Content uploaded by Roy Mersland
Author content
All content in this area was uploaded by Roy Mersland on Nov 04, 2017
Content may be subject to copyright.
Working Paper
2009-2
Women and Repayment in
Microfinance
Bert D'Espallier
Isabelle Guérin
Roy Mersland
1
RUME Working Papers Series
This series is published by the Rural Microfinance and Employment project (RUME). The project,
RUME, has been selected in December 2007 by the French National Agency for Research
(programme: Les Suds, Aujourd’hui). The main objective of this research is to explore the linkages
between rural finance and rural employment - including diversification and migration - with a view to
contributing to the ongoing discussions and interventions in the fields of rural development and
poverty and vulnerability reduction. The project methodology relies on the following features: a
pluridisciplinary approach, a combination of tools for data collection and analysis, a comparative
perspective across three countries (Madagascar Mexico, South-India), a strategic collaboration with
microfinance institutions. From an outcome perspective, the proposal will seek to achieve a balance
between academic and applied results. Further details about the project and its work can be viewed on
our web site at www.rume-microfinance.org
2
Rural Microfinance and Employment Project
LPED – IRD – Université de Provence
Case 10 – 3, Place Victor Hugo
13331 Marseille cedex 3
Tel: 00 33 (0) 6 72 06 52 66
www.rume-microfinance.org
3
Women and Repayment in Microfinance∗
Bert D'Espallier, Lessius Hogeschool, Belgium
Isabelle Guérin, Institute of Research for Development, France
Roy Mersland1, University of Agder, Norway
Working paper
This version, March 2009
Abstract
This paper analyzes gender-differences with respect to microfinance
repayment-rates using a large global dataset covering 350 Microfinance
Institutions (MFIs) in 70 countries. The results indicate that more women
clients is associated with lower portfolio-at-risk, lower write-offs, and
lower credit-loss provisions, ceteris paribus. These findings confirm
common believes that women in general are a better credit-risk for MFIs.
Interaction effects reveal that the effect is stronger for NGOs, individual-
based lenders, ‘finance plus’-providers and regulated MFIs. This indicates
that two types of MFIs benefit more than others from focussing on
women: First, those MFIs that develop hands-on, women-friendly
procedures tailored to individual women’s need, and Second, those MFIs
that apply coercive enforcement methods to which women are more
responsive.
Key-words: Microfinance, Gender, Repayment, Portfolio-at risk, Write-offs, Provisions
JEL-classification codes: O10,O12
∗ All remaining errors remain our own responsibility.
1 Roy Mersland is the corresponding author. All correspondence can be addressed to roy.mersland@uia.no
4
1. Introduction
Microfinance, financial services tailored for poor people, has been celebrated for its
ability to reach out to women and enhance their welfare. From the starting point of
experimental schemes in Asia and Latin America in the 1970ties microfinance has been above
all a matter of women. Even today, the gender argument continues to be at the forefront. The
objective of the Microcredit Summit Campaign, which plays a central role in the promotion of
microfinance, is “to ensure that 175 million of the world’s poorest families, especially
women, receive credit for self-employment and other financial and business services” [our
emphasis]2. When the Nobel Prize was awarded to Mohammad Yunus and the Grameen Bank
the Nobel committee highlighted the role of microcredit in women liberation (Norwegian
Nobel Committee, 2006).
Among many Morduch (1999) argues that one of the main reasons for the success of
microfinance in the public eye is because the targeting of women. Indeed, Micro Finance
Institutions (MFIs) do target women. In this study’s dataset covering 350 MFIs from 70
countries women represent 73% of microfinance customers on average. This figure is close to
what has been found in previous literature (see for instance Cull et al., 2007; Daley-Harriss
2007: among others).
Another strong appeal of microfinance is the success of achieving high repayment
records. From an historical perspective this is not surprising. After all, modern microfinance
was born as a response to the frustrated development resulting from subsidised rural credit in
the 1950s-1980s. For example, Hulme and Mosley (1997) report default rates of up to half the
loan amount on small loans in Indian state banks in the late 1980’s. In addition to the use of
group collateral (Ghatak and Guinnane, 1999) and dynamic incentives like sequential loans
(Aghion and Morduch, 2003), the targeting of women has been put forward as a main
determinant of microcredit repayment. In this paper we test this assertion empirically: Do
MFIs targeting women experience higher repayment rates than other MFIs?
The assertion of women being good credit risks is regularly put forward by
microfinance advocacy networks and sponsors. For example ever since its first report in 1997
the Microcredit Summit reports that “women are consistently better in promptness and
reliability in payment” (Result, 1997: 8). The argument is repeatedly taken up by bilateral and
multilateral development aid agencies, including the World Bank ”[…] experience has shown
2 http://www.microcreditsummit.org/
5
that repayment is higher among female borrowers, mostly due to more conservative
investments and lower moral hazard risk” (World Bank, 2007: 124). Armendariz and
Morduch (2005: 139 sq), when they assess the different techniques to reduce repayment
defaults, consider targeting women as a technique in its own right alongside group lending or
dynamic incentives.
Strangely, the question of the link between the efficiency of the MFI and the targeting
of women has barely been rigorously studied and the existing empirical evidence is mixed.
Sponsors seem to evoke the repayment argument without ever furnishing it with empirical
evidence. Beside a lot of anecdotic evidence as well as the analysis of one or two MFIs in a
given country (see for instance Khandker, Khalily and Kahn, 1995 in Bangladesh or Kevane
and Wydick, 2001 in Guatemala) no paper of which we know has provided a detailed
empirical analysis of the gender-repayment issue within an international and longitudinal
context.
This paper makes use of a global dataset spanning 350 MFIs in 70 countries over 10
years to study whether the targeting of female customers significantly influences the MFIs’
repayment rates. Repayment is being studied through a wide variety of measures such as
portfolio-at risk, write-offs and loan-loss provision expenses. Thereby, we take into account
the specific methodological problems related to this type of estimation such as a. isolating the
gender-effect from other MFI or institutional effects influencing repayment and b. the time-
invariant nature of many covariates. The paper thus claims to be the first rigorous global
empirical study searching for evidence to the argument put forward that women are a good
credit risk for the MFI.
The findings indicate that more women clients is significantly associated with lower
portfolio-at-risk and lower portfolio write-offs, after properly controlling for a number of
MFI-specific factors as well as institutional factors. The gender-effects, which are not only
statistically significant, but are also economically relevant, indicate that women in general are
indeed a better credit risk for the MFI. Additionally, MFIs with more women clients carry less
loan-loss provisions ceteris paribus, providing additional evidence that focus on women
significantly reduces the MFI’s perceived credit-risk.
Looking at the issue more in detail, interaction effects indicate that the positive
repayment effect is particularly strong for certain categories of MFIs. Specifically, NGOs,
individual-based lenders, ‘finance-plus’-providers and regulated MFIs seem to benefit more
actively from focussing on women. This suggest that, although there is a consistent overall
positive impact of female clients on repayment, two types of MFIs are able to benefit even
6
more from focusing on women: a. MFIs who have developed individual women-friendly
procedures tailored to women’s needs and b. MFIs facing regulatory constraints and who
might use coercive enforcement methods to which women are more sensitive. These findings
might be of interest to policymakers and practitioners concerned with MFI-development and
repayment.
The rest of this paper proceeds as follows. A brief literature review on women and
microfinance is put forward in section 2 followed by section 3 where the literature on gender
and repayment is reviewed and the specific hypotheses to be tested are laid out. Section 4
explains the dataset and the statistical methodologies employed while section 5 reports the
findings. Section 6 concludes and presents the main implications of this study.
2. Women and microfinance
Various arguments that relate to both supply and demand for microcredit can explain
the targeting of women by microfinance organisations (Armendariz and Morduch, 2005;
Mayoux, 1999; World Bank, 2001; World Bank, 2007).
Firstly, demand for microfinance services is probably higher among women for a
number of reasons. In many countries, women are more credit constrained than men. They are
more restricted in their access to finance and control over land (Agarwal, 1994) and capital
(Fletschner, 2009). Consequently they are considered less creditworthy by traditional banks.
Lower education levels, as well as limited time and mobility also prevent them from engaging
with the complex and lengthy procedures usually requested by the formal banking sector.
