Sie sind auf Seite 1von 290

The Over-Indebtedness of Microfinance Customers

An Analysis from the Customer Protection Perspective

DISSERTATION

Presented in Partial Fulfillment of the Requirements for


The PhD in Economics and Management

By
Jessica Schicks, MPhil, Dipl.Oec

At the
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management
Universit libre de Bruxelles

Members of the Jury:


Prof. James Copestake, Centre for Development Studies, University of Bath
Prof. Isabelle Gurin, IRD, Universit Paris I Sorbonne and French Institute of Pondicherry
Prof. Marek Hudon, Supervisor, CERMi, Universit libre de Bruxelles
Prof. Marc Labie, CERMi, Universit de Mons, Universit libre de Bruxelles
Prof. Ariane Szafarz, Secretary of the Jury, CERMi, Universit libre de Bruxelles

Academic Year 2012/2013


UNIVERSITE LIBRE DE BRUXELLES

On a Personal Note
In the beginning of a PhD one knows that one will only be able to cover a small part of a
topic. That the research will lack data. That no answers will come without new questions. The
further the research progresses, the more apparent become its limitations. The things it
cannot cover. The shortcomings of the empirical study. The ambivalence of its findings. A
PhD is a powerful reminder of the limitations of our knowledge and methodologies. Of the
lack of an absolute truth.
Nevertheless, these three years have been an exciting personal experience. I have learned so
much. I have been able to follow my passion for this topic with almost unlimited freedom. My
work has been received with unanticipated amounts of recognition, both in academia and
among the practitioners of the microfinance industry. It has been given a lot of attention in
the Ghanaian microfinance sector and in our local results workshops with MFIs, investors,
the Ghanaian Ministry of Finance, and the Ghana Central Bank. It sometimes seems like I
have been able to give something small back to all those borrowers who talked to us in the
course of this research, revealing their personal hardships in the hope of helping
microfinance practitioners design better products in the future.
I am immensely grateful for this opportunity. For this experience. For the trust that people
put in me at the beginning of this project. My thoughts go to all of you who have helped me to
make this project come true and who have been close to me during the last three years, no
matter if you were in Belgium, Ghana, Germany, France, the UK, or the US. You know who
you are.
Thank you!

-i-

Table of Contents
INTRODUCTION

CHAPTER 1

The Definition and Causes of Microfinance Over-Indebtedness: A


Customer Protection Point of View
1. Introduction

10

2. Defining and measuring microfinance over-indebtedness: When is too


much too much?

12

3. The reasons for over-indebtedness: why do people overborrow?

19

4. Conclusion

32

CHAPTER 2

41

From a supply gap to a demand gap? The risk and consequences of overindebting the underbanked
1. Introduction

42

2. The consequences of over-indebtedness

45

3. Empirical research on over-indebtedness in microfinance to date

56

4. Conclusion

63

CHAPTER 3

71

The sacrifices of microborrowers in Ghana: A customer-protection


perspective on measuring over-indebtedness
1. Introduction

72

2. Predicting over-indebtedness

75

3. The data: An in-depth survey in urban Ghana

79

4. The sacrifices of microborrowers

83

5. Estimation and empirical results

88

6. Conclusion

94

- ii -

CHAPTER 4

114

Over-indebtedness in Microfinance: An Empirical Analysis of Related


Factors on the Borrower Level
1. Introduction

115

2. Factors related to over-indebtedness

117

3. Data and econometric framework

127

4. Results and discussion

134

5. Conclusion and Implications

141

AN OVERALL CONCLUSION TO THIS THESIS

162

1. Key findings of the dissertation

164

2. Implications for other fields of research

172

3. Concluding remarks: Limitations and further research

187

FULL BIBLIOGRAPHY

197

APPENDIX:

214

Schicks, J. and R. Rosenberg (2011) 'Too Much Microcredit? A Survey of the


Evidence on Over-Indebtedness'. CGAP Occasional Paper 19. Washington
DC.
Schicks, J. (2011) 'The over-indebtedness of microborrowers in Ghana - An
empirical study from a customer protection perspective'. Center for
Financial Inclusion Publication No. 15. Washington DC.

- iii -

Introduction
Beginning in the 1970s with a scalable microcredit model based on social collateral
(Armendriz and Morduch, 2010), microfinance has developed from a credit focused nonprofit activity into a semi-commercial industry offering a range of financial services to lowincome populations. However, in spite of the broadened product portfolio offered by
microfinance,

the focus

of the industry remains

lending small-scale loans

to

microentrepreneurs. In spite of its commercialisation, the mission of the microfinance


industry remains to benefit a target group whose poverty usually excludes it from formal
financial services. A large share of microfinance institutions (MFIs) remain subsidy
dependent (Hudon and Traca, 2011). These institutions require a social mission of positively
impacting the lives of the poor to justify their existence. Even fully profit-oriented MFIs use
the label microfinance to position themselves as investments with a positive social impact
and are usually managed as double bottom-line organisations (Labie, 2004; Copestake,
2007).
Given that credit remains the flagship product of microfinance and that the industrys
existence is based on the social mission of benefitting people in poverty, surprisingly little is
known about the impact that microlending has on the lives of the poor. Increasingly rigorous
impact studies challenge the original narrative of poverty alleviation through microloans
(Duvendack et al., 2011; Karlan and Goldberg, 2011). They show that the benefits of
microcredit may be much more limited than originally assumed. Although there is ample
indication that microfinance can be highly beneficial for the poor, there is no consistent and
robust proof to date that microfinance effectively promotes microenterprise development, that
it increases the asset ownership of microborrowers or that it positively affects borrowers
income. Similarly, recent theoretical developments and empirical research highlight that the
contribution of microfinance may lie not in facilitating productive activities that earn high
returns on investment but rather in helping the poor smoothen their consumption in the face of
volatile incomes and frequent emergencies (Collins et al., 2009; Karlan and Zinman, 2009).
Again, further research has yet to prove the benefits of microloans in terms of consumption
smoothing. In the meantime, the concept of microfinance having an impact through
consumption smoothing rather than through microenterprise development and poverty
alleviation defeats the idea that high returns on productive investments made with microloans
allow borrowers to repay their microloans without difficulty.

-1-

Even if the above studies on the social impact of microfinance leave room for future research,
their findings on what the positive impact of microfinance may be represent crucial insights
for both microfinance scholars and for policy makers. For example, less belief in the benefits
of microfinance may reduce tolerance for risks created by means of microfinance. It may
reduce the tolerance of the microfinance industry for borrowers struggling severely to repay
their loan. However, microfinance research often does not take the potential negative effects
of microfinance and the struggles of microborrowers into account. There remains an
important need for more research in microfinance on the potential downsides of microloans
and on the difficulties that microborrowers face in repaying their debt.
Over the past 40 years of the microfinance industrys history of growth and
commercialisation, the downsides of credit for the poor have been largely ignored by
academic research. Scholars have identified a risk of mission drift for microfinance
institutions that attempt to combine financial and social objectives. When trade-offs exist
between achieving social results and financial results, MFIs may reduce their efforts to pursue
their social mission and focus on more easily measured financial results instead (Copestake,
2007; Mersland and Strm, 2010; Armendriz and Szafarz, 2011). Some sociologists and
anthropologists have noted the damage that social pressure in group lending could do to the
poor (Besley and Coate, 1995; Montgomery, 1996; Rahman, 1999). More recently, scholars
began highlighting the substantial harm that a debt over-load and repayment problems could
imply for the poor (Brett, 2006; Dichter, 2007; Hulme, 2007; Gurin et al., 2009). Hudon
(2009) highlighted that the idea of a human right to credit follows the tradition of blindness
to credit risks and ignores the danger of over-indebtedness that credit implies. Nevertheless,
the over-indebtedness risks of microfinance remain one of the most urgent and underresearched topics of microfinance.
Today, over-indebtedness is one of the most important challenges faced by the microfinance
industry. It is the most important risk according to Microfinance Banana Skins 2012, a
global report on risks in the microfinance industry (Lascelles and Mendelson, 2012). Several
microfinance markets have reached extreme levels of over-indebtedness that have
materialised in the form of market-wide repayment crises and have put the local microfinance
industry at risk. For example, in the Indian state of Andhra Pradesh, microfinance has almost
come to a halt as the result of an over-indebtedness crisis that has not only threatened the
sustainability of lenders because of widespread default but also has had strong political
repercussions. This crisis triggered a regulatory reaction that fundamentally challenges the
business model of MFIs and has put the future of the local microfinance sector in question.
-2-

Other microfinance markets that in recent years faced substantial crises that included
problems with over-indebtedness are, for example, Bosnia, Nicaragua, Morocco and Pakistan.
Furthermore, the industry does not know to what extent over-indebtedness is prevalent in
microfinance markets around the world that have not yet reached a stage of crisis and thus
whether more markets risk going through similar experiences.
The industry needs to perform research to agree on a definition of over-indebtedness and to
understand the extent to which over-indebtedness is prevalent in microfinance markets
(Kappel et al., 2010). It seeks to understand and be able to mitigate the harm that overindebtedness might cause. To develop solutions to the problem of over-indebtedness, the
microfinance industry must analyse what factors are related to over-indebtedness and may be
potential causes of the phenomenon. It needs to identify the borrowers who are most at risk.
The recent rebalancing in the microfinance industry that has led to a focus on protecting
clients and working towards the social mission of microfinance requires a parallel shift to
research through the lens of customers (Rozas et al. 2011; Copestake, 2007). Contrary to the
wide range of microfinance research at the institutional level, there is now an urgent need for
borrower-level research that considers the perspective of microborrowers and gives
borrowers, as the main stakeholders of microfinance, a voice in the industrys future
development.
It is this practical and scholarly need for answers to the microfinance industrys questions of
over-indebtedness that has set the research agenda for my PhD. The first paper addresses the
conceptual challenges of defining and measuring over-indebtedness. Based on a thorough
analysis of the interdisciplinary literature on the over-indebtedness of consumers in developed
countries and on the existing research within microfinance, this paper develops an innovative
definition of over-indebtedness. It defines over-indebtedness according to the sacrifices
microborrowers experience related to their loans. Moreover, in addition to proposing the first
definition of over-indebtedness in the microfinance literature that is appropriate for customer
protection purposes, the paper suggests a practicable way to use this definition in empirical
research.
In a second step, the paper analyses the potential causes of over-indebtedness. It unites the
contributions of economics, sociology and psychology into a comprehensive framework of
the causes of over-indebtedness that, to the authors knowledge, does not exist in the literature
to date. The framework identifies the role of lenders in pushing borrowers into overindebtedness due to an exaggerated marketing and growth focus, due to inappropriate lending
-3-

products or due to sub-optimal lending procedures. It equally highlights the role of borrowers,
working with insights from behavioural economics and psychology on the impact of cognitive
biases on borrowing decisions and taking the sociological influences of, for example, societal
pressures into account. The paper identifies economic and socio-demographic borrower
characteristics related to consumer over-indebtedness in developed countries, which are likely
to at least partially represent causal effects. Moreover, the framework of the causes of overindebtedness identifies external influences on over-indebtedness, which include both those in
the institutional and macroeconomic environment that may prevent or facilitate overindebtedness ex-ante and those in the form of adverse economic shocks that affect borrowers
and turn a healthy debt situation into over-indebtedness ex-post.
The second paper of this PhD develops the customer protection perspective on overindebtedness further by analysing in detail what consequences over-indebtedness may have on
borrowers and on lending institutions. The paper shows that from the viewpoint of customer
protection, the consequences of over-indebtedness reach far beyond the risk management
concerns that primarily affect MFIs and investors. It details the material effects of overindebtedness on borrowers, the potential social cost of being over-indebted and the substantial
effects that over-indebtedness can have on psychological well-being. With regard to
microfinance institutions, this paper indicates that the potential consequences of overindebtedness on lenders are much more complex than the common focus on portfolio
problems in the form of delinquency and default suggest. It points out that there are also
significant indirect consequences of over-indebtedness on earnings, operational cost, staff
retention, and customer satisfaction. The analysis extends to other stakeholders in identifying
briefly the second-order effects that over-indebtedness can have on third parties that are not
directly part of the lending contract between an MFI and its borrowers. It recognises the
effects that microfinance over-indebtedness can have on borrowers who are not overindebted, on MFI customers who are not borrowing, on other MFIs and the financial system,
as well as on the microfinance industrys investors, donors and support organisations.
The second paper then moves on to comprehensively review the empirical research on
microfinance over-indebtedness to date. It identifies early research on over-indebtedness in
microfinance from the first microfinance crisis in Bolivia but notes that microfinance scholars
did not grasp the broader structural lessons of this crisis at the time. Instead, the interest in
over-indebtedness as a microfinance topic quickly faded away again, followed only by
individual pieces of research that were not specific to over-indebtedness but recognised the
potential downsides of the group-lending methodology. The paper then reviews a number of
-4-

papers related to microfinance markets with debt problems in the aftermath of the global
financial crisis but also points out that the phenomenon of over-indebted microborrowers does
not appear to be specific to crisis markets. Looking through a lens of competitive equilibrium
and identifying the challenges of a market characterised by a simultaneous demand gap and
supply gap, the paper analyses the existing empirical studies with regard to our knowledge on
the extent of over-indebtedness in microfinance markets. It shows that there are alternative
scenarios for how over-indebtedness in microfinance will develop in the future and notes a
need for further research, where the findings from the existing literature are not yet
conclusive.
The third paper works with a unique empirical data set from Ghana collected by means of an
in-depth survey among 531 microborrowers in Accra, Ghana, at the end of 2010. Respondents
represent a random sample from five of Ghanas most important microfinance institutions:
ProCredit Ghana, Opportunity International Ghana, Sinapi Aba Trust, EB-ACCION and
Advans Ghana. They were interviewed anonymously in a research project that was enabled by
the cooperation of the microfinance industrys global customer protection organisation, the
Smart Campaign, and the German development bank KfW. The paper measures the overindebtedness of microborrowers from a customer protection perspective, putting into practice
the definition and measurement developed in Chapter 1. It identifies the share of overindebtedness in the local population of microborrowers and analyses what sacrifices
borrowers experience related to their loans, how frequently they repeat these sacrifices and
how acceptable the various sacrifices are from the subjective perspective of microborrowers.
The paper proceeds to compare the common risk management indicators of over-indebtedness
to the customer-protection definition based on borrower sacrifices. Using econometric
regression techniques, it develops a model to predict over-indebtedness from the customer
protection point of view by means of more easily available indicators. At the same time, it
analyses to what extent the risk management perspective and the customer protection
perspective are in line or whether a risk management approach may be inadequate for
customer protection purposes.
Finally, the fourth and last paper of this dissertation identifies borrower-level factors related
to over-indebtedness. It tests factors identified in the first chapter of this dissertation for their
relationship to over-indebtedness among microfinance borrowers in Accra. The paper uses the
primary database from Ghana to run a logistic regression model that tests the relationship of
poverty, adverse economic shocks, returns on investment, and financial literacy to over-5-

indebtedness. It addresses other economic, socio-demographic, and loan- or lender-specific


factors by means of control variables. Whereas a cross-sectional dataset cannot establish
causal relationships, the paper argues that the factors it identifies as related to overindebtedness are most likely causes of over-indebtedness. It further breaks down their
relationship to the individual sacrifices that borrowers experience and, using ordered logit
regressions, to the repetition of sacrifices over time and to the acceptability of sacrifices from
the borrower perspective. It deducts policy recommendations for the development of solutions
to the over-indebtedness challenge in microfinance.
With these four papers, this dissertation makes an important contribution to over-indebtedness
research in microfinance. It suggests a definition of over-indebtedness that meets the
requirements of a customer protection perspective. It develops a comprehensive framework of
the causes of over-indebtedness. Although this dissertation does not challenge that
microfinance may have a substantial positive impact on the lives of the poor and that
microcredit is often urgently needed and highly valued by the poor, it shows that the
microfinance industry should be cautious about the potential dangers of microcredit. It
analyses over-indebtedness risks with regard to their magnitude, identifying the potential
consequences of over-indebtedness in detail, and with regard to their likelihood, assessing our
knowledge of the extent of over-indebtedness in microfinance markets around the world.
Based on a unique empirical dataset with borrower-level survey data, the PhD puts the
customer-protection oriented over-indebtedness definition into practice and measures the
over-indebtedness of borrowers in an African microfinance market, revealing what sacrifices
borrowers go through and what their subjective experience of these sacrifices are. It contrasts
this customer protection perspective to the risk management perspective that is common in
over-indebtedness research to date. Finally, it provides empirical insights into what factors are
related to over-indebtedness on the borrower level, factors that are likely to be causes of overindebtedness. It thus contributes to the development of solutions meeting the challenge of
over-indebtedness in microfinance.

-6-

References
Armendriz, B. and J. Morduch (2010) 'The economics of microfinance (2nd ed.).
Cambridge, Mass: MIT Press.
Armendriz, B. and A. Szafarz (2011) 'On Mission Drift In Microfinance Institutions' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 341366). London,
Singapore: World Scientific Publishing.
Besley, T. and S. Coate (1995) 'Group lending, repayment incentives and social collateral'.
Journal of Development Economics, 46(1), 118.
Brett, J. A. (2006) '"We Sacrifice and Eat Less": The Structural Complexities of Microfinance
Participation'. Human Organization, 65(1), 8.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How
the World's Poor Live on $2 a Day'. Princeton, NJ: Princeton Univ. Press.
Copestake, J. (2007) 'Mainstreaming microfinance: social performance management or
mission drift?'. World Development, 35(10), 17211738.
Dichter, T. (2007) 'Can microcredit make an already slippery slope more slippier?: Some
lessons from the social meaning of debt' In T. Dichter & M. Harper (Eds.), What's wrong
with Micofinance? (pp. 918). Warwickshire: Intermediate Technology Publications.
Duvendack, M., R. Palmer-Jones, J. Copestake, L. Hooper, Y. Loke and N. Rao (2011) 'What
is the evidence of the impact of microfinance on the well-being of poor people?'. EPPICentre, Social Science Research Unit, Institute of Education, University of London,
London.
Gurin, I., M. Roesch, O. Hlis and Venkatasubramanian (2009) 'Microfinance, Endettement
et Surendettement: Une tude de cas en Inde du Sud [Microfinance, Indebtedness and
Over-indebtedness: A case study in Southern India]'. Revue Tiers Monde, 197, 131146.
Hudon, M. (2009) 'Should Access to Credit be a Right?'. Journal of Business Ethics, 84, 17
28.
Hudon, M. and D. Traca (2011) 'On the Efficiency Effects of Subsidies in Microfinance: An
Empirical Inquiry'. World Development, 39(6), 966973.
Hulme, D. (2007) 'Is microdebt good for poor people? A note on the dark side of
microfinance' In T. Dichter & M. Harper (Eds.), What's wrong with Micofinance? (pp. 19
22). Warwickshire: Intermediate Technology Publications.

-7-

Kappel, V., A. Krauss and L. Lontzek (2010) 'Over-Indebtedness and Microfinance Constructing an Early Warning Index'. Center for Microfinance, University of Zurich,
responsAbility, Council of Microfinance Equity Funds, Triodos Investment Management
Zurich.
Karlan, D. and N. Goldberg (2011) 'Microfinance Evaluation Strategies: Notes on
Methodology and Findings' In B. Armendriz & M. Labie (Eds.), The Handbook of
Microfinance (pp. 1758). London, Singapore: World Scientific Publishing.
Karlan, D. and J. Zinman (2009) 'Expanding Microenterprise Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts in Manila'. Financial Access
Initiative Cambridge, MA.
Labie, M. (2004) 'Introduction'. Mondes en Dveloppement, 126(2), 78.
Lascelles, D. and Mendelson, S. (2012) #Microfinance Banana Skins 2012, The CSFI survey
of microfinance risk: Staying Relevant. Centre for the Study of Financial Innovation
(CSFI), Kent, UK.
Mersland, R. and R.. Strm (2010) 'Microfinance Mission Drift?'. World Development,
38(1), 2836.
Montgomery, R. (1996) 'Disciplining or protecting the poor? Avoiding the social costs of peer
pressure in micro-credit schemes'. Journal of International Development, 8(2), 289305.
Rahman, A. (1999) 'Micro-credit Initiatives for Equitable and Sustainable Development: Who
Pays?'. World Development, 27(1), 6782.
Rozas, D., I. Barrs, C. Connors and E. Rhyne (2011) 'Implementing Client Protection in
Microfinance: The State of the Practice, 2011'. Center for Financial Inclusion Center for
Financial Inclusion Publication No. 14. Washington DC.

-8-

Chapter 1
The Definition and Causes of
Microfinance Over-Indebtedness:
A Customer Protection Point of View
Jessica Schicks*
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management
Centre Emile Bernheim, Universit Libre de Bruxelles

With over-indebtedness emerging among microfinance customers, the industrys sustainability and
social impact are both at risk. Filling a void in the literature, this paper develops a definition of
over-indebtedness that is appropriate for customer-protection purposes. It provides a framework for
the causes of over-indebtedness that highlights the role of external influences and the responsibility
of lenders. It recognises the role borrowers play in their own over-indebtedness. This paper
challenges several misconceptions and oversimplifications about microfinance over-indebtedness.
These misconceptions include the belief that default-based risk management indicators can present
the customer protection perspective, the dangers of consumption loans, and the benefits of
competition, regular instalment schedules, the zero-tolerance policy, and Annual Percentage Rates
(APRs). By enhancing our understanding of microfinance over-indebtedness and its causes, this
paper provides the means for measuring over-indebtedness and tailoring solutions to its root causes.
The analysis shows that combating over-indebtedness does not automatically mean reducing access
to microcredit. While a sound evaluation of repayment capacity is essential, adapting products to
clients needs also reduces over-indebtedness.

Paper forthcoming in Oxford Development Studies

*The author is grateful to Marie-Christine-Adam Foundation for its financial support. She thanks
Prof. Marek Hudon, Richard Rosenberg and Christoph Kneiding, William Steel, Meritxell Martinez
and Katja Kirchstein as well as two anonymous referees for their valuable comments and
suggestions. Further thanks go to seminar and conference participants at the University of Oxfords
CSAE conference, the 2nd European Research Conference in Microfinance, and at ULB.
-9-

1. Introduction
1

Over the last decade, the microfinance industry has achieved immense growth and public
recognition. Its aim to reduce global poverty has been complemented by sizeable profits for some
microfinance institutions (MFIs) and their investors. However, the fairy tale is beginning to show
cracks. The impact on poverty alleviation may be an illusion, and the social microfinance industry
faces increasing criticism for exploiting and over-indebting poor customers in the service of profits.
Although the social mission of microfinance places particular importance on protecting customers,
the 2008/2009 global banking crisis has shown that unsustainable financial services based on
irresponsible lending imply much broader risks. Over-indebtedness is currently one of the most
serious risks in microfinance, endangering both social impact and the industrys stability (Hudon,
2009). Over-indebtedness can push customers further into poverty, accompanied by the material,
psychological, and social consequences of debt (see Schicks (in press) for details). It puts MFIs, their
portfolio quality, and institutional stability at risk. The consequences can spill over to investors,
donors and the industry as a whole. While the individual experiences differ, several countries such as
Bosnia, Morocco, Nicaragua, Pakistan and India have already experienced crises with important
elements of over-indebtedness among microfinance clients.

From a customer protection perspective, this paper contributes to our understanding of microfinance
over-indebtedness and its causes. There is widespread ambiguity surrounding the definition and
indicators of over-indebtedness. Scholars end up implicitly discussing different things depending on
whether they focus on risk management or customer protection. A clear definition is required to
assess the extent of the problems in microfinance markets around the world and to provide empirical
insights into the phenomenon of over-indebtedness. However, the definitions and indicators that
researchers have relied on to date are grounded in a risk management framework. The current need
for customer protection creates different requirements. This paper develops an innovative definition
of over-indebtedness based on borrower sacrifices. It suggests a practical approach for using this
definition in empirical research.
Understanding the risks of over-indebtedness and developing solutions also requires an inclusive
understanding of the causes of over-indebtedness. To the author's knowledge, no studies
comprehensively analyse the drivers of debt problems. In the consumer finance literature on

1 This paper focuses on institutional microfinance following the solidarity banking, village banking or individual lending models.
Although microfinance has expanded to a broad range of financial services, this paper focuses on its original product: microloans.
2 Earlier crises, although different in many regards, took place in Bolivia and South Africa in the late 1990s and early 2000s.

- 10 -

developed countries , our knowledge of the causes of over-indebtedness is spread across different
strands of research. We need a comprehensive framework that unites the insights of economists,
sociologists and psychologists. In the microfinance literature, academic work on the causes of over4

indebtedness hardly exists. The literature on repayment problems focuses on repayment incentives

and ignores many of the findings that consumer finance research offers regarding the origins of debt
problems. Therefore, a framework that informs microfinance scholars and practitioners and enables
them to better address the current over-indebtedness challenges must gather our knowledge
regarding over-indebtedness among microfinance customers and

combine it with the

interdisciplinary findings from the consumer finance literature.


By providing an account of the most important factors that impact over-indebtedness, this paper
pinpoints the roles that external influences, lenders, and borrowers play in causing this problem. It
challenges several myths and misconceptions that, for example, treat lender competition as a silver
bullet for clients, condemn consumption loan use and insist on a zero-tolerance policy or lowfrequency instalment schedules. It raises doubts about Annual Percentage Rates (APRs) and shows
why MFI customers strong demand for credit does not prove that the loans are beneficial. This
paper reveals unexpected answers regarding solutions to over-indebtedness and shows that reducing
over-indebtedness does not automatically imply reducing access to finance.
The next section categorizes the various approaches to defining and measuring over-indebtedness in
the consumer finance and microfinance literatures. Based on this analysis, it develops a new
definition of over-indebtedness that is appropriate for microfinance research. In contrast to the
existing approaches, this definition incorporates the industrys current focus on customer protection.
The article goes on to analyse the causes of over-indebtedness and the responsibilities of lenders and
borrowers respectively. The final section concludes and provides directions for further research.

3 In contrast to the paradigm of microenterprise finance, this paper uses the consumer finance literature as a reference. Microfinance
primarily relies on character-based lending techniques, and a sizable amount of microcredit is used for consumption purposes.
Microborrowers make personal rather than professional credit decisions. Furthermore, being personally liable for their loans,
microborrowers face the same personal consequences of over-indebtedness that are discussed in the consumer credit literature.
4 See Godquin (2004) for a review of the empirical literature.

- 11 -

2. Defining and measuring microfinance over-indebtedness: When is


too much too much?
Despite the importance of the topic, no universal definition of personal over-indebtedness exists.
Definitions in the consumer credit and microfinance literatures vary for example depending on the
5

scientific lens, data availability, and the threshold for over-indebtedness. There have been several
attempts to analyse the different types of definitions and their advantages (e.g., Betti et al., 2007).
However, to the author's knowledge, no structured analysis of all options exists.
Figure 1. Dimensions of over-indebtedness definitions

Figure 1 provides a systematic overview of the dimensions that distinguish the definitions of overindebtedness used in the microfinance and consumer finance literatures. Definitions differ depending
on the purpose of the research, their academic orientation, their precision in distinguishing between
6

actual definitions versus measurements or indicators of over-indebtedness , and their units of


reference. There are aggregate measurements that include all customers of an institution or entire
country. Other scholars consider over-indebtedness an individual phenomenon. Disney et al. (2008)

5 There is an equivalent macroeconomic debate on sustainable debt on a country level, especially with regards to debt relief for
Heavily Indebted Poor countries (HIPC). For details see e.g., Benno (2008).
6 As they are often subject to discussion, Appendix 1 provides an overview of the different dimensions of quantitative overindebtedness measurements.

- 12 -

argue that all of a households cash flows should be considered. In the microfinance context, the
appropriate unit may even be a borrowers extended kin support network (Gurin et al., 2009b).
In terms of method, over-indebtedness definitions apply either single or multiple criteria. Applying
multiple criteria selectively risks extending a definition too far (Disney, Bridges and Gathergood,
2008), applying them simultaneously increases precision. Over-indebtedness can either be defined as
a quantitative threshold (e.g. a debt-to-income ratio) or in qualitative terms such as a perceived
burden. Although, in legal contexts, qualitative definitions produce uncertainty for lenders seeking to
comply, these definitions are more flexible and better able to account for the circumstances of
individual borrowers. Definitions are considered more reliable if they are based on objective
information. However, many researchers emphasise the merits of a subjective definition of overindebtedness (Lea et al., 1993; Betti et al., 2007; Gurin et al., 2009b). They consider individuals the
best judges of their own debt situation. Gurin et al. (2009a) point out that for borrowers, overindebtedness may be a question of perceptions and social consequences rather than a material
problem of disequilibrium among assets, revenues and debt. Debt conditions are heterogeneous in
meaning. Whether debt is a burden depends on the nature of the debt relationship rather than the
amount (Gurin et al., 2009b). As a result, for many poor borrowers, the total amount of debt is not
meaningful. Self-reports are considered less reliable than external data and provide better
information about a borrowers circumstances and future income prospects, as well as the meanings
of debt. For example, in their study on over-indebtedness and financial literacy in the US, Lusardi
and Tufano (2009) completely rely on self-reports.
The essential dimension of an over-indebtedness definition is severity. Most researchers only
consider borrowers over-indebted if their debt problems are structural and persistent over a certain
time horizon (Canner and Luckett, 1991). According to the Life Cycle Hypothesis (Modigliani and
Brumberg, 1980) and the Permanent Income Hypothesis (Friedman, 1957), consumers aim for stable
consumption over a long time horizon. Permanently high debt levels at a young age may be rational,
if income is expected to increase. The correct reference for measuring over-indebtedness would thus
be a permanent, life-long estimate of a borrowers income (Betti et al., 2007). Conversely, given the
short-term cash management challenges of microborrowers and their high level of uncertainty about
the future, microborrowing focuses on a rather short time horizon.
Researchers use various severity levels of repayment problems as the threshold of over-indebtedness.
Some only count legally bankrupt borrowers, whereas others include cases of default or arrears (e.g.
Disney, Bridges and Gathergood, 2008; Kappel et al., 2010) or even all borrowers who struggle with
an unhealthy debt balance (e.g. Collins, 2008). Bankruptcy is easy to measure on an aggregate level,
- 13 -

but represents a limited understanding of over-indebtedness that is difficult to compare among legal
systems. Defaults or permanent arrears are common criteria of over-indebtedness in developed
markets, but Gurin et al. (2009b) point out that they may not be suitable in developing countries,
where informal lending agreements (that often do not specify explicit payment deadlines) play a
large role. Conversely, a system of juggling debts, where borrowers continuously borrow anew to
repay old loans, may hide over-indebtedness before it appears in the form of arrears. Manifest
7

repayment irregularities are thus rather a lagging indicator of over-indebtedness. For unhealthy debt
balances, the main challenge is that a no threshold can account for the circumstances of all
borrowers. A debt-to-income ratio that is unsustainable for one borrower can be manageable for
another. Furthermore, all of these measurements fundamentally focus on repayment and, thus, riskmanagement. Although using unhealthy balances as measurements leaves room to bring in borrower
experiences, their typical application is institution centric.
A common feature in definitions of over-indebtedness is that problems exist despite the will of the
borrower, who gets into trouble innocently or at least unintentionally. Gonzalez (2008) provides an
exception, and asserts that overindebtedness occurs when the repayment outcome of a loan contract
does not correspond to the original expectations of either the borrower or the lender or both. Thus,
over-indebtedness can result from an unwillingness to repay, an inability to repay, or costly actions
that are required to repay. In fact, most indicators of over-indebtedness cannot distinguish the
borrowers intentions.
Definitions of over-indebtedness also differ in the sacrifices they expect from borrowers. In many
countries, bankruptcy proceedings require debtors to give up their incomes and assets above a
minimum existence level (see, e.g., European Commission, 2008). Similarly, microfinance
practitioners display the attitude that microcredit must not deprive borrowers of their basic survival
capacity (DeVaney, 2006). In developing his concept of debt-capacity, von Pischke (1991)
implies that microborrowers should be able to repay on time without suffering hardship. He states
that debt-capacity is not only about loan recovery but also about determining the loan amount that
can be safely provided to borrowers while protecting their interests and avoiding unacceptable risks.
A less severe definition considers borrowers over-indebted if they need to reduce expenditures
beyond the level they are accustomed to or make more sacrifices than expected (Murray, 1997). The
deprivation of any liquidity buffer for emergencies (European Commission, 2008) is a similar

7 This makes over-indebtedness from the customer protection point of view a potential lead indicator of arrears in MFI portfolios,
similar to earlier attempts to use drop-out rates or exit monitoring from lending programms as lead indicators for both social and
financial performance of MFIs, including arrears (see Copestake (2002)).

- 14 -

feature. The most generous definitions count every sacrifice by the borrower as a sign of overindebtedness (Canner and Luckett, 1991). This expectation is common in microfinance, because the
loans aim to make borrowers better off. In many definitions of over-indebtedness (e.g., insolvency
regulations) the sacrifice limit is only implicit. Canners explicit sacrifice considerations come
closest to meeting the needs of the customer protection movement and represent a borrower-centric
approach. However, this approach ignores the possibility that borrowers may deliberately sacrifice
their consumption levels to reach a certain goal.
This paper requires a definition that promotes a customer protection perspective. It should specify
what customers, not lenders, are to be protected from. Default or arrears are inappropriate criteria as
they do not sufficiently represent the borrowers perspective: For example, borrowers who manage
to meet their repayment obligations but who must take children out of school or skip meals to do so
should count as over-indebted.
Definitions of consumer insolvency for developed countries at least partly account for this customer
protection element by guaranteeing borrowers a minimum existence level. Many microfinance
markets however lack defined minimum existence levels and most borrowers already live below any
objective minimum level of income before assuming a loan. Therefore, they are the only ones who
can decide on an acceptable level of sacrifices. The definition requires qualitative elements to
account for individual borrower circumstances and will use the borrowers subjective perceptions as
the best judgments of their debt situations. The current EU definition of over-indebtedness includes a
similar element in considering the criterion that borrowers perceive their fixed monthly payments as
a heavy burden and their payment capacity as difficult or very difficult (European Commission,
2008). Subjective approaches which account for suffering as a subjective experience are in line with
the academic literature on subjective, self-reported measures of well-being and happiness (see Dolan
et al., 2008; Oshio and Kobayashi, 2010; Angner, 2010). Viewing the aim of development practice and for our case of microfinance- though a lens of impact on well-being, allows to distinguish
several mental models of well-being and thus of desired and non-desired outcomes (see Copestake,
2008, 2011). A subjective, person-centred view of over-indebtedness recognizes that well-being
importantly depends on peoples own feelings and that mere material measures ignore important
aspects of the broader needs, rights or local solidarity-based mental models of well-being.
This paper considers over-indebtedness a household level or individual phenomenon because
broader kin support should not be the reason that borrowers manage to repay their debt. It does not
refer to one-off struggles but implies that debt problems are more than transitory. At the same time,

- 15 -

our definition should account for the limited time horizons of microloans and the implications of
severe sacrifices even over short periods of time.
Therefore, this paper proposes a sacrifice-based definition of over-indebtedness that is unique in the
academic literature8: A microfinance customer is over-indebted if he/she is continuously
struggling to meet repayment deadlines and has to make unduly high sacrifices related to
his/her loan obligations that have more than transitory effects.9
This definition excludes borrowers who strategically default or deliberately accumulate
unsustainable amounts of debt. Although the boundaries are blurred, the criterion of unduly high
sacrifices precludes deliberation. To allow for sacrifices willingly incurred by borrowers, we deem
sacrifices to be unduly high if they exceed the costs that borrowers consider acceptable for the
purpose of the loan. The definition excludes mere short-term difficulties by specifying that, for
borrowers to count as over-indebted, their struggles need to be a continued phenomenon over time.
Sacrifices only represent signs of over-indebtedness if they have more than just transitory effects.
The definition is similar to an impact-based approach. If loans make borrowers suffer and worse off
because the sacrifices exceed the benefits, then these loans can cause over-indebtedness. However,
as in Western insolvency regulations, the definition does not postulate causality between the loan
and over-indebtedness. First, establishing causality between loans and struggles is highly impractical
due to the general attribution problems related to impact assessments (Hulme, 2000). Second,
beneficial loans that an emergency renders unbearable to repay also fall under the definition. The
next section will show that borrowers can be over-indebted for a complex set of reasons and that the
original cause of the problem is not necessarily a loan that is too large. Third, non-beneficial loans
for customers who do not get into serious struggle in spite of the negative loan impact do not fall
under the definition. There are thus cases of negative loan impact that are not cases of overindebtedness. This approach implies that a low level of over-indebtedness exists in any lending
market, particularly markets with poor and vulnerable borrowers. Preventing all loans for

8 The notion of sacrifice in microcredit repayments was first suggested by Brett (2006) who finds that microborrowers rely on cash
support from their families and social networks, assume additional debt, sell assets, reduce the quantity and quality of their food, and
take on additional paid labour to repay their debts on time. Gonzalez (2008) develops a broader definition that accounts for sacrifices
but includes strategic defaults. In Bolivia, he identifies similar sacrifices as well as reductions in human capital investments which can
reduce income generating capacity and household welfare in the long run. This research is in line with earlier findings on the coping
strategies of poor households by Corbett (1988).
9 Other versions of the papers based on this PhD use the term structural rather than more than transitory. See Appendix 2 for the
categories applied to this definition of over-indebtedness.

- 16 -

investments that fail or to borrowers who get hit by emergencies is impossible and would restrict
credit supply below its social optimum. However, such cases should be minimised.
Finally, the definition has parallels with an approach of rational choice under uncertainty that would
consider borrowers as over-indebted if, with hindsight, they regret the amount of debt they took on.
Under an ontology of complexity and uncertainty, borrowers who make rational loan decisions may
take loans that they would not take if all information was available and understandable to them
loans that they will regret ex-post. Behavioural economics allows for further deviations from the
original homo oeconomicus assumptions that explain why individuals may take debt that they will
regret (see next section for related literature and detailed arguments). If borrowers could be
considered rational in the sense of regretting loans as soon as their impact was negative, this
argument would be the same as the impact argument, but regret allows for some more flexibility and
circumvents the challenges of impact measurement. Nevertheless, a borrower may regret borrowing
without having reached a stage of over-indebtedness. And even an over-indebted borrower may not
regret having borrowed if for example a medical emergency drove him to get indebted or if the
borrowers mental representations see the cause of over-indebtedness rather in, lets say, a failed
investment than in the act of borrowing for investment. Regretting ones situation with ones debt
does not automatically equal attributing ones regret to the debt.
To put this definition into practice and empirically measure over-indebtedness in microfinance,
Figure 2 suggests using a funnel of over-indebtedness criteria that filters out borrowers at each step
until only the over-indebted remain. Researchers can collect the information by conducting
household surveys, interviewing respondents about their debt struggles and collecting detailed lists
of the sacrifices that respondents experience related to their loans. Respondents subjectively weight
their sacrifices by acceptability and indicate the frequency of each sacrifice they experience. We
consider sacrifices to be unduly high if they are unacceptable to the borrower. They are considered to
have more than transitory effects, if they have been performed repeatedly (e.g., more than three
times) or if even a one-off occurrence is a sign of non-transitory problems (e.g., asset seizures).

- 17 -

Figure 2. A survey-based measurement of microfinance over-indebtedness from a customer


protection perspective

All Borrowers

Over-indebted
Borrowers

Make
unacceptable
Struggle to
always repay sacrifices
on time

Experience
sacrifices that
indicate structural
problems1 or
Make non-structural
sacrifices
repeatedly

1 For example suffering an asset seizure, or selling/pawning ones assets

As in any research that leverages the power of subjective measures, subjectivity reduces the claims
to universality that can be made. Over-indebtedness is a context-specific phenomenon and the
borrowers subjective perceptions of what constitutes a sacrifice and how acceptable the different
sacrifices are depend on cultures, local contexts and individual borrower characteristics. This
reduces the comparability of over-indebtedness based on this measurement across settings, as well as
the validity of findings based on this measurement for the microfinance industry in general.
Nevertheless, the subjective measure allows for important insights into the real nature of overindebtedness that remain obscured by looking at more simplified, objectivised measures of overindebtedness. In an unavoidable trade-off, the more universal and interculturally comparable a
measurement is, the less it reflects local realities (Copestake, 2008).
Currently a number of research projects are using similar approaches to gather empirical data about
the sacrifices of over-indebted borrowers. These projects will reveal the degree to which
microborrowers struggle with debt in various markets. As extensive survey work is costly, future
research should analyse to what extent simpler indicators (e.g., debt-service ratios, delinquency, or
the number of credit arrangements) represent workable proxies for this definition. Doing so will also
show to what extent credit risk management reduces over-indebtedness from a customer protection
perspective. Until then, simply relying on the existing risk management indicators for customer
protection purposes is inappropriate and risks ignoring many struggling borrowers who manage to
avoid delinquency by making heavy sacrifices.
- 18 -

3. The reasons for over-indebtedness: why do people overborrow?


Microfinance customers have good reasons to take loans but little reason to borrow to the extent of
over-indebtedness. Similarly, lenders, especially MFIs with social missions, should have no interest
in over-burdening their customers with debt. With a few exceptions10, over-indebtedness is an
undesirable consequence for both parties and should not exist in perfect markets. This section
analyses why people borrow more than they can handle by examining the influence of external
factors, lenders, and the borrowers themselves. We develop an inclusive framework of the causes of
over-indebtedness that will allow future researchers to analyse the full scope of influence factors, be
more comprehensive in their empirical approaches, and develop solutions to over-indebtedness
based on their full understanding of the issues.
In applying empirical findings from the consumer credit literature to the microfinance context, there
are several challenges to keep in mind:
Studies differ in their definitions of over-indebtedness. Some are not specific to overindebtedness but analyse the causes of debt, the accumulation of credit, delinquency, or
financial difficulties.
Studies vary in the loan products that they consider. Borrower behaviour may differ between
for example credit card use and decisions on instalment loans.
Borrower behaviour and characteristics differ among Western markets (Betti et al., 2007) and
even within countries (Collins, 2008). The applicability of the findings to microfinance markets
may be limited.
Few studies identify causal relationships (e.g., Webley and Nyhus, 2001, based on a
longitudinal approach or Brown et al., 2005, with the Granger causality test). Most empirical
studies use cross-sectional data to identify characteristics and behaviour associated with debt
or over-indebtedness. They are unable to exclude spurious correlations.
Consequently, existing research is a suitable starting point but microfinance specific research will
be required to confirm and extend the findings below.
Outside influences: One set of factors that may push borrowers into over-indebtedness lies outside
the control of the lending parties. Despite sound lending decisions, external shocks to the incomes
or expenses of microborrowers can render their debt loads unsustainable. An empirical study by

10 From a profitability perspective, lenders may accept some arrears and defaults as long as the costs of improving risk management
exceed the costs of delinquency. Apart from the indirect impact through customer satisfaction, lenders only consider repayment
performance, not the sacrifices of borrowers. In some cases, even repayment problems may be in a lender's interest, as they can
increase the interest earned on a given loan or keep the borrower dependent on the lender.

- 19 -

Bouquet et al. (2007) in Madagascar confirms that credit problems are most frequently due to
declining income or rising expenses. In addition to personal shocks (e.g., illness or job loss),
macroeconomic developments (e.g., financial crises) can drive borrowers into difficulties. Typical
factors for developing countries include natural disasters, changes in government policies (e.g. the
displacement of street vendors), and political crises.
Furthermore, the institutional and legal environment influences the behaviour of lenders and
borrowers. For example, the existence of credit bureaus, the efficiency of the judicial system, and the
level of competition can enhance or reduce the risks of over-indebtedness. A common misconception
states that competition among MFIs will always benefit customers by making MFIs more
professional and efficient, bringing down costs and improving customer service. In fact, competition
increases asymmetric information by making information sharing more difficult (McIntosh and
Wydick, 2005), reduces repayment incentives due to the availability of alternative credit sources
(Gonzalez, 2008) and creates pressure on MFIs to overlend, particularly if new market entrants are
aggressive and pay less attention to repayment capacity. Hellmann et al. (2000) show that by
reducing the capitalised value of expected future profits, competition lowers the incentives for
making good loans and incites opportunistic lender behaviour. For clients of Grameen Bank, Matin
(1997) and Chaudhury and Matin (2002) find that access to multiple lending NGOs reduces
repayment performance. Vogelgesang (2003) finds that in Bolivia, competition has positive effects
on repayment in good times but negative effects in times of crisis and over-indebtedness. Only
Krishnaswamy (2007) finds better repayment rates in more competitive branch locations in India.
This may result from an initial effect of multiple borrowing on hiding repayment difficulties.
Incentive structures also result from the priorities of donors and investors. A strong focus on
outreach and sustainability may lead to excessive growth and profit-seeking behaviour from MFIs,
as discussed below. Understanding these external influences on over-indebtedness is a prerequisite
for developing appropriate solutions. The answers may include regulatory measures, promoting
credit bureaus, improving safety nets for borrowers, or simply factoring the likelihood of shocks into
the evaluation of repayment capacity.
Lender behaviour: In all institutional environments, over-indebtedness is ultimately created by the
parties that make the credit decisions, i.e., the lenders and the borrowers. Three major levers of
lending behaviour exacerbate over-indebtedness risks: a) an excessive focus on marketing and
growth, b) unsuitable product characteristics, and c) lending procedures conducive to overindebtedness. All of these factors may be exacerbated by extreme profit-seeking behaviour from
MFIs but are not limited to for-profit institutions. Well-managed for-profit MFIs that take a long- 20 -

term view, rather have reasons to avoid such behaviours and take both portfolio quality and customer
satisfaction into account.
a) MFIs can push borrowers beyond their limits through an exaggerated focus on portfolio growth
and aggressive marketing techniques. The lending focused business model of MFIs can create
pressure on customers to continue borrowing instead of focussing on the protective services that they
need or taking a break when they do not require credit (Collins et al., 2009).11 Another mechanism
driving MFIs to over-indebt clients is the volume-focused incentive system (Rahman, 1999).
Volume-focused incentives reward credit officers for disbursing loans, even if clients struggle to
repay them. Many MFIs mitigate this effect by providing complementary incentives for portfolio
quality to prevent loans to borrowers who will default. However, in existing lending relationships,
portfolio quality incentives can trigger new loans to delinquent borrowers, to enable borrowers to
maintain their repayments beyond their capacity. Finally, aggressive sales techniques can incite
borrowers to borrow beyond reasonable limits. A randomized control trial by Bertrand et al. (2010)
in South Africa shows that advertising content significantly affects demand, particularly by
appealing to intuition rather than reason. Small features such as limiting the numbers of examples
shown, not suggesting what loans should be used for, and adding pictures of attractive women
trigger the same increase in demand as a 25% lower interest rate. This does not yet include
particularly aggressive techniques such as door-step sales, limited time offers, or hiding loan costs.
At the root of the MFIs sales focus are the industry's growth pressure and the desire for financial
returns that may overrule an institutions social mission. In extreme cases, rapid growth may exceed
the institutional capacity of the MFIs information systems, lending policies and governance
structures. Particularly, growth rates above 250 per cent, local growth in existing branch locations,
and growth in countries with a penetration rate above 10 per cent or high aggregate microfinance
growth present risks to the portfolio quality of MFIs (Gonzalez, 2010). According to pre-crisis MIX
Market data12, the gross loan portfolios of Morocco's 10 reporting MFIs grew at a compound annual
growth rate of 110 per cent in 2006 and 2007, with portfolios more than doubling every twelve
months. In 2007, the fastest growing MFI (ARDI) grew at 330 per cent, multiplying its loan
portfolio by more than four in the course of one year.
This growth speed creates certain challenges for an institutions management. Rapid growth leads to
a high share of inexperienced loan officers, who may make poorer loan decisions than their
experienced colleagues and identify less with the values of their institutions. ARDIs personnel grew
11 Banerjee and Mullainathan (2009) explain that MFIs have an incentive to keep profitable borrowers trapped in a cycle of debt.
12 www.mixmarket.org, self-reports of limited liability but sufficient for the purposes of this argument

- 21 -

by approximately 350 per cent in 2007 such that nearly 80 per cent of its employees were new to the
institution. Two years after Moroccos MFIs experienced strong growth, the >30 day portfolio at risk
(PAR) rose from below 1 per cent to an average of 7 per cent. This data excludes ARDI which
stopped reporting its PAR.
Nevertheless, there are also examples of institutions that maintained high growth rates for years
without exceeding their institutional capacities. Fast growth is likely to be most problematic on a
market level when markets approach saturation. To date the concepts of market saturation and
penetration are still poorly conceptualized in relation to microfinance. The above estimates only
provide rough rules of thumb.
b) MFIs further increase over-indebtedness risks if they offer products that are inappropriate to
the borrower's situation. In this case, lenders are not inducing borrowers to take on more credit
than is reasonable. Rather, the impact chain centres on the debtors ability to meet their obligations
from existing contracts. For example, the typical repayment schedule starts too early for most
investments to generate returns. And if maturities are too short and instalment schedules are too
inflexible, it can be difficult for borrowers with volatile incomes to repay their debts on time (Collins
et al., 2009; Gurin et al., 2009a). A high-frequency repayment schedule helps borrowers to repay
because small, regular amounts of money are easier to assemble than larger ones. It also is an
important disciplining mechanism for borrowers and highlights repayment problems early on.
Nevertheless, microfinance practitioners need to abandon the myth that only standardised, shortterm, high-frequency loans work, and better adapt to the needs of borrowers and their businesses.
Even if only testing their hypothesis in a low-delinquency environment in urban India, Field and
Pande (2008) find that MFIs can switch from weekly to monthly instalments without reducing the
borrowers repayment discipline. More flexible options should be tested in future research.
The difficulties become exacerbated if MFIs do not reschedule loans for honest borrowers who face
liquidity problems. The zero-tolerance policy promotes the early recognition of repayment
difficulties and keeps borrowers from accumulating obligations (Gonzalez, 2008), differentiates MFI
lending from frequently forgiven government lending, and avoids strategic defaults and spillover
effects of delinquency on non-delinquent borrowers (Krishnaswamy, 2007). The policy improves an
MFIs portfolio quality and has operational benefits in helping MFIs to avoid handling large
numbers of rescheduling requests. However, it prevents loan officers from rescheduling loans for
borrowers facing short-term liquidity problems. By forcing these borrowers into delinquency and
potentially adding late fees or increased interests to their obligations, the zero-tolerance policy may
not only put obligors under avoidable repayment pressures but also turn short-term liquidity gaps
- 22 -

into longer-term problems. For example, in a South African MFI, the zero-tolerance policy was so
pervasive that MFIs proudly reported borrowers going hungry and selling their food to repay on time
(Smets and Bhre, 2004). According to our sacrifice-based definition, these borrowers are overindebted.
A regular instalment schedule and the zero-tolerance policy serve a good purpose and might help
some borrowers to stay out of over-indebtedness. At the same time, these policies are absolutely
inappropriate for other borrowers. Unless the microfinance industry develops a more nuanced
approach, the policies may create the same debt problems that they aim to prevent.
Repayment capacity is also a function of loan use. In Bouquet et al.'s empirical work in Madagascar
(2007), the most frequent product feature that borrowers cite as a reason for repayment problems is
the timing of disbursements. If MFIs disburse too late for the borrowers to exploit their business
opportunities, customers may not earn enough returns. Similarly, scholars frequently identify loans
used for consumption purposes as a source of over-indebtedness as these loans do not provide
debtors with returns for repayment (Vogelgesang, 2003). This argument however should be handled
with care: First, as money is fungible and a distinction hardly ever exists between household and
microenterprise cash flows, many microloans go to household purposes. Second, distinguishing
between consumption and production loan use can be difficult, as in the example of educational
expenses (Collins, 2008). Third, although increasingly rigorous impact studies are challenging the
empirical results of microfinance, microfinance theory is gradually moving away from the
microenterprise approach to a household finance approach. It considers the benefits of microfinance
to be in short-term consumption smoothing and managing the risks of low and volatile incomes
(Collins et al., 2009; Karlan and Zinman, 2009). Therefore, consumption loan use can be both a
cause of and protection from over-indebtedness.
c) Microlenders also contribute to over-indebtedness with their lending procedures. Particularly lax
evaluations of repayment capacity and automatic increases in loan sizes over time create overindebtedness risks. In Bangladesh and Bolivia, larger loan amounts and higher debt-to-asset ratios
are positively related to borrowers repayment problems (Sharma and Zeller, 1997; Vogelgesang,
2003; Godquin, 2004).
Moreover, lending procedures promote over-indebtedness through a lack of transparency. An Indian
study discovered that more than 70 per cent of microcredit clients believed all or most of the rules
regarding their loan contracts not to be communicated in written form (Tiwari et al., 2008).
Excessive interest rates and fees contribute to debt problems, especially if borrowers are not fully
aware of them. There is a misconception in customer protection debates that MFIs should simply
- 23 -

state Annual Percentage Rates (APRs), as APRs are fair measures of borrowing costs that allow for
comparisons among MFIs and are less subject to lenders attempts to hide the full cost of credit. In
reality, the APR is not well aligned with the mental processes of human beings.
People rarely think about spending and borrowing decisions via discount rates as economic logic
suggests. Instead, humans make their spending decisions by utilising simpler mental representations
that fit their cognitive capacities. Ranyard and Craig (1995) combine the original theories of mental
accounts (Thaler, 1985) and psychological accounts (Kahneman and Tversky, 1984) in the concept
of dual accounting: People consider the total cost of a loan and the affordability of the instalments in
every future loan period to make borrowing decisions. In most cases, discounting does not play a
role. Empirical findings confirm that microfinance clients in India think about their loans in terms of
how much they owe on a weekly basis but know little about interest rates or total interest expenses
(Tiwari, Khandelwal and Ramji, 2008). Working with students in Britain, Ranyard and Craig (1993)
find that without information on the total interest charged, borrowers strongly underestimate the
repayment periods and total amount to be repaid. Therefore, they may borrow more than they can
handle.
Finally, Collins et al. (2009) point out that APRs may not represent appropriate measures of price,
especially for short-term microloans. If borrowers make loan decisions for only a few weeks, they
decide based on the actual impact of the loans on their budgets rather than a hypothetical annual
price. Pointing out the high APR of a loan to borrowers equates to dissuading someone from taking a
taxi to the airport based on the cost that they would face if they used taxis all year long. Instead,
many microfinance borrowers perceive the interest on loans as a fee for a service (bridging the
liquidity gap) rather than the intertemporal cost of funds. Therefore, using APRs to represent the cost
of credit is not compatible with borrowers cognitive approaches to loan decisions. In developing
countries with weak educational systems, where many borrowers do not understand percentages, this
misconception is likely to be even stronger. If loan information is incompatible with borrowers
mental processes, it may contribute to over-indebtedness.
The final trade-off concerning lending policies is related to collection practices. Microfinance
institutions in India and elsewhere increasingly face accusations of employing collection practices
that increase the burden of indebtedness for borrowers by humiliating and intimidating them or
depriving them of their assets. Borrowers are not necessarily aware of their rights and, in India,
many borrowers accept that MFIs force borrowers to discuss their delinquency in public, extend
group meetings at the cost of business time or punishment by husbands, or seize assets in spite of
uncollateralised loan contracts (Tiwari, Khandelwal and Ramji, 2008). At the same time, if
- 24 -

collections are too lax, repayment difficulties may only materialise when over-indebtedness has
already become unavoidable (Vogelgesang, 2003).
This analysis of the lenders contributions to over-indebtedness prompts a strong message to those
seeking to address over-indebtedness risks: reducing access is not the only answer. The term overindebtedness bears the connotation of too much debt and conveys the idea that the only solution to
too much debt is less debt. This approach is only suitable for addressing cases of excessive growth,
aggressive marketing and lax credit distribution policies. Where over-indebtedness results, for
example, from unsuitable product features or is aggravated by inappropriate collection practices,
borrowers do not need less credit but better credit. The microfinance industry will have to continue
developing its methodology and increasingly shift its focus from pure outreach to delivering quality
services. This entails a broader product portfolio that features more flexible lending products and
broader access to savings and insurance products.
Borrower behaviour: There are two parties to every credit decision: a lender and a borrower. Both
are obliged to make responsible choices. Both can render a credit agreement harmful. A customer
protection perspective should resist the temptation of finding fault only with lenders and adverse
circumstances. It should also consider the role that borrowers themselves play in destructive credit. It
should recognise that, to a certain extent, protecting borrowers from over-indebtedness may mean
protecting them from themselves.
a) The psychological and behavioural economics literatures have revealed numerous cognitive and
psychological biases that can cause borrowers to over-burden themselves with debt. Contrary to the
postulates of traditional economic theory, the rationality of human decision makers is bounded by
imperfect information and limited cognitive abilities such that errors in judgement occur. For
example, individuals with limited information tend to rely on proxies for their decision making.
Inexperienced consumers tend to view their credit limit as a signal of future earnings and
affordability (Notani, 1997; Soman and Cheema, 2002). Being offered more credit may encourage
individuals to borrow beyond what they can afford. People do not always understand financial
choices well. Their cognitive resources mediate the impact financial services have on their wellbeing (Nio-Zarazua and Copestake, 2009). Lusardi and Tufano (2009) discover strikingly low
levels of debt literacy across the USA. Those who are less debt literate are more likely to be overindebted. Additionally, decision makers in Western cultures tend to suffer from an overconfidence
bias and from habit persistence, which makes them reduce consumption too slowly in response to
income fluctuations (Brown, 1952). They may overconsume and underestimate the risk of
borrowing. From the perspective of the poor, Gurin (2000) reveals the cognitive effects of
- 25 -

permanently living in a precarious situation: the constant focus on managing daily emergencies
changes people's relationship to time so that the short term assumes disproportionate importance.
Individuals may lose the ability to plan for the longer term.
According to the theory of procrastination by Akerlof (1991) and the hyperbolic discount function
(Laibson, 1997), human beings display time-inconsistent behaviour. Near-term events are subject to
higher discount rates than distant events. The losses from saving money one day later and consuming
today seem negligible. However, individuals fail to foresee that the dilemma will be the same in all
the following instances. They eventually consume beyond their means and are left without any
savings. For borrowers, the effect may be the same. A small purchase on credit today does not cause
problems, but the chronic accumulation of credit leads to over-indebtedness. Gul and Pesendorfer
(2004) and Fudenberg and Levine (2006) provide similar explanations with their theory of
temptation and the dual self model, respectively.
These general human characteristics are likely to apply to microborrowers as well. As a result,
microborrowers who usually have low educational levels and little experience with formal lending
institutions, struggle to understand credit offers and make sound borrowing decisions. If these
challenges are combined with the attractiveness of the Western lifestyle and microfinance
development aid rhetoric, MFI customers may be unable to restrict their borrowing to sustainable
levels. Ashraf et al. (2006) and Bauer et al. (2009) provide empirical support for time inconsistent
behaviour among the poor. Being present-biased, the poor use loans as commitment devices to
replace savings. Their behavioural biases are the mains reason they get into debt.
Banerjee and Mullainathan (2009) and Strotz (1956) show that the poor may be even more
vulnerable to temptation than the rich because poor people are unlikely to improve their financial
situations even if they behave rationally. For example, their potential daily savings are very small
and the likelihood high that the savings get consumed before having accumulated enough to make an
investment. In this situation, even starting to save may not seem worthwhile (Banerjee and Duflo,
2007), an effect later labelled "temptation tax" (Banerjee and Mullainathan, 2009). Also, for the rich,
temptation spending may remain at unproblematic levels because of declining temptations, but that
same level may be critical for the poor (Banerjee and Mullainathan, 2009). Bertrand et al. (2004) add
that if the poor are prone to the same biases as everybody else, in their vulnerable situations these
biases have more dire consequences. In summary the above findings from behavioural economics
imply that microfinance customers are at risk of over-indebtedness because their rationality is
limited by a number of biases and weaknesses. They may not always make smart choices about debt.

- 26 -

Another relevant factor in the psychology literature is the concept of attitude. A borrowers attitudes
towards the future (Brown et al., 2005), debt (Livingstone and Lunt, 1992; Lea, Webley and Levine,
1993), money and money management (Webley and Nyhus, 2001) and his/her locus of control
(Livingstone and Lunt, 1992; Webley and Nyhus, 2001) all impact the risk of over-indebtedness.
Microborrowers and their attitudes may differ from the findings of British studies, but the
mechanisms are likely to be similar. For microfinance investment loans, this may include attitudes
towards risk taking in business decisions. If a certain attitude implies risks of over-indebtedness,
MFIs might want to be careful in lending to these customers or introduce countermeasures, such as
awareness campaigns and training programs.
b) Sociological factors can also drive borrowers towards over-indebtedness. The pressures of
consumer society and materialism may lead consumers to overspend and borrow beyond their limits.
In France (probably like any other materialistic society), instead of satisfying basic needs,
consumption is required to build ones identity (Viaud and Roland-Lvy, 2000). If sociologists see a
causal relationship between a culture of consumption and the level of consumer debt, lenders should
be aware that borrowing decisions may result from societal pressures. This phenomenon is likely to
be similar among the poor, at least in cultures with consumer societies.
A closely related influence factor is inequality and social comparison. US research shows that this
factor can drive lower income groups to borrow irresponsibly to keep up with the consumption
levels of their peers (Duesenberry, 1949; Christen and Morgan, 2005; Luttmer, 2005). Future
microfinance specific research should confirm this effect in different cultural settings.
As a third point, economic socialisation influences peoples readiness to take on debt. In Britain,
societal support for debt determines to what extent the average borrower perceives debt as a burden
or as an easy solution (Lea, Webley and Levine, 1993). According to a US study, the attitudes and
experiences of parents with debt and money are relevant (Stone and Maury, 2006), and few
microborrowers have learned handling formal credit from their parents. Furthermore, social
obligations to support relatives and neighbours may drive people into indebtedness. These
obligations are likely to be strong in microfinance environments.
c) Finally, empirical studies analyse socio-demographic and economic characteristics for their
associations with repayment problems and over-indebtedness. Some relationships may not be causal
in nature, others probably are. The most common socio-demographic factors in consumer finance
and some microfinance studies include young age, more and younger children, low levels of
education and literacy and low or unstable labour market status (Livingstone and Lunt, 1992; Lea,
Webley and Levine, 1993; Lea et al., 1995; Vogelgesang, 2003; Godquin, 2004). In line with the
- 27 -

findings on education, studies in Burkina Faso and Bolivia show that a lack of borrower training and
experience with past loans is related to a lower payment performance (Paxton et al., 2000; Schreiner,
2004). Conversely, repayment performance in group lending may decrease over loan cycles if repeat
loans no longer meet the needs of every group member to the same extent (i.e., the matching
problem) (Paxton, Graham and Cameron, 2000; Godquin, 2004).
Single adult households and those with recent changes in family composition (e.g., divorce) face
over-indebtedness risks (Lea, Webley and Levine, 1993; Vogelgesang, 2003). Similarly, ill health,
gender and ethnicity are related to indebtedness (Lea, Webley and Levine, 1993; Godquin, 2004;
Disney, Bridges and Gathergood, 2008). Many of these characteristics are prominent among
microfinance customers and might also be related to indebtedness in the developing country context.
Even if they should not lead to the exclusion of the target group, they should be taken into account in
avoiding over-indebtedness.
The most common economic driver of indebtedness is low income (Livingstone and Lunt, 1992; Lea,
Webley and Walker, 1995). In a developing country setting, Collins (2008) finds a correlation
between indebtedness and income in urban but not rural areas. A study from Burkina Faso finds that
borrowers in rural areas may be more prone to experiencing repayment problems, potentially
because of higher poverty levels (Paxton, Graham and Cameron, 2000). Other research points
towards the influences of income instability (Webley and Nyhus, 2001), low wealth (Sharma and
Zeller, 1997; Godquin, 2004; Disney, Bridges and Gathergood, 2008) and low returns on the
borrowers investment (Gonzalez, 2008). The role of returns probably explains Schreiner s finding
of a difference in risk between various economic sectors and the borrowers business activities
(2004). While MFIs are aware of lending to customers with low, unstable incomes, and low wealth
levels, they may overestimate their customers returns on investment and fail to consider the fact that
a large share of their loans is not actually put to productive use.
The impact of borrower characteristics on over-indebtedness does not only point towards selecting
the right borrowers to reduce over-indebtedness. They also suggest that, for example, financial
literacy training, awareness campaigns, and the right to withdraw from spontaneous credit decisions
might help reduce over-indebtedness. Additionally, they show that MFIs should match their
communication with the borrowers educational levels and that advertisements should not exploit the
borrowers psychological biases.
To sum up, Figure 3 provides an overview of the causes of over-indebtedness. External influences,
such as adverse shocks to income or expenses, can render a debt load unmanageable. The
institutional and legal environment can enhance or reduce the risks of over-indebtedness. In
- 28 -

microfinance markets, institutional protection from over-indebtedness may be weak. A significant


share of the responsibility for over-indebtedness lies with lenders. MFIs can push borrowers beyond
their limits if they focus excessively on portfolio growth and utilise aggressive marketing techniques.
MFIs sometimes offer products that are inappropriate for the borrower's situation, enforce unrealistic
instalment schedules, resist the need to reschedule loan agreements, or artificially limit maturities.
Microlenders also contribute to over-indebtedness through their operating procedures, by being lax
when evaluating repayment capacity, offering nontransparent terms and conditions, and using
coercive collection practices. Finally, microborrowers play an important role in their own overindebtedness. Because of cognitive limitations, difficulties in resisting temptation, and social
pressures, they sometimes make irresponsible borrowing decisions. The tendency towards overborrowing also depends on the borrowers' socio-demographic and economic characteristics, and
many of those that are related to over-indebtedness are particularly prominent among microfinance
clients.
Figure 3. Drivers of over-indebtedness
1

External factors

Adverse shocks
Institutional

environment (e.g.
macroeconomic or
legal)

Over-indebtedness as a
consequence of
interacting factors
2

Lender behaviour
Marketing and
growth focus
Inflexible products
Unfair lending procedures / collections

Borrower behaviour

Cognitive and

psychological biases

Sociological influences
Socio-demographic /
economic attributes

If a credit relationship leads to over-indebtedness, the cause is usually a combination of interacting


factors. If borrowers made perfectly informed, rational decisions and did not give in to temptation or
social pressures, MFIs would face little risk of over-burdening them with debt. At the same time, if
MFIs prioritised their social mission and the customers interests above all other goals, designed
products that perfectly matched their customers needs, carefully evaluated repayment capacities and
- 29 -

accounted for the psychological limitations and biases of borrowers in their communication and
lending decisions, even imprudent borrowers would hardly be at risk.
In reality, both parties are bound to make mistakes because of their complex situations and
asymmetric information. Adverse shocks may turn even the best borrowing and lending decisions
into over-indebtedness triggers. Therefore, measures against over-indebtedness should address the
full range of drivers of over-indebtedness. The relative importance of the respective influence factors
remains to be determined.
Based on this more detailed understanding of the causes of over-indebtedness, the solutions extend
far beyond the standard risk management reactions to over-indebtedness trends. The risk
management-oriented reactions of reducing access to loans, decreasing loan sizes, and intensifying
due diligence are appropriate in situations where lender-side causes are predominant, for example
because MFIs are pushing loans to customers too aggressively, are over-prioritizing portfolio growth
and are not conducting sound evaluations of repayment capacity. However, this paper has revealed a
much broader range of potential over-indebtedness causes that suggests a wider range of measures to
be taken.
MFIs should make loan schedules more flexible if repayment struggles arise predominantly from
inappropriate products such as loans with rigid instalment schedules that do not match the
borrowers cash flows. Particularly in markets with low financial literacy levels, social pressures and
a credit culture that is conducive to over-indebtedness, MFIs should ensure transparent
communication that is suitable to the mental processes of borrowers and work towards
improvements in financial literacy among their borrowers. Literacy trainings should be tailored
towards the mental needs of borrowers and address psychological and behavioural biases directly by
pointing them out to borrowers and suggesting strategies for borrowers to deal with them. Lenders
can also counter over-indebtedness trends by offering and promoting savings products to create a
safety buffer for borrowers in difficulty. Deposit products with commitment features may reduce the
use of loans as commitment devices for present-biased borrowers.13
The mere awareness on the part of MFI managers and loan officers that over-indebtedness begins
long before default, and that they should consider the borrowers sacrifices, can already go a long

13 Not in relation to over-indebtedness but as measures to enhance debt capacity, some items such as rescheduling, flexibilising
instalment schedules or adapting them to cash flows, and mobilizing savings have been mentioned in von Pischke (1991). Taking the
risk of adversities into account constitutes an essential feature of his concept of debt capacity. He further pointed to more general
measures that increase the cash flows of the poor, such as improving infrastructure, investing in education, or influencing market
prices, indicating that debt capacity depends on social and institutional infrastructure and is created by people through markets.

- 30 -

way. It should trigger changes in the institutional culture, in the prioritization of impact targets
versus portfolio quality goals, in the incentives for loan officers and in the MFIs lending policies.
The same is true for recognising which type of lender behaviour puts clients at risk, for example
aggressive marketing, not allowing for breaks between loans cycles, and insisting on a zero-tolerance
policy. Understanding the underlying causality is the first step for MFIs to avoid the behaviour in
question. Nevertheless, MFIs will not be able to reduce over-indebtedness to zero, unless they stop
lending and deprive all of their customers from the benefits of access to credit. For the remaining
microborrowers that still cannot be protected from over-indebtedness, curative and rehabilitative
customer protection measures (e.g., debt relief) are required that are largely outside of the lenders
responsibility. Financial market regulators, governments as those responsible for social security
networks, and microfinance donors and investors should revisit what their role can be in reducing
over-indebtedness given this papers analysis of drivers behind the over-indebtedness phenomenon.

- 31 -

4. Conclusion
For several reasons, over-indebtedness poses a serious risk to the microfinance industry. First, it
contradicts the sector's social mission. Over-indebtedness can exacerbate the poverty of
microfinance customers, threaten their social positions and lead to psychological disorders and
health problems. Second, over-indebtedness can threaten the sustainability of MFIs by triggering
defaults. Third, over-indebtedness may have spillover effects that would affect even healthy
institutions and their customers. It risks damaging the general reputation of the microfinance
industry. Given the trend to interpret the impact of microfinance as a reduction in the vulnerability of
poor households, it is interesting to note that for certain customers the evidence for overindebtedness suggests an increase in household vulnerability due to microborrowing. Based on an
improved understanding of the causes of over-indebtedness, the industry must implement customer
protection measures to reduce the risk of over-indebtedness for MFI customers while simultaneously
providing relief to those in trouble.
This paper develops an innovative definition of over-indebtedness based on the concept of sacrifice.
For purposes of customer protection, we consider microborrowers over-indebted if they are
experiencing unacceptable sacrifices with more than just transitory effects. The definition considers
their subjective knowledge about their future prospects and the implications of the different types of
debt. Our definition represents a customer-protection-based alternative to the usual definitions of
over-indebtedness which, based on risk management, focus on default (the late stages of overindebtedness), or on somewhat arbitrary debt-to-income ratios.
Furthermore, this paper provides an unprecedented overview of the causes of over-indebtedness
based on previous microfinance research and the interdisciplinary consumer finance literature. It
highlights the interacting factors and responsibilities behind the phenomenon of over-indebtedness.
The causal framework shows that lenders exacerbate the risks of over-indebtedness via their focus
on marketing and growth, unsuitable product characteristics and inappropriate lending procedures.
Borrowers contribute to the problem through cognitive biases, psychological misrepresentations, and
social pressures. In addition, external influences, such as adverse shocks to borrowers incomes and
expenses, can drive borrowers into over-indebtedness, especially if they are facilitated by the
institutional and legal environment. The paper challenges a number of conventional beliefs, such as
the inappropriateness of consumer loans for microfinance customers, the requirements of tight
repayment schedules in the microfinance methodology, and the universal benefits of competition for
customers.

