics
Independent Variable: Educational Attainment
Source: MHSS 1996
*Primary = Grades 1-5; Secondar = Grades 6-10; Unattached Secondary = Grades 11 & 12