Social norms are another factor: restrictions exerted by in-laws (in many countries the
financial dependence of women is fully integral to patriarchy) combine with discrimination
from bank staff. In some countries women do not even have the legal right to open a bank
account.
Gender aspects of the labour market are a second cause. A growth in the numbers of
women in self-employment and entrepreneurial activities explains increased demand for
microcredit (Armendariz and Morduch, 2005; Kevane and Wydick, 2001). Moreover, women
are more likely to pay the high interest rates required by many MFIs since they are more
restricted in their access to the formal labour market (Emran, Morshed, and Stiglitz, 2006).
As far as supply is concerned, three main arguments are usually put forward by donors
or practitioners in favour of targeting women: gender equality, poverty reduction and
efficiency (Mayoux, 2001). With respect to gender equality, microfinance is considered an
7
effective means to promote the empowerment of women. Drawing on the findings of
household economics developed over the last three decades3, it is suggested that gender
inequalities result in great part from inequalities in bargaining power in the context of
decision-making within the household. It is also suggested that women’s weaker bargaining
power results from their smaller contribution (real or perceived) to household cash flows and
to market-based income generating activities. By enabling women to develop or strengthen
income generative activities, microfinance is likely to increase their monetary income, their
control over their income and their bargaining power. These effects are then expected to lead
to various social, psychological and even political effects which are mutually reinforcing:
better self-esteem and self-confidence, improvement in status within the family and the
community, better spatial mobility and visibility of women in public spaces, etcetera.
As far as poverty reduction and efficiency is concerned, it is argued that women invest
their income to nurture the well being of their families– and this is supported by various
empirical studies conducted all over the world4: therefore one dollar loaned to a woman has
greater development impact than one dollar loaned to a man (World Bank, 2007: 165).
3. Gender and repayment
The relation between gender and repayment has been analyzed in a number of studies.
However, the evidence is mixed and usually anecdotic or very limited in geographical and/or
institutional scope. On the one hand, a number of studies find that women consistently
outperform men in terms of repayment. For instance, Armendariz and Morduch (2005) report
that in its initial faze the Grameen Bank also included men as customers. However, the bank
decided to move over to a nearly full concentration on women due to repayment problems
related to male customers. In a first empirical investigation, Hossain (1988) reports that in
Bangladesh 81 % of women encountered no repayment problems compared to 74 % of men.
Similarly, Khandker, Khalily and Kahn (1995) find that 15.3 % of Grameen’s male borrowers
had repayment problems compared to only 1.3 % of the women. Also from Bangladesh,
Sharma and Zeller (1997) report that credit groups with higher percentage of women had
significantly better repayment rates. From Malawi, Hulme (1991) reports that 92 % of women
pay on time, compared to 83 % for men, and Gibbons and Kasim (1991) find that in Malaysia
3 See for instance Sen (1990).
4 See for instance Chant (1985) ; Kabeer (1997) ; Haddad et Hoddinod (1995) ; Senauer (1990) ; Thomas
(1990).
8
95 % of women repay their loans compared to 72 % of the men. Finally, in a study from
Guatemala Kevane and Wydick (2001) report that female credit groups performed better than
male groups.
On the other hand, a number of studies find that there is no significant relation
between gender and repayment. In Bangladesh, the analysis carried out by Godquin (2004)
shows that correlation between gender and repayment is positive but not significant after
controlling for a number of MFI-specific effects. In a study reporting from four of the oldest
microfinance programs in the US Bhatt and Tang (2002) find that gender is in fact not a
significant determinant for loan repayment. The work done in Ethiopia by Brehanu and Fufa
(2008) leads to similar conclusions. Finally, BRI, a most reputed MFI in Indonesia, has never
had any specific focus on women, but still it has achieved nearly perfect repayment rates over
several years (Aghion and Morduch, 2005: 139). So in spite of popular belief, perhaps the
women repayment argument is not as clear cut as promoters seem to believe?
When it comes to theory, a number of arguments have been put forward to explain
gender-differences with respect to repayment rates (Armendariz and Morduch, 2005). For
instance, based upon her experience in Grameen-villages in Bangladesh, Todd (1996) argues
that women are more conservative or cautious in their investment strategies, and therefore
have better repayment records. Also from Bangladesh, Rahman (2001) and Goetz and Gupta
(1996) argue that women are more easily influenced by peer-pressure and the interventions of
the loan-managers. For matters pertaining to reputation and honor, women are believed to be
more sensitive to verbal hostility on the part of the loan-manager, while men are able to
default with a sense of impunity.
Another argument put forward is that female customers tend to stay closer to their
homes rather than going out to work. They can therefore be more easily monitored and
followed up by the MFI (Aghion and Morduch, 2005; Goetz and Gupta, 1996). Aghion and
Morduch (2005) also argue that since men can more easily access credit in other formal or
informal channels, women have more at stake when enrolling in a credit-program. They thus
have to repay to ensure continued access to credit. Ameen (2004) argues that women have a
lower opportunity cost of time than men making their time less valuable than for their male-
counterparts. As a result, they are more inclined to have more contact with the MFI (including
group-meetings, meetings with loan-officers) which altogether has a positive impact on their
repayment. Goetz and Gupta (1996) suggest that women may have a higher incentive than
men for loan repayment since it allows them to retain access to village groups, whereas men
have many more opportunities for social contact.
9
However, theoretical counter-arguments of women being better credit risk can also be
put forward. For instance, Philips and Bhatia-Panthaki (2007) argue that women entrepreneurs
tend to be over-represented in traditional sectors with relatively lower profits, fewer growth
opportunities and harsher competition. This should make them less able to honour credit
contracts. Somewhat in line with this argument, various studies point out that many women
borrowers don’t have any control over their own microcredit: loans are in fact used and
controlled by men within the household (Goetz and Gupta, 1996; Rahman, 1999; Kabeer,
2001; Mayoux, 2001; Montgomery, 1996). This could have a negative impact on women
repayment-rates. The previous discussion shows that the relation between gender and
repayment remains largely unresolved.
Based upon the arguments put forward in the above literature review we propose the
following main hypothesis:
H1. The proportion of female customers in an MFI has a negative impact on its default rates.
The proportion of portfolio overdue with more than 30 days (portfolio-at-risk) and the
percentage of the loan portfolio that is written-off because of non-repayment (write-off ratio)
are used as proxies for the default rate.
For robustness, we also analyze the relation between gender and the perceived credit-
risk directly as follows:
H2. The proportion of female customers in an MFI has a negative impact on its perceived
credit risk.
We use loan-loss provisions (measured in terms of the provision expense rate) as an
indicator of the MFI’s perceived credit risk5.
The impact of women on repayment might be more prevalent in certain categories of
MFIs or be more prevalent under certain conditions. In other words, while the previous
hypotheses investigate whether gender influences repayment in general, we also want to
provide a more detailed analysis by looking whether the gender-repayment relation differs for
5 For some of the regulated MFIs the regulator defines minimum levels of provisions while in some countries the
local tax authorities may indicate maximum levels. Most MFIs do however themselves define their final level of
risk provisions.
10
certain categories or features of the MFI. Therefore, we analyze whether the effect differs
with experience, legal status, activities, lending methodology and regulation.
a. MFI experience
Experience might be a variable influencing the gender-repayment relation. For
instance, MFIs might adapt their internal procedures and rules as they gain more experience
over time. Similarly, more experienced staff members might influence customers’ repayment
behavior. Therefore we expect that the gender-repayment relation might differ with MFI-
experience which can be described by following hypothesis:
H3a. The impact of the proportion of female borrowers on repayment differs with MFI-
experience.
b. Legal status
Gender-differences with respect to repayment might differ with the MFI’s legal-status.
For instance, non-profit organizations like NGOs have broader objectives and governance
forms that make them more likely to service efficiently marginalized customers like women
(Mersland, 2009). As a result, they are also more likely to develop specific gender policies
which might influence the relation of gender on repayment. Therefore, the following
hypothesis can be derived:
H3b. The impact of the proportion of female borrowers on repayment differs with the MFI’s
legal status.
c. Activities
The different activities that an MFI is involved in might have an impact on the gender-
repayment relation. For instance, the provision of non-financial services alongside
microfinance – often referred to as ‘microfinance plus’ - like health services and basic literacy
training (Godquin, 2004), or business training (Khandker, Khalily and Kahn, 1995) could
improve repayment performance. As suggested by Edgcomb and Barton (1998) non-financial
services not only improve the economic ability of the borrower to repay but also the quality of
the relationships between MFIs and their clients, helping to strengthen trust and confidence.