- 32 -

By analysing the causes of over-indebtedness we provide the basis for designing appropriate
customer protection mechanisms to address the phenomenon at its roots. Only with an understanding
of the different factors that create over-indebtedness can regulators, MFIs, and other parties craft
effective over-indebtedness protection measures for the microfinance market. This framework
reveals the breadth of the potential solutions to the challenge of over-indebtedness and shows that
reducing over-indebtedness does not only imply a reduction of access to credit. There are many
possible solutions such as improvements to product design and lending policies, borrower training,
or appropriate marketing and communications. Often, borrowers do not need less credit but better
credit.
Further research is required in this young field, to confirm to what extent the drivers of overindebtedness among the consumers of developed countries apply to microfinance and to determine
which of these drivers are the most important. Empirical research should measure the extent of overindebtedness in specific geographies and test the causes of debt problems in these markets.
Additionally, qualitative research needs to provide insights into the experiences of over-indebted
customers and the sacrifices that they make. On the basis of these findings, the microfinance industry
will have to develop customer protection measures.
By refocusing on the interests of clients and making customer protection a top priority, microfinance
has made important progress. Academic research needs to support this step by pointing out the
implications of shifting from a repayment performance focus, i.e., an institution-centric risk
management perspective, to a customer focus. This includes an appropriate definition of overindebtedness and a more comprehensive view on the origins of the over-indebtedness phenomenon.
While an over-indebtedness definition based on subjective perceptions implies automatic limitations
with regards to universality, it can reveal a perspective on microlending that the industry has long
ignored and allows for a much richer understanding of the causes of over-indebtedness than a more
limited definition would suggest.

- 33 -

References
Akerlof, G. A. (1991) 'Procrastination and Obedience'. The American Economic Review, 81(2), 1
19.
Angner, E. (2010) 'Subjective well-being: Financial Crisis: Capabilities and Happiness'. Journal of
Socio-Economics, 39(3), 361368.
Ashraf, N., D. Karlan and W. Yin (2006) 'Tying Odysseus to the Mast: Evidence from a
Commitment Savings Product in the Philippines'. Quarterly Journal of Economics, 121(2), 635
667.
Banerjee, A. and S. Mullainathan (2009) 'The shape of temptation: Implications for the economic
lives of the poor'. Harvard University & MIT.
Banerjee, A. V. and E. Duflo (2007) 'The Economic Lives of the Poor'. Journal of Economic
Perspectives, 21(1), 141167.
Bauer, M., J. Chytilov and J. Morduch (2009) 'Behavioral Foundations of Microcredit:
Experimental and Survey Evidence from Rural India'. Financial Access Initiative, New York.
Benno, F. (2008) 'Proposal for a Contingency Debt Sustainability Framework: Special Section:
Social Movements and the Dynamics of Rural Development in Latin America (pp. 2874-2952)'.
World Development, 36(12), 25472565.
Bertrand, M., D. Karlan, S. Mullainathan, E. Shafir and J. Zinman (2010) 'What's Advertising
Content Worth? Evidence from a Consumer Credit Marketing Field Experiment'. Quarterly
Journal of Economics, 125(1), 263305.
Bertrand, M., S. Mullainathan and E. Shafir (2004) 'A Behavioral-Economics View of Poverty'. The
American Economic Review, 94(2), 419423.
Betti, G., N. Dourmashkin, M.C. Rossi and Y.P. Yin (2007) 'Consumer over-indebtedness in the EU:
measurement and characteristics'. Journal of Economic Studies, 34(2), 136156.
Bouquet, E., B. Wampfler, E. Ralison and M. Roesch (2007) 'Trajectoires de crdit et vulnrabilit
des mnages ruraux: Le cas des Cecam de Madagascar [Credit trajectories and vulnerability of
rural households: the case of Cecam in Madagascar]'. Autrepart, 44, 157172.
Brett, J. A. (2006) '"We Sacrifice and Eat Less": The Structural Complexities of Microfinance
Participation'. Human Organization, 65(1), 8.

- 34 -

Brown, S., G. Garino, K. Taylor and S. Wheatley Price (2005) 'Debt and Financial Expectations: An
Individual- and Household-Level Analysis'. Economic Inquiry, 43(1), 100120.
Brown, T. M. (1952) 'Habit Persistence and Lags in Consumer Behaviour'. Econometrica, 20(3),
355371.
Canner, G. B. and C.A. Luckett (1991) 'Payment of household debts'. Federal Reserve Bulletin,
77(4), 218.
Chaudhury, I. A. and I. Matin (2002) 'Dimensions and dynamics of microfinance membership
overlap: A micro study from Bangladesh'. Small Enterprise Development, 13(2), 4655.
Christen, M. and R.M. Morgan (2005) 'Keeping Up With the Joneses: Analyzing the Effect of
Income Inequality on Consumer Borrowing'. Quantitative Marketing and Economics, 3(2), 145
173.
Collins, D. (2008) 'Debt and household finance: Evidence from the Financial Diaries'. Development
Southern Africa, 25(4), 469479.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How the
World's Poor Live on $2 a Day, Princeton, NJ: Princeton Univ. Press.
Copestake, J. (2002) 'Unfinished business: The need for more effective microfinance exit
monitoring'. Journal of Microfinance, 4(2), 130.
Copestake, J. (2008) 'Wellbeing and international development: whats new?'. Journal of
International Development, 29(4).
Copestake, J. (2011) 'Well-Being in Development: Comparing Global Designs with Local Views in
Peru'. European Journal of Development Research, 23(1), 94110.
Corbett, J. (1988) 'Famine and household coping strategies'. World Development, 16(9), 10991112.
DeVaney, P. L. (2006) 'Bringing Pro-Consumer Ideals to the Client, A Consumer Protection Guide
for Financial Institutions Serving the Poor'. Monograph Series: Vol. 14, Washington, DC:
ACCION Publications.
Disney, R., S. Bridges and J. Gathergood (2008) 'Drivers of Over-Indebtedness: Report to the
Department for Business, Enterprise and Regulatory Reform'.
Dolan, P., T. Peasgood and M. White (2008) 'Do we really know what makes us happy?A review of
the economic literature on the factors associated with subjective well-being'. Journal of Economic
Psychology, 29(1), 94122.
- 35 -

Duesenberry, J. S. (1949) 'Income, Saving and the Theory of Consumer Behavior, Cambridge, MA:
Harvard University Press.
European Commission (2008) 'Towards a Common Operational European Definition of OverIndebtendess'. OEE Studies, Brussels.
Field, E. and R. Pande (2008) 'Repayment Frequency and Default in Microfinance: Evidence from
India'. Journal of the European Economic Association, 6(2/3), 501509.
Friedman, M. (1957) 'A Theory of the Consumption Function, Princeton: Princeton University Press.
Fudenberg, D. and D.K. Levine (2006) 'A Dual-Self Model of Impulse Control'. The American
Economic Review, 96(5), 14491476.
Godquin, M. (2004) 'Microfinance Repayment Performance in Bangladesh: How to Improve the
Allocation of Loans by MFIs'. World Development, 32(11), 19091926.
Gonzalez, A. (2008) 'Microfinance, Incentives to Repay, and Overindebtedness: Evidence from a
Household Survey in Bolivia. Doctoral thesis'. Ohio State University, Ohio.
Gonzalez, A. (2010) 'Is Microfinance Growing Too Fast?'. MIX Data Brief No. 5.
Gurin, I. (2000) 'Pratiques montaires et financires des femmes en situation de prcarit: Entre
autonomie et dpendance. Doctoral thesis in economics [Monetary and financial practices of
women in precarious situations: Between autonomy and dependance]'. Universit Lumire Lyon
II, Lyon.
Gurin, I., M. Roesch, O. Hlis and Venkatasubramanian (2009a) 'Microfinance, Endettement et
Surendettement: Une tude de cas en Inde du Sud [Microfinance, Indebtedness and Overindebtedness: A case study in Southern India]'. Revue Tiers Monde, 197, 131146.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2009b) 'The Social Meaning of OverIndebtedness and Creditworthiness in the Context of Poor Rural South-Indian Households (Tamil
Nadu), Draft paper presented at the RUME / IEDES workshop on over-indebtedness in Paris,
December 2009.
Gul, F. and W. Pesendorfer (2004) 'Self-Control and the Theory of Consumption'. Econometrica,
72(1), 119158.
Hellmann, T. F., K.C. Murdock and J.E. Stiglitz (2000) 'Liberalization, Moral Hazard in Banking,
and Prudential Regulation: Are Capital Requirements Enough?'. The American Economic Review,
90(1), 147165.
Hudon, M. (2009) 'Should Access to Credit be a Right?'. Journal of Business Ethics, 84, 1728.
- 36 -

Hulme, D. (2000) 'Impact Assessment Methodologies for Microfinance: Theory, Experience and
Better Practice'. World Development, 28(1), 7998.
Kahneman, D. and A. Tversky (1984) 'Choices, values and frames'. American Psychologist, 39(4),
341350.
Kappel, V., A. Krauss and L. Lontzek (2010) 'Over-Indebtedness and Microfinance - Constructing
an Early Warning Index'. Center for Microfinance, University of Zurich, responsAbility, Council
of Microfinance Equity Funds, Triodos Investment Management, Zurich.
Karlan, D. and J. Zinman (2009) 'Expanding Microenterprise Credit Access: Using Randomized
Supply Decisions to Estimate the Impacts in Manila'. Financial Access Initiative, Cambridge,
MA.
Krishnaswamy, K. (2007) 'Competition and Multiple Borrowing in the Indian Microfinance Sector'.
Institute for Financial Management and Research, Centre for Micro Finance Working Paper.
Laibson, D. (1997) 'Golden Eggs and Hyperbolic Discounting'. Quarterly Journal of Economics,
112(2), 443477.
Lea, S. E. G., P. Webley and R.M. Levine (1993) 'The Economic Psychology of Consumer Debt'.
Journal of Economic Psychology, 14(1), 85119.
Lea, S. E. G., P. Webley and C.M. Walker (1995) 'Psychological Factors in Consumer Debt: Money
Management, Economic Socialization, and Credit Use'. Journal of Economic Psychology, 16(4),
681701.
Livingstone, S. M. and P.K. Lunt (1992) 'Predicting personal debt and debt repayment:
Psychological, social and economic determinants'. Journal of Economic Psychology, 13(1), 111
134.
Lusardi, A. and P. Tufano (2009) 'Debt Literacy, Financial Experiences, and Overindebtedness'.
National Bureau of Economic Research NBER Working Paper Series No. 14808, Cambridge,
MA.
Luttmer, E. F. P. (2005) 'Neighbors as Negatives: Relative Earnings and Well-Being'. Quarterly
Journal of Economics, 120(3), 9631002.
Matin, I. (1997) 'Repayment performance of Grameen Bank borrowers: the unzipped' state'. Savings
and Development, 4, 451473.
McIntosh, C. and B. Wydick (2005) 'Competition and microfinance'. Journal of Development
Economics, 78(2), 271298.
- 37 -

Modigliani, F. and R. Brumberg (1980) 'Utility Analysis and the Consumption Function. An
Interpretation of Cross-Section Data. Post Keynesian Economics, edited by Kenneth K. Kurihara,
pp. 388-436 Rutgers University Press 1954' In A. Abel (Ed.): Vol. 2, The Collected Papers of
Franco Modigliani. The Life Cycle Hypothesis of Saving, pp. 79127. Cambridge, MA; London:
MIT Press.
Murray, A. P. (1997) 'Debt and "the" consumer'. Business Economics, 32(2), 4145.
Nio-Zarazua, M. and J. Copestake (2009) 'Financial inclusion, vulnerability, and mental models:
From physical access to effective use of financial services in a low income area of Mexico City'.
Savings and Development, 23(4), 353380.
Notani, A. S. (1997) 'Perceptions of Affordability: Their Role in Predicting Purchase Intent and
Purchase'. Journal of Economic Psychology, 18(5), 525546.
Oshio, T. and M. Kobayashi (2010) 'Income inequality, perceived happiness, and self-rated health:
Evidence from nationwide surveys in Japan'. Social Science & Medicine, 70(9), 13581366.
Paxton, J., D. Graham and T. Cameron (2000) 'Modeling Group Loan Repayment Behavior: New
Insights from Burkina Faso'. Economic Development and Cultural Change, 48(3), 639655.
Pischke, J. D. von (1991) 'Finance at the frontier, Debt capacity and the role of credit in the private
economy, Washington, D.C: World Bank.
Rahman, A. (1999) 'Micro-credit Initiatives for Equitable and Sustainable Development: Who
Pays?'. World Development, 27(1), 6782.
Ranyard, R. and G. Craig (1993) 'Estimating the duration of a flexible loan: The effect of
supplementary information'. Journal of Economic Psychology, 14(2), 317335.
Ranyard, R. and G. Craig (1995) 'Evaluating and budgeting with installment credit: An interview
study'. Journal of Economic Psychology, 16(3), 449467.
Schicks, J. (in press) 'From a supply gap to a demand gap? The risk and consequences of overindebting the underbanked' In J.-P. Gueyie, R. Manos, & J. Yaron (Eds.), Microfinance in
Developing Countries. Issues, Policies and Performance Evaluation. Palgrave Macmillan.
Schreiner, M. (2004) 'Scoring Arrears at a Microlender in Bolivia'. Journal of Microfinance, 6(2),
6588.
Sharma, M. and M. Zeller (1997) 'Repayment performance in group-based credit programs in
Bangladesh: An empirical analysis'. World Development, 25(10), 17311742.

- 38 -

Smets, P. and E. Bhre (2004) 'When Coercion Takes Over: The Limits of Social Capital in
Microfinance Schemes' In H. Lont & O. Hospes (Eds.), Livelihood and microfinance.
Anthropological and sociological perspectives on savings and debt (pp. 215236). Delft: Eburon
Academic Publishers.
Soman, D. and A. Cheema (2002) 'The Effect of Credit on Spending Decisions: The Role of the
Credit Limit and Credibility'. Marketing Science, 21(1), 3253.
Stone, B. and R.V. Maury (2006) 'Indicators of personal financial debt using a multi-disciplinary
behavioral model'. Journal of Economic Psychology, 27(4), 543556.
Strotz, R. H. (1956) 'Myopia and Inconsistency in Dynamic Utility Maximization'. The Review of
Economic Studies, 23(3), 165180.
Thaler, R. (1985) 'Mental Accounting and Consumer Choice'. Marketing Science, 4(3), 199214.
Tiwari, A., A. Khandelwal and M. Ramji (2008) 'How Do Microfinance Clients Understand their
Loans?'. Institute for Financial Management and Research, Centre for Micro Finance Working
Paper Series No. 25.
Viaud, J. and C. Roland-Lvy (2000) 'A positional and representational analysis of consumption.
Households when facing debt and credit'. Journal of Economic Psychology, 21(4), 411432.
Vogelgesang, U. (2003) 'Microfinance in Times of Crisis: The Effects of Competition, Rising
Indebtedness, and Economic Crisis on Repayment Behavior'. World Development, 31(12), 2085
2114.
Webley, P. and E.K. Nyhus (2001) 'Life-Cycle and Disproportionate Routes into Problem Debt'.
British Journal of Psychology, 92, 423446.

- 39 -

Appendix 1: Dimensions of objective over-indebtedness measurements

Appendix 2: Characteristics of the new over-indebtedness definition


Type of choice

Economic

Sociological

Precision

Definition

Indicator

Proxy

Individual

Household

Network of kin

Composition

Single criterion Multiple criteria

Scale

Quantitative

Qualitative

Perspective

Objective

Subjective

Data source

External

Self-reported

Time horizon

Current

Structural

Permanent

Debt condition

Bankruptcy

Default

Arrears

Role of the borrower

Innocent

Unintended

Deliberate1

Level of sacrifice

3 Severity

Legal

Reference unit

2 Method

Categories

Scientific lens
1 Purpose

Dimension of choice

To minimum
More than
existence level expected

1 E.g., Strategic default or fraud

2 Inability to meet unexpected expenses

- 40 -

Other

Aggregate

Imbalance

Liquidity buffer2 No sacrifice

Chapter 2
From a supply gap to a demand gap?
The risk and consequences of over-indebting the underbanked
Jessica Schicks*
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management (SBS-EM)
Universit Libre de Bruxelles (ULB)
In the past, the microfinance industry focused mainly on growth and outreach. Financial
exclusion implied a large supply gap. Recent over-indebtedness in several countries has
shown that this gap can turn into oversupply. The industry urgently requires research on the
magnitude and consequences of this shift. This chapter reveals the broad spectrum of
consequences that over-indebtedness can have for borrowers and other stakeholders, mainly
MFIs. It emphasises that the consequences of over-indebtedness reach far beyond the risk
management concerns that primarily affect MFIs and investors. The chapter also reviews the
existing empirical research on microfinance over-indebtedness and examines the extent of
over-indebtedness in microfinance markets today. It highlights the evidence of overindebtedness in crisis markets and non-crisis markets, especially when markets mature. Only
if the extent of the problem is known and if its effects are properly understood, can the
microfinance industry develop appropriate measures to combat over-indebtedness and adapt
to the challenge of oversupply.

Paper forthcoming in Jean-Pierre Gueyie, Ronny Manos, and Jacob Yaron,


eds., Microfinance in Developing Countries:
Issues, Policies and Performance Evaluation, Macmillan

*The author would like to extend her warmest gratitude to Richard Rosenberg for his
generous support, his work summarising the existing over-indebtedness research, and his
many valuable ideas that inspired her thinking on over-indebtedness. She is grateful for
comments from Marc Labie and from the referees. Additionally, she would like to thank the
Marie Christine Adam Foundation and the German National Merit Foundation for their
financial support.
- 41 -

1. Introduction
The microfinance sector is used to extensive debates about its mission.1 However, in one
regard, the goal has always been clear: microfinance, whether or not it was impact focused
and poverty focused, has always been about extending small-scale financial services to the
underbanked. The aim has been to reach those who are not normally served by the formal
financial system.
This goal had implications for the competitive situation in microfinance markets, especially in
the industrys early years. Microfinance institutions (MFIs) were working in an environment
characterized by a permanent supply gap2. Despite the impressive growth that the industry
achieved over the past decades, enormous figures frequently reminded the microfinance
community of the remaining supply gap in the industry: estimates of the underserved centred
around two or three billion individuals worldwide (Chaia et al., 2009; Karlan and Morduch,
2010).
As a result, the efforts of researchers and practitioners alike focused above all on growth and
outreach. The numbers of clients served often represented the main proxy for an institutions
impact. Efforts concentrated on expanding MFIs into larger institutions that would continue to
serve the poor sustainably in the long term. Under the paradigm of growth as priority number
one, the industry commercialised, arguing that only self-financed institutions will persist and
that only tapping private capital will allow the microfinance industry to reach the billions of
underserved people who still lack access to finance.
Since its beginning, the flagship product of microfinance was microcredit. The industry
developed based on the assumption that the poor can work themselves out of poverty if given
access to the capital that they require to expand their income-generating activities into more
profitable businesses. It assumed that, the poor, if they had any access to credit at all, were
borrowing from usurious moneylenders. Replacing those high cost loans with cheaper funds
from MFIs would make businesses more profitable and would make them grow more rapidly,
generating a better living for the owners families and potentially creating employment
opportunities and contributing to the development of the local economy. The strong demand

1 For background information on these debates see for example Copestake (2007); Labie (2007); Armendriz and Szafarz
(2011).
2 See Armendriz and Labie (2011) for an encompassing volume of microfinance papers that centre on the question of
supply and demand mismatch.

- 42 -

for credit among the poor seemed to confirm the benefits of microfinance and to validate the
design of loans according to the microcredit methodology.
Over the past two years, this picture has changed fundamentally. Some of the above
assumptions have come under increasing criticism, for example the positive impact of
microcredit on poverty (Banerjee et al., 2009; Karlan and Zinman, 2009, 2010; Odell, 2010),
the impact chain through enterprise development (Collins et al., 2009), and the substitution
effect of microfinance replacing informal loans offered at worse conditions (Gurin et al.,
2009; Morvant-Roux, 2009). Behavioural economics highlighted that continued demand for a
product does not that prove it is beneficial. Typical human challenges such as temptation and
the inconsistent evaluation of benefits over time (hyperbolic discounting) can lead to
unreasonable spending, under-saving and over-borrowing (Ashraf et al., 2006; Banerjee and
Duflo, 2007; Banerjee and Mullainathan, 2009; Schicks, in press). The risk of overindebtedness gave rise to an argument against the right to credit that was suggested by
promoters of the microfinance industry (Hudon, 2009).
The most blatant signal that the strong focus of the microfinance industry on expansion might
be flawed, at least in some areas of the world, emerged from a number of countries that
experienced serious crises of over-indebtedness. Like one of the worlds most mature
microfinance markets, Bolivia, in the late 1990s, certain regions of India, Nicaragua, Bosnia,
Morocco, and Pakistan recently experienced severe crises that included problems with overindebtedness. Although each country has its own story and although to varying degrees they
all suffered from the global financial crisis or from political backlashes against high-interest
lending, these crises have one characteristic in common: many borrowers were no longer able
to repay their loans because they had accumulated too much debt. The supply gap had turned
into oversupply.
This does not mean that the previous assumptions governing the development of microfinance
are completely wrong. Nor does it mean that there is no longer a need for microfinance to
grow and expand its outreach. However, it does mean that we need to develop a more
balanced perspective, paying attention to the nuances of the impact of microfinance and
taking the needs of clients in their various contexts more seriously. Instead of putting
institutional growth first, microfinance needs to put the needs of clients first. Institutional
objectives are a means to that end.
Microfinance has started long ago to grow beyond the one-size-fits-all approach of its original
group-lending methodology. Now is the time to take the next step in diversifying approaches
- 43 -

to microfinance, tailoring credit offers to the needs of customers, tailoring the operations of
MFIs and investor expectations to different market environments and speeding the
development of non-credit products. The industry must recognise that not everybody is a
potential borrower and that even microfinance markets can become saturated.
Given that the question of over-indebtedness is relatively new to microfinance and contradicts
earlier thinking about outreach and the impact of credit, there is an urgent need for more
research into the extent, causes and consequences of over-indebtedness. Every actor in the
microfinance industry needs to remain aware in his or her daily work that the positive impact
of credit is not automatic and that lending to the poor can cause harm. If we want to realise
the full potential of microfinance and leverage the power of credit for the poor, we need to
learn how to extend the right products to the right people. This will require careful
experimentation with product features and continually improved assessments of the
creditworthiness of loan applicants. It will require educating clients as well as investors, loan
officers, and the managers of MFIs.
Microfinance scholars can learn from developed country consumer finance research on overindebtedness but will have to test the transferability of these findings to different cultural and
economic contexts and consider the specifics of over-indebtedness in its target group. This
chapter first reveals the broad spectrum of consequences of over-indebtedness for borrowers
and other stakeholders, mainly MFIs. It emphasises that the consequences of overindebtedness reach far beyond the typical risk management concerns that MFI managers and
investors tend to emphasise.
The chapter then reviews the empirical findings on over-indebtedness in the microfinance
industry to date. It shows that there is reason for concern regarding microfinance overindebtedness but that we cannot determine whether such problems are likely to arise in many
microfinance markets or only in a few exceptional cases. Neither can we tell if microfinance
over-indebtedness is a growing concern or reached its peak with the financial crisis, or,
alternatively, will periodically rise and fall as markets enter different phases of development.
Finally, it is unclear to what extent customer experiences of over-indebtedness always
precipitate portfolio problems for MFIs; over-indebtedness at the customer level maybe a
much more wide-spread and permanent challenge than larger delinquency crises.

- 44 -

2. The consequences of over-indebtedness


Over-indebtedness is a complex phenomenon that still lacks a universal definition. Debates
vary substantially between the over-indebtedness of nation states, companies, or consumers.
Despite of the original paradigm of microenterprise finance, the appropriate reference for
microfinance over-indebtedness is that of consumer over-indebtedness - households being
personally liable for their loans and not usually separating household and business cash flows
(Schicks, in press).
Second, even within the category of consumer over-indebtedness, the definitions and
measurements of over-indebtedness vary considerably. Schicks (in press) provides an
overview of the various approaches and distinguishes between institution-centric and
borrower-centric approaches. While the main impact of over-indebtedness on lending
institutions occurs through repayment irregularities, for microborrowers, the consequences of
too much debt tend to manifest in the form of personal struggle and sacrifice before they
reach the stage of delinquency or default. It may be difficult to determine objectively, what
level of repayment difficulty is acceptable. And personal struggle is more difficult to measure
than straightforward concepts such as delinquency. Nevertheless, from a client-protection
perspective, focusing exclusively on outcomes of over-indebtedness such as repayment
irregularities is insufficient because it occludes the full extent of over-indebtedness and can
encourage an incomplete analysis of the causes and consequences of the problem, and of the
solutions that can be used to resolve it.
This chapter therefore uses the borrower-centric approach to define over-indebtedness.
Microborrowers are defined as over-indebted when they continuously struggle to meet their
repayment deadlines and have to make unduly high sacrifices related to their loan obligatons,
that have more than transitory effects.3 Because client protection is intended to avoid
customer suffering as a result of product use, and because suffering is a subjective
phenomenon, this definition is based on the borrowers subjective understanding of whether
they are struggling and whether their level of sacrifices is acceptable.
This section focuses first on the consequences of over-indebtedness for microborrowers.
Based on the above definition these include the struggle to repay as well as the consequences

3 See Schicks (in press) for a detailed discussion of this definition and how to put it into practice as a measurement based on
survey data.

- 45 -

of involuntary delinquency and default. We then pinpoint the impact of over-indebtedness on


lending institutions and other stakeholders.

2.1 The impact of over-indebtedness on microborrowers


In the microfinance literature, research on borrower over-indebtedness is scarce. However,
research has been conducted on the over-indebtedness of private borrowers in developed
countries. The disciplines of economics, sociology and psychology have each contributed
relevant insights to our understanding of the consequences of over-indebtedness for
borrowers. Although most of this research has been performed in environments that are
culturally very different from the typical microfinance setting, the basic findings are likely to
be transferable. The underlying mechanisms of credit and basic needs of human beings should
be similar.4 In any case, the impact of over-indebtedness and the intensity of the various
effects will vary not only between developed and developing countries but also between
microfinance markets and based on individual circumstances. This chapter therefore presents
an overview of the potential consequences of over-indebtedness rather than a list of the effects
of each real-life case of over-indebtedness.
This chapter sorts the consequences of over-indebtedness for microborrowers into three
categories: the material, social and psychological consequences of debt. In each category, the
consequences result mainly from the cost of repayment and from the cost of delinquency or
default. In addition, indirect effects and the general cost of being in debt may play a role.

Material consequences
The most obvious cost of being over-indebted is the need to repay a large amount of debt.
When heavily indebted borrowers still manage to make the payments, they already suffer
from the cost of repayment. Because they have greater debt expenses relative to their income,
over-indebted borrowers live on a lower consumption level than their peers (Betti et al., 2007)
and have lower buffers against shocks (Burton, 2008). According to Brett (2006), in addition
to sacrificing income for instalments, microborrowers in Bolivia rely on cash support from
their families and social networks, take on additional debt, sell assets, reduce the quantity and
quality of their food, and take on additional paid labour to be able to repay their debt on time.
4 The section therefore reports theoretical and empirical findings from developed and developing countries in parallel.
Further research should analyse to what extent non-microfinance findings apply to microfinance settings.

- 46 -

Hardly any of Bretts subjects were able to cover instalments from the returns on their
investment projects. Gonzalez (2008) identifies similar repayment sacrifices in Bolivia, as
well as reductions in human capital investments, indicating that the immediate consequences
of over-indebtedness can lead to lower education levels, a lower income generating capacity
and lower household welfare in the long term. Schicks (2012) provides a detailed account of
the sacrifices of over-indebted microborrowers in Ghana, revealing precisely which sacrifices
are most common, how often they tend to be made, and which sacrifices borrowers perceive
as most severe.5 The sacrifices identified in microfinance markets are quite similar to those
isolated in Canner and Luckett (1991) and Drentea and Lavrakas (2000) in developed
countries.
When, in spite of sacrifices, over-indebted borrowers are no longer able to repay their loans
on time and delay, skip or stop repayments, this might result in some material relief.
However, these actions trigger new consequences: the costs of delinquency and default
(Stearns, 1991; DeVaney and Lytton, 1995). The direct material costs of delinquency include
late fees and higher or longer interest payments on the loan amount. The most significant cost
of default is the seizure of collateral or other assets. Asset seizures in microfinance are
performed either by loan officers or by solidarity groups and can include the seizure of
productive assets or essential items such as land, roofing or cooking pots (Hulme, 2007). Such
an event usually represents a serious loss to the household and in some cases reduces the
borrowers future capacity to generate income, especially when an essential business asset
such as a new sewing machine is seized. In addition, depending on the context, legal fees and
insolvency fees may accrue. Another factor that is more difficult to quantify but highly
relevant for some borrowers is the loss of the client relationship with the MFI. In industrial
countries with credit bureaus, losing ones creditworthiness can even precipitate complete
exclusion from credit, paid labour and other contracts such as apartment rentals, car purchases
or mobile phone plans (Lyons and Fisher, 2006).
Given the cost of delinquency and default, many borrowers try to avoid this stage by any
means. Some employ strategies of multiple borrowing to gain additional time. However, the
material consequences of over-indebtedness can worsen if over-indebted borrowers sink
further into debt by using new loans to help them repay old ones. The costs of delinquency

5 Some of the sacrifices presented in Schicks (2012) refer to the non-material consequences of debt that are discussed later in
this chapter, but many are the result of material pressures such as those discussed by Brett (2006) and Gonzalez (2008).

- 47 -

and default will probably only be postponed to a later stage when they will be even more
severe and debt problems will have become too serious to be solved using simple measures.
Finally, over-indebtedness can have indirect material consequences via non-material channels.
As a result of the psychological difficulties generated by their financial stress, borrowers may
experience a loss of income due to reduced workplace performance (Bagwell, 2001) and
higher workplace absenteeism (Jacobson et al., 1996; Kim et al., 2006). If microborrowers
suffer from their repayment pressure to the extent that they can no longer fulfil their daily
responsibilities, it may destroy their livelihood. Moreover, over-indebtedness can lead to a
loss of business opportunities if business partners withdraw from over-indebted individuals
when their reputation deteriorates (Besley and Coate, 1995; Smets and Bhre, 2004).
Overall, the material consequences of over-indebtedness include a wide range of additional
costs and deprivations for individuals who are already struggling with financial difficulties. In
the worst cases, over-indebtedness actually leads to further impoverishment or, as Gurin et
al. (2011) call it, pauperization through debt.

Social consequences
Societal perceptions regarding indebtedness vary from one culture to the next and over time.
According to research in France by Viaud and Roland-Lvy (2000), depending on the type,
debt is sometimes perceived as necessary but is often considered negative and dangerous.
Being in debt can have harmful effects on ones reputation and can be a source of social
differentiation and shame for borrowers. Roesch and Hlis (2007) find that in Southern
India, even in a microfinance environment in which credit is very common, people consider
debt to be negative and try hard to avoid it. Similarly, Gurin et al. (2011, p. 15) explain that
in Tamil Nadu the term "to be involved in debt" (kadangaran/kadangar) has pejorative moral
connotations, implying surrender, dependence and even servility. Whereas the image of
indebtedness in industrial societies has become much more positive, Dichter (2007) claims
that most microfinance markets today are similar to Western societies in the 19th century: the
mere fact of being in debt is charged with negative symbolism and creates an emotional
burden for debtors and a social stigma against them.
In certain cultures, the social stigma associated with debt may be even stronger for women.
According to the fieldwork of Gurin et al. (2011) in India, a single woman who borrows
from a man outside her family is normally forced to offer sexual favours in return. This being
widely known in society, female debt automatically carries the same stigma as prostitution.
- 48 -

If debt as such carries a social cost, there are further social consequences of struggling with
loan repayment. They range from the burden of asking others for help (Canner and Luckett,
1991) and the pain of peer pressure in solidarity groups, to marital tensions and violence
towards women borrowers within households (Rahman, 1999). It is important to take such
intra-household dynamics into account in analysing credit decisions and their consequences
(Kirchler et al., 2008).
The strongest social mechanisms emerge in cases of delinquency and default. Historical
analysis shows that Western nations had a long history of turning defaulters into servants of
their creditors or imprisoning them (Muldrew, 2000; Burton, 2008). Today, there are reports
of coercive collection practices and even imprisonment in microfinance markets
(Montgomery, 1996; Hulme, 2007). MFIs deliberately employ social pressure as an
enforcement mechanism.6 This social pressure in group lending creates a social cost for
borrowers (Besley and Coate, 1995). Simple measures such as detaining borrowers at group
meetings to force a delinquent member to repay can have severe consequences for the
delinquent (Rahman, 1999). They lose business time, get into trouble with their spouses, or
are forced to leave their children unattended for too long. Another strong mechanism in group
lending is peer pressure. Peer pressure tends to begin with gossiping about delinquent group
members, insulting and humiliating and sometimes threatening them. Eventually peer
pressure may progress as far as physical violence and lead to the destruction of the defaulters
belongings (Montgomery, 1996; Smets and Bhre, 2004; Hulme, 2007).
Even outside solidarity groups, repayment is a question of respect and non-repayment a
source of shame and coercion (Gurin, 2006). The stigma of default can lower self-confidence
and affect ones social network and safety net (Smets and Bhre, 2004; Gurin et al., 2011).
These changes, in turn, may not only affect the delinquent borrower but may also erode the
social trust and mutual support that exists in the community, disturbing informal credit
relationships and social networks (Montgomery, 1996; Smets and Bhre, 2004; Banerjee and
Mullainathan, 2009; Morvant-Roux, 2009).
As suggested by Sen (1999), if capability deprivation is a form of poverty, it is not only the
material costs of over-indebtedness that lead to the further impoverishment of borrowers.
Instead, social consequences such as loss of respect, independence and freedom of choice, are
also mechanisms of impoverishment.
6 See Stearns (1991) for the explicit advice suggesting that MFIs use shame as an enforcement mechanism and make the
names of delinquent borrowers public.

- 49 -

Psychological consequences
To the list of financial and social consequences of over-indebtedness among poor borrowers,
psychologists have added the subtle but very real dimension of psychological struggles
resulting from over-indebtedness. The mere fact of being in debt can already have
psychological costs for the borrower. These consequences tend to worsen when loans become
problem debt. According to empirical research by Brown (1952), who employs data from
Canada, household debt is correlated with the lower psychological well-being of the head of
household, the effect increasing for unsecured debt and for larger amounts of debt. A British
study by Bridges and Disney (2005) finds that debt is related to psychological stress and can
even cause of depression. Similarly, in the USA and Britain, debt is positively associated with
behaviours that indicate low self-control, including smoking, drinking and obesity (Drentea
and Lavrakas, 2000; Webley and Nyhus, 2001). In all three examples, the cause and effect
chains are ambiguous and are likely to be bidirectional; debt may reduce well-being, causing
depression and undermining self-control, which may feed back into higher indebtedness. In
microfinance markets, similar psychological effects may occur.
These psychological mechanisms are expected to be the underlying reasons for the abovementioned effect of debt and financial stress on workplace performance (Bagwell, 2001) and
absenteeism (Jacobson et al., 1996; Kim, Sorhaindo and Garman, 2006). Negative effects on
physical health can also accrue, probably resulting from a combination of psychological stress
with the reduction in self-control and fewer available funds for medical expenses (Drentea
and Lavrakas, 2000). In extreme cases, both in developed countries and in microfinance
settings in developing countries, these consequences of over-indebtedness have pushed
defaulters into crime or suicide (Sarthou-Lajus, 1997; Fouillet, 2006; Dichter, 2007; Dossey,
2007; Hulme, 2007; Burton, 2008).
There are a number of reasons why being debt is so detrimental to psychological well-being.
According to Sarthou-Lajus (1997), debt relationships are asymmetrical and create
perceptions of dependence and a lack of self-sufficiency in the borrower. Not being able to
repay may mean not being able to reconcile ones personal identity and may instil a feeling of
alienation and guilt in debtors. Dichter (2007) cites Nietsche in confirming that debtors
internalise a sense of guilt imposed by the debt, may feel unequal or unworthy, and may
even lose their sense of identity. However, the effect of debt varies from one borrower to the
next. Some manage to rationalise and mitigate their sense of shame, whereas others suffer

- 50 -

from guilt and shame to the extent that it deteriorates their physical and mental health
(Gloukoviezoff, 2008).
Debt and repayment problems alone have serious negative consequences, but the effects are
exacerbated by collection practices. These practices can significantly increase the pressure
and shame that a borrower experiences. The abusive language that lenders sometimes use
with delinquent borrowers can have psychological effects. In Bangladesh and India, lenders
have even been reported to harass and threaten delinquent borrowers and to pressure them to
sell their clothes or even their children (Karnani, 2009). Similarly, solidarity groups can turn
against delinquent borrowers, especially in cases where MFIs apply collective sanctions to all
group members. This has the potential to create severe tensions in groups that exacerbate the
psychological effects of delinquency for the borrowers in trouble, and can lead to group
erosion, reductions in well-being, and further exclusion of the poorest (Marr, in press).
In summary, over-indebtedness can impact many spheres of life and may result in increased
poverty, social exclusion, psychological issues and physical illness among over-indebted
microborrowers. Some of these findings have been transferred to microfinance from a
developed market consumer finance context. Others are specific to particular microfinance
environments. Empirical research should confirm these effects in microfinance markets more
generally. For better comparability, such research should optimally be based on a universally
accepted definition of over-indebtedness or at least on a list of generally accepted indicators.
In the meantime, microfinance practitioners require heightened awareness that debt carries
risks and may have severe negative consequences for borrowers. This insight is not specific to
microfinance, nor does it negate the potential positive impact of microlending on borrowers
who are not over-indebted. However, cultivating an awareness of the downsides of debt may
help lenders to develop a more nuanced and careful approach in determining to whom they
extend what amount of credit and how they do so.

2.2 The impact of microfinance over-indebtedness on other stakeholders


Over-indebtedness is not just a concern for customers and for customer protection efforts. It is
also an important concern for microfinance institutions. The next section analyses the possible
effects of over-indebtedness on MFIs. It then considers how wide-spread over-indebtedness
can affect other stakeholders, including borrowers who are not over-indebted, donors,
investors, support organisations and the general public. This chapter will identify these
- 51 -

second-order effects, but its focus is on the key parties in lending contracts: borrowers and
lenders.

Consequences for lending institutions


For MFIs over-indebtedness is mainly a risk management concern. Especially when
delinquency and default occur, over-indebtedness is a threat to profitability and selfsustainability. In terms of financial cost, the most direct effect of over-indebtedness on MFIs
is that of delinquency deteriorating portfolio quality. It first requires higher loan-loss
provisions and then the write-offs ultimately resulting from default create permanent losses
for lending institutions. This is especially true for microloans without effective collateral.
Additional financial consequences may include higher screening expenses, collection costs,
and other operating costs associated with over-indebted customers (Canner and Luckett, 1991;
DeVaney and Lytton, 1995).
According to microfinance practitioners, the development of portfolio risk in the mostly
uncollateralised loan portfolios of MFIs does not occur in a linear fashion. Instead, it seems
that seemingly moderate delinquency levels spin out of control when losses reach
approximately 5 per cent of an outstanding portfolio or when 10 per cent of the portfolio is
overdue by at least one repayment period (Rosenberg, 1999). Once the level of delinquency
becomes unmanageable, it not only decreases profitability or self-sustainability but threatens
the existence of the lending institution.
With regard to income, the effects begin with the reduction of market potential that occurs
when many clients in a market are over-indebted. MFIs cannot sustainably or profitably lend
to customers who have insufficient repayment capacity. The loss of creditworthiness via a bad
credit history may also exclude some customers from credit for longer than is justified by
their financial situation. In addition, over-indebtedness in microfinance markets tends to
encourage MFIs to be over-careful in their lending decisions and to restrict their own target
markets. Even if over-indebtedness has not yet led to wide-spread defaults because the clients
of the MFI are making sacrifices to repay their loans, this is likely to reduce the pool of
borrowers who will be willing to apply for loans again. Even without repayment irregularities,
over-indebtedness among microborrowers can thus impact the profitability and growth of
lending institutions via negative effects on customer satisfaction.
Ultimately, for MFIs, delinquent interest equals postponed income. Income is also lost
because of the slower rotation of the loan portfolio. In the case of defaults, write-offs
- 52 -

permanently reduce the income earning portfolio (Stearns, 1991). The negative effects of
over-indebtedness are partially offset by late fees and extended interest payments on given
loan amounts. Therefore, one might argue that MFIs may accept over-indebtedness as long as
the cost is mainly born by the clients and doesnt result in poor repayment performance. MFIs
may even choose to accept some delinquency in exchange for late fees or reduced monitoring
and selection costs. However, in most business models used by MFIs, these gains are not
sufficient to compensate for serious repayment problems in a portfolio. Risk management
remains essential for the success of MFIs, and over-indebtedness can ultimately threaten an
MFIs existence.
In addition to the direct financial effects of over-indebtedness on MFIs, over-indebtedness
represents a reputation risk in several respects.7 First, the reputation of the institutions with
customers is at risk. A critical level of defaults will erode repayment discipline among other
borrowers and may trigger strategic defaults. In the Bolivian crisis, debtors even formed
associations whose aim was to release borrowers from their repayment obligation (Gonzalez,
2008). More recent examples from Nicaragua and India indicate the occurrence of similar
phenomena there. Furthermore, there is a risk of adverse selection: good borrowers may leave
an institution with a tainted reputation, while risky borrowers stay. Depositors may withdraw
their savings when rumours of high default levels spread.
Second, over-indebtedness puts the public reputation of an MFI at risk, particularly because of
its social mission and the negative social consequences of over-indebtedness. The general
public might support defaulters, negative media coverage can threaten MFIs, and ultimately,
governments may close branches or institutions, require debt forgiveness, or impose more
restrictive regulations. The current events in Andra Pradesh in India provide a prominent
example. Reputation effects may also alienate donors and investors. All of these reputation
effects can also have repercussions for other MFIs in the market that do not have significant
problems with over-indebtedness. They could ultimately affect the microfinance industry as
such.
Over-indebtedness can also have internal reputation effects. According to Dichter (2007),
collecting bad debt may challenge the organizational identity of an MFI. When collecting bad

7 Measuring reputation risk is difficult for MFIs (and for companies or public institutions in general). However, by regularly
tracking customer satisfaction, encouraging client feedback for example with complaint mechanisms, engaging proactively
and constructively with regulators, competitors and industry associations, and monitoring the media, MFIs should be able to
roughly assess their reputation risk with regards to over-indebtedness.

- 53 -

debt represents a significant share of the MFIs activities, it risks making a bad impression on
good borrowers and lowering staff morale. Especially in institutions with intrinsically
motivated staff who have chosen their work place based on its mission, daily interactions with
over-indebted clients may lower staff motivation and encourage attrition. Even among loan
officers who consider identification with a social mission a lower priority and see themselves
foremost as professional bankers , collecting bad debts can be difficult and tedious and offers
few rewards. Thus, the risk of low motivation and high staff turn-over remains.

Consequences on other stakeholders


Just as the consequences of over-indebtedness may spill over from one MFI to others in the
same market, over-indebtedness can also have repercussions for other stakeholders. The
damage to the reputation of MFIs in some markets can create reputational damage for the
microfinance industry as a whole, including its funders, donors and investors. Higher default
in microfinance portfolios reduces investments returns. In addition to creating direct costs for
microfinance investors and the sectors support industry, the effects of over-indebtedness on
reputation and on investment returns could reduce the availability of capital for microfinance
locally or even globally.
In a specific market, when over-indebtedness spreads, borrowers who default on one loan may
also default on their other loans, potentially including loans or bills from other parties besides
institutional lenders. Such defaults could reduce the stability of other MFIs or institutions that
are not even directly involved in microlending. The instability of a few institutions can have
systemic consequences, and in contexts in which the MFI sector is sufficiently large or is
closely connected with the formal banking industry, such instability can potentially reduce the
stability of the financial system more generally.
Furthermore, if over-indebtedness weakens lending institutions, this affects even customers
who are not over-indebted and even non-borrowing clients. MFIs may set higher interests and
less favourable terms and conditions. In extreme cases the instability may eliminate the MFIs
product and service offer and put savings at risk (Stearns, 1991; DeVaney and Lytton, 1995;
Burton, 2008). Additionally, the higher risk aversion of lenders in markets with overindebtedness is likely to restrict the supply of credit even to good borrowers (Canner and
Luckett, 1991). To the extent that over-indebtedness results from poor evaluations of
repayment capacity and poor borrower selection, it channels credit (and, for non-sustainable
MFIs, subsidies) to unproductive uses and activities with negative welfare effects. Thus the
- 54 -

effects of over-indebtedness spread far beyond over-indebted microborrowers and their


lenders.
In conclusion, the range of direct and indirect effects of over-indebtedness is broad and
reaches far beyond individual MFIs and over-indebted borrowers. Some of the above effects
are unlikely to occur except in extreme cases. Others are common effects of overindebtedness that occur quite frequently there is simply a lack of awareness of debt as their
cause, especially if debt is one possible cause among many.
The main message of this section is therefore twofold. First, researchers and practitioners
should be aware that the full spectrum of effects of over-indebtedness is much broader than
the simple effects that are usually covered in public debate. It is particularly important to
acknowledge the risk of tipping points, in which a previously manageable level of overindebtedness turns into a serious crisis that affects other borrowers, institutions and external
stakeholders. Second, even on an individual level, the phenomenon of over-indebtedness is of
most severe consequences. Again, the public debate tends to focus on certain elements of the
issue, while ignoring others. The most common reduction is that of over-emphasising the
institutional perspective of risk management and not taking borrower experiences into
account. In addition, even when addressing borrower concerns, scholars and practitioners
rarely consider the full spectrum of material, social and psychological damage that overindebtedness may cause.

- 55 -

3. Empirical research on over-indebtedness in microfinance to


date
In the current stage of the microfinance industry, many stakeholders are alert to the threat of
over-indebtedness, but the seriousness of the issue is still unclear. One missing component is
a comprehensive understanding of the consequences of over-indebtedness. The other is
empirical evidence of the extent of over-indebtedness in microfinance markets. The next
section will review what we know about the extent of microfinance over-indebtedness to
date.8 It will review in a very brief format the main studies that have analysed overindebtedness situations in both crisis and non-crisis markets in the microfinance industry.
Instead of analysing the respective cases and their differences and similarities in detail, it will
summarize them on a high level to distil the key learnings that we can draw from them with
regards to over-indebtedness in microfinance.
For most of the industrys growth years, measuring over-indebtedness has not been a priority
among microfinance scholars. As this chapter has indicated, the main reason focus was the
significant supply gap created by the lack of access to finance among the poor. The possibility
of market saturation seemed remote. However, the industrys first crisis that came with overindebtedness in microfinance actually occurred in the late 1990s in Bolivia, when a
microfinance market had first attained a relevant size within a national financial system and
began to overlap with the consumer finance industry.

The first crisis with over-indebtedness in microfinance


The first large repayment crisis in microfinance9 occurred in 1999 and 2000 in Bolivia,
triggered not only by strong growth of microfinance institutions but also by an economic
recession in Bolivia and the entry of consumer lenders as new competitors. The new entrants
introduced lending practices that differed substantially from traditional microfinance and

8 Also see Schicks and Rosenberg (2011) who review the empirical findings on over-indebtedness from a slightly different
angle.
9 According to this papers customer-protection oriented definition of over-indebtedness, delinquency and default are not
automatically equal to over-indebtedness, However, this literature review considers delinquency crises signs of overindebtedness as long as the existing research does not attribute the repayment problems to wilful default but rather to an
inability of borrowers to pay. On a detailed level, empirical findings may differ according to the definitions used, but for the
purposes of this review, research that is based on different definitions of over-indebtedness provides a sufficient indication of
the general prevalence of over-indebtedness in microfinance markets.

- 56 -

implied a higher risk of over-indebtedness for their borrowers (Rhyne, 2001; Gonzlez-Vega
and Villafani-Ibarnegaray, 2011). Initially, both lenders and the regulator struggled with their
reactions to the new competitive situation. It became common for microborrowers to borrow
from multiple lenders in parallel, often exceeding their repayment capacity. Working with
data from Caja Los Andes, Vogelgesang (2003) finds significant delinquency in 2000, with
27 per cent of repayments being made too late. Arguing from the motivation of developing
stronger repayment incentives, Vogelgesang raised many of todays most recent
recommendations nearly ten years ago: she explains that the microfinance industry should
adapt to increasingly saturated markets by means of well thought-out regulations, functioning
credit bureaus, and risk-adjusted growth strategies and lending policies among MFIs. As
much as the crisis resulted from the complex interplay of a number of influence factors,
implicitly this early analysis suggests that as microfinance markets mature, over-indebtedness
and defaults are likely to occur more widely unless the industry adopts appropriate
preventative measures. It also shows that saturation is a relative phenomenon. On the national
level, there remained ample opportunity for microfinance in Bolivia, and the industry has
continued to grow and develop successfully since the 1999 crisis (Gonzlez-Vega and
Villafani-Ibarnegaray, 2011). However, the segments that were mainly targeted by MFIs and
consumer lenders at the time were already overbanked.
In addition to illustrating that microfinance markets can saturate and may do so faster than
expected, the Bolivian crisis revealed another fundamental message about over-indebtedness:
more clients may be over-indebted than are paying late. Applying a very broad definition of
over-indebtedness to Bolivian household survey data from 1997 to 2000, Gonzalez (2008)
identifies 85 per cent of all microborrowers in his sample as over-indebted. These include
borrowers who are strategically unwilling to repay and those who incur just one costly action
to repay which they had not anticipated at the time of taking the loan (e.g., working more than
under their ordinary schedule). While not everyone may agree that some extra effort or even
wilful default are signs of over-indebtedness (Gonzalez does not distinguish between
reasonable efforts and intolerable efforts by debtors), the most important contribution of the
study is that it recognises costly borrower actions: the strong repayment performance of
microfinance borrowers is often a result of hardship, as borrowers sometimes go to great
lengths to meet their repayment deadlines.

- 57 -

Concerns about debt problems for customers


Following the Bolivian crisis, over-indebtedness became less of a focus again and the
industrys focus continued to be on growth. However, researchers increasingly pointed to the
troubles that loans can create for microborrowers. They recognised that on a smaller scale,
debt problems even occur to individual borrowers in non-crisis markets.
Already in 1995 and 1996 in Bangladesh and Sri Lanka, empirical research indicated the costs
of the repayment incentives and social collateral used in microfinance group lending,
particularly peer pressure, aggressive collection practices, and reputational or material loss for
defaulters (Besley and Coate, 1995; Montgomery, 1996). In 2004, Smets and Bhre analysed
social capital and coercion in Indian and South African microlending. More recent research
highlights the severe social implications that debt can have for borrowers and points out that
microfinance may contribute to poor borrowers sliding into debt problems (Dichter, 2007;
Hulme, 2007). An ethnographic analysis from Bolivia in 2006, long after the crisis, reveals
that most microenterprises in their local context do not earn sufficient returns for borrowers to
repay their loans (Brett, 2006). Instead, borrowers struggle to repay and make severe
sacrifices.
Together with over-indebtedness research conducted in 2003/04 in South Africa that is not
specific to typical microfinance contexts (Collins, 2008), the above studies increased
awareness of the downsides of credit. The focus of the industry, however, remained on the
large supply gap that remained and the increase in outreach required to address this gap.
Scholars did not connect the messages about maturing markets with the findings about
borrowers struggling to repay their debt.

Delinquency after the global financial crisis


In the years 2008 and 2009 the global financial crisis hit the worlds capital markets and
slowed down real economies. Although microfinance markets are usually resilient to
macroeconomic cycles and shocks, a number of microfinance markets experienced severe
crises that were marked by over-indebtedness, high delinquency and, in India, by a series of
borrower suicides. In most cases, the financial crisis was probably the immediate but not the
underlying cause of the problems. Three empirical studies shed light on the extent of overindebtedness in such markets.

- 58 -

The greatest over-indebtedness crisis for the microfinance industry to date, especially given
its political consequences, took place in Andhra Pradesh in India. While these very recent
events have not yet been researched, Krishnaswamy and Ponce (2010) provide an analysis of
the mass defaults in the neighbouring state of Karnataka that preceded the larger crisis in
Andhra Pradesh. A preliminary report from the study offers strong evidence of the spillover
risks of microfinance over-indebtedness. It shows that significant repayment stress among a
share of MFI customers precipitated the mass defaults. 21 per cent of borrowers in default
towns reported their repayments as a burden compared to only 3 per cent in non-default
towns. For specific sacrifices such as skipping meals, the percentages are even higher and
display the same divide between crisis and non-crisis towns. However, not all of the
defaulters were over-indebted. More than 90 per cent of defaulters stated that they refused to
pay even though they would have had the means to do so. Instead, the over-indebtedness of
some triggered a ban on MFI repayments by religious organisations and thus the default of a
much larger group of borrowers. Once the defaults started to spread, second-level effects
made even more borrowers default because peers in their group or centre had defaulted as
well. Overall, the practitioner-oriented study by Krishnaswamy and Ponce (2010) attributes
this crisis to a mix of institutional factors such as weakened processes of lending institutions
(e.g., reductions on personal visits of borrowers and in monitoring of loan use), more
vulnerable borrowers, and the religious influence factors of a Muslim ban on repayment.
Although the exact ratio of over-indebted defaulters to opportunistic defaulters remains
unclear, Krishnaswamy and Ponce (2010) reinforce the existing evidence that there are critical
levels of over-indebtedness in certain microfinance markets. Their findings underline the
difference between over-indebtedness from a customer protection perspective and actual
repayment behaviour: not only can there be more borrowers struggling with their debt than
are actually defaulting, but there can also be more defaults than over-indebtedness.
In Bosnia and Herzegovina, another crisis country, a study from 2009 uses MFI and credit
bureau data for 1000 microborrowers and a smaller sample of personal interviews (Maurer
and Pytkowska, 2011). The study defines over-indebtedness as a net debt-service ratio10
above 100 per cent and finds that 17 per cent of microborrowers are over-indebted. Another
11 per cent are at risk of over-indebtedness with a net debt-to-income ratio of 75-100 per cent.
Over-indebtedness especially affects clients with lower incomes and those with several loans

10 Monthly debt instalments divided by monthly household income net of other expenses. This definition basically declares a
non-saver as over-indebted and does not determine if the current level of expenses is exaggerated, reasonable or precarious.

- 59 -

outstanding. Nearly all borrowers have multiple loans simultaneously; almost half of
borrowers have five or more loans. 27 per cent of all microborrowers are already delinquent.
Delinquency only partially overlaps with over-indebtedness defined as a net debt-to-income
ratio above 100 per cent.
A third study is based on 2010/11 data on 1,200 microborrowers from four MFIs and three
banks in Kosovo (Pytkowska and Spannuth, 2012). A smaller sample of 778 clients was
personally interviewed. The situation is less alarming than in Bosnia but still concerning, 25
per cent of clients having more than one loan, with five as the maximum. 8.4 per cent of loans
are being paid more than 30 days late. Multiple borrowers are in arrears more frequently. Half
of the borrowers feel that loan repayments are a major burden. According to the definition of
over-indebtedness as a 100 per cent net debt-service ratio, 25 per cent of borrowers are
seriously over-indebted or at risk. Again, multiple borrowers are over-indebted more often.11

Over-indebtedness in non-crisis markets


The issues of the poor with debt do not seem to be limited to markets with explicit default
crises. Instead, based on data from 2005 to 2009, there is also evidence of over-indebtedness
in the third largest Southern state of India, Tamil Nadu. Although there is no repayment crisis,
Gurin et al. (2011) show that 91 per cent of households in their sample villages have debt.
On average, a households outstanding debt amounts to its total income for one year, and its
monthly repayments require half of its monthly income. Gurin et al. (2011) focus their more
detailed analysis on the 20 per cent of households with the highest debt load.

Using

impoverishment through debt as their definition of over-indebtedness, they find that all of
these households are over-indebted. For 19 per cent of them (4 per cent of the total sample),
this over-indebtedness is probably only transitory; for another 38 per cent (8 per cent of total
sample) there seemed no hope to escape debt in the near future, and the remaining 43 per cent
(9 per cent of total sample) had reached a stage of extreme dependence on others and were
unable to keep their repayments up.
A study by Grammling in Ghana in 2009 also finds signs of over-indebtedness in a non-crisis
market in Africa. The study employs various research methodologies that differ in their

11 Some of the studies reviewed in this section are oriented towards practitioners. Not all studies report statistical
significance.

- 60 -

rigorousness to several samples of borrowers12 and indicates that at least half of microfinance
borrowers have more than one loan outstanding. Subjective perceptions of borrowers indicate
that respondents in local markets believe 46 per cent of borrowers in their area to face serious
repayment problems (19 per cent when interviewed in branches). Regarding their own
situation, 60 per cent of respondents say that their expenses temporarily exceeded their
incomes and 15 per cent admit a permanent income gap. The study considers borrowers to be
over-indebted if they are de-capitalising and if their business assets no longer exceed their
liabilities. The results indicate that 12 per cent of borrowers are over-indebted and that
another 16 per cent are at risk. Over-indebted borrowers according to this definition are more
likely to have multiple loans in parallel and are more likely to be delinquent. Overall, 14 per
cent of borrowers were delinquent.
While the results of the Ghana study provoke concerns about an approaching overindebtedness crisis, our recent work in Ghana indicates that the situation may have partially
improved (Schicks, 2012). The rates of multiple borrowing and delinquency in the 2010
sample among the countrys top MFIs are acceptable. At the same time, the personal
repayment situation of many borrowers remains challenging, and over-indebtedness according
to the sacrifice-based definition of this paper remains at 30%. Sound repayment records do
therefore not guarantee the absence of over-indebtedness.

So what do we know to date?


In summary, the existing empirical studies on over-indebtedness show that over-indebtedness
is a reoccurring phenomenon in microfinance markets, sometimes at worrisome levels.
Despite the large numbers of underbanked poor, some microfinance markets have produced at
least local oversupply. There have been cases of open crises at several points in time during
the history of the industry, especially when (sub-national) markets have matured and the
economic environment has been weak. Under these circumstances, over-indebtedness can
lead to crises of delinquency and default that, because of strategic defaults, exceed the scope
of the original over-indebtedness problem. Conversely, some markets may not experience an
open crisis but may include a significant share of microborrowers who are struggling with
their debt. Finally, a certain share of over-indebted borrowers might be an unavoidable
12 Our summary of this study reports the findings obtained using the various approaches without identifying the
methodologies used. The results are not perfectly comparable. Refer to the original paper for a better understanding of the
reliability of these findings and of their shortcomings.

- 61 -

phenomenon in all lending markets. As the above studies on the downsides of social capital in
Asia or on microborrowers returns in Latin America suggest, microfinance markets might
exhibit the downsides of debt burdens even if a large supply gap persists.
The empirical research on microfinance over-indebtedness to date remains too limited to
determine how widespread the phenomenon is in microfinance markets in general. The
sample of studies reviewed in this chapter is skewed because most of them were conducted
because of ex-ante concerns about over-indebtedness in their respective markets. Further
research needs to analyse the extent of over-indebtedness in microfinance markets more
generally. Also, further research should seek to understand the dynamics that led to each of
the crises that the microfinance industry has experienced so far. Understanding the
management dynamics, competitive environments and complex dynamics of financial distress
that lead to crises in the past will help researchers to assess whether the 2008/09 financial
crisis represented the peak of over-indebtedness, whether over-indebtedness will increase as
the industry matures, or whether it peaks at a certain stage in a markets life cycle but
decreases as MFIs adapt to competition. Alternatively, instead of each market going through
the same cycle at a different time, the industry may collectively adapt to avoid future crises
and to improve the debt experiences of microborrowers in general.

- 62 -

4. Conclusion
Globally, the microfinance industry still faces a large supply gap. However, in an increasing
number of sub-markets, situations of oversupply emerge or a mix of simultaneous oversupply
for certain customers and undersupply for others can develop. These cases of oversupply can
be detrimental to microfinance customers and lending institutions alike.
This chapter sheds light on the over-indebtedness risks inherent in microfinance lending.
Risks can be measured in terms of magnitude and likelihood. The first part of the chapter
therefore analyses the magnitude of the over-indebtedness problem. It reveals the broad
spectrum of potential consequences of over-indebtedness for borrowers and lending
institutions.
The chapter suggests that borrowers are likely to suffer from over-indebtedness on a material
level, experiencing further impoverishment in extreme cases. They may also experience social
consequences such as social stigma, peer pressure, domination in the household, shame and
coercion and the loss of their social networks. Based on Sens concept of development as
freedom (1999), this loss of respect and support, which leads to a reduction in the borrowers
personal freedom of choice and ability to determine his life, could also be interpreted as a
form of impoverishment. Finally, borrowers can experience psychological effects of overindebtedness, which range from alienation and guilt to psychological stress. These effects
may cause depression or deteriorate physical health. The consequences of over-indebtedness
thus extend far beyond the challenges to portfolio quality that are important from the
suppliers perspective of risk management. However, insofar as the findings are specific to
non-microfinance research settings, future research should confirm their transferability to
other cultures and to the microfinance context.
In a second step, this section identifies the consequences of over-indebtedness for lending
institutions. In addition to the obvious costs of loan loss provisions and write-offs, overindebtedness can result in increased operating costs, reductions in market size, loss of
customer satisfaction, postponed and lost income, and internal and external reputation effects.
There can also be spillover effects on other MFIs or even other stakeholders, possibly
endangering an entire market.
The second part of the chapter focuses on the likelihood of over-indebtedness in microfinance
markets and reviews what we know about the extent of over-indebtedness to date. It points
out that over-indebtedness may occur in maturing markets, at least on a sub-national level and
for a period of adaptation to the changes in the competitive environment. Bolivia and several
- 63 -

regional markets in India seem to be such examples where, among other influence factors, the
saturation that came with the approaching maturity of certain market segments contributed to
the emergence of over-indebtedness among microborrowers. This section also shows that
crises of over-indebtedness may arise in connection with other economic crises in a country.
For example, this was probably the case in Bosnia, even if the global financial crisis is
unlikely to have been the only driver of this over-indebtedness wave. Finally, a certain level
of over-indebtedness may be prevalent in any debt market. For example, studies on non-crisis
markets in Tamil Nadu in India and in Ghana and research on the mechanisms of social
collateral in Bangladesh and Sri Lanka both indicate that some borrowers suffer from their
debt, even if there is no crisis in the market.
More research must be conducted if we are to understand how many microfinance markets are
currently affected by various levels of over-indebtedness and what the likely future trends are.
We need to understand if over-indebtedness is generally increasing, or has already peaked due
to the global financial crisis and the current stage of development of the industry, or whether
it will continue to increase and decrease in different regions in line with the development of
these microfinance markets. This analysis requires a systematic understanding of the causes of
over-indebtedness and the magnitude of the various factors that influence it, from market
structure to institutional environments and to economic cycles.13 Moreover, it is necessary to
disentangle the concepts of over-indebtedness and delinquency crises. The two can go hand in
hand but are not necessarily equivalent, as the empirical review in this paper shows. Customer
experiences with over-indebtedness may not always be related to portfolio problems for MFIs
if the borrowers who are struggling with too much debt are still repaying. Some customers
may also not repay even though they are not over-indebted - for example during crises and in
the case of spillover delinquency from clients who are over-indebted to those who are not.
In spite of the many unanswered questions and the urgent need for research on overindebtedness in microfinance, existing research clearly indicates that the actors in the
microfinance industry need to take the possibility of oversupply and the downsides of debt for
borrowers into account. Such awareness is likely to improve the products offered to
microfinance customers and may help to regulate the equilibrium of under- and oversupply.

13 See Schicks (in press) for a conceptual framework of the causes of over-indebtedness. Future research will need to
confirm to what extent the factors identified in consumer finance markets can be transferred to microfinance markets and
what the relative importance of the different influence factors is.

- 64 -

Such awareness will also lead to the necessary experimentation with product features14 and to
the development of codes of conduct and customer protection mechanisms. An assessment of
the risk of over-indebtedness could become part of the evaluation of repayment capacity at the
borrower level, as well as of the governance of MFIs and the due diligence of investors on an
institutional and a sector level. The microfinance industry will continue to develop and, with
sufficient attention to the customer perspective and to over-indebtedness, will increasingly
extend the right products to the right people.

14 Hamp and Laureti (2011) analyse the possibility of enhancing product flexibility to better meet the needs of clients. They
show that flexibility can be combined with the necessary repayment incentives but in turn may increase cost.

- 65 -

References
Armendriz, B. and A. Szafarz (2011) 'On Mission Drift In Microfinance Institutions' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 341366). London,
Singapore: World Scientific Publishing.
Armendriz, B.; M. Labie (Eds.) (2011) 'The Handbook of Microfinance, London, Singapore:
World Scientific Publishing.
Ashraf, N., D. Karlan and W. Yin (2006) 'Tying Odysseus to the Mast: Evidence from a
Commitment Savings Product in the Philippines'. Quarterly Journal of Economics, 121(2),
635667.
Bagwell, D. C. (2001) 'Hierarchical regression analysis of work outcomes with personal and
financial factors'. Writings of the Western Region Home Management Family Economics
Educators, 4554.
Banerjee, A. and S. Mullainathan (2009) 'The shape of temptation: Implications for the
economic lives of the poor'. Harvard University & MIT.
Banerjee, A. V. and E. Duflo (2007) 'The Economic Lives of the Poor'. Journal of Economic
Perspectives, 21(1), 141167.
Banerjee, A. V., E. Duflo, R. Glennerster and C. Kinnan (2009) 'The miracle of microfinance?
Evidence from a randomized evaluation'. MIT Poverty Action Lab, Cambridge, MA.
Besley, T. and S. Coate (1995) 'Group lending, repayment incentives and social collateral'.
Journal of Development Economics, 46(1), 118.
Betti, G., N. Dourmashkin, M.C. Rossi and Y.P. Yin (2007) 'Consumer over-indebtedness in
the EU: measurement and characteristics'. Journal of Economic Studies, 34(2), 136156.
Brett, J. A. (2006) '"We Sacrifice and Eat Less": The Structural Complexities of Microfinance
Participation'. Human Organization, 65(1), 8.
Bridges, S. and R. Disney (2005) 'Debt and depression'. Centre for Finance and Credit
Markets, School of Economics, Nottingham Working Paper No. 06/02.
Brown, T. M. (1952) 'Habit Persistence and Lags in Consumer Behaviour'. Econometrica,
20(3), 355371.
Burton, D. (2008) 'Credit and Consumer Society, Oxon: Routledge.

- 66 -

Canner, G. B. and C.A. Luckett (1991) 'Payment of household debts'. Federal Reserve
Bulletin, 77(4), 218.
Chaia, A., A. Dalal, T. Goland, M.J. Gonzalez, J. Morduch and R. Schiff (2009) 'Half the
World is Unbanked'. Financial Access Initiative Framing Note, New York.
Collins, D. (2008) 'Debt and household finance: Evidence from the Financial Diaries'.
Development Southern Africa, 25(4), 469479.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How
the World's Poor Live on $2 a Day, Princeton, NJ: Princeton Univ. Press.
Copestake, J. (2007) 'Mainstreaming microfinance: social performance management or
mission drift?'. World Development, 35(10), 17211738.
DeVaney, S. A. and R.H. Lytton (1995) 'Household Insolvency: A Review of Household Debt
Repayment, Delinquency, and Bankruptcy'. Financial Services Review, 4(2), 137.
Dichter, T. (2007) 'Can microcredit make an already slippery slope more slippier?: Some
lessons from the social meaning of debt' In T. Dichter & M. Harper (Eds.), What's wrong
with Micofinance? (pp. 918). Warwickshire: Intermediate Technology Publications.
Dossey, L. (2007) 'Debt and Health'. Explore: The Journal of Science and Healing, 3(2), 8390.
Drentea, P. and P.J. Lavrakas (2000) 'Over the Limit: The Association Among Health, Race,
and Debt'. Social Science & Medicine, 50(4), 517529.
Fouillet, C. (2006) 'La microfinanceserait-elle devenue folle? Crise en Andrha Pradesh'.
Gret/Cirad, Espace Finance Bulletin d'Information du Mardi No. 24-04-2006, Paris.
Gloukoviezoff, G. (2008) 'De lexclusion linclusion bancaire des particuliers en France:
Entre ncessit sociale et contrainte de rentabilit. Doctoral thesis'. Universit Lumire
Lyon 2, Lyon.
Gonzalez, A. (2008) 'Microfinance, Incentives to Repay, and Overindebtedness: Evidence
from a Household Survey in Bolivia. Doctoral thesis'. Ohio State University, Ohio.
Gonzlez-Vega, C. and M. Villafani-Ibarnegaray (2011) 'Microfinance in Bolivia: Foundation
of the Growth, Outreach and Stability of the Financial System' In B. Armendriz & M.
Labie (Eds.), The Handbook of Microfinance (pp. 203251). London, Singapore: World
Scientific Publishing.

- 67 -

Grammling, M. (2009) 'Cross-borrowing and Over-indebtedness in Ghana: Empirical


Evidence from Microfinance Clientele and Small Enterprises. Technical draft for
discussion'. ProCredit Holding, Frankfurt am Main.
Gurin, I. (2006) 'Women and Money: Lessons from Senegal'. Development and Change,
37(3), 549570.
Gurin, I., M. Roesch, O. Hlis and Venkatasubramanian (2009) 'Microfinance, Endettement
et Surendettement: Une tude de cas en Inde du Sud [Microfinance, Indebtedness and
Over-indebtedness: A case study in Southern India]'. Revue Tiers Monde, 197, 131146.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2011) 'The social meaning of
over-indebtedness and creditworthiness in the context of poor rural South India households
(Tamil Nadu)'. RUME Working Paper No. 2011-1, Paris.
Hamp, M. and C. Laureti (2011) 'Balancing flexibility and discipline in microfinance:
Innovative financial products that benefit clients and service providers'. Centre for
European Research in Microfinance, Universit Libre de Bruxelles CEB Working Paper
No. 11/044.
Hudon, M. (2009) 'Should Access to Credit be a Right?'. Journal of Business Ethics, 84, 1728.
Hulme, D. (2007) 'Is microdebt good for poor people? A note on the dark side of
microfinance' In T. Dichter & M. Harper (Eds.), What's wrong with Micofinance? (pp. 19
22). Warwickshire: Intermediate Technology Publications.
Jacobson, B. H., S.G. Aldana, R.Z. Goetzel, K.D. Vardell, T.B. Adams and R.J. Pietras
(1996) 'The relationship between perceived stress and self-reported illness-related
absenteeism'. American Journal of Health Promotion, 11(1), 5461.
Karlan, D. and J. Morduch (2010) 'Access to Finance' In D. Rodrik & M. R. Rosenzweig
(Eds.), Handbook in economics: Vol. 5. Handbook of Development Economics (pp. 4703
4784). Amsterdam: Elsevier.
Karlan, D. and J. Zinman (2009) 'Expanding Microenterprise Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts in Manila'. Financial Access
Initiative, Cambridge, MA.
Karlan, D. and J. Zinman (2010) 'Expanding Credit Access: Using Randomized Supply
Decisions to Estimate the Impacts'. Review of Financial Studies, 23(1), 433464.

- 68 -

Karnani, A. (2009) 'Regulate Microcredit to Protect Borrowers'. University of Michigan Ross


School of Business Working Paper No. 1133.
Kim, J., B. Sorhaindo and E.T. Garman (2006) 'Relationship between Financial Stress and
Workplace Absenteeism of Credit Counseling Clients'. Journal of Family and Economic
Issues, 27(3), 458478.
Kirchler, E., E. Hoelzl and B. Kamleitner (2008) 'Spending and credit use in the private
household'. Journal of Socio-Economics, 37(2), 519532.
Krishnaswamy, K. and A. Ponce (2010) 'An analysis of the Karnataka mass defaults'.
Working Paper.
Labie, M. (2007) 'Rflexions prliminaires pour une approche thique de la gestion des
organisations en microfinance'. Ethics and Economics, 5(1), 18.
Lyons, A. C. and J. Fisher (2006) 'Gender Differences in Debt Repayment Problems after
Divorce'. Journal of Consumer Affairs, 40(2), 324346.
Marr, A. (in press) 'Impact of group dynamics on microfinance sustainability: the case of
Peru'. Journal of Development Effectiveness, (Special Issue).
Maurer, K. and J. Pytkowska (2011) 'Indebtedness of Microcredit Clients in Bosnia and
Herzegovina: Results from a comprehensive field study'. EFSE Development Facility.
Montgomery, R. (1996) 'Disciplining or protecting the poor? Avoiding the social costs of peer
pressure in micro-credit schemes'. Journal of International Development, 8(2), 289305.
Morvant-Roux, S. (2009) 'Accs au microcrdit et continuit des dynamiques d'endettement
au Mexique: Combiner l'anthropologie conomique avec l'conomtrie [Access to
microcredit and continuity of indebtedness dynamics in Mexico: Combining economic
anthropology with econometrics]'. Revue Tiers Monde, 196, 109130.
Muldrew, C. (2000) 'The economy of obligation, The culture of credit and social relations in
early modern England'. ([Nachdr.]). Early modern history, Basingstoke: Macmillan [u.a.].
Odell, K. (2010) 'Measuring the Impact of Microfinance: Taking Another Look'. Grameen
Foundation USA Grameen Foundation USA Publication Series, Washington, DC.
Pytkowska, J. and S. Spannuth (2012) 'Indebtedness of Microcredit Clients in Kosovo:
Results from a comprehensive field study'. Finance in Motion; EFSE.
Rahman, A. (1999) 'Micro-credit Initiatives for Equitable and Sustainable Development: Who
Pays?'. World Development, 27(1), 6782.
- 69 -

Rhyne, E. (2001) 'Mainstreaming microfinance, How lending to the poor began, grew, and
came of age in Bolivia, Bloomfield, CT: Kumarian Press.
Roesch, M. and O. Hlis (2007) 'La microfinance: outil de gestion du risque ou de mise en
danger par surendettement? [Microfinance: A tool for risk management or for the creation
of over-indebtendness risks]'. Autrepart, 44, 199-140.
Rosenberg, R. (1999) 'Measuring Microcredit Delinquency: Ratios can be harmful to your
health'. CGAP CGAP Occasional Paper No. 3, Washington DC.
Sarthou-Lajus, N. (1997) 'L'thique de la Dette, Paris: Presses Universitaires de France.
Schicks, J. (2012) 'The over-indebtedness of microborrowers in Ghana - An empirical study
from a customer protection perspective'. Center for Financial Inclusion Center for
Financial Inclusion Publication No. 15, Washington DC.
Schicks, J. (in press) 'Microfinance Over-Indebtedness: Understanding its Drivers and
Challenging the Common Myths'. Oxford Development Studies,
Schicks, J. and R. Rosenberg (2011) 'Too Much Microcredit? A Survey of the Evidence on
Over-Indebtedness'. CGAP Occasional Paper 19, Washington DC.
Sen, A. (1999) 'Development as Freedom: Oxford University Press.
Smets, P. and E. Bhre (2004) 'When Coercion Takes Over: The Limits of Social Capital in
Microfinance Schemes' In H. Lont & O. Hospes (Eds.), Livelihood and microfinance.
Anthropological and sociological perspectives on savings and debt (pp. 215236). Delft:
Eburon Academic Publishers.
Stearns, K. (1991) 'The Hidden Beast: Delinquency in Microenterprise Credit Programs'.
ACCION International Discussion Paper Series No. 5, Washington DC.
Viaud, J. and C. Roland-Lvy (2000) 'A positional and representational analysis of
consumption. Households when facing debt and credit'. Journal of Economic Psychology,
21(4), 411432.
Vogelgesang, U. (2003) 'Microfinance in Times of Crisis: The Effects of Competition, Rising
Indebtedness, and Economic Crisis on Repayment Behavior'. World Development, 31(12),
20852114.
Webley, P. and E.K. Nyhus (2001) 'Life-Cycle and Disproportionate Routes into Problem
Debt'. British Journal of Psychology, 92, 423446.

- 70 -

Chapter 3
The sacrifices of microborrowers in Ghana
A customer-protection perspective on measuring over-indebtedness
Jessica Schicks*
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management
Centre Emile Bernheim, Universit Libre de Bruxelles
This paper measures the over-indebtedness of microborrowers in Ghana. It defines overindebtedness from a customer-protection perspective, considering borrowers over-indebted if they
continuously struggle with repayment and experience unacceptable sacrifices related to their debt.
We find that 30% of borrowers in our urban African population of microborrowers are overindebted. The paper provides a detailed analysis of the sacrifices borrowers experience. In a second
step, it tests the risk-management indicators of debt problems as predictors of the customerprotection measurement of over-indebtedness. Over-indebtedness is strongly related to delinquency
and to the debt-to-income ratio but not to total debt amounts or to multiple borrowing. We construct
a model that correctly predicts 72.6% of cases. However, even the best indicators for overindebtedness identify only a small portion of cases of over-indebtedness. To protect customers from
unacceptable struggles, the industry needs to measure customer experiences directly. Sound risk
management is not enough to protect customers against over-indebtedness.

Paper forthcoming in Journal of Development Studies

* The author is grateful to the Marie-Christine-Adam Foundation for its financial support. She would
like to thank the Independent Evaluation Department of KfW Entwicklungsbank and the ACCION
Center for Financial Inclusions Smart Campaign for their contribution to study design and for
financing the empirical research project. She is indebted to the Ghanaian microfinance institutions
that made this data collection possible by sharing both their data and access to their clients. Valuable
comments were provided by Prof. Oscar Bernal, Prof. Marek Hudon, Prof. Marc Labie, Prof. Silvia
Prina, Prof Khalid Sekkat, Prof. Francois Rycx and Prof. Ariane Szafarz. Further thanks go to the
seminar and conference participants at the 2nd European Research Conference in Microfinance, the
Academy of Management Annual Meeting, the Global Microcredit Summit, UPC Kinshasa, the
University of Namur and ULB. Finally two anonymous referees have provided helpful comments.
- 71 -

1. Introduction
Microfinance the provision of financial services to the poor has been celebrated for its
potential to reduce poverty while simultaneously being financially sustainable or even
profitable. However, a number of crises in the industry have recently threatened both financial
sustainability and the industrys social reputation.
There are concerns about microfinance institutions drifting away from their original social
mission (Labie, 2007; Mersland, Strm, 2010; Armendriz, Szafarz, 2011). Increasingly
rigorous impact studies are questioning the original impact and poverty alleviation claims of
microfinance (Banerjee et al., 2009; Karlan, Zinman, 2010). Some microfinance markets have
clearly overheated (Chen, Rasmussen, 2010), with Andra Pradesh in India representing the
most blatant example. After a decade of focusing on commercialization, the microfinance
industry is currently undergoing a turn-around: client-focused products and services have
moved back into the spotlight and customer protection has become the industrys primary
concern. Given the risks to the social impact of microfinance as well as to institutional
sustainability, protecting customers against over-indebtedness has become the top priority.
However, there is no data on the extent of over-indebtedness. Although some national or subnational markets in India, Bosnia, Pakistan, Nicaragua and Morocco (and earlier, Bolivia)
have clearly suffered from client over-indebtedness, there remains immense uncertainty
regarding the extent of over-indebtedness in global microfinance. Moreover, there is no
accepted indicator to measure over-indebtedness. While previous over-indebtedness indicators
come mainly from the risk management perspective of avoiding delinquency, for purposes of
customer protection the industry needs to develop indicators that recognize debt problems
before customers become delinquent, but rather when they struggle with their debt to an
unacceptable extent.
Therefore, this paper deploys a definition of over-indebtedness that is appropriate for
customer protection purposes: A microfinance customer is over-indebted if he/she is
continuously struggling to meet repayment deadlines and has to make unduly high sacrifices
related to his/her loan obligations that have more than transitory effects.1 This is the first
academic study to measure over-indebtedness from the customer protection perspective, i.e. a
client welfare perspective rather than an economic perspective. It pin-points the debt struggles
of microborrowers in an urban African setting, and it reveals that even in markets with good

1 See Schicks (in press)for the details of this definition and for how to measure it in practice. Also see figure 1 in section 2.

- 72 -

repayment performance, microborrowers might suffer from excessive struggles to repay their
debt.
Additionally, this study evaluates the relationship between existing risk management
indicators of over-indebtedness and over-indebtedness based on customer sacrifices. It
evaluates the measurements of over-indebtedness that are commonly used in the geographical
and historical context of consumer finance in developed countries (Betti et al., 2007;
Vandone, 2009), examining the relationship of delinquency, debt amounts and the debt-toincome ratio (i.e. monthly repayment obligations over monthly income) to excessive
sacrifices. It also considers indicators specifically suggested for microfinance environments
(Krishnaswamy, 2007; Roesch, Hlis, 2007; Gurin et al., 2009; Morvant-Roux, 2009),
exploring the correlation between cross-borrowing and undue sacrifices.
The research is based on an extensive customer survey of 531 microborrowers in the
Ghanaian microfinance market of Accra. The respondents represent a random sample of
microdebtors from five of Ghanas most important microfinance institutions (MFIs). The
survey includes demographic and loan data, as well as detailed information on the experience
of borrowers with their loans. In addition, the participating MFIs have contributed loan
information from their MIS. 10 qualitative interviews with over-indebted borrowers
contribute background information on borrowers perceptions and local circumstances that
helped calibrate the researchers approach to analysis.
This is the first academicstudy that quantifies over-indebtedness in Ghana. It reveals that 30%
of the sampled microborrowers in Ghana experience unacceptable sacrifices related to their
debt. The paper develops a model of over-indebtedness that correctly predicts 72.6% of the
cases.2 On the respondent level, the best risk management indicators of extreme borrower
sacrifices are delinquency and the debt-to-income ratio. On an aggregate level, a debt-toincome ratio of 50% is the best indicator predicting the level of over-indebtedness in the
sample. However, as expected, the predictive power of even the best indicators remains
limited. Because there are many borrowers that make unacceptable sacrifices to avoid
delinquency, debt problems do not always manifest in arrears. A delinquency indicator alone

2 In this paper, the term prediction refers to the ability of easily available indicators to foretell, i.e., predict, how much
over-indebtedness one would find in a sample if the survey data was available that is required to measure over-indebtedness
precisely. The paper takes the perspective of MFIs and regulators that have access to data as commonly provided by the
lenders MIS. Prediction does not refer to an inter-temporal prediction of trends but to the prediction of the results a more
precise measurement would deliver.

- 73 -

recognizes only around 5% of the over-indebted cases. A debt-to-income ratio of 50%


correctly recognizes only 11% of the individual over-indebted cases.
Our findings challenge conventional wisdom concerning over-indebtedness. The common risk
management indicators for over-indebtedness are not appropriate to indicate the suffering of
borrowers. From a customer protection perspective, the argument that high repayment rates or
low levels of multiple borrowing prove the absence of over-indebtedness is invalid. Sound
risk management by MFIs does not guarantee strong customer protection.
The following section develops the hypotheses which common risk management indicators
might predict borrower sacrifices. Section 3 presents the empirical data. Section 4 provides
unprecedented insight into the sacrifices that microborrowers make related to their loans. It
analyzes the frequency of each sacrifice and describes how microborrowers perceive that
burden. Section 5 develops the quantitative methodology. This section regresses the indicators
that are potential predictors of over-indebtedness on the sacrifice-based over-indebtedness
measurement. Section 6 offers a conclusion and develops recommendations.

- 74 -

2. Predicting over-indebtedness
This paper responds to the need to measure over-indebtedness from a customer protection
perspective. It acknowledges that client protection must understand over-indebtedness in
terms of harm to borrowers. We cannot reduce over-indebtedness to issues of repayment
performance and thus harm to lending institutions. There is a need for an interdisciplinary
rather than purely economic approach.
However, measuring borrower sacrifices requires survey work. If the customer protection and
thus client welfare perspective and the risk management and thus economic perspective
correspond sufficiently, then risk management indicators that are easier to track for MFIs and
policy makers can act as predictors of the sacrifice-based over-indebtedness concept. If the
two perspectives are substantially different, this entails policy implications for regulators as
well as investors and MFIs that want to avoid harm to clients but have so far looked at overindebtedness through the lens of risk management indicators. In the latter case, investors and
MFIs have not been measuring what they are trying to manage.
Although over-indebtedness is context specific, as borrower sacrifices are a form of debt
problems we expect the common indicators of debt problems in the literature to at least
partially predict over-indebtedness in terms of borrower sacrifices. The most basic indicator
of debt problems in the literature is the amount of debt a borrower holds. Sharma, Zeller
(1997) or Godquin (2004) in Bangladesh and Vogelgesang (2003) in Bolivia all find larger
loans per microborrower positively related to repayment irregularities. Betti et al. (2007) list
the stock of debt per capita as one of the main indicators of consumer over-indebtedness in
developed countries. Moreover, a study by Brown et al. (2005) indicates that in Britain higher
amounts of debt relate negatively to an individuals level of psychological well-being. This
finding could also apply to microfinance environments, and psychological well-being is most
likely related to borrower sacrifices. Hypothesis 1 therefore postulates a relationship between
absolute debt amounts and over-indebtedness. As the total burden of the loan is more likely to
have caused sacrifices than the remaining loan balance at the time of the interview (which
might have taken place at the end of the repayment schedule), we measure debt amounts in
terms of the original loan size at disbursement.3 We include informal repayment obligations.
H1: Larger absolute debt amounts are positively related to over-indebtedness.

3 Disbursement information is also more correct and complete as many borrowers do not know the current amount of their
outstanding debt.

- 75 -

Depending on their financial and household situation, borrowers vary significantly in their
capacity to handle debt. The indicators of absolute debt amounts do not take these differences
into account. It is therefore common in the consumer finance literature to use the ratio of a
borrowers debt burden relative to his or her financial capacity as an indicator of debt and
repayment problems. The most common measure is the debt-to-income ration, i.e.

borrowers monthly repayment burden related to monthly income (Rinaldi, Sanchis-Arellano,


2006; Betti et al., 2007). One study even finds a relationship of the relative debt load to
subjective measures of debt as a burden (Del-Ro, Young, 2005). Microfinance-specific
research equally uses debt-to-income ratios as measures of over-indebtedness (Collins, 2008;
Maurer, Pytkowska, 2011) and confirms that the debt-service ratio is correlated with
subjective debt stress (Krishnaswamy, Ponce, 2010).4
H2: The debt-to-income ratio as a measure of repayment burden relative to financial
capacity is positively related to over-indebtedness.

In microfinance, where credit bureaus are rare and MFIs often do not know if their loan
applicants already have other outstanding loans, scholars assume that over-indebtedness is
related to the number of loans a borrower holds (Matin, 1997; Paxton et al., 2000; Chaudhury,
Matin, 2002; Vogelgesang, 2003; McIntosh, Wydick, 2005). The phenomenon of one
borrower holding several debt contracts at the same time is called multiple borrowing.
Over-indebtedness is especially likely to be linked to cases where borrowers exploit
information asymmetries to borrow from several lenders in parallel, rather than to cases of
one MFI extending e.g. an investment loan and a working capital loan to the same person.
While scholars do not always distinguish between the two, the phenomenon of one person
accumulating debt from several institutions at the same time is more precisely termed crossborrowing.
H3: Cross-borrowing is positively related to over-indebtedness.

4 Another measure of relative debt burden would be the debt-to-asset ratio. However, this measure is less relevant in a lowasset environment such as microfinance. Also, assets are mostly illiquid and do not facilitate repayment. We have tested the
indicator in our main model and while it does not change the other findings the debt-to-asset ratio itself is not significant.

- 76 -

Finally, delinquency is a common indicator of debt problems, the fact of repaying a loan later
than it is due or in the worst case not repaying all or part of it at all (Vogelgesang, 2003;
Godquin, 2004; Kappel et al., 2010). Delinquency or default represent the usual risk
management definition of over-indebtedness, and a standard criterion of over-indebtedness in
regulatory frameworks. Given that the sacrifices borrowers make due to difficulties in
meeting their repayment obligations are not always sufficient to ensure repayment, the
sacrifices are likely related to actual delinquency evident in the portfolio quality of MFIs.
Delinquency represents a late stage manifestation of debt problems.
H4: Delinquency is positively related to over-indebtedness.

Each of the above indicators has its challenges. Absolute amounts do not consider the
variations in borrowers repayment capacity. The ratios of debt burden to income hardly take
into account the wide range of individual circumstances that determine the share of income a
borrower can free up for repayment purposes. Cross-borrowing may not necessarily be a sign
of over-indebtedness but also result e.g., from the usual liquidity management practices of the
poor or from product limitations and credit rationing at MFIs (Krishnaswamy, 2007;
Gonzalez, 2008; Gurin et al., 2011). The relationship of cross-borrowing to debt problems
might therefore only be significant at high (and thus unhealthy) levels of cross-borrowing.
Delinquency can result from fraudulent behavior rather than from over-indebtedness. At the
same time, over-indebted borrowers might incur significant suffering in repaying their loans
and, as a result of these struggles, might manage to avoid delinquency. Section 4 reveals that
many more customers suffer in making their loan repayments than actually pay late.
So far, the above indicators have mostly been used and tested from a risk management
perspective. They have not yet been tested as predictors of over-indebtedness in the form of
borrower sacrifices. As a result, this paper expects the quality of risk management indicators
in approximating over-indebtedness to be limited. In fact, the literature that views debt
problems through the lens of repayment performance is likely to have created an over-reliance
on the indicators tested in this paper. These indicators might very well explain part of the
phenomenon of borrower struggles, but they have probably underrepresented the dimension
of customer protection in the favor of risk management.
H5: The risk-management indicators of over-indebtedness tested in H1 to H4 have a low
predictive power of over-indebtedness defined through a customer protection lens of
borrower sacrifices.
- 77 -

The following sections will use our unique data set from Ghana to empirically test the above
hypotheses. They will reveal to what extent the existing risk management indicators are
sufficient to represent the customer protection perspective of avoiding unacceptable levels of
sacrifice among microborrowers.

- 78 -

3. The data: An in-depth survey in urban Ghana


According to the MIX market5, the Ghanaian microfinance sector experienced a major growth
period between 2001 and 2007, with a Compound Annual Growth Rate (CAGR) of 173% on
gross loan portfolio and 48% in terms of active borrowers. With time, concerns about
increasing competition between microlenders and potential over-indebtedness risks started to
emerge (Grammling, 2009; Steel, 2010). However, growth slowed down and became negative
for the period 2007 to 2010, with a -19% CAGR on gross loan portfolio and -14% on the
number of borrowers. As of 2010, the industry served 194,786 active borrowers with a gross
loan portfolio of USD 71.7 million. In addition, Ghana has an active microsavings offer, with
a total of 488.633 million depositors and USD 62.0 million in deposits outstanding in 2010.
Compared to other countries, the market penetration of microfinance in Ghana is still low.
Even at the industrys peak in 2009, only 9% of Ghanas working age population below the
poverty line were using microloans, compared to 13% in Mexico, 14% in Kenya, 21% in
Ecuador, and 51% in Mongolia or Azerbaijan.6 A study comparing 12 countries for the
purposes of developing an over-indebtedness early warning index finds that Ghana has the
lowest microfinance market penetration rate of all countries in the sample (Kappel et al.,
2010). A recent representative survey on financial access in Ghana finds that 44% of
Ghanaians older than 18 use no financial products, neither formal nor informal. In the area of
Greater Accra, where our study was conducted, this group of the unbanked still amounts to
31% of the population (Finmark Trust, 2010). The overall financial access in Ghana
(including informal access) is thus lower than in Namibia or Botswana but higher than in
Nigeria, Kenya or Uganda.7
In spite of the growth and chances for saturation at some point, Steel (2010) concludes that
there is still a supply gap, i.e. unserved demand, and ample room left for growth in the
Ghanaian microfinance industry. Finally, the MFIs in our sample are the best institutions in

5 www.mixmarket.org. This website provides self-reports of limited liability but, given the lack of more reliable data sources,
is sufficient for the purposes of this sector overview. Part of the decline in recent years might be due to a reduction of MFIs
reporting their data to the MIX Market.
6 Rough estimates based on MIX market and CIA World Fact Book data. They do not take into account the coverage of a
population by the formal financial sector, nor adjust for the borrower overlap between reporting institutions.
7 Data from Ghana and Nigeria are from 2010. For Botswana, Kenya and Uganda data is from 2009, and for Namibia from
2007.

- 79 -

the market.8 They are following careful and restrictive lending practices, routinely reducing
loan sizes compared to applications. Moreover, they mostly stick to the old paradigm of
(officially) lending for productive use only. As a result, our sample can be considered roughly
representative of a normal microfinance market that is experiencing competition but is not in
crisis at the moment.
In cooperation with the Smart Campaign hosted by ACCION International, a global effort to
unite microfinance leaders around a common set of client protection principles, and with the
Independent Evaluation Department of KfW Entwicklungsbank, the German development
bank, we have conducted an in-depth survey among 531 microdebtors. Borrowers were
sampled from five of Ghanas leading microfinance institutions: ProCredit Ghana,
Opportunity International Ghana, Sinapi Aba Trust, EB-ACCION and Advans Ghana. Among
all the MFIs reporting to the MIX Market in 2010, these MFIs account for 83% of
microborrowers in Ghana (43% in 2009) and for 95% of Ghanas gross microloan portfolio
(43% in 2009). The respondents represent a random sample from the institutions
microborrowers in Accra, the heart of Ghanas microfinance industry.9 To balance their
expected lower response rate and to ensure sufficient data points and variation of customers in
serious problems with their loans, we oversampled delinquent customers.10 We used sample
weights to correct for this bias and for a variation in response rates between lending
institutions, delinquency status, and lending methodology. There were no corrections required
for disparities in gender. Table 1 provides an overview of the sample characteristics.11 To
encourage honest replies, interviews were conducted anonymously at a site of the
respondents convenience, independent of the MFIs.

8 Given the high quality of the lending institutions that contributed to our sample, we consider the amount of overindebtedness identified by this study as a lower estimate. Customers of weaker MFIs are likely to be selected less carefully
and to experience more multiple borrowing, delinquency and sacrifices.
9 To apply a common threshold for micro- and SME-borrowers across all MFIs, we consider as microborrowers all
customers with active personal loans below 5000 Ghana Cedis (GH; 1 GH=0.7 USD). For most MFIs in the sample, all of
their borrowers fall into this category. In the sample, 87% of the disbursed loan amounts are less than 2000 GH.
10 With some MFIs, over-sampling for groups implied over-sampling delinquent groups rather than individuals.
11 As the research methodology relies on self-reports, the statistical data regarding the borrowers economic situation is
subject to limitations. Besides questions of honesty, the respondents often experience difficulty estimating the monetary
value of their assets and their average incomes, given the volatility of such incomes. However, on average there is no reason
to assume a general upward or downward bias of the estimates and we consider the data sufficiently reliable for our level of
analysis. We have analyzed the key variables such as debt-to-income ratio, total assets and debt outstanding for outliers and
there are no outliers to be excluded.

- 80 -

<insert table 1 about here>


In addition to socio-demographic and economic information, as well as the details of all the
respondents formal or informal loans that were outstanding at the time of the interview, the
questionnaire obtained information on the sacrifices the borrowers experienced related to their
loans over the course of one year. Without any influence of the interviewer, respondents
answered first with the sacrifices that came to their mind. In a second step, the interviewers
checked for additional sacrifices using a pre-defined list based on sacrifices identified in the
interdisciplinary literature by e.g., (Brett, 2006), Corbett (1988) and (Rahman, 1999), and then
piloted in Ghana (see Appendix for list of sacrifices). On an individual basis, the respondents
indicated how many times they experienced each of the sacrifices12, and they weighted
sacrifices subjectively by how acceptable the experience was, given the purpose of the loan.13
Cutting down on food could therefore be acceptable for one person (e.g., if it meant
substituting meat for cheaper food), while for another borrower it would be unacceptable (e.g.
because it implied going hungry on a single meal per day).
Based on this sacrifice data, all respondents enter the funnel displayed in figure 1. Borrowers
are considered over-indebted only if they meet all criteria in our sacrifice-based overindebtedness definition, i.e., they struggle to repay on time, they make unacceptable sacrifices
and these unacceptable sacrifices occur repeatedly. Given the short-term orientation of
microloans and the severe consequences of sacrifices even over short periods of time, we
employ repetition as the criterion for debt problems being of a more than transitory nature.
For severe sacrifices such as assets seizures, loan recycling, and selling or pawning ones
assets, a single unacceptable experience is a sign of more than transitory debt problems. These
sacrifices typically occur only once per loan, but they nevertheless have effects that persist
over a longer time. A seizure occurs only after permanent delinquency; taking new loans to
repay old ones or selling assets usually augments repayment capacity for several installments
and represents a long-term cost for the borrower.14

12 Once in past year, 1-3 times in past year, >3 times but not often, or Frequently in past year. For a respondent
cutting down on their food at several points for a week at a time, instead of every individual day, each week would count as
one occurrence.
13 Easily acceptable, Only just acceptable, Not really acceptable, or Not acceptable. In this paper we summarize the
first two categories under acceptable and the latter two under unacceptable or not acceptable.
14 The measurement only counts unacceptable sacrifices and thus does not include assets sales or loan swaps that borrowers
simply employ as liquidity management tools. Nevertheless, the calibration of this measurement might have to be different in
a different cultural context.

- 81 -

<insert Figure 1 about here>


To avoid the possibility that the econometric results in section 5 might depend on the
threshold set by our over-indebtedness definition, we develop an alternative measure of
borrower sacrifices as a robustness check that is free of any threshold. We construct a discrete
score of sacrifices from zero to 71: Each sacrifice a borrower has experienced is recorded with
two to eight points, depending on the frequency and acceptability of the experience. The
harder a sacrifice was to accept and the more frequent the experience, the more it increases
the score. This measurement of sacrifices is independent of any definition of overindebtedness. The downside of the sacrifice score is that it does not differentiate between
loans that require an acceptable level of sacrifice from borrowers and those that require
serious debt struggles. Customer protection does not aim to prevent any sacrifice but to
prevent unacceptable levels of sacrifice. Even if working with a binary variable entails losing
information, the sacrifice score therefore remains a robustness check only.

- 82 -

4. The sacrifices of microborrowers


This section reveals to what extent the microborrowers in our urban Ghanaian sample struggle
with their debt. It quantifies the level of over-indebtedness according to our customer
protection definition of over-indebtedness. The section provides a detailed analysis of the
sacrifices borrowers encounter and of the frequency of the various sacrifice experiences. It
shows which sacrifices the average borrower perceives as most severe.
In spite of the low level of delinquency (see table 1), repayment struggles are very common
among Ghanaian microborrowers. In our sample, only a quarter (26%) of the microborrowers
report that they are not struggling with their loan repayments. Some of these respondents
report that they make certain sacrifices for their repayments but they do not consider them a
serious struggle. 31% of borrowers struggle with an installment once in a while, and 43%
struggle regularly or at (almost) every single installment. According to the detail of sacrifices
that respondents described during the interviews, respondents on average did not overstate
their struggles or sacrifices but rather had a high tolerance for sacrifice and a tendency to
underreport their personal hardships.15
Struggles with debt as such are not a sufficient indicator of over-indebtedness. However the
sacrifices of borrowers in our sample indicate that, according to the customer-protection
definition of over-indebtedness, over-indebtedness in Ghana is substantial. One third of
microborrowers meet all three criteria of over-indebtedness as specified in our sacrifice-based
definition: they struggle to repay their loans on time, they experience unacceptable sacrifices,
and this experience can be considered more than transitory on the grounds of repetition or of
the sacrifices longer-term impact (as in the case of seizures, assets sales and loan recycling).
As seen in table 2, the level of over-indebtedness among the microborrowers in our urban
Ghanaian sample is at 30%.
<Insert Table 2 around here>
This confirms the implicit knowledge of MFI managers and loan officers in Ghana who have
indicated their impression that there might be over-indebtedness in the Ghanaian microfinance
market. However, their concerns had never been confirmed or quantified to date. Our
empirical findings also highlight that the phenomenon of over-indebtedness is not limited to

15 In addition, triangulating the survey data on delinquency and multiple borrowing with the objective information from the
participating MFIs management information systems (MIS) indicates that borrowers self-reports were mostly honest.

- 83 -

crisis markets. From a customer protection perspective, over-indebtedness also exists in noncrisis environments such as Ghana.
These findings relate to but do not measure the impact of microcredit. The borrowers report
only sacrifices that they perceive as related to their repayment obligations. However, this does
not automatically imply causality. It is normal for cash demands to represent a difficulty for
the poor, yet loan impact may still be positive.16
<Insert Figure 2 around here>
Analyzing the sacrifices of microborrowers in more detail, we find that, collectively and on
average, microborrowers act according to the assumptions of rational behavior: most
frequently, they make those sacrifices that they experience as the least costly. Figure 2
displays this almost linear relationship between the prevalence of the various sacrifices among
microborrowers and the perceived acceptability of sacrifices. On the X-axis sacrifices appear
ranked by their prevalence among all respondents in the sample. The most prevalent sacrifices
are those that only a few of those borrowers who experience the respective sacrifices perceive
as unacceptable. The more sacrifices seem unacceptable to the average borrower, the less
common they are; this is most likely because borrowers first prefer to employ easier coping
strategies first and avoid the hardest sacrifices by all means.
Table 3 presents the sacrifices of microborrowers in more detail. The sacrifices are again
ranked by their prevalence in the total sample. This reveals the distinction of four categories
of sacrifices grouped according to their prevalence and acceptability. Borrowers most
frequently work more than usual (61% of total sample), postpone important expenses (45%)
and, if available, deplete their savings (34%). Only a little more than a third of the borrowers
who make these sacrifices consider them unacceptable. These are probably the more severe
cases of these sacrifices, such as working at times of serious illness or in dangerous
environments or where fundamental expenses for housing and daily survival cannot be met.
Apart from these exceptions the borrowers are very willing to sacrifice for example some free
time or some consumption for the purpose of repaying their loans. Thus, these minor
sacrifices, when acceptable, represent adequate coping strategies and do not constitute signs
of over-indebtedness. The sacrifice-based definition of over-indebtedness only takes
unacceptable and an more than transitory sacrifices into account.

16 For details on the relationship between over-indebtedness and negative loan impact see Schicks (in press).

- 84 -

With the exception of the depletion of savings (29%), sacrifices in this first category usually
become frequent experiences to those that go through them once. They represent the typical
coping strategies that borrowers employ on a regular basis (71% for working more and 60%
for postponing expenses). The frequency of depleting savings is probably simply limited by
the lack of savings once depleted.
<Insert Table 3 around here>
The second category of sacrifices consists of those sacrifices that each affect between 10 and
20% of borrowers and that are unacceptable to most respondents. Microborrowers reduce
their food (18%), rely on friends or family to help them out (13%) and suffer from
psychological stress (10%). Between 70 and 80% of borrowers who experience these
sacrifices rate them as unacceptable. Nevertheless, once this stage of debt problems is
attained, food reductions (62%) and psychological stress (52%) tend to become repeated
experiences. The frequency of relying on external support (21%) is most likely limited by the
supports availability. Gurin et al. (2011) note the high cost of family support in India, where
kin debts form part of the most severe debts. The low acceptability of the sacrifice of relying
on support by friends and family in Ghana confirms the findings of Gurin et al. in the
Ghanaian cultural context.
A third category of sacrifices is much less prevalent, but is unacceptable to almost all
borrowers (80%-90%) if it occurs. These sacrifices are reductions in education, e.g., removing
children from school (5% of borrowers), trying to keep up repayments on one loan by taking
on a new debt obligation (4%), and selling or pawning assets (4%). Once this stage of
sacrifice is reached, for some but not for the majority of borrowers, loan recycling or assets
sales become repeated experiences (49% and 38% respectively). Probably, single experiences
of these sacrifices already facilitate several periods of repayment at once. The cuts to
education rarely occur more than three times (13%).
Finally, the most severe experiences from the perspective of borrowers are experiences of
threats or harassment (3% of borrowers), of shame or insult (3%) and of asset seizures (1%).
These sacrifices are unacceptable to all borrowers who have experienced them. This
subjective evaluation of the severity of sacrifices indicates that a borrowers reputation and
personal honor might be more important than monetary privation and material sacrifices. The
numbers of observations for these items are too low to make qualified statements about their
repetition, but they tend to occur less frequently than most other sacrifices. A repeated seizure
happened in only a single case.
- 85 -

The data also allows us to analyze how the borrowing experiences of women differ from the
sacrifices of men (table 4), and how group lending customers differ from borrowers in the
individual lending methodology (table 5). The sample size and the high number of sacrifices
do not always allow for statistically significant findings. However, based on a chi-square
contingency analysis of gender differences, Cramers-V (Backhaus et al., 2011) as a measure
of association between sacrifices and gender is statistically significant with regards to loan
recycling (8% of male borrowers experiences this sacrifice versus 3% of female borrowers),
the selling or pawning assets (7% versus 3%), and the depletion of savings (39% versus 31%).
Men resort to these solutions more often than women, which might be due to better
economic/borrowing opportunities and to mens greater control of assets. Additionally, fewer
women report experiences of threats and harassment (7% male versus 2% female). However,
this difference in reporting might simply derive from womens different perception of their
rights towards loan officers. In all cases, Cramers-V is approximately 0.1, which indicates a
relevant but rather low influence of gender on the prevalence of these sacrifices.
<Insert Table 4 around here>
Research on the effects and incentives of peer pressure suggests that group customers make
high sacrifices to avoid being responsible for the delinquency of their group. In our sample,
compared to individual borrowers, group customers are significantly more prone to depleting
their savings (41% versus 28%), relying on the support of friends and family (18% versus
7%), and suffering psychological stress (11% versus 7%). They also sell or pawn their assets
more often (7% versus 3%). Cramers-V is low to moderate and the strongest influence of the
lending methodology exists with regards to support by family and friends (-0.17). While the
enhanced prevalence of psychological stress rather points to the downsides of the group
lending methodology of enhancing pressure on borrowers (rather as an incentive mechanism
than as an advantage to customers), the higher reliance of group borrowers on support by
friends and family might result inter alia from the better availability of support within the
groups. To the extent that borrowers assess this support as acceptable it might thus partially
be a positive effect of group lending.
<Insert Table 5 around here>
In sum, this section has revealed insights into the borrowing experiences of microfinance
borrowers in urban Ghana. It identifies which sacrifices are most prevalent among
microborrowers, which are most severe from the subjective perspective of the debtors and
how frequently the average borrower experiences the various sacrifices. The majority of
- 86 -

microborrowers work harder than usual to repay their loans. Many postpone important
expenses. They usually consider these efforts absolutely acceptable. However, there are also
borrowers who experience more severe sacrifices, from going hungry over taking their
children out of school to selling their assets. The most painful and least frequent experiences
are those of threats or harassment, shame or insults, and asset seizures. A few sacrifices differ
in prevalence by gender and by lending methodology. Women reduce their food more often,
while men more often result to external mechanisms such as asset sales and repeat borrowing.
Group customers deplete their savings more often, suffer from psychological stress and rely
on the help of friends or family.
The section finds that, from a customer protection perspective, 30% of borrowers in our urban
Ghanaian sample are over-indebted. They struggle to repay their loans on time, and they make
sacrifices that they experience as unacceptable, and that can be considered of a more than
transitory nature. The sample consists of borrowers from five of Ghanas largest and most
professional MFIs. The over-indebtedness in our sample is therefore likely to represent a
lower bound estimate for the total microfinance market of Accra.

- 87 -

5. Estimation and empirical results


The previous section has shown that a relevant share of microborrowers in Ghana experiences
severe sacrifices related to their repayment obligations. Measuring over-indebtedness in terms
of borrower experiences is therefore an important practical requirement to protect
microborrowers. This section tests the hypotheses developed in section 2 to identify potential
predictors for the sacrifice-based measurement of over-indebtedness that do not require
survey work. It analyzes how the customer protection (and thus client welfare) perspective of
over-indebtedness relates to the risk management (and thus economic) perspective.
Estimations
To test our hypotheses on the potential over-indebtedness predictors, we construct a logistic
regression model that predicts our binary indicator of over-indebtedness based on severe
sacrifices. We control for borrower characteristics and the main features of the respondents
loans. Asset, gender, maturity and lender controls follow the repayment focused regression
analysis of Godquin (2004). Because Godquin had group borrowers only, we introduce an
additional control for the lending methodology. As common determinants of repayment
capacity in developed countries (Bridges, Disney, 2004), we add controls for age, household
size, and income. Loan characteristics such as variations in interest rates, the type of group
lending methodology, and the repayment schedule tend to vary on the institutional level. We
therefore capture the influence of these loan features on over-indebtedness by means of our
lender dummies.
For reasons of multicollinearity with our cross-borrowing indicator, we exclude control
dummies for lenders other than our five partner MFIs. Instead, we use dummies only for the
main lenders among our partner institutions in terms of loan size. This difference is negligible
because only 3.6% of our sample indicated larger loans from non-partner institutions than
from the partner MFIs. There is a risk of endogeneity with regards to delinquency in so far as
delinquency can trigger sacrifices for borrowers but at the same time it may be the unbearable
amount of sacrifices experienced that triggers an over-indebted borrowers decision to no
longer pay on time. However, we do not analyze causality but are testing the quality of
indicators as predictors of over-indebtedness. Neither endogeneity nor omitted variable bias
affect our interpretation regarding the use of risk management indicators as predictors for
over-indebtedness.

- 88 -

We test hypothesis 1 to 5 by estimating the following logit model:


Oi = 0 + 1 DAi + 2 DIRi + 3 CBi + 4 DEi + 5 Zi + 6 Xi + ui

(1)

where for each respondent i, Oi is our binary measurement of over-indebtedness according to


the sacrifice-based definition, DAi is the total formal and informal amount of debt disbursed,
DIRi is the debt-to-income ratio, CBi is cross-borrowing i.e., the number of loans from
different lenders, and DEi is delinquency, i.e. a dummy that takes the value one for a borrower
who is in arrears of at least one day on an outstanding loan at the time of the survey. Zi is a
matrix of borrower-specific controls, and Xi represents the loan specific controls.
<insert table 6 about here>
Table 6 displays the logistic regression results in the form of odds ratios and robust standard
errors in parenthesis. Column 4 shows the results for the main model including all controls.
Across all models, the debt-to-income ratio (H2) is consistently significant and positive. In
terms of marginal effects, a 1% increase in debt-to-income ratio for an average borrower
corresponds to an increase of the probability of over-indebtedness by 0.002.17 Similarly,
delinquency is positive and highly significant at the 1%-level in all models (H4). With all
other factors equal at their means, the probability of over-indebtedness is by 0.313 higher if a
borrower is delinquent. The number of lenders a borrower is indebted to (H3) is not a
significant predictor of over-indebtedness. The absolute amount of debt disbursed to a
borrower (H1) is only significant in the models without borrower controls. In the main model
in column 4 and in the parsimonious model in column 5, the effect disappears. In sum, the
debt-to-income ratio and delinquency (Hypotheses 2 and 4) are confirmed as significant
predictors of over-indebtedness. Debt amounts and numbers of loans are not confirmed
(Hypotheses 1 and 3). Further research should analyze however, to what extent debt amounts
could work as predictors from the perspective of regulators and MFIs who may not have the
data for borrower specific controls.
The main model in column 4 correctly predicts 72.6% of all cases. This is significantly greater
than the 55 to 60% of cases that random selection predicts correctly at a predetermined level
of 30% over-indebtedness. However, it is only marginally greater than the maximum random
probability of 70% (Backhaus et al., 2011), and confirming Hypothesis 5, the predictive
17 See Appendix 2 for the marginal effects table that corresponds to the main model displayed in column (4).

- 89 -

power of the models remains low with Nagelkerkes R at approximately 10%. The sensitivity
analysis below sheds more light on the schism between a highly significant correlation of
predictors to over-indebtedness but low predictive power.
Robustness checks and further analyses
Additionally, we analyze the differences between male and female borrowers on the one hand
and individual and group borrowers on the other. Table 7 indicates that both the gender and
the lending methodology display differences. While high-level results remain the same,
delinquency is a better predictor for female borrowers than for men, and the debt-to-income
ratio is a better predictor for male borrowers. Also, for male borrowers the amount of debt
disbursed becomes significant. According to the data in section 4, men resort to external
sacrifices that require control over assets more often and may thus have more options to avoid
delinquency. In line with these findings, the link to economic indicators such as debt amounts
and income may be more direct for men. Column 2 indicates similar differences between
group and individual borrowers. Again, the findings of our main model remain robust.
However, the debt-to-income ratio and total amount of debt are predictors only for group
customers and delinquency is a significant predictor only for individual borrowers. The link
between delinquency and over-indebtedness in groups might be weakened by the
intermediation of group members. In cases of group solidarity or forced joint liability, the
contributions of other group members might avoid formal arrears when a borrower is in
trouble but the pressure on the borrower to reimburse his peers remains and can trigger
additional sacrifices. Future research should further examine the differences identified in this
section.
<insert table 7 about here>
As there are no standardized measurements for delinquency, we test the robustness of our
findings to several alternative measurements for delinquency (see table 8). The most obvious
delinquency measurement and the one we use in our main model, is that a borrower is
currently in arrears: At the time of the survey he/she is at least one day late with at least one
installment on an outstanding debt. We measure this indicator as a binary variable. However,
the sacrifices a borrower reports might have been related to an instant of delinquency in the
recent past. Therefore we also test a measure of delinquency over time rather than a spot
indicator at the time of the interviews: We introduce an alternative binary variable if the
- 90 -

respondent has been delinquent at any point over the course of their current loans. Finally, we
test intensity measures of delinquency, i.e., being in arrears for a longer time or having been
delinquent more frequently. As the latest stage and most severe indicator, we test default,
approximated by experiences of assets seizures or forgiven loans.18
<insert table 8 about here>
Table 9 displays the effects of varying the measure of delinquency. In all models, the debt-toincome ratio and delinquency remain highly significant predictors of over-indebtedness.
Cross-borrowing remains without effect. The total amount of debt remains insignificant with
the exception of the model that uses default as the most severe delinquency measurement.
Moreover, in column 2 of table 9 we cluster standard errors by lending institutions. Clustering
standard errors has no impact on our findings and all results remain robust as discussed above.
<insert table 9 about here>
Finally, to test the sensitivity of the above analysis to the dependent variable overindebtedness, we run an ordinary least square regression of our predictors on a threshold-free
score of sacrifices as a robustness check (2).
Si = 0 + 1 DAi + 2 DIRi + 3 CBi + 4 IRi + 5 Zi + 6 Xi + ui

(2)

where Si is a discrete score of sacrifices from zero to 71, the score increasing with each
additional sacrifice, with more frequent repetition of sacrifices, and with the subjective
severity of sacrifices. The score of sacrifices makes it trivial to interpret coefficients because
an increase of the score by one is a meaningless measurement to the reader. However, it
avoids any potential arbitrariness of the over-indebtedness threshold and therefore provides a
useful robustness check.
<insert table 10 about here>
Table 10 confirms that the debt-to-income ratio and delinquency are consistently highly
significant across all models. Additionally, the total amount of debt disbursed to a borrower is
now significant even in the full model although at a low level of confidence, and not in the
parsimonious model. Interestingly, its coefficient is negative, which might indicate sound
18 Information for delinquency >30 days is not available but would be situated between the 1 week late and default indicator.

- 91 -

lending decisions at MFIs: larger absolute amounts are lent to those customers who have a
higher capacity to repay and who will struggle less with installments. There is a need for
further research to analyze the differences in the logit and OLS models with regards to debt
amounts. Cross-borrowing remains insignificant and the predictive power of the model
remains low.
Sensitivity analysis
Given that delinquency and the debt-to-income ratio have proven the best predictors of overindebtedness among the tested indicators, even if their predictive power remains limited, we
analyze the use of these indicators in practice. The next paragraphs analyze the number of
cases each of these predictors individually can predict correctly. As all measurements of
delinquency in table 9 provided similar results, we test the sensitivity of results to the
alternative indicators. Because delinquency in the sample is rather low, the various indicators
of delinquency recognize only a small share of over-indebtedness even if almost all
delinquents and defaulters are rightly classified as over-indebted. Table 11 demonstrates that
none of the measurements recognizes more than 21% over-indebtedness in the sample; in fact
most indicators recognize significantly less than 10%. Most measurements recognize only 1
to 5% of the over-indebted respondents correctly (maximum 9%). The correct prediction of
approximately 70% of cases in the full sample is thus due to the large group of borrowers who
are not over-indebted and to the high probability that a delinquent borrower is over-indebted.
The majority of the over-indebted in contrast are not delinquent and the indicator does not
recognize them.
<insert table 11 about here>
For the debt-to-income ratio, the threshold is more flexible. Table 12 indicates that a 10%
threshold of the debt-to-income ratio recognizes a third of the over-indebted. However,
because it does so by classifying a total of 89% of all borrowers as over-indebted, the
threshold is not useful. The higher the threshold gets, the lower the share of the over-indebted
that is correctly identified. At a threshold of 60%, only 7% of borrowers are still correctly
classified as over-indebted, compared to approximately 30% that should be in this category.
The best indicator to predict the overall level of over-indebtedness in the sample is a debt-toincome ratio of 50%. Considering every borrower with a higher than 50% debt-to-income
ratio to be over-indebted approximates the level of over-indebtedness in the sample rather
- 92 -

well at 28%. Nevertheless, in recognizing only 11% of those borrowers that are over-indebted
correctly, this indicator does not work well to categorize individuals. Given the low
effectiveness on the individual level, this threshold might not hold in other markets or future
situations, even on the over-all sample level. More research is required to determine if a 50%
debt-to-income ratio holds as a policy recommendation for regulators to determine the
average level of over-indebtedness in a market.
<insert table 12 about here>
The empirical results in this section indicate that the debt-to-income ratio and delinquency are
highly significant predictors of over-indebtedness from the customer protection point of view.
In contrast, a borrowers number of loans outstanding are not significant. Neither is the
amount of debt disbursed to a borrower a consistently significant predictor of overindebtedness. Based on the debt-to-income ratio, delinquency, and several controls on the
borrower and institutional level we can correctly predict the over-indebtedness status of
72.6% of respondents. However, good prediction relies mainly on the large group of
borrowers that are not over-indebted. It is improved by the small group of over-indebted
borrowers who are already delinquent. Beyond that, on an individual level, neither the debtto-income ratio nor delinquency are good measurements to identify who the over-indebted
borrowers in the sample are. On an aggregate level, a debt-to-income ratio of 50% recognizes
the correct level of over-indebtedness in the population. This indicator requires further
research. Further research is also required into the gender differences and the differences
between group and individual borrowers.

- 93 -

6. Conclusion
Over-indebtedness is one of the major risks for the microfinance industry but there is hardly
any research on its extent in microfinance markets, nor are there accepted indicators for
measuring it. This paper has measured the over-indebtedness of microborrowers in urban
Ghana, defining over-indebtedness from a perspective of customer protection, i.e., as an
unacceptable level of sacrifices that borrowers experience related to their debt. It revealed
that, in Accra, 30% of the borrowers of Ghanas top five microlenders are over-indebted, i.e.,
they suffer from their debt at a level that they consider unacceptable and that has more than
just transitory effects. The study provided unprecedented insights into the sacrifices of these
borrowers. It highlighted the most common and the most frequent sacrifices and it identified
which sacrifices are the most difficult for borrowers to bear.
Additionally, the paper tested four potential predictors of this perspective on overindebtedness, borrowing from the microfinance and high-income country consumer finance
literature on debt problems. We consistently found that the debt-to-income ratio (the monthly
repayment burden over income) and delinquency are highly significant predictors of sacrificebased over-indebtedness (H2 and H4 confirmed). The amount of disbursed debt is not
consistently significant, and in our robustness check debt disbursed has a negative coefficient,
possibly as a sign of the MFIs sound lending decisions (H1 not confirmed). Cross-borrowing
is not related to our measure of over-indebtedness (no confirmation for H3). Note that
multiple and cross-borrowing are not significant predictors of delinquency either. It is
possible that the relationship only comes into effect at extreme levels of multiple borrowing
as in Maurer, Pytkowska (2011).
Our model correctly predicts 72.6% of cases. Nevertheless, we find hypothesis 5 confirmed
that even those predictors that are highly significant at a one percent level recognize only a
low share of over-indebtedness through the customer protection lens. A debt-to-income ratio
of 50% is the best predictor of the aggregate level of over-indebtedness in a market. However,
on an individual respondent level, it recognizes only 11% of the borrowers that are overindebted correctly. Similarly, while almost all borrowers that are delinquent are equally
suffering from severe sacrifices, the reverse does not hold true: there are many struggling
borrowers who have not (yet) reached the stage of delinquency. Due to extensive sacrifice,
many might never get to that stage. As late stage of over-indebtedness, delinquency is
strongly related but not very useful for prognosis. This analysis refutes the common argument

- 94 -

that good repayment statistics prove the benefits of microloans to poor borrowers and that
they defy allegations of over-indebtedness in the sense of severe borrower sacrifices.
Over-indebtedness harms microborrowers before it affects the portfolio quality of lending
institutions. The current focus of over-indebtedness research on risk management indicators is
therefore inappropriate for customer protection purposes. The client welfare perspective and
the economic perspective of over-indebtedness are closely related but clearly separate. The
microfinance methodology, praised for reducing the risk of lending to the poor and thus
making them bankable, has not made risk disappear; it has shifted it from the portfolios of
institutions to their vulnerable customers. To act on its claims of protecting clients, the
microfinance industry needs to directly take the client welfare perspective, the customers
point of view into account.
These findings result in policy recommendations for all actors involved in client protectionrelated activities, from MFIs to investors to regulators.19 The most basic and most important
recommendation refers to the actors frame of attention: risk management and portfolio
quality measures do not represent an appropriate frame of attention for customer protection
purposes. In the future, actors with a customer protection responsibility should measure overindebtedness in terms of borrower experiences, even if that will require survey work. Future
research should continue to identify predictors of the customer protection perspective on overindebtedness that are less costly to measure. According to the Attention Based View (Ocasio,
1997), managements mere awareness of this schism is likely to have a significant impact on
the everyday lives of customers on the ground. Examples of possible actions include a
reduction in inappropriate marketing activities with regards to debt, and a relaxation of the
zero-tolerance policy, thus increasing the flexibility in restructuring loans for borrowers
undergoing honest temporary difficulties. Similarly, awareness of over-indebtedness risks
might entail more careful lending decisions in cases where a borrower is likely to repay the
lender (and thus contribute to his profit) but can most likely do so only at an unacceptable
personal cost.
To balance the pressures for high disbursements and harsh collections, MFIs may introduce a
customer satisfaction element to loan officer incentives. Loan officers should be incentivized
not to hide borrowers problems, but to reveal them before they reach the stage of

19 As many of the recommendations refer to the behaviour of MFIs and thus to the supply side, a thourough analysis of
supply-side factors should accompany our analysis on the demand side to develop the measures in more detail and assess
their feasibility.

- 95 -

delinquency. For details on the requirements of such incentives in terms of monitoring and
hiring policies, see Agier, Szafarz (2011), who recommend similar incentives to counter
borrower discrimination by loan officers. Consistent with the South African National Credit
Act, by which lenders cannot collect on loans if they have not conducted an affordability
assessment and their loan falls in the category of reckless lending, MFIs could introduce expost bonus reductions for troubled loans: penalties would apply to loans for which the loan
officer could have anticipated that repayment would become too difficult for the borrower.
Even though such individual measures may be useful, the Attention Based View suggests that
complex phenomena such as client protection are best addressed by focusing attention on the
phenomenon at hand. For example, MFIs could try to promote a strong organizational culture,
which in our case would consist of a welfare orientation above economic efficiency and
profits.
Similarly, regulators and investors should understand the need for customer protection that
goes beyond ensuring stable financial institutions. A mix of regulation and self-regulatory
codes of conduct is likely to be most effective. Donors should put as much focus on
promoting the measurement of impact, of the client welfare perspective, as they have put on
economic self-sustainability and reliable economic performance measurements in the past.
This study makes clear that in measuring the client welfare perspective, the focus should not
only be on positive impact. Instead impact measurement should explicitly consider the
potential downsides of debt.
Government and regulators could also play an important role in developing systems of debt
relief and personal insolvency. Such safety nets reduce borrower sacrifices. Moreover, even if
the potential for abuse render subjective borrower sacrifices futile as a legal overindebtedness definition, our study also has implications for the indicators regulators use to
measure over-indebtedness. Regulators need indicators that are easier to track and more
objective than a measurement based on borrower sacrifices. They should represent borrower
sacrifices as closely as possible. Delinquency and the debt-to-income ratio are the best
indicators we identified. At the same time they do not represent the customer-protection
perspective very well: Our sensitivity analysis of these predictors indicates that they correctly
predict only a small share of over-indebted borrowers. The debt-to-income ratio of 50%
predicts the aggregate level of over-indebtedness in the market, but this threshold might vary
in other markets and does not work reliably on an individual borrower level. Potentially a
debt-to-income ratio can be used to monitor trends on the aggregate market level and a rising
ratio could represent an early warning signal for markets where a closer analysis of over- 96 -

indebtedness is necessary. This indicator therefore requires further research as do the


differences between men and women and between group and individual borrowers.
In general, the microfinance industry should measure customer satisfaction and impact as a
standard management tool and place the same importance on these factors as it places on
economic indicators. For more precise recommendations and for a better generalization of
findings, similar studies should be repeated in other market environments. Over-indebtedness,
especially if measured according to sacrifices, is highly context dependent and it is unclear to
what extent the above findings would hold in other geographical, cultural and historical
contexts. Also, more potential indicators for over-indebtedness might have to be developed
and tested.

- 97 -

References
Agier, I. and A. Szafarz (2011) 'Subjectivity in Credit Allocation to Micro-Entrepreneurs:
Evidence From Brazil'. Universit Libre de Bruxelles CEB Working Paper No. 11/016,
Brussels.
Armendriz, B. and A. Szafarz (2011) 'On Mission Drift In Microfinance Institutions' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 341366). London,
Singapore: World Scientific Publishing.
Backhaus, K., B. Erichson, W. Plinke and R. Weiber (2011) 'Multivariate Analysemethoden,
Eine anwendungsorientierte Einfhrung (13th ed.), Berlin: Springer.
Banerjee, A. V., E. Duflo, R. Glennerster and C. Kinnan (2009) 'The miracle of microfinance?
Evidence from a randomized evaluation'. MIT Poverty Action Lab, Cambridge, MA.
Betti, G., N. Dourmashkin, M.C. Rossi and Y.P. Yin (2007) 'Consumer over-indebtedness in
the EU: measurement and characteristics'. Journal of Economic Studies, 34(2), 136156.
Brett, J. A. (2006) '"We Sacrifice and Eat Less": The Structural Complexities of Microfinance
Participation'. Human Organization, 65(1), 8.
Bridges, S. and R. Disney (2004) 'Use of Credit and Arrears on Debt among Low-Income
Families in the United Kingdom'. Fiscal Studies, 25(1), 125.
Brown, S., K. Taylor and S. Wheatley Price (2005) 'Debt and distress: Evaluating the
psychological cost of credit'. Journal of Economic Psychology, 26(5), 642663.
Chaudhury, I. A. and I. Matin (2002) 'Dimensions and dynamics of microfinance membership
overlap: A micro study from Bangladesh'. Small Enterprise Development, 13(2), 4655.
Chen, G. and S.R.X. Rasmussen (2010) 'Growth and Vulnerability in Microfinance'. CGAP
Focus Note No. 61, Washington DC.
Collins, D. (2008) 'Debt and household finance: Evidence from the Financial Diaries'.
Development Southern Africa, 25(4), 469479.
Del-Ro, A. and G. Young (2005) 'The impact of unsecured debt on financial distress among
British households'. Bank of England Working Paper No. 262, London.
Finmark Trust (2010) 'FinScope Ghana 2010: Making financial markets work for the poor.
Launch Presentation, Accra.

- 98 -

Godquin, M. (2004) 'Microfinance Repayment Performance in Bangladesh: How to Improve


the Allocation of Loans by MFIs'. World Development, 32(11), 19091926.
Gonzalez, A. (2008) 'Microfinance, Incentives to Repay, and Overindebtedness: Evidence
from a Household Survey in Bolivia. Doctoral thesis'. Ohio State University, Ohio.
Grammling, M. (2009) 'Cross-borrowing and Over-indebtedness in Ghana: Empirical
Evidence from Microfinance Clientele and Small Enterprises. Technical draft for
discussion'. ProCredit Holding, Frankfurt am Main.
Gurin, I., M. Roesch, O. Hlis and Venkatasubramanian (2009) 'Microfinance, Endettement
et Surendettement: Une tude de cas en Inde du Sud [Microfinance, Indebtedness and
Over-indebtedness: A case study in Southern India]'. Revue Tiers Monde, 197, 131146.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2011) 'The social meaning of
over-indebtedness and creditworthiness in the context of poor rural South India households
(Tamil Nadu)'. RUME Working Paper No. 2011-1, Paris.
Kappel, V., A. Krauss and L. Lontzek (2010) 'Over-Indebtedness and Microfinance Constructing an Early Warning Index'. Center for Microfinance, University of Zurich,
responsAbility, Council of Microfinance Equity Funds, Triodos Investment Management,
Zurich.
Karlan, D. and J. Zinman (2010) 'Expanding Credit Access: Using Randomized Supply
Decisions to Estimate the Impacts'. Review of Financial Studies, 23(1), 433464.
Krishnaswamy, K. (2007) 'Competition and Multiple Borrowing in the Indian Microfinance
Sector'. Institute for Financial Management and Research, Centre for Micro Finance
Working Paper.
Krishnaswamy, K. and A. Ponce (2010) 'An analysis of the Karnataka mass defaults'.
Working Paper.
Labie, M. (2007) 'Rflexions prliminaires pour une approche thique de la gestion des
organisations en microfinance'. Ethics and Economics, 5(1), 18.
Matin, I. (1997) 'Repayment performance of Grameen Bank borrowers: the unzipped' state'.
Savings and Development, 4, 451473.
Maurer, K. and J. Pytkowska (2011) 'Indebtedness of Microcredit Clients in Bosnia and
Herzegovina: Results from a comprehensive field study'. EFSE Development Facility.

- 99 -

McIntosh, C. and B. Wydick (2005) 'Competition and microfinance'. Journal of Development


Economics, 78(2), 271298.
Mersland, R. and R.. Strm (2010) 'Microfinance Mission Drift?'. World Development,
38(1), 2836.
Morvant-Roux, S. (2009) 'Accs au microcrdit et continuit des dynamiques d'endettement
au Mexique: Combiner l'anthropologie conomique avec l'conomtrie [Access to
microcredit and continuity of indebtedness dynamics in Mexico: Combining economic
anthropology with econometrics]'. Revue Tiers Monde, 196, 109130.
Ocasio, W. (1997) 'Towards an attention-based view of the firm'. Strategic Management
Journal, 18(S1), 187206.
Paxton, J., D. Graham and T. Cameron (2000) 'Modeling Group Loan Repayment Behavior:
New Insights from Burkina Faso'. Economic Development and Cultural Change, 48(3),
639655.
Rahman, A. (1999) 'Micro-credit Initiatives for Equitable and Sustainable Development: Who
Pays?'. World Development, 27(1), 6782.
Rinaldi, L. and A. Sanchis-Arellano (2006) 'Household debt sustainability. What explains
household non-performing loans?: An empirical analysis'. European Central Bank Working
Paper Series No. 570, Frankfurt am Main.
Roesch, M. and O. Hlis (2007) 'La microfinance: outil de gestion du risque ou de mise en
danger par surendettement? [Microfinance: A tool for risk management or for the creation
of over-indebtendness risks]'. Autrepart, 44, 119140.
Schicks, J. (in press) 'Microfinance Over-Indebtedness: Understanding its Drivers and
Challenging the Common Myths'. Oxford Development Studies
Sharma, M. and M. Zeller (1997) 'Repayment performance in group-based credit programs in
Bangladesh: An empirical analysis'. World Development, 25(10), 17311742.
Steel, W. (2010) 'Ghanas Microfinance Industry: Evolution and Prospects. Draft August
2010'. ISSER, University of Ghana, Prepared for IFC Access to Finance Advisory Microfinance. Accra.
Vandone, D. (2009) 'Consumer credit in Europe, Risks and opportunities of a dynamic
industry'. Contributions to economics, Heidelberg: Physica.

- 100 -

Vogelgesang, U. (2003) 'Microfinance in Times of Crisis: The Effects of Competition, Rising


Indebtedness, and Economic Crisis on Repayment Behavior'. World Development, 31(12),
20852114.

- 101 -

Figure 1: Measuring over-indebtedness by customer protection standards

Source: Schicks (forthcoming).

Figure 2: Prevalence and acceptability of sacrifices among microborrowers


100%

Percent

80%

60%

% of borrowers in total sample that


experience this sacrifice

40%

% of borrowers that make this sacrifice


and consider it unacceptable

20%

0%
1

10

Sacrifices ranked by their prevalence

- 102 -

11

12

13

Table 1: Descriptive statistics of borrower sample


Variable
Gender (F%)
Age
Household_size
Avg_monthly_income_approx_value
Total_assets
Average_maturity_weighted
Lending methodology (Group%)
Total_amount_of_debt_disbursed
Debttoincome_ratio_flow
Number_of_MFIs_crossborrowing
Delinquency_status_spot (Delinquent%)
Delinquency_status_weeks
Delinquency_status_whole_loan_term (Delinquent%)
Aggregate_number_of_late_payments_all_loans
Default_over_year_rough_proxy (Default%)

Units
Percent female
Years
Number of persons
Ghana Cedis
Ghana Cedis
Months
Percent group
Ghana Cedis
Percent
Number of MFIs
Percent delinquent
Weeks
Percent delinquent
Number of payments
Percent defaulted

Observations
531
520
531
530
524
527
531
529
509
531
518
518
522
522
529

Global Mean
Standard Deviation
72.2
0.449
40.1
8.596
4.7
2.148
645.1
575.481
14,997.1
21,908.090
8.0
4.118
47.6
0.500
1,408.7
1,286.748
41.6
37.075
1.1
0.282
7.0
0.256
0.2
1.243
21.3
0.410
0.4
0.928
0.9
0.096

Mean
Over-indebted Not Over-Indebted
71.0
72.7
40.3
40.0
4.9
4.7
544.8
687.8
11375.8
16500.2
8.1
8.0
51.0
46.1
1339.7
1437.8
48.0
38.8
1.1
1.1
13.8
4.2
0.6
0.1
31.4
17.2
0.6
0.3
2.7
0.2

* T-test for equal means between over-indebted and not over-indebted group
** Contingency analysis in Stata is unweighted. We have checked weighted Chi-Square results with SPSS which implies no substantial changes to results

- 103 -

T-test*
-0.34
-1.21
2.75***
2.87***
-0.12
0.88
-2.32**
-0.98
-2.83***
-3.34***
-

Chi-Square**
Cramer's V
0.04
-0.05
0.16***
0.16***
0.15***

Table 2: The struggles and over-indebtedness of Ghanaian micro-borrowers

Numbers of borrowers
Percent of borrowers

Not struggling
137
26%

Struggling but not


over-indebted
236
44%

Over-indebted
158
30%

Total
531
100%

Table 3: Analysis of the sacrifices of microborrowers by prevalence, acceptability and frequency

Sacrifices
Work more than usual
Postpone important expenses
Deplete savings
Reduce food quantity/quality
Use family/friends support
Suffer psychological stress
Reduce education
Sell or pawn assets
Borrow anew to repay
Feel threatened/harassed
Suffer from shame or insults
Other
Seizure of assets

Numbers of borrowers
325
240
179
96
67
51
26
21
20
15
14
6
4

Percentage of borrowers in
sample
61%
45%
34%
18%
13%
10%
5%
4%
4%
3%
3%
1%
1%

Percentage of borrowers
among those making this
sacrifice that find it
unacceptable
32%
33%
38%
73%
72%
80%
80%
90%
85%
100%
100%
89%
100%

- 104 -

Percentage of borrowers
among those making this
sacrifice, that experience it
more than 3 times
71%
60%
29%
62%
21%
52%
13%
38%
49%
44%
24%
51%
24%

Table 4: Differences in sacrifices between male and female microborrowers


Sacrifices
Work more than usual
Postpone important expenses
Deplete savings
Reduce food quantity/quality
Use family/friends support
Suffer psychological stress
Reduce education
Sell or pawn assets
Borrow anew to repay
Feel threatened/harassed
Suffer from shame or insults
Other
Seizure of assets

% of Male
57.4
41.3
38.7
13.6
9.7
10.3
3.9
6.5
7.7
7.1
3.9
1.3
1.3

% of Female
60.1
45.5
31.1
18.4
12.5
8.2
4.8
2.7
2.9
1.9
2.7
1.1
0.8

Cramer's-V*
-0.02
-0.04
0.07*
-0.06
-0.04
0.03
-0.02
0.09**
0.11**
0.13***
0.03
0.01
0.02

* Unweighted contingency analysis. We have checked weighted Chi-Square results with SPSS.
The only relevant difference is that the difference for reducing food quantity/quality becomes
significant at a 10% level. In turn, the difference in depleting savings is not significant.

Table 5: Differences in sacrifices between group lending customers and individual


borrowers
Sacrifices
Work more than usual
Postpone important expenses
Deplete savings
Reduce food quantity/quality
Use family/friends support
Suffer psychological stress
Reduce education
Sell or pawn assets
Borrow anew to repay
Feel threatened/harassed
Suffer from shame or insults
Other
Seizure of assets

% of Indiviual
58.3
42.6
27.9
16.4
7.1
7.1
3.5
6.5
4.8
3.9
2.6
1.0
1.3

% of Group
60.7
46.6
41.1
17.8
18.3
11.4
5.9
2.7
3.7
2.7
3.7
1.4
0.5

Cramer's-V*
-0.02
-0.04
-0.14***
-0.02
-0.17***
-0.08*
-0.06
0.09**
0.03
0.03
-0.03
-0.02
0.04

* Unweighted contingency analysis. We have checked weighted Chi-Square results with SPSS.
The only relevant difference is that the difference for selling/pawning assets is not significant.

- 105 -

Table 6: Logistic regression on over-indebtedness


Dep. Var. Over-indebtedness
Total_amount_of_debt_di_in_hds
Debttoincome_ratio_flow
Number_of_M FIs_crossborrowing
Delinquency_status_rank

1
0.979**
(0.010)
1.009***
(0.003)
1.552

2
0.980
(0.015)
1.011***
(0.004)
1.294
(0.514)
3.646***

(0.595)
3.453***
(1.213)

(1.439)

Controls
Gender_rank

1.042
(0.273)
0.999
(0.014)
1.109*
(0.061)
0.997
(0.057)
0.988**
(0.006)
1.033
(0.026)
0.820
(0.204)

Age
Household_size
Avg_monthly_income_ran
Total_assets_in_thds
Average_maturity_weighted
Group_or_individual_customer_ran

1.097*
(0.054)

0.986**
(0.006)

Lender
controls

0,070
498

0,105
462

0,085
494

0,113
460

0,100
489

71,5%

71,9%

72,1%

72,6%

71,3%

_IMain_lend_6

Cases correctly predicted

3.599***
(1.289)

Borrower
controls

_IMain_lend_5

Nagelkerke's R
Observations (N)

1.087
(0.284)
0.998
(0.014)
1.098*
(0.060)
0.992
(0.060)
0.988**
(0.006)
1.036
(0.027)
0.712
(0.229)
0.844
(0.335)
1.018
(0.468)
1.537
(0.612)
0.942
(0.348)
0.157**
(0.140)
Added

5
0.988
(0.010)
1.010***
(0.003)

0.150**
(0.118)

_IMain_lend_3

Controls

4
0.977
(0.015)
1.011**
(0.004)
1.406
(0.565)
3.819***
(1.500)

0.975
(0.285)
1.075
(0.402)
1.753*
(0.596)
0.861
(0.291)
0.179***
(0.083)

_IMain_lend_2

_cons

3
0.974**
(0.010)
1.009***
(0.003)
1.877
(0.777)
3.583***
(1.231)

0.224***
(0.091)
Excluded

Odds Ratios. Robust Standard Errors in parenthesis.


* ** and *** denote significance at the 10% 5% and 1% level.

- 106 -

1.015
(0.301)
1.088
(0.419)
1.481
(0.520)
0.862
(0.296)
0.214***
(0.077)
Added
if relevant

Table 7: Logistic regression split by gender and lending methodology


Dep. Var. Over-indebtedness
Total_amount_of_debt_di_in_hds
Debttoincome_ratio_flow
Number_of_MFIs_crossborrowing
Delinquency_status_rank

1
Female
0.985
(0.020)
1.007
(0.005)
1.376
(0.661)
5.218***
(3.089)

Male
0.932**
(0.031)
1.035***
(0.011)
2.213
(3.391)
2.755
(1.772)

Controls
Gender_rank
Age
Household_size
Avg_monthly_income_ran
Total_assets_in_thds
Average_maturity_weighted
Group_or_individual_customer_ran

_IMain_lend_2
_IMain_lend_3
_IMain_lend_5
_IMain_lend_6
_cons
Controls
Nagelkerke's R
Observations (N)

0.993
(0.017)
1.092
(0.074)
0.998
(0.074)
0.984**
(0.008)
0.985
(0.052)
1.070
(0.417)
1.168
(0.594)
1.423
(0.837)
1.893
(0.952)
1.211
(0.569)
0.218
(0.232)
Added

1.019
(0.032)
1.054
(0.106)
1.087
(0.119)
1.004
(0.012)
1.094**
(0.045)
0.218**
(0.156)
0.210*
(0.174)
0.257
(0.285)
0.794
(0.621)
0.378
(0.254)
0.076
(0.169)
Added

0.108
324

0.280
136

Odds Ratios. Robust Standard Errors in parenthesis.