MFIs providing non-financial services normally do so in order to service poorer and more
11
marginalized customers (Lensink and Mersland, 2009). It also argued that that women more
readily accept non-financial services, whilst also needing them more (Armendariz and
Morduch, 2005; Mayoux, 2001). In conclusion, the impact of gender on repayment might be
influenced by the MFI’s activities which can be described by the following hypothesis:
H3c. The impact of the proportion of female borrowers on repayment differs with the MFI’s
activities.
d. Lending methodology
MFI’s employ different lending methodologies such as village-banking, solidarity
groups or individual-based lending (Sharma and Zeller, 1997; Kevane and Wydick, 2001).
The relation between gender and repayment might differ with the lending methodology that is
being used. Group lending is usually considered as a ’female’ method: women would accept
more easily to join groups and to spend time in group meetings. But the group-lending
methodology as such is more and more criticized (Harper 2007) and recently MFIs switch
towards individual-based methods. Indeed, there is no clear evidence that group-lending lead
to better repayment for women6. Individual-based lending methodologies might give a more
personal monitoring and direct enforcement which could influence the repayment rates for
either male or female borrowers. In conclusion, the gender-repayment relation might be
influenced by the MFI lending methodology described by the following hypothesis:
H3d. The impact of the proportion of female borrowers on repayment differs with the MFI’s
lending methodology.
e. Regulation
Regulated MFIs are monitored by banking authorities. Regulation sometimes leads to
a “mission drift” if demands to fulfil requirements divert attention away from serving the poor
(e.g. by shifting the focus from serving poor clients to improve capital adequacy ratios).
Similarly, regulation may hold back innovation in lending technology that has been the
driving force behind MFIs’ ability to expand outreach and serve poor clients (Dichter, 1997,
6 Some studies focused on group lending and repayment have paid a specific attention to gender (see for instance
Sharma & Zeller (1997); Wydick et al. (1999); Kevane & Wydick 2001). But in many studies it is not an issue
since all or almost all borrowers are women. See for instance Paxton et al.,. (2000); Karlan & Giné (2006);
Karlan (2007); Cassar & al. (2007). There is no information about gender in Alhin & Towndsend (2007).
12
Hardy, Holden and Prokopenko, 2003). Alternatively, regulated MFIs might have more
pressure on behalf of the regulating authority than non-regulated ones, which could have an
influence on their gender and repayment-strategies. As a result, the effect of gender on
repayment might differ between regulated and unregulated MFIs as described by the
following hypothesis:
H3e. The impact of the proportion of female borrowers on repayment differs between
regulated and non-regulated MFIs.
4. Data and estimation methods
4.1 Data and summary statistics
We use observations of 350 rated MFIs from 70 countries. Specialized rating agencies
perform the assessment reports. A main motive behind submitting to a rating is the potential
access to external funding from investors. A major advantage of the assessment reports is that
they are worked out by a third-party and cover a wide range of organizational features
alongside financial data and social and financial indicators. For example the dataset contains
information on both conscious gender bias towards women and proportion of customers being
female in order to study the gender-issue. At each rating four years of data are obtained, at
best. The ratings are performed in the period 2001 to 2008, which means that we have data
from 1998 to 2008. Most data are from the period 2001 to 2006.
No dataset is perfectly representative of the microfinance field. In particular, our
dataset contains relatively fewer of the mega-sized MFI, and it does not cover the virtually
endless number of small savings and credit cooperatives. The former are rated by such
agencies as Moody’s and Standard and Poor, while the latter are not rated.
Different inflation rates in 70 countries make comparisons difficult for all monetary
variables. We solve this by converting the monetary variables into USD amounts at the going
exchange rate. From the purchasing power parity theorem of international finance (Solnik and
McLeavy, 2003), the conversion into USD implies that the local inflation has been adjusted
for.
In table 1 we report mean, standard deviation, minimum, maximum and quartiles for a
number of key-variables for our sample. We see that the average MFI has total assets of $
6,519,000; a total loan portfolio of $ 4,225,000; serves 17,111 clients and has 9 years of
experience in the sector. The average loan size is $787 dollar and annual net results are
13
$209,000.The portfolio yield is 0.39 on average and the OSS is 1.12 indicating that net
income exceeds operating expenses on average.
Looking at the repayment-variables, we see that PaR30 equals 0.06 on average which
means that 6% of the total loan portfolio is 30 days or more on arrear. However, the median
value is only 3%, which indicates that the distribution is somewhat skewed with a number of
MFIs having high values for PaR30. On average, write-offs are around 1% of total loan
portfolio and the median value is also 1%. The provision expense rate is 0.03 on average,
which means that provisions are held for 3% of the total loan portfolio.
For the gender variables, we see that on average, MFIs have 73% women clients,
indicating that there is in general a substantial focus on women within our sample. Moreover,
the 75th percentile in the distribution of female clients is 1, indicating that at least 25% of the
MFIs focus exclusively on women. The dummy on conscious gender bias indicates that
around 40% of the MFIs report to have a conscious bias towards women.
Table 1. Summary statistics
14
In this table, we present summary statistics for key-variables of the sample under study. Q1, Q2 and Q3 are the first, second and third
quartile, respectively. Total assets is inferred from the balance sheet and measured in $ 1,000. Loan portfolio is the total loans
outstanding and is measured in $ 1,000. Total clients is the total number of credit and savings clients active with the MFI. Loan size
is the average loan outstanding measured in $ and is defined as gross outstanding portfolio per client. Experience measures the
number of years the institution is active in microfinance activities. Employers is the number of full-time employers active in the MFI.
Staff Efficiency is the number of total clients divided by the number of employers. Loan Officer Efficiency is the number of total
clients divided by the number of loan officers. Annual return is the net result before extra-ordinary income and expenses, donations
and taxation and measured in $ 1,000. Portfolio yield is the percentage yield on the MFI’s total portfolio. OSS is operational self-
sufficiency measured as net income dividend by operating expenses. PaR30 is portfolio at risk measuring the part of the loan
portfolio more than 30 days in arrears. Write-offs measures the part of the total loan portfolio that has been written of and therefore is
accepted as a loss. Provision expense rate is loss loan provision as a percentage of total loan portfolio. Risk coverage rate is the loss
loan provisions divided by PaR30. DUMgender is a dummy variable that is 1 if the MFI has a conscious gender bias and 0 otherwise.
Women clients measures which percentage of the MFI’s clients is female.
Variables N mean Q1 Q2 Q3 st.dev min max
General
Total assets 1,201 6,519 1,036 2,593 6,876 1,470 19 250,000
Loan portfolio 1,217 4,225 752 1,918 4,921 6,222 12 59,700
Total clients 1,001 17,111 2,329 5,780 14,625 41,924 113 534,342
Loan size 1,155 787 147 381 886 976 0 28,693
Experience 3,208 9 4 7 12 8 0 84
Employers 1,147 89 24 50 94 140 2 1,842
Staff efficiency 1,138 129 67 108 170 100 2 1,893
Loan officer
efficiency 1,083 289 160 239 358 270 5 4,591
Annual result 1,191 209 -14 48 254 701 -3,533 11,800
Portfolio yield 1,147 0.39 0.24 0.34 0.49 0.24 0.02 5
OSS 716 1.12 0.95 1.11 1.32 0.38 0.07 2.94
Repayment
PAR30 1,100 0.06 0.01 0.03 0.07 0.1 0 0.98
Write-offs 1,020 0.01 0.00 0.01 0.02 0.1 0.00 0.74
Provision expense
rate 1,075 0.03 0.01 0.02 0.04 0.04 -0.06 0.63
Gender
DUMgender 2,934 0.4 0 0 1 0.49 0 1
Women clients 1,267 0.73 0.55 0.76 1 0.25 0.08 1
Notes:
- Obvious special cases have been omitted from the analysis.
- We have data on DUMgender and women clients only in the MFI’s rating year. Therefore, we have assumed
DUMgender and women clients to be constant over time.