* ** and *** denote significance at the 10% 5% and 1% level.

- 107 -

2
Group
Individual
0.936*
0.994
(0.033)
(0.015)
1.004
1.022***
(0.008)
(0.005)
1.821
1.149
(1.107)
(0.714)
2.643
4.033***
(1.652)
(2.155)
1.440
(0.601)
0.997
(0.024)
0.981
(0.092)
1.072
(0.106)
0.984
(0.011)
1.106**
(0.050)

0.755
(0.267)
0.995
(0.019)
1.252***
(0.098)
0.952
(0.074)
0.988*
(0.007)
0.982
(0.036)

3.147
(3.397)
3.960
(4.374)
4.552
(6.147)
3.870
(4.480)
0.029**
(0.044)
Added

0.395
(0.269)
1.325
(1.104)
1.124
(0.488)
0.763
(0.327)
0.212
(0.278)
Added

0.177
191

0.139
269

Table 8: Alternative measurements of delinquency


Measurements for repayment irregularities
Delinquency_status_rank
Delinquency_status_whole_loan_term
Delinquency_status_weeks
Aggregate_number_of_late_payments_all_loans
Default_rough_proxy

Time horizon
At time of survey
Over loan term
At time of survey
Over loan term
Over last year

Definition
Units
Respondent is at least 1 day late on any outstanding loan
Dummy
Respondent was at least 1 day late on any outstanding loan
Dummy
Number of weeks respondent was late on currently outstanding loans Weeks
Number of times the respondent was late on currently outstanding loans Number of payments
Respondent has experienced an assets seizure or was forgiven a loan
Number of defaults

- 108 -

Table 9: Clustered standard errors and alternative indicators of delinquency


Dep. Var. Over-indebtedness
Total_amount_of_debt_di_in_hds
Debttoincome_ratio_flow
Number_of_MFIs_crossborrowing
Delinquency_status_rank

1
0.977
(0.015)
1.011**
(0.004)
1.406
(0.565)
3.819***
(1.500)

2
0.977
(0.015)
1.011**
(0.005)
1.406
(0.614)
3.819***
(1.604)

Delinquency_status_weeks

3
0.977
(0.015)
1.011**
(0.004)
1.361
(0.537)

4
0.976
(0.015)
1.012***
(0.004)
1.471
(0.592)

5
0.977
(0.015)
1.012***
(0.004)
1.363
(0.594)

1.309***
(0.120)

Delinquency_status_whole_loan_te

2.153***
(0.561)

Aggregate_number_of_late_payment

1.446***
(0.180)

Default_over_year_rough_proxy
Controls
Gender_rank
Age
Household_size
Avg_monthly_income_ran
Total_assets_in_thds
Average_maturity_weighted
Group_or_individual_customer_ran

_IMain_lend_2
_IMain_lend_3
_IMain_lend_5
_IMain_lend_6
_cons
Controls
Nagelkerke's R
Observations (N)

6
0.968**
(0.016)
1.012***
(0.004)
1.345
(0.528)

23.193***
(19.890)
1.087
(0.284)
0.998
(0.014)
1.098*
(0.060)
0.992
(0.060)
0.988**
(0.006)
1.036
(0.027)
0.712
(0.229)
0.844
(0.335)
1.018
(0.468)
1.537
(0.612)
0.942
(0.348)
0.157**
(0.140)
Added

1.087
(0.253)
0.998
(0.008)
1.098***
(0.033)
0.992
(0.065)
0.988**
(0.006)
1.036
(0.030)
0.712
(0.178)
0.844
(0.216)
1.018
(0.306)
1.537**
(0.288)
0.942
(0.152)
0.157**
(0.116)
Added

1.160
(0.301)
1.001
(0.015)
1.095
(0.060)
0.992
(0.059)
0.988**
(0.006)
1.033
(0.028)
0.695
(0.225)
0.812
(0.322)
0.934
(0.432)
1.449
(0.582)
0.879
(0.325)
0.162**
(0.147)
Added

1.279
(0.326)
1.002
(0.015)
1.096
(0.062)
0.999
(0.059)
0.988**
(0.006)
1.029
(0.025)
0.672
(0.208)
0.795
(0.302)
0.883
(0.388)
1.490
(0.585)
0.911
(0.329)
0.122**
(0.109)
Added

1.241
(0.320)
1.002
(0.015)
1.097
(0.062)
1.005
(0.059)
0.988**
(0.006)
1.028
(0.025)
0.687
(0.218)
0.789
(0.308)
0.928
(0.407)
1.474
(0.576)
0.943
(0.338)
0.130**
(0.118)
Added

1.340
(0.339)
1.003
(0.015)
1.087
(0.060)
1.000
(0.060)
0.987**
(0.006)
1.044*
(0.026)
0.745
(0.238)
0.943
(0.368)
1.022
(0.458)
1.563
(0.625)
0.900
(0.327)
0.142**
(0.127)
Added

0.113
460

0.113
460

0.108
460

0.106
462

0.113
462

0.106
467

Odds Ratios. Robust standard errors in parenthesis. In regression 2, standard errors are clustered by lending institutions.
* ** and *** denote significance at the 10% 5% and 1% level.

- 109 -

Table 10: OLS regression on a threshold-free sacrifice score


Dep. Var. Sacrifice Score

1
2
3
4
5
-0.00388**
-0.00366* -0.00478***
-0.00408*
-0.002
(-0.00154) (-0.00215) (-0.00165) (-0.00218) (-0.00157)
0.00179*** 0.00216*** 0.00178*** 0.00210*** 0.00195***
(-0.000567) (-0.000789) (-0.000577) (-0.000797) (-0.000559)
0.095
0.052
0.132
0.068
(-0.0834)
(-0.0907)
(-0.0838)
(-0.0851)
0.287***
0.295***
0.292***
0.303***
0.291***
(-0.0826)
(-0.0907)
(-0.0808)
(-0.0902)
(-0.0848)

Total_amount_of_debt_di_in_hds
Debttoincome_ratio_flow
Number_of_MFIs_crossborrowing
Delinquency_status_rank
Controls
Gender_rank

0.007
(-0.0498)
0.000
(-0.00274)
0.0206*
(-0.0109)
0.001
(-0.00951)
-0.00194**
(-0.000785)
0.006
(-0.00527)
-0.039
(-0.0495)

Age
Household_size
Avg_monthly_income_ran
Total_assets_in_thds
Average_maturity_weighted
Group_or_individual_customer_ran

_IMain_lend_2

0.082
(-0.154)

-0.006
(-0.0565)
0.014
(-0.073)
0.115
(-0.0716)
-0.030
(-0.0635)
0.118
(-0.0963)

Borrower
controls

Lender
controls

0.080
0.050
462

0.060
0.050
494

_IMain_lend_3
_IMain_lend_5
_IMain_lend_6
_cons

0.160*
(-0.0879)
Excluded

Controls

0.050
0.040
498

R
Adjusted R
Observations (N)
Robust Standard Errors in parenthesis.
* ** and *** denote significance at the 10% 5% and 1% level.

- 110 -

0.016
(-0.0503)
0.000
(-0.00276)
0.0186*
0.0205**
(-0.0111)
(-0.0101)
0.000
(-0.0101)
-0.00191** -0.00224***
(-0.000801) (-0.000742)
0.007
(-0.00576)
-0.067
(-0.063)
-0.030
(-0.0767)
0.000
(-0.0904)
0.084
(-0.0804)
-0.011
(-0.0704)
0.088
0.160***
(-0.177)
(-0.0567)
Added
Added
if relevant

0.080
0.050
460

0.070
0.060
492

Table 11: Sensitivity analysis on repayment irregularities as an indicator of over-indebtedness


Indicator of repayment irregularities
Delinquency_status_ Delinquent_more_th Delinquency_status_ Aggregate_late_payme Default_over_year_r
rank
an_1_week
whole_loan_term
nts_more_than_2
ough_proxy
Correct overindebtedness status

72%

71%

68%

71%

71%

Correctly classified IF
over-indebted

4%

2%

9%

2%

1%

% over-indebtedness in
sample

7%

4%

21%

4%

1%

Table 12: Sensitivity analysis on debt-to-income ratio as an indicator of over-indebtedness


10%

Threshold for debt-to-income ratio


20%
30%
40%

50%

60%

Correct over-indebtedness
status

36%

47%

52%

57%

63%

65%

Correctly classified IF
over-indebted

28%

24%

17%

13%

11%

7%

% over-indebtedness in
sample

89%

71%

52%

39%

28%

19%

- 111 -

Appendix 1: List of borrower sacrifices


Interviewers asked each respondent about the following list of sacrifices
1) Reduce food quantity/quality (cut down eating)
2) Reduce education (e.g. taking children out of school)
3) Work more than usual (e.g. take additional labor, work longer hours, on Sundays, and when ill)
4) Postpone important expenses (e.g. for health, housing, business assets etc.)
5) Deplete your financial savings (e.g. money in the house or in a savings account)
6) Borrow anew to repay (take an additional loan from another lender)
7) Sell or pawn assets (e.g. jewelry, cattle, productive or household assets)
8) Seizure of assets (MFI takes property by force to make up for missed payment)
9) Use family/friends' support to repay
10) Suffer from shame or insults (also gossip about you/exclusion from a contract)
11) Feel threatened/harassed by peers/family/loan officer
12) Suffer psychological stress yourself or in your marriage
13) Other

Respondents ranked the acceptability and frequency of each sacrifice on a scale from 1 to 4.
Acceptability: Easily acceptable, Only just acceptable, Not really acceptable, Not acceptable at
all.
Frequency: Once in past year, 1-3 times in past year, > 3 times but not often, Frequently in past
year

- 112 -

Appendix 2: Marginal effects for full sample logistic regression


Variable
Total_amount_of_debt_di_in_hds
Debttoincome_ratio_flow
Number_of_MFIs_crossborrowing
Delinquency_status_rank*
Gender_rank*
Age
Household_size
Avg_monthly_income_ran
Total_assets_in_thds
Average_maturity_weighted
Group_or_individual_customer_ran*
_IMain_lend_2*
_IMain_lend_3*
_IMain_lend_5*
_IMain_lend_6*

dy/dx
Std.Err.
z
-0.005
0.003
0.002
0.001
0.068
0.081
0.313
0.094
0.017
0.053
0.000
0.003
0.019
0.011
-0.002
0.012
-0.003
0.001
0.007
0.005
-0.068
0.065
-0.034
0.077
0.004
0.093
0.091
0.089
-0.012
0.073

P>|z|
-1.490
2.590
0.850
3.320
0.320
-0.100
1.700
-0.140
-2.140
1.350
-1.050
-0.430
0.040
1.030
-0.160

(*) dy/dx is for discrete change of dummy variable from 0 to 1

- 113 -

95% C.I.

0.136
0.010
0.397
0.001
0.751
0.917
0.090
0.887
0.033
0.177
0.293
0.664
0.968
0.305
0.870

-0.011
0.001
-0.090
0.128
-0.087
-0.006
-0.003
-0.025
-0.005
-0.003
-0.196
-0.185
-0.178
-0.083
-0.155

0.001
0.004
0.226
0.498
0.121
0.005
0.040
0.022
0.000
0.017
0.059
0.118
0.185
0.266
0.131

X
14.265
42.761
1.069
0.065
0.282
40.209
4.750
3.688
15.025
8.042
0.516
0.306
0.161
0.179
0.146

Chapter 4
Over-indebtedness in Microfinance
An Empirical Analysis of Related Factors on the Borrower Level
Jessica Schicks*
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management
Centre Emile Bernheim, Universit Libre de Bruxelles

This paper analyses the over-indebtedness of microborrowers in Ghana. It defines over-indebtedness


from a customer protection perspective and considers borrowers over-indebted if they continuously
struggle with repayment and experience unacceptable sacrifices related to their debt. It finds that
poorer microborrowers are more likely to be over-indebted. The risk of over-indebtedness further
increases with the occurrence of adverse economic shocks to a borrowers income or expenses. The
likelihood of over-indebtedness is higher for borrowers with low returns on their investment and if
borrowers use loans, at least in part, for non-productive purposes. It is higher for borrowers with a
low, debt-specific financial literacy. General financial literacy has negative effects on overindebtedness. We find no effect for mere numeracy. The paper also breaks down the relationship of
the above factors to the specific sacrifices that borrowers make, to how frequently they repeat them
and to how acceptable sacrifices are to borrowers.

Paper under revise and resubmit at World Development

* The author would like to thank the Independent Evaluation Department of KfW Entwicklungsbank
and the ACCION Center for Financial Inclusions Smart Campaign for their contribution to study
design and for financing the empirical research project. She is indebted to the Ghanaian
microfinance institutions who have shared their data and access to their clients to make this database
possible. Prof. Marek Hudon, Prof. Ariane Szafarz, and Prof. Oscar Bernal, as well as seminar
participants at the University of Lubumbashi, Congo, have provided valuable comments.

- 114 -

1. Introduction
The literature on consumer finance in developed countries literature has extensively analysed
why borrowers enter into burdensome levels of debt and for what borrowers the risk of
personal over-indebtedness is particularly high (Lea et al., 1995; Webley and Nyhus, 2001;
Bridges and Disney, 2004; Lusardi and Tufano, 2009). The microfinance industry, which
provides financial services to poor populations in developing countries, has only recently
become aware of the risk of over-indebtedness. There is a limited amount of research on
delinquency and default in microfinance (Vogelgesang, 2003; Schreiner, 2004; Godquin,
2004), but hardly any research on over-indebtedness. With increasing concerns about negative
effects of microfinance on customers, at a time where trust in the positive impact of
microfinance is weakening (Duvendack et al., 2011; Karlan and Goldberg, 2011), overindebtedness is one of the most pressing challenges facing the microfinance industry. It
endangers at the same time the sustainability of microfinance institutions (MFIs) and their
social impact. The lack of research on over-indebtedness in microfinance thus presents a
substantial research gap.
Developing measures to avoid over-indebtedness requires a sound understanding of what
drives it. To tailor effective solutions to the challenge, we need to understand the situations
where over-indebtedness is most likely to exist and which borrowers are most at risk. To our
knowledge, there are no academic studies that analyse which factors are related to overindebtedness in microfinance. It has not been tested which of the factors identified in
consumer finance research could equally apply to the microfinance low-income country
context.
This paper analyses factors related to over-indebtedness in microfinance. Taking a customer
protection perspective to enhance our understanding of the phenomenon, this paper
contributes to the development of customer protection measures to combat over-indebtedness.
We work with a unique primary dataset from a survey among 531 microborrowers in Accra,
Ghana. We use logistic regression analysis to identify how poverty, adverse shocks, loan
returns and financial literacy relate to over-indebtedness. We detail the relationship of these
factors to the individual sacrifices that borrowers experience. By means of ordered logit
regressions, we shed light on the role that these factors play concerning how acceptable
sacrifices are to borrowers and how frequently borrowers repeat specific sacrifices.

- 115 -

For the geographical and historical context of this study, we find that poorer borrowers, in
terms of income and assets, are more likely to be over-indebted, as are borrowers who
experience adverse economic shocks, who use their loans, at least in part, for non-productive
purposes, and whose investments into microenterprises produce low returns. According to
their own perceptions, for nearly sixty percent of borrowers, their returns are not sufficient to
repay their loans. A lack of debt-specific financial literacy also comes with a higher likelihood
of being over-indebted, but surprisingly, the opposite is true for general financial literacy. The
same factors that relate to over-indebtedness also relate to which sacrifices borrowers employ
as coping strategies for their debt, to how acceptable sacrifices are to borrowers, and to how
frequently borrowers repeat sacrifices.
The correlations of borrower-level factors with over-indebtedness confirm prior expectations
based on theory and on empirical research in non-microfinance environments, for what might
be the causes of over-indebtedness. These related factors are therefore likely to be causes of
over-indebtedness. Because measures against over-indebtedness should address the borrowers
that are most at risk and would optimally address the causes of the over-indebtedness
phenomenon, our findings have important implications for the development of measures
against over-indebtedness. They indicate that the spectrum of measures of over-indebtedness
may be broader than expected.
The next section reviews the literature for factors related to over-indebtedness. In the absence
of longitudinal studies, these factors represent the best candidates for potential causes of overindebtedness. Section 3 describes our dataset and our econometric approach. Section 4
contains a presentation and discussion of our empirical findings. Concluding remarks and
recommendations are provided in section 5.

- 116 -

2. Factors related to over-indebtedness


To develop measures against the problem of over-indebtedness, it is necessary to identify the
most effective levers. These measures should address the cases with the highest risk for overindebtedness; optimally, they would address the main drivers of over-indebtedness. The
former requires knowledge concerning which customers are most at risk of over-indebtedness,
i.e., we need to identify borrower characteristics related to over-indebtedness. The latter
requires knowledge about the causes of over-indebtedness. However, there are no panel data
for dealing with the challenges of endogeneity and causal interactions. The best possible
approximation of the causes of over-indebtedness is thus to depart from theory and our
knowledge of the causes of over-indebtedness in the context of high-income country
consumer finance and empirically analyse factors that are related to over-indebtedness in the
given microfinance context. A statistically significant relationship in multivariate analysis that
confirms theoretical expectations and prior empirical findings is currently the best available
indication that a factor might be a cause of over-indebtedness.1
This section identifies the socio-demographic and economic borrower characteristics that are
related to over-indebtedness according to the microfinance literature and high-income country
consumer finance research. It also analyses business and loan-related factors and financial
literacy, exploring the theory of what might be the causes of over-indebtedness. It develops
four hypotheses for factors related to over-indebtedness in microfinance. We empirically test
these hypotheses in section 4.
There is no consensus on a definition of personal over-indebtedness in the literature. From the
viewpoint of risk management, delinquency or default are common indicators for overindebtedness. From the viewpoint of customer protection, these indicators represent only late
stage consequences of over-indebtedness. Over-indebted borrowers often struggle
significantly with their loans before they reach the stage of manifest repayment problems.
Following Schicks (in press), this paper therefore uses an over-indebtedness definition that
takes the borrowers experiences with debt into account: A microfinance customer is overindebted if he/she is continuously struggling to meet repayment deadlines and has to make
unduly high sacrifices related to his/her loan obligations that have more than transitory
effects.

1 See Schicks (in press) for a framework of the causes of over-indebtedness.

- 117 -

The empirical research analysed in this section works with a range of measurements of overindebtedness, problem debt, repayment problems, or high indebtedness. While the
measurement is prone to influence empirical findings, factors that are related to any of these
phenomena are also likely candidates for factors related to over-indebtedness according to the
above sacrifice-based definition. This section therefore reviews factors that the literature
identifies to be related to debt problems, independently of the exact measure of debt problems
used in each study. The studies we review also vary according to their geography and cultural
context, the type of debt products they include, and the type of borrowers they analyse. This
paper is the first to test the transferability of their findings to a microfinance setting.

Socio-demographic factors
A borrowers risk of personal over-indebtedness varies with his/her socio-demographic
background. For example, the borrowers age is a frequently studied characteristic in
consumer finance research is age. Although evidence is mixed, in the UK and the US,
younger people seem to have a higher risk of indebtedness (Livingstone and Lunt, 1992; Lea
et al., 1993; Drentea and Lavrakas, 2000; Bridges and Disney, 2004). This is consistent with a
life-cycle approach to borrowing that assumes consumption smoothing over an individuals
life time with debts acquired at a young age based on expected future income (Friedman,
1957; Modigliani, 1980). Vogelgesang (2003) confirm this finding in a microfinance-specific
study in Bolivia. In two studies by Canner and Luckett (1991) and Tokunaga (1993) with
subjects from the US, the age effect is not significant.
Another influence factor is the borrowers household situation. Livingstone and Lunt (1992)
find that in the UK, having more children reduces risk, assuming that children create
discipline in their parents. The common view is, however, that having many children and
younger children is positively associated with debt risks (Canner and Luckett, 1991; Lea,
Webley and Levine, 1993; Lea, Webley and Walker, 1995). This may be because children
augment the fixed expenses of a household, while at the same time reducing its income
generating capacity. Except for Tokunaga (1993), most studies find increased overindebtedness among single adult households, particularly after a divorce (Canner and Luckett,
1991; Lea, Webley and Levine, 1993; Webley and Nyhus, 2001; Bridges and Disney, 2004;
Del-Ro and Young, 2005). The problem might be the household dependence on a single
income, combined with lingering habits and expenses that, in divorce cases, were appropriate
for a different economic situation. This finding holds in the microfinance study conducted in
- 118 -

Bolivia by Vogelgesang (2003). Webley and Nyhus (2001) get closest to actually establishing
a causal relationship because their study uses a longitudinal approach based on panel data.
Several UK studies find higher debt risk for renters as opposed to home owners (Lea, Webley
and Walker, 1995; Bridges and Disney, 2004; Disney et al., 2008), an effect that is not
significant in the US-based multivariate analysis by Canner and Luckett (1991). According to
Del-Ro and Young (2005), Drentea and Lavrakas (2000), and Disney et al. (2008), ill health
is related to debt problems; however cause and effect are ambiguous. Finally, gender (Lea,
Webley and Levine, 1993) and ethnicity (Del-Ro and Young, 2005) have been correlated to
indebtedness.
Concerning education, Bridges and Disney (2004), Canner and Luckett (1991) and Tokunaga
(1993) find a higher risk of debt when a borrowers level of education is low. According to
Livingstone and Lunt (1992) the effect of education is not significant in the UK. Gonzalez
(2008) confirms the relationship of low education to over-indebtedness in microfinance in
Bolivia. More educated households might have a better economic standing in general but may
also be better able to make more advantageous financial decisions. The risk of overindebtedness could result from behavioural biases that make borrowers take on more debt
than is reasonable. Humans do not make fully rational decisions, sometimes due to a lack of
information, or due to a lack of cognitive abilities. They may be subject to temptations and
over-value present benefits. Laibson (1997) explains this with the theory of hyperbolic
discounting, where decision makers apply time-inconsistent discount rates. For example, in
taking a loan, a borrower may give too much weight to the pleasure of receiving immediate
cash and too little weight to the future burden of repayment. To a certain extent, a higher
educational level may help borrowers deal with these biases.

Economic factors
The academic debate on economic factors influencing over-indebtedness focuses, most
importantly, on income. Having a low income is a major cause of indebtedness in many
consumer finance studies in the US and UK. Lea et al. (1993) and Lea et al. (1995) categorise
British respondents into individuals without debt mild, with mild debt, and with serious debt.
They find that individuals on lower incomes are more likely to have any debt at all and that
borrowers in more serious debt situations have significantly lower incomes than mild debtors.
A study specifically among low-income households in the UK finds that income is one of the
- 119 -

main causes of arrears and that sustained low income is a key factor for debt problems to
become permanent (Bridges and Disney, 2004). Another UK study identifies low income as a
major factor related to various measurements of over-indebtedness, the main measurement
being one of individual instances of arrears on loans (Disney, Bridges and Gathergood, 2008).
Income is not significant in the US studies by Canner and Luckett (1991) and Tokunaga
(1993). Livingstone and Lunt (1992) find that the absolute amount of borrowing among US
households increases with income, but confirm that serious debt situations and repayment
difficulties mainly exist among low-income borrowers. Earning a low income probably
reduces a borrowers repayment capacity, due to a low absolute amount of cash inflow at a
borrowers disposition, and due to a larger share of total income being bound for essential
living expenses. In microfinance, Paxton et al. (2000) find a higher risk for repayment
problems among rural borrowers in Burkina Faso that could be related to their lower incomes.
In contrast, in South Africa, Collins (2008) finds that indebtedness is negatively related to
income only in urban areas.
Alternatively, income uncertainty can be a driver of debt problems (Webley and Nyhus,
2001). If income is instable but the instalment schedule for a loan is fixed, borrowers may not
have the means to serve their debt in low-income periods, even if their average annual income
might be sufficient to deal with a given debt load.
Moreover, the literature points to an association between wealth levels and debt. With an
ordered logit regression based on the British Household Panel Survey, Del-Ro and Young
(2005) find that a households financial wealth is one of the key explanatory variables of selfreported financial distress. Savings are negatively associated with over-indebtedness or
arrears and the possession of savings makes it much more likely for over-indebted borrowers
to get out of their difficulties (Disney, Bridges and Gathergood, 2008). Sharma and Zeller
(1997) confirm the relationship of wealth to debt among microfinance borrowers. Based on a
survey among three group lending programmes in Bangladesh, they find that land ownership
as a proxy for wealth on a group level is significantly and positively related to repayment
performance: the lower the groups wealth the higher the probability of repayment problems.
Analysing data from a quasi-experimental survey, equally from Bangladesh, Godquin (2004)
show that the total productive assets of a household are positively related to repayment
performance. They include self-employment in agriculture and the number of landed relatives
as further proxies of wealth, reinforcing the finding that wealth prevents repayment problems.
Wealthier borrowers are more likely to have assets to liquidate that they can spare without
- 120 -

incurring significant sacrifice as a result. In line with the findings of Disney et al. (2008), the
positive effect of wealth on a borrowers repayment capacity is likely to be most important for
liquid assets, such as cash savings that could be used to reimburse a loan in periods with
insufficient income. Less liquid assets may still enhance repayment capacity if a borrower
sells or pawns them. In addition to the risk buffering effects of wealth, it may be that
wealthier borrowers in the microfinance context have access to investment projects with a
better risk-return profile (Godquin, 2004).
In terms of labour market status, being unemployed or in an instable or part-time position is
related to higher debt in the UK (Lea, Webley and Walker, 1995). This effect is probably
related to that of a borrowers income and income volatility.
Aside from the economic background characteristics of a borrower, adverse shocks to the
income or expenses of borrowers reduce their capacity to repay debt and can trigger a
situation of over-indebtedness. Among UK consumers, Disney et al. (2008) identify loss of
employment and marital breakup as key examples of adverse shocks to the income and
expenses of borrowers that have a high risk of leading to over-indebtedness in terms of
several arrears-based and self-reported measures. Stone and Maury (2006) analyse consumer
indebtedness among enlisted US Airforce members, even if those may be little representative
for microfinance borrowers. They find that life altering events such as childbirth or relocation
were significantly related to consumer indebtedness, but also instances of not being able to
pay medical expenses in the family. Bouquet et al. (2007) find that the main reasons for credit
problems mentioned by microborrowers in Madagascar are sudden increases in expenses or
loss of income. Gonzalez (2008) provides an extensive discussion of the influence of adverse
shocks on repayment capacity in microfinance and tests this factor empirically, using data
from a household survey in Bolivia that was conducted in 1997 to 2000, around the times of
the Bolivian microfinance crisis. The research identifies adverse shocks as one of the few
factors that explain the difference between borrowers who are not over-indebted and those
who are over-indebted while willing to repay (over-indebtedness defined as unexpected
outcomes of loan contracts, mainly in the form of repayment irregularities). However, in other
regressions, explaining costly actions taken by borrowers, adverse shocks were not
significant. In this paper, we define shocks as irregular events such as sudden expenses for
emergencies, expected but irregular lump sum expenses (e.g., wedding ceremonies), or
sudden impermanent drops in income (e.g., due to illness). While borrowers may be able to
adapt to longer-term reductions in income by reducing their consumption or shifting their
- 121 -

income generating capacity, they may not have the means or financial buffer to deal with
sudden income reductions that occur unexpectedly. Human beings tend to display habit
persistence that prevents them from immediately adjusting their consumption in response to
income cuts (Brown, 1952). This effect is similar to but distinct from that of generally volatile
incomes. Similarly, adverse shocks to a borrowers expenses, such as large lump sums
required either for expected events (e.g., school fees, marriages), or unexpected shocks (e.g.,
medical expenses, funerals), may consume most of a borrowers income in a given period and
may not leave sufficient cash flow for debt repayment. In the self reports of Madagascan
microborrowers about the causality behind debt problems, sudden drops in expected income
appear nearly twice as much (35%) as rises in expenses (18%) (Bouquet et al., 2007).
There are two main factors that emerge from this analysis: material poverty and adverse
economic shocks. The findings discussed above suggest that over-indebtedness is related to
material poverty in the form of low incomes and low asset ownership. They also indicate a
higher risk of over-indebtedness following unexpected reductions in income or sudden lump
sum expenses. We test two hypotheses relating to economic factors to confirm that both a
borrowers general economic situation and sudden events that negatively affect their income
and expenses are independently related to over-indebtedness. We test the influence of material
poverty separately for income and for asset ownership.
H1: Poorer borrowers are more likely to be over-indebted.
H1a: A borrowers income is negatively related to his/her likelihood of overindebtedness.
H1b: A borrowers ownership of assets is negatively related to his/her likelihood of
over-indebtedness.
We test the influence of the following adverse economic events: unexpected shocks to
expenses, lump sums such as school fees that are expected but nevertheless difficult to
manage, and shocks to income.
H2: Borrowers who experience adverse economic shocks are more likely to be overindebted.
H2a: Unexpected shocks to a borrowers expenses are positively related to his/her
likelihood of over-indebtedness.

- 122 -

H2b: Expected lump sum expenses are positively related to a borrowers likelihood of
over-indebtedness.
H2c: Unexpected shocks to a borrowers income are positively related to his/her
likelihood of over-indebtedness.

Business and loan-related factors


Finally, as microcredit was originally (and often still is) primarily directed towards productive
purposes, i.e., the investment into microenterprises, returns on this investment were supposed
to enable microborrowers to repay their debt. Low returns on this investment can increase the
difficulties of borrowers coping with their debt. Gonzalez (2008) finds that only 24% of
borrowing households can repay their debt from their general income and loan returns,
without having to undertake costly actions. Karlan and Zinman (2009) conduct a randomized
field experiment in Manila to analyse the impact of microcredit and find only limited
evidence that the profits of microborrowers increase at all. If so, that is rather through
shrinking their businesses by laying off unproductive staff than through growing as a results
of the microloan being invested. The return on an investment (ROI) depends on the specific
investment, for example the borrowers choice of what to invest in, the market conditions, the
general condition of the microenterprise, the borrowers entrepreneurial skill and other factors
(de Mel et al., 2008). Additionally, the ROI may depend on the general characteristics of the
enterprise, for example on its sector of activity. Some sectors of activity may have higher
returns than others, i.e., they may be more profitable. This may result, for example, from a
relaxed competitive situation in a sector, due to entry barriers that allow businesses to exploit
monopoly rents. Hulme (2007) claims that the high returns on the investment of
microborrowers are nothing but a myth, most borrowers investing into "low-return activities
in saturated markets. Sectors of activity might also differ in the timing of cash flows. A loan
for working capital used to restock a small shop with a quick turn over generates returns faster
than a loan invested in agriculture where farmers must wait until harvest to generate returns.
A highly profitable investment does not increase a microborrowers repayment capacity if
returns only materialise after the end of the instalment schedule. Schreiner (2004) finds
manufacturers in Bolivia to be nearly twice as risky borrowers as traders, i.e., they are less
likely to repay their loans. This might be a result of the profitability of trading businesses, but
it is also due to their faster turnover.

- 123 -

Furthermore, the ROI on the total loan amount goes down if a borrower decides to use the
loan (at least partially), for non-productive purposes. On the one hand, scholars increasingly
argue for the benefits of microlending for consumption purposes and income smoothing
(Collins et al., 2009). A loan that does not generate financial returns may, nevertheless, be
very beneficial to a borrower, for example, in avoiding periods of hunger by financing food in
a low-income period, or in avoiding a deterioration in health by financing medical expenses.
While the above mentioned randomized control trial by Karlan and Zinman (2009) does not
confirm high returns on microloans, it argues instead that microcredit has positive impact
through risk management and household-level investments. On the other hand, nonproductive loan use is considered a risk factor for repayment (Vogelgesang, 2003). A
practitioner-oriented analysis of MIX Market data indicates that MFI portfolios that contain
higher shares of consumption loans display a lower portfolio quality (Gonzalez, 2009).
Buying goods for immediate consumption does not create returns and thus does not help to
repay a loan on time and without excessive sacrifice. In certain cases, it is difficult to
distinguish which loan use is productive and which one is not (Collins, 2008). A loan for
medical expenses may avoid the loss of ones income generating capacity and may therefore
have an indirect positive financial return. Nevertheless, the loan inflicts the cost of interest
and fees on the borrower and does not create any extra repayment capacity compared to the
base scenario that prevails without illness/medical expenses and without the loan.
Finally, loan use is not the only loan-specific factor related to over-indebtedness. The
characteristics of a loan, such as the amount of interest charged, the instalment schedule, and
the lending policies of MFIs, also influence a borrowers ability to repay their debt. In this
papers sample population in Ghana, these factors vary mainly on the institutional level, i.e.,
between MFIs, rather than from one borrower to the other. Where they vary within MFIs, it is
mainly between group borrowers and individual borrowers. As a result, this paper does not go
into the detail of loan and lender influences on the risk of over-indebtedness. Instead, along
with the most common borrower level controls for loan characteristics, the empirical analysis
will control for general loan characteristics on the MFI level and for the lending methodology
(i.e., group or individual).
We test the hypothesis that microborrowers over-indebtedness is related to the return on
investment on their loans. We test if there is an independent influence of the borrowers
decision about loan use (i.e., at least partially for consumption or exclusively for productive
purposes) and of the returns on investing the loan into a productive activity.
- 124 -

H3: Borrowers with low returns on their loan are more likely to be over-indebted.
H3a: Using a loan for non-business purposes is positively related to a borrowers
likelihood of over-indebtedness.
H3b: Returns on investment in a borrowers productive activity are negatively related
to the borrowers likelihood of over-indebtedness.

Financial literacy
More specifically than education, a borrowers financial literacy, his/her understanding of
financial products, may be related to over-indebtedness. Empirical research indicates that the
cognitive abilities of an average individual are low compared to the complexity of financial
decisions. Tiwari et al. (2008) find that microfinance clients in India think about their loans in
terms of how much they owe on a weekly basis but know very little about their interest rate or
total interest expenses. This is in line with the theory of mental accounting (Thaler, 1985),
which shows that individuals think about their financial decisions in terms of budgets for
various mental expense categories but not in terms of interest rates and the time value of
money. Similarly, Atkinson et al. (2006) provide empirical support from the UK for people
making poorly informed financial choices, by for example not reading the terms and
conditions of contracts and by not comparing providers. Lusardi and Tufano (2009) discover
strikingly low levels of debt literacy across the U.S. population, for example serious difficulty
in grasping percentages. Those who are less debt literate bear a disproportionally large share
of avoidable costs (e.g., late fees). They are also more likely to be over-indebted. Schreiner
(2004) conducts a microfinance specific analysis of loan data to develop a scorecard for
predicting loans with more than 15 days arrears in an MFI portfolio in Bolivia. He finds that
the number of previous loans a borrower has had decreases the risk of arrears on a loan and
considers this indicator a proxy for the borrowers experience with loans and thus for their
loan literacy. On the other hand, the number of months since the first disbursement that can
also be an indicator of loan literacy is positively related to arrears. In a study by Paxton et al.
(2000) in Burkina Faso, experience with past loans has a negative relationship to repayment
performance, probably because other factors outweigh the effect of increasing literacy.
However, groups that received good loan literacy training and had good leadership had a
higher repayment performance, confirming the positive effect of financial literacy in avoiding
repayment problems.
- 125 -

This paper therefore tests the hypothesis that financial literacy is negatively related to
microborrowers over-indebtedness risk. The more borrowers understand loans, the better
borrowing decisions they can make and the less likely they are to take on debt beyond their
means.
H4: Borrowers with a higher level of financial literacy are less likely to be overindebted.
H4a: A borrowers numeracy is negatively related to his/her risk of over-indebtedness.
H4b: A borrowers general financial literacy is negatively related to his/her risk of overindebtedness.
H4c: A borrowers debt literacy is negatively related to his/her risk of overindebtedness.

- 126 -

3. Data and econometric framework


This paper is based on a unique primary database from Accra, Ghana. The Ghanaian
microfinance sector is relatively well developed for an African microfinance market. It is
dominated by regulated non-bank financial institutions (Savings & Loans Companies) and
features both a strong credit and substantial deposit offer. It addresses relatively typical
microfinance clients, mostly female, married, microentrepreneurs, on low incomes but not in
extreme poverty. Clients tend to have low to medium levels of education, but even those who
lack basic literacy skills tend to have the math skills required for their daily business
operations. The sector has seen substantial growth and is starting to experience competition
but is not known for signs of crisis (Steel, 2010).
In cooperation with the Independent Evaluation Department of the German development bank
KfW and with the Smart Campaign hosted by ACCION International, we conducted 531
structured interviews among microfinance borrowers2 at the end of 2010. The interviews were
anonymous. Five of Ghanas leading microfinance institutions, ProCredit Ghana, Opportunity
International Ghana, Sinapi Aba Trust, EB-ACCION and Advans Ghana contributed to this
random sample of microborrowers. They account for 83% of Ghanas microborrowers as
reported to the MIX Market3 in 2010 and for 95% of Ghanas gross microloan portfolio. We
oversampled delinquent customers as they were likely to augment the number of overindebted respondents in the sample and would allow for more variation in the over-indebted
subgroup.4 We corrected for both oversampling and response rates with sample weights for a
borrowers lending institutions, a borrowers delinquency status, and the lending methodology
of the borrowers main loan. There was no need to correct for gender.
Respondents reported details on all of their outstanding formal or informal loans at the time of
the interview, including the purpose for which they used the loans and their subjective
categorisation of return on their investment. Because borrowers would not be able to indicate
the ROI in quantitative terms, they could choose from three categories of returns: the first

2 To apply a common threshold for micro- and SME-borrowers across all MFIs, we designate all customers with active
personal loans below 5000 Ghana Cedis (GH; 1 GH=0.7 USD) as microborrowers. For most MFIs in the sample, all of
their borrowers fall into this category. A total of 87% of disbursed loan amounts in the sample are below 2000 GH.
3 www.mixmarket.org. The site provides self-reports of limited liability and is likely to overstate the role of our partner MFIs
in the local market as not all of the smaller MFIs may be reporting to the MIX.
4 With some MFIs, for groups over-sampling implied over-sampling delinquent groups rather than individuals.

- 127 -

indicating that the respondent did not permanently increase his/her earnings due to the
investment; the second indicating that earnings increased but not enough or not stable enough
to cover the instalments on the loan; and the third indicating that, on a regular basis, the
borrower earned significantly more income due to the investment or productive activity
financed with the loan. A fourth category captures the few loans (3% of our sample) that have
no productive component at all. Although a subjective measurement of returns is much less
precise than a quantitative percentage, the relatively broad categories and their verbal
description are tuned to avoiding endogeneity, i.e. a respondent will not indicate lower returns
because of being over-indebted. A borrower can report positive returns even if he is overindebted, if, for example, his repayment difficulties are related to adverse shocks rather than
to low returns (23% of the over-indebted in table 2 categorise ROI as a permanent and
significant increase in income). Or a borrower who repaid his loan without problems can
report having experienced no (12% of borrowers who are not over-indebted) or too little
increases in income (36% of borrowers who are not over-indebted). These borrowers may
have had sufficient other resources at their disposal to not need the loan returns for
investment. Our measurement of returns therefore focuses on the impact of returns on overindebtedness.
Moreover,

respondents

reported

their

socio-demographic,

economic and

business

characteristics, and took a test of financial literacy. Appendix 1 contains the financial literacy
questions based on Lusardi and Mitchell (2007) and Lusardi and Tufano (2009). In
consultation with local MFI staff and after a pilot of the questionnaire, the questions were
adapted to the local environment and to the aptitude of Ghanaian low-income borrowers. For
our measurement of over-indebtedness, respondents listed all sacrifices that they experienced
related to their loans over the course of one year (see Appendix 2 for the list of sacrifices).
They indicated how many times they experienced each of the sacrifices (Once in past year,
1-3 times in past year, >3 times but not often, or Frequently in past year)5 and weighted
their sacrifices according to their subjective judgment of how acceptable the experience was
to them (Easily acceptable, Only just acceptable, Not really acceptable, or Not
acceptable). The first two categories summarise as acceptable, and the latter as
unacceptable or not acceptable.

5 For a respondent cutting down on their food at several points for a week at a time, instead of every individual day, each
week would count as one occurrence.

- 128 -

Based on this data on sacrifices, we can apply the sacrifice-based over-indebtedness definition
and determine which borrowers are over-indebted in our sample. Respondents are overindebted if they indicate that they struggle to repay their loans on time, make unacceptable
sacrifices related to their loans, and that their experiences of unacceptable sacrifices are more
than just transitory. To determine if sacrifices are a sign of more than transitory debt
problems, we only consider borrowers who experience a minimum of three repeated,
unacceptable sacrifices as over-indebted. Only the structural one-off sacrifices, asset seizures,
unacceptable loan recycling, and selling or pawning assets to be able to repay a loan, do not
need to be repeat experiences.6
As a robustness check, we introduce a threshold free measurement of a borrowers level of
sacrifice. We construct a score that increases from zero to 72 for each sacrifice that a
borrower makes. Sacrifices weigh more the less they are acceptable (one to four points) and
the more frequently they occur (one to four points). This score does not allow us to determine
which factors are related to over-indebtedness, as it does not distinguish between low,
acceptable levels of sacrifice, and sacrifices at the level of over-indebtedness. However, if
empirical results for factors related to over-indebtedness and to the general score of sacrifices
are similar, this proves that our findings are not sensitive to the exact threshold of our binary
over-indebtedness measurement.
<insert table 1 about here>
Tables 1 and 2 provide an overview of the sample characteristics.7 On average, the borrowers
in our sample are representative of typical microborrowers. They are mostly female (72%)
who are poor but economically active and not living in extreme poverty. They are
predominantly self-employed (98%) and are mainly traders (81%). The sample is almost
evenly split between the group lending methodology (48%) and the individual lending
methodology (52%), with some MFIs following only one or the other model, and some
offering both types of lending products. According to borrowers reports, almost all loans are,
at least in part, used for productive purposes (97%), although many of them have not gone
exclusively into a microenterprise but have also been used for consumption purposes (30%).
6 The measurement is not sensitive to these exceptions. They affect only very few cases.
7 As the research methodology relies on self-reports, statistical data regarding the borrowers economic situation are subject
to limitations. Aside from questions of honesty, respondents often experience difficulty estimating the monetary value of
their assets and, given their volatility, their average incomes. However, on average there is no reason to assume a general
upward or downward bias of the estimates and we consider the data sufficiently reliable for our level of analysis.

- 129 -

The majority of borrowers (59%) report only low investment returns on their loan use and
state that their returns were insufficient to repay their loans.
<insert table 2 about here>
The descriptive statistics in table 1 and 2 provide a first indication of which factors are related
to over-indebtedness. Using t-tests for the difference in means between the over-indebted
group and the borrowers who are not over-indebted, we find significant differences in terms
of a borrowers income and assets, as well as adverse shocks to income. Similarly, all three
dimensions of financial literacy have significantly different means among the over-indebted.
In contrast to income shocks, the amount of adverse shocks a borrower has experienced to
his/her expenses is not significantly different between groups. Using a chi-squared test on our
categorical variables, we find a significant correlation between the returns on investment that
a borrower reports and his over-indebtedness status. There is no significant relationship
between over-indebtedness status and loan use. However, neither t-tests nor the chi-squared
contingency analysis can control for the influence of other variables on over-indebtedness.
We therefore regress our indicators of financial literacy, poverty, adverse economic shocks
and loan returns on our binary over-indebtedness measurement in a multinomial logistical
regression. As identified in section 2, we control for the socio-demographic and economic
characteristics of borrowers and the business- and loan-related factors that may be related to
over-indebtedness.
(1) Oi = 0 + 1 Ii + 2 Ai + 3 SUEi + 4 SEEi + 5 SIi + 6 CUi + 7 ROIi + 8 LNi + 9 LGi
+ 10 LDi + 11 Xi + 12 Yi + 13 Zi + ui
For each respondent, i, Oi is a dummy variable that takes the value one for a borrower who is
over-indebted and zero otherwise. I represents income, and A total assets. Regarding adverse
shocks, we distinguish unexpected expense shocks (SUEi), expected large lump sum shocks
(SEEi) and adverse shocks to income (SIi). For returns, CUi stands for at least a portion of the
loan being used for consumption and ROIi for the return on investment. LNi is the result of
the literacy test specific to numeracy, LGi is the result for general financial literacy related to
savings accounts and inflation, and LDi denotes the result of the debt-specific literacy test. Xi
is a vector of controls for socio-demographic characteristics that include the following
categories: gender, age, ethnic background, marital status, the number of children, household

- 130 -

size, housing type, employment status, a subjective categorical measurement of health


problems and a borrowers level of education.8
Yi is a vector of the economic characteristics, income volatility and amount of savings owned,
as well as the key loan characteristics of the number of lenders to whom a borrower is
indebted, the total amount of debt disbursed on current outstanding loans, the average
maturity of their loans and the lending methodology of their main loan (group or individual).
The variable Main_occupation controls for the sector of activity in which the microborrower
is mainly active. Yi includes the borrowers subjective assessment on a five-point Likert scale
of the fairness of the lenders treatment. To retain the information on the order of categories,
we assume equal distances between the different categories of agreement and treat the Likert
scale as continuous. This control ensures that the subjective judgment on the acceptability of
sacrifices in our over-indebtedness measurement is not distorted by the borrowers perception
of having been treated unfairly by a lender. Finally, Zi is a vector of controls for a borrowers
main lender among our partner MFIs that covers all loan characteristics that differ on the
institutional level. This variable introduces an aggregate explanatory variable into our
regression on the individual level. To account for common variance effects on the MFI level
that would bias our estimated standard errors and thus the significance of our coefficients, we
cluster standard errors according to Zi (Greenwald, 1983; Moulton, 1990).
We employ a five-step estimation procedure for the logistic regression model suggested
above. The first model serves as a basic benchmark for analysis. It includes all potential
variables and controls but does not yet take common variance effects into account. We report
odds ratios and Whites heteroscedasticity consistent standard errors (White, 1980). In model
(2) we add the correction for common variance components on the lender level. Given that
our controls describe borrowers individual situations quite comprehensively, there is the
potential for interactions between some of our explanatory variables. We expect women to
differ from men in how they use microloans (D'Espallier et al., 2011; Agier et al., 2012).
Women might be more likely to use loans for household purposes rather than for productive
investments (Garikipati, 2008). We also expect that lending methodology may have an effect
on loan use, because the pressure to invest a loan into a microenterprise as well as the extent

8 In the regression analysis, for reasons of limited observations in certain uncommon categories of socio-demographic
variables, we summarize the categories Gurma/Other for Ethnic_background, No formal education/PreSchool for
Level_of_education, Minor/Strong health problems for Health_problems, Own house/Other for Housing_type,
Permanent/Temporary employment/Student for Employment_status, and Production of goods/Other for Main_occupation.

- 131 -

to which loan use is monitored are likely to be higher in the group lending model (Ghatak and
Guinnane, 1999). Model (3) therefore expands on model 2 by adding interaction terms for
gender with loan use and for lending methodology with loan use.9 Model (4) drops all
controls that were not significant in the previous models, eliminating the noise that potential
non-significant variables may have created. This parsimonious model represents our main
model and is the basis for all consecutive analyses. Model (5) shows the effects of running
our regression on the main hypotheses only, excluding control variables.
As a robustness check, we repeat the same regressions (1) to (5) with a different dependent
variable that does not depend on our over-indebtedness definition and on the specific
threshold applied in our measurement of over-indebtedness. We regress the factors that are
likely to be related to over-indebtedness on the 72-point sacrifice score, as described earlier in
this section. The results are unlikely to be exactly the same, given that different factors may
be related to minor sacrifices than to severe ones. Also, factors that explain if a borrower is
over-indebted may not explain small variations in the intensity of his struggles. Nevertheless,
we expect the robustness check to deliver roughly similar results to those obtained from the
main regression. It thus confirms that our findings are not specific to the details of the overindebtedness measurement but are capable of explaining borrower sacrifices in general.
To understand the relationship of poverty, adverse shocks, loan returns and financial literacy
to sacrifices in more detail, we run additional regressions that consider the sacrifices
underlying over-indebtedness separately. We run our main parsimonious model (4) for the
prevalence of each sacrifice in the borrower sample, independently of the severity of
sacrifices and of a borrowers over-indebtedness status. We then run the same model for the
four most common sacrifices as an ordered logistic regression on the 4-point scale of sacrifice
acceptability. Due to the limited numbers of observations for less common sacrifices (the
ordered logit model with all variables requires approximately 100 observations per sacrifice),
this analysis is possible only for the following sacrifices: cutting down on food, working
harder, postponing important expenses, and depleting ones savings. We repeat the same

9 We also tested the following interaction terms: whether loan use differs between the customers of different MFIs, whether
gender or group lending had any impact on returns on investment, whether there was an interaction of either gender or the
lending methodology with a borrowers debt literacy, and whether gender and methodology interacted with each other. None
of these interaction terms were significant. We tested if the debt-to-income ratio had an independent relationship to overindebtedness in addition to absolute income and absolute debt amounts, but it was not significant.

- 132 -

analysis for the 4-point scale that measures how frequently a borrower repeats the same
sacrifice.
This analysis allows us to understand the mechanisms through which poverty, adverse shocks,
loan returns and financial literacy relate to over-indebtedness; how factors relate to the
severity of sacrifices; which factors increase the repetition of sacrifices; and which factors are
related to specific sacrifices rather than to others. It provides relevant insights for developing
policy implications for microfinance and identifying the best responses to address the risk of
over-indebtedness in microfinance.

- 133 -

4. Results and discussion


Main regressions
Table 3 displays the results of the logistic regressions on over-indebtedness, with the main
model represented by column (4). It confirms that poverty is positively related to overindebtedness (H1). Average monthly income (H1a) is consistently related to overindebtedness at a 1% significance level. According to the marginal effects for model (4) as
represented in table 4, marginal income increases correspond to a reduction in overindebtedness risk by 2%. The picture is less clear for a borrowers wealth in terms of assets,
but in our main model assets are significant at the 5%-level (H1b). A marginal increase of
assets (in thousands) implies that the likelihood of over-indebtedness is lower by 0.2%. As the
control variables in Appendix 3 show, this does however not seem to be a question of savings
as one of the more liquid categories of assets. The variable for the amount of savings a
borrower holds is not significant. This may be due to the lack of savings for many borrowers.
Given the large and significant difference between the amount of savings held by the overindebted and those who are not that is evident from the descriptive statistics in table 1, the
non-significance of savings may also be due to the influence of other variables such as income
on over-indebtedness that, once controlled for, do not leave an independent influence of
savings.
<insert table 3 about here>
<insert table 4 about here>
Aside from the general effects of poverty in terms of low income and little asset ownership,
the analysis confirms that adverse economic shocks have an independent effect on overindebtedness (H2). While the evidence is not strong for unexpected shocks to a borrowers
expenses (for example funerals or broken assets that have to be replaced) (H2a), expected
shocks to expenses such as lump sum expenses for school fees or marriages are related to
over-indebtedness with a 1% significance level (H2b). Surprisingly expected lump sums
appear to be more important than unexpected shocks. For each marginal Cedi of expected
shock, the likelihood of over-indebtedness increases by 0.1%. The data are even stronger
regarding shocks to a borrowers income (H2c). Leaving out the benchmark regression that
had not yet taken the common variance effects on the MFI level into account, all models
consistently display coefficients for income shocks that are significant at the 1%-level.

- 134 -

Similar to expense shocks, a marginal increase in income shocks increases the risk of overindebtedness by 0.1%.
These regressions also relate returns of investment (H3) to over-indebtedness. They confirm
that both loan use (H3a) and the returns on an investment made with the loan (H3b) are
related to over-indebtedness, with significance levels at 1%.10 The results are slightly less
consistent for loan use, but the main model shows that for women (when gender is at its
reference value of 0 and there is no effect of the interaction term), the likelihood of being
over-indebted increases by 12.2% for those who used a loan, at least in part, for a nonproductive purpose. This effect is independent of our measurement of investment returns,
potentially because the classification of returns into four categories does not cover the full
amount of variation of returns and because having to use debt for non-productive purposes
might indicate a lower ex-ante capacity to cover debt instalments that cannot be covered by
investment returns from other sources. It could also relate to the reverse effect of overindebtedness increasing the risk that an individual needs to use part of their loan for nonproductive purposes. Moreover, there is a significant interaction effect between loan use and
gender. Regarding returns on investments, loan returns that are not sufficient to cover
instalments come with a 15.4%

higher likelihood of being over-indebted compared to

productive loan uses that produce permanent earnings increases at a level sufficient to repay
the loan. If there are only impermanent increases in earnings from the productive loan use,
this factor increases to 36.5%, and when no returns are possible (i.e., no investment, not even
in part) it increases to 42.9%.11
For our forth hypothesis, the regressions confirm H4c that borrowers with higher debtspecific financial literacy are less likely to be over-indebted. Marginal effects indicate that an
infinitesimally small increase of a borrowers score in our debt literacy test corresponds to a
0.2% lower likelihood of over-indebtedness. The result is significant at a 1%-level.
Contradicting our prior expectations, the opposite is true for general financial literacy (H4b):
while the effect is small, an increase in general financial literacy corresponds to a higher risk
of over-indebtedness. Mere numeracy (H4a), the ability to perform simple mathematical

10 There is no significant multicollinearity between partial non-productive loan use and ROI.
11 The strength of these results may partly be due to the fact that we have measured ROI in four rough categories as reported
by borrowers. However, section 3 has shown that borrowers are able to distinguish between repayment difficulties and
returns and that both high returns with over-indebtedness (e.g., due to shocks) and low returns without over-indebtedness
(e.g. due to other resources) occur in our sample.

- 135 -

operations, is not related to over-indebtedness. This may be due to the complexity of


borrowing decisions, where knowledge of simple calculations is far from enough to make
good decisions. It may also be due to the relatively good performance of respondents with
regards to the numeracy skills, many of them being market women or involved in other
businesses where basic math is a requirement for success.
In sum, we find strong confirmation for our hypotheses with the expected signs, except for
general financial literacy, for which the relationship to over-indebtedness is positive, and
except for numeracy, which has no relationship to over-indebtedness. Concerning asset
ownership and unexpected shocks to expenses, the results are not always consistent across
models. As Appendix 3 shows, many of our control variables for socio-demographics,
economic situation and lenders are also significant. Interestingly, a borrowers level of
education is not significant, nor is his debt amount, the number of MFIs he borrows from, or
the lending methodology (group/individual).
While a cross-sectional dataset without an experimental design and with a risk of endogeneity
does not allow for an analysis of causality, for most of the hypotheses tested, there is likely to
be a causal chain of the factors impacting over-indebtedness. Over the short time horizon of
the study, it is more likely that low income has contributed to over-indebtedness than that
income has gone down due to over-indebtedness. Adverse economic shocks have not occurred
due to over-indebtedness but are much more likely to have triggered it. The returns on the
productive activity financed with the loan are unlikely to be low because a borrower is overindebted but it is probable that low returns have made repayment more difficult and thus
contributed to over-indebtedness. Debt literacy has not changed when over-indebtedness hit,
but has probably influenced a borrowers choices in taking on debt. More debt literate
individuals seem to make smarter borrowing choices. Further research is required to confirm
and analyse the opposing finding for general financial literacy. In terms of assets and loan use,
the evidence for causality is less clear. Over-indebtedness might be caused by a lack of assets
but might also reduce a borrowers assets (for example, in the case of a seizure or asset sales
to repay a debt). Similarly, consumption loan use can create over-indebtedness but if there is
an effect that is independent of the ROI this may also be due to a higher likelihood that an
over-indebted borrower needs to take on debt for non-productive purposes.

- 136 -

Robustness check
We test the robustness of our findings by replacing the over-indebtedness measure with a
general threshold-free score of borrower sacrifices; table 5, column (4) displays the main
model. The same results are obtained for income, but not for assets and unexpected expense
shocks that were already weaker findings for to over-indebtedness. Out findings also hold
true for the following factors: expected expense shocks and income shocks, loan use and the
interaction of gender with loan use, ROI, debt literacy, and the lack of a relationship to
numeracy. The opposite effect of general financial literacy is confirmed, although weakly.
Given that not all variation in borrower sacrifices is necessarily determined by the same
factors as over-indebtedness, the robustness check provides confirmation for our main
findings. They are not dependent on our specific measurement of over-indebtedness.12
<insert table 5 about here>

Further analyses
To deepen our understanding of how financial literacy, poverty, adverse shocks and loan
returns are related to over-indebtedness, we run separate regressions of our main model (4) on
each sacrifice to determine which of these factors are related to the incidence of each specific
sacrifice.13 This analysis is independent of the acceptability of a sacrifice; thus, not every
sacrifice contributed to a borrower being over-indebted. The analysis allows for a refined
approach to developing solutions to over-indebtedness. Table 6 indicates that income is
related to all sacrifices except to increases in work efforts and shame. Potentially, the poorest
borrowers might already exert their maximum possible work effort and are very accustomed
to financial problems, making sacrifices less of a source of shame. In line with its over-all
relationship to over-indebtedness, the relationship of income to the various sacrifices is
typically negative. Asset ownership is related to exactly those sacrifices where income does

12 The findings are however specific to the customer protection perspective of defining over-indebtedness through the lens of
sacrifices. If over-indebtedness was measured as delinquency, assets, income, expense shocks and loan use would not be
significantly related. Nevertheless, income shocks, returns on investment, and financial literacy are also highly significant
factors related to delinquency (see Appendix 4). While delinquency can contribute to a borrowers sacrifices, from the
customer protection point of view it is a (reinforcing) consequence of over-indebtedness rather than an original influence
factor. We therefore do not include this highly endogenous factor in our regressions on over-indebtedness. We did, however,
test that doing so would not change our findings.
13 Marginal effects available upon request.

- 137 -

not play a role. It prevents extra work and the depletion of savings and has a very small but
positive impact on shame. When adverse shocks hit a borrower, the typical responses to
unexpected expense shocks appear to delay postponing other expenses, including food costs.
Large, expected expenses, in contrast, allow for longer-term coping strategies, such as lining
up extra work, using savings and selling assets. They come with a higher likelihood for shame
for the borrowers, maybe because they are more foreseeable and solicit less pity from others.
When borrowers experience sudden reductions in income, they primarily react by taking on
new debt, or by selling or pawning assets, and they tend to suffer from psychological stress.
Using ones loan, at least in part, for non-productive purposes is very significantly related to
experiences of shame. This may be due to the borrowers bad conscience if he was supposed
to fully invest the loan. At a 10% significance level, non-productive loan use is also positively
related to psychological stress and, interestingly, to reductions in the education of the
borrowers children. Independent of loan use, the different levels of ROI are jointly related to
all sacrifices, except for taking on new debt and experiencing threats. The reason for the
former might be that new debt is more difficult to obtain with a business that is not running
well. The relationship is consistently positive. Finally, good numeracy reduces the risk of
food reductions but increases the risk of reducing ones savings or of relying on external help.
Similarly, good general financial literacy is positively related to savings depletion, taking new
debt, shame and threats. This provides further detail to our counterintuitive finding that
general financial literacy is positively related to over-indebtedness. Literacy could not only
influence the quality of borrowing decisions but could also be related to a borrowers decision
to make certain sacrifices. For example, borrowers with higher levels of general literacy may
have built more savings that they can deplete. Debt-specific financial literacy is related
mainly to reducing food, working harder, and suffering from shame and stress.
<insert table 6 about here>
This analysis of how the factors related to over-indebtedness relate to each of the sacrifices is
an early step in identifying how microborrowers cope with debt. More research must be
performed but we can already conclude that borrowers with different economic backgrounds,
loan uses and levels of financial literacy respond to repayment difficulties in different ways.
This impacts the effect of potential measures against over-indebtedness. For example,
insurance against unexpected expense shocks such as medical expenses could help borrowers
to maintain their general consumption patterns and reduce reductions in food. Different
measures would be required to address other sacrifices.
- 138 -

Regarding the acceptability of sacrifices, table 7 shows that for the most common sacrifices,
the factors related to over-indebtedness are not only related through a higher or lower
likelihood of certain sacrifices occurring, but that they are also related to how acceptable
different sacrifices are to the borrowers.14 For example borrowers in more difficult economic
situations might experience the same sacrifices more severely because they have no further
buffer for e.g., reductions in consumption, or they may experience more severe forms of these
sacrifices. At the same time they could experience some sacrifices as less severe because they
are more used to financial difficulties. The data indicate that a lack of income and assets is
positively related to the unacceptability of having to work harder, i.e., it reduces acceptability.
Expense shocks do not seem to change the acceptability of any of these common sacrifices.
Using a portion of a loan for a non-productive purpose equally reduces the acceptability of
having to work harder to repay it, perhaps because productive loan use links directly to the
expectation to also invest more time in the enterprise. A lack of returns on the investment
reduces the acceptability of both extra work and postponed expenses in the household and
business. Not having invested the loan at all however comes with a higher acceptability of
postponing important expenses, potentially because there were no return expectations linked
to the loan and the borrower was willing to postpone other expenses ex ante. In terms of
financial literacy, general financial literacy seems related to a higher acceptability to postpone
expenses and a lower acceptability to deplete ones savings. Increased debt literacy comes
with a higher acceptability for extra work to repay the loan. Future research with larger
sample sizes would need to increase the precision of these estimates and extend this analysis
to encompass the less common sacrifices.
<insert table 7 about here>
Table 8 analyses how frequently borrowers repeat sacrifices. For those borrowers who have
cut down on food consumption, increased work efforts, or depleted their savings to pay a loan
instalment, poorer borrowers are likely to repeat these sacrifices more frequently than richer
borrowers. In terms of reductions in food and savings, the same is true for borrowers who
have experienced unexpected expense or income shocks. Loan use does not impact the
frequency of the most common sacrifices. A lack of ROI is negatively related to the frequency
of working harder and postponing expenses, i.e., these sacrifices are repeated less frequently.
For financial literacy the picture is twofold: a higher score on the numeracy test comes with

14 Of course cut-off values for the latent variable of acceptability vary significantly between the models.

- 139 -

less repetition of sacrifices three and five. A higher score on general financial literacy seems
to correspond to more frequent sacrificing of food and savings, and debt-specific literacy to
more frequent food cuts and taking on extra work. The best way to understand these findings
would be with qualitative research that can disentangle the reasons behind these effects. For
example, literacy can be positively related to the frequency of some sacrifices because
borrowers who understand loans better are better able to anticipate repayment challenges and
thus identify the necessary coping strategies. At the same time, it can be negatively related to
the frequency of other sacrifices because more literate borrowers may get into less severe
repayment problems and thus be able to avoid the more severe sacrifice experiences.
<insert table 8 about here>
The analysis of how factors related to over-indebtedness relate to the frequency and
acceptability of individual sacrifices is in its nascency, but it sheds some initial light on the
complexity of factors related to over-indebtedness through the lens of customer protection.
Depending on the situation of the individual borrower and on the trigger for the debt
problems, borrowers may experience the same sacrifices as more or less severe or experience
a more or less severe form of them. A borrowers poverty, adverse shocks, loan returns and
financial literacy are also related to the question of how frequently borrowers experience
specific sacrifices. This provides an indication of more prolonged problems with overindebtedness. While it is too early to deduce clear policy recommendations, our analysis
indicates that there may be a means of reducing the customer experience of over-indebtedness
by addressing specific levers that reduce the duration of the sacrifices or that help borrowers
make their situation more acceptable.

- 140 -

5. Conclusion and implications


This paper analyses over-indebtedness from a perspective of customer protection.
Endangering both MFI sustainability and the social impact of microfinance, overindebtedness is currently one of the most important challenges in microfinance. Customer
protection efforts are one of the industrys top priorities. Nevertheless, to date, overindebtedness, especially from the customers perspective, remains highly underresearched.
Insights on factors related to over-indebtedness are urgently needed for the development of
measures against this phenomenon.
This paper indicates that microborrowers in the context of this study are more likely to be
over-indebted when they are living on lower incomes and with lower assets, and when they
experience adverse economic shocks to their income and expenses. There are independent
effects of over-indebtedness risk being higher for borrowers with low returns on the
investment in their microenterprise and for borrowers who, in part, use their loans for nonproductive purposes. Borrowers are more likely to be over-indebted when they lack the
financial literacy to understand debt products. General financial literacy seems to have an
opposite effect, contradicting our expectations. For mere numeracy, there is no statistically
significant effect on over-indebtedness.
While cross-sectional regression cannot analyse causality and contains risks of endogeneity,
the factors we identify to be related to over-indebtedness are likely to be causes of overindebtedness. The relationships we identify confirm prior expectations concerning causality,
and with the exception of asset ownership and loan use a reverse causal relationship is
improbable. Based on a unique database from Ghana, this is the first research to confirm these
relationships in a microfinance context. By identifying the relationships of factors to overindebtedness that are likely candidates for causal influence factors, this paper provides an
important contribution to our understanding of over-indebtedness in microfinance and to the
development of solutions to this major challenge for the industry.
We further deepen the analysis by breaking down how the above factors relate to the question
of which sacrifices borrowers employ as coping strategies for their debt. Borrowers with
different economic backgrounds, loan uses, and financial literacy respond to repayment
difficulties in different ways. For example, when hit by unexpected shocks to their expenses,
borrowers rather react with reductions in their food consumption and with postponing
important other expenses, but they tend to employ longer-term coping strategies such as
- 141 -

working harder, creating and using savings, and selling household or business assets when
faced with lump sum expenses that are expected.
Moreover, this paper provides the first insights into the relationship of economic background,
adverse shocks, loan returns and financial literacy to how acceptable the most common
sacrifices are to borrowers and how frequently these sacrifices are repeated. For example, low
investment returns on a loan are related to a lower acceptability of extra work and of
postponing expenses, but having used a loan exclusively for non-productive purposes comes
with a higher readiness to work harder to repay it. Regarding the repetition of specific
sacrifices, borrowers who are faced with unexpected expenses or with income shocks, for
example, sacrifice food and savings more frequently, while other sacrifices remain unaffected.
Bearing potential limitations to the universality of the findings in mind, our findings have
important policy implications. The microfinance industry must identify ways to reduce overindebtedness among its customers. From a customer protection point of view, it is important
to not just reduce default but to reduce the suffering of customers from over-indebtedness,
i.e., their sacrifices. Knowing that borrowers are more likely to be over-indebted when they
live on lower incomes and own fewer assets, MFIs might be able to reduce over-indebtedness
by reducing credit to poorer borrowers, lending smaller amounts, or choosing to not lend to
the poorest borrowers.15 However, it is important to strike a good balance between the
inclusion of the poor and avoiding over-indebtedness. Our findings should not lead to a return
to financial exclusion but excluding people who would experience harm from debt may be in
their interest.16
Similarly, if adverse shocks to income and expenses represent a main cause of overindebtedness among microborrowers, MFIs should take the likelihood of adverse shocks into
account when calculating repayment capacity. Adverse shocks are frequent among the target
customers of MFIs and can turn an apparently healthy debt balance into a case of suffering
and over-indebtedness. There is, again, reason for caution to avoid financial exclusion. At

15 The control variables in the Appendix paint a more precise picture of the general socio-demographic and economic
characteristics of the borrowers with the highest risk of over-indebtedness. They could contribute to some type of credit
scoring but the same limitations apply that apply to the poverty criterion.
16 Note that our measurement of over-indebtedness does not imply causality. Microdebt may be the cause of the sacrifices
but, even if respondents have listed the sacrifices that they experience as being related to their loan, the loans may also have
helped the borrowers deal with an otherwise even more unbearable situation. Or the loan may have been useful but an
adverse shock can have triggered over-indebtedness ex post.

- 142 -

times, loans can be the best available tool for borrowers to cope with adverse economic
shocks. Borrowers will be left with an unmanageable debt burden but may yet benefit, rather
than having to face, for example, high medical expenses without access to credit. MFIs should
therefore anticipate the occurrence of adverse shocks when determining creditworthiness and
deciding on loan amounts and instalment schedules. They may, however, provide emergency
loans, knowing that repayment might be difficult for the borrower. In certain cases, instead of
reducing access to loans, more flexible instalment schedules could be the best solution to
avoid over-indebtedness in borrowers who experience adverse shocks. This includes ex-ante
flexibility of repayment conditions as much as ex-post flexibility with renegotiations and
rescheduling (Hamp and Laureti, 2011). Promoting savings instead of credit or providing
insurance to borrowers could also go a long way toward preventing over-indebtedness.
There is an ongoing debate in microfinance concerning the industrys original requirement
that loans be used productively. For the context of this study, our paper confirms that nonproductive loan use increases the risk of over-indebtedness. However, these loans are, at
times, worthwhile. Repaying a loan for emergency medical expenses is almost automatically
more difficult than repaying a loan with good returns on investment. Nevertheless, the
emergency loan may be beneficial to the borrower. Similarly, lower ROI on invested loans is
related to higher over-indebtedness risks. Based on the perceptions of borrowers, our paper
indicates that ROI is often not sufficient for borrowers to repay their loans. It seems that the
returns on the investment of microborrowers do not match the cost of credit. The high interest
rates in microfinance have often been challenged for ethical reasons (Hudon, 2007), but at the
same time, they have been justified with high returns on borrowers investments. While
arguments concerning the high costs of providing microloans, the sustainability of MFIs and
risks of exclusion due to interest rate caps remain valid, our findings challenge the assumption
that high returns enable microborrowers to pay high interest rates. Nevertheless, the
microenterprise investment may pay off in the long run, even if returns are not sufficient over
the short instalment schedule of the typical microloan. Instead of reducing access to credit, the
response might again be to better tailor products to borrowers needs and to develop
instalment schedules appropriate to the cash flows of microenterprises.
Finally, our paper indicates that improving borrowers financial literacy can reduce their risk
of over-indebtedness. It shows, however, that financial literacy training needs to be specific to
debt literacy to have the desired effect. Trainings in computational skills and general financial
literacy would not be effective.
- 143 -

Moreover, in developing solutions to over-indebtedness, the industry should be aware of the


complexity of the phenomenon. The factors analysed in this paper do not only relate to
borrowers risk of over-indebtedness, but they are equally related to which sacrifices
borrowers experience, how frequently they repeat these sacrifices and how acceptable
different sacrifices are to them. Further research needs to shed light on the details of these
complex relationships for policy measures to effectively take into account the fact that
specific measures address only specific sacrifice experiences among borrowersand that
there might be levers tailored to reducing the repetition of sacrifices or to increasing the
acceptability of sacrifices to borrowers. This research points to a much broader spectrum of
potential policy measures aimed at over-indebtedness, than just a focus on reducing
delinquency and default would suggest.
These policy implications promise progress in the fight against over-indebtedness in
microfinance but leave substantial room for ambiguity. It is unclear to what extent findings
can be generalized to other microfinance contexts. It will remain a question of judgment, in
specific situations, to what extent restricting credit for the poorest customers, being
conservative in calculating repayment capacity, limiting loan use, flexibilising loan products,
and offering financial literacy trainings is the correct response to over-indebtedness risks.
Extreme forms of these solutions could potentially eliminate over-indebtedness but would do
so at the cost of excluding many customers from debt who could sincerely benefit, or (in the
case of flexibility) at the cost of eroding repayment performance and putting the sustainability
of institutions at risk. A certain level of over-indebtedness is probably unavoidable in any
lending market. However, over-indebtedness measures should aim to reduce this level to a
responsible minimum.