In Table 2 we report the correlation matrix of the gender and repayment variables
(Panel A) as well as discrete mean values for some key-variables in different classes of
proportion of female borrowers. As can be seen from Panel A, there is a negative correlation
between the proportion of female borrowers and both PaR30 (-0.02) and write-offs (-0.09).
Although this is only a univariate correlation, this is a first indication that MFIs who have
more female borrowers obtain better repayment rates (hypothesis 1). Additionally, the
correlation between the proportion of female borrowers and the provision expense rate is also
negative (-0.14) which provides a first indication that MFIs who have a higher proportion of
15
female clients carry less provisions and therefore are potentially a better credit risk
(hypothesis 2).
In Panel B, we analyze mean values for a number of key variables in classes that differ
in their proportion of female borrowers going from ‘very low’ to ‘very high’. The different
cut-off points for the classes correspond to the quartiles in the distribution of the proportion of
female clients. As can be seen from Panel B, mean PaR30 is 0.03 and 0.05 for the classes
‘very low’ and ‘low’, respectively. This is higher than for the classes ‘high’ (0.02) and ‘very
high’ (0.01). Mean write-offs are also somewhat lower in the classes ‘high’ (0.007) and ‘very
high’ (0.004) when compared to the classes ‘low’ (0.02) and ‘very low’ (0.01). Again, this
seems to support the findings of the correlation matrix that higher proportion of female
borrowers is associated with lower portfolio-at-risk and lower loan-loss write-offs.
For the provision expense rate we see that the mean values are lower in the classes
‘high’ and ‘very high‘indicating that MFIs with more female clients carry less provisions.
This supports our second hypothesis that female clients are perceived as better credit risk.
Table 2. Univariate statistics
Panel A. Correlations
This table reports the correlations between the gender and repayment variables.
Women clients Par30 write-offs provision expense rate
Women clients 1
Par30 -0.02 1
Write-offs -0.09 0.12 1
Provision expense rate -0 .14 0.23 0.58 1
Panel B. Median values in different gender-classes
This table reports median values for a number of key-variables in different classes that differ in the proportion of
female borrowers in columns 1-4. The different classes correspond to different quartiles in the distribution of women
clients going from very low (women clients < Q1 =0.55), to low (Q1=0.55 < women clients< Q2 = 0.76) to high
(Q2=0.76 < women clients < Q3 = 1) to very high (women clients > Q4 = 1). Column 5 (6) reports median values for
MFIs with (without) a conscious gender bias.
Women clients Conscious gender bias
Very low
(1)
low
(2)
High
(3)
very high
(4)
yes
(5)
no
(5)
PaR30 0.03 0.05 0.02 0.01 0.02 0.04
Write-offs 0.01 0.02 0.007 0.004 0.008 0.01
Provision expense
rate 0.03 0.02 0.02 0.01 0.02 0.02
Total assets 3,149 1,866 2,506 1,225 1,973 2,648
Portfolio yield 0.3 0.38 0.47 0.4 0.38 0.33
OSS 1.16 1.14 1.11 0.99 1.1 1.12
Loan size 931 522 218 101 157 653
Loan portfolio 2,092 1,457 2,051 878 1,241 1,964
16
4.2 Estimation methods
To test our hypotheses, we use panel data regression techniques where the independent
variables are the proportion of female borrowers and a wide variety of MFI-specific and
institutional controls. This enables us to quantify the impact of the proportion of female
borrowers on the different repayment-variables while holding constant other variables that
potentially affect repayment. For instance, to test hypothesis 1, we regress PaR30 on the
proportion of female borrowers, controlling for a wide range of MFI-specific factors as well
as institutional factors as follows:
(1)
where PaR30i,t is the portfolio-at-risk for MFI i in year t, FEMi,t is the percentage of female
clients, Zi,t is a matrix of MFI-specific controls such as size, experience, average loan size,
dummies for lending-methodology, rural/urban market activity etc, and Xi,t is a matrix of
controls that capture the conditions of country C in which the MFI is active.
Two important methodological issues surrounding this kind of estimation require close
attention7. Firstly, it is important that all control variables that potentially affect repayment
should be included explicitly as controls in the regression equation. If variables are omitted
that affect repayment and which are also correlated with the proportion of female borrowers,
the OLS-estimates could be biased as a result of omitted variables bias (see for instance Stock
and Watson, 2007, p.186). While we take up controls for a wide variety of MFI-specific and
institutional factors, it is possible that some unobserved factors related to both repayment and
proportion of female borrowers could cause the OLS-estimates to be inconsistent and
potentially biased. Therefore, besides OLS-estimates we also analyze pooled random
coefficients models (RE) that takes up a MFI-specific unobserved effect as follows:
(2)
where
μ
i is the unobserved MFI-specific effect. The main benefit of such a random
coefficients model is that all unobserved heterogeneity potentially affecting the dependent
variable is taken up by the MFI-specific effect and therefore a potential omitted variable bias
is avoided (see for instance Stock and Watson, 2007, p 349).
7 These methodological issues are more rule than exception within the microfinance-context. For more
information, see for instance Mersland and Strøm (2009b) or Hatarska (2005).
17
Secondly, many of the controls that need to be added are time-invariant such as
institutional dummies, dummies on rural or urban market, etcetera. Incorporating time-
invariant covariates in the context of panel-data requires the additional assumption that the
time-invariant covariates are always uncorrelated with the unit-fixed effect. When this
assumption does not hold, the random effects estimator might yield inconsistent and biased
estimates (see Baltagi et al., 2003 for a detailed discussion). Therefore, besides performing
OLS and RE, we also report the Fixed Effects Vector Decomposition-estimator (FEVD)
developed by Plümper and Troeger (2007). This estimator is designed to tackle time-invariant
covariates and unit- fixed effects in the context of panel data and employs a three-stage
estimation procedure. The first stage estimates a pure fixed effects model to obtain an estimate
of the unit- fixed effect. The second stage decomposes the fixed effects into a part explained
by the time-invariant variables and an unexplained part. The third stage re-estimates the
model including the time-invariant variables and the error term of the second stage using
pooled OLS.
5. Results and discussion
5.1 The gender-repayment relation
In Table 3 we analyze the impact of gender on the repayment-measures in terms of
portfolio-at-risk (Panel A) and write-offs (Panel B). For gender we use two proxies, namely
the proportion of female clients and the conscious gender dummy that is 1 if the MFI reports
to have a conscious bias towards women and 0 otherwise. The different columns correspond
to the different estimation methods that have been used for robustness (OLS, RE and FEVD).
Looking at columns (1) to (3) we see that the proportion of female borrowers is
negatively related to the portfolio-at-risk and the coefficients are quite robust over the
estimation methods. These estimated effects are statistically significant, although confidence
levels may vary somewhat between estimation methods8. Additionally, the regression
statistics (F-stat for OLS and FEVD and χ²-stat for RE) always denote joint significance of the
models.
Looking at the other controls we see that mainly size, portfolio growth, dumRURAL
and HDI are significantly related to PaR30. In particular, a lower PaR30 is associated with
8 This is not surpirsing given the fact that the different estimation methods require a different number of
parameters to be estimated and hence, loss in degrees-of-freedom may vary substantially over the different
stilation methods.
18
larger MFIs, MFIs who have a higher portfolio-growth, MFIs who operate in rural areas and
MFIs operating in richer countries. These effects are robust over the estimation methods and
fully in line with our expectations. MFIs that grow will normally have lower PaR as a
significant part of the portfolio is new and still uncontaminated and rural clients are easier to
monitor and control.
The coefficients on experience and the efficiency measures (staff efficiency, credit
officer efficiency) are insignificant and close to zero. This suggests that there is no significant
effect of experience or efficiency measures on the portfolio-at-risk. Similarly, whether the
MFI is an NGO or not, or whether the MFI practice group or individual lending have little or
no effect on portfolio-at-risk. This is in line with Mersland and Strøm (2008) who find that
performance differences between NGOs and non-NGOs is minimal, and Mersland and Strøm
(2009a) who demonstrate that individual and group lenders don’t differ much when it comes
to repayment records.