- 144 -

References
Agier, I., I. Gurin and A. Szafarz (2012) 'Child gender and parental borrowing: Evidence
from India'. Economics Letters, 115, 363365.
Atkinson, A., S. McKay, E. Kempson and S. Collard (2006) 'Levels of Financial Capability in
the UK: Results of a baseline survey'. Financial Services Authority Consumer Research
No. 47, London.
Bouquet, E., B. Wampfler, E. Ralison and M. Roesch (2007) 'Trajectoires de crdit et
vulnrabilit des mnages ruraux: Le cas des Cecam de Madagascar [Credit trajectories
and vulnerability of rural households: the case of Cecam in Madagascar]'. Autrepart, 44,
157172.
Bridges, S. and R. Disney (2004) 'Use of Credit and Arrears on Debt among Low-Income
Families in the United Kingdom'. Fiscal Studies, 25(1), 125.
Brown, T. M. (1952) 'Habit Persistence and Lags in Consumer Behaviour'. Econometrica,
20(3), 355371.
Canner, G. B. and C.A. Luckett (1991) 'Payment of household debts'. Federal Reserve
Bulletin, 77(4), 218.
Collins, D. (2008) 'Debt and household finance: Evidence from the Financial Diaries'.
Development Southern Africa, 25(4), 469479.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How
the World's Poor Live on $2 a Day, Princeton, NJ: Princeton Univ. Press.
de Mel, S., D.J. McKenzie and C. Woodruff (2008) 'Returns to Capital in Microenterprises:
Evidence from a Field Experiment'. Quarterly Journal of Economics, 123(4), 13291372.
Del-Ro, A. and G. Young (2005) 'The impact of unsecured debt on financial distress among
British households'. Bank of England Working Paper No. 262, London.
D'Espallier, B., I. Gurin and R. Mersland (2011) 'Women and Repayment in Microfinance: A
Global Analysis'. World Development, 39(5), 758772.
Disney, R., S. Bridges and J. Gathergood (2008) 'Drivers of Over-Indebtedness: Report to the
Department for Business, Enterprise and Regulatory Reform', Centre for Policy
Evaluation, University of Nottingham.

- 145 -

Drentea, P. and P.J. Lavrakas (2000) 'Over the Limit: The Association Among Health, Race,
and Debt'. Social Science & Medicine, 50(4), 517529.
Duvendack, M., R. Palmer-Jones, J. Copestake, L. Hooper, Y. Loke and N. Rao (2011) 'What
is the evidence of the impact of microfinance on the well-being of poor people?'. EPPICentre, Social Science Research Unit, Institute of Education, University of London,
London.
Friedman, M. (1957) 'A Theory of the Consumption Function, Princeton: Princeton University
Press.
Garikipati, S. (2008) 'The Impact of Lending to Women on Household Vulnerability and
Women's Empowerment: Evidence from India: Special Section: Social Movements and the
Dynamics of Rural Development in Latin America (pp. 2874-2952)'. World Development,
36(12), 26202642.
Ghatak, M. and T.W. Guinnane (1999) 'The economics of lending with joint liability: theory
and practice'. Journal of Development Economics, 60(1), 195228.
Godquin, M. (2004) 'Microfinance Repayment Performance in Bangladesh: How to Improve
the Allocation of Loans by MFIs'. World Development, 32(11), 19091926.
Gonzalez, A. (2008) 'Microfinance, Incentives to Repay, and Overindebtedness: Evidence
from a Household Survey in Bolivia. Doctoral thesis'. Ohio State University, Ohio.
Gonzalez, A. (2009) 'Consumption, Commercial or Mortgage Loans: Does It Matter for MFIs
in LAC?'. Microfinance Information Exchange, Inc. Data Brief No. 3.
Greenwald, B. (1983) 'A general analysis of bias in the estimated standard errors of least
square coefcients'. Journal of Econometrics, 22(3), 323338.
Hamp, M. and C. Laureti (2011) 'Balancing flexibility and discipline in microfinance:
Innovative financial products that benefit clients and service providers'. Centre for
European Research in Microfinance, Universit Libre de Bruxelles CEB Working Paper
No. 11/044.
Hudon, M. (2007) 'Fair interest rates when lending to the poor'. Ethics and Economics, 5(1).
Hulme, D. (2007) 'Is microdebt good for poor people? A note on the dark side of
microfinance' In T. Dichter & M. Harper (Eds.), What's wrong with Micofinance? (pp. 19
22). Warwickshire: Intermediate Technology Publications.

- 146 -

Karlan, D. and N. Goldberg (2011) 'Microfinance Evaluation Strategies: Notes on


Methodology and Findings' In B. Armendriz & M. Labie (Eds.), The Handbook of
Microfinance (pp. 1758). London, Singapore: World Scientific Publishing.
Karlan, D. and J. Zinman (2009) 'Expanding Microenterprise Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts in Manila'. Financial Access
Initiative, Cambridge, MA.
Laibson, D. (1997) 'Golden Eggs and Hyperbolic Discounting'. Quarterly Journal of
Economics, 112(2), 443477.
Lea, S. E. G., P. Webley and R.M. Levine (1993) 'The Economic Psychology of Consumer
Debt'. Journal of Economic Psychology, 14(1), 85119.
Lea, S. E. G., P. Webley and C.M. Walker (1995) 'Psychological Factors in Consumer Debt:
Money Management, Economic Socialization, and Credit Use'. Journal of Economic
Psychology, 16(4), 681701.
Livingstone, S. M. and P.K. Lunt (1992) 'Predicting personal debt and debt repayment:
Psychological, social and economic determinants'. Journal of Economic Psychology, 13(1),
111134.
Lusardi, A. and O.S. Mitchell (2007) 'Baby Boomer retirement security: The roles of
planning, financial literacy, and housing wealth'. Journal of Monetary Economics, 54(1),
5224.
Lusardi, A. and P. Tufano (2009) 'Debt Literacy, Financial Experiences, and
Overindebtedness'. National Bureau of Economic Research NBER Working Paper Series
No. 14808, Cambridge, MA.
Modigliani, F. (1980) 'The Collected Papers of Franco Modigliani: The Life Cycle Hypothesis
of Saving'. (Abel, A., Ed.), Cambridge, MA; London: MIT Press.
Moulton, B. (1990) 'An illustration of a pitfall in estimating the effects of aggregate variables
on micro units'. Review of Economics and Statistics, 72(2), 334338.
Paxton, J., D. Graham and T. Cameron (2000) 'Modeling Group Loan Repayment Behavior:
New Insights from Burkina Faso'. Economic Development and Cultural Change, 48(3),
639655.

- 147 -

Schicks, J. (in press) 'Microfinance Over-Indebtedness: Understanding its Drivers and


Challenging the Common Myths'. Oxford Development Studies,
Schreiner, M. (2004) 'Scoring Arrears at a Microlender in Bolivia'. Journal of Microfinance,
6(2), 6588.
Sharma, M. and M. Zeller (1997) 'Repayment performance in group-based credit programs in
Bangladesh: An empirical analysis'. World Development, 25(10), 17311742.
Steel, W. (2010) 'Ghanas Microfinance Industry: Evolution and Prospects. Draft August
2010'. ISSER, University of Ghana, Prepared for IFC Access to Finance Advisory Microfinance. Accra.
Stone, B. and R.V. Maury (2006) 'Indicators of personal financial debt using a multidisciplinary behavioral model'. Journal of Economic Psychology, 27(4), 543556.
Thaler, R. (1985) 'Mental Accounting and Consumer Choice'. Marketing Science, 4(3), 199
214.
Tiwari, A., A. Khandelwal and M. Ramji (2008) 'How Do Microfinance Clients Understand
their Loans?'. Institute for Financial Management and Research, Centre for Micro Finance
Working Paper Series No. 25.
Tokunaga, H. (1993) 'The use and abuse of consumer credit: Application of psychological
theory and research'. Journal of Economic Psychology, 14(2), 285316.
Vogelgesang, U. (2003) 'Microfinance in Times of Crisis: The Effects of Competition, Rising
Indebtedness, and Economic Crisis on Repayment Behavior'. World Development, 31(12),
20852114.
Webley, P. and E.K. Nyhus (2001) 'Life-Cycle and Disproportionate Routes into Problem
Debt'. British Journal of Psychology, 92, 423446.
White, H. (1980) 'A heteroscedasticity-consistent covariance matrix estimator and a direct test
for heteroscedasticity'. Econometrica, 48(4), 817830.

- 148 -

Table 1: Descriptive statistics of borrower sample. Metric Variables


Variable
Avg_monthly_income_approx_value
Total_assets_in_thds
Unexpected_expense_shocks_over_i
Expected_expense_shocks_over_inc
Unexpected_income_shocks_over_in
Score_numeracy_test
Score_general_financial_literacy
Score_debt_literacy_test
Age
Number_of_children
Household_size
Amount_of_savings_self_reported
Number_of_MFIs_crossborrowing
Total_amount_of_debt_disbursed
Average_maturity_weighted

Units
Ghana Cedis
Ghana Cedis (thds)
% of income
% of income
% of income
% of correct answers
% of correct answers
% of correct answers
Years
Number of persons
Number of persons
Ghana Cedis
Number of MFIs
Ghana Cedis
Months

Expected
relationship to over+
+
+
+
+
+
+
+/-

Standard
Over-indebted (n= 158)
N Global Mean Deviation (SD)
Mean
SD
530
645.14
575.48
544.81
456.40
524
15.00
21.91
11.38
14.97
531
12.22
31.00
13.81
24.78
531
20.84
37.34
22.42
37.19
530
11.76
28.35
17.01
35.94
530
80.06
24.86
76.04
27.86
527
57.31
36.92
51.64
37.79
527
28.31
29.39
20.17
24.63
520
40.09
8.60
40.31
8.73
531
2.48
1.57
2.54
1.65
531
4.75
2.15
4.94
2.41
525
406.80
765.55
286.89
638.45
531
1.07
0.28
1.09
0.29
529
1408.70
1286.75
1339.66
1007.79
527
8.04
4.12
8.07
4.48

T-test for equal means between over-indebted and not over-indebted group; *** p<0.01, ** p<0.05, * p<0.1
Income data was collected in categories of 200 Ghana Cedis. The approximate values refer to the middle of each category.

- 149 -

Not over-indebted (n=373)


Mean
SD
687.80
614.83
16.50
24.07
11.54
33.29
20.18
37.44
9.53
24.15
81.75
23.31
59.69
36.34
31.74
30.57
40.00
8.55
2.45
1.53
4.66
2.02
457.41
808.74
1.06
0.28
1437.75
1387.70
8.02
3.96

T-test
2.75***
2.87***
-0.84
-0.59
-2.27**
2.02**
2.05**
4.3***
-0.34
-0.59
-1.21
2.31**
-0.98
0.88
-0.12

Table 2: Descriptive statistics of borrower sample. Categorical Variables


Variable
Any_non-productive_loan_use
Productive only
At least one non-productive loan use
Returns_on_investment
Permanent significant increase in earnings
Increase not sufficient to repay loan
No permanent increase in earnings
No investment
Gender
Female
Male
Ethnic_background
Akan
Dagbone-Dagomba
Ewe
Ga
Gurma
Other
Marital_status
Divorced
Married
Separated
Single
Widowed
Level_of_education
No formal education
PreSchool
Primary
JSS/JHS/Middle School
Secondary/vocational/technical/commerc
Tertiary, university graduate
Health_problems
No health problems
Minor health problems
Strong heath problems
Housing_type
Living for rent
Living in family housing for free
Living in your own house
Other
Employment_status
Permanent employment
Self-employed
Student
Temporary employment
Main_occupation
Offering services
On-selling
Production of goods
Other
Income_volatility
Stable income
Little volatility
Strong volatility
Group_or_individual_customer
Group
Individual
General_fairness_of_MFIs
Strongly disagree
Somewhat disagree
Neither agree nor disagree
Somewhat agree
Strongly agree
Main_lender
MFI 1
MFI 2
MFI 3
MFI 4
MFI 5

N %

Over-indebted (n= 158) Not over-indebted (n=373)


N
%
N
%

372
157

70.2
29.8

106
52

67.0
33.0

266
105

71.6
28.4

186
177
74
13

41.4
39.3
16.4
2.9

31
64
37
5

22.6
46.7
27.2
3.5

156
113
37
8

49.6
36.1
11.7
2.6

383
148

72.2
27.8

112
46

71.0
29.1

271
102

72.7
27.3

346
21
73
66
2
23

65.1
4.0
13.8
12.4
0.4
4.4

100
5
22
20
1
10

63.2
2.9
14.0
12.9
0.4
6.6

246
16
51
45
1
13

65.9
4.4
13.7
12.2
0.3
3.4

23
382
10
87
30

4.2
71.9
1.8
16.5
5.6

7
108
3
32
9

4.4
68.2
1.8
20.2
5.5

16
274
7
56
21

4.2
73.5
1.8
14.9
5.7

30
6
70
269
119
37

5.6
1.1
13.3
50.6
22.5
6.9

10
2
27
80
31
9

6.2
1.0
16.9
50.4
19.8
5.7

20
4
44
189
88
28

5.4
1.2
11.7
50.8
23.6
7.4

408
117
6

76.9
22.0
1.2

116
39
3

73.5
24.4
2.1

292
78
3

78.3
20.9
0.8

333
75
121
2

62.7
14.2
22.7
0.3

102
20
36
0

64.7
12.6
22.8
0.0

231
56
85
2

61.9
14.9
22.7
0.5

7
522
1
2

1.3
98.2
0.1
0.4

2
155
0
1

1.2
98.0
0.0
0.8

5
367
1
1

1.3
98.3
0.2
0.2

65
430
35
1

12.3
80.9
6.7
0.1

16
131
11
0

9.9
82.9
7.2
0.0

50
299
24
1

13.3
80.1
6.5
0.2

12
336
183

2.3
63.2
34.5

1
103
54

0.9
65.0
34.1

11
233
129

2.9
62.5
34.6

253
278

47.6
52.4

81
77

51.0
49.0

172
201

46.1
53.9

19
26
17
141
325

3.6
4.8
3.3
26.7
61.6

10
11
9
42
86

6.1
6.8
5.5
26.9
54.8

9
15
9
98
239

2.5
4.0
2.3
26.6
64.5

110
166
81
97
70

21.0
31.7
15.5
18.5
13.4

29
51
23
35
19

18.6
32.7
14.6
22.2
11.9

80
115
58
62
51

Chi-Square
Cramer's V

2.23
0.065

22.0
31.3
15.9
16.9
14.0

36.36***

0.284

0.8

0.039

3.11

0.077

1.8

0.058

4.31

0.090

2.65

0.071

2.14

0.064

1.85

0.059

1.99

0.061

1.48

0.053

1.44

-0.052

14.6***

0.166

2.1163

0.064

Chi-Square significance levels indicate statistical significance *** p<0.01, ** p<0.05, * p<0.1. Cramer's V indicates strength of association.
Contingency analysis in Stata is unweighted. Weighted Chi-Square results imply no substantial changes to results

- 150 -

Table 3: Logistic regression of borrower characteristics on over-indebtedness


Dependent variable: Over-indebtedness (dummy)
Average_monthly_income

(1)
0.843**
(0.057)
0.988
(0.008)
1.002
(0.003)
1.003
(0.004)
1.007
(0.004)
1.304
(0.440)
2.857***
(1.087)
6.150***
(2.761)
5.160*
(5.024)
0.995
(0.008)
1.005
(0.005)
0.987**
(0.006)

(2)
0.843***
(0.025)
0.988
(0.009)
1.002
(0.002)
1.003***
(0.001)
1.007***
(0.001)
1.304
(0.238)
2.857***
(0.911)
6.150***
(3.983)
5.160***
(2.921)
0.995
(0.008)
1.005***
(0.002)
0.987***
(0.005)

Controls

0.000***
(0.000)
Added

0.000***
(0.000)
Added

(3)
0.856***
(0.030)
0.986
(0.009)
1.002*
(0.001)
1.004***
(0.001)
1.007***
(0.001)
2.513***
(0.658)
2.824***
(0.912)
6.229***
(4.210)
5.410**
(4.103)
0.993
(0.008)
1.005**
(0.002)
0.987**
(0.006)
0.364***
(0.065)
0.457
(0.290)
0.000***
(0.000)
Added

Interaction terms

Excluded

Excluded

Added

Observations (N)
Pseudo R
Nagelkerke's R

401
0.2241
0.3370

401
0.2241
0.3370

401
0.2347
0.3509

Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Interaction_gender_with_loan_use
Interaction_methodology_with_loan_use
Constant

(4)
0.875***
(0.016)
0.987**
(0.007)
1.002*
(0.001)
1.005***
(0.002)
1.008***
(0.001)
2.089***
(0.287)
2.579***
(0.693)
6.073***
(3.798)
7.089***
(3.547)
0.997
(0.006)
1.005**
(0.002)
0.985***
(0.005)
0.314***
(0.090)

(5)
0.880***
(0.028)
0.992*
(0.004)
1.001
(0.002)
1.003**
(0.001)
1.006***
(0.001)
1.513**
(0.284)
2.924***
(0.755)
5.930***
(2.770)
5.049***
(1.314)
0.996
(0.007)
1.004**
(0.002)
0.985***
(0.003)

0.000***
(0.000)
Added if
relevant
Added if
relevant
414
0.2196
0.3311

0.380*
(0.190)
Excluded
Excluded
417
0.1332
0.2107

Odds Ratios. Robust Standard Errors in parenthesis. Standard Errors for models 2-4 clustered by Main_lender_among_my_MFIs
*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first category according to table 2.

- 151 -

Table 4: Marginal effects for the parsimonious model (4)


Variable
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test

dy/dxStd.Err.
-0.020 0.002
-0.002 0.001
0.000 0.000
0.001 0.000
0.001 0.000
0.122 0.027
0.154 0.040
0.365 0.136
0.429 0.112
0.000 0.001
0.001 0.000
-0.002 0.001

1 dy/dx is for discrete change of dummy variable from 0 to 1

- 152 -

z
-8.670
-1.970
1.590
3.250
7.980
4.520
3.890
2.690
3.830
-0.410
2.220
-3.210

P>z
0.000
0.048
0.113
0.001
0.000
0.000
0.000
0.007
0.000
0.681
0.026
0.001

95% C.I.

-0.025 -0.016
-0.004 0.000
0.000 0.001
0.000 0.001
0.001 0.001
0.069 0.175
0.076 0.232
0.099 0.631
0.210 0.649
-0.002 0.001
0.000 0.001
-0.004 -0.001

X
3.540
14.845
13.299
21.565
12.941
0.318
0.394
0.161
0.026
79.781
56.908
28.724

Table 5: Robustness check - OLS regression of models 1-5 on a discrete score of sacrifices
Dependent variable: Sacrifice score (0-72)
Average_monthly_income

(1)
-0.0248**
-0.00985
-0.00158
-0.00111
0.000152
-0.00066
0.000553
-0.00073
0.00136*
-0.000797
0.0411
-0.056
0.149**
-0.0603
0.320***
-0.0791
0.176
-0.158
-0.000964
-0.00135
0.000981
-0.000908
-0.00226**
-0.000943

(2)
-0.0248***
-0.00455
-0.00158
-0.00098
0.000152
-0.000448
0.000553*
-0.000248
0.00136***
-0.000162
0.0411
-0.0364
0.149*
-0.054
0.320*
-0.118
0.176
-0.134
-0.000964
-0.00143
0.000981**
-0.000295
-0.00226**
-0.000726

Controls

-0.176
-0.314
Added

-0.176
-0.215
Added

(3)
-0.0224**
-0.00513
-0.00152
-0.000944
0.000166
-0.000413
0.000644*
-0.000241
0.00135***
-0.000155
0.144**
-0.0489
0.149**
-0.0536
0.321*
-0.119
0.189
-0.142
-0.00115
-0.00142
0.000912*
-0.00036
-0.00219*
-0.000827
-0.153**
-0.0509
-0.102
-0.0916
-0.148
-0.233
Added

Interaction terms

Excluded

Excluded

Added

Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Interaction_gender_with_loan_use
Interaction_methodology_with_loan_use
Constant

N
R-sq
adj. R-sq

401
0.2350
0.1356

401
0.2350
0.1356

(4)
-0.0204***
-0.00332
-0.00152
-0.000814
0.00023
-0.000335
0.000825**
-0.000287
0.00145***
-0.000108
0.121**
-0.0308
0.141**
-0.0423
0.322**
-0.104
0.267**
-0.0634
-0.000505
-0.00111
0.000823
-0.000389
-0.00254**
-0.000565
-0.180**
-0.0518

(5)
-0.0205**
-0.00514
-0.00101*
-0.000436
0.00000727
-0.000545
0.000564
-0.000331
0.00125***
-0.000169
0.0744
-0.0373
0.181***
-0.0391
0.347**
-0.0937
0.305***
-0.0626
-0.000806
-0.00131
0.000791*
-0.000359
-0.00263***
-0.000423

0.0606
-0.189
Added if
relevant
Added if
relevant

0.293**
-0.0966
Excluded

401
0.2439
0.1407

Robust Standard Errors in parenthesis. Standard Errors for models 2-4 clustered by Main_lender_among_my_MFIs
*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first category according to table 2.

- 153 -

Excluded
414
0.233
0.1597

417
0.1515
0.1263

Table 6: Logistic regressions of factors related to over-indebtedness on individual sacrifices


Dependent variable: Sacrifices 1-12 (dummy)
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Constant
Controls
Observations (N)
Pseudo R
Nagelkerke's R

(1)
Food
0.789***
(0.054)
0.989
(0.014)
0.989**
(0.006)
1.001
(0.003)
1.005
(0.003)
2.057
(0.925)
2.206
(1.296)
5.495***
(3.535)
0.227
(0.525)
0.992
(0.009)
0.996
(0.007)
0.987***
(0.003)
21.643*
(37.695)
Added
414
0.2314
0.3154

(2)
Education
0.864***
(0.029)
0.977
(0.027)
0.982
(0.019)
1.002
(0.005)
0.993
(0.010)
1.647*
(0.418)
14.184***
(10.296)
55.345***
(23.852)
1.038***
(0.009)
1.001
(0.004)
0.984
(0.014)
0.000***
(0.000)
Added
301
0.3764
0.4340

(3)
Work
0.977
(0.047)
0.979***
(0.005)
1.000
(0.006)
1.002***
(0.001)
1.000
(0.005)
0.984
(0.393)
1.237
(0.266)
1.731***
(0.190)
13.001**
(13.300)
1.002
(0.003)
1.004
(0.003)
0.991***
(0.001)
29.585***
(26.812)
Added
414
0.1742
0.2802

(4)
Expenses
0.789***
(0.054)
0.989
(0.014)
0.989**
(0.006)
1.001
(0.003)
1.005
(0.003)
2.057
(0.925)
0.626
(0.302)
0.967
(0.325)
0.195**
(0.123)
1.007
(0.008)
1.006
(0.005)
0.998
(0.002)
0.027**
(0.040)
Added
414
0.1645
0.2688

(5)
Savings
1.085*
(0.053)
0.974***
(0.009)
1.006
(0.004)
1.007**
(0.003)
1.000
(0.004)
1.208
(0.742)
1.451***
(0.093)
1.448***
(0.126)
0.561
(0.619)
1.019***
(0.006)
1.005*
(0.003)
0.998
(0.003)
1.245
(2.409)
Added
414
0.1320
0.2131

Odds Ratios. Robust Standard Errors in parenthesis. Standard Errors clustered by Main_lender_among_my_MFIs
*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first category according to table 2.
The last category of ROI had to be dropped in three models for reasons of multicollinearity.
Models numbered according to list of sacrifices in Appendix 2.

- 154 -

(6)
New debt
0.410***
(0.117)
1.021
(0.022)
0.995
(0.006)
1.006
(0.011)
1.011**
(0.005)
1.971
(1.784)
5.117
(9.318)
9.951
(16.103)
40.642
(136.967)
0.981
(0.021)
1.024**
(0.012)
0.983
(0.019)
0.000***
(0.000)
Added
390
0.4177
0.4590

(7)
Sell assets
0.343***
(0.133)
0.990
(0.022)
1.013
(0.010)
1.011**
(0.005)
0.978**
(0.010)
2.332
(1.704)
293.635**
(766.868)
202.491
(667.616)
283.411*
(889.877)
0.976
(0.026)
1.026
(0.019)
0.998
(0.010)
0.000***
(0.000)
Added
380
0.5649
0.6100

(9)
(10)
External help Shame
0.410***
1.176
(0.117)
(0.388)
1.021
1.045***
(0.022)
(0.004)
0.995
1.037
(0.006)
(0.052)
1.006
1.029***
(0.011)
(0.010)
1.011**
1.003
(0.005)
(0.016)
1.971
47.422***
(1.784)
(62.280)
2.794***
480.398**
(0.917)
(1429.347)
0.915
128.608*
(0.468)
(365.289)
2.308*
(1.061)
1.020**
0.997
(0.008)
(0.022)
1.007
0.983***
(0.007)
(0.004)
0.994
0.972**
(0.010)
(0.012)
0.000***
0.062
(0.000)
(0.663)
Added
Added
372
184
0.3021
0.5249
0.3828
0.5788

(11)
Threats
1.483**
(0.283)
1.008
(0.028)
0.972
(0.041)
0.984
(0.050)
1.007
(0.031)
2.872
(2.465)
1.730
(2.298)
4.602
(7.541)
0.963
(0.027)
1.045***
(0.015)
1.001
(0.031)
0.000***
(0.000)
Added
302
0.6182
0.6581

(12)
Stress
0.755**
(0.104)
0.988
(0.015)
0.994
(0.010)
1.003
(0.005)
1.009***
(0.003)
3.416*
(2.478)
2.447***
(0.595)
2.724
(1.790)
2.613***
(0.884)
1.004
(0.011)
1.004
(0.005)
0.988**
(0.006)
0.051**
(0.068)
Added
367
0.2285
0.2923

Table 7: Ordered logit regression on the acceptability of the most common sacrifices
Dependent variable: 4-point scale for acceptability
of sacrifices
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Controls
Latent cut-off 1
Latent cut-off 2
Latent cut-off 3
Observations (N)
Pseudo R
Nagelkerke's R

(1)
Food
0.388
(0.349)
-0.034
(0.027)
-0.020
(0.059)
0.006
(0.011)
0.013
(0.029)
1.045
(1.232)
2.430
(2.155)
1.919
(1.880)
5.522
(3.742)
0.028
(0.034)
-0.004
(0.010)
0.017
(0.029)
Added
26.037
(5.946)
27.146
(5.72)
29.706
(6.064)
66
0.2973
0.5778

(3)
Work
0.064**
(0.030)
0.033**
(0.014)
-0.011
(0.008)
0.004
(0.009)
0.007
(0.008)
0.663**
(0.302)
1.608**
(0.706)
1.574**
(0.727)
0.397
(0.572)
0.001
(0.007)
-0.002
(0.003)
-0.018***
(0.006)
Added
4.517
(1.89)
5.786
(2.063)
7.681
(2.015)
241
0.2690
0.5271

(4)
Expenses
-0.064
(0.081)
0.031
(0.021)
-0.002
(0.009)
0.010*
(0.006)
-0.005
(0.003)
0.474
(0.651)
1.708**
(0.789)
2.097***
(0.409)
-15.247***
(1.275)
0.003
(0.009)
-0.012*
(0.007)
-0.001
(0.004)
Added
17.553
(2.854)
18.527
(2.162)
20.816
(2.734)
187
0.2871
0.5392

(5)
Savings
0.388
(0.349)
-0.034
(0.027)
-0.020
(0.059)
0.006
(0.011)
0.013
(0.029)
1.045
(1.232)
-0.149
(0.228)
0.052
(0.440)
0.607
(2.475)
0.014
(0.009)
0.008*
(0.004)
-0.010
(0.007)
Added
0.130
(1.024)
2.033
(1.134)
3.960
(1.285)
139
0.1450
0.3406

Regression coefficients. Robust Standard Errors in parenthesis, clustered by Main_lender_among_my_MFIs.


*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first according to table 2.
Models numbered according to list of sacrifices in Appendix 2.

- 155 -

Table 8: Ordered logit regression on the frequency of the most common sacrifices
Dependent variable: 4-point scale for frequency of
sacrifices
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Controls
Latent cut-off 1
Latent cut-off 2
Latent cut-off 3
Observations (N)
Pseudo R
Nagelkerke's R

(1)
Food
1.652***
(0.411)
-0.066
(0.042)
0.161**
(0.078)
0.000
(0.010)
0.070*
(0.036)
3.503
(2.883)
0.049
(3.034)
-1.211
(3.257)
-1.018
(3.000)
0.020
(0.020)
-0.019
(0.017)
0.061**
(0.030)
Added
0.513
(2.703)
4.629
(3.633)
8.334
(3.902)
66
0.5041
0.7902

(3)
Work
-0.194**
(0.089)
-0.004
(0.014)
0.003
(0.008)
-0.005
(0.006)
0.005
(0.013)
-0.324
(0.654)
-0.806**
(0.358)
-0.449
(0.458)
-1.789*
(1.058)
-0.014*
(0.008)
0.008*
(0.004)
0.011**
(0.005)
Added
-4.509
(2.338)
-2.288
(2.033)
-1.116
(1.994)
241
0.1622
0.3438

(4)
Expenses
-0.187
(0.118)
-0.010
(0.011)
0.004
(0.014)
-0.006
(0.006)
0.003
(0.005)
-0.177
(0.742)
-0.998***
(0.326)
-0.659**
(0.331)
-0.855
(0.575)
0.005
(0.012)
-0.005
(0.006)
-0.001
(0.003)
Added
-19.745
(1.38)
-18.021
(1.342)
-17.071
(1.393)
187
0.1959
0.4239

(5)
Savings
1.652***
(0.411)
-0.066
(0.042)
0.161**
(0.078)
0.000
(0.010)
0.070*
(0.036)
3.503
(2.883)
0.619
(0.660)
0.577
(0.440)
-0.488
(0.733)
-0.028**
(0.013)
0.009*
(0.005)
-0.005
(0.010)
Added
-5.410
(2.437)
-2.851
(2.401)
-1.035
(2.676)
139
0.1639
0.3604

Regression coefficients. Robust Standard Errors in parenthesis, clustered by Main_lender_among_my_MFIs.


*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first according to table 2.
Models numbered according to list of sacrifices in Appendix 2.

- 156 -

Appendix 1: Survey questions on financial literacy

Numeracy
1. Compute 5+8
2. Compute 7*4
3. If the chance of getting a disease is 10%, how many people out of 100 would be expected to get
the disease?
4. If 5 people all have the winning number in the lottery and the prize is 2,000 Cedis, how much
will each of them get?

General financial literacy


5. Suppose you had 100 in a savings account and the interest rate was 2% per year (2 in year
one). After 5 years, how much do you think you would have in the account if you left the money
to grow: A: more than 102, B: exactly 102, C: less than 102?
6. Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per
year. After 1 year, would you be able to A: buy more than, B: exactly the same as, or C: less than
today with the money in this account?
7. Suppose that in the year 2011, your income has doubled and prices of all goods have doubled
too. In 2011, how much will you be able to buy with your income: A: more than today, B: the
same, or C: less than today?

Debt literacy
8. You owe 100 and you are charged 20% interest per year compounded annually. If you never
pay anything off, you pay 20 interest in the first year. In the second year, do you pay A: less
interest, B: the same interest, C: more interest?
9. You owe 300 with 1% interest charged each month (i.e. 3). You pay an installment of 3
each month. How many years would it take to eliminate your debt: A: less than 5 years, B: 5-10
years, or C: never (you will continue to be in debt)?
10. You take a loan of 100. To repay, you are given the following two options: 1) Pay 12 monthly
installments of 10 each; 2) Pay back 120 a year from now. Which is the better offer: A:
option 1, B: option 2, or C: are they the same?

(Questions were developed based on {Lusardi 2007 #343 /footcit} and {Lusardi 2009 #166
/footcit}).

- 157 -

Appendix 2: List of borrower sacrifices


Interviewers asked each respondent about the following list of sacrifices
1)

Reduce food quantity/quality (cut down eating)

2)

Reduce education (e.g. taking children out of school)

3)

Work more than usual (e.g. take additional labor, work longer hours, on Sundays, and when ill)

4)

Postpone important expenses (e.g. for health, housing, business assets etc.)

5)

Deplete your financial savings (e.g. money in the house or in a savings account)

6)

Borrow anew to repay (take an additional loan from another lender)

7)

Sell or pawn assets (e.g. jewelry, cattle, productive or household assets)

8)

Seizure of assets (MFI takes property by force to make up for missed payment)

9)

Use family/friends' support to repay

10) Suffer from shame or insults (also gossip about you/exclusion from a contract)
11) Feel threatened/harassed by peers/family/loan officer
12) Suffer psychological stress yourself or in your marriage
13) Other

Respondents ranked the acceptability and frequency of each sacrifice on a scale from 1 to 4.
Easily acceptable, Only just acceptable, Not really acceptable, Not acceptable at all.
Once in past year, 1-3 times in past year, > 3 times but not often, Frequently in past year

- 158 -

Appendix 3: Full regressions as in table 4, including detailed controls


Dependent variable: Over-indebtedness
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Gender
Age
Ethnic_background_Dagbone-Dagomba
Ethnic_background_Ewe
Ethnic_background_Ga
Ethnic_background_Gurma/Other
Marital_status_Married
Marital_status_Separated
Marital_status_Single
Marital_status_Widowed
Number_of_children
Household_size
Housing_type_Living in family housing for free
Housing_type_Living in your own house/Other
Employment_status_Not self-employed

(1)
0.843**
(0.057)
0.988
(0.008)
1.002
(0.003)
1.003
(0.004)
1.007
(0.004)
1.304
(0.440)
2.857***
(1.087)
6.150***
(2.761)
5.160*
(5.024)
0.995
(0.008)
1.005
(0.005)
0.987**
(0.006)
2.073**
(0.680)
1.010
(0.021)
0.548
(0.462)
0.786
(0.369)
1.871
(0.867)
2.331
(1.457)
0.355
(0.263)
0.265
(0.314)
0.617
(0.516)
0.562
(0.522)
1.017
(0.144)
1.127
(0.102)
0.464*
(0.209)
0.998
(0.412)
1.718
(1.645)

(2)
0.843***
(0.025)
0.988
(0.009)
1.002
(0.002)
1.003***
(0.001)
1.007***
(0.001)
1.304
(0.238)
2.857***
(0.911)
6.150***
(3.983)
5.160***
(2.921)
0.995
(0.008)
1.005***
(0.002)
0.987***
(0.005)
2.073*
(0.802)
1.010
(0.023)
0.548
(0.420)
0.786
(0.485)
1.871***
(0.409)
2.331***
(0.236)
0.355**
(0.187)
0.265
(0.355)
0.617
(0.397)
0.562
(0.376)
1.017
(0.194)
1.127**
(0.066)
0.464*
(0.205)
0.998
(0.185)
1.718
(2.309)

- 159 -

(3)
0.856***
(0.030)
0.986
(0.009)
1.002*
(0.001)
1.004***
(0.001)
1.007***
(0.001)
2.513***
(0.658)
2.824***
(0.912)
6.229***
(4.210)
5.410**
(4.103)
0.993
(0.008)
1.005**
(0.002)
0.987**
(0.006)
3.068***
(1.195)
1.008
(0.024)
0.645
(0.440)
0.800
(0.532)
1.950***
(0.481)
2.527***
(0.383)
0.356**
(0.176)
0.259
(0.352)
0.577
(0.373)
0.617
(0.403)
1.021
(0.193)
1.123**
(0.062)
0.468*
(0.184)
1.135
(0.204)
1.997
(2.437)

(4)
0.875***
(0.016)
0.987**
(0.007)
1.002*
(0.001)
1.005***
(0.002)
1.008***
(0.001)
2.089***
(0.287)
2.579***
(0.693)
6.073***
(3.798)
7.089***
(3.547)
0.997
(0.006)
1.005**
(0.002)
0.985***
(0.005)
2.618**
(1.009)

0.393
(0.242)
0.928
(0.540)
1.959***
(0.318)
2.059***
(0.512)
0.297**
(0.165)
0.234
(0.384)
0.518
(0.222)
0.401
(0.268)

1.142***
(0.030)
0.376***
(0.139)
1.165
(0.223)

(5)
0.880***
(0.028)
0.992*
(0.004)
1.001
(0.002)
1.003**
(0.001)
1.006***
(0.001)
1.513**
(0.284)
2.924***
(0.755)
5.930***
(2.770)
5.049***
(1.314)
0.996
(0.007)
1.004**
(0.002)
0.985***
(0.003)

Appendix 3 continued.
Main_occupation_On-selling

2.528*
2.528***
2.603***
1.976**
(1.320)
(0.527)
(0.504)
(0.608)
Main_occupation_Production of goods/Other
2.165
2.165
2.228*
1.695
(1.622)
(1.021)
(1.052)
(0.687)
Health_problems_Minor/Strong heath problems
1.938*
1.938*
2.011*
1.922**
(0.698)
(0.722)
(0.765)
(0.593)
Level_of_education_Primary
1.840
1.840
2.020
(1.454)
(2.169)
(2.253)
Level_of_education_JSS/JHS/Middle School
1.457
1.457
1.601
(1.196)
(1.261)
(1.334)
Level_of_education_Secondary/vocational/technical/commerc
2.301
2.301
2.585
(2.146)
(2.287)
(2.515)
Level_of_education_Tertiary, university graduate
0.500
0.500
0.722
(0.585)
(0.677)
(0.945)
Income_volatility_Little volatility
9454958.000*** 9454958.000*** 5827024.000*** 5392843.000***
(14800000.000) (9325882.000) (7143987.000) (8835741.000)
Income_volatility_Strong volatility
6031196.000*** 6031196.000*** 3596522.000*** 3798558.000***
(9562575.000) (6412996.000) (4558067.000) (6590799.000)
1.000
1.000
1.000
Amount_of_savings_self_reported
(0.000)
(0.000)
(0.000)
1.511
1.511
1.556
Number_of_MFIs_crossborrowing
(0.768)
(0.999)
(0.932)
1.000
1.000
1.000
Total_amount_of_debt_disbursed
(0.000)
(0.000)
(0.000)
1.074*
1.074**
1.056*
1.057
Average_maturity_weighted
(0.045)
(0.035)
(0.034)
(0.041)
0.644
0.644
0.835
Group_or_individual_customer_ran
(0.255)
(0.173)
(0.325)
0.783*
0.783
0.770*
0.795*
General_fairness_of_MFIs_rank
(0.109)
(0.119)
(0.111)
(0.094)
0.364***
0.314***
Interaction_gender_with_loan_use
(0.065)
(0.090)
0.457
Interaction_methodology_with_loan_use
(0.290)
Main_lender_MFI 2
1.355
1.355**
1.260
1.560***
(0.663)
(0.161)
(0.179)
(0.159)
Main_lender_MFI 3
1.289
1.289
1.240
1.494***
(0.753)
(0.293)
(0.239)
(0.222)
Main_lender_MFI 4
2.082
2.082**
2.091**
1.787***
(1.001)
(0.640)
(0.632)
(0.395)
Main_lender_MFI 5
1.235
1.235
1.260
1.277**
(0.593)
(0.281)
(0.290)
(0.159)
Constant
0.000***
0.000***
0.000***
0.000***
0.380*
(0.000)
(0.000)
(0.000)
(0.000)
(0.190)
401
401
401
414
417
Observations (N)
0.2241
0.2241
0.2347
0.2196
0.1332
Pseudo R
0.3370
0.3370
0.3509
0.3311
0.2107
Nagelkerke's R

Odds Ratios. Robust Standard Errors in parenthesis. Standard Errors for models 2-4 clustered by Main_lender_among_my_MFIs
*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first category according to table 2.

- 160 -

Appendix 4: Logistic regression on a binary indicator of delinquency


Dependent variable: Delinquency (dummy)
Average_monthly_income
Total_assets_in_thds
Unexpected_expense_shocks
Expected_expense_shocks
Unexpected_income_shocks
Loan_use:_At_least_one_non-productive_loan_use
Returns:_Increase_not_sufficient_to_repay_loan
Returns:_No_permanent_increase_in_earnings
Returns:_No_investment
Literacy:_Score_numeracy_test
Literacy:_Score_general_financial_literacy
Literacy:_Score_debt_literacy_test
Constant
Controls
Observations (N)
Pseudo R
Nagelkerke's R

(1)
1.077
(0.111)
0.981
(0.017)
0.990
(0.007)
1.002
(0.008)
1.017***
(0.005)
1.291
(0.621)
10.384**
(10.793)
25.337***
(25.580)
1.275
(3.361)
1.045**
(0.018)
0.978**
(0.009)
0.980**
(0.009)
0.004*
(0.013)
Added
386
0.4158
0.4781

(2)
1.077
(0.124)
0.981
(0.017)
0.990
(0.010)
1.002
(0.006)
1.017***
(0.006)
1.291
(0.702)
10.384***
(4.615)
25.337***
(21.229)
1.275
(4.804)
1.045***
(0.011)
0.978*
(0.012)
0.980***
(0.007)
0.004**
(0.010)
Added
386
0.4158
0.4781

(3)
1.060
(0.108)
0.979
(0.019)
0.991
(0.010)
1.002
(0.006)
1.018***
(0.006)
1.083
(0.760)
10.850***
(4.797)
27.580***
(21.263)
1.082
(4.472)
1.046***
(0.011)
0.979*
(0.012)
0.980***
(0.006)
0.002**
(0.006)
Added
386
0.4180
0.4804

(4)
1.077
(0.111)
0.981
(0.017)
0.990
(0.007)
1.002
(0.008)
1.017***
(0.005)
1.291
(0.621)
10.717***
(5.046)
20.623***
(13.022)
2.304
(5.315)
1.041**
(0.017)
0.989
(0.018)
0.980***
(0.006)
0.003***
(0.003)
Added
413
0.3069
0.3652

(5)
0.962
(0.060)
0.991
(0.009)
0.996
(0.003)
0.999
(0.003)
1.010***
(0.001)
1.139
(0.376)
3.857***
(0.903)
10.189***
(7.691)
2.064
(3.202)
1.031**
(0.013)
0.990
(0.008)
0.991
(0.005)
0.004***
(0.004)
Added
410
0.1560
0.1919

Odds Ratios. Robust Standard Errors in parenthesis, for models 2-4 clustered by Main_lender_among_my_MFIs
*, ** and *** denote significance at the 10% 5% and 1% level.
For each categorical variable we have dropped one reference category, i.e. the first category according to table 2.
Delinquency is a binary variable if a respondent was at least 1 day late on a loan at the day of the survey.

- 161 -

An overall conclusion to this thesis


This PhD addresses one of the most important challenges of the microfinance industry at
present: the over-indebtedness of microborrowers. It responds to the need for both research
and practice to understand microfinance over-indebtedness from a perspective of customer
protection. It suggests a definition and measurement of over-indebtedness that is appropriate
for customer protection purposes, analyses the potential causes and consequences of overindebtedness in microfinance and, putting the customer-protection definition into practice,
empirically identifies the extent of over-indebtedness in an African microfinance market. It
gives a voice to microborrowers, revealing their coping strategies for debt repayment and
exposing unparalleled detail about the sacrifices they experience related to their debt. It
contrasts the risk management and customer protection perspectives and develops a model of
risk-management oriented indicators that predicts the over-indebtedness of microborrowers
according to the customer protection perspective. It provides unprecedented empirical insights
into which borrower-level factors are related to over-indebtedness and may be causes of overindebtedness.
Due to the origins of microfinance as a social industry that addressed the exclusion of a large
proportion of the worlds poor population from the formal financial system and particularly
from access to credit, over-indebtedness did not seem to be a topic of relevance to the
microfinance industry until very recently. The target group of microfinance was characterised
explicitly as lacking credit. Furthermore, the extremely high repayment rates of
microfinance borrowers around the world seemed to confirm the repayment capacity of this
target group. Only with the repayment crises that several microfinance markets experienced in
the aftermath of the global financial crisis, did over-indebtedness emerge as a topic that might
have broader relevance for the microfinance industry. This period of crisis coincided with
increasing criticism of the commercialisation of the microfinance industry and a shift back
from profitability concerns to concerns of customer protection and social impact. It became
necessary to analyse over-indebtedness not only from the risk management point of view but
also with the primary aim of protecting microborrowers.
This dissertation represents a first response to these two recent developments in microfinance,
the broad recognition of concerns about over-indebtedness, and the industrys focus on
protecting microfinance customers from harm. It frames the concern for customer protection
as a question of over-indebtedness, bringing it to the attention of the industry in a new and
potentially more powerful way. Being one of the first pieces of research in microfinance on
- 162 -

the topic of over-indebtedness, particularly with a focus on customer protection, the


dissertation works within the limitations of a lack of previous literature in the field, a lack of
an agreed definition of over-indebtedness, and a lack of data on over-indebtedness in
microfinance markets. It therefore leaves much room for further analysis of the topic of
microfinance over-indebtedness, for better datasets including time-series data and data from
randomized experiments, for research with higher levels of external validity. This dissertation
focuses instead on pioneering work on the following main research questions:
a) The definition and measurement of over-indebtedness from a perspective of customer
protection
b) The consequences of over-indebtedness
c) The extent of over-indebtedness in microfinance markets
d) The identification of easier-to-measure predictors of over-indebtedness comparing the
risk management perspective on over-indebtedness to the customer-protection
perspective
e) The identification of potential causes of over-indebtedness and, as a first step, factors
that are empirically related to the over-indebtedness of microborrowers
The first section of this conclusion will briefly summarize the key findings of the dissertation
and formulate first answers to the above questions. The second section will indicate
implications of this PhD dissertation for other fields of research. A third, concluding section
will comment on the limitations of the dissertations approach to research and indicate needs
for further research.

- 163 -

1. Key findings of the dissertation


a) The definition and measurement of over-indebtedness
According to Chapter 1 of this dissertation, a microfinance customer is over-indebted if
he/she is continuously struggling to meet repayment deadlines and has to make unduly high
sacrifices related to his/her loan obligations that have more than transitory effects. This
definition takes the subjective experiences of borrowers with their debt into account and
acknowledges that microborrowers often go to great lengths to avoid defaulting on their loans
or paying instalments past their due date. While subjectivity limits the universal validity of
findings based on this definition, the definition is suitable for the purpose of customer
protection research that aims to avoid suffering and harm to microborrowers and not solely to
the portfolio quality of MFIs. It reflects that suffering is a subjective phenomenon.
In spite of parallels, the definition differs from an approach that considers over-indebtedness a
question of negative loan impact or of regret on behalf of borrowers regarding their borrowing
decision. In some cases, negative impact or regret may not be severe enough to define a
borrower as over-indebted, in others over-indebtedness may exist without having been caused
directly by the loan or, as for example with emergency loans for urgent medical expenses,
without the borrower regretting the loan.
This definition can be applied as a practical measurement of over-indebtedness by means of
survey work. It requires detailed data on the sacrifices of microborrowers, their persistence
over time and their acceptability based on the borrowers own evaluation. It is not suitable for
legal purposes or any situations where borrowers have strong incentives for dishonesty in
reporting their sacrifices. For research purposes with a customer-protection however, it is the
best way to understand the suffering of microborrowers as perceived and experienced by them
in their historical, geographical and individual circumstances.

b) The consequences of over-indebtedness


Chapter 2 of this dissertation shows that the consequences of over-indebtedness on the overindebted individual range from material costs and the potential of further impoverishment to
psychological and physical health effects. They also include social consequences such as
social stigma, peer pressure, and the loss of borrowers social networks.
For MFIs, aside from direct costs to lenders in the form of loan loss provisions and write-offs
on the lending portfolio, there also exist many secondary consequences of over-indebtedness
- 164 -

for MFIs. Operating costs may increase, market size may decrease, customer satisfaction may
deteriorate, and income may be postponed or even lost. Furthermore, there are internal and
external reputation effects that can lead to, for example, problems of staff attrition, the
uncontrolled spreading of delinquency problems, funding cuts, and political constraints faced
by the MFI.
In addition to the lenders and borrowers directly affected by over-indebtedness, third-party
stakeholders can face negative consequences. Spillover effects can affect borrowers who are
not over-indebted or MFIs that do not initially have an over-indebtedness problem in their
portfolio. Large over-indebtedness crises can affect complete microfinance markets, the
financial system, and the wider industry.
Chapter 2 shows that the consequences of over-indebtedness reach far beyond the direct
effects of delinquency and default that are the most commonly discussed in the debate about
over-indebtedness. It suggests that cultivating an awareness of these complex potential
downsides of debt will help MFIs to develop a more fine-tuned approach to lending, both in
terms of selecting borrowers and determining repayment capacity and in terms of designing
loan product features.

c) The extent of over-indebtedness in microfinance markets


The analysis of the consequences of over-indebtedness exposes the magnitude of the overindebtedness risk in terms of its possible severity. A review in Chapter 2 of the existing
studies on over-indebtedness in microfinance and the related research about the negative
effects of microlending demonstrates that the second dimension of risk assessment, its
likelihood, is also not as negligible as the microfinance industry had long assumed.
In spite of the large supply gap that microfinance is addressing, there have been overindebtedness crises in several microfinance markets to date. In maturing markets, the supply
gap of microfinance can turn into oversupply, at least at the sub-national level. Furthermore,
macroeconomic crises can trigger substantial over-indebtedness in markets that previously
appeared healthy and growing. Eventually, even in non-crisis markets, where delinquency and
default are low, there can be an underlying over-indebtedness problem from the perspective of
customer protection. A relevant share of borrowers may be suffering considerably from their
debt even if they are still managing to repay their loans on time.

- 165 -

To anticipate future developments of the over-indebtedness risk in microfinance and draw


conclusions about the best strategy to address the challenge of over-indebtedness, more
research is required that analyses the extent of over-indebtedness in microfinance markets.
Research must investigate whether over-indebtedness is a rising problem in microfinance
markets globally, whether it has reached its over-all peak for the industry, whether it will
return periodically in relation to economic crises or whether it will peak separately in
microfinance markets as they mature. It should determine a typical level of over-indebtedness
in non-crisis markets so that the industry can determine the required response.
In one non-crisis market, in Accra (Ghana), Chapter 3 of this dissertation has determined the
level of over-indebtedness among microborrowers according to the sacrifice-based definition
that defines over-indebtedness from the customer protection perspective. Although
delinquency and default in this market are low, 30% of microborrowers are over-indebted;
i.e., they suffer from their debt to an extent that they consider unacceptable. The borrowers in
our sample display a relatively high willingness to sacrifice for their loans and tend to
consider the most common sacrifices such as working harder or postponing important
expenses as fully acceptable. Such acceptable coping strategies do not contribute to our
measurement of over-indebtedness. However, borrowers also make more severe sacrifices
such as going hungry, selling their assets and taking their children out of school. They
experience threats, harassment and shame related to their debt. From the viewpoint of
customer protection efforts, the more than transitory prevalence of such sacrifices among
microfinance borrowers at an unacceptable level is a reason for concern.
Nevertheless, although borrowers clearly attribute the sacrifices they report to their debt, the
study does not establish causality between microlending and sacrifices or over-indebtedness.
Future research would have to determine counterfactual sacrifice conditions without access to
microloans. Optimally, future impact studies, including upcoming randomised control trials,
would explicitly take the sacrifices of microborrowers into account and measure potential
negative impacts in addition to potential positive impacts. In the meantime, it is already
highly important for customer protection efforts to know to what extent and in what way
microborrowers struggle with their loans and how conditions can be improved
independently of the underlying causal relationship.

- 166 -

d) Risk-management predictors of the customer-protection perspective on over-indebtedness


Given that the measurement of over-indebtedness from the customer protection perspective
requires survey work that is costly and time-intensive, the microfinance industry needs to
identify indicators that are easier to measure and predict how much over-indebtedness one
would find in a sample if customer survey data were available. Testing the four most common
risk management indicators of over-indebtedness as predictors of a dummy variable for overindebtedness according to the customer protection definition, the third chapter of this
dissertation develops a logistic regression model that predicts 72.6% of cases in the Ghana
sample correctly. The debt-to-income ratio and delinquency are highly significant predictors
of over-indebtedness. The amount of disbursed debt is not consistently significant, and at least
at the low level of multiple borrowing in our sample, cross-borrowing is not related overindebtedness.
However, the models correct prediction of 72.6% of cases is primarily due to the structure of
the sample (70% of borrowers are not over-indebted) and the actual predictive power of the
risk management based indicators remains low. Even the highly significant predictors, the
debt-to-income ratio and delinquency, recognise only few of the over-indebtedness cases
correctly. At best, a debt-to-income ratio of 50% may function as an indicator of the aggregate
level of over-indebtedness in a population. Future research would have to confirm whether
this indicator holds in other markets.
The risk management perspective and the customer protection perspective are thus closely
related but different nevertheless. As many borrowers go through significant sacrifices to
avoid delinquency, many more customers can be over-indebted than the risk management
point of view would recognise. Contrary to common assumptions in the microfinance
industry, good repayment statistics do not prove the benefits of microloans.
To a certain extent, the microfinance methodology has reduced risk (e.g., in the form of peer
selection in group lending and in instances where social collateral prevents moral hazard).
However, a significant part of the risk that microlending appeared to have eliminated has in
fact just been shifted from lending institutions back to borrowers. The personal sacrifices of
microborrowers are often what buffer income volatilities or adverse shocks and prevent them
from affecting the portfolio quality of MFIs at a high cost from the customer protection
perspective. Instead of focusing on risk management and sustainability or even profitability,
the microfinance industry thus must directly consider the sacrifices of its customers. Doing so
not only will ensure customer protection but also, as the results of the analysis of over- 167 -

indebtedness consequences on MFIs indicate, is likely to make business sense in the long
run.1
Furthermore, recent repayment crises in microfinance markets have shown that when the risk
absorption capacity of microborrowers is over-extended and even severe sacrifices are no
longer sufficient to buffer repayment difficulties, the risk that had been shifted to borrowers
can shift back into MFI portfolios. Good customer protection therefore contributes to good
risk management. Identifying over-indebted borrowers is an excellent means for identifying
borrowers at risk of future delinquency.
Providing that high-level findings can be generalized to the microfinance industry in spite of
the context dependency of over-indebtedness and the subjective elements of our
measurement, the main policy implication of these findings in Chapter 3 is that microfinance
institutions must foster awareness of the downsides of debt and listen better to their
customers. This awareness is likely to have practical repercussions, for example, on loan
officers incentives, MFI regulation, investor and donors expectations, and on impact
measurement in microfinance. In general, the microfinance industry should place at least as
much importance on managing its impact as it does on managing its financial sustainability,
and it should focus on not only the potential positive impact but also the potential negative
consequences of debt on microborrowers.

e) Related factors to and potential causes of over-indebtedness


According to the conceptual analysis in Chapter 1, over-indebtedness can be caused by
irresponsible decisions on behalf of borrowers that may be due to cognitive biases and
psychological misrepresentations such as hyperbolic discounting, imperfect mental
representations of borrowing decisions, and debt-conducive attitudes. It can originate from
borrowers reactions to social pressures such as materialism and social comparison. The risk
of over-indebtedness varies with a borrowers socio-demographic and economic
characteristics. Over-indebtedness risks can also result from external influence factors such as
an institutional environment that lacks legal enforcement mechanisms or personal debt-relief
procedures, macroeconomic crises, and adverse economic shocks at the borrower level.

1 This dissertation does not expand on the detailed practicalities of MFIs measuring over-indebtedness according to customer
satisfaction. While it encourages MFIs to work with this level of customer data from time to time, there are significant
challenges for MFIs to do sacrifice interviews themselves and total anonymity needs to be guaranteed to respondents for data
to be sufficiently reliable. Cooperation on an market level with qualified research firms might be useful and are in fact
already on their way in several microfinance markets.

- 168 -

Finally, lenders can increase the risk of over-indebtedness by means of, for example, an
aggressive marketing and sales focus, with an inappropriate product offer such as instalment
schedules that are not appropriate for a borrowers repayment capacity, an exaggerated zerotolerance policy, and conducive lending and collections procedures such as lax evaluations of
repayment capacity, nontransparent terms and conditions, and coercive collection practices.
This analysis of the causes of over-indebtedness in Chapter 1 indicates that there exists a
broad range of levers for measures against over-indebtedness. It shows that the response to the
challenge of over-indebtedness is not necessarily a reduction of credit and thus a return to a
higher degree of financial exclusion. Although some causes are best addressed with more
conservative lending policies, others suggest that over-indebtedness can be reduced by means
of, among others, more appropriate credit products, flexibility in instalment schedules,
reasonable options for rescheduling, customer communication that is more appropriate for
borrowers mental abilities. Some borrowers may need less credit than the microfinance
industry currently offers but many may rather need better credit. The microfinance industry
needs to further develop its lending methodology and non-credit product offerings.
Although, for some of these measures, MFIs will have to accept trade-offs in the interest of
social performance or at least in the interest of maintaining a minimum level of customer
protection, many of these measures will make good business sense, improving portfolio
quality and contributing to business growth in the long run through enhanced customer
satisfaction.
This conceptual analysis of over-indebtedness causes is based however on the literature and
many of the findings have been transferred from the historical and geographical context of
high-income country consumer finance to the microfinance context. To develop more
concrete policy recommendations with regard to effective measures against overindebtedness, the microfinance industry needs to know which factors are empirically related
to over-indebtedness and which borrowers are most at risk. Optimally, it would tailor
solutions to the root causes of the challenge of over-indebtedness in a given context. Research
on the factors related to over-indebtedness is thus both in the interest of the risk management
of MFIs and in the interest of the promoters of microfinance customer protection.
Based on econometric regression analysis, Chapter 4 of this dissertation investigates four
hypotheses of potential factors related to over-indebtedness. It indicates that in the study
context in Ghana, poorer microborrowers in terms of income as well as assets are more likely
to be over-indebted. In addition, there is an independent effect of adverse economic shocks to
- 169 -

a borrowers income and expenses. The likelihood of over-indebtedness strongly increases


among borrowers who attain only low returns on their investment and among borrowers who
use their loans at least in part for non-productive purposes. Over-indebtedness risks are also
higher for borrowers with low levels of debt literacy. Surprisingly, general financial literacy is
positively related to over-indebtedness, a finding that requires further qualitative research to
analyse what the exact relationship between general financial literacy and over-indebtedness
may be and why. Mere numeracy does not have a significant relationship with overindebtedness. With the exceptions of loan use and asset ownership, where causality could
move in both directions, the above factors have a considerable likelihood of being causes of
over-indebtedness. Nevertheless the risk of endogeneity in the econometric analysis and the
lack of time series or experimental data prevent statistically reliable conclusions about
causality.
Breaking the logistic regression on the binary over-indebtedness variable down into separate
regressions for the different sacrifices that microborrowers make, Chapter 4 shows that the
same factors that are related to over-indebtedness are also related to which sacrifices
borrowers experience. Depending on their poverty level, the types of adverse shocks they
experience, their loan returns and their financial literacy, borrowers react to debt problems
with different coping strategies and display a different likelihood to experience the various
sacrifices. Moreover, in ordered logistic regressions on the four-point scales of sacrifice
acceptability and frequency, the same factors are related to how acceptable certain sacrifices
are to borrowers and how frequently a borrower repeats specific sacrifices. Poverty, for
example, makes extra work efforts harder to bear while increasing the likelihood that a
borrower who must make extra work efforts once will repeat them more frequently over time.
Based on the findings of Chapter 4, provided that they can be generalised to the wider
microfinance industry in spite of the context dependency of over-indebtedness and our
subjective measurement, there are policy recommendations for MFIs trying to avoid overindebtedness. MFIs should improve their evaluations of repayment capacity and may decide
to lend less to borrowers who are more at risk of over-indebtedness. In addition to the main
hypotheses tested in Chapter 4, the control variables paint a relatively precise picture of the
borrowers who are most at risk of over-indebtedness. However, the policy implications of the
above results reach far beyond recommendations to exclude high-risk borrowers from credit.
In many cases, a more appropriate response to over-indebtedness risks may be to enhance the
flexibility of loan products and better tailor them to the needs of microborrowers. Depending
on individual circumstances, longer maturities can reduce over-indebtedness in the face of
- 170 -

low investment returns, more flexible instalment schedules allow borrowers to deal better
with low and volatile incomes and with adverse economic shocks. Temporary shocks may
also represent reasons for loan rescheduling or even for loan upgrades. Other approaches
could be based on the product mix of microfinance, the promotion of savings and insurance,
on financial literacy training, or - in countries where this may not be overly ambitious - on
government action with regard to safety nets and debt relief for over-indebted borrowers.

- 171 -

2. Implications for other fields of research


In addition to its direct implications on over-indebtedness in microfinance, this dissertation
might feed back into some of the other fields of research from which it initially borrowed. The
validity of its findings relative to other areas of development or for industrial countries is
subject to clear limitations, but the dissertation may create an impulse upon other threads of
literature and suggest possible explanations for phenomena in other disciplines that would
have to be tested by scholars within these respective domains. This section will reflect on the
potential implications of the present research on over-indebtedness in microfinance, covering
the topics of
a) Gender in microfinance
b) The financial literacy of microfinance customers
c) Credit bureaus as an example of meso-level financial sector development policies
d) Development policy beyond microfinance
e) Financial services consumer protection in industrialised countries.

a) Gender in microfinance
There is extensive literature on gender issues in microfinance, investigating the impact of
microfinance on gender empowerment (see Mayoux (2011) for a comprehensive framework)
and, although less frequently, the impact of gender on microfinance operations. With regards
to the former, microfinance is suggested to have a positive effect on gender equality,
empowering women borrowers to take more (financial) responsibility in the household and to
organise the interests of women on a community level (Hashemi et al., 1996; Kabeer, 2001;
Pitt et al., 2006). While this dissertation does not analyse the impact of microloans or include
data on the relationship between men and women, its findings about over-indebtedness are
more in line with the critical studies that suggest a potential negative effect of microfinance
on womens empowerment at least in the case of some borrowers, increasing the vulnerability
of female debtors (Goetz, Sen Gupta, 1996; Mayoux, 2001; Garikipati, 2008). The overindebtedness that this dissertation revealed is a sign of vulnerability. It underlines the request
by Mayoux (2011) to take gender issues more seriously, as opposed to simply expanding
access to finance by women, by ensuring real empowerment and an enhancement of well-

- 172 -

being rather than capturing female savings for the sake of MFI profitability and triggering a
feminisation of debt.
With regards to the latter, the gender of microborrowers is said to influence the impact of
microlending. Women represent a particularly vulnerable and needy target group and use
microloans and their income more to the benefit of their households than do men. (Thomas,
1990; Engle, 1993; Khandker, 2003). Even the gender of children in a borrowers household
can impact borrowing behaviour (Agier et al., 2012). These streams of research are likely to
relate to the findings of Chapter 3 that differences exist between the over-indebtedness of
male and female borrowers. Chapter 3 revealed that women experience different sacrifice
situations than are men, for example resorting to economic coping strategies such as assets
sales or loan recycling less often. It also showed that for women, over-indebtedness based on
sacrifices seems to have a slightly different relationship to the typical risk management
indicators of over-indebtedness than for men. This observation may relate to the different
usage of loans and to the different level of attention to household welfare observed between
genders. Together with the above mentioned gender research, these findings may indicate that
the consequences of over-indebtedness on the household are more severe in the case of female
borrowers.
Additionally, the control variables in Chapter 4 show that a borrowers gender is significantly
related to the likelihood of over-indebtedness. A supplementary analysis of the marginal
effects of the control variables in our parsimonious model of the main logistic regression
indicates that this relationship is in favour of women: ceteris paribus a female borrower has a
16.5% lower likelihood of being over-indebted than does a male borrower. While overindebtedness and a visibly poor repayment performance are not the same, they are related.
This finding may thus relate to the gender research that has proven the repayment rates of
women to be better than those of men (Khandker et al., 1995; Armendriz, Morduch, 2010),
confirming it from a customer protection point of view. The causes for this difference could
of course be manifold. Women may differ from men in terms of the sociological and
psychological influence factors on over-indebtedness that Chapter 1 has identified but that,
with the given data set, could not be tested in Chapter 4. For example, the gender literature
suggests that women behave differently in financial decisions and tend to be more risk averse
(Jianakopolos, Bernasek, 1998; Olsen, Cox, 2001). This risk aversion may help women avoid
certain risks of over-indebtedness, making them use their loans more carefully or inciting
them to better prepare for the risk of adverse shocks. The next paragraph adds the reverse
- 173 -

perspective that differences may be due to gender effects on the treatment of borrowers by
MFIs.
With regards to the effects of gender on microfinance operations, the gender of borrowers has
been shown to impact the lending decisions of MFIs. Specifically, MFIs seem to discriminate
against women by granting them lower loan amounts (Agier, Szafarz, 2010, 2011). This may
influence over-indebtedness risks. For example, they could explain the lower overindebtedness risk of women borrowers, as loan officers seem to be stricter in their assessment
of repayment capacity when a borrower is a woman. Of course, the over-indebtedness
problem applies only to certain borrowers and other women may have benefitted from higher
loan amounts if they had not been discriminated against. Nevertheless, the higher overindebtedness risk for men might indicate that the positive discrimination of male borrowers by
MFIs is not always in the interest of male clients.
Finally, taking gender differences seriously can help MFIs design their products and
operations in a way that fits the needs of clients better. Gurin (2011) suggests that
microfinance services should be built on an improved understanding of the specific needs and
motivations of women in dealing with financial services and that they should take the preexisting informal financial practices into account that tend to display strong gender influences.
This is in line with this dissertations suggestion to tailor products better to clients needs and
to understand the complexity of the financial lives of microborrowers and the vulnerability
this implies, even when repayment rates are high. A more detailed analysis of gender
differences in the sacrifices that borrowers make and how acceptable the different coping
strategies are to men and women could help in developing the required understanding of
gendered finance needs.

b) The financial literacy of microfinance customers


One of the favourite solutions of todays policy makers to combat over-indebtedness is
consumer education to increase the financial literacy of microfinance customers (Gardeva,
Rhyne, 2011). See Gurin (in press) for a description of recent policy initiatives and their
importance in customer protection policies. The field of financial literacy research, both in
developed countries and in the microfinance context, comprises studies that reveal the
shockingly low levels of financial literacy in various countries and that relate a lack of
financial literacy to unfavourable financial decisions and financial difficulties (Lusardi,
Mitchell, 2007; Tiwari et al., 2008; Lusardi, Tufano, 2009; Bucher-Koenen, Ziegelmeyer,
- 174 -

2011; Gathergood, 2012). Another stream of research measures the effects of financial
literacy trainings and tries to identify the most effective type of training (DeLaune et al.,
2010; Drexler et al., 2010; Altman, 2012).
The present dissertation contributes to this literature by highlighting the relationship of debtspecific literacy to over-indebtedness. It suggests that debt literacy trainings that teach
borrowers to better understand loan contracts and repayment obligations can indeed reduce
the risk of over-indebtedness. At the same time, this dissertation sheds more light on the type
of training required: according to the available data, financial literacy education programs
need to be specifically tailored to the goal of promoting the understanding of loan contracts.
Numeracy training does not seem likely to have any effects on over-indebtedness, at least in
the Ghanaian sample where numeracy is already relatively good. And general knowledge
about savings and inflation even seemed to have negative effects potentially because people
who feel more knowledgeable about financial matters in general become less careful in their
financial decision making.
The fact that debt-literacy specific knowledge seems to be more helpful than general financial
skills could be in line with the studies that find general financial literacy trainings to have
very limited effects on behaviour and on reducing financial problems (Willis, 2008; Karlan,
Valdivia, 2011; Altman, 2012; Meier, Sprenger, in press). It is in line with research on the
intersection of financial literacy studies and behavioural economics: simple rules of thumb
that help borrowers with their decision making may be more effective than complicated
financial trainings that overload borrowers with knowledge that is hard to memorise and
difficult to put into practice (Tiwari et al., 2008; Drexler et al., 2010; Altman, 2012). The
same applies to financial literacy trainings being more effective in an informal format rather
than in the form of standard classroom trainings. For example theatre plays and soap operas
on TV convey indirect messages about the importance of savings and about caution with
regards to debt.
Finally, there is important criticism against relying on financial education as the sole solution
to customer protection against over-indebtedness. The following paragraphs are substantially
based on the critical analysis of Gurin (in press). Firstly, financial education cannot replace
financial market regulation or poor social security systems. Placing emphasis on financial
education to avoid over-indebtedness implicitly gives the responsibility for debt problems to
the customers alone and does not recognise most of the main causes of over-indebtedness.
This is in line with the framework of Chapter 1 that has identified the broad range and
- 175 -

complex interrelatedness of the causes of over-indebtedness. Lenders exert important


influences upon over-indebtedness that should rather be addressed by other means such as
financial services regulation (Willis, 2008), codes of conduct and pressure by investors. There
are also important external, environmental influences and borrower-level influences that are
not under the borrowers control. Addressing them may require for example social safety nets.
Holding only borrowers who may make erroneous decisions responsible for the overindebtedness problem thus seems to be an easy way out.
Secondly, an over-emphasis on the need for financial literacy among the poor ignores the
findings of economic anthropology that poor individuals tend to have sophisticated
mechanisms in place to manage their financial situations (Rutherford, 2000; Collins et al.,
2009; Morvant-Roux, 2009b; Gurin et al., 2011a). Resulting from the social embeddedness
of finance, their mechanisms often follow a different logic than the straight-forward economic
rationality that literacy trainings teach. They include for example the importance of social
interactions and community-based safety nets, non-material implications of debt relationships,
transaction costs of formal financial services, commitment features of in-kind savings, and so
on (Morvant-Roux, 2009a; Gurin et al., 2011b; Gurin et al., 2012). The complexity of
financial literacy is therefore much higher than what normal literacy training suggests. This
may partly explain the findings in Chapter 4 that numeracy and general financial literacy do
not seem to work as solutions against over-indebtedness. Taking the findings of economic
anthropology further one could assume that training programs that impose a pure economic
rationale on borrowers which is detached from their local social realities can even have
negative effects in the long term. They may lead to an isolation of individuals from the
circulation of wealth in local communities and, in the long run, to a destruction of such interpersonal and sometimes inter-generational mechanisms of insurance for emergencies and
savings for old age.

c) Credit bureaus as an example of meso-level financial sector development policies


The second measure against microfinance over-indebtedness that is currently favoured by
many policy makers is the promotion of credit bureaus, a measure on a meso-level of
financial sector development. Credit bureaus are industry-level bodies that collect and match
borrower information across lending institutions, optimally integrating the data from the
formal financial sector and from microfinance lenders (Christen et al., 2012). They provide
information back to their reporting members, enabling them to improve their assessment of
- 176 -

creditworthiness. This section will briefly review the literature on credit bureaus and then
focus on details where this dissertations findings on over-indebtedness relate to the research
on credit bureaus.
Credit bureaus are considered one of the most important opportunities for the microfinance
sector and their lack one of the biggest current obstacles (Gardeva, Rhyne, 2011). The
promoters of credit bureaus argue that credit bureaus expand access to credit as they reduce
the risk for lending institutions and help good borrowers to build a positive credit history and
improve their access to loans (Christen et al., 2012). Also, national credit registries are said to
lower the interest rates on loans by reducing the cost of risk assessment and improving client
segmentation (Gardeva, Rhyne, 2011). Based on a randomised experiment with a microlender
in Guatemala, de Janvry et al. (2010) provide empirical evidence for both the credit expansion
and the efficiency effects in a microfinance context.2 This is in line with a study from the nonmicrofinance corporate lending sector in Eastern Europe and the Soviet Union that also finds
increased access to credit for companies and lower credit costs (Brown et al., 2009). It has
also been confirmed on a macroeconomic level, by a panel analysis of the effects of national
identification systems, which result in better functioning credit bureaus, on the extent and cost
of lending in 172 countries between 2000 and 2008 (Giannetti, Jentzsch). A macro analysis by
Ahlin et al. (2011) in contrast shows that the evidence on the effects of credit bureaus is not
clear. They analyse data from 373 MFIs merged with macroeconomic national data. With
regards to the effects of credit bureaus, they find that private credit bureau coverage (i.e.,
breadth of information) is correlated with faster loan-size growth, while the credit information
index (i.e., quality of information) is strongly associated with slower loan-size growth.
However, the main function of credit bureaus is currently seen in a reduction of multiple
borrowing and thus over-indebtedness (Gardeva, Rhyne, 2011; Lyman et al., 2011; Christen
et al., 2012). This dissertation has underlined the need for measures against over-indebtedness
in microfinance as an important priority for the microfinance sector. At the same time, the
framework of the causes of over-indebtedness in Chapter 1 highlights the complexity of the
matter and notes that the spectrum of factors that can lead to over-indebtedness is very broad.
Many of the causes identified in Chapter 1 cannot be addressed by credit bureaus. In fact, the
only cause of over-indebtedness that credit bureaus combat in a direct form is the
unreasonable behaviour of microborrowers to a certain extent, a credit information registry

2 These findings for the microfinance context may be specific to group-level information sharing rather than to individual
information in credit bureaus (McIntosh et al., in press).

- 177 -

could help lending institutions identify borrowers that are more likely than others to make
unreasonable decisions with their loan applications. It can reduce lending to such borrowers.
At the same time, credit bureaus are far from perfect for this purpose. They may provide
certain incentives for microborrowers to avoid a negative credit history and take loan
obligations more seriously (Chomsisengphet, Elul, 2006; de Janvry et al., 2010), but that is a
reduction of moral hazard, not of repayment difficulties among honest borrowers. Empirical
results from the introduction of a credit bureau in Albania in 2008 confirm this disciplinary
effect as the main impact of credit bureaus (Behr, Sonnekalb, 2012). Credit bureaus do not
address the underlying causes of the questionable decisions made by borrowers. For example,
they do not address temptation, or a lack of understanding of loan contracts, or social
pressures, and so on. Nor are credit bureaus perfect instruments to identify the borrowers that
are likely to run into difficulties. They may not estimate the repayment capacity and
willingness to repay of all borrowers correctly and can lead to the exclusion of certain
borrowers that could have benefited from credit and who would have repaid without
difficulties, while others may continue borrowing to their detriment. In the microfinance
context, even the correct identification of individuals as the same debtors with several
different lenders is already very challenging; furthermore, data protection standards are low,
thus raising additional customer protection concerns (Lyman et al., 2011).
With regards to lending institutions and the causes of over-indebtedness that Chapter 1
identified on the side of MFIs, credit bureaus do not improve the incentives for MFIs to lend
more responsibly. To a certain extent, they improve the ability of MFIs to lend responsibly,
but they do not address the pressures for growth that tend to push MFI management into
irresponsible lending, nor do credit bureaus help to improve unfair lending procedures or
promote MFIs to tailor products to the needs of clients. In contrast, the effect tends to be the
other way around: As information sharing between lenders by contributing to and using a
credit bureau can entail certain risks for the market share of participating lenders, the
willingness of lenders to share information depends on the extent of competition and the level
of information asymmetries in a market (Brown, Zehnder, 2010; Lyman et al., 2011).
Eventually, as concerns external factors that can cause over-indebtedness, credit bureaus
represent an element of the institutional environment and can have an effect on the
competitive situation in a microlending market. They provide incentives against moral hazard
on behalf of borrowers and against adverse selection in borrower groups (de Janvry et al.,
2010). However, they cannot address the most frequently cited external cause of overindebtedness: they do not protect borrowers from adverse shocks to their income or expenses.
- 178 -

Supplementing the theoretical analysis of how credit bureaus relate to the causes of overindebtedness, Chapter 3 of this dissertation provides empirical evidence on the relationship
between multiple borrowing and over-indebtedness. This is important insofar as, according to
the literature cited above, the reduction of multiple borrowing is the major mechanism of how
credit bureaus are supposed to combat over-indebtedness. The findings of Chapter 3 are rather
disappointing to those who consider credit bureaus the silver bullet solution. In the Ghanaian
case study the number of loans a borrower has outstanding is not related to the likelihood of
being over-indebted. Chapter 3 suggests that there are good reasons why borrowers might
want to take multiple loans, so a direct relationship between multiple borrowing and overindebtedness is not to be expected. This chapter also hypothesises that multiple borrowing
may become a problem only at very high, unmanageable levels. The relationship between the
two variables may therefore emerge only in markets with much higher levels of multiple
borrowing than is the case in Ghana. This interpretation of the findings in Chapter 3 suggests
that credit bureaus may be a good solution against unsustainable levels of loan accumulation
in highly competitive and overbanked markets (or market segments). However, the Ghanaian
case proves that a reduction of multiple borrowing is not always the key to fighting overindebtedness: the level of multiple borrowing in this market is very low and nevertheless 30%
of borrowers are over-indebted. A credit bureau and reduction of multiple borrowing would
not be able to address this over-indebtedness.
Finally, the empirical findings reported at the beginning of this section indicate an expansion
of credit as a result of credit bureaus. This could also be interpreted as a sign of an increased
debt burden on individuals. The expansion is generally seen as positive based on the
assumption that credit is given to the more creditworthy individuals and is thus desirable. At
the same time, this dissertation has shown that the short-term risk management considerations
of MFIs who want to keep repayment discipline high are not always in line with the customer
protection perspective. Individuals may be creditworthy from a repayment point of view but
struggling excessively with their repayment obligations. The expansion of credit may partly
be based on the above-mentioned discipline effects of credit bureaus. According to this PhD
dissertation, pressure for repayment discipline may imply an increase in borrower sacrifices
and thus in over-indebtedness from a customer protection point of view. Also, the empirical
study by (Ahlin et al., 2011) cited above contrasts the common belief that credit bureaus
reduce default. All three indicators, private credit bureau coverage, public credit bureau
coverage, and the credit information index, are related to higher default levels among
microborrowers. The authors are unsure how to interpret these findings but suggest a potential
- 179 -

positive explanation that credit bureaus lead to the funding of more risky ventures and thus
higher default rates. This can be a desirable development if the levels of risk financing were
previously suboptimal. From a perspective of over-indebtedness it looks rather worrisome.
Of course, the above critical paragraphs about the effects of credit bureaus and their
relationship to the causes of over-indebtedness do not contradict the potential of promoting
credit bureaus in microfinance markets. Credit bureaus are most likely very useful and, given
all the obstacles to their development and the lengthy process of establishing functioning
credit bureaus (Lyman et al., 2011), the industry should build credit bureaus long before the
competitive situation in a market gets too critical. Nevertheless, this dissertation is a warning
that credit bureaus may not be a perfect solution to the over-indebtedness challenge, and it
proves that in cases like that of Ghana, other measures are necessary to prevent overindebtedness among microborrowers.

d) Development policy beyond microfinance


The above sections have identified several streams of research that are related to the findings
of this dissertation on microfinance over-indebtedness. All three examples represent fields of
literature that are situated within the larger topic of microfinance. However, the approach of
this dissertation may have implications beyond microfinance. It is related to development
policy and development research in general. Given the breadth of the development policy
field, this section will touch on the different elements only very briefly to highlight where the
main value of the over-indebtedness research for broader development policy lies.
The first contribution of this dissertation is to highlight the vulnerability of poor populations
in developing countries from a new angle. The fact that development policy that directly
targets the poor addresses a highly vulnerable target group is new neither in microfinance nor
in general development policy research. However, the implications of this vulnerability for
development policy are not always obvious and can be overlooked due to the generally good
intentions of development work, as has been the case in microfinance for a long time (see
Chapter 2). This is in line with the empirical research by Bunce et al. (2010) from Tanzania
and Mozambique who note that well-intended development policy can be perceived by
vulnerable target groups to further erode resilience and exacerbate vulnerabilities. The
researchers point to the risk of a policy misfit when policies do not explicitly address the risks
in terms of target group vulnerability and call for more adaptive forms of governance of
development policies.
- 180 -

The explicit consideration of vulnerabilities is important not only for the design of
development policies it has equally important implications for the measurement of the
impact of development policies. Impact measurement is an intensely researched question in
microfinance (Duvendack et al., 2011; Karlan, Goldberg, 2011) as well as in most other fields
of development policy and in the broader literature on aid effectiveness for poverty alleviation
(Collier, Dollar, 2001; Cogneau, Naudet, 2007; Feeny, de Silva, 2012). Although this
dissertation does not measure the impact of microfinance, it raises pertinent questions about
the role of microcredit in the causal chain behind borrower sacrifices and over-indebtedness.
It suggests that future impact measurements of microfinance should not focus only on the
upside potential of microfinance impact but should also explicitly measure the potential
downsides. This is likely to contribute to the general discussion about commercialization in
development and may render potential trade-offs between commercialization and customer
protection apparent.
The need for explicit attention in microfinance impact studies to potential downsides applies
to the selection of variables for impact evaluation: Certain variables such as income or assets
can display both improvements and deteriorations. Other downsides such as stress and
pressure, negative health effects, and social repercussions of bad debt would usually go
undetected unless they were measured explicitly. The need for explicit attention to negative
effects of interventions also applies to the level of detail in impact analysis: Impact
evaluations tend to focus on average developments in the treatment group. However, to
determine whether an intervention that is useful for some recipients but harmful to others,
impact evaluations need to disaggregate the average developments and identify segments in
the groups of the recipients that have been impacted in different ways. Both the variable
selection effect and the disaggregation effect apply to all development policy impact
evaluations that track impact variables on an individual level far beyond microfinance.
This thesis suggests explicitly addressing the potential negative impact of microfinance in
future randomised control trials, the most precise type of impact evaluation in terms of
avoiding selection-biases (Duflo et al., 2007). However, it underlines the case both for
complementing randomised control trials with and basing them on qualitative work and
macro-analysis (Rodrik, 2008) as well as for linking theory (in this dissertation for example
behavioural economics) to empirical research (Duflo et al., 2007; Deaton, 2010). Once again,
the implications for measuring the impact of microfinance are applicable to the individuallevel impact measurement of development policies in general.
- 181 -

Finally, the implications of this dissertation for general development research result from the
use of a subjective approach to understanding the experiences of recipients. This approach is
still unusual and tends to encounter a certain resistance by other researchers who are used to
more objective measurements and therefore are less sensitive to the shortcomings of standard
measurements3 while remaining highly sensitive to the shortcomings of indicators with
subjective elements.
There are two different perspectives on over-indebtedness in microfinance, the risks
management perspective and the customer protection perspective. For the risk management
point of view, over-indebtedness can be defined based on straight forward, objective concepts.
For the customer protection perspective, researchers recognize that the phenomenon is more
complex and has personal elements on the individual level. Nevertheless, many researchers
doubt that a subjective approach to measuring over-indebtedness can provide useful results.
This dissertation demonstrates that a subjective definition of over-indebtedness works in
practice and that it provides new insights that previous research based on objective indicators
was not able to address.
The subjective approach contributes to a growing literature from other fields that consider
subjective indicators more reliable for questions of well-being and happiness than the usual
material proxies, such as income or consumption (Dolan et al., 2008; Oshio, Kobayashi, 2010;
Angner, 2010). It is in line with Sens concept of development as freedom that promotes a
much larger view on the desired impact of development than just improving the material
situation of the poor (Sen, 1999). Rather than just at increasing consumption, the concept of
development as freedom aims to increase the capabilities of people, their freedom to live
according to their own choices.4 Copestake (2008) equally promotes a focus on well-being
during the evaluation of the desired and non-desired outcomes of development practices,
highlighting that well-being is influenced by peoples feelings and perceptions rather than
solely depending on material indicators. Engelbrecht (2009) argues for a new welfare
economics of sustainability based on subjective measures of well-being. On a
macroeconomic cross-country level, the paper makes the case for incorporating the subjective

3 Deaton (2010) make the point that the usual material measurements of poverty and inequality are also much less objective,
reliable, and comparable than we like to think.
4 In this dissertation, not only the definition of over-indebtedness is related to Sens concept of development but also the
resulting findings. Chapter 2 establishes the link of this dissertations findings about over-indebtedness among microfinance
borrowers to the concept of development as freedom, showing that the consequences of over-indebtedness represent
reductions in capabilities and could thus be interpreted as a form of impoverishment.

- 182 -

approach to measuring well-being into the discourse on sustainable development and natural
capital.
The question of subjectivity is related to a trend towards interdisciplinarity, questioning the
role of economics in development research. According to Fine (2002), development research
is prone to a risk of economic imperialism, where the economic discipline has too much
weight compared to other social sciences and forecloses the analytical agenda of development
research. Similarly, education and curriculum development research by Gervedink Nijhuis et
al. (2012) expresses a need for more culturally sensitive approaches in development policy.
Along the same lines, Kanbur (2002) points towards the benefits of considering non-economic
disciplines such as political science, sociology and anthropology in development research. He
argues in favour of interdisciplinarity. Given the difficulties that interdisciplinary research
tends to pose in practice, he hopes that future studies will increasingly serve as examples of
interdisciplinary research on development policy, showing how the influence of multiple
disciplines changes the common thinking based on a single discipline.
This PhD thesis implements his recommendation. It provides multiple examples where, taking
the insights from other disciplines than economics into account, common misconceptions of
the microfinance industry do not hold against local realities. It challenges common thinking
about the relationship between delinquency and over-indebtedness, about the riskiness of
multiple borrowing, about the benefits of tight, high-frequency instalment schedules, about
the effects of the zero-tolerance-policy, and so on. Importantly, the vulnerability revealed by
the Ghana case study in this dissertation is only partly a question of financial vulnerability.
Several of the sacrifices that make up the vulnerability experiences of the microborrowers in
Chapter 3 are culturally specific vulnerabilities. They result from the social interaction
between individuals and from the local, culturally dependent perception of honour, reputation
and personal dependency. The extent of over-indebtedness that this dissertation identifies in
Ghana is therefore dependent on its subjective, interdisciplinary approach and would have
gone over-looked by an objective economics approach based on risk management indicators.
Interdisciplinarity tends to be complicated, take time and reduce the comparability of results.
Nevertheless, based on its interdisciplinary approach, this PhD provides a new answer to the
measurement of over-indebtedness from the customer protection perspective and goes further
into the analysis of the causes, consequences and solutions than would have been possible
with a purely economic approach. By highlighting the difference that interdisciplinarity

- 183 -

makes to the analysis of over-indebtedness, it aspires to reinforce the point for


interdisciplinary approaches in development research.

e) Financial services consumer protection in industrialised countries


Last but not least, this dissertation has worked with the consumer finance literature from
industrialised countries. This raises the question of what it can contribute back to financial
services consumer protection in the developed country context. The following section will
discuss the transferability of the dissertations main findings to the protection of nonmicrofinance borrowers against over-indebtedness.
The most important element in that regard is the use of the sacrifice-based definition of overindebtedness. If microfinance customer protection is about avoiding unacceptable sacrifices of
borrowers, then financial services customer protection in industrialised countries might also
benefit from a perspective on over-indebtedness through the lens of sacrifices. After all, it is
likely to pursue the same goal of preventing consumers from suffering too much from their
use of loan products. And indeed, this is the case and has been documented by several authors
in the policy-oriented literature on over-indebtedness: They consider a borrowers perception
of a high debt burden at least a partial indicator of over-indebtedness (Lea et al., 1993;
Bridges, Disney, 2005; BERR, 2007; Betti et al., 2007).
At the same time, there is a substantial difference between the developing country context and
the developed country context. Developed countries usually have reliable judicial systems that
include personal insolvency and bankruptcy regulations. Borrowers who can prove that they
do not have the resources to repay their debt can be exempt from their repayment obligations.
As personal insolvency regulations are based on minimum existence levels (Haas, 2006;
European Commission, 2008), the existence of such an option for personal insolvency
changes the over-indebtedness condition significantly: there is an inbuilt limit to borrower
sacrifices. Minimum existence levels define the income and assets that any borrower is
allowed to keep, no matter how indebted he may be. No lender has the right to take revenues
or assets below this level from a debtor. No lender can oblige a borrower to continue repaying
his loans if his resources do not exceed the minimum existence level.
The developed nations personal insolvency regulations, which are based on minimum
existence levels, therefore represent a customer protection tool that implicitly includes the
avoidance of unacceptable sacrifices. This reduces the need for a sacrifice-based definition of
over-indebtedness significantly. Using a default based on personal bankruptcy filings as an
- 184 -

indicator of over-indebtedness in a developed country is much closer to a customer protection


perspective than a default indicator from the risk management perspective in the microfinance
context where a personal insolvency option is usually not accessible to borrowers and no
protection of a locally defined minimum existence level exists.
At the same time, there are still many developed country borrowers who are struggling
significantly with their debt (Kempson, 2002; BERR, 2007). The above-mentioned inclusion
of a perceived debt burden in definitions of over-indebtedness in developed countries is a
reaction to this concern and an attempt to strengthen the customer protection perspective in
the over-indebtedness discourse in developed country consumer finance. It recognises that a
materially defined minimum existence level in terms of income and assets cannot prevent all
serious sacrifices. Additionally, the stigma of personal insolvency, the difficulties related to
the judicial procedure, and the loss of access to credit that can follow after bankruptcy, tend to
incite developed country borrowers to accept significant sacrifices rather than having to file
for personal bankruptcy. There may therefore also be room to use the sacrifice-based
approach to measuring over-indebtedness in developed countries for purposes of research,
thus enhancing our understanding of the situation of borrowers with personal debt.
Such a transfer of this dissertations methodology would of course require adaptations to the
different context. As much as the realities of borrowers in developed countries are different
from those of developing country borrowers, the indicator for subjective over-indebtedness
may need to be different, for example with regards to the list of sacrifices. Similarly, the
subjective thresholds for acceptability are likely to vary substantially, in line with the use of
relative rather than absolute measures of poverty. Nevertheless, the normative core of this
dissertations approach, the avoidance of structural, unacceptable sacrificing on behalf of
indebted borrowers, is universal in its use.
This leads us to the question of the consequences of over-indebtedness. This dissertation
provides an unprecedented analysis of the effects that over-indebtedness can have on
microborrowers. It broadens our understanding of the full range of consequences of overindebtedness, including not only material effects but also psychological and sociological
consequences. In part, these have been transferred to the microfinance context from existing
developed country consumer finance research (e.g., Drentea, Lavrakas, 2000; Viaud, RolandLvy, 2000; Bridges, Disney, 2005; Kim et al., 2006; Lyons, Fisher, 2006). When they are
drawn from microfinance specific research, they may to a certain extent apply to developed
country environments as well. In any case, the claim of Chapter 2 that researchers and policy
- 185 -

makers should take the full spectrum of potential consequences of over-indebtedness into
account applies to consumer finance in industrialised nations as much as it does to
microfinance in developing countries.
Understanding the full breath of consequences of over-indebtedness also feeds back into the
definition of a minimum existence level in developed country insolvency regulations.
Recognising certain psychological and social risks of over-indebtedness (and poverty in
general), the concept of a minimum existence level in developed countries already reaches
beyond the mere guarantee of survival. Minimum existence levels are not only supposed to
ensure that borrowers will keep the resources to eat and live in decent accommodations, but
they tend to be sufficiently generous to allow for a minimum level of social participation. The
analysis of the consequences of over-indebtedness in this PhD thesis can contribute to the
definition of what a minimum existence level should comprise, pointing out the full range of
problems that borrowers experience, and highlighting the importance of financial and social
participation. Also, the data from Ghana provides local information on the subjective
evaluation of hardship in terms of acceptability. Finally, the analysis of consequences can
identify the potential consequences of over-indebtedness that no minimum existence level can
prevent.
Finally, this dissertation has developed a framework of the causes of over-indebtedness. This
framework is built in large part on consumer finance research from developed countries (e.g.,
Webley, Nyhus, 2001; Soman, Cheema, 2002; Brown et al., 2005; Christen, Morgan, 2005;
Stone, Maury, 2006; Lusardi, Tufano, 2009). It is therefore likely to apply to this context. The
framework of the causes of over-indebtedness unites previous research into a comprehensive
tool for analysis that was not previously available, neither in microfinance nor in consumer
finance in industrialised countries. It can guide future research on over-indebtedness in
developed countries as much as it can in microfinance, ensuring a comprehensive perspective
on the causes of over-indebtedness and the complex interaction of influence factors. It can
enrich the debate about whose responsibility it is to prevent over-indebtedness. And it can
broaden the solution space for measures to prevent and alleviate over-indebtedness - also in
the context of developed countries.

- 186 -

3. Concluding remarks: Limitations and further research


In its summary of the dissertations findings, this conclusion has already noted a number of
limitations of the present research and has indicated room for further research. Another
important limitation is that the analysis focuses on the client perspective only. While this
represents one of the main strengths of the research, it implies that this dissertation does not
analyse the functioning of the supply side in detail. This dissertation is therefore limited in the
recommendations it can provide for MFIs. The schism between risk management and
customer protection that this dissertation discusses has implications for MFIs on every level,
from lending procedures in the field to governance processes. However, an in-depth analysis
of the supply side is required to determine where the main levers for change in an MFI lie
with regards to over-indebtedness. Such an analysis would have to analyse the credit process
and especially the cash flow analysis of MFIs, the organisational structure and resulting
clientloan officer relationships, and the incentive structures for loan officers and their
alignment with MFIs missions. It should examine the missions of different institutions, the
differences between NGOs and licensed financial intermediaries, and how over-indebtedness
relates to MFI profits (both via direct financial impact and via long-term effects on customer
satisfaction) and to potential impact related elements of the missions of MFIs. A supply side
analysis would have to address how all of these factors play together and how they depend on
the management and governance structures of lending institutions, in interaction with the
influence of investors and donors.
In addition, Chapter 4 has analysed factors of over-indebtedness that are related to overindebtedness at the level of microborrowers, without covering all the potential psychological
and sociological influence factors identified in the causal framework of Chapter 1. Studies
that wanted to address such factors would have to develop valid constructs and experiments to
measure the respective psychological and sociological factors such as hyperbolic discounting
or conspicuous consumption in the respective local contexts and test to what extent such
variables are related to over-indebtedness in microfinance from the perspective of customer
protection.
Also, Chapter 4 focused on the borrower level by controlling for the influence of lenders only
as aggregates. Future research should extend the analysis to the institutional characteristics of
lenders. Instead of controlling for the general influences of lenders with MFI dummies, future
studies should test potential causes of over-indebtedness on the institutional level for their
relationship to over-indebtedness.
- 187 -

Furthermore, the research on the extent of over-indebtedness among microborrowers and on


the sacrifices they experience should be replicated in other geographies. The same is true for
the analysis of easier-to-measure indicators of over-indebtedness and for that of borrowerlevel factors related to over-indebtedness. The research in Ghana is limited in its external
validity, as the extent of over-indebtedness, the quality of risk management indicators as
predictors of over-indebtedness, and the factors related to over-indebtedness in microfinance
may differ between countries, historical contexts and cultures. This is especially true for overindebtedness defined according to borrower sacrifices and based on the subjective value
judgements of respondents. Not all respondents may draw the line of acceptability at the same
level of suffering and the borrowers views about their own well-being and sacrificing may be
inconsistent and biased. Suffering is a subjective phenomenon that, as it differs between
contexts and between individuals, may effectively require different responses in different
contexts. Nevertheless, this subjectivity and context dependency limits the universal validity
of findings with regards to the predictors of over-indebtedness or of factors related to overindebtedness.
Indeed, in terms of replication, there are already a number of on-going studies in microfinance
that replicate parts of this PhD research in other microfinance markets such as Kosovo,
Mexico, Kyrgyzstan and Tajikistan. These studies use a similar definition of overindebtedness, work with a similar methodology, and at least one study empirically tests the
framework of over-indebtedness causes. In addition, the Appendix contains two practitioneroriented publications that build on the findings of this PhD to communicate its key insights
about the definition, extent and causes of over-indebtedness to decision makers in the
microfinance sector.
Finally, this dissertation analyses microfinance as a tool of international development by
focussing on a narrowly defined set of stakeholders, most importantly the borrowers of MFIs
and MFIs as lending institutions. It sheds light on an important phenomenon that influences
these stakeholders and their credit relationship. However, according to the categorization of
Copestake (in press) it represents policy-specific or instrument-oriented research, rather than
broad research that would consider microfinance an endogenous component of a complex
institutional environment under uncertainty. The dissertation recognizes the partly qualitative
and subjective character of the phenomenon of over-indebtedness and acknowledges that
over-indebtedness research cannot fully abstract from history and context as is often the case
in instrument-oriented microfinance research. It takes certain interactions of microfinance
with the social environment of microborrowers into account. But it nevertheless abstracts
- 188 -

from the embeddedness of microfinance into institutional processes and historical


developments. More research is needed to understand for example how local systems of
power relationships based on indebtedness interact with the development of the microfinance
industry and how they influence the risk of microfinance over-indebtedness. It should
examine the interaction of microfinance over-indebtedness with inequality, mutual exchange
relationships in communities, and the economic and political aspirations of poor populations.
Future research should more explicitly take into account the meso and macro-level
implications of microfinance over-indebtedness for non-borrowers, for the informal economy,
and for society as a collective, including its shaping of mental models and institutional
developments.
To conclude, if the microfinance industry wants to avoid over-indebtedness, ensure a
minimum level of customer protection, and live up to its claims of positively impacting the
lives of the poor, it must learn to better listen to the voice of its borrowers. This dissertation
hopes to contribute to this goal and to give customers a voice in revealing their struggles and
sacrifices. It does not deny the positive impact that microfinance may have for so many;
instead, it aims to promote a future development of the microfinance industry that ensures the
highest standards of customer protection and reduces the potential dangers that credit may
imply to a minimum. This research hopes to contribute to the development of a client-focused
understanding of over-indebtedness in microfinance and to help the industry to identify
indicators that allow MFIs and regulators to track the risks of over-indebtedness for
microfinance customers. With its analysis of factors related to over-indebtedness, it hopes to
contribute to the development of measures against over-indebtedness and thus to an
improvement in the borrowing experiences of microfinance customers in the long run.

- 189 -

References
Agier, I. and A. Szafarz (2010) 'Microfinance and Gender: Is There a Glass Ceiling in Loan
Size?'. Universit Libre de Bruxelles CEB Working Paper No. 10-047, Brussels.
Agier, I. and A. Szafarz (2011) 'Subjectivity in Credit Allocation to Micro-Entrepreneurs:
Evidence From Brazil'. Universit Libre de Bruxelles CEB Working Paper No. 11/016,
Brussels.
Agier, I., I. Gurin and A. Szafarz (2012) 'Child gender and parental borrowing: Evidence
from India'. Economics Letters, 115, 363365.
Ahlin, C., J. Lin and M. Maio (2011) 'Where does microfinance flourish? Microfinance
institution performance in macroeconomic context'. Journal of Development Economics,
95(2), 105120.
Altman, M. (2012) 'Implications of behavioural economics for financial literacy and public
policy'. The Journal of Socio-Economics, 41(5), 677690.
Angner, E. (2010) 'Subjective well-being: Financial Crisis: Capabilities and Happiness'.
Journal of Socio-Economics, 39(3), 361368.
Armendriz, B. and J. Morduch (2010) 'The economics of microfinance (2nd ed.), Cambridge,
Mass: MIT Press.
Behr, P. and S. Sonnekalb (2012) 'The effect of information sharing between lenders on
access to credit, cost of credit, and loan performance Evidence from a credit registry
introduction: International Corporate Finance Governance Conference'. Journal of Banking
& Finance, 36(11), 30173032.
BERR (2007) 'Tackling over-indebtedness: Annual Report 2007'. Department for Business
Enterprise & Regulatory Reform, London.
Betti, G., N. Dourmashkin, M.C. Rossi and Y.P. Yin (2007) 'Consumer over-indebtedness in
the EU: measurement and characteristics'. Journal of Economic Studies, 34(2), 136156.
Bridges, S. and R. Disney (2005) 'Debt and depression'. Centre for Finance and Credit
Markets, School of Economics, Nottingham Working Paper No. 06/02.
Brown, M. and C. Zehnder (2010) 'The emergence of information sharing in credit markets'.
Journal of Financial Intermediation, 19(2), 255278.

- 190 -

Brown, M., T. Jappelli and M. Pagano (2009) 'Information sharing and credit: Firm-level
evidence from transition countries'. Journal of Financial Intermediation, 18(2), 151172.
Brown, S., G. Garino, K. Taylor and S. Wheatley Price (2005) 'Debt and Financial
Expectations: An Individual- and Household-Level Analysis'. Economic Inquiry, 43(1),
100120.
Bucher-Koenen, T. and M. Ziegelmeyer (2011) 'Who lost the most? Financial literacy,
cognitive abilities, and the financial crisis'. European Central Bank Working Paper Series
No. 1299.
Bunce, M., K. Brown and S. Rosendo (2010) 'Policy misfits, climate change and cross-scale
vulnerability in coastal Africa: how development projects undermine resilience'.
Environmental Science & Policy, 13(6), 485497.
Chomsisengphet, S. and R. Elul (2006) 'Bankruptcy exemptions, credit history, and the
mortgage market'. Journal of Urban Economics, 59(1), 171188.
Christen, M. and R.M. Morgan (2005) 'Keeping Up With the Joneses: Analyzing the Effect of
Income Inequality on Consumer Borrowing'. Quantitative Marketing and Economics, 3(2),
145173.
Christen, R., K. Lauer, T. Lyman and R. Rosenberg (2012) 'A Guide to Regulation and
Supervision of Microfinance: Consensus Guidelines'. CGAP, Washington, D.C.
Cogneau, D. and J.-D. Naudet (2007) 'Who Deserves Aid? Equality of Opportunity,
International Aid, and Poverty Reduction'. World Development, 35(1), 104120.
Collier, P. and D. Dollar (2001) 'Can the World Cut Poverty in Half? How Policy Reform and
Effective Aid Can Meet International Development Goals'. World Development, 29(11),
17871802.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How
the World's Poor Live on $2 a Day, Princeton, NJ: Princeton Univ. Press.
Copestake, J. (2008) 'Wellbeing and international development: whats new?'. Journal of
International Development, 29(4), 577597.
de Janvry, A., C. McIntosh and E. Sadoulet (2010) 'The supply- and demand-side impacts of
credit market information'. Journal of Development Economics, 93(2), 173188.
Deaton, A. (2010) 'Understanding the Mechanisms of Economic Development'. Journal of
Economic Perspectives, 24(3), 316.
- 191 -

DeLaune, L. D., J.S. Rakow and K.C. Rakow (2010) 'Teaching financial literacy in a cocurricular service-learning model'. Journal of Accounting Education, 28(2), 103113.
Dolan, P., T. Peasgood and M. White (2008) 'Do we really know what makes us happy? A
review of the economic literature on the factors associated with subjective well-being'.
Journal of Economic Psychology, 29(1), 94122.
Drentea, P. and P.J. Lavrakas (2000) 'Over the Limit: The Association Among Health, Race,
and Debt'. Social Science & Medicine, 50(4), 517529.
Drexler, A., G. Fischer and A.S. Schoar (2010) 'Keeping it Simple: Financial Literacy and
Rules of Thumb'. Centre for Economic Policy Research CEPR Discussion Papers No.
7994, London.
Duflo, E., R. Glennerster and M. Kremer (2007) 'Chapter 61 Using Randomization in
Development Economics Research: A Toolkit' In T. Paul Schultz and John A. Strauss
(Ed.), Handbook of Development Economics (pp. 38953962). Elsevier.
Duvendack, M., R. Palmer-Jones, J. Copestake, L. Hooper, Y. Loke and N. Rao (2011) 'What
is the evidence of the impact of microfinance on the well-being of poor people?'. EPPICentre, Social Science Research Unit, Institute of Education, University of London,
London.
Engelbrecht, H.-J. (2009) 'Natural capital, subjective well-being, and the new welfare
economics of sustainability: Some evidence from cross-country regressions: Special
Section: Analyzing the global human appropriation of net primary production - processes,
trajectories, implications'. Ecological Economics, 69(2), 380388.
Engle, P. (1993) 'Influences of mothers and fathers income on childrens nutritional status in
Guatemala'. Social Science and Medicine, 37, 13031312.
European Commission (2008) 'Towards a Common Operational European Definition of OverIndebtendess'. OEE Studies, Brussels.
Feeny, S. and A. de Silva (2012) 'Measuring absorptive capacity constraints to foreign aid'.
Economic Modelling, 29(3), 725733.
Fine, B. (2002) 'Economics Imperialism and the New Development Economics as Kuhnian
Paradigm Shift?'. World Development, 30(12), 20572070.

- 192 -

Gardeva, A. and E. Rhyne (2011) 'Opportunities and Obstacles to Financial Inclusion: Survey
Report'. Center for Financial Inclusion at Accion Center for Financial Inclusion Publication
No. 12, Washington, D.C.
Garikipati, S. (2008) 'The Impact of Lending to Women on Household Vulnerability and
Women's Empowerment: Evidence from India: Special Section: Social Movements and the
Dynamics of Rural Development in Latin America (pp. 2874-2952)'. World Development,
36(12), 26202642.
Gathergood, J. (2012) 'Self-control, financial literacy and consumer over-indebtedness'.
Journal of Economic Psychology, 33(3), 590602.
Gervedink Nijhuis, C. J., J.M. Voogt and J.M. Pieters (2012) 'The cultural complexity of
international collaboration: Conditions for sustainable curriculum development in Ghana'.
International Journal of Intercultural Relations, 36(5), 647658.
Giannetti, C. and N. Jentzsch 'Credit Reporting, Financial Intermediation and Identification
Systems: International Evidence'. Journal of International Money and Finance, (0).
Goetz, A. M. and R. Sen Gupta (1996) 'Who takes the credit? Gender, power, and control
over loan use in rural credit programs in Bangladesh'. World Development, 24(1), 4563.
Gurin, I. (2011) 'The Gender of Finance and Lessons for Microfinance' In B. Armendriz &
M. Labie (Eds.), The Handbook of Microfinance (pp. 589612). London, Singapore: World
Scientific Publishing.
Gurin, I. (in press) 'L'ducation financire ou comment apprendre aux pauvre bien
consommer' In I. Gurin & Selim (Eds.), Comment et quoi dpenser son argent?. Paris:
L'Harmattan.
Gurin, I., S. Morvant-Roux and J.-M. Servet (2011a) 'Understanding the Diversity and
Complexity of Demand for Microfinance Services: Lessons from Informal Finance' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 101122). London,
Singapore: World Scientific Publishing.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2011b) 'The social meaning of
over-indebtedness and creditworthiness in the context of poor rural South India households
(Tamil Nadu)'. RUME Working Paper No. 2011-1, Paris.

- 193 -

Gurin, I., M. Roesch, O. Venkatasubramanian and B. D'Espallier (2012) 'Credit for whom
and for what? The diversity of borrowing sources and borrowing uses in south-India'.
Journal of International Development, (24 Issue Supplement S1), 122S137.
Haas, O. J. (2006) 'Overindebtedness in Germany'. Social Finance Programme, International
Labour Organisation Working Paper No. 44, Geneva.
Hashemi, S. M., S.R. Schuler and A.P. Riley (1996) 'Rural credit programs and womens
empowerment in Bangladesh'. World Development, 24(4), 635653.
Jianakopolos, N. A. and A. Bernasek (1998) 'Are women more risk averse?'. Economic
Inquiry, 36(4), 620630.
Kabeer, N. (2001) 'Conflicts Over Credit: Re-Evaluating the Empowerment Potential of
Loans to Women in Rural Bangladesh'. World Development, 29(1), 6384.
Kanbur, R. (2002) 'Economics, Social Science and Development'. World Development, 30(3),
477486.
Karlan, D. and N. Goldberg (2011) 'Microfinance Evaluation Strategies: Notes on
Methodology and Findings' In B. Armendriz & M. Labie (Eds.), The Handbook of
Microfinance (pp. 1758). London, Singapore: World Scientific Publishing.
Karlan, D. and M. Valdivia (2011) 'Teaching Entrepreneurship: Impact of Business Training
on Microfinance Clients and Institutions'. Review of Economics and Statistics, 93(2), 510
527.
Kempson, E. (2002) 'Over-indebtedness in Britain: A report to the Department of Trade and
Industry'. Personal Finance Research Centre, University of Bristol, Bristol.
Khandker, S. R. (2003) 'Microfinance and poverty: Evidence using panel data from
Bangladesh'. World Bank World Bank Policy Research Working Paper No. 2945,
Washington, D.C.
Khandker, S. R., B. Khalily and Z. Kahn (1995) 'Grameen Bank: Performance and
sustainability'. World Bank Discussion Paper No. 306, Washington, D.C.
Kim, J., B. Sorhaindo and E.T. Garman (2006) 'Relationship between Financial Stress and
Workplace Absenteeism of Credit Counseling Clients'. Journal of Family and Economic
Issues, 27(3), 458478.
Lea, S. E. G., P. Webley and R.M. Levine (1993) 'The Economic Psychology of Consumer
Debt'. Journal of Economic Psychology, 14(1), 85119.
- 194 -

Lusardi, A. and O.S. Mitchell (2007) 'Baby Boomer retirement security: The roles of
planning, financial literacy, and housing wealth'. Journal of Monetary Economics, 54(1),
5224.
Lusardi, A. and P. Tufano (2009) 'Debt Literacy, Financial Experiences, and
Overindebtedness'. National Bureau of Economic Research NBER Working Paper Series
No. 14808, Cambridge, MA.
Lyman, T., T. Lythgoe, M. Miller, X. Reille and S. Sankaranarayan (2011) 'Credit Reporting
at the Base of the Pyramid: Key Issues and Success Factors'. CGAP and IFC FORUM No.
1, Washington, D.C.
Lyons, A. C. and J. Fisher (2006) 'Gender Differences in Debt Repayment Problems after
Divorce'. Journal of Consumer Affairs, 40(2), 324346.
Mayoux, L. (2001) 'Tackling the down side: Social capital, women's empowerment and
micro-finance in Cameroon'. Development and Change, 32(3), 421450.
Mayoux, L. (2011) 'Taking Gender Seriously: Towards a Gender Justice Protocol for
Financial Services' In B. Armendriz & M. Labie (Eds.), The Handbook of Microfinance
(pp. 613642). London, Singapore: World Scientific Publishing.
McIntosh, C., E. Sadoulet, S. Buck and T. Rosada (in press) 'Reputation in a public goods
game: Taking the design of credit bureaus to the lab'. Journal of Economic Behavior &
Organization,
Meier, S. and C.D. Sprenger (in press) 'Discounting financial literacy: Time preferences and
participation in financial education programs'. Journal of Economic Behavior &
Organization
Morvant-Roux, S. (2009a) 'Accs au microcrdit et continuit des dynamiques d'endettement
au Mexique: Combiner l'anthropologie conomique avec l'conomtrie [Access to
microcredit and continuity of indebtedness dynamics in Mexico: Combining economic
anthropology with econometrics]'. Revue Tiers Monde, 196, 109130.
Morvant-Roux, S. (2009b) 'L'Anthropo-conomie: un dtour indispensable pour comprendre
l'appropriation de la microfinance [Anthropo-economy: An unavoidable detour to
understand the appropriation of microfinance]'. Revue Tiers Monde, (197), 109130.

- 195 -

Olsen, R. A. and C.M. Cox (2001) 'The Inuence of Gender on the Perception and Response to
Investment Risk: The Case of Professional Investors'. Journal of Psychology and Financial
Markets, 2(1), 2936.
Oshio, T. and M. Kobayashi (2010) 'Income inequality, perceived happiness, and self-rated
health: Evidence from nationwide surveys in Japan'. Social Science & Medicine, 70(9),
13581366.
Pitt, M. M., S.R. Khandker and J. Cartwright (2006) 'Empowering Women with Micro
Finance: Evidence from Bangladesh'. Economic Development and Cultural Change, 54(4),
791831.
Rodrik, D. (2008) 'The New Development Economics: We Shall Experiment, but How Shall
We Learn?'. HKS Working Paper No. RWP08-055.
Rutherford, S. (2000) 'The Poor and Their Money, An Essay about Financial Services for
Poor People, Oxford: Oxford University Press.
Sen, A. (1999) 'Development as Freedom: Oxford University Press.
Soman, D. and A. Cheema (2002) 'The Effect of Credit on Spending Decisions: The Role of
the Credit Limit and Credibility'. Marketing Science, 21(1), 3253.
Stone, B. and R.V. Maury (2006) 'Indicators of personal financial debt using a multidisciplinary behavioral model'. Journal of Economic Psychology, 27(4), 543556.
Thomas, D. (1990) 'Intrahousehold allocation: An inferential approach'. Journal of Human
Resources, 25(4), 635664.
Tiwari, A., A. Khandelwal and M. Ramji (2008) 'How Do Microfinance Clients Understand
their Loans?'. Institute for Financial Management and Research, Centre for Micro Finance
Working Paper Series No. 25.
Viaud, J. and C. Roland-Lvy (2000) 'A positional and representational analysis of
consumption. Households when facing debt and credit'. Journal of Economic Psychology,
21(4), 411432.
Webley, P. and E.K. Nyhus (2001) 'Life-Cycle and Disproportionate Routes into Problem
Debt'. British Journal of Psychology, 92, 423446.
Willis, L. E. (2008) 'Against financial literacy education'. Iowa Law Review, 94, 197285.

- 196 -

Full bibilography
Agier, I. and A. Szafarz (2010) 'Microfinance and Gender: Is There a Glass Ceiling in Loan
Size?'. Universit Libre de Bruxelles CEB Working Paper No. 10-047, Brussels.
Agier, I. and A. Szafarz (2011) 'Subjectivity in Credit Allocation to Micro-Entrepreneurs:
Evidence From Brazil'. Universit Libre de Bruxelles CEB Working Paper No. 11/016,
Brussels.
Agier, I., I. Gurin and A. Szafarz (2012) 'Child gender and parental borrowing: Evidence
from India'. Economics Letters, 115, 363365.
Ahlin, C., J. Lin and M. Maio (2011) 'Where does microfinance flourish? Microfinance
institution performance in macroeconomic context'. Journal of Development Economics,
95(2), 105120.
Akerlof, G. A. (1991) 'Procrastination and Obedience'. The American Economic Review,
81(2), 119.
Altman, M. (2012) 'Implications of behavioural economics for financial literacy and public
policy'. The Journal of Socio-Economics, 41(5), 677690.
Angner, E. (2010) 'Subjective well-being: Financial Crisis: Capabilities and Happiness'.
Journal of Socio-Economics, 39(3), 361368.
Armendriz, B. and J. Morduch (2010) 'The economics of microfinance (2nd ed.), Cambridge,
Mass: MIT Press.
Armendriz, B. and A. Szafarz (2011) 'On Mission Drift In Microfinance Institutions' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 341366). London,
Singapore: World Scientific Publishing.
Armendriz, B.; M. Labie (Eds.) (2011) 'The Handbook of Microfinance, London, Singapore:
World Scientific Publishing.
Ashraf, N., D. Karlan and W. Yin (2006) 'Tying Odysseus to the Mast: Evidence from a
Commitment Savings Product in the Philippines'. Quarterly Journal of Economics, 121(2),
635667.
Atkinson, A., S. McKay, E. Kempson and S. Collard (2006) 'Levels of Financial Capability in
the UK: Results of a baseline survey'. Financial Services Authority Consumer Research
No. 47, London.
- 197 -

Backhaus, K., B. Erichson, W. Plinke and R. Weiber (2011) 'Multivariate Analysemethoden,


Eine anwendungsorientierte Einfhrung (13th ed.), Berlin: Springer.
Bagwell, D. C. (2001) 'Hierarchical regression analysis of work outcomes with personal and
financial factors'. Writings of the Western Region Home Management Family Economics
Educators, 4554.
Banerjee, A. and S. Mullainathan (2010) 'The shape of temptation: Implications for the
economic lives of the poor'. NBER Working Paper Series No. 15973.
Banerjee, A. V. and E. Duflo (2007) 'The Economic Lives of the Poor'. Journal of Economic
Perspectives, 21(1), 141167.
Banerjee, A. V., E. Duflo, R. Glennerster and C. Kinnan (2009) 'The miracle of microfinance?
Evidence from a randomized evaluation'. MIT Poverty Action Lab, Cambridge, MA.
Bauer, M., J. Chytilov and J. Morduch (2009) 'Behavioral Foundations of Microcredit:
Experimental and Survey Evidence from Rural India'. Financial Access Initiative, New
York.
Behr, P. and S. Sonnekalb (2012) 'The effect of information sharing between lenders on
access to credit, cost of credit, and loan performance Evidence from a credit registry
introduction: International Corporate Finance Governance Conference'. Journal of Banking
& Finance, 36(11), 30173032.
Benno, F. (2008) 'Proposal for a Contingency Debt Sustainability Framework: Special
Section: Social Movements and the Dynamics of Rural Development in Latin America (pp.
2874-2952)'. World Development, 36(12), 25472565.
BERR (2007) 'Tackling over-indebtedness: Annual Report 2007'. Department for Business
Enterprise & Regulatory Reform, London.
Bertrand, M., D. Karlan, S. Mullainathan, E. Shafir and J. Zinman (2010) 'What's Advertising
Content Worth? Evidence from a Consumer Credit Marketing Field Experiment'.
Quarterly Journal of Economics, 125(1), 263305.
Bertrand, M., S. Mullainathan and E. Shafir (2004) 'A Behavioral-Economics View of
Poverty'. The American Economic Review, 94(2), 419423.
Besley, T. and S. Coate (1995) 'Group lending, repayment incentives and social collateral'.
Journal of Development Economics, 46(1), 118.

- 198 -

Betti, G., N. Dourmashkin, M.C. Rossi and Y.P. Yin (2007) 'Consumer over-indebtedness in
the EU: measurement and characteristics'. Journal of Economic Studies, 34(2), 136156.
Bouquet, E., B. Wampfler, E. Ralison and M. Roesch (2007) 'Trajectoires de crdit et
vulnrabilit des mnages ruraux: Le cas des Cecam de Madagascar [Credit trajectories
and vulnerability of rural households: the case of Cecam in Madagascar]'. Autrepart, 44,
157172.
Brett, J. A. (2006) '"We Sacrifice and Eat Less": The Structural Complexities of Microfinance
Participation'. Human Organization, 65(1), 8.
Bridges, S. and R. Disney (2004) 'Use of Credit and Arrears on Debt among Low-Income
Families in the United Kingdom'. Fiscal Studies, 25(1), 125.
Bridges, S. and R. Disney (2005) 'Debt and depression'. Centre for Finance and Credit
Markets, School of Economics, Nottingham Working Paper No. 06/02.
Brown, M. and C. Zehnder (2010) 'The emergence of information sharing in credit markets'.
Journal of Financial Intermediation, 19(2), 255278.
Brown, M., T. Jappelli and M. Pagano (2009) 'Information sharing and credit: Firm-level
evidence from transition countries'. Journal of Financial Intermediation, 18(2), 151172.
Brown, S., G. Garino, K. Taylor and S. Wheatley Price (2005a) 'Debt and Financial
Expectations: An Individual- and Household-Level Analysis'. Economic Inquiry, 43(1),
100120.
Brown, S., K. Taylor and S. Wheatley Price (2005b) 'Debt and distress: Evaluating the
psychological cost of credit'. Journal of Economic Psychology, 26(5), 642663.
Brown, T. M. (1952) 'Habit Persistence and Lags in Consumer Behaviour'. Econometrica,
20(3), 355371.
Bucher-Koenen, T. and M. Ziegelmeyer (2011) 'Who lost the most? Financial literacy,
cognitive abilities, and the financial crisis'. European Central Bank Working Paper Series
No. 1299.
Bunce, M., K. Brown and S. Rosendo (2010) 'Policy misfits, climate change and cross-scale
vulnerability in coastal Africa: how development projects undermine resilience'.
Environmental Science & Policy, 13(6), 485497.
Burton, D. (2008) 'Credit and Consumer Society, Oxon: Routledge.

- 199 -

Canner, G. B. and C.A. Luckett (1991) 'Payment of household debts'. Federal Reserve
Bulletin, 77(4), 218.
Chaia, A., A. Dalal, T. Goland, M.J. Gonzalez, J. Morduch and R. Schiff (2009) 'Half the
World is Unbanked'. Financial Access Initiative Framing Note, New York.
Chaudhury, I. A. and I. Matin (2002) 'Dimensions and dynamics of microfinance membership
overlap: A micro study from Bangladesh'. Small Enterprise Development, 13(2), 4655.
Chen, G. and S.R.X. Rasmussen (2010) 'Growth and Vulnerability in Microfinance'. CGAP
Focus Note No. 61, Washington DC.
Chomsisengphet, S. and R. Elul (2006) 'Bankruptcy exemptions, credit history, and the
mortgage market'. Journal of Urban Economics, 59(1), 171188.
Christen, M. and R.M. Morgan (2005) 'Keeping Up With the Joneses: Analyzing the Effect of
Income Inequality on Consumer Borrowing'. Quantitative Marketing and Economics, 3(2),
145173.
Christen, R., K. Lauer, T. Lyman and R. Rosenberg (2012) 'A Guide to Regulation and
Supervision of Microfinance: Consensus Guidelines'. CGAP, Washington, D.C.
Cogneau, D. and J.-D. Naudet (2007) 'Who Deserves Aid? Equality of Opportunity,
International Aid, and Poverty Reduction'. World Development, 35(1), 104120.
Collier, P. and D. Dollar (2001) 'Can the World Cut Poverty in Half? How Policy Reform and
Effective Aid Can Meet International Development Goals'. World Development, 29(11),
17871802.
Collins, D. (2008) 'Debt and household finance: Evidence from the Financial Diaries'.
Development Southern Africa, 25(4), 469479.
Collins, D., J. Morduch, S. Rutherford and O. Ruthven (2009) 'Portfolios of the Poor, How
the World's Poor Live on $2 a Day, Princeton, NJ: Princeton Univ. Press.
Copestake, J. (2002) 'Unfinished business: The need for more effective microfinance exit
monitoring'. Journal of Microfinance, 4(2), 130.
Copestake, J. (2007) 'Mainstreaming microfinance: social performance management or
mission drift?'. World Development, 35(10), 17211738.
Copestake, J. (2008) 'Wellbeing and international development: whats new?'. Journal of
International Development, 29(4), 577597.
- 200 -

Copestake, J. (2011) 'Well-Being in Development: Comparing Global Designs with Local


Views in Peru'. European Journal of Development Research, 23(1), 94110.
Copestake, J. (in press) 'Research into rural microfinance and development: the case for
breadth illustrated by an unsystematic review of the India literature'. Oxford Development
Studies,
Corbett, J. (1988) 'Famine and household coping strategies'. World Development, 16(9),
10991112.
de Janvry, A., C. McIntosh and E. Sadoulet (2010) 'The supply- and demand-side impacts of
credit market information'. Journal of Development Economics, 93(2), 173188.
de Mel, S., D.J. McKenzie and C. Woodruff (2008) 'Returns to Capital in Microenterprises:
Evidence from a Field Experiment'. Quarterly Journal of Economics, 123(4), 13291372.
Deaton, A. (2010) 'Understanding the Mechanisms of Economic Development'. Journal of
Economic Perspectives, 24(3), 316.
DeLaune, L. D., J.S. Rakow and K.C. Rakow (2010) 'Teaching financial literacy in a cocurricular service-learning model'. Journal of Accounting Education, 28(2), 103113.
Del-Ro, A. and G. Young (2005) 'The impact of unsecured debt on financial distress among
British households'. Bank of England Working Paper No. 262, London.
D'Espallier, B., I. Gurin and R. Mersland (2011) 'Women and Repayment in Microfinance: A
Global Analysis'. World Development, 39(5), 758772.
DeVaney, P. L. (2006) 'Bringing Pro-Consumer Ideals to the Client, A Consumer Protection
Guide for Financial Institutions Serving the Poor'. Monograph Series: Vol. 14,
Washington, DC: ACCION Publications.
DeVaney, S. A. and R.H. Lytton (1995) 'Household Insolvency: A Review of Household Debt
Repayment, Delinquency, and Bankruptcy'. Financial Services Review, 4(2), 137.
Dichter, T. (2007) 'Can microcredit make an already slippery slope more slippier?: Some
lessons from the social meaning of debt' In T. Dichter & M. Harper (Eds.), What's wrong
with Micofinance? (pp. 918). Warwickshire: Intermediate Technology Publications.
Disney, R., S. Bridges and J. Gathergood (2008) 'Drivers of Over-Indebtedness: Report to the
Department for Business, Enterprise and Regulatory Reform'.

- 201 -

Dolan, P., T. Peasgood and M. White (2008) 'Do we really know what makes us happy? A
review of the economic literature on the factors associated with subjective well-being'.
Journal of Economic Psychology, 29(1), 94122.
Dossey, L. (2007) 'Debt and Health'. Explore: The Journal of Science and Healing, 3(2), 83
90.
Drentea, P. and P.J. Lavrakas (2000) 'Over the Limit: The Association Among Health, Race,
and Debt'. Social Science & Medicine, 50(4), 517529.
Drexler, A., G. Fischer and A.S. Schoar (2010) 'Keeping it Simple: Financial Literacy and
Rules of Thumb'. Centre for Economic Policy Research CEPR Discussion Papers No.
7994, London.
Duesenberry, J. S. (1949) 'Income, Saving and the Theory of Consumer Behavior, Cambridge,
MA: Harvard University Press.
Duflo, E., R. Glennerster and M. Kremer (2007) 'Chapter 61 Using Randomization in
Development Economics Research: A Toolkit' In T. Paul Schultz and John A. Strauss
(Ed.), Handbook of Development Economics (pp. 38953962). Elsevier.
Duvendack, M., R. Palmer-Jones, J. Copestake, L. Hooper, Y. Loke and N. Rao (2011) 'What
is the evidence of the impact of microfinance on the well-being of poor people?'. EPPICentre, Social Science Research Unit, Institute of Education, University of London,
London.
Engelbrecht, H.-J. (2009) 'Natural capital, subjective well-being, and the new welfare
economics of sustainability: Some evidence from cross-country regressions: Special
Section: Analyzing the global human appropriation of net primary production - processes,
trajectories, implications'. Ecological Economics, 69(2), 380388.
Engle, P. (1993) 'Influences of mothers and fathers income on childrens nutritional status in
Guatemala'. Social Science and Medicine, 37, 13031312.
European Commission (2008) 'Towards a Common Operational European Definition of OverIndebtendess'. OEE Studies, Brussels.
Feeny, S. and A. de Silva (2012) 'Measuring absorptive capacity constraints to foreign aid'.
Economic Modelling, 29(3), 725733.
Field, E. and R. Pande (2008) 'Repayment Frequency and Default in Microfinance: Evidence
from India'. Journal of the European Economic Association, 6(2/3), 501509.
- 202 -

Fine, B. (2002) 'Economics Imperialism and the New Development Economics as Kuhnian
Paradigm Shift?'. World Development, 30(12), 20572070.
Finmark Trust (2010) 'FinScope Ghana 2010: Making financial markets work for the poor.
Launch Presentation, Accra.
Fouillet, C. (2006) 'La microfinanceserait-elle devenue folle? Crise en Andrha Pradesh'.
Gret/Cirad, Espace Finance Bulletin d'Information du Mardi No. 24-04-2006, Paris.
Friedman, M. (1957) 'A Theory of the Consumption Function, Princeton: Princeton University
Press.
Fudenberg, D. and D.K. Levine (2006) 'A Dual-Self Model of Impulse Control'. The
American Economic Review, 96(5), 14491476.
Gardeva, A. and E. Rhyne (2011) 'Opportunities and Obstacles to Financial Inclusion: Survey
Report'. Center for Financial Inclusion at Accion Center for Financial Inclusion Publication
No. 12, Washington, D.C.
Garikipati, S. (2008) 'The Impact of Lending to Women on Household Vulnerability and
Women's Empowerment: Evidence from India: Special Section: Social Movements and the
Dynamics of Rural Development in Latin America (pp. 2874-2952)'. World Development,
36(12), 26202642.
Gathergood, J. (2012) 'Self-control, financial literacy and consumer over-indebtedness'.
Journal of Economic Psychology, 33(3), 590602.
Gervedink Nijhuis, C. J., J.M. Voogt and J.M. Pieters (2012) 'The cultural complexity of
international collaboration: Conditions for sustainable curriculum development in Ghana'.
International Journal of Intercultural Relations, 36(5), 647658.
Ghatak, M. and T.W. Guinnane (1999) 'The economics of lending with joint liability: theory
and practice'. Journal of Development Economics, 60(1), 195228.
Giannetti, C. and N. Jentzsch 'Credit Reporting, Financial Intermediation and Identification
Systems: International Evidence'. Journal of International Money and Finance, (0).
Gloukoviezoff, G. (2008) 'De lexclusion linclusion bancaire des particuliers en France:
Entre ncessit sociale et contrainte de rentabilit. Doctoral thesis'. Universit Lumire
Lyon 2, Lyon.
Godquin, M. (2004) 'Microfinance Repayment Performance in Bangladesh: How to Improve
the Allocation of Loans by MFIs'. World Development, 32(11), 19091926.
- 203 -

Goetz, A. M. and R. Sen Gupta (1996) 'Who takes the credit? Gender, power, and control
over loan use in rural credit programs in Bangladesh'. World Development, 24(1), 4563.
Gonzalez, A. (2008) 'Microfinance, Incentives to Repay, and Overindebtedness: Evidence
from a Household Survey in Bolivia. Doctoral thesis'. Ohio State University, Ohio.
Gonzalez, A. (2009) 'Consumption, Commercial or Mortgage Loans: Does It Matter for MFIs
in LAC?'. Microfinance Information Exchange, Inc. Data Brief No. 3.
Gonzalez, A. (2010) 'Is Microfinance Growing Too Fast?'. MIX Data Brief No. 5.
Gonzlez-Vega, C. and M. Villafani-Ibarnegaray (2011) 'Microfinance in Bolivia: Foundation
of the Growth, Outreach and Stability of the Financial System' In B. Armendriz & M.
Labie (Eds.), The Handbook of Microfinance (pp. 203251). London, Singapore: World
Scientific Publishing.
Grammling, M. (2009) 'Cross-borrowing and Over-indebtedness in Ghana: Empirical
Evidence from Microfinance Clientele and Small Enterprises. Technical draft for
discussion'. ProCredit Holding, Frankfurt am Main.
Greenwald, B. (1983) 'A general analysis of bias in the estimated standard errors of least
square coefcients'. Journal of Econometrics, 22(3), 323338.
Gurin, I. (2000) 'Pratiques montaires et financires des femmes en situation de prcarit:
Entre autonomie et dpendance. Doctoral thesis in economics [Monetary and financial
practices of women in precarious situations: Between autonomy and dependance]'.
Universit Lumire Lyon II, Lyon.
Gurin, I. (2006) 'Women and Money: Lessons from Senegal'. Development and Change,
37(3), 549570.
Gurin, I. (2011) 'The Gender of Finance and Lessons for Microfinance' In B. Armendriz &
M. Labie (Eds.), The Handbook of Microfinance (pp. 589612). London, Singapore: World
Scientific Publishing.
Gurin, I. (in press) 'L'ducation financire ou comment apprendre aux pauvre bien
consommer' In I. Gurin & Selim (Eds.), Comment et quoi dpenser son argent?. Paris:
L'Harmattan.
Gurin, I., M. Roesch, O. Hlis and Venkatasubramanian (2009a) 'Microfinance,
Endettement et Surendettement: Une tude de cas en Inde du Sud [Microfinance,

- 204 -

Indebtedness and Over-indebtedness: A case study in Southern India]'. Revue Tiers Monde,
197, 131146.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2009b) 'The Social Meaning of
Over-Indebtedness and Creditworthiness in the Context of Poor Rural South-Indian
Households (Tamil Nadu), Draft paper presented at the RUME / IEDES workshop on overindebtedness in Paris, December 2009.
Gurin, I., S. Morvant-Roux and J.-M. Servet (2011a) 'Understanding the Diversity and
Complexity of Demand for Microfinance Services: Lessons from Informal Finance' In B.
Armendriz & M. Labie (Eds.), The Handbook of Microfinance (pp. 101122). London,
Singapore: World Scientific Publishing.
Gurin, I., M. Roesch, Venkatasubramanian and S. Kumar (2011b) 'The social meaning of
over-indebtedness and creditworthiness in the context of poor rural South India households
(Tamil Nadu)'. RUME Working Paper No. 2011-1, Paris.
Gurin, I., M. Roesch, O. Venkatasubramanian and B. D'Espallier (2012) 'Credit for whom
and for what? The diversity of borrowing sources and borrowing uses in south-India'.
Journal of International Development, (24 Issue Supplement S1), 122S137.
Gul, F. and W. Pesendorfer (2004) 'Self-Control and the Theory of Consumption'.
Econometrica, 72(1), 119158.
Haas, O. J. (2006) 'Overindebtedness in Germany'. Social Finance Programme, International
Labour Organisation Working Paper No. 44, Geneva.
Hamp, M. and C. Laureti (2011) 'Balancing flexibility and discipline in microfinance:
Innovative financial products that benefit clients and service providers'. Centre for
European Research in Microfinance, Universit Libre de Bruxelles CEB Working Paper
No. 11/044.
Hashemi, S. M., S.R. Schuler and A.P. Riley (1996) 'Rural credit programs and womens
empowerment in Bangladesh'. World Development, 24(4), 635653.
Hellmann, T. F., K.C. Murdock and J.E. Stiglitz (2000) 'Liberalization, Moral Hazard in
Banking, and Prudential Regulation: Are Capital Requirements Enough?'. The American
Economic Review, 90(1), 147165.
Hudon, M. (2007) 'Fair interest rates when lending to the poor'. Ethics and Economics, 5(1).

- 205 -

Hudon, M. (2009) 'Should Access to Credit be a Right?'. Journal of Business Ethics, 84, 17
28.
Hudon, M. and D. Traca (2011) 'On the Efficiency Effects of Subsidies in Microfinance: An
Empirical Inquiry'. World Development, 39(6), 966973.
Hulme, D. (2000) 'Impact Assessment Methodologies for Microfinance: Theory, Experience
and Better Practice'. World Development, 28(1), 7998.
Hulme, D. (2007) 'Is microdebt good for poor people? A note on the dark side of
microfinance' In T. Dichter & M. Harper (Eds.), What's wrong with Micofinance? (pp. 19
22). Warwickshire: Intermediate Technology Publications.
Jacobson, B. H., S.G. Aldana, R.Z. Goetzel, K.D. Vardell, T.B. Adams and R.J. Pietras (1996)
'The relationship between perceived stress and self-reported illness-related absenteeism'.
American Journal of Health Promotion, 11(1), 5461.
Jianakopolos, N. A. and A. Bernasek (1998) 'Are women more risk averse?'. Economic
Inquiry, 36(4), 620630.
Kabeer, N. (2001) 'Conflicts Over Credit: Re-Evaluating the Empowerment Potential of
Loans to Women in Rural Bangladesh'. World Development, 29(1), 6384.
Kahneman, D. and A. Tversky (1984) 'Choices, values and frames'. American Psychologist,
39(4), 341350.
Kanbur, R. (2002) 'Economics, Social Science and Development'. World Development, 30(3),
477486.
Kappel, V., A. Krauss and L. Lontzek (2010) 'Over-Indebtedness and Microfinance Constructing an Early Warning Index'. Center for Microfinance, University of Zurich,
responsAbility, Council of Microfinance Equity Funds, Triodos Investment Management,
Zurich.
Karlan, D. and N. Goldberg (2011) 'Microfinance Evaluation Strategies: Notes on
Methodology and Findings' In B. Armendriz & M. Labie (Eds.), The Handbook of
Microfinance (pp. 1758). London, Singapore: World Scientific Publishing.
Karlan, D. and J. Morduch (2010) 'Access to Finance' In D. Rodrik & M. R. Rosenzweig
(Eds.), Handbook in economics: Vol. 5. Handbook of Development Economics (pp. 4703
4784). Amsterdam: Elsevier.

- 206 -

Karlan, D. and M. Valdivia (2011) 'Teaching Entrepreneurship: Impact of Business Training


on Microfinance Clients and Institutions'. Review of Economics and Statistics, 93(2), 510
527.
Karlan, D. and J. Zinman (2009) 'Expanding Microenterprise Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts in Manila'. Financial Access
Initiative, Cambridge, MA.
Karlan, D. and J. Zinman (2010) 'Expanding Credit Access: Using Randomized Supply
Decisions to Estimate the Impacts'. Review of Financial Studies, 23(1), 433464.
Karnani, A. (2009) 'Regulate Microcredit to Protect Borrowers'. University of Michigan Ross
School of Business Working Paper No. 1133.
Kempson, E. (2002) 'Over-indebtedness in Britain: A report to the Department of Trade and
Industry'. Personal Finance Research Centre, University of Bristol, Bristol.
Khandker, S. R. (2003) 'Microfinance and poverty: Evidence using panel data from
Bangladesh'. World Bank World Bank Policy Research Working Paper No. 2945,
Washington, D.C.
Khandker, S. R., B. Khalily and Z. Kahn (1995) 'Grameen Bank: Performance and
sustainability'. World Bank Discussion Paper No. 306, Washington, D.C.
Kim, J., B. Sorhaindo and E.T. Garman (2006) 'Relationship between Financial Stress and
Workplace Absenteeism of Credit Counseling Clients'. Journal of Family and Economic
Issues, 27(3), 458478.
Kirchler, E., E. Hoelzl and B. Kamleitner (2008) 'Spending and credit use in the private
household'. Journal of Socio-Economics, 37(2), 519532.
Krishnaswamy, K. (2007) 'Competition and Multiple Borrowing in the Indian Microfinance
Sector'. Institute for Financial Management and Research, Centre for Micro Finance
Working Paper.
Krishnaswamy, K. and A. Ponce (2010) 'An analysis of the Karnataka mass defaults'.
Working Paper.
Labie, M. (2004) 'Introduction'. Mondes en Dveloppement, 126(2), 78.
Labie, M. (2007) 'Rflexions prliminaires pour une approche thique de la gestion des
organisations en microfinance'. Ethics and Economics, 5(1), 18.

- 207 -

Laibson, D. (1997) 'Golden Eggs and Hyperbolic Discounting'. Quarterly Journal of


Economics, 112(2), 443477.
Lascelles, D. and Mendelson, S. (2012) #Microfinance Banana Skins 2012, The CSFI survey
of microfinance risk: Staying Relevant. Centre for the Study of Financial Innovation
(CSFI), Kent, UK.
Lea, S. E. G., P. Webley and R.M. Levine (1993) 'The Economic Psychology of Consumer
Debt'. Journal of Economic Psychology, 14(1), 85119.
Lea, S. E. G., P. Webley and C.M. Walker (1995) 'Psychological Factors in Consumer Debt:
Money Management, Economic Socialization, and Credit Use'. Journal of Economic
Psychology, 16(4), 681701.
Livingstone, S. M. and P.K. Lunt (1992) 'Predicting personal debt and debt repayment:
Psychological, social and economic determinants'. Journal of Economic Psychology, 13(1),
111134.
Lusardi, A. and O.S. Mitchell (2007) 'Baby Boomer retirement security: The roles of
planning, financial literacy, and housing wealth'. Journal of Monetary Economics, 54(1),
5224.
Lusardi, A. and P. Tufano (2009) 'Debt Literacy, Financial Experiences, and
Overindebtedness'. National Bureau of Economic Research NBER Working Paper Series
No. 14808, Cambridge, MA.
Luttmer, E. F. P. (2005) 'Neighbors as Negatives: Relative Earnings and Well-Being'.
Quarterly Journal of Economics, 120(3), 9631002.
Lyman, T., T. Lythgoe, M. Miller, X. Reille and S. Sankaranarayan (2011) 'Credit Reporting
at the Base of the Pyramid: Key Issues and Success Factors'. CGAP and IFC FORUM No.
1, Washington, D.C.
Lyons, A. C. and J. Fisher (2006) 'Gender Differences in Debt Repayment Problems after
Divorce'. Journal of Consumer Affairs, 40(2), 324346.
Marr, A. (in press) 'Impact of group dynamics on microfinance sustainability: the case of
Peru'. Journal of Development Effectiveness, (Special Issue).
Matin, I. (1997) 'Repayment performance of Grameen Bank borrowers: the unzipped' state'.
Savings and Development, 4, 451473.

- 208 -

Maurer, K. and J. Pytkowska (2011) 'Indebtedness of Microcredit Clients in Bosnia and


Herzegovina: Results from a comprehensive field study'. EFSE Development Facility.
Mayoux, L. (2001) 'Tackling the down side: Social capital, women's empowerment and
micro-finance in Cameroon'. Development and Change, 32(3), 421450.
Mayoux, L. (2011) 'Taking Gender Seriously: Towards a Gender Justice Protocol for
Financial Services' In B. Armendriz & M. Labie (Eds.), The Handbook of Microfinance
(pp. 613642). London, Singapore: World Scientific Publishing.
McIntosh, C. and B. Wydick (2005) 'Competition and microfinance'. Journal of Development
Economics, 78(2), 271298.
McIntosh, C., E. Sadoulet, S. Buck and T. Rosada (in press) 'Reputation in a public goods
game: Taking the design of credit bureaus to the lab'. Journal of Economic Behavior &
Organization
Meier, S. and C.D. Sprenger (in press) 'Discounting financial literacy: Time preferences and
participation in financial education programs'. Journal of Economic Behavior &
Organization,
Mersland, R. and R.. Strm (2010) 'Microfinance Mission Drift?'. World Development,
38(1), 2836.
Modigliani, F. (1980) 'The Collected Papers of Franco Modigliani: The Life Cycle Hypothesis
of Saving'. (Abel, A., Ed.), Cambridge, MA; London: MIT Press.
Modigliani, F. and R. Brumberg (1980) 'Utility Analysis and the Consumption Function. An
Interpretation of Cross-Section Data. Post Keynesian Economics, edited by Kenneth K.
Kurihara, pp. 388-436 Rutgers University Press 1954' In A. Abel (Ed.): Vol. 2, The
Collected Papers of Franco Modigliani. The Life Cycle Hypothesis of Saving, pp. 79127.
Cambridge, MA; London: MIT Press.
Montgomery, R. (1996) 'Disciplining or protecting the poor? Avoiding the social costs of peer
pressure in micro-credit schemes'. Journal of International Development, 8(2), 289305.
Morvant-Roux, S. (2009a) 'Accs au microcrdit et continuit des dynamiques d'endettement
au Mexique: Combiner l'anthropologie conomique avec l'conomtrie [Access to
microcredit and continuity of indebtedness dynamics in Mexico: Combining economic
anthropology with econometrics]'. Revue Tiers Monde, 196, 109130.