In columns (4)-(6), gender preference is measured through the dummy on conscious
gender bias. As can be seen, the dummy on conscious gender bias is, like the proportion
female clients, negatively related to the portfolio-at-risk with the coefficients being highly
significant. MFIs who report to have a conscious gender bias towards women have a
significantly lower Par30. Looking at the control variables, we see again a negative
association with size, portfolio growth, dumRURAL, and HDI.
In panel B, repayment is measured by the actual loan loss write-offs, and again gender
is measured through both the proportion of female borrowers and the dummy on a conscious
gender bias. As can be seen from columns (1) to (3) the proportion of female borrowers is
significantly negatively related with the write-off rate and estimated coefficients are highly
significant at the 1% significance level. This means that MFIs who focus more on women
have significantly lower write-offs, ceteris paribus. Looking at the control variables we see
that size, portfolio growth, DumNGO and dumRURAL are significant determinants of the
MFI’s write-off rate. Specifically, a higher write-off rate is associated with smaller MFIs,
lower portfolio growth, NGOs and MFIs operating in rural areas.
Overall, the results from Table 3 point towards a negative association between the
number of female clients and repayment, confirming hypothesis 1 that the proportion of
female clients reduces the MFI’s default rate. This effect holds for several measures of
repayment (par30 and write-offs), several gender-measures (percentage female borrowers and
conscious gender bias) and for several estimation methods (OLS, RE and FEVD).
19
Table 3. Gender and loan repayment
In this table we analyze the impact of gender on loan repayment both in terms of PaR30 (panel A) and write-offs (panel B).
DumNGO is a dummy that is 1 if the MFI is an NGO and 0 otherwise, DumGroup is a dummy that is 1 if the MFI provides loans on
a group basis (such as village-bankers or group-lenders). DumRural is 1 if the MFI operates mainly in rural areas and 0 otherwise.
Dum performance pay is 1 if the MFI pays incentive-based salaries and 0 otherwise, HDI is the human development index. All other
variables are defined as in Table 1. OLS indicates that pooled OLS has been used as the estimation method. RE means that a pooled
random effects model has been estimated and FEVD means that the Fixed Effects Vector Decomposition-estimator has been used.
Robust standard errors are provided in parentheses. *, ** and *** denote statistical significance at the 10%, 5% and 1% significance
level, respectively.
Panel A. Repayment in terms of PaR30
Dep.var. PaR30
(1)
OLS
(2)
RE
(3)
FEVD
(4)
OLS
(5)
RE
(6)
FEVD
gender
women clients -0.02 -0.05 -0.05
(0.015)* (0.038)* (0.003)***
conscious gender bias -0.01 -0.02 -0.02
(0.005)*** (0.012)* (0.001)***
MFI-controls
general
Experience 0.002 0.00 0.00 0.00 0.00 0.00
(0.001)*** (0.001) (0.000) (0.002) (0.000) (0.000)
lnTA -0.02 -0.01 -0.01 -0.01 -0.01 -0.01
(0.004)*** (0.002)*** (0.001)*** (0.002)*** (0.002)*** (0.004)***
Loansize 0.02 -0.01 -0.02 0.01 0.01 0.01
(0.006) (0.005)*** (0.002)*** (0.004) (0.005) (0.001)
Portfolio growth -0.05 -0.02 -0.02 -0.07 -0.02 -0.02
(0.008)*** (0.004)*** (0.002)*** (0.007)*** (0.004)*** (0.002)***
Legal status
DumNGO 0.00 0.00 0.00 0.01 0.01 0.02
(0.007) (0.016) (0.001) (0.004)*** (0.011) (0.001)***
Loan methodology
DumGroup 0.00 -0.01 -0.01 0.01 0.00 0.00
(0.006) (0.018) (0.002)*** (0.005) (0.012) (0.002)
DumRural -0.04 -0.01 -0.03 -0.03 -0.03 -0.04
(0.008)*** (0.009) (0.003)*** (0.005)*** (0.011)*** (0.002)***
Efficiency
Staff efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Credit officer efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Dum performance pay 0.01 -0.01 -0.01 0.02 0.01 0.01
(0.005) (0.014) (0.001) (0.003) (0.011) (0.001)
Country controls
HDI -0.13 -0.14 -0.11 -0.13 -0.12 -0.1
(0.026)*** (0.067)** (0.006)*** (0.019)**** (0.044)*** (0.006)***
Model statistics
N 830 830 830 1748 1748 1748
R² 0.21 0.1 0.96 0.24 0.16 0.92
F-stat / Wald ?² 13.41*** 100.22*** 1408.59*** 38.04*** 67.19*** 1419.38***
20
Panel B. Repayment in terms of write-offs
Dep.var. Write-off rate
(1)
OLS
(2)
RE
(3)
FEVD
(4)
OLS
(5)
RE
(6)
FEVD
gender
women clients -0.02 -0.03 -0.03
(0.005)*** (0.013)*** (0.013)***
conscious gender bias
-0.01 -0.01 -0.01
(0.003)*** (0.005) (0.002)***
MFI-controls
general
Experience 0.00 0.00 0.00 0.00 0.00 0.00
(0.000 (0.000 (0.000 (0.000 (0.000 (0.000
lnTA 0.00 -0.01 -0.01 0.00 0.00 -0.01
(0.001)** (0.002)*** (0.001)*** (0.001)*** (0.002)** (0.006)***
Loansize 0.00 0.00 0.00 0.00 0.00 0.00
(0.001) (0.004) (0.001) (0.001) (0.002) (0.001)
Portfolio growth -0.05 -0.01 -0.01 -0.05 -0.04 -0.03
(0.004)*** (0.004)*** (0.002)*** (0.007)*** (0.01)** (0.002)***
Legal status
DumNGO 0.02 0.01 0.02 0.01 0.00 0.01
(0.002)*** (0.005)*** (0.001)*** (0.002)*** (0.001) (0.001)***
Loan methodology
DumGroup 0.00 0.00 0.01 0.01 0.00 0.01
(0.003) (0.006) (0.001)*** (0.003)*** (0.005) (0.002)***
DumRural 0.00 0.00 -0.01 0.00 0.00 0.00
(0.002)** (0.006) (0.001)*** (0.002) (0.005) (0.002)
Efficiency
Staff efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Credit officer efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Dum performance pay 0.01 0.00 0.01 0.01 0.01 0.01
(0.002) (0.001) (0.005) (0.002) (0.004) (0.001)
Country controls
HDI -0.01 -0.01 -0.03 -0.04 -0.04 -0.05
(0.009) (0.024) (0.006)*** (0.012)*** (0.017)*** (0.006)***
Model statistics
N 773 773 773 1621 1621 1621
R² 0.35 0.28 0.81 0.22 0.30 0.62
F-stat / Wald χ² 23.91*** 54.62*** 239.56*** 16.86*** 131.01*** 199.02***
Notes:
- Sometimes, especially in the case of RE and FEVD, the number of parameters to be estimated exceeds the number of observations
which makes the exercise an underdetermined problem. As Fraser (2000) points out, classical estimation methods do not apply. In
that case, missing values have been imputed with averages over the time-period in order to boost the number of observations to
reasonable levels.
- Loan size has been scaled by a factor 1/1000 in order to make the coefficients easier to read.
- For robustness, we have experimented with other country-controls such as GDI (gender-development-index). Similar results obtain.
- For robustness, we have also divided dumGROUP into separate dummies for village-bankers (DumVILL) and solidarity-groups
(DumSOL). Similar results obtain.
In Table 4 we analyze the impact of gender on the provisions measured as the loan-
loss provision expense rate. Again gender is measured both in terms of the proportion of
21
female borrowers as well as a dummy for conscious gender bias and the different columns
represent different estimation methods that have been used. As can be seen from columns (1)
to (3) the coefficient on the proportion of women clients is always negative and significance
levels vary between 10% for RE and 1% for OLS and FEVD. This means that an MFIs with
more women clients carry significantly lower provisions.
From columns (4) to (6) we see that the coefficient on the conscious gender bias is also
negative regardless of the estimation method that is being used. However, significance levels
vary somewhat with the estimation method (1% significance level for FEVD, 5% significance
level for OLS and insignificance for RE). Regarding control variables in table 4 the main
difference with table 3 is that NGOs carry more provisions compared to non-NGOs. This is
not surprising since they are generally not regulated and often don’t pay taxes. NGOs can
therefore fix their own level of provisions and write-offs.
Overall, we find consistent evidence that MFIs who focus more on women clients
carry lower provisions, controlling for other MFI-specific and country-specific effects. This
finding supports hypothesis 2 that the proportion of female customers has a negative impact
on its perceived credit risk.
22
Table 4. Gender and provisions
We analyze the impact of gender on the loan loss provisions measured in terms of provision expense rate. DumNGO is a
dummy that is 1 if the MFI is an NGO and 0 otherwise, DumGroup is a dummy that is 1 if the MFI provides loans on a group
basis (such as village-bankers or group-lenders). DumRural is 1 if the MFI operates mainly in rural areas and 0 otherwise.
Dum performance pay is 1 if the MFI pays incentive-based salaries and 0 otherwise, HDI is the human development index.
All other variables are defined as in Table 1. OLS indicates that pooled OLS has been used as the estimation method. RE
means that a pooled random effects model has been estimated and FEVD means that the Fixed Effects Vector
Decomposition-estimator has been used. Robust standard errors are provided in parentheses. *, ** and *** denote statistical
significance at the 10%, 5% and 1% significance level, respectively.
Dep.var. provision
expense rate
(1)
OLS
(2)
RE
(3)
FEVD
(4)
OLS
(5)
RE
(6)
FEVD
gender
women clients -0.02 -0.02 -0.02
(0.005)*** (0.012)* (0.003)***
conscious gender bias -0.01 -0.01 -0.01
(0.002)** (0.004) (0.001)***
MFI-controls
general
Experience 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.001) (0.000) (0.001) (0.000) (0.000)
lnTA -0.002 -0.006 -0.01 -0.01 -0.01 -0.01
(0.001)* (0.002)*** (0.007)*** (0.001)*** (0.001)*** (0.005)***
Loansize 0.00 0.00 0.00 0.00 0.00 0.00
(0.002) (0.004) (0.001) (0.016) (0.001) (0.006)
Portfolio growth -0.03 -0.01 -0.01 -0.02 -0.01 -0.004
(0.004)*** (0.004)** (0.005)*** (0.003)*** (0.003)** (0.002)**
Legal status
DumNGO 0.010 0.010 0.010 0.010 0.000 0.003
(0.002)*** (0.005)** (0.002)*** (0.002)** (0.003) (0.011)***
Loan methodology
DumGroup 0.00 0.00 0.00 0.00 0.00 0.00
(0.002) (0.006) (0.002) (0.002) (0.001) (0.001)
DumRural 0.00 0.00 0.00 0.00 0.00 0.00
(0.001) (0.006) (0.001) (0.002) (0.005) (0.001)
Efficiency
Staff efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Credit officer efficiency 0.00 0.00 0.00 0.00 0.00 0.00
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Dum performance pay 0.01 0.01 0.01 0.01 0.01 0.01
(0.002)*** (0.005)** (0.001)*** (0.002)*** (0.003)*** (0.001)***
Country controls
HDI -0.02 -0.03 -0.04 -0.05 -0.04 -0.05
(0.009)** (0.023)* (0.006)*** (0.009)*** (0.015)*** (0.005)***
Model statistics
N 751 751 751 1624 1624 1624
R² 0.25 0.27 0.72 0.15 0.18 0.59
F-stat / Wald χ² 20.66*** 43.18*** 149.04*** 22.13*** 53.49*** 181.74***
Notes:
- Sometimes, especially in the case of RE and FEVD, the number of parameters to be estimated exceeds the number of
observations which makes the exercise an underdetermined problem. As Fraser (2000) points out, classical estimation
methods do not apply. In that case, missing values have been imputed with averages over the time-period in order to boost the
number of observations to reasonable levels.
- Loan size has been scaled by a factor 1/1000 in order to make the coefficients easier to read.
- For robustness, we have experimented with other country-controls such as GDI (gender-development-index). Similar results
obtain.
23
- For robustness, we have also divided dumGROUP into separate dummies for village-bankers (DumVILL) and solidarity-
groups (DumSOL). Similar results obtain.
5.2 Interaction effects
The previous discussion shows that focus on women has a positive impact on
repayment in general. However, as we have argued in the literature review section, this effect
might be more prevalent under certain conditions, or apply more for certain categories of
MFIs. Analyzing these hypotheses amounts to including additional interaction terms in the
regression equations as follows:
(3)
where all variables are defined as in equation (2) and (FEMi,t * INTi,t) is the interaction term
that measures whether the effect of female clients on repayment differs with the different
interaction variables INTi,t.
Regression outputs with respect to the interaction terms are provided in table 5. The
different columns represent the different interaction terms that were added. In the last column,
all interaction variables were taken up simultaneously to isolate the impact of the each of the
interactions. The coefficient on women clients now only represents the impact of female
clients on repayment in the reference category, whereas the sum of the reference coefficient
and the coefficient on the interaction term returns the gender-repayment effect for the
different categories.
As can be seen from column (1) the interaction term (women clients * age) returns an
insignificant coefficient that is close to zero. This indicates that the gender-repayment relation
does not differ with the MFI’s experience. Therefore, hypothesis 3a cannot be supported.
Column (2) indicates that for non-NGOs the effect is insignificant and close to zero whereas
for NGOs the effect is around -0.09 (0.01 + -0.10) which is in support of hypothesis 3b. This
effect remains when the other interactions are added as can be seen from column (6).
Looking at column (3) we see that providing financial services only reduces the
positive impact of female customers on repayment with 0.01. For MFIs who provide financial
services only, an increase in women clients by 1 is associated with a decrease in PaR30 of
4%, whereas for MFIs who provide additional activities, an increase by 1 of the proportion
female clients is associated with a decrease in Par30 of 5%.
24
From columns (4) and (5) the following main results can be derived: The effect of
women on repayment is stronger for MFIs who provide loans on an individual basis (-0.11) in
comparison to village bankers (-0.02) and MFIs who operate on the basis of solidarity groups
(-0.05). The effect is stronger for regulated MFIs (-0.08) then for non-regulated MFIs (-0.04).
Overall, we find that the general impact of women on repayment is indeed different for
different categories of MFIs. We find substantial support for hypotheses 3b, 3d, 3e while
finding only weak support of hypothesis 3c and no support for 3a.
Specifically, the positive effect of women on repayment seems significantly stronger
for NGOs, MFIs who provide loans on an individual basis, MFIs who provide other services
alongside pure financial services and regulated MFIs. These findings suggest that there are
two mechanisms that actively enhance the positive effect of women on repayment. First, MFIs
who apply more personal individual methods such as NGOs, finance plus providers and
individual-based lenders benefit more from their focus on women. This suggests that a more
personalised, tailor-made approach that is better adapted to the specificities of women
increases the positive impact of women on repayment. For instance: gender awareness in the
staff recruitment, working hours adapted to women’s domestic obligations, easy procedures,
repayment schedules appropriate to businesses activities which are specific to women, seem to
enhance women’s repayment rates.
In contrast to commonly held ideas, it is interesting to see that the positive impact of
women on repayment is greater in the case of individual loans. This kind of result could be
illustrative of the poor quality of particular forms of group-lending. As argued by Harper
(2007), group lending is all too often used as a « second-hand » method. This seems to be
especially true for women, for whom group-lending is mainly a way to shift transaction costs
onto female borrowers (Mayoux 2001 ; Molyneux 2002; Rankin, 2002 ; Rao 2008 ; Wright
2006).
However, this argument does not explain why the positive women-repayment effect is
stronger for regulated MFIs. Therefore, there might be a second mechanism that effectively
enhances the positive women-repayment effect. Regulation is often associated with ‘mission
drift’, which is usually understood as less attention being paid to poor clients and towards
women. We suggest that individual lending and regulation can also translate into more
coercive enforcement methods (social pressure, verbal hostility, harassment, etc.). Moreover,
as suggested earlier, it seems that women are more responsive to coercive practices. In
Bangladesh for instance, the regulation and financial sustainability constraints of the 1990s
have strongly influenced credit officers’ behaviors (Goetz and Gupta 1996; Rahman 1999,
25
2004; Huq 2004). Faced with increasing productivity constraints, credit officers have not only
used more robust enforcement methods, but also primarily target women, since they know
from experience that they repay better. In the description given by Goetz and Gupta (1996),
microcredit officers clearly explain why they avoid men: it is owing to their lack of
commitment, their greater ability to evade both development agents and the law, the threat of
violence, and the greater range of opportunities available to men for squandering credit in the
context of urbanization and Westernisation (Goetz and Gupta 1996, p. 55). In India, where
women represent 95% of clientele, regulation and competition constraints have also led to
‘abusive collection methods’ (APMAS 2006, Ghate 2007, p. 97). Drawing on field work in
various parts of the world, Fernando et al.,. (2006) report similar findings and offer various
examples where regulation constraints translate into increased pressure on microfinance
clients, especially women.