- 209 -

Morvant-Roux, S. (2009b) 'L'Anthropo-conomie: un dtour indispensable pour comprendre


l'appropriation de la microfinance [Anthropo-economy: An unavoidable detour to
understand the appropriation of microfinance]'. Revue Tiers Monde, (197), 109130.
Moulton, B. (1990) 'An illustration of a pitfall in estimating the effects of aggregate variables
on micro units'. Review of Economics and Statistics, 72(2), 334338.
Muldrew, C. (2000) 'The economy of obligation, The culture of credit and social relations in
early modern England'. ([Nachdr.]). Early modern history, Basingstoke: Macmillan [u.a.].
Murray, A. P. (1997) 'Debt and "the" consumer'. Business Economics, 32(2), 4145.
Nio-Zarazua, M. and J. Copestake (2009) 'Financial inclusion, vulnerability, and mental
models: From physical access to effective use of financial services in a low income area of
Mexico City'. Savings and Development, 23(4), 353380.
Notani, A. S. (1997) 'Perceptions of Affordability: Their Role in Predicting Purchase Intent
and Purchase'. Journal of Economic Psychology, 18(5), 525546.
Ocasio, W. (1997) 'Towards an attention-based view of the firm'. Strategic Management
Journal, 18(S1), 187206.
Odell, K. (2010) 'Measuring the Impact of Microfinance: Taking Another Look'. Grameen
Foundation USA Grameen Foundation USA Publication Series, Washington, DC.
Olsen, R. A. and C.M. Cox (2001) 'The Inuence of Gender on the Perception and Response to
Investment Risk: The Case of Professional Investors'. Journal of Psychology and Financial
Markets, 2(1), 2936.
Oshio, T. and M. Kobayashi (2010) 'Income inequality, perceived happiness, and self-rated
health: Evidence from nationwide surveys in Japan'. Social Science & Medicine, 70(9),
13581366.
Paxton, J., D. Graham and T. Cameron (2000) 'Modeling Group Loan Repayment Behavior:
New Insights from Burkina Faso'. Economic Development and Cultural Change, 48(3),
639655.
Pischke, J. D. von (1991) 'Finance at the frontier, Debt capacity and the role of credit in the
private economy, Washington, D.C: World Bank.
Pitt, M. M., S.R. Khandker and J. Cartwright (2006) 'Empowering Women with Micro
Finance: Evidence from Bangladesh'. Economic Development and Cultural Change, 54(4),
791831.
- 210 -

Pytkowska, J. and S. Spannuth (2012) 'Indebtedness of Microcredit Clients in Kosovo:


Results from a comprehensive field study'. Microfinance Centre, Finance in Motion; EFSE.
Rahman, A. (1999) 'Micro-credit Initiatives for Equitable and Sustainable Development: Who
Pays?'. World Development, 27(1), 6782.
Ranyard, R. and G. Craig (1993) 'Estimating the duration of a flexible loan: The effect of
supplementary information'. Journal of Economic Psychology, 14(2), 317335.
Ranyard, R. and G. Craig (1995) 'Evaluating and budgeting with installment credit: An
interview study'. Journal of Economic Psychology, 16(3), 449467.
Rhyne, E. (2001) 'Mainstreaming microfinance, How lending to the poor began, grew, and
came of age in Bolivia, Bloomfield, CT: Kumarian Press.
Rinaldi, L. and A. Sanchis-Arellano (2006) 'Household debt sustainability. What explains
household non-performing loans?: An empirical analysis'. European Central Bank Working
Paper Series No. 570, Frankfurt am Main.
Rodrik, D. (2008) 'The New Development Economics: We Shall Experiment, but How Shall
We Learn?'. HKS Working Paper No. RWP08-055.
Roesch, M. and O. Hlis (2007) 'La microfinance: outil de gestion du risque ou de mise en
danger par surendettement? [Microfinance: A tool for risk management or for the creation
of over-indebtendness risks]'. Autrepart, 44, 119140.
Rosenberg, R. (1999) 'Measuring Microcredit Delinquency: Ratios can be harmful to your
health'. CGAP CGAP Occasional Paper No. 3, Washington DC.
Rozas, D., I. Barrs, C. Connors and E. Rhyne (2011) 'Implementing Client Protection in
Microfinance: The State of the Practice, 2011'. Center for Financial Inclusion Center for
Financial Inclusion Publication No. 14. Washington DC
Rutherford, S. (2000) 'The Poor and Their Money, An Essay about Financial Services for
Poor People, Oxford: Oxford University Press.
Sarthou-Lajus, N. (1997) 'L'thique de la Dette, Paris: Presses Universitaires de France.
Schicks, J. (2012) 'The over-indebtedness of microborrowers in Ghana - An empirical study
from a customer protection perspective'. Center for Financial Inclusion Center for
Financial Inclusion Publication No. 15, Washington DC.
Schicks, J. (in press) 'From a supply gap to a demand gap? The risk and consequences of
over-indebting the underbanked' In J.-P. Gueyie, R. Manos, & J. Yaron (Eds.),
- 211 -

Microfinance in Developing Countries. Issues, Policies and Performance Evaluation.


Palgrave Macmillan.
Schicks, J. (in press) 'Microfinance Over-Indebtedness: Understanding its Drivers and
Challenging the Common Myths'. Oxford Development Studies,
Schicks, J. and R. Rosenberg (2011) 'Too Much Microcredit? A Survey of the Evidence on
Over-Indebtedness'. CGAP Occasional Paper 19, Washington DC.
Schreiner, M. (2004) 'Scoring Arrears at a Microlender in Bolivia'. Journal of Microfinance,
6(2), 6588.
Sen, A. (1999) 'Development as Freedom: Oxford University Press.
Sharma, M. and M. Zeller (1997) 'Repayment performance in group-based credit programs in
Bangladesh: An empirical analysis'. World Development, 25(10), 17311742.
Smets, P. and E. Bhre (2004) 'When Coercion Takes Over: The Limits of Social Capital in
Microfinance Schemes' In H. Lont & O. Hospes (Eds.), Livelihood and microfinance.
Anthropological and sociological perspectives on savings and debt (pp. 215236). Delft:
Eburon Academic Publishers.
Soman, D. and A. Cheema (2002) 'The Effect of Credit on Spending Decisions: The Role of
the Credit Limit and Credibility'. Marketing Science, 21(1), 3253.
Stearns, K. (1991) 'The Hidden Beast: Delinquency in Microenterprise Credit Programs'.
ACCION International Discussion Paper Series No. 5, Washington DC.
Steel, W. (2010) 'Ghanas Microfinance Industry: Evolution and Prospects. Draft August
2010'. ISSER, University of Ghana, Prepared for IFC Access to Finance Advisory Microfinance. Accra.
Stone, B. and R.V. Maury (2006) 'Indicators of personal financial debt using a multidisciplinary behavioral model'. Journal of Economic Psychology, 27(4), 543556.
Strotz, R. H. (1956) 'Myopia and Inconsistency in Dynamic Utility Maximization'. The
Review of Economic Studies, 23(3), 165180.
Thaler, R. (1985) 'Mental Accounting and Consumer Choice'. Marketing Science, 4(3), 199
214.
Thomas, D. (1990) 'Intrahousehold allocation: An inferential approach'. Journal of Human
Resources, 25(4), 635664.
- 212 -

Tiwari, A., A. Khandelwal and M. Ramji (2008) 'How Do Microfinance Clients Understand
their Loans?'. Institute for Financial Management and Research, Centre for Micro Finance
Working Paper Series No. 25.
Tokunaga, H. (1993) 'The use and abuse of consumer credit: Application of psychological
theory and research'. Journal of Economic Psychology, 14(2), 285316.
Vandone, D. (2009) 'Consumer credit in Europe, Risks and opportunities of a dynamic
industry'. Contributions to economics, Heidelberg: Physica.
Viaud, J. and C. Roland-Lvy (2000) 'A positional and representational analysis of
consumption. Households when facing debt and credit'. Journal of Economic Psychology,
21(4), 411432.
Vogelgesang, U. (2003) 'Microfinance in Times of Crisis: The Effects of Competition, Rising
Indebtedness, and Economic Crisis on Repayment Behavior'. World Development, 31(12),
20852114.
Webley, P. and E.K. Nyhus (2001) 'Life-Cycle and Disproportionate Routes into Problem
Debt'. British Journal of Psychology, 92, 423446.
White, H. (1980) 'A heteroscedasticity-consistent covariance matrix estimator and a direct test
for heteroscedasticity'. Econometrica, 48(4), 817830.
Willis, L. E. (2008) 'Against financial literacy education'. Iowa Law Review, 94, 197285.

- 213 -

Appendix
Schicks, J. and R. Rosenberg (2011) 'Too Much Microcredit? A Survey of the Evidence on
Over-Indebtedness'. CGAP Occasional Paper 19. Washington DC.

Schicks, J. (2011) 'The over-indebtedness of microborrowers in Ghana - An empirical


study from a customer protection perspective'. Center for Financial Inclusion Publication
No. 15. Washington DC.

- 214 -

OCCASIONAL PAPER
No. 19
September 2011
Jessica Schicks and
Richard Rosenberg

Too Much Microcredit? A


Survey of the Evidence on
Over-Indebtedness
Important Notes about the Scope of This Paper
This paper is written primarily for microlenders, along with the institutions that fund and support them. CGAP is
exploring development of a separate paper addressing regulatory aspects of over-indebtedness.
We use the terms microcredit, microloan, and microborrower in a narrow sense, to refer to loans to
low-income peoplemainly unsalaried peopleusing new techniques developed over the past 30 years (see
Christen, Lauer, Lyman, and Rosenberg [2011] for a list) and made by lenders who usually describe themselves
as providing microcredit or microfinance. We dont imply that such lenders provide all, or even most, of
the formal-sector loans to poor or low-income households. In many countries, microfinance institutions offering
microcredit compete with other forms of low-income retail credit, such as consumer lending and merchandise
finance, all of which can have a bearing on the indebtedness levels of microcredit clients.
This is an exploratory paper. We examine conceptual issues and the limited empirical evidence about overindebtedness in microcredit markets. That evidence does yet not permit a general conclusion about the extent
of microcredit over-indebtedness worldwide. We also offer a rudimentary checklistcertainly not a how-to
manualof possible approaches to dealing with over-indebtedness risk. Most of these are presented only as
options, not recommendations, because their feasibility and usefulness can depend heavily on local circumstances.

Executive Summary
Microcredit has long had an enviable repayment
recordlevels of delinquency and default have
been very low. But more recently, collection
problems have appeared in some major markets.
In a review of four countries, Chen, Rasmussen,
and Reille (2010) reported that delinquent loans,
which averaged 2 percent of portfolio in 2004,
skyrocketed to 2009 levels of 7 percent in Bosnia
Herzogovina, 10 percent in Morocco, 12 percent
in Nicaragua, and 13 percent in Pakistan. In some
of these countries, subsequent levels have risen
quite a bit higher. More recently, collection has
collapsed in the Indian state of Andhra Pradesh.1
Delinquency and default threaten the viability of
microlending institutions. But this paper looks at
repayment problems mainly from the perspective
of the clients, not the lenders. We examine
definitions of over-indebtedness in some detail,

but as a rough, provisional definition to begin the


discussion, let us say that borrowers are overindebted if they have serious problems repaying
their loans.2 This definition implies a view that
borrowers can be over-indebted even if they are
repaying their loans.
Over-indebtedness often implies heightened
vulnerability and further impoverishment of
borrowers. Material effects include reduced
consumption levels, late fees, asset seizures,
downward spirals of ever-increasing debt, and
eventually, a loss of creditworthiness. There are
sociological effects related to peer pressure and
a loss of social position, as well as psychological
effects on mental and physical health. In extreme
cases, borrowers desperation can even lead to
suicide.
The objective should be to reduce the prevalence of
over-indebtedness to reasonable levels. Complete

1 In Nicaragua, Pakistan, and Andhra Pradesh, at least some of the repayment problem was due to strategic default: borrowers who were able to
repay choosing not to do so.
2 Well address what serious means later in the discussion.

elimination of over-indebtedness is not a practical


goal: the only way to accomplish this would be
to drastically restrict access to microloans. Many
measures to reduce over-indebtedness can have
some degree of negative impact on borrower
access and cost. For instance, tightening credit
standards will lower over-indebtedness, but it will
also prevent some loans to borrowers who would
have had no difficulty repaying. Careful balancing
is needed.

Warning Light or Crisis


For most of its history, the main concern of the
microfinance movement has been to rapidly
increase delivery of financial servicesespecially
creditto low-income clients. However, the recent
crises in a few markets have fueled growing concern
that microcredit may be getting borrowers into
trouble. Rather than assuming that the problems in
these markets are isolated occurrences, there may
be reason to be more alert to this risk worldwide.
Present evidence doesnt permit a conclusion about
the actual prevalence of over-indebtedness in most
markets. But several factors suggest that managers,
regulators, and funders ought to devote much more
attention and resources than they currently do to
investigating and perhaps reducing the number of
clients who are getting into trouble with microloans.
To reiterate for the sake of clarity, we are making an
assertion about the appropriate level of vigilance,
not about the proportion of borrowers who are in
fact over-indebted.
There are two main reasons for increased vigilance
about microcredit over-indebtedness: it might be
more prevalent than suspected in the past, and its
consequences may be more serious than we used
to think.

Prevalence
A rapidly increasing number of microcredit
markets, especially regional markets within
a country, are becoming competitive and

even approaching saturation, at least for the


typical current credit products.3 (Demand
estimates about the number of people who
want a typical microloan and dont yet have
access tend to be inflated.) In retail credit
markets, competition and saturation tend to
increase over-indebtedness. In their quest
for expansion and market share, lenders
turn to higher risk borrowers, and may get
sloppy with their internal risk management
systems. In the absence of a credit reporting
service, lenders find it harder to assess client
repayment capacity, because they have no
way of knowing about clients repayment
obligations to competing lenders.
Borrowers dont always make smart choices.
The emerging field of behavioral economics
is challenging the classical assumption that
borrowers can be counted on to behave
rationally and make decisions that serve their
own best interests.
Strong repayment statistics dont assure us
that all is well.
Of the fewwe have located sixempirical
studies of microcredit over-indebtedness so
far in particular countries, most have found
significant levels of over-indebtedness.
(At the same time, these results cannot
be generalized. Not only is the number
of studies very small, but the sample is not
representative. Most of the studies were done
because there were pre-existing concerns
about over-indebtedness in the particular
country or local markets.)

Consequences and Implications


The most important consequences of overindebtedness are the impact on borrowers, as
discussed above.
Over-indebtedness can lead to political
backlash and damaging over-reaction by
policy makers. Such over-reaction can
destabilize the industry and deny access to
many potential borrowers. There is also a risk
of backlash from donors and public and social

3 The final qualifying clause is important: this statement and much of this paper have reference to the simple and relatively rigid microcredit
products that have made up most of the supply until now. Many microlenders are now looking to supplement their offerings with more flexible
and client-responsive products, including not just loans but also voluntary and commitment savings, insurance, and money transfer services.

investors, who may turn away from funding


not only microcredit, but microfinance more
broadly.
Today there is less confidence in assertions
that microcredit can raise millions of people
out of poverty. The actual benefits may be
considerably more modest. If the quantum of
benefit we expect is lower, then the potential
downsides for clients that we are willing to
tolerate should be lower.

too far out of step with borrowers cash flows can


create serious repayment distress even when the
debt amount is reasonable, especially if there is
rigid enforcement of a zero tolerance policy
toward delinquency. Once borrowers get into
trouble, over-aggressive collection practices can
worsen their problem. Finally, there is a complex
debate about whether it is unduly risky to lend
to borrowers who wind up using their loans for
consumption rather than for investment in an
income-generating activity.

Causes of Over-Indebtedness
Multiple factors contribute to over-indebtedness:
lender behavior can put borrowers at undue risk,
clients themselves make bad borrowing decisions,
and external factors beyond either partys control
(e.g., illness or natural disaster) can push borrowers
into situations where its very difficult or impossible
to repay.
It seems plausible that very rapid growth of an
individual lender could strain its systems and lead
to loan portfolio problems. But it is hard to find
support for this proposition in the statistical data.
Rather, it appears more likely that repayment
deterioration is associated with characteristics of
the aggregate market (which is not necessarily a
nationwide market), including the growth rate in
aggregate number or amount of active loans, as
well as the penetration ratei.e., the percentage
of the population in the market who have loans.
As noted, microcredit providers may relax their
lending standards or stray from proven loan
management methods under conditions of
competition in markets approaching saturation.
Over-aggressive marketinge.g., pressuring
borrowers to take out a new loan after they
have paid off an old oneadds to risk. Lenders
sometimes fail to give borrowers clear and
accurate information about loan costs and terms,
communicated in a format that supports good
decision making. The common system of gradually
increasing loan sizes sometimes becomes
practically automatic, which eventually puts clients
at risk if there has not been sufficient investigation
of their ability to repay. Loan products that are
too inflexible and repayment schedules that are

Some of the impetus behind over-indebtedness


comes from borrowers. The emerging field of
behavioral economics has mounted a strong
challenge to the proposition that borrowers actually
behave like the fully rational homo economicus of
classical economic theory. Behavioral experiments
confirm and extend commonsense perceptions
about borrower biases. Many borrowers put too
much weight on present gratification, because it
is salient, and pay too little attention to future
consequences because they seem less real. Peoples
predictions of the future tend to be over-optimistic,
and habit persistence causes them to reduce
consumption too slowly when net income declines.
External shocks can turn a perfectly manageable
repayment situation into an impossible one. Poor
people often experience sudden reductions in
income (e.g., a job loss or illness in the household)
or large unexpected expenses (e.g., accidents,
medical expenses, or funeral obligations). Other
shockse.g., natural disasters or manmade
conflicts that destroy livelihoodscan affect many
borrowers at the same time.

Defining and Measuring


Over-Indebtedness
Coming up with a precise definition of overindebtednesse.g., for research or regulatory
purposesis a surprisingly complex challenge. We
look at six different approaches that are used to
define, or proxy for, over-indebtedness. All have
limitations.
Negative impact. Calling people over-indebted
would seem to assert that they have too much

debt, which would ordinarily mean more debt


than is good for them. In this view, the question
would come down to whether loans are making
borrowers worse off. This might be a good
definition from a theoretical perspective, but as
a practical matter, reliably determining whether
loans are helping or hurting individual borrowers
is too difficult, expensive, and time-consuming
to be used for monitoring purposes. Also, most
people wouldnt describe a borrower as overindebted if the loan had a minor negative impact
but was easy to repay.

Default and arrears (late payments). These are


the most commonly used indicators of overindebtedness because they are the easiest to
measure. The main problem with these indicators
is that they fail to embrace a situation that most
people would think represented too much debt.
For instance, to avoid being socially humiliated by
default, some borrowers might repay their loans
by making drastic sacrifices, such as going without
food, taking children out of school, or selling off
productive assets. In addition, repayment failure
isnt a leading indicator; rather, its a trailing
indicator that flags a problem that may have
reached a point of no return months or even years
earlier.

Debt ratios. Over-indebtedness is often


measured by ratios that compare borrowers
total debt, or periodic debt service payments,
to their income or assets. Such ratios are clearly
meaningful and usefulthey are at the core of
many lenders systems for appraising borrowers
creditworthiness. One limitation, though, is
that it can be hard to determine an appropriate
threshold ratio beyond which the borrower is
regarded as over-indebted. Because there are
wide differences in borrower circumstances,
there are also wide differences in the level of
debt they can comfortably manage. Furthermore,
most microcredit markets dont yet have credit
reporting systems that allow determination of
borrowers total debt from formal lenders (not to
mention informal loan sources), and even where

such systems exist, they dont contain income


information for microborrowers.
Multiple borrowing. Most (but not all) of the
available empirical studies find that multiple
concurrent borrowing is associated with some
degree with higher risk of default. But multiple
debt is a very common cashflow management
technique for poor households, many of whom
manage such borrowing with minimal difficulty.
Multiple borrowing by itself is not a reliable
indicator of over-indebtedness.
Borrower struggle and sacrifice. The advantage
of this kind of definition is that it captures
situations where loans are repaid, but only at
the cost of severe distressa situation that most
people would regard as too much debt. One
such definition posits that the threshold of overindebtedness is reached when a microborrower
is continuously struggling to meet repayment
deadlines and structurally has to make unduly high
sacrifices related to his or her loan obligations.
The authors think that this is a strong and useful
definition for purposes of survey research. But
when are sacrifices unduly high? There is a good
argument for letting borrowers themselves make
that judgment. The more subjective approach
entails risks in terms of borrower idiosyncracy,
honesty, and bias, and would not be sharp-edged
enough for regulatory use. But these risks can be
managed in a research setting.
A crucial clarification is in order: the fact that a
borrower cant repay without severe sacrifice does
not automatically mean that the loan has hurt the
borrower. For instance, if a woman can repay only
by going without food for two days, its tempting
to conclude that she would have been better off
without the loan. But perhaps she borrowed the
money in the first place to avoid going without
food for two weeks, in which case the loan has
been a great help, and we would not want to
prevent it from being made.4
Loan repayment is only one of the cash demands
that poor households face, and many of those

4 In this clear cut hypothetical example, the borrower would probably not declare her skipped meals as an unduly high sacrifice related to the loan
repayment. But real situations are often less clear cut.

demands can be met only with struggle and


sacrifice, with or without microloans. This
consideration might temper our judgment as
to what level of struggle-and-sacrifice overindebtedness is reasonable in microlending.
Composite indicators. Given that all definitions
or proxies for over-indebtedness have their
limitations, some analysts construct indicators that
quantify and weight several different elements,
yielding a composite score.
Broader measures. More generally, its important
to recognize that debt is only one of the financial
vulnerabilities facing poor households, who also
have other obligatory payments. And there is a
strong argument for combining debt vulnerability
with other forms of financial vulnerabilitye.g.,
low and undependable incomes, exposure to
unexpected expenses that are large in relation
to income, etc.into a composite financial
vulnerability index. We do not explore this
broader approach because this paper focuses only
on microcredit clients, lenders, and funders.

The Empirics of Microcredit


Over-Indebtedness: What Do We
Know to Date?
As noted, weve been able to locate only six studies
that quantified microcredit over-indebtedness in
specific markets, using various definitions and
methods. Most of these studies found levels of
debt problems, usually in a subnational market,
that seem worrisome. However, the evidence
so far doesnt justify any broad generalization
about the situation worldwide. The number of
studies is small. More importantly, they represent
a very skewed sample. Most of the studies were
conducted in a given market precisely because
there were pre-existing danger signs of overindebtedness in that market. Table 1 offers a
concise (and therefore oversimplified) view of
study results.

What Can Be Done?


The list of practical steps by lenders and funders
that might prevent or remedy over-indebtedness

Table 1. Summary of empirical studies


Setting
(Author)

Methodology

Definition
of Over-indebtedness

Findings

Bolivia 19972001
(Gonzalez 2008)

Household (HH) survey

Costly unanticipated
actions to repay

85% of HH had at least


one occurrence during the
four years

Ghana 2009
(Grammling 2009)

Rapid market assessment,


borrower surveys, info
exchange between MFIs

12% over-indebted,
Microbusiness
decapitalizing (as assessed another 16% at risk
by researchers)

Country X 2009
(restricted report)

Debt service and expense


data from credit bureau
and loan files of lenders

Debt service greater than 17% over-indebted and


100% of HH income net of another 11% at risk (debt
service 75100% of net
other expenses
income)

Karnataka, India 2010


(Krishnaswamy 2011)

Loan records and HH


survey

Subjective report of stress; Over-indebtedness


reportedly high in mass
sacrifices to repay
default towns, but low in
nondefault towns

Tamil Nadu, India


20052009
(Gurin, Roesch,
Subramanian, and
Kumar 2011)

Qualitative interviews,
observation, and HH
survey (N=344)

Impoverishment through
debt

More (possibly much


more) than 20% overindebted

Multi-country study
(Kappel, Krauss, and
Lontzek 2010)

Preliminary composition
of an early warning index,
largely based on signaling
analysis with an MFI
survey + MIX market data

Chronic and involuntary


inability to meet all
payment obligations by
means of the households
excess cash. Proxies:
arrears, write-offs, loan
losses, debt-service ratio

14 potential early warning


indicators for overindebtedness. Highest
risk countries in sample:
BosniaHerzogovina,
Cambodia, Peru

is a long one. This paper can provide only a brief,


broad-brush survey of themnot a how-to manual,
but rather a checklist of actions to consider along
with a few factors to bear in mind. For most of
these topics, we imply no recommendation, not
only because the evidence base is thin, but also
because their feasibility and usefulness depend on
local circumstances.

Lender actions
Product design
Flexible product offerings, including savings
Reduction or tailoring of loan increments in
graduated lending ladders
Sales
Marketing practices and follow-on loans
Fair and intelligible disclosure of loan terms
Measures to improve clients financial literacy/
capability
Expansion into new areas rather than already
served ones
Loan underwriting
Evaluation of affordability in light of borrower
cash flows
Specific limits on debt-service ratios
Verification of borrowers repayment history
and other debts
Collection
Restraining of abusive collection practices
Appropriate rules for renegotiation of loans
Avoidance of inappropriate use of penalty
interest
Redress mechanisms for customer complaints
Other microlender practices
Staff incentives that dont encourage overlending
Staff training on avoiding over-indebtedness
Written policies, enforced through internal
audit
Specialized portfolio audits
Early warning systems for over-indebtedness

Marketwide tools
Industry codes
Credit reporting systems

Funders actions
Some observers think that an over-supply
of funding, along with funders pressure on
microlenders for growth and profitability, has
contributed to over-indebtedness crises. Donors
or socially oriented investors should factor
over-indebtedness risk into their evaluation of
microlenders as potential grantees or investees.
This includes looking at the saturation level of the
particular market, and whether the microlenders
are dealing appropriately with the range of options
presented above.
At a minimum, donors as well as public and socially
motivated investors ought to assure themselves
that any microlender they fund is
not using deceptive or high-pressure
marketing tactics
not structuring staff incentives in ways that
encourage over-lending
taking reasonable measures to check
on borrowers repayment capacity, past
repayment history, and outstanding
obligations with other lenders
maintaining and communicating clear written
policies to guide employees in addressing
over-indebtedness risk
not using collection techniques that are
abusive, given the local setting.
Donors that have grant funding for supporting
public goods can also finance some of the
above measures directlye.g., development of
credit reporting services, early warning systems,
or financial capability campaignsor support
regulators efforts to control over-indebtedness risk.

Introduction
Microcredit has long had an enviable repayment
recordlevels of delinquency and default in
competent institutions have been very low for

decades.5 But more recently, collection problems


have appeared in some major national or
subnational markets. In a review of four countries,
Chen, Rasmussen, and Reille (2010) reported that
delinquent loans, which averaged 2 percent of
portfolio in 2004, skyrocketed to 2009 levels of
7 percent in BosniaHerzogovina, 10 percent
in Morocco, 12 percent in Nicaragua, and 13
percent in Pakistan. In some of these countries,
subsequent levels have risen even higher. More
recently, collection has collapsed in the Indian
state of Andhra Pradesh.6 These and other
developments have focused a lot of attention on
over-indebtedness among microborrowers.
The topic of microcredit over-indebtedness raises a
lot of questions, the first of which is what we mean
when we use the term. This definitional issue is a
complex and difficult one that we explore at length
later in this paper. For present purposes, we can
make do with the imprecise notion that borrowers
are over-indebted if they have serious problems
paying off their loans. This provisional definition
reflects an important underlying choice. We do
not restrict the concept of over-indebtedness
to situations where the borrower is completely
unable to repay. If our main concern were the
sustainability of the lenders, we might view overindebtedness simply as debt that cant be repaid,
because that is what hurts a microlender most.
Rather, our primary focus is on the well-being
of clients, so our definition incorporates client
distress, even in cases where the loan gets paid.
Over-indebtedness can seriously damage clients.
They struggle to make repayments, cutting back
on basic consumption as well as other important
household expenditures, such as education or
healthcare. Then, of course, over-indebtedness
has material costs, such as late fees and, in
default cases, asset seizures. The resulting loss
of creditworthiness can deprive the household
of crucial cash management tools. On another

level, over-indebtedness can have sociological


implications, including peer pressure in groups, loss
of ones dignity and social position, and violence
in the household. Finally, over-indebtedness can
have other long-lasting mental and physical health
effects, including suicide in extreme cases.7
Over-indebtedness clearly hurts people. But
elimination of over-indebtedness is not a practical
goal. Everyone knows that borrowing does not
always work out well for the borrower. Sometimes
people get in over their heads because they make
an imprudent borrowing choice, or the microcredit
provider makes an imprudent lending choice. But
even when the borrowing and lending decisions have
been perfectly sensible, unpredictable circumstances,
such as emergencies or failed investments, can make
repayment difficult or impossible. The only way to
eliminate over-indebtedness altogether would be to
eliminate lending.
This basic point is worth emphasizing. If there is
microlending, some borrowers will inevitably wind
up with repayment problems. Thus, there is some
degree of trade-off: we cannot maximize borrowers
access and minimize debt stress at the same time.
For instance, the most obvious way to lower overindebtedness is to tighten lending standards. This
will prevent loans to some people who would wind
up having serious repayment problems. But it will
also prevent loans to some other people who could
have repaid without difficulty, even though they fell
short of the new loan requirements. As anyone who
makes credit decisions can tell us, its not always
easy to predict whos going to fall in which group.
The useful question is not whether microcredit is
over-indebting borrowers, but whether it is overindebting too many borrowers. How many is too
many? This belongs on the long list of questions
we cant settle in this paper. We dont try to
develop an a priori norm for acceptable levels
of over-indebtedness. The approach that seems

5 For historical collection data, see MIX Market (www.mixmarket.org).


6 E.g., CGAP (2010). It is unclear how much of the Andhra Pradesh problem stems from borrower inability to repay, as opposed to strategic
default.
7 See Schicks (2011) for a detailed discussion of the consequences over-indebtedness can have on borrowers and on other stakeholders.

more useful is to develop a working definition, or


definitions, of over-indebtedness; conduct more
research to determine actual levels in various
markets; and then attempt judgments (necessarily
fuzzy ones) about whether the level found in a
given market is reasonable or not.
Most people would have an instinctive sense that
over-indebting, say, a third of the borrowers of a
microfinance institution (MFI) would be unacceptable
in any circumstance.8 Likewise most would agree
that if only one out of every hundred borrowers
has serious difficulty in repaying, that would be
unfortunate but acceptable because the only way to
eliminate this problem would probably be to deny
credit to the 99 who can repay without difficulty.
Between 1 percent and 33 percent, where does the
tipping point of acceptability lie? Ones answer may
depend more on instinct than analysis. In any event,
we think there is a better chance of getting some
consensus about whether a given level is too high
or not when the discussion is focused on the actual
circumstances of a specific market.
This paper offers a broad-brush survey of a wide
range of issues associated with over-indebtedness.
It is neither exhaustive nor conclusive, but we hope
it advances the discussion of this crucial issue. In
particular we caution the reader not to expect
clear, concrete answers to the question, What
should I do about over-indebtedness right
now? We wish we could offer such answers,
but the state of the evidence does not permit it,
especially because so much depends on widely
differing local circumstances. The best we can do
is offer a menu of options to consider.
One further caveat: we discuss over-indebtedness
in the context of microcredit, but it is important
to recognize that many of the same dynamics
and issues apply to other forms of retail lending,
including consumer credit. In some countries,
there may be considerable overlap between the
markets for microcredit and other retail lending, so
the behavior of one set of lenders can over-indebt
people who also borrow from other lenders.

Section 1 asks whether lenders and funders


really need to be all that concerned with overindebtedness at this stage of microfinance
development. (Our answer is yeswe do not
know the extent of over-indebtedness worldwide,
but there are strong reasons to be alert for the
occurrence of such problems.)
Section 2 looks at causes of over-indebtedness.
We discuss some of the probable causes, though
we dont have the evidence to say much about
how widespread each of those causes is.
Section 3 addresses the harder-than-it-looks question
of how to define and measure over-indebtedness.
In Section 4, we look at the results of six
empirical studies that have tried to quantify overindebtedness and over-indebtedness risk.
Section 5 is a brief survey of potential responses
to over-indebtedness by lenders and funders.
When we find a problem, or anticipate one, what
if anything can be done about it?
Finally, a brief conclusion distils the core messages
of the paper.

1. Why Worry about


Over-Indebtedness?
For a long time, the main concern of microfinance
was expanding accessdelivering formal financial
services, mainly credit, to as many people as
possible. Few people were thinking about overindebtedness. As late as April 2008, Deutsche
Bank, the Boulder Institute, and CGAP convened
a group of microfinance practitioners and experts
to reflect together about the state of the industry,
its risks, and its future. As an early exercise,
participants each listed a few things that worried
them most about the industry. Over-indebtedness
was mentioned by only three of the 35 participants.9
In Microfinance Banana Skins 2008, hundreds of
practitioners, investors, analysts, and regulators

8 In this paper, the terms microlender, microcredit provider, and MFI are used indiscriminately, referring to any formal institution that makes
microloans, whether or not it also provides other financial services or serves other target clients.
9 After discussion, the conferees did go on to list over-indebtedness as a major threat (http://www.db.com/csr/en/docs/Pocantico_Declaration.pdf).

rated credit risk as a less serious danger than nine


other threats to the industry (Laschelles 2008).
Three years later, the picture is very different.
Driven in large part by repayment crises in Bosnia
Herzogovina, Morocco, Nicaragua, Pakistan,
and India, microcredit over-indebtedness and
its causes are the subject of intense discussion
and a growingif still smallbody of research.
In Microfinance Banana Skins 2011, credit risk
had climbed to the top of the list of microfinance
threats (Laschelles and Solomon 2011).
At one level, the answer to the question of whether
to be concerned about over-indebtedness is
obvious. Over-indebtedness hurts poor clients,10
whose welfare is the declared objective of most
microlenders, funders, and governments. And overindebtedness sooner or later tends to produce
delinquency and default, which threaten the lending
institutions own viability.11 So over-indebtedness
should always be a concern, in principle.
The more meaningful question is whether we have
reason to be concerned that over-indebtedness
may be at, or be approaching, problematic
levels. Or, to put the question in practical terms,
should we be devoting much more attention and
resources than we currently are to identifying
over-indebtedness problems and correcting or
preventing them?
Before we address this question, we need to
make an important clarification, starting with an
automotive analogy. When the check-engine
warning indicator on an autos dashboard lights
up, it may mean that there has been a major
malfunction that will destroy the engine if it is not
repaired promptly. But it might just as well mean
that there is nothing more than a minor glitch
in the engines computer. Either way, a smart
driver will want to have the engine checked by a
competent mechanic.

It is one thing to assert that there are reasons


for particular concern about over-indebting
microborrowers right now. We do assert thisthe
check-engine light is onand offer our reasons
below. It is quite another thing to assert that there
is a worldwide problem with over-indebtedness
that affects more than a few markets. We make no
such assertion, because the evidence to confirm
or refute it simply isnt available yet. Indeed, a
central contention of this paper is that we urgently
need much more research so we can find out how
much of a problem exists. Readers who forget this
clarification may walk away from the paper with a
picture containing more gloom and doom than the
authors intend or the evidence justifies.
Let us turn then to the reasons for particular
concern.
Competition and market saturation create
over-indebtedness risk. Since the inception of
the microfinance revolution, its rallying cry has
been the need to bring formal financial services
(mainly loans, in the early years) to the hundreds
of millions of poor people who have had no
access to them. Today, there are more and more
markets where that goal is being reached, at
least in the limited sense of making standard
microloan products available to almost all of
the people who can be served sustainably. In
other words, more and more microcredit markets
are starting to reach saturation. (Typically, local
markets in a country will become saturated
before the national market as a whole does.) As
this phenomenon spreads, it marks a momentous
shift in the development of the industry, requiring
re-examination of earlier conventional wisdom. In
particular, over-indebtedness risk rises, and MFIs
can expand their market only by developing new
financial products.
For many people, talk of market saturation seems
odd when most of the low-income population dont

10 We use poor throughout this paper as shorthand for poor and low-income. New tools have been developed for testing the poverty status of
microfinance clients. (See http://www.povertytools.org; http://progressoutofpoverty.org/.) Application of these tools over the past five years has
amply confirmed what experienced practitioners have been saying for a long time: most MFIs serve not only customers below the official poverty
line, but also substantial numbers of low-income customers who are, at least at a given point in time, above the poverty line.
11 Microlenders face a tight margin for error when it comes to collecting their loans. Delinquency (delay in payment) tends to spin out of control
if more than 10% of a portfolio becomes late by more than one repayment period. The corresponding critical maximum for annual rates of loan
losses (default) tends to be about 5% of portfolio. See Rosenberg (1999).

10

have a microloan. This is because there has been


a tendency to overestimate demand for standard
microcredit, sometimes drastically. Many poor
people have income sourcese.g., farmingthat
dont match up well with the microfinance loan terms
presently on offer. And among microentrepreneurs
for whom the standard loan terms are a better fit,
large proportions simply do not want to borrow.
Even those who want to borrow may not want to
borrow all the time. And finally, some who might
want to borrow cant be given loans because their
income is too small, unreliable, or irregular to safely
support repayment. There is growing evidence that
the actual number of people who want and qualify
for a standard microloan at any given time may be
considerably lower than the demand estimates that
MFIs and industry analysts have put forward in the
past.12
Adrian Gonzalez (2010) finds that growth in
country-level delinquency and default, which may
reflect market saturation, is statistically associated
with penetration rates as low as 10 microloans per
100 of population.13 At any rate, microlenders in
an increasing number of competitive markets are
finding that good new customers are getting harder
to recruit, and that most of their existing customers
have access to one or more other microlenders.
Gabriel Davel, the recently retired chief of South
Africas National Credit Regulator, argues that
when competitive retail credit markets approach
saturation, over-indebtedness problems will
arise almost inevitably, not just when a few badapple lenders irresponsibly ruin the market.14 As
competition intensifies, competitors are likely to
step up their efforts to capture as large a share
as they can of the remaining untapped market.
Once lenders have picked most of the lowhanging fruiti.e., low-risk borrowersthey
may relax their standards and start lending to a
higher risk clientele.15 Over-aggressive marketing

can also extend to current customers, who may be


encouraged, or at least permitted, to increase their
loan size beyond safe limits, or discouraged from
resting between loans (Guerin 2006; Hulme 2007;
Collins, Morduch, Rutherford, and Ruthven 2009;
Gonzalez 2008).16 When this kind of race to the
bottom occurs, even a careful and responsible
lender is at riskespecially if there is no sharing
of credit informationbecause its borrowers
ability to repay may be compromised by reckless
behavior of other lenders.
In most credit systems, there is a considerable
delay between the point where borrowers are
getting into trouble and the point where the
problem becomes apparent in the collection
statistics of lenders. David Roodman (2011, ch. 8)
points out that systemsnot just economic ones
but also biological ones like reindeer herds
are inherently prone to unsustainable expansion
and recurring crises if they combine high growth
pressure with a delayed feedback loop.
Borrowers often take advantage of their access to
multiple lenders, and accumulate concurrent loans.
Especially in settings where there is no credit
bureau, even a very careful lender will have trouble
evaluating clients repayment capacity, because it
is hard to find out about their other debt service
obligations. As all of this happens, more borrowers
risk getting in over their heads, eventually resulting
in serious delinquency and default (McIntosh and
Wydick 2005). (This does not mean that multiple
indebtedness is the same as over-indebtedness.
See Section 3.)
All these dynamics are exacerbated by the
common pattern of MFI expansion observed
in many countries. When MFIs open new
branches, they prefer, where possible, to do it
in safe places where a competitor has already
developed customer awareness of microcredit

12 For a fuller discussion of demand estimates, see Anand and Rosenberg (2008).
13 Gonzalez focuses on national population and penetration because the MIX data he relies on do not give regional breakouts. In fact, at the present
time microcredit markets in most countries are more likely to be regional rather than national.
14 Presentations at CGAP and at Deutschebank over-indebtedness roundtable, January 2011. Retail credit is aimed at the level of households
(including, among other things, their consumption and their informal income-producing activities) rather than at formal firms.
15 E.g., Robb (2007), Schoell (2010), Lupica (2009), White (2007), and Dick and Lehnert (2010).
16 Hellman, Murdock, and Stiglitz (2000) link increasing competition to opportunistic behavior by lenders. Their argument is that competition
reduces profits, lower profits imply lower franchise or charter values (namely, the capitalized value of expected future profits), and lower franchise
values reduce the incentives for making good loans, as bank owners would have a lower stake in the outcome. (Gonzalez 2008).

11

and demonstrated a strong market, rather than


expanding into previously unserved localities.17
Lenders may get sloppy with their internal systems
for tracking and managing delinquency. Borrowers
who have amassed more debt than theyre able
to pay can paper over the situation, for a while at
least, by using new loans from one source to pay
off old loans from another. This kind of concealed
over-indebtedness is especially likely when large
amounts of new lending capital are flowing into
the market. And even when borrowers in trouble
have no other formal lender to go to, or have
exhausted their other borrowing options, their
MFIs loan officers may try to conceal the situation
by rescheduling bad loans so they dont show up
as delinquent.
This pattern is not restricted to for-profit
commercial MFIs. Not-for-profit MFIs have
experienced the same problems, for instance in
Morocco and Pakistan.18 As the list of microcredit
markets with heavy competition grows, and as
they get closer to saturating the demand that can
be met with present methods and products, we
should expect to see widening problems with overindebtedness. Lenders will have to improve their
marketing, underwriting, and loan management.
Financial authorities will need to craft better
consumer protection regulations. And many clients
will have to learn how to deal better with their
new credit access. If these things happen, overindebtedness wont disappear, but it will probably
drop back toward more acceptable levels.
In the meantime, the natural dynamics of a
maturing retail credit market give us good reason
to worry about over-indebtednessi.e., to
devote much more energy and resources to
finding ways to identify it as early as possible and
prevent it as much as possible.19
Borrowers dont always make smart choices.
Some observers are troubled because they sense
a tinge of neocolonial paternalism in all the hand-

wringing about over-indebtedness. They argue


that the concern subtly implies that poor people in
developing countries arent smart enough to know
whats good for them, and wont make sensible
borrowing decisions for their households unless
the state or some other wise and benevolent
nanny limits their options.
The proposition that people, including poor
people, generally know whats best for them and
act accordingly is a good default assumption. But
in the case of credit behavior, recent developments
in economics and psychology suggest there may
be substantial evidence to the contrary.
The default assumption is that borrowers act
in their own best interestin other words, that
they behave like the rational homo economicus
of classical economic theory. A new school of
behavioral economics is using the insights and
tools of psychology to argue that the classical
model of rational motivation fails to account for
important dimensions of real behavior, including
behavior of borrowers whether theyre rich or poor.
Rather, the thesis is that real people are subject to
systematic biases pushing them in the direction
of judgments and behaviors that damage, rather
than promote, their long-term welfare. Borrowers
consistently make choices that, they themselves
agree, diminish their own well-being in significant
ways (Barr, Mullinaithan, and Shafir 2008).
We discuss specific biases in Section 2 on the
causes of over-indebtedness. The important point
for now is that behavioral economics has raised
serious doubts about the proposition that we can
safely rely on microborrowers native good sense
to keep their borrowing at healthy levels. There
are plausible theoretical models, and a growing
body of empirical results, suggesting that overindebtedness may be a permanent structural
concern for any kind of retail credit.
Strong repayment statistics dont assure us that
all is well. Many peopleincluding one of the

17 E.g., Krishnaswamy (2007) and references cited therein (p. 4)


18 This does not necessarily mean that commercial motives are irrelevant to the problem. Even not-for-profit managers may be tempted into
intemperate growth by the prospect of eventually converting their MFIs into for-profit form.
19 By focusing on the special challenges when competitive retail credit markets approach saturation, we do not mean to imply that over-indebtedness
is not a risk at other times as well.

12

authors of this paperhave argued that when


MFIs in a country show very high repayment rates
over the years, they probably havent been overindebting too many borrowers. Even if this view
has any merit, it should be subject to substantial
qualifications. We discuss the interpretation of
repayment statistics in Section 3. For now, we
summarize a few core issues:
Even a borrower who repays a loan faithfully
may encounter very serious problems in
making the payments.
Default and even delinquency rates are trailing
indicators. Borrowers may be having problems
right now that wont show up in the lenders
repayment statistics until latersometimes
much later, especially where borrowers can
juggle loans from several sources.
Some MFIs published repayment statistics
are unreliable.
Finally, as noted, competition and market
saturation change the whole game. Good
loan collection before that change occurs is
no assurance of continuing good collection
afterward.
For all these reasons, reports of strong past
and current repayment are not a good reason
to discount the possibility of serious overindebtedness among current borrowers.
Studies so far have found serious overindebtedness in some markets. In Section
4, we review six empirical studies that quantify
microcredit over-indebtedness and/or explore
ways to predict it. There are not enough of these
studies to permit much generalization. But most of
them have found levels of over-indebtedness that
many people would regard as worrisome.
Perception of over-indebtedness can lead
to political backlash. Over-indebtedness is of
concern not only for its possible prevalence
but also for the severity of its consequences. In
addition to the above-mentioned consequences
for borrowers, over-indebtedness can trigger

political backlash with wide-ranging consequences


for the industry and for future borrowers.
In many countries, microcredit will always be
politically sensitive, even when MFIs are behaving
perfectly responsibly. At first blush, it shocks the
conscience to learn that poor borrowers are being
saddled with interest rates that are a lot higher
than what wealthy borrowers are being charged
by banks. The rationale for the higher rates (i.e.,
higher costs per unit lent) takes some explaining,
and there will always be many people who dont
understand or accept it.20 This interest rate issue
makes microcredit a tempting political target,
especially when it is combined with occasional
(or frequent) examples of over-aggressive loan
collection by MFIs. When a political backlash
occurs, it is often driven not only by sincere concern
for clients but also by other political motives.
Taking the political crises in Bolivia, Nicaragua,
and Andhra Pradesh as examples, it seems that at
least some of the storm was fueled by factors that
had little to do with borrower welfare.
When a political backlash occurs, some borrowers
who are perfectly able to repay may decide
opportunistically that they can avoid repaying, and
the MFIs viability is threatened. More importantly,
there is always a danger of regulatory over-reaction:
governments may impose policy measures that not
only restrict bad lending but also prevent a much
larger amount of good lending.
Given the inherent political vulnerability of
microcredit, those who want to promote it should
be especially vigilant about over-indebtedness,
becauseamong other reasonsit can fuel a
public over-reaction that hurts a lot of potential
borrowers who need the credit and are fully able
to repay it. There is also a risk of backlash from
donors and public and social investors, who may
turn away from funding not only microcredit, but
microfinance more broadly.
We should have an even lower tolerance for
over-indebtedness if our expectations of what

20 Lending costs are not the only component of microcredit interest rates. Shareholders profits also contribute to the price paid by borrowers, and
there is controversy about how much profit is appropriate for institutions serving poor people.

13

microcredit can achieve have become more


modest. In a reasonable weighing of costs and
benefits, the amount of over-indebtedness were
willing to accept should depend on the size of the
benefit that we think is being delivered to all the
other borrowers who are not over-indebted.
The prevailing narrative about microcredit used
to be that borrowers were investing their loan
proceeds in new or expanded microenterprises,
and that the extra income produced by these
enterprises was raising people out of poverty
by the millions. If the benefit is really that great
for most of the borrowers, then one ought to be
willing to accept some collateral damage to a small
minority of the borrowers if it is an unavoidable
part of the process.21
The point can be illustrated by a stylized example.
Suppose that, in a given setting, microcredit is
instrumental in allowing 75 percent of its borrowers
to double their incomes and escape from poverty.
Further suppose that 10 percent of the borrowers get
over-indebted: some of them struggle hard to repay
their loan, and some default, but all are worse off to
some extent because of the loan. The circumstances
of the remaining 15 percent are unchanged.
Before settling for this situation, we would look
for ways to lower the damage for the unfortunate
10 percent. But even if improvement is possible,
we will reach a point where the only way to
further reduce over-indebtedness would be to
tighten lending standards so much that plenty of
otherwise qualified borrowers will be frozen out, or
to relax collection standards past the point where
delinquency can be kept manageable. At this
point, many people would feel that the remaining
level of over-indebtedness is acceptable if it is
the price of continuing the huge benefits that 75
percent of the borrowers are getting.
However, there are growing doubts about
the raising-millions-out-of-poverty narrative.

A few recent and rigorous impact studies


have not found such results, and some older,
more optimistic studies are being challenged.
The issue is far from settled. But recent
developments raise the distinct possibility that
the benefits of microcredit may be less than
previously advertised.22 It may be, for instance,
that the principal benefit of microcredit lies not
in getting people out of poverty but in helping
them cope better with poverty, giving them
useful cash management tools for a variety of
purposes, including keeping their consumption
stable, coping with shocks, and accumulating
larger sums to pay for business and nonbusiness
investments. If research confirms this, it is a
valuable benefit, but considerably smaller than
liberating households from poverty.
Returning to our hypothetical example, if the
benefit to the 75 percent is a much more modest
one, then the acceptable amount of collateral
damage to other borrowers should be much
smaller. Because our confidence about huge
microcredit benefits is weakening (getting more
realistic, some would say) these days, our concern
about over-indebtedness should be growing
because the amount of it were willing to tolerate
should be shrinking.
Putting all these considerations together, there
is a strong case for intensifying the attention,
research, and resources devoted to overindebtedness issues. In particular, there is a strong
message for most MFIs in competitive markets
especially markets that may be approaching
local or national saturation levels. If possible,
they need to develop early warning systems to
identify an over-indebtedness problem before it
shows up in a collapse of repayment. And they
need to adjust their loan marketing, approval,
management, and collection practices with a
view to keeping over-indebtedness at acceptable
levels. Section 5 looks at some of the steps that
may be needed.

21 This is a utilitarian approach, trying to achieve the best collective outcome for those involved. Some people might argue that there are levels of
damage (e.g., suicide) that are not acceptable under any circumstances, no matter what the benefit to others may be.
22 E.g., Rosenberg (2010); Collins, Morduch, Rutherford, and Ruthven (2009); and Karlan and Zinman (2007).

14

2. The Causes of
Over-Indebtedness
Over-indebtedness is a complex phenomenon; in
most cases, multiple factors are at work. Lender
behavior can put clients at undue risk, clients
themselves make mistakes in their borrowing
decisions, and sometimes external factors beyond
either partys control push debt to unsustainable
levels.23 We can describe some of these dynamics,
but there isnt enough evidence yet to make strong
assertions about how prevalent most of them are or
how heavily they contribute to over-indebtedness.

The Responsibility of Microlenders


When trying to understand what can go wrong, we
tend to learn best from examples where problems
are most visible. There have been complaints about
microlender behavior in a number of countries with
recent repayment crises. MFIs are said to have
marketed too aggressively, expanded too fast,
used staff incentives that encouraged over-lending,
offered the wrong products, obscured their loan
terms, and used abusive collection practices. We
look first at the challenges entailed by fast growth,
and then at the products and procedures of MFIs
more specifically.
Since its inception, the microfinance industry has
pursued growthindeed, rapid growth. Growth
implies outreach to more customers, andit is
hopedcompetition that brings down interest
rates and improves customer service. These
arguments may be valid, but they need to be
handled with care. The industry is learning more
about the challenges that come along with growth.
It is important to distinguish between growth of
an individual MFI and growth of a microcredit
market. It seems plausible that too rapid
expansion of an MFI could overstretch its lending
and management systems, lowering the quality

of preloan analysis as well as post-loan follow-up,


and resulting in serious collection problems.24
However, Gonzalezs analysis of MIX data finds no
correlation between individual MFI growth rates
and portfolio deterioration, except at extreme
(and very unusual) levels (Gonzalez 2010).25
The picture is different, however, when one looks
at aggregate market-level growth rates. Gonzalez
finds that high aggregate growth of a countrys
number of microfinance borrowers (greater
than 63% per year) and an active-loans-to-totalpopulation rate above 10 percent are significantly
related to repayment problems. His findings are
consistent with our discussion of the dynamics
of competition in saturating markets in Section
1. MFIs push harder and harder to capture the
remaining market share, they tie too much of loan
officer compensation to making more and bigger
loans, they start lending to borrowers who are
likely to have more trouble repaying, their ability to
assess repayment capacity declines because their
borrowers are also carrying debt from competing
lenders, and borrowers use their multiple access
to refinance loans that are ultimately unpayable.26
Some observers think that an over-supply of
funding, along with funders pressure on MFIs for
growth and profitability, has contributed to overindebtedness crises. Investor behavior affects the
incentives of MFIs. In addition, the fast growing
microfinance investment by commercial and quasicommercial players has tended to concentrate on
markets that have mature MFIs with strong profits
and high growthin other words, markets that are
more likely to be saturated.27
Regardless of growth rates and saturation, overindebtedness can be influenced by specific
products and practices of individual MFIs before,
during, and after loan approval.28
Ex ante, before the loan is made, overindebtedness risks may emerge from a lack of

23 This section is based on the analysis in Schicks (2010).


24 E.g., Chen, Rasmussen, and Reille (2010), analyzing debt crises in Nicaragua, Pakistan, Morocco, and BosniaHerzogovina.
25 The study finds that intensive growth (adding new borrowers to existing branches) is more dangerous than expansive growth (adding branches in
new locations), but neither form of growth appears to correlate with collection problems except at growth rates that are very rare in large MFIs.
26 E.g., Abed (2011).
27 E.g., Chen, Rasmussen, and Reille (2010) and Roodman (2011).
28 Some of these products and practices are discussed further in Section 5.

15

transparency. MFIs may reduce borrowers ability


to make smart choices if their advertising and
face-to-face marketing do not communicate loan
conditionsincluding pricingin accurate, fair,
and easy-to-understand terms.
Next, in giving loans, MFIs that are too growthfocused may relax lending standards and invest less
in a sound evaluation of repayment capacity. This
is especially likely when loan officers incentives
are strongly linked to adding new clients or making
larger loans.29
There is a complex debate about offering loans
that will be used for consumption purposes.
Some MFIs offer loan products that are explicitly
focused on consumption uses. But even where
loans are declared to be for business uses,
manysometimes mostborrowers use them for
nonincome-generating purposes. Following up
with borrowers to enforce business use of loan
proceeds raises costs (and thus interest rates),
and may not be effective. When loans finance
consumption, the loan use does not produce
returns that compensate for the interest expense.
On the other hand, nonbusiness uses may be very
helpful for managing emergencies, smoothing the
volatility of poor peoples incomes, and funding
nonbusiness investment, such as education.30 The
relationship to over-indebtedness is ambiguous.
Paradoxically, a standard microcredit feature that
was directed at testing repayment capacity of
borrowers and reducing over-indebtedness risks
can, in some cases, create the opposite result.
Automatic loan size increases may lead some
clients to take larger loans than are appropriate
for them. Especially in group lending, where a
loan officer is not assessing repayment capacity,
ever-increasing loan sizes can eventually outstrip
borrowers ability to repay. At the other end of
the graduated-loan-size ladder, many MFIs control
credit risk by testing out new borrowers with

loans that are a lot smaller than those borrowers


want and can handle. Where possible, borrowers
go to other lenders to make up the total loan
amount they want. This adds to the borrowers
transaction costs, and may even contribute to
over-indebtedness by making it harder for each
individual lender to assess borrowers ability to
handle their total indebtedness.
During the course of the loan contract, the
installment schedule can affect the borrowers
ability to repay. Products that are too inflexible
and repayment schedules that are too far out
of step with borrowers cash flows can drive
borrowers into over-indebtedness even when
the debt amount is reasonable. When individual
borrowers are struggling with liquidity problems
rather than structural debt problems, the zerotolerance policy that has been the mantra of
many MFIs may need to be relaxed.31 The policy
is valuable in teaching borrowers the importance
of regular repayments, especially compared to
typical government lending. It promotes the early
recognition of repayment difficulties and keeps
borrowers from piling up obligations (Gonzalez
2008). It reduces the risk of spillover effects of
delinquency on nondelinquent borrowers and
avoids the operational cost of handling large
numbers of rescheduling requests. Nevertheless,
if the zero-tolerance policy is so strict that there
is no room to accommodate legitimate individual
circumstances, the result can be an unnecessary
increase in the struggles of honest borrowers.32
A related issue concerns collection methods. Once
borrowers have gotten into trouble, inappropriate
collection practices can worsen their situation.
While a certain level of firmness is necessary to
provide sufficient repayment incentives, collection
needs to respect the basic rights of borrowers.
Also, if the MFI charges substantial penalties for
late payments (over and above continued accrual
of regular interest), it may place itself in the

29 E.g., Chen, Rasmussen, and Reille (2010).


30 E.g., Collins, Morduch, Rutherford, and Ruthven (2009). The authors caution against the assumption that borrowing for nonbusiness purposes is
usually a recipe for over-indebtedness.
31 Some MFIs point with pride to their perfect 100% collection rate. In fact, that level of collection is anything but ideal. It usually means that the
MFI is being too conservative in extending loans, which reduces both borrower access and MFI profits. Long experience has shown that most
MFIs can tolerate low levels of nonrepayment (e.g., annual loan losses below 34%) without the problem spinning out of control.
32 Several studies have found that microborrowers resort to expensive informal moneylenders to make loan payments during low-income periods.
E.g., Jain and Mansuri (2003) and the studies cited therein; CARE/Bangladesh (2005 and 2005a).

16

dubious position of making higher profits when


clients cant or dont pay on time, which is hardly
an incentive to minimize over-indebtedness.
Finally, after a loan cycle has been completed, it
is difficult for some MFIs to accept that borrowers
sometimes need a break from credit rather than
an immediate follow-up loan. This is especially the
case if MFIs are not offering deposits or other
services that would bind clients to the institution
while they are not borrowing.
This list of the ways MFIs can contribute to
over-indebtedness is not an indictment but
rather an illustration of the difficult challenges
involved in lending to poor people, especially in
an ever-changing and increasingly competitive
market. Some of the core methodologies that
make microlending successful and safer for
both institutions and borrowers (e.g., fixed
weekly repayment schedules or graduated loan
sizes) can have unexpected effects in specific
situations. The microfinance industry is currently
in the process of learning about, and dealing
with, risks that are becoming more apparent as
the industry matures.

can lead borrowers to take on more debt than


is good for them. Some of the biases described
by behavioral economists and confirmed by
their experiments dont surprise us very much,
because they are consistent with our everyday
commonsense experience.
To begin with, many people put too much weight
on present gratification, because it is salient,
and pay too little attention to future consequences
because they seem less real. 34 Hyperbolic
discounters have a strong bias toward receiving
rewards (e.g., loan disbursements) now and
deferring costs (e.g., repayment) into the future.
They regularly make present decisions that their
future selves wind up regretting. In one study,
they were found to borrow three times as much as
rational consumers.35 There are other theoretical
approaches as well to the universally recognized
phenomenon of temptation interfering with selfcontrol.36 See Box 1.

Some people dont repay their loans simply


because they would prefer to keep the money
they are able but unwilling to repay. Our focus in
this paper is not on these opportunistic defaulters,
but on borrowers who have genuine difficulties
in repayingthose who are willing but unable
to repay.33 Borrowers need protection from
irresponsible lender behavior, but this should not
obscure the role borrowers themselves play. Every
over-indebting loan contract represents a decision
not only by the lender but also by the borrower.

There are other relevant biases. For instance,


peoples predictions of the future tend to be overoptimistic. They underestimate the likelihood of
shocks, underestimate the amounts they will wind
up borrowing, and overestimate their ability to
pay on time (Barr, Mullinaithan, and Shafir 2008;
Kilborn 2005; Vandone 2009). Habit persistence
causes people to reduce consumption too slowly
when income declines, with obvious consequences
for the likelihood of over-borrowing (Brown 1952).
In general, human beings have difficulties with
processing all the complex information related to
loan contracts. They tend not to fully understand
all the implications of loans and sometimes rely
on rather weak proxies in their decision making.
For example, people are inclined to think that the
credit limit they are given by financial institutions
is an indicator of their ability to repay (Soman and
Cheema 2002).

As noted in Section 1, the emerging field of


behavioral economics has identified biases that

Some behavioral biases affect poor people more


than others. It seems plausible to assume that

Why Borrowers Get Themselves into


Trouble

33 We would probably include borrowers who make dishonest statements in their loan applications, embellishing their financial situation or not
admitting the real purpose of a loan, as long as they borrow in good faith that they will repay the debt to the MFI and, if problems arise, will make
serious efforts to fulfill their repayment obligations.
34 E.g., Mullinaithan and Krishnan (2008)
35 E.g., Laibson (1997); Laibson, Repetto, and Tobacman (2003).
36 E.g., Akerlof (1991), Gul and Pesendorfer (2004), Fudenberg and Levine (2006)

17

Box 1. Poor People Recognize Their


Problems with Self-Discipline
Paradoxically, many poor people take out loans
precisely for the sake of external discipline.
They use the loan proceeds for some important
purpose, knowing that the repayment schedule
will force them to regularly set money aside,
protecting it from the insistent temptation of
current consumption.a Accumulating a usefully
large lump sum of cash through borrowing is more
expensive than doing the same thing through
saving, but many poor people are happy to pay the
price to shore up their self-control. When savings
products are designed with a commitment feature
that deliberately prevents everyday access to the
money, there tends to be strong demand from
poor people, and their asset accumulation and
investment are likely to improve.b
a

E.g., Collins, Morduch, Rutherford, and Ruthven (2009);


Rutherford and Arora (2009); Rutherford (2000)
E.g., Ashraf, Karlan, and Yin (2005); Duflo, Kremer, and
Robinson (2009).

resisting the temptation to over-borrow is


particularly challenging for poor people, who
are more likely to find themselves in desperate
situations. Inequality and social comparison can
lead lower income groups to borrow irresponsibly
in a desire to keep up with the consumption levels
of peers (Duesenberrry 1949, Luttmer 2005,
Christen and Morgan 2005, Brown 2004). Social
obligations to support relatives and neighbors
at any cost may drive people into indebtedness
(Cressy 2006, Guerin 2000). Obviously, the pain
of deferring a given quantum of consumption is
worse for the poor than for the rich (Bannerjee
and Mullainathan 2009). And mistakes in
predicting the future tend to have more drastic
consequences for people living closer to the
edge of survival. We emphasize that none of this
necessarily means that the poor are less rational
than the rich. For the most part, poor people
probably have the same biases as rich people, but
because of their more vulnerable circumstances,
those biases tend to produce more damaging
results (Bannerjee and Duflos 2007, Bertrand and
Mullainathan 2010).

At the same time, it is important not to overstate


the case about cognitive and behavioral biases.
To begin with, while everyone acknowledges the
existence of such biases, there is some scholarly
disagreement about how much they skew real
borrowing behavior.37 Also, many of the empirical
results cited here are from rich countries. One
cannot automatically assume they apply equally
well in poor country settings.
In addition to the biases discussed above,
microborrowers have to deal with a learning curve.
Many of them have gone quite quickly from having
no formal loan access to having simultaneous
access to multiple sources. Some speculate that
borrower inexperience is a major source of debt
problems.
In sum, when we look for the causes of overindebtedness, we need to examine borrower
decision making, not just the behavior of lenders.
Indeed, lenders policies should be based on a
realistic recognition of the biases of their borrowers.
The existence of borrower bias does not absolve
lenders from their responsibility for prudent, fair
lending that does not leave undue numbers of
borrowers in trouble.

The Power of External Factors


In addition to behavior by lenders and borrowers,
circumstances beyond the control of either of them
contribute to over-indebtedness.
Individual borrowers are hit by household-specific
external shocks, most commonly a sudden
reduction in income (e.g., a job loss or illness in the
household) or a large unexpected expense (e.g., an
accident, medical costs, or funeral obligations).38
Other shocks affect many borrowers at the same
timee.g., natural disasters or conflicts that destroy
livelihoods. Changes in government policies
can affect input and market prices, or displace
street vendors (Stearns 1991). Macroeconomic
contractions and currency fluctuations can impair
many borrowers repayment capacity.

37 E.g., Ellihausen (2010).


38 E.g., Bouquet, Wampfler, Ralison, and Roesch (2007), Krishnaswamy (2007).

18

One key characteristic of poverty is the difficulty


of building safety buffers for such situations.
These shocks are very common in the lives of
vulnerable populations, but it is difficult to account
for them at the time of borrowing and lending,
so they subject all credit decisions to a higher
level of uncertainty. Before the shock occurs,
the debt level can seem quite manageable. But
pre-existing debt tends to limit the households
maneuvering room in handling a shock when it
occurs. What originally seemed like a reasonable
debt burden can quickly turn into unsustainable
over-indebtedness. And the more rigid the loan
repayment schedule is, the harder it is for the
borrower to manage the shock.
One by one, each of these shocks is unpredictable.
But the overall prevalence of such shocks among
the target population is eminently predictable.
Lenders need to factor in these risks when they
design their loan products, and when they decide
how much to lend to whom.
Finally external sources of over-indebtedness
risks can also include the regulatory environment,
infrastructure for information exchange among
lenders, etc. They impact the behavior of lenders
and borrowers.
To conclude, lenders, borrowers, and external
influences all contribute to over-indebtedness,
and each of the factors discussed in this section
represents a potential lever to reduce the problem.
Section 5 presents a brief survey of commonly
proposed approaches.

3. Defining and Measuring


Over-Indebtedness
When people in the microfinance industry
discuss over-indebtedness, they usually dont
have to interrupt the conversation to figure out
whether theyre talking about the same thing. In
this sense only, the term is not problematic. But
when the time comes to design research or draft
regulations, being precise about the meaning of
over-indebtedness raises tricky questions, not
all of which have clear answers.

We begin by distinguishing definitions from


proxy indicators. A definition tries to clarify the
core meaning of a term. For instance, our rough
provisional definition at the start of this paper was
that borrowers are over-indebted if they have
serious problems in paying off their loans. It is hard
to do direct measurement of over-indebtedness
defined this way. So we might look for a more
easily measurable proxye.g., we could measure
borrowers debt payments as a percentage of their
income. One could reasonably assume that this
correlates pretty well with over-indebtednessi.e.,
people whose debt service ratio is high are more
likely to encounter serious difficulty in repaying.
Its useful to keep this distinction between
definitions and proxies in mind while reading this
section, which discusses both. But its not always
clear which is which: one persons proxy can be
anothers definition.
The choice of an indicator (whether as definition
or proxy) depends on its purpose. Regulators,
for instance, look for legal definitions that can be
applied more or less mechanically, with as little
ambiguity as possible, so that the application
of the regulation will be as predictable as
possible for all parties. Academics or market
researchers need indicators that can be
measured by the data or investigative tools
available to them.
This section looks at the following definitions and
proxies:
Negative impact
Default and arrears
Debt and debt service ratios
Multiple borrowing
Borrower struggles and sacrifices
Composite indicators
Negative impact. Saying people are overindebted would seem to be an assertion that
they have too much debtwhich would ordinarily
mean more debt than is good for them. In this
view, the question would come down to whether
or not the loan is making the borrowers worse
off than they would have been without the loan.

19

We might think that this should be the core


definition from a policy perspective. After all, we
would like as much as possible to prevent loans
that make borrowers worse off. As for loans that
make borrowers better off, we should prefer not to
prevent such loans, even if (for instance) repaying
them ties up a big portion of the borrowers
income, or the borrowers have to struggle to
repay. The test, then, would be whether the loan
has a negative impact on the borrower.
But this definition doesnt fit well with common
usage. Lets imagine someone who is temporarily
short on cash and takes out a very small loan
to buy a shirt he doesnt really need; the loan
is repaid very easily when payment falls due.
Perhaps the loan made the borrower worse off
it might have been better not to spend money on
the shirt, and the loan interest was a further cost
but this is not the sort of image most of us have
in mind when we talk about over-indebtedness.
A further problem is that credibly determining
the impact of loans is difficult and expensive.39
For example, at present one of the most widely
recognized tools for testing the impact caused
by loans is the randomized controlled trial (RCT),
an approach that (1) imposes severe demands
on the lenders who participate, (2) can cost
hundreds of thousands of dollars per study, and
(3) can take years to produce results. RCTs are
not a practical tool for ongoing monitoring of
over-indebtedness.40 Thus, the negative impact
definition may not be of much practical use,
whatever its theoretical merits.
Default and arrears.41 The most common proxy
indicator for over-indebtedness is default. Clearly,
when microborrowers have loans they can no
longer repay, they are over-indebted. This is

the moment when over-indebtedness puts the


sustainability of MFIs at risk, and when borrowers
may be exposed to public humiliation, asset
seizures, loss of creditworthiness, or erosion of
the social networks on which they rely.
However, borrowers who default dont usually do
so because they are suddenly struck with too much
debt on the day payment falls due. The problems
have most likely been around for quite some time.
The over-indebtedness has usually started much
earlier and has probably been accompanied by
signs of the problem getting worse. Also, some
borrowers default not because they have an
unmanageable debt situation, but because they
find it more convenient not to pay. Default is not
a definition of over-indebtedness, but rather it a
trailing indicator of it.
Arrears (delinquency), as an indicator, have the
same advantages and disadvantages as default.
They start flagging the problem a little earlier, when
a payment is late, but before its clear the loan wont
be repaid. But the borrowers debt situation may
have been unmanageable well before a payment
is actually missed. Arrears and default rates often
measure a consequence of over-indebtedness,42
but not over-indebtedness itself. They are the most
widely used indicators only because they are the
easiest to measure. (See Box 2.)
Debt ratios. Another common set of indicators
takes the term over-indebted more literally.
It defines as over-indebted borrowers who have
borrowed over a certain limit. The limit is often
framed as the ratio of individual or household debt
service to income (or take-home pay, or disposable
income after deducting minimum consumption
expense). One is over-indebted if loan payments
eat up more than X percent of ones income. This

39 See, e.g., Rosenberg (2010).


40 In theory, it might be desirable for impact RCTs to collect data about over-indebtedness indicators, so as to determine how well they correlate
with negative impact. E.g., if someone fails to repay a loan, how strong an inference (if any) can we draw that the loan made them worse off?
However, there are challenges to the practical implementation of such a strategy.
41 Default as used here refers to a situation where the borrower never repays all of the loan, and the lender is left with a loss of principal. Arrears
or delinquency refers to a situation where the borrower fails to make one or more payments on time, but might later repay the loan in full.
42 Of course, arrears and defaults occur not only when borrowers are unable to repay, but also when borrowers are able but unwilling to repay.
In particular, when most MFIs are collecting well, high arrears at an individual MFI more often reflects a management problem than overindebtedness.