Table 5. Interactions on the gender-repayment relation
In this table, we analyze whether the positive effect of women on repayment differs with experience, legal status, activities,
lending methodology and regulation by investigating various interaction effects on the percentage of women clients. Dum
fin.only is a dummy variable that is 1 if the MFI provides financial services only and 0 if the MFI is engaged in other
activities as well. Dum Vill is a dummy that is 1 if the MFI is a village bank and 0 otherwise. Dum Sol is a dummy that is 1 if
the MFI provides loans on the basis of solidarity groups and 0 otherwise. Dum regulated is a dummy variable that is 1 if the
MFI is regulated by banking authorities and 0 otherwise. All other variables are defined as before. Robust standard errors are
provided in parentheses. *, ** and *** denote statistical significance at the 10%, 5% and 1% significance level, respectively.
The estimation method is pooled Random Effects (RE).
Dep Var. PaR30 (1) (2) (3) (4) (5) (6)
women clients -0.05 0.01 -0.05 -0.11 -0.04 -0.08
(reference category) (0.038)* (0.055) (0.038)* (0.059)** (0.038)* (0.075)*
experience
(women * experience) 0.00 0.00
(0.001) (0.001)
legal status
(women * Dum NGO) -0.10 -0.11
(0.064)* (0.067)*
Activities
(women * Dum fin.only) 0.01 0.01
(0.022) (0.023)
Lending methodology
(women * DumVill) 0.09 0.11
(0.078) (0.080)*
(women * Dum Sol) 0.06 0.09
(0.086) (0.088)
Regulation
(women * Dum
Regulated) -0.04 -0.05
(0.031) (0.032)*
26
Other controls added added added added added added
N 830 830 830 830 830 830
R² 0.1 0.13 0.11 0.14 0.12 0.23
Wald χ² 100.15*** 102.82*** 100.38*** 104.26*** 101.68*** 113.83***
6. Conclusions
In this paper we use a large global dataset covering 350 MFIs in 70 countries to test
whether there is a gender effect on microfinance repayment. This is important given the
undocumented popular believe that women honour their microfinance loans more than men.
Repayment is studied from a variety of measures such as portfolio-at-risk, loan-loss write-offs
and provisions, and gender is studied through the proportion of female clients as well as a
dummy variable that indicates whether the MFI has a conscious gender bias towards women
or not.
The findings indicate that MFIs with higher proportions of female borrowers have a
lower portfolio-at-risk. A dummy indicating whether the MFI consciously practice a woman
gender bias yields similar results. Using loan-loss write-offs and loan-loss provisions as
alternative dependent variables yield similar results. These combined findings provide
compelling evidence that that focus on women clients enhances microfinance repayment, and
that women in general are a better credit risk.
Interaction terms reveal that the positive repayment effect is stronger for NGOs, MFIs
that practice individual lending, MFIs that provide additional services alongside financial
services and regulated MFIs. These findings suggest two theoretical predictions that
significantly affect the women-repayment relation. First, MFIs who offer a more personalised,
individual-based service to their client (such as NGOs, finance plus providers and individual-
based lenders) benefit more from focussing on women. Therefore developing individual
procedures tailored to women’s needs might significantly increase repayment rates. This
finding is in line with the research by Rahman (2001) and Goetz and Gupta (1996) who argue
that more intensive contact and individual monitoring seriously improves repayment-rates.
Secondly, MFIs who face greater pressure might apply more coercive enforcement methods to
which women are more responsive. This prediction could explain why regulated MFIs benefit
more from focussing on women which is in line with the research by Rahman (2001) and Huq
(2004) who suggest that women are more responsive to coercive enforcement. This might also
27
be an alternative explanation to why individual lenders experience increased repayment
benefits from focusing on women clients.
All together the paper proves what policy makers and practitioners have long argued, that
women are better payers of microfinance loans than men. Is this good news or bad news? First
of all, it is interesting to observe that despite a lower objective credit-worthiness, women
prove to be good borrowers and good payers. But the issue of repayment should not obscure a
much more fundamental question: the well-being of female clients. Does high repayment rate
mean higher women welfare? As suggested by Susan Cheston, vice-president of Opportunity
International (one leading organisation in the field of microfinance): “women are good for
microfinance but is microfinance good for women?” (Cheston 2007: 15). Microfinance might
lead to women empowerment but also to feminization of debt (Mayoux 2002). Besides, good
repayment performance and loyalty do not necessarily mean clients’ satisfaction: it might be
the result of a debt trap (Cull et al.,. 2007). As suggested here, it might also be the result of
enforcement practices that women fear more than men. The interesting question to ask is
whether women repay better because they are more successful in their enterprises or simply
because they are more sensitive to MFI enforcement practices and social pressure. Besides,
the finding that women repay their loans better than men does not mean that women are better
customers than men. Maybe servicing women is more costly than servicing men? Or maybe
women take smaller loans and thereby reduce MFI scale economies?
28
References
Ahlin, C. and Townsend, R. (2007) Using repayment data to test across models of joint
liability lending, Economic Journal, vol. 117, pp. F11–F51.
Agarwal B. (1994), A field on one’s own. Gender and land rights in south Asia, Cambridge:
Cambridge University Press.
Aghion, B. A. D. & Morduch, J. (2003) Microfinance beyond group lending. Economics of
Transition, 8, 401-420.
Ameen, F., (2004) Loan Repayment in the Grameen Bank: The importance of borrower
opportunity cost and gender. In Research in Banking and Finance 5: Bank and Financial
Market Efficiency: Global Perspectives, 109-136.
Apmas, (2006) Voice of the People on Lending Practices of Microfinance Institutions in
Krishna District of Andhra Pradesh, Hyderabad, Andhra Pradesh Mahila Abhivruddhi
Society.
Armendariz, B. & J. MORDUCH (2005) The Economics of Microfinance, Cambridge: MIT.
Baltagi, B.H., SONG, S.H. and KOH, W. (2003) Testing panel data regression models with
spatial error correlation. Journal of Econometrics 117(1), 123-150.
Bhatt, N. & TANG, S. Y. (2002) Determinants of Repayment in Microcredit: Evidence from
Programs in the United States. International Journal of Urban and Regional Reseach 26,
360.
Brehanu A, FUFA, B. (2008) Repayment rate of loans from semi-formal financial institutions
among small-scale farmers in Ethiopia: Two-limit Tobit analysis, The Journal of Socio-
Economics, 37: 2221–2230
Cassar, A., CROWLEY, L. and WYDICK, B. (2007) The effect of social capital on group
loan repayment: evidence from field experiments, Economic Journal, vol. 117, pp. F85–
F106.
Chant, S. (1985) Single Parent Families: Choice or Constraint ? The Formation of Female-
Headed Households in Mexican Shanty Towns, Development and Change, October, 16
(4), pp.635-656.
Cheston, S. (2007) Just the facts, Ma’am: gender stories from unexpected sources with morals
for microfinance, Dialogue, 37, 13-27 2007
Cull R., DEMIGÜC-KUNT A., MORDUCH, J. (2008) Microfinance meets the market, Policy
Research Working Paper No 4630, Washington: World Bank.
_____. (2007) Financial performance and outreach: A global analysis of leading microbanks.
Economic Journal 17 (517), 107–133.
Daley-harris, S. (2007) State of the microcredit summit campaign. Report 2007, Washington,
DC: Microcredit summit campaign, 66 p.
http://microcreditsummit.org/pubs/reports/socr/EngSOCR2007.pdf
Dichter, T. (1997) Appeasing the Gods of sustainability: the future of international NGOs in
microfinance’. IN HULME, D. & MICHAEL, E. (Eds.) NGOs, states and donors: Too
close for comfort? International political economy series. New York, St. Martin’s Press in
association with Save the Children.