20

Box 2. What Do Repayment Rates Tell Us?


High delinquency or default in a market is usually
though not alwaysa good sign that there has been
serious over-indebtedness.a But what about the
converse proposition: do high repayment rates that
characterize microcredit in most markets tell us that
theres little over-indebtedness there? As noted in
Section 1, this is a more dubious proposition.
To begin with, the fact that borrowers repay loans
doesnt tell us whether theyve had to struggle with
serious difficulties to do so. Over-indebted people
sometimes repay, albeit at the cost of major sacrifices.
(See Borrower struggles and sacrifices on p. 23.)
The stronger the external incentives to repay (e.g.,
peer pressure or aggressive collection practices),
the more likely it is that borrowers will make serious
sacrifices to repay.
Also, a borrower may be in trouble for some time
before a failure to pay shows up in the loan collection
statistics. This risk of hidden delinquency is aggravated
by the fact that, especially in competitive markets,
borrowers who are unable to repay their loan can
postpone the day of reckoning by taking out a new
loan from some other sourcee.g., a competing MFI
or an informal lenderto pay off the original loan.
On the one hand, juggling different credit sources is
a standard survival skill of the poor in certain markets
and does not automatically imply over-indebtedness.b
On the other hand, such juggling can conceal overindebtedness for a long time until it eventually hits
some lenders books as a failure to pay.

rates. Taking into account the possibility of such


rollovers of unpayable debt, and assuming the
correctness of the MFIs repayment statistics, we might
repose more confidence in the situation as reported
two years ago than in the situation reported last year
or during the current year. (Remember that we are
talking here about confidence that most borrowers
were willing and able to pay, not confidence that few
borrowers were being over-indebted.)
Looking at repayment statistics from a different
perspective, the sad truth is that, even at this late
stage in the evolution of management information
systems for microfinance, some big MFIs repayment
statistics cant be trusted. There may be a gap
between reporting and reality because managers
are deliberately concealing problems. Probably more
often, the problem is inadequate information systems,
so the managers are as much in the dark as anyone
else.

How long can borrowers inability to pay be


camouflaged this way? Clearly, the more lenders a
borrower has access to, the longer an unpayable debt
can be shifted around among them, at least if there is
no credit bureau. (Note that borrowers also roll over
their unpayable debt with informal moneylenders, who
dont show up in any credit bureau.) Some practitioners
will hazard a guess that its hard to keep such a house
of cards in place for more than a couple of years. But
we dont know of any empirical evidence on the topic.

A striking example can be found in the rating firm


M-CRILs on-the-ground testing of likely loan loss
levels in branches of several big Indian MFIs. The
MFIs financial reports showed annual loan loss
levels of about 0.5 percent of portfolio. M-CRIL
investigators estimated that real losses were likely
to be 57 percent (Sinha 2010). (The prevailing rule
of thumb for uncollateralized microcredit has been
that something like 5 percent tends to be an upper
limit of sustainability. An annual loss rate above that
level must be reduced quickly or it will tend to spin
out of control, rising rapidly to levels that cannot be
compensated for through higher interest rates.) In
the Indian MFIs, the gap between report and reality
happened mainly because loan officers were simply
rescheduling or otherwise renegotiating the loans
of borrowers who couldnt pay, even when eventual
repayment was very unlikely. This left the loans
looking like they were being paid on schedule, and
protected loan officers bonuses that were linked to
low delinquency.c The MFIs should have had internal
audit systems that would flag such practices.

Imagine an MFI in a competitive market that has


reported a continuous history of very high collection

All in all, then, high repayment rates are not a clean


bill of health.

b
c

High default in a market usually involves serious over-indebtedness. But high default for a particular lender, when other lenders
in the market are collecting successfully, is more likely to be a symptom of poor loan management and insufficient attention to
collection systems and performance. Also note that in some repayment crises, political agitation may be a major contributor, leading
to opportunistic default by many borrowerse.g., Krishnaswamy and Ponce (2009a). Of course, the political agitation may have
resulted from a real over-indebtedness problem in the first place.
E.g., Collins, Morduch, Rutherford, and Ruthven (2009) and Morvant-Roux (2009).
Rescheduling or renegotiating loans is appropriate when borrowers have temporary cash flow problems but are likely to be able to
pay off their obligations somewhat later. Rescheduling is inappropriate, indeed dangerous, when it is done to postpone the day of
reckoning on loan amounts that are unlikely ever to be recovered.

21

measurement is based on an admittedly crude


estimate that such a situation is unsustainable, or
at least that it is bad, for the average borrower.43
The debt service ratio compares scheduled loan
payments with the income available to make those
payments. This comparison is a fundamental tool
of risk management for many microlenders. In
most individual lending programs, a key part of the
loan officers job is supposed to be sizing up a loan
applicants likely repayment capacity by looking
at income and household expense, and seeing
whether there is enough surplus income to cover
the periodic payments on the proposed loan.44
Sometimes a microlenders loan policy states an
explicit numerical thresholde.g., loans should
normally not be approved if the payments would
exceed X percent of the borrowers disposable
household income. The percentage threshold
may be a seat-of-the-pants guesstimate, or it may
be determined by analyzing historical repayment
experience with thousands of borrowers: if
borrowers have a debt service ratio of X percent,
what percent of the time do they default?

ratio that is perfectly manageable for one borrower


may be too much of a burden for another.
Even if debt ratios were perfect proxies for overindebtedness, the necessary data are usually not
readily available, especially in less developed
markets. Most microcredit markets do not yet
have credit bureaus that cover microcredit. And
even where such credit bureaus exist, they have
no income information for microborrowers and
cant give a complete picture of their debt either,
because so many borrowers also use informal
lenders that dont report to credit bureaus. This
problem is especially acute if there is no way to
determine the borrowers consumption expenses,
so that the ratio has to be based on income rather
than income net of basic expenses.

So debt service ratios can help estimate the odds


of loan repayment. Indeed, debt or debt service
ratios are often used in efforts to quantify overindebtedness. But such measures have their
limitations.

Finally, the use of the loan proceeds can make a


huge difference. For example, suppose we are told
that a given borrowers total monthly debt service
is 75 percent of monthly income. At first blush
that sounds terribly high, because it seems as if
only 25 percent is left for essential consumption
expenses. This analysis would be near the mark
if, for instance, the loan proceeds are used to pay
for a wedding. But suppose the loan proceeds are
used to finance basic food, clothing, and shelter. In
that case, the inflow (loan disbursement) and later
outflow (loan payment) of principal doesnt change
the total amount available for consumption. It is
only the interest expense on the loan that reduces
consumption possibilities. The same debt service
ratio might be completely unsustainable when
financing a funeral, but very manageable when
financing basic consumption.45

One challenge is that there is no one-size-fits-all ratio


that is appropriate for everyone. Circumstances of
borrowers and their households vary a lote.g.,
number of children, absolute size of income, etc. A

So, it is impossible to set a threshold debt service


percentage that fits all borrower circumstances well.
The problems with debt-to-asset ratios are similar.
But this certainly does not mean that the ratios

Other microlenders require the same cash flow


analysis, but do not define a numerical threshold.
Either way, when the debt service ratio is judged
to be too high, the loan is refused.

43 Debt-to-asset ratios are also used: over-indebtedness starts when a borrowers total debt exceeds Y% of the borrowers assets. This approach is close
to the common definition of corporate bankruptcy as a situation where liabilities exceed assets. But debt-to-asset ratios tend to be less relevant in
microfinance. Microborrowers tend to have few assets, their loans are usually uncollateralized, and the lenders typically expect to be repaid out
of current income, not by liquidation of assets.
44 In group lending programs, decisions about a borrowers creditworthiness are usually made by fellow group members, not a loan officer. But when
group members make such judgments, surely it is not unusual for them think about whether the borrower has enough income to make the loan
payments.
45 For an extreme illustration, imagine someone who spends his entire take-home pay every month on consumption, uses a credit card for all
purchases, and pays off the card each month without incurring any interest. The debt service ratio is 100% of take-home pay, yet the use of credit
rather than cash is completely manageable, and leaves the borrower no worse off.

22

are useless for market research. Not surprisingly,


some studies have established a robust correlation
between a given debt ratio and high delinquency.46
High delinquency is not the same as having serious
problems in repaying (our provisional definition
of over-indebtedness), but it seems reasonable to
assume that there equally is a correlation between
the debt-to-income ratio and over-indebtedness.
Gurin, Roesch, Subramanian, and Kumar (2011)
point out another limitation of debt ratios,
especially for poor borrowers in developing
countries. Working with rural villages in Tamil
Nadu in India, they find that debts differ in their
social meaning, depending on their conditions and
the relationships involvedespecially given the
diversity of informal loans that microborrowers may
also be carrying. Certain debt contracts, or failure
to pay those contracts, are more dishonoring than
others, more difficult to handle, or more costly in
terms of reciprocal obligations (e.g., expressing
gratitude or making a return loan at a time that
may be inconvenient). Some have tight repayment
schedules with high penalties on late payments,
while others pose less of a problem even if their
amounts may be higher. The concept of a total
amount of debt is therefore not intuitive for many
poor borrowers and can feel like comparing apples
to oranges. The more we aggregate information
into ratios and unified thresholds across a market,
or even an industry, the further we get from the real
world of microborrowers and their experiences of
over-indebtedness.
Multiple borrowing. Once borrowers have
access to competing lenders, some take loans
from more than one of those lenders at the same
time. Multiple borrowingor to be more precise,
cross-borrowingis often seen as a proxy or
early warning signal for over-indebtedness.47 The
standard picture includes two problems: (1) if there
is no credit bureau, borrowers accumulate more
debt than they can handle because lenders have no
way of knowing how much other debt the borrower

is carrying, and (2) borrowers postponeand


deepentheir problem by taking out new loans to
pay off old ones, even though they may never be
able to retire all their debt.
Conceptually, there are two problems with multiple
borrowing as an indicator of over-indebtedness.
First, someone can be over-indebted with even a
single loan. Second, and more importantly, there
are many settings where multiple indebtedness is
a common, and perfectly manageable, cash flow
management technique for low-income households.
In some cultures, the phenomenon has existed
for a long time in the informal sector. It is part of
the sophisticated system of money management
poor people have developed, where one person
is often a borrower and a lender at the same time
(Collins, Morduch, Rutherford, and Ruthven 2009;
Morvant-Roux 2009). Poor borrowers may consider
it totally normal to juggle their different sources
of cash, including a number of lenders. They
may make smart choices within their systems of
multiple indebtedness, aiming not to pay off the
total amount they owe, but rather to keep their
creditworthiness with as many lenders as possible
as a safety net for future cash needs. They may not
consider themselves at all over-indebted (Gurin,
Roesch, Subramanian, and Kumar 2011). There are
other good reasons for multiple borrowing that
have nothing to do with having too much debt.
Good borrowers may want, and be able to handle,
larger loans than individual MFIs are ready to
offer,48 or an additional opportunity may come up
in the course of a loan cycle. An emergency can be
a good reason to borrow anew without implying
over-indebtedness. Finally, additional loans may
help manage an unfavorable timing of disbursement
or a strict repayment schedule that doesnt fit a
borrowers cash flow. In short, multiple borrowing
options are beneficial for many borrowers.
Empirical findings are mixed. Most studies have
found at least some degree of correlation between

46 E.g., Maurer and Pytkowska (2010), Vogelgesang (2001), Rinaldi and Sanchez-Arellano (2006).
47 Strictly speaking, multiple borrowingcarrying more than one loan at a timewould include having two or more loans from a single lender.
Cross-borrowing refers to simultaneous loans from different lenders. This section focuses on the latter situation. But we retain the term multiple
borrowing because it is the one most commonly used.
48 E.g., Grammling (2009).

23

multiple loans and repayment problems.49 But


studies in other settings have found no such
correlation. For instance, Gonzalez (2008) found
that the Bolivian over-indebtedness crisis in the late
1990s was not associated with multiple borrowing.
The indicator would not have revealed that an overindebtedness crisis was going on. Another study
actually found a negative relationship between
multiple loans and repayment failuresi.e., multiple
borrowers repaid better than those with single loans
(Krishnaswamy 2007). This somewhat surprising
finding might be explained by hypothesizing that
the more skilled money managers were more likely
to take on multiple debts. Alternatively, it might
be that those with multiple loans were temporarily
postponing an inevitable delinquency or default;
longitudinal research would be required to test this
explanation.
To sum up, there is reason to believe that, for some
borrowers, multiple borrowing can be a step on the
path to over-indebtedness, but, by itself, it is not a
reliable measurement to identify over-indebtedness
problems in a population. At best, it might be a
useful piece of a multi-factor over-indebtedness
index.
Borrower struggles and sacrifices. The above
measures leave out some people we normally
would think of as over-indebted, and include
people we would not think of as over-indebted. Our
earlier provisional definitionborrowers are overindebted if they have serious problems in paying
off their loanswas chosen because we think it
corresponds to what most people have in mind
when they use the term. If borrowers simply cant
repay, almost everyone would regard them as overindebted. If certain borrowers do manage to repay
their loans, but have to make extreme sacrifices to
do so, most of us would think they probably have
too much debt. (More later on whats extreme.)

Some researchers have looked at the sacrifices


borrowers make to repay.50 Others have framed
a formal over-indebtedness definition along these
lines.51
Gonzalezs (2008) definition is that overindebtedness occurs when the repayment
outcome of a loan contract does not correspond
to the original expectations of either the borrower
or the lender or both. Even if they eventually
repay, microborrowers are over-indebted already
when payment requires more costly actions than
expected. This is quite a broad concept of overindebtedness. It includes costly actions even if the
borrower might not view them as sacrifices.52
According to Schicks (2010), the threshold of overindebtedness is reached when a microborrower
is continuously struggling to meet repayment
deadlines and structurally has to make unduly
high sacrifices related to his/her loan obligations.
Especially in the context of an industry that says
its purpose is to help the poor, microborrowers
who manage to repay only by sacrificing minimum
nutrition levels or their childrens education should
be counted as over-indebted. In the context of
short-term microlending, this definition treats
over-indebtedness as a structural phenomenon.
Only repeated sacrificing or severe sacrifices that
indicate persistence of debt problems meet the
threshold.
But when are sacrifices unduly high? In the
absence of any objective standard that would
apply in all settings, Schickss approach is to let
the borrowers themselves indicate the outer limits
of the level of sacrifice they feel is acceptable. If
the focus is on the level of distress experienced by
borrowers, then their subjective judgment is the only
source that can take all individual circumstances and
complexities of the debt phenomenon into account.

49 E.g., Maurer and Pytkowska (2010), Chaudhury and Matin (2002), CARE/Bangladesh (2005), Martinez and Gaul (2011), Grammling (2009),
Krishnaswamy and Ponce (2009a).
50 E.g., CARE/Bangladesh (2005, 2005a).
51 Cf. Rhyne (2010).
52 Gonzalezs definition includes opportunistic default (in this case, it is the lenders expectations that are not met), even though most people would
not describe such defaulters as over-indebted.

24

Like any other approach with subjective elements,


using borrower judgments about the acceptability
of sacrifices has its downsides. It has to deal with the
idiosyncrasies, biases, and honesty of respondents.
This can be an acceptable challenge in a research
context but would be problematic for regulatory
purposes. Box 3 illustrates one approach to
implementing this struggle-and-sacrifice definition
in on-the-ground research, showing that it can
deliver useful results.
There is also an ethical complexity. If overindebtedness is defined in terms of subjective
acceptability of sacrifices, then one might
conceivably make the perverse argument that
the way to reduce over-indebtedness is to
condition borrowers to accept greater sacrifices.
Nevertheless, the idea that borrowers are not
over-indebted if they are happy with their loans in
spite of some struggle is powerful from a consumer
protection perspective. For example, reducing
the stigmatization of borrowers in difficulty and
enhancing their safety network may be a valid
measure to ease over-indebtedness.
Finally, the notion that borrowers are suffering from
serious sacrifices to repay their loans might lead to
an inference that the loans have hurt the borrowers
and that wed therefore prefer to prevent such
loans. This inference is not necessarily true. For
instance, a borrower who has to go without food for
two days to repay a loan may have taken the loan
in the first place to avoid going without food for a
full week. In this scenario, the loan has helped the
borrower notwithstanding the serious and repeated
repayment sacrifices. Struggle and sacrifice are
not the same as negative impact.53 Given the low,
variable, and vulnerable incomes of poor people,
coming up with the money for any cash need,
not just loan repayments, often requires serious
sacrifice.54 This consideration might temper our
judgment as to what level of struggle-and-sacrifice
over-indebtedness is reasonable in microlending.
Even if measuring struggle and sacrifice doesnt
automatically tell us that borrowers are being

hurt by their loans, it is still very useful as an


indicator. First, if a large proportion of borrowers
are struggling to repay their loans, we would want
to know that regardless of whether the loans are
ultimately pluses or minuses in their lives. Better
loan analysis, more flexible loan policies, or other
measures (even alternatives to credit) might reduce
the problem.
Second, repayment struggles may not prove
negative impact, but it seems reasonable to think
that, all other things being equal, higher levels of
borrower struggle increase the probability that the
loans are actually making people worse off. And the
borrowers judgment that sacrifices were related
to the loan and unacceptable given the loans
purpose implies a certain relationship to impact.
The link between struggle and negative impact
gets stronger when the definition includes only
those sacrifices that borrowers themselves judge
as unacceptable. Presumably, borrowers would
be less likely to feel that repayment sacrifices are
unacceptable in situations where they think the
loan has made them better off, notwithstanding
the sacrifices.
Finally, tracking this kind of indicator (like some of
the other indicators) over time can reveal important
trends and perhaps reveal growing problems before
they hit the lenders delinquency reports. We would
argue that, for survey work oriented to consumer
protection, it is the most powerful of the definitions.
Composite indicators. Finally, when no definition
is free of challenges, it makes sense to adapt the
definition and measurements we are working with
to the questions we ask. It is not one solution we
are looking for but a set of responses to the various
purposes of measuring over-indebtedness. A client
protection question is likely to work best with one
of the latter definitions; a regulatory definition
is more likely to use fixed criteria that might be
imperfect but at least provide clear guidelines. And
a risk management definition for lenders may look
slightly different again.

53 In this simple example, the borrower would probably not rate the sacrifice of going without food as unduly high. But real situations are usually
more complex.
54 Cf. footnote 56.

25

One approach to deal with the limitations of the


various criteria is to combine several measures
into a composite indicator. One such example, in
a developed nation context, has been suggested
by the European Commission (2008). For the
political purpose of approximating the number of
consumers with over-indebtedness problems in
a society, it combines subjective elements, such
as the households perception that repayment is
very difficult and the debt a heavy burden,
with objective criteria, such as persistent arrears
and debt or other required payments that push a
household below the poverty line. Applying several
indicators simultaneously reduces the number
of subjects who fall into the over-indebtedness
category, while applying them selectively increases
the number. See Box 3 for a brief discussion of
financial vulnerability indices.
The microfinance industry needs empirical research
to identify proxies, or proxy combinations, that not
only correlate well with over-indebtedness but are
practical for ongoing collection. Kappel, Krauss,
and Lontzek (2010), described in the next section,
report on work toward an early warning index that
incorporates multiple indicators.

Box 3. Financial Vulnerability Indices


The University of South Africas Bureau of Market
Research has developed a Consumer Financial
Vulnerability Index (CFVI), which combines income
vulnerability (household income and savings) with
expenditure vulnerability (consumption and debt
service) (van Aardt and Moshoeu 2009).a CFVI
incorporates a broad range of financial obligations,
not just debt service. Conceptually, this is a superior
approach. Being short of money to pay a debt is
not that different from being short of money to
make any other required payment. Likewise, the
inclusion of income in the equation makes sense.
Difficulties spring not from payment obligations,
but from a mismatch between payment obligations
and income. Collecting this fuller set of information
is, of course, more challenging, especially in less
developed countries.
a

See also European Commission (2008).

55 Bolivia, Ghana, Country X, Karnataka, Tamil Nadu, South Africa

4. The Empirics of Microcredit


Over-Indebtedness: What We
Know to Date
We have been able to locate six field studies that
try to quantify microcredit over-indebtedness.
Given the small number of studies, our ability to
generalize is limited. But most of them have found
levels of over-indebtedness (variously defined)
that seemed higher than what the conventional
wisdom about microcredit might have suggested.
As noted earlier, a normal, healthy level of
over-indebtednessi.e., a level that couldnt be
lowered without undue restriction of loan access
is probably higher than what we would expect or
want to see at first intuition. But even with that
caution in mind, the levels of over-indebtedness
found in most of the studies seem worrisome. On
the other hand, most of the markets studied were
selected precisely because local observers were
concerned about over-indebtedness problems,55
so the small sample of countries is highly skewed.
A brief discussion of each of the studies follows,
with key points summarized in Table 1.
Bolivia. Gonzalez (2008) relied on a 19972001
household survey. He identified 1,256 households
that had at least one microloan during the period
and were willing to repay it. Three quarters of
these households had to resort at least once
to costly, unanticipated measures to repay the
loan. The list of costly measures included having
to work more than ones regular schedule
(66%), liquidating financial savings (47%), having
remittances specially sent for the purpose (29%),
selling productive assets (23%), getting a new loan
to repay another (10%), and others.
Gonzalezs definition of over-indebtedness was
an unusually expansive one. It included some
repayment measures (e.g., working extra hours
or drawing on savings) that most people would
not regard as drastic, even if the household didnt
anticipate them at the time of the loan. And a
single occurrence of such actions over a fouryear span was enough to count the household
as over-indebted. So it is not surprising that the

26

proportion of people who were over-indebted by


this definition was quite high85 percent.56
Gonzalez found that, for his sample as a whole,
there was no significant association between
multiple borrowing and over-indebtedness.
Surprisingly, in spite of the comprehensive
information (including past repayment problems)
the Bolivian credit bureau collects, he couldnt
detect a significant relation between use of a
credit bureau and lower loan delinquency.
Ghana. Grammling (2009) studied crossindebtedness and over-indebtedness among
borrowers at an MFI affiliate in Ghana, using a
combination of techniques, including rapid
market appraisals of microfinance clientele in
markets, villages, and the MFIs premises; in-depth
interviews with the MFIs clients; and an exchange
of information on a sample of clients of the MFI
and two competitor MFIs.57
Over-indebtedness was broken into three
categories:
Not over-indebted.
At risk of becoming over-indebtedthe
borrower is decapitalizing, but business assets
exceed liabilities.
O v e r- i n d e b t e d t h e
b o r ro w e r
is
decapitalizing, and business assets do not
exceed liabilities. However, private assets
are not included in the analysis, so it can be
assumed that some borrowers in this category
will manage to pay their loan by selling
nonbusiness assets.
The researchers assessed the over-indebtedness
level of each borrower, based on interviews, loan
files, and information exchanged among MFIs.
Twelve percent of the MFIs borrowers were
found to be over-indebted, and 16 percent were
judged to be at risk of over-indebtedness

levels much higher than what the MFIs ongoing


collection statistics would have suggested
at the time of the study. More than half the
borrower respondents had more than one loan
at a time, and cross-borrowing did correlate with
over-indebtedness, even though many crossborrowers were not over-indebted. Grammling
concluded that multiple and over-indebtedness
were growing fast, and that a credit bureau was
urgently needed. See Box 4 for a discussion of a
later Ghana study.
Country X. A restricted-distribution study used
a sample of roughly a thousand microborrowers
at a half-dozen institutions. Levels of overindebtedness were assessed using data on client
debt service and income from a credit bureau and
the lenders loan files. In a second phase, problem
clients and lender staff were interviewed to shed
light on the causes of over-indebtedness.
Over-indebtedness was based on a debt-service
ratio: households monthly loan payments divided
by gross income net of expenses:58
Not over-indebteddebt service below 75
percent of net income.
At risk of becoming over-indebteddebt
service between 75 and 100 percent of net
income.
Over-indebteddebt service more than 100
percent of net income.
Among the borrowers studied, 17 percent were
classified as over-indebted and 10 percent as
at risk of becoming over-indebted. Multiple
borrowing correlated strongly with overindebtedness. Poorer clients were more likely to
be over-indebted. The average monthly income of
over-indebted households was less than half that
of the not-over-indebted households.
Karnataka, India. In 2009, there were mass
defaults, largely by Muslims, in several towns

56 Gonzalezs focus was not to identify problem lending, but rather to explore clients willingness to make extra efforts to repay loans.
57 In Ghana (and no doubt in some other countries as well), borrowers are culturally reluctant to discuss loans, and especially their difficulties in
repaying loans. Grammling provides a useful description of techniques he used to overcome this problem.
58 This analysis treats household expenses as a fixed amount per month (probably drawn from loan officers appraisal of borrowers cash flow as
recorded in the MFIs loan files). In fact, household expenditures are somewhat flexible and can rise or fall to accommodate loan payments.

27

Box 4. Recent Over-Indebtedness


Study in Ghana
Two years following the Grammling study, after the
Ghanaian microfinance market stopped growing
and MFIs became more careful with disbursements,
one of the authors of this paper returned to Ghana
to analyze over-indebtedness in more detail. In
cooperation with the Independent Evaluation
Department of the German development bank KfW
and the Smart Campaign, the study is the first to
apply an over-indebtedness definition that is based
on excessive sacrifices in quantitative empirical
research.
The researchers asked more than 500 microborrowers
from five of Ghanas top MFIs about their
experiences with their loans. They collected detailed
information on the borrowers personal situation,
financial literacy, risk attitude, and borrowing
behavior. Most importantly, they mapped current
outstanding loans for every borrower and collected
a detailed list of whatever sacrifices borrowers had
to make to repay. Borrowers rated the acceptability
of each sacrifice on a subjective scale from easily
acceptable to not acceptable at all. For example,
they would rate cut down on eating as acceptable
if it was about buying less meat in a week and as
rather unacceptable if it was about going hungry
and eating only a single meal per day. Their
subjective judgment showed a fairly high tolerance
for sacrifice, some borrowers finding it acceptable to
eat less or sell their belongings. Borrowers indicated
how frequently they experienced each sacrifice.
The study sheds new light on over-indebtedness,
tackling it from a thoroughgoing customer
protection perspective. By quantifying sacrifices,
the study gives struggling borrowers a voice and
provides insights into the efforts microborrowers
make to meet their payment obligations.
Additionally, the data will enable us to reassess
whether over-indebtedness is a significant problem
in Ghana and to test some of the potential drivers
that we have discussed in Section 2 of this paper for
their relevance in the Ghanaian microfinance setting.
Watch upcoming Smart Campaign publications from
the Centre for Financial Inclusion for results.

in Karnataka, prompted in large part by orders


from local Muslim organizations banning Muslims
from continuing contact with MFIs. Krishnaswamy
(2011) reports on a CGAP study, conducted by the
consulting firm EDA, of borrowers in two towns

that had mass defaults and two similar towns


where such defaults did not occur.
In the mass-default towns, 21 percent of
respondents said repayment was a burden (versus
only 3% in the nondefault towns); 34 percent said
they had skipped important expenses, including
meals, or sold/mortgaged assets to repay loans
(versus 2% in the nondefault towns); and the amount
of weekly loan service actually paid at the time
of the crisis averaged 27 percent higher than the
borrowers said was affordable (versus 4% less in the
nondefault towns). In the mass-default towns, some
of the default was probably opportunistic. One can
only speculate as to how much the opportunistic
default and the environment of political agitation
may have skewed borrowers accounts of their
sacrifices in the mass-default towns.
An index of subjectively reported debt stress was
found to be correlated with debt service ratio,
income shocks during the preceding year, income
variability, and numerical literacy, among others.
Tamil Nadu, India. Combining quantitative and
qualitative research methodologies, Gurin,
Roesch, Subramanian, and Kumar (2011) find that
the average household in the sample villages
in South India has about one years household
income outstanding in debt and is making monthly
repayments that amount to half of its income.59
The paper argues that over-indebtedness is
a complex social concept that has multiple
meanings. For the purpose of the study, it defines
over-indebtedness as a process of impoverishment
through debt, distinguishing three different levels:
Transitional over-indebtednessdebt
servicing leads to a poverty trap, preventing
any accumulation of assets, but households
have effective strategies in place that promise
a reduction of debt in the future. Average
debt levels of 1.4 annual household incomes.
Debt service around one-third of monthly
household income.

59 Some alert readers may wonder how the math in this sentence works. It makes sense if some of the debt is amortized over a period greater than
a year. The most common sources for loans in Gurins study were well-known people and pawnbrokers. Self-help groups, which are often
included within the boundaries of microfinance, ranked third, being used by 41% of the households.

28

Pauperizationin spite of assets sales, debt


levels continue increasing, just to service
existing debts and ensure household survival.
Households survive on multiple borrowing
and have no realistic prospect of meeting
their repayment obligations in the long run.
Average debt levels of 3.2 annual household
incomes. Debt service ratio around 100
percent.
Extreme dependencehouseholds depend
on kin support and charity for daily survival
and have no possibility of ever repaying their
debt. Many cases lead to complete social
isolation and loss of self-dignity. The average
debt level of this group is 13 times its average
annual household income.
Of the original sample of 344 households, Gurin,
Roesch, Subramanian, and Kumar studied only
the most indebted 20 percent in detail. Out of
these 68 households, 13 (19%) have fallen into
transitory over-indebtedness, 26 (38%) represent
cases of pauperization, and 29 (43%) suffer from
extreme dependence.60 The findings suggest
that over-indebtedness is also prevalent among
households that didnt fall into this subsample.
Twenty percent is, therefore, the absolute
minimum estimate for overall over-indebtedness
in the original sample.
Analyzing the causes that lie behind the overindebtedness phenomenon, Gurin, Roesch,
Subramanian, and Kumar make out two major
forces: over-indebtedness seems to result from
the combination of material poverty and growing
social aspirations.
Multi-country study. Kappel, Krauss, and Lontzek
(2010) develop a preliminary early warning index
for over-indebtedness on a country level, using
primary data from an MFI survey combined
with secondary data from MIX Market61 and
macroeconomic databases. The sample is limited
to 13 countries and is based on the experience with

only 34 crisis countries. The authors acknowledge


that this imposes severe restrictions on the
reliability of the index, but the paper takes an
important first step in developing a methodology
for predicting country-level repayment crises.
The study focuses on over-indebtedness in the
sense of repayment problems, defining it as
a chronic and involuntary inability to meet all
payment obligations by means of the households
excess cash. It approximates over-indebtedness
on the country level, using 30- and 90-day arrears
as a measurement of crisis outbreak and the loan
loss rate and write-off ratio as measurements of
continued crisis.62 On an individual level, it uses a
debt service ratio as proxy for over-indebtedness.
Drawing on the available data for a modified
approach of signaling analysis and sometimes on
hypotheses from the literature or the judgment
of microfinance practitioners, Kappel, Krauss, and
Lontzek (2010) identify 14 potential early warning
indicators for over-indebtedness:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.

Remittances
Market penetration
Growth rates of total volume of loan portfolios
Quality and use of credit information sharing
systems
Perceived commercial bank involvement
Perceived level and trends in competition
Perceived investment flows
MFI liquidity
Average loan balance per borrower
Loan requirements and lending methodologies
Productivity
Growth and market targets
Multiple lending
Consumer lending

The study concludes that countries such as Bosnia


Herzegovina, Cambodia, and Peru currently have
the highest over-indebtedness risks among the 13
sample countries. Bolivia, Ecuador, El Salvador,

60 We are providing percentages to give the reader an idea of the magnitude of problems identified by this study. However, we cannot draw reliable
statistical conclusions about the distribution of households within this smaller subsample of only 68 respondents.
61 Microfinance Information Exchange, an online information portal based on self-reports but offering the broadest global database on MFIs to date.
www.mixmarket.org
62 The loan loss rate and write-off ratio measure unrecoverable loan amounts as a percentage of total loan portfolio.

29

and Georgia display the lowest risks of overindebtedness according to the selected indicators.
[There have been several studies of overindebtedness among low-income and other
borrowers in South Africa. We have not included
them here because little of the lending involved
was the kind of microcredit that is the focus of this
paper. Nevertheless, its worth noting that these
South Africa studies have generally found worrisome
levels of over-indebtedness.63] As noted, the list of
quantitative studies available so far seems much too
short. We hope that many more researchers will work
on such studies, including not only academic research
but also less rigorous market research conducted by
MFIs and regulators. See Table 2 for a summary of
the empirical studies discussed in this section.

5. What Can Be Done?


Finally, we will look at practical steps by lenders
(and, secondarily, by donors and investors that

support access to finance) that might prevent or


remedy over-indebtedness. The list of such steps
is fairly long. We can provide only a brief, broadbrush survey herefull discussion of any one of
the topics would require a separate paper.
Note that this paper does not address actions by
regulators.
Despite the desire for concrete, down-to-earth
advice on what to do and what not to do, on
most topics we avoid recommendations, not
only because the evidence base is thin, but also
because the usefulness of the action depends
heavily on local circumstances. We cannot offer a
how-to manual, but rather we can offer a checklist
of options to consider along with a few factors to
bear in mind when considering them. We begin
with measures that MFI managers can take, and
then turn to possible actions by funders. (See
Box 5 for a discussion of the Smart Campaigns
consumer protection resources.)

Table 2. Summary of empirical studies


Setting
(Author)

Methodology

Definition
of Over-indebtedness

Findings

Bolivia 19972001
(Gonzalez 2008)

HH survey

Costly unanticipated
actions to repay

85% of HH had at least


one occurrence during the
four years

Ghana 2009
(Grammling 2009)

Rapid market assessment,


borrower surveys, info
exchange between MFIs

Business decapitalizing (as


assessed by researchers)

12% over-indebted,
another 16% at risk

Country X 2009
(restricted report)

Debt service and expense


data from credit bureau
and loan files of lenders

Debt service greater than 17% over-indebted and


100% of HH income net of another 11% at risk (debt
service 75100% of net
other expenses
income)

Karnataka, India, 2010


(Krishnaswamy 2011)

Loan records and HH


survey

Subjective report of stress; Over-indebtedness high


sacrifices to repay
in mass-default towns, but
low in nondefault towns

Tamil Nadu, India 2005


2009
(Gurin, Roesch,
Subramanian, and Kumar
2011)

Qualitative interviews,
observation, and HH
survey (N=344)

Impoverishment through
debt

More (possibly much


more) than 20% overindebted

Multi-country study
(Kappel, Krauss, and
Lontzek 2010)

Preliminary composition
of an early warning index,
largely based on signaling
analysis with an MFI
survey and MIX Market
data

Chronical and involuntary


inability to meet all
payment obligations by
means of the households
excess cash. Proxies:
arrears, write-offs, loan
losses, debt service ratio

14 potential early warning


indicators for overindebtedness. Highest
risk countries in sample:
BosniaHerzegovina,
Cambodia, Peru

63 E.g., Collins (2008)

30

Box 5. The Smart Campaign


The Smart Campaign is a global coalition of
hundreds of MFIs, investors, donors, and other
industry players who are committed to keeping
clients welfare as the driving force in microfinance.
Its Client Protection Principles include a heavy
focus on avoidance of over-indebtedness. For a list
of normative recommendations in this regard for
MFIs, see Smart Campaign (2011). More generally,
the Campaigns Web site (www.smartcampaign.
org) offers a wealth of client-protection resources.

We begin with measures that MFI managers can


take, and then turn to possible actions by funders.

MFI actions
MFIs can consider revisions to product design:
Flexible loan product offerings that better
meet client needs. The incomes and expenses of
microborrowers tend to be irregular and unreliable.
The harder it is for them to match their actual cash
flow with their loan repayment schedule, the more
debt stress is likely to occur.
In the early years of microcredit, most MFIs settled
on one-size-fits-all loan products with lockstep
installments due every week or every month.
The MFIs understood that this rigid approach
didnt fit their borrowers cash flows very well.
They chose the model because a single cookiecutter product helped keep costs low (cost was
the central challenge of early microcredit), and
frequent regular installments reduced repayment
risk for uncollateralized borrowers.
But much of the industry has moved a long way up
the learning curve since then. Lending institutions
are more experienced and solid. Numerous MFIs
have demonstrated the skill needed to offer
multiple loan products, including more flexible
products, without delinquency or costs spinning
out of control. Grameen Bank, for instance, has
had great success with its Grameen II product
line up, which includes options for borrowers
who have temporary problems making their
64 E.g., Rutherford, Maniruzzaman, Sinha, and Acnabin & Co (2004).

payments.64 Note that if an MFI allows penalty-free


prepayments, borrowers are more likely to pay off
their loans during good times, which obviously
eliminates the risk that they will run into trouble
meeting later payments.
Whatever the relation to over-indebtedness,
improving product flexibility and, more generally,
matching products better to client needs are
important overall goals for the microcredit
industry. And it seems strongly plausible that
over-indebtedness would be lower if amortization
schedules could fit better with the timing of
borrowers incomes, and if more flexibility could
be allowed in repayment. However, we dont
want to imply that all microlenders should be
moving to flexible loan products right now. All
other things being equal, flexible products tend
to be more complex and expensive to administer,
and there are situations where insisting on too
much flexibility can actually limit access. General
discussions like this one cant go very far toward
defining the proper balance among flexibility,
cost, risk, and access. That question should be
addressed in specific settings, based on the
particular characteristics of individual lenders and
their borrowers.
More conservative loan amounts and increments
on graduated lending ladders. Lenders might
lower the size of new clients first loans, or slow
the growth in loan sizes when previous loans are
repaid, or not grant the same size increase to
everyone automatically. Steps like these should
be approached cautiously, because they directly
reduce loan access. As noted, clients are often
driven to multiple borrowing (including informal
borrowing) because they want, and can handle,
loan amounts that are larger than an MFI is willing
to give them.
Savings. Voluntary savings products might also
play a role in reducing borrower stress. If the MFI
offers convenient, safe, liquid savings vehicles, its
customers will use them. A customer with a liquid
savings cushion would seem less likely to have to
make extreme sacrifices, such as selling a business

31

asset or taking a child out of school, to make a loan


payment.
Nonliquid commitment savings products, with
restrictions on withdrawal, and/or incentives to
make regular deposits, can play a role as well. As
discussed in Section 2, many poor people take
out loans to help themselves with the discipline
to savei.e., to set aside a regular amount each
period and protect it from the temptations of
immediate consumption. They find this service so
valuable that they are willing to pay high interest
rates to save this way. A good commitment savings
product would be much better than debt for this
purpose. The borrowers cost and risk are both
much lower that way.65
MFI managers have reported that when loans are
funded largely by local community deposits, both
loan officers and borrowers are more responsible
about putting those funds at risk. And finally, a
loan officer may be less likely to twist a borrowers
arm to take out an inappropriate follow-on loan if
the MFI will have a continuing deposit relationship
with the customer.
MFIs can consider adjustments to their sales
process:
Marketing practices and follow-on loans. Some
marketing practicese.g., targeting active clients
of another microlender with offers to add to
their debt levelcan intensify the risk of overindebtedness.
Loan officers should avoid pressuring existing
clients who have repaid a loan to take out another
one immediately, especially if the clients dont
need a new loan just then. Particularly, they should
not say, or imply, that these clients will lose access
to future loans if they dont borrow constantly.
Clear disclosure of loan terms. It seems plausible
that people who understand the costs and other

requirements of their loan clearly would be less


likely to over-indebt themselves. And indeed, there
is a straightforward case for informing borrowers
about the obligations they are undertaking,
whether or not it reduces over-indebtedness.
Even in the absence of regulatory requirements,
MFIs should ensure that borrowers get a brief,
understandable statement of how much net cash
will be disbursed to them, the expected repayment
schedule, any fees and penalties, and perhaps
other key terms and risks (e.g., the consequences
of default).66 The format for such disclosure needs
to be tested for intelligibilityinformation that
is complete and accurate may still be useless if
it is stated in language that borrowers cannot
digest, or is buried in a lengthy and complicated
disclosure document.
When it comes to disclosure of interest costs, the
picture is more complex. Theoretically, the ideal
interest rate disclosure format would capture all
information about amount and timing of cash flows
in a single number that can be used to compare
the cost of various forms of credit that may be
structured quite differently. Discounted cash flow
calculations, such as the annual percentage rate
(APR), do this, and APR is the standard disclosure
tool in most of the world.
But an emerging body of evidence suggests that
APR disclosure doesnt work very well, at least by
itself. However accurate it might be, it seems to fall
short of the desired effect on borrower behavior,
especially for lower income and less sophisticated
borrowers, in both rich and poor countries.67 It
seems likely that in most microcredit settings,
clients will be better served by complementing
or replacing APR with some other price disclosure
methode.g., a simple statement of the total
amount of interest and fees that the borrower will
pay over the life of the loan. The Philippine central
bank is involved in field research testing various
forms of interest disclosure with microborrowers.

65 See Abed (2011).


66 Although it is less common among microlenders, late-payment penalties can be a major source of income for some lenders. But such penalties are
usually not included in APR calculations, because there is no way to predict their amount at the beginning of a loan.
67 E.g., Collins, Morduch, Rutherford, and Ruthven (2009); Bertrand and Morse (2010); Robb (2007); Barr, Mullinaithan, and Shafir (2008);
Elliehausen (2010); Brix and McKee (2010).

32

Financial literacy. Growing concern about overindebtedness is matched by growing interest


in programs that help cultivate the knowledge,
skills, attitudes, and behavior people need to make
sound financial decisions. Some MFIs deliver such
material to customers at the preloan stage. Often
such programs are also promoted by governments,
educators, and civil society actors. Evidence so far
about the effectiveness of these programs seems
mixed.68 More research and experimentation is
needed.
Expand into new areas rather than already
served ones. As noted in Section 1, MFIs often
prefer to expand into areas where some competitor
has already developed customer awareness of
microcredit and demonstrated a strong market,
rather than expanding into previously unserved
localities. As MFIs become more aware of the
risks to themselves and their customers posed by
oversaturated local markets, we hope that they
will channel their expansion more often into unor underserved areas. This could reduce overindebtedness and improve outreach at the same
time.
Many MFIs face key issues in their loan
underwriting (i.e., analysis of borrower creditworthiness).
Cash flow evaluation. The most obvious way to
reduce overindebtedness risk is to strengthen
the assessment of a borrowers repayment
capacity before approving loans. This starts with
determining the amount, regularity, and reliability
of the borrowers (or, more typically, the borrowing
households) income. Income has to be matched
against outgoing cash flows, including not only
the debt service on the proposed loan, but also

consumption and other expenses, including,


where possible, debt service on other borrowings.
The risk of adverse shocks to income and expenses
should be considered.
However, information about the borrowers other
debts may be difficult to get if there is no credit
reporting system, and even a good credit reporting
system will not capture debts to informal lenders.69
We discuss credit reporting later in this section.
This kind of cash flow analysis is a common feature
of individual and solidarity group lending,70 even
though it is not always vigorously implemented.
It is far less common in group lending. In some
circumstances, the threat of over-indebtedness
might require substantial adjustments to group
lending techniques. After an over-indebtedness
crisis, Kashf Foundation in Pakistan eliminated
group lending and moved to an individual model
where loan officers analyzed each borrowers cash
flow, though this roughly doubled the cost of
lending.71 As a less drastic alternative, the Negros
Womens Trust for Finance in the Philippines is
experimenting with training group borrowers to
analyze each others cash flow.72
MFIs sometimes fail to refresh the cash flow
analysis when borrowers take out follow-on loans
that are much larger than the original loan for
which the analysis was performed.
Specific limits on debt service ratios. Once the
cash flow analysis is done, an MFI may constrain
loan officers discretion by imposing a cap on the
debt service for the proposed loan as a percentage
of net incomee.g., a loan will be denied if the
repayments would constitute more than X percent
of household income minus consumption and other

68 E.g., Bilal (2010).


69 Clients are usually not forthcoming about their other debt. They perceive (quite correctly) that disclosing other debt may hurt their chances of
getting their loan approved. In the absence of a credit reporting system, one possible approach to improving clients incentives to disclose might
be for microlenders to make it clear to borrowers that (1) carrying multiple loans is fine as long as total debt service doesnt exceed repayment
capacity, and (2) failure to disclose other active loans (perhaps only loans from formal lenders) will, if detected, result in the cancellation of the
current loan and the borrowers permanent exclusion from the MFIs services. E.g., the lenders could try to locate occasional examples of false
disclosure by informal consultations with other lenders, and make sure that its broad clientele is aware that the sanctions are in fact enforced
vigorously in these cases. This approach sounds harsh, but it could reduce over-indebtedness.
70 In solidarity group lending, borrowers are organized into small groups, but the MFI lends to individuals rather than the group, and assesses each
individual loan separately.
71 Roshaneh Zafar at the 2011 Boulder MicroFinance Training. Kashfs Web site is www.kashf.org.
72 Roque Caseres at the 2011 Boulder MicroFinance Training. The MFIs Web site is www.nwtf.ph. For suggestions on loan underwriting in group
lending, see Smart Campaign (n.d.). See also Smart Campaign (2010) for a recommendation against approving loans based solely on guarantees
by others, when there has been no appraisal of the borrowers repayment capacity.

33

expenses (including debt service on other loans).


The MFI may or may not allow exceptions to the
policy based on a specific approval process.73
Verifying borrowers repayment history and
other debts. Where there is a credit reporting
system, MFIs need to be sure their loan officers
use it to identify borrowers who have experienced
problems with repayment in the past, as well as to
find out what debts the borrower has with other
formal lenders. Even in the absence of a formal
or informal credit reporting system, MFIs may be
able to tap community knowledge, or use other
techniques, to identify borrowers who are in debt
to another lender. And in any circumstances an
MFI can, of course, check its own collection history
with the borrower; surprisingly, a few MFIs fail to
do this systematically.
An MFIs collection process is another arena for
possible action.
Appropriate policies for renegotiation of
loans. When borrowers cannot meet a payment,
loan officers or managers may renegotiate the
delinquent loan, either by amending its terms to
stretch out the repayments (rescheduling) or by
giving the borrowers new loans they can use to
pay off the old ones (refinancing). The motivation
for the renegotiation may be to accommodate
individual borrowers who are likely to be able to
repay eventually. This is appropriate. As discussed
earlier, literal enforcement of a zero-tolerance
policy is seldom desirable. But often, staffwith
or without the collusion of branch managers
will extend or roll over loans for a borrower who
has no realistic prospect of eventual repayment,
to protect salary bonuses that are tied to loan
repayment. This kind of renegotiation is extremely
dangerous, because it can conceal from central
management a serious outbreak of repayment
problems until it spins out of control. Perversely, a
policy that prohibits loan rescheduling altogether
can make it more likely that loan officers roll over
uncollectible loans by issuing new ones to the same

borrower. In addition, inappropriate rescheduling


may allow borrowers to dig themselves deeper
into debt problems instead of facing them at an
early stage where less drastic solutions might still
be available.
Getting the rules right involves some tricky
balancing. Opening the door to loan rescheduling
introduces ambiguity and may increase costs, but
keeping this door totally shut hurts borrowers who
have run into honest repayment difficulties but are
likely to be able to pay eventually. MFIs need clear,
carefully thought out renegotiation policies.
Restraining abusive collection. Over-aggressive
loan collection practices are doubly dangerous.
They increase the likelihood that borrowers will
make draconian sacrifices to repay, which may
be good for the lender but can be bad for the
borrowers. And conscience-shocking collection
practices have often fueled major political
backlash.
Which practices are abusive? Somee.g., physical
threatsshould be unacceptable in any setting.
But as long as they do not reach the extent of
harassment, the acceptability of otherse.g.,
repeated visits to a borrowers house, publicizing
the names of nonpaying borrowers, or pressure
from other members of a borrowers groupmay
depend on local attitudes and culture.
Collection practices need to be effective, without
trespassing on the courtesy and respect clients
deserve. This can be a delicate balancing act. MFIs
need to define acceptable practices with care and
specificity.74
Penalty interest. Some lenders charge higher
interest on late payments. This has obvious value
as a repayment incentive, but it can be dangerous.
Not only does it add to the payment obligation of
already overburdened borrowers, but in addition, if
a lender is making a substantial portion of its profit
from late fees, it may have created, intentionally

73 The Smart Campaign (2011) cautions that if the lender is calculating debt service ratios, [i]t is also useful to have a qualitative definition of overindebtedness to help staff keep the main objective in mind and to avoid the rote use of numeric tools. An example of such a definition is a state
in which a borrower has to make significant sacrifices to his or her standard of living or business affairs in order to repay debts.
74 The Smart Campaigns checklist on collection practices is at http://www.smartcampaign.org/storage/documents/Tools_and_Resources/
Collections_Guidelines_FINAL.pdf. For a good example of a collections policy, see Swadhaar (2011).

34

or unintentionally, an incentive to get borrowers


in trouble.
Internal redress mechanisms. Problems and
misunderstandings inevitably arise in the course
of collections as well as other aspects of credit
delivery. They can be addressed better if the
lender has a clear and fair dispute process
that is known to customers, readily accessible,
and efficient. Some complaints turn out to
be inquiries or misunderstandings. If they are
handled well, the borrower ends up more loyal
to the lender; if not, the borrower may be less
inclined to meet his or her loan obligations. In
other cases, legitimate complaints may point to
inconsistent or inappropriate behavior by staff
(e.g., pressure to borrow more than desired,
overly aggressive collections) or their failure
to follow established policy. If customers know
where to go in such cases and have confidence
that they will get an even-handed hearing and
that the matter will be dealt with promptly,
the lender has a chance to identify credit risk
trouble spots, ensure compliance with policy,
and take appropriate action against infractions.
The lender also receives valuable feedback that
can inform improvement in loan products and
processes.
Some lenders place internal dispute resolution
with a person or unit that is separate from lending
operations, and reports to senior management or
the board. Special care should be taken to ensure
that customers know their options for recourse.
Some lenders procedures for handling borrowers
with serious repayment problems include one-onone debt counseling and a well-articulated process
for deciding when rescheduling or refinancing
might be justified.
Redress mechanisms could also be lodged
externally, for instance with a federation of
microlenders.
Other MFI measures that can affect overindebtedness are not associated with a particular
stage of the loan process.

Written policies, and enforcement through


internal audit. Whatever policies the lender
adopts on each of the topics discussed so far
in this section, they should be stated in writing,
communicated clearly to loan officers, and
reinforced periodically.
Even when an MFI puts its policies in writing
and communicates them clearly, the parts that
are inconvenient for loan officers will not be
implemented consistently unless incentives
are properly aligned. One strong incentive is a
vigorous system of internal audit that regularly
checks on compliance with these policies.
Most MFIs use people with accounting backgrounds
to staff their internal audit department. But for
testing loan officer behaviors like the ones weve
been discussing, managers should strongly
consider adding former loan officers into their
internal audits. Former loan officers know all
the tricks, and they will be much more effective
at interviewing loan officers and borrowers
an essential component in testing compliance
with the policies discussed above. Staff whose
background is accounting are more likely to focus
on paper documentation.
Specialized portfolio audits. Normal external
audits of MFIs, and sometimes even bank
examinations by prudential supervisors, do not
look at loan portfolio quality intensively enough
to provide solid assurance that the reported
collection performance reflects reality. This is a
serious gap. Loan collection is far and away the
biggest business risk facing most MFIs. And while
good collection is not a guarantee that there is
little over-indebtedness, deteriorating collection
is a sign that there may be serious borrower
distress.75
An MFI that wants solid independent portfolio
testing usually needs to supplement its standard
annual audit with additional agreed procedures
that instruct auditors to conduct specified tests
and report the results. There are at least two
portfolio testing tools available that focus on the

75 In some cases, poor collection results not so much from borrower distress as from managements failure to keep their staff strongly focused on loan
repayment.

35

specific risks presented by microcredit portfolios


(Christen and Flaming 2009, MicroSave 1999).
Both of these tools can also be used by internal
auditors. They examine not just loan balances but
also MFI policies, procedures, and information
systems. Note that we are speaking here of
auditing issues connected with loan repayment.
We are not suggesting audits as a tool to directly
determine client over-indebtedness.
Staff training. Training can raise loan officers
awareness of the problem of over-indebtedness,
educate them about the lenders formal policies,
and illustrate practical ways to deal with commonly
encountered situations. One good way to find out
whether management is serious about preventing
over-indebtedness is to see whether this topic
features prominently in employee trainingboth
initial training at the time of hiring and follow-on
training thereafter.76
Staff incentives. Employees respond to incentives.
Many MFIs offer cash bonuses for certain kinds
of results, for instance recruiting new borrowers,
increasing the loan portfolio, or maintaining strong
collection. Even when there are no cash bonuses,
employees have expectations about what kind of
results lead to promotion and salary raises.
At the risk of some over-simplification, we can
divide staff incentives into two groups: those
that push toward expansion of the MFIs or the
individual loan officers number of borrowers
or amount loaned, and those that focus on
collection of loans. It has long been argued that
expansion incentives are dangerous unless they
are balanced with strong collection incentives.
The concern was mainly for the well-being of
the MFI. Without enough staff attention to
maintaining repayment, delinquency could spin
out of control. But even if the concern is client
welfare rather than MFI welfare, a strong focus on
loan repayment makes sense. When loan officers
know their compensation or promotion depends
on high repayment, they are less likely to structure
loans that hurt borrowers by over-straining their
repayment capacity.
76 Sample training resources can be found at www.smartcampaign.org.

One of the Smart Campaigns assessment criteria is


Portfolio quality valued: Productivity targets and
incentive systems value portfolio quality at least
as highly as other factors, such as disbursement or
client growth. Growth is rewarded only if portfolio
quality is high (Smart Campaign 2010).
But collection targets can be a double-edged
sword. If the targets are so high that they amount
to zero tolerance for delinquency and default, then
three problems occur. First, the MFI is probably
restricting access by lending too conservatively.
Maintenance of perfect collection is often a sign
that the MFI is denying loans to many people
whose odds of repayment are very high. A modest
level of delinquencysay, maybe a PAR of 12
percentis safe and sustainable, and consistent
with serving a less restricted range of borrowers.
Second, expectations of zero delinquency can
encourage loan officers to engage in abusive
collection practices. Third, such an expectation
can lead loan officers to reschedule or refinance
loans that are ultimately unpayable, depriving
management of critical information about portfolio
problems.
Staff quality. It has been suggested that higher
pay may attract more qualified loan officers, who
would exercise better judgment in assessing
repayment ability. The suggestion seems plausible,
but we have no empirical evidence to offer on
the subject. Of course, higher loan officer pay will
usually mean higher interest rates for borrowers.
Early warning systems for over-indebtedness.
Many MFI managers are caught unawares by an
over-indebtedness crisis that would have been
much easier to fix if it had been spotted earlier.
This strongly suggests the need for formal or
informal early warning systems. The approaches
discussed here could be implemented by MFIs,
MFI associations, or government bodies.
First consider nonsurvey approaches that are
possible if microlenders and their borrowers are
covered by a functioning credit reporting system.
Such a credit database can be used to track multiple

36

indebtedness, the amount of indebtedness per


customer, the number of loan commitments, or the
number of credit inquiriesat least for debt with
formal creditors. This information can be useful for
watching trend lines even if we arent prepared
to identify particular thresholds as unacceptable.
As credit bureaus are unlikely to have income
information for microborrowers, an MFI could
draw on its own loan files to produce a debtservice-to-income ratio. At a minimum level,
national household surveys may have average
income estimates for various occupational groups
that could be combined with debt information
from the credit bureau.
Loan officers often know their customers well
and could provide useful indications of overindebtedness trends if this information can be
collected in ways that dont threaten their bonuses,
raises, and promotions. Another approach that
doesnt require formal survey work is to find a way
to collect loan officer views about how many clients
are over-indebted, and whether things are getting
better or worse. For MFIs that offer voluntary
savings services, it may be useful to monitor
for deposit withdrawals that are used to pay a
loan installment. And, obviously, management
and information systems that flag delinquency
immediately and control inappropriate loan
renegotiation will bring problems to light more
quickly.
Now turn to approaches that involve surveying
going out and asking an appropriately sized
sample of clients. In most cultures, money and
especially debt are sensitive subjects. In addition,
respondents may worry that disclosing their
debt problems will make it harder for them to
get new loans. Interviewers need to be taught
how to win respondents confidence, guarantee
confidentiality, and frame tactful questions to get
the desired information.

In the end, responses dont have to be totally


honest for the results to be useful. The usual bias
will be to understate ones debt or difficulty in
repaying. So the reported result can be treated
as a lower bound, with actual over-indebtedness
likely to be higher rather than lower. Also, for all
kinds of trend analyses, as long as the same biases
are present from quarter to quarter or year to year,
trend data can be meaningful even if we cant
quantify the amount of the biases. (See Box 6 for
advice on conducting client surveys.)
Finally, some initiatives are implemented at the
level of the market rather than of individual MFIs.
Credit reporting systems.77 Credit reporting
systems allow lenders to share information on
borrowers debt, debt service, and/or repayment
performance. By informing lenders about a loan
applicants other obligations, these systems reduce
the risk of over-indebting the borrower, and the
consequent risk to the lenders own viability.
Just as important, a credit reporting system lets
borrowers convert their good repayment behavior
with one lender into a reputational asset that gives
them access to other credit sources as well. Many
people consider credit reporting to be the single
most powerful weapon to fight over-indebtedness
in competitive markets. At the same time, credit
bureaus can have their downsides for clients and
arent a silver bullet that fixes over-indebtedness
alone without attention to the other kinds of
measures discussed here.78
The ideal reporting system for low-income
borrowers would allow (or require) participation
not only by licensed banks, but also unlicensed
nondepository lenders, such as MFIs and consumer
credit companies, as well as other providers that
clients owe regular payments to (e.g., telephone
providers or appliance and furniture merchants),
and include both positive and negative repayment
information.

77 Our discussion of this topic draws on Christen, Lauer, Lyman, and Rosenberg (2011). This paper distinguishes three broad approaches to the
sharing of borrower information among lenders: government-run credit registries, privately owned credit bureaus, and MFI-specific databases
that are usually set up because credit registries and bureaus wont incorporate lower income borrowers or lending institutions that are not licensed
and prudentially regulated.
78 Data accuracy and client privacy are common issues. And in one sense, credit bureaus can increase repayment stress for borrowers. Without a
credit bureau, the borrower can default with one MFI but then get loans from its competitors. Once the credit bureau is in place, the pressure on
the borrower to pay is higher, because default could reduce access to finance and potentially other types of transactions in the long run.

37

Box 6. Client Surveys: Advice from Two


Field Researchers
Surveying is complex work that calls for detailed
expertise. In an interview for this paper, Marguerite
Robinson and Daryl Collins, two very experienced
client research specialists, were willing to offer
some general pointers:

work usually cant expect meaningful answers


on questions like these if they just turn over the
whole task to a local research firm. The actual
surveying can be contracted to a research firm,
but the MFI or other commissioning institution
should have its own in-house expertise to
develop and pilot the questionnaire, to monitor
the survey work, and, if possible, to include the
outside firms lead field researcher during the
pilot testing to make sure that expectations are
understood.
training, not just a few general guidelines, if
they are expected to get honest and open
answers on sensitive questions like household
debt.
market research should begin with intensive
qualitative interviews with a few respondents,
to understand the dynamics of the behavior
theyre investigating, before they launch
statistical surveys of large samples.
In addition, Collins was willing to venture an orderof-magnitude guesstimate of survey costs, subject
to situational caveats. To survey the clients of a
single MFI, a sample of around 7501000 might
typically be required. Depending on local survey
firm pricing, the cost might range from $30,000
to $150,000, which should be within the means
of a medium or large MFI. For a national survey,
one might multiply those estimates by a factor
of something like five. Of course, these figures
depend on the size of the MFI or country, on the
type of survey, and on the desired results and
statistical significance levels. Other researchers
suggest that costs may be lower.

It would be a mistake to assume that every


microcredit market needs a credit bureau. Among
other circumstances, MFIs in some early stage
markets arent yet making enough loans for credit
reporting to be cost-effective. But based on past
experience, MFIs are much more likely to start
thinking about credit bureaus too late rather than
too early.

Industry codes. Local MFI networks in a number


of countries have worked on voluntary codes of
behavior over the years. The earlier attempts were
often driven, in large part, by a desire to forestall
government regulation. Getting agreement
on, and compliance with, the codes has often
been difficult. As over-indebtedness and other
consumer problems have become more prominent
recently, the motivation behind the codes seems
to be increasing, and the codes are paying more
detailed attention to consumer protection issues,
including over-indebtedness. The international
Smart Campaign has made considerable progress
in developing substantive consumer protection
principles, offering tools to implement those
principles, and securing endorsements from
hundreds of institutions around the world. It is
clearly a promising initiative, though more time
will be needed before the impact on MFI behavior
can be assessed.

Funder actions
Some observers have concluded that the behavior
of fundersi.e., the donors and investors that
finance MFIscan contribute to credit crises.
For instance, Chen, Rasmussen, and Reille (2010)
found that in four crisis countries rapid growth,
market saturation, and in some cases over-lending
were fueled by the large supply of funding
mainly debt fundingfrom international investors
and domestic apex wholesalers. Some of this
funding was purely commercial, but most of it
came from sources that included social welfare in
their objectives. Naturally, fundersespecially the
more commercially oriented oneswant to invest
in strong MFIs with solid track records in dynamic
markets. This biases them toward markets where
the risk of saturation-induced over-indebtedness
may be higher.
Many of the institutions that fund MFIs face
disbursement pressure of their own. The supply of
money they have to move sometimes exceeds the
demand from appropriate investees. If they overfund an MFI or a market, they are solving their own
problem in a way that can hurt the very clients the
money is supposed to be helping.

38

Here is one point on which we are prepared to


make a definite recommendation: for all funders,
and especially for donors or socially oriented
investors, evaluation of microlenders as potential
grantees or investees should always include an
explicit assessment of over-indebtedness risk. This
includes trying to gauge the saturation level of the
overall markets into which they are investing, and
assessing whether the microlender investees are
dealing appropriately with the range of options
laid out above. In Section 1, we developed an
argument for regarding over-indebtedness as
a clear and present danger for microlending. If
that argument is convincing, then any funder who
professes a social objective should avoid financing
microlenders that are not taking credible steps to
address that risk.79

regulators efforts to control over-indebtedness


risk. In addition, we encourage institutions
who finance microcredit research to emphasize
studies that examine the extent and dynamics of
overindebtedness.

Credible steps vary from one setting to another,


but at a minimum, a funder ought to assure itself
that the investee microlender is

We looked at some of the causes of overindebtedness, finding that lender practices,


borrower mistakes, and external factors all
contribute to the problem.

not using deceptive or high-pressure


marketing tactics
not structuring staff incentives in ways that
encourage over-lending
taking reasonable measures to check
on borrowers repayment capacity, past
repayment history, and outstanding
obligations with other lenders
maintaining and communicating clear written
policies to guide employees in addressing
over-indebtedness risk
not using collection techniques that are
abusive, given the local setting.
Roodman (2011, ch. 9) elaborates on the risk of
investor-driven credit bubbles, and argues that
microfinance funders should set up reporting
systems that allow transparency and exchange of
information about levels of investment in MFIs and
markets.
Public development agencies can finance some of
the above measures directlye.g., implementation
of credit reporting services, early warning systems,
or financial literacy initiativesor by supporting

Conclusion
We began this paper by listing reasons for paying
close attention to the risk of over-indebting
microborrowers, especially as more and more
markets become competitive and eventually
approach saturation. We pointed out that there is
often a trade-off between over-indebtedness on
one hand and access or cost for borrowers on the
other: the only way to eliminate over-indebtedness
completely is to stop lending.

We then investigated various definitions or proxy


indicators for over-indebtedness, finding that all
of them suffer from limitations. For survey work,
we preferred an indicator based on borrower
struggles and sacrifices to repay loans. At the
same time, we noted that poor people often have
to struggle and sacrifice to come up with many
kinds of cash payments, even without microloans,
and we cautioned that, when one finds borrowers
struggling to repay their loans, one cannot
automatically conclude that the loans are making
those borrowers worse off.
We reviewed a short list of studies that have
tried to quantify microcredit over-indebtedness
levels and over-indebtedness risk. Most of them
found levels of over-indebtedness, variously
defined, that seem worrisome, but the study
countries were not representative of worldwide
microcredit markets. Most of these studies were
implemented because there was a pre-existing
concern about an over-indebtedness crisis in the
particular markets. Overall, the evidence is too
skimpy so far to draw general conclusions about

79 Cf. UNPRI, Principles for Investors in Inclusive Finance, http://www.unpri.org/files/2011_01_piif_principles.pdf.

39

the degree of microcredit over-indebtedness


worldwide, but at least it shows that the topic
needs more attention.
Finally, we looked all too briefly at a wide
range of possible approaches to preventing
over-indebtedness, focusing on measures for
microlenders and for those who fund them.
Among the most important are improvements in
MFI marketing and underwriting, products that
better match client needs and cognitive abilities,
credit reporting, and early warning systems.
If managers, loan officers, and funders become
more alert to microcredits potential downsides,
that alone should go a long way to helpalong
with a lot more research on the extent and
dynamics of over-indebtedness. In too many
places, we are simply flying blind right now.

Bannerjee, Abhijit, and Sendhil Mullinathan.


2009. The Shape of Temptation: Implications for
the Economic Lives of the Poor. Boston: Harvard
University and MIT.
Barr, Michael, Sendhil Mullainathan, and Eldar
Shafir. 2008. Behaviorally Informed Financial
Services Regulation. http://www.newamerica.
net/publications/policy/behaviorally_informed_
financial_services_regulation
Bertrand, Marianne, and Adair Morse. 2010.
Information Disclosure, Cognitive Biases and
Payday lending. http://www.poverty-action.org/
sites/default/files/Morse-Paper.pdf
Bilal, Zia. 2010. The Fad of Financial Literacy?
Blog, World Bank Financial and Private Sector
Network. 12 August. http://blogs.worldbank.org/
allaboutfinance/the-fad-of-financial-literacy

References
Abed, Sharmeran. 2011. Over-Indebtedness: A
Practitioners Perspective. CGAP Microfinance
Blog. 22 February. http://microfinance.cgap.org/
author/Shameran-Abed/
Akerlof, G. A. 1991. Procrastination and
Obedience. The American Economic Review,
81(2): 119.
Anand, Malika, and Richard Rosenberg. 2008. Are
We Overestimating Demand for Microloans?
Brief. Washington, D.C.: CGAP, April. http://
www.cgap.org/gm/document-1.9.2724/BR_
Overestimating_Demand_Microloans.pdf
Ashraf, N., D. Karlan, and W. Yin. 2005.
Tying Odysseus to the Mast: Evidence from a
Commitment Savings Product in the Philippines.
Yale University Economic Growth Center
Discussion Paper No. 917. New Haven, Conn.: Yale
University. http://people.hbs.edu/nashraf/SEED.
pdf
Bannerjee, Abhijit, and Esther Duflo. 2007. The
Economic Lives of the Poor. Journal of Economic
Perspectives, 21/1 Winter: 14167.

Bouquet, E., B. Wampfler, E. Ralison, and M.


Roesch. 2007. Trajectoires de crdit et vulnrabilit
des mnages ruraux: Le cas des Cecam de
Madagascar. Autrepart, 44: 15772.
Brix, Laura, and Katherine McKee. 2010. Consumer
Protection Regulation in Low-Access Environments:
Opportunities to Promote Responsible Finance.
Focus Note 60. Washington, D.C.: CGAP. http://
www.cgap.org/gm/document-1.9.42343/FN60.
pdf
Brown, T. M. 1952. Habit Persistence and Lags
in Consumer Behaviour. Econometrica, 20(3):
35571.
Brown, Christopher. 2004. Does Income
Distribution Matter for Effective Demand?
Evidence from the United States. Review of
Political Economy, 16(3): 291.
CARE Bangladesh. 2005. Debt and Vulnerability
in Northwest Bangladesh. Discussion Paper,
Livelihoods Monitoring Unit, Rural Livelihoods
Program.
CARE Bangladesh. 2005a. Debt and Vulnerability
in Northwest and Southeast Bangladesh: A

40

Cross-Regional Comparison. Discussion Paper,


Livelihoods Monitoring Unit, Rural Livelihoods
Program.

Duesenberry, James S. 1949. Income, Savings and


the Theory of Consumer Behavior. Cambridge,
Mass.: Harvard University Press.

CGAP. 2010. Andra Pradesh 2010: Global


Implications of the Crisis in Indian Microfinance.
Focus Note 67. Washington, D.C.: CGAP. http://
www.cgap.org/gm/document-1.9.48945/FN67.
pdf

Duflo, Esther, Michael Kremer, and Jonathan


Robinson. 2009. Nudging Farmers to Use
Fertilizer: Theory and Experimental Evidence from
Kenya. National Bureau of Economic Research
Working Paper No. 15131.

Chaudhury, Iftekhar, and Imran Matin. 2002.


Dimensions and Dynamics of Microfinance
Member OverlapA Micro Study from
Bangladesh.

Ellihausen, Gregory. 2010. Implications of


Behavioral Research for the Use and Regulation
of Consumer Credit Products. Washington, D.C.:
Federal Reserve Board and George Washington
University. http://www.federalreserve.gov/pubs/
feds/2010/201025/201025pap.pdf

Chen, Greg, Stephen Rasmussen, Xavier Reille.


2010. Growth and Vulnerabilities in Microfinance.
Focus Note 61. Washington D.C.: CGAP.
Christen, Markus, and Ruskin M. Morgan. 2005.
Keeping up with the Joneses: Analyzing the Effect
of Income Inequality on Consumer Borrowing.
Quantitative Marketing and Economics, 3(2): 14573.
Christen, Rebert Peck, Kate Lauer, Timothy
Lyman, and Richard Rosenberg. 2011. A Guide
to Regulation and Supervision of Microfinance.
Consensus Guidelines. Washington, D.C.: CGAP.
Draft for public comment.
Christen, Robert Peck, and Mark Flaming. 2009.
Due Diligence Guidelines for the Review of
Microcredit Loan Portfolios. Technical Guide.
Washington, D.C.: CGAP. http://www.cgap.
org/gm/document-1.9.36521/DueDiligence_
TechGuide_ENG.pdf
Collins, Daryl, Jonathan Murdoch, Stuart Rutherford,
and Orlanda Ruthven. 2009. Portfolios of the Poor.
Princeton, N.J.: Princeton University Press.
Cressy, Robert. 2006. Debt Finance and Credit
Constraints on SMEs. In S. Parker, ed., The
Life Cycle of Entrepreneurial Ventures, Boston:
Springer, 187225.
Dick, Astrid A., and Andreas Lehnert. 2010.
Personal Bankruptcy and Credit Market
Competition. Journal of Finance 65(2): 755686.

European Commission. 2008. Towards a Common


Operational European Definition of Overindebtedness. Brussels: European Commission
Directorate-General for Employment, Social
Affairs