Edgcomb, E. & L. BARTON, (1998) Social intermediation and microfinance programs: A
literature review, USAID, Washington: Microenterprise Best Practices.
Emran, M. S., MORSHED A. M., STIGLITZ J. (2006) Microfinance and Missing Markets,
Working paper, Columbia University, Department of Economics, New York.
Fernando, J. (2006) Microfinance. Perils and Prospects, Routledge, Londres.
29
Fletschner, D. (2009) Rural women’s access to credit: market imperfections and
intrahouseholds dynamics, World Development, 37 (3): 618-631.
Ghatak, M. & GUINNANE, T. W. (1999) The economics of lending with joint liability:
theory and practice. Journal of development economics 60, 195-228.
Ghate, P. (2007) Indian microfinance. The challenges of rapid growth, New-Delhi: Sage.
Gibbons, D., KASIM, S., (1991) Banking on the Rural Poor. Center for Policy Research,
University Sains, Malaysia.
Giné, X., & KARLAN, D. (2006) Group vs. Individual Liability: A Field Experiment in the
Philippines, Policy Research Working Paper 4008, World Bank, Washington, DC.
Godquin, M. (2004) Microfinance Repayment Performance in Bangladesh: How to Improve
the Allocation of Loans by MFIs, World Development 32 (11): 1909–1926, 2004
Goetz, A., and GUPTA, R. S. (1996) Who takes the credit? Gender, power and control over
loan use in rural credit programs in Bangladesh. World Development, 24(1): 45–63.
Haddad L. HODDINOD, J. (1995) Women’s Income and Boy Girl Anthropometric Status in
Cote d’Ivoire, World Development, vol. 22, n°4, pp. 543-553.
Hardy, D. C. D. I. O., HOLDEN, P. & PROKOPENKO, V. (2003) Microfinance institutions
and public policy. Policy Reform 6, 147.
Hashemi, S., Schuler, S., and Riley, A. (1996) Rural credit programs and women’s
empowerment in Bangladesh. World Development, 24(4), 635–654.
Harper, M. (2007) What’s wrong with groups? in Dichter & Harper (eds) What’s wrong with
microfinance, Practical Action Publishing, 35-49.
Hossain, M., (1988) Credit for Alleviation of Rural Poverty: The Grameen Bank of
Bangladesh. International Food Policy Research: Institute Research Report 65,
Washington DC.
Hulme, D., (1991) The Malawi Mudzi Fund: Daughter of Grameen, Journal of International
Development 3(3), 427-431.
Kabeer, N. (1995) Reversed Realities. Gender Hierarchies in Development Thought, London /
New-York: Verso 1st Edition 1994, 346 p.
Kabeer, N. (1997) Woman, Wages and Intra-household Power relations in Urban Bangladesh,
Development and Change, vol. 28, pp. 261-302.
Kabeer, N. (2001) Conflicts Over Credit: Re-evaluating the Empowerment Potential of Loans
to Women in Rural Bangladesh, World Development 29(1): 63–84.
Kevane, M. & WYDICK, B. (2001) Microenterprise Lending to Female Entrepreneurs:
Sacrificing Economic Growth for Poverty Alleviation? World Development, 29, 1225-
1236.
Khandker, S.R., KHALILY, B., KAHN, Z., (1995) Grameen Bank: Performance and
sustainability. World Bank Discussion Paper 306, Washington DC.
Karlan, D. (2007) Social connections and group banking, Economic Journal, vol. 117, pp.
F52–F84.
Lensink, R. & MERSLAND, R. (2009) Microfinance Plus. Kristiansand, Norway, University
of Agder.
Mayoux, L. (1999) Questioning virtuous spirals: microfinance and women’s empowerment in
Africa. Journal of International Development, 11(7): 957–984.
Mayoux, L. (2001) Tackling the Down Side: Social Capital, Women’s Empowerment and
Micro-Finance in Cameroon, Development and Change 32(3): 421–50.
Mersland, R. (2009) The cost of ownership in microfinance organizations. World
Development, 37, 469-478.
30
Mersland, R. & STRØM, R. Ø. (2008) Performance and trade-offs in microfinance
organizations - does ownership matter? Journal of International Development, 20, 598-
612.
Mersland, R. & STRØM, R. Ø. (2009a) The empirics of microfinance lending methodologies.
Kristiansand, Norway, University of Agder.
Mersland, R. & STRØM, R. Ø. (2009b) Performance and governance in microfinance
institutions. Journal of Banking and Finance, 33, 662-669.
Montgomery, R., (1996) Disciplining or protecting the poor: avoiding the social costs of peer
pressure in microcredit schemes Jounal of International Development, 8(2): 289-305.
Morduch, J. (1999) The microfinance promise, Journal of Economic Literature, 37 (4): 1569–
614.
Molyneux, M. (2002) Gender and the silences of social capital, Development and Change, 33
(2): 167-188.
Norwegian Nobel Committee (2006). The Nobel Peace Prize 2006, Press release, Oslo,
October. weblink: http://nobelprize.org/nobel_prizes/peace/laureates/2006/press.html.
Paxton J., GRAHAM, D. and THRAEN, C. (2000) Modeling group loan repayment
behaviour: new insights from Burkina Faso, Economic Development and Cultural
Change, 48(3): 639–55.
Phillips, Ch. BHATIA-PANTHAKI S. (2007) Enterprise development in Zambia: reflections
on the missing middle, Journal of International Development, 19(6):793-802.
Plümper, T., TROEGER, V.T. (2007) Efficient estimation of time-invariant and rarely
changing variables in finite sample panel analyses with unit fixed effects, Political
Analyses, 15(2): 124-139.
Huq H. (2004) Surviving in the world of microdebt: a case from rural Bangladesh,
BOUMAN & HOSPES (eds) Financial Landscape Reconstructed: The Fine Art of
Mapping Development. Boulder, CO: Westview Press, pp. 43-54.
Rahman, A. (2004) Microcredit and poverty reduction : trade-off between building
institutions and reaching the poor, in BOUMAN & HOSPES (eds) Financial Landscape
Reconstructed: The Fine Art of Mapping Development. Boulder, CO: Westview Press, pp.
27-43.
Rahman, A., 2001. Women and Microcredit in Rural Bangladesh: An Anthropological Study
of Grameen Bank Lending. Boulder, CO: Westview Press.
Rahman, A. (1999). Micro-credit initiatives for equitable and sustainable development: who
pays? World Development, 27(1): 67–82.
Rankin K. N. (2002), Social capital, microfinance and the politics of development, Feminist
Economics, 8 (1):1-24.
Rao, S. (2008) Reforms with a female face: gender, liberalization, and economic policy in
Andhra Pradesh, India, World Development, 36 (7): 1213-1232.
Sen, A. (1990) Gender and Cooperative Conflicts, in TINKER (ed), Persistant Inequalities.
Woman and World Developement, New-York/Oxford : Oxford University Press, pp. 123-
149.
Senauer, B. (1990) The Impact of the Value of Women’s Time on Food and Nutrition, in
TINKER (ed), Persistant Inequalities. Woman and World Developement, New-
York/Oxford: Oxford University Press, pp. 150-161.
Stock, J.H., & WATSON, M.W., 2007. Introduction to econometrics. Pearson Education,
Boston.
Thomas, D. (1990) Intra-household resource allocation: an inferential approach, Journal of
31
Human Resources, 25 (4), pp. 635-664.
Todd, H., (1996) Women at the Center: Grameen Bank Borrowers after One Decade. Dhaka
University Press Ltd.
Sharma M., & ZELLER, M. (1997) Repayment performance in group-based credit programs
in Bangladesh: An empirical analysis. World Development, 25(10): 1731–1742.
World Bank (2001) World Development Report: Engendering Development: Through Gender
Equality in Rights, Resources and Voice. New York: World Bank and Oxford University
Press.
World Bank (2007 Finance for All? Policies and Pitfalls in Expanding Access. A World Bank
Policy Research Report, The World Bank: Washington.
Wydick, B. (1999) Can social cohesion be harnessed to repair market failures? Evidence from
group-based lending in Guatemala, Economic Journal, 109(457): 463–75.
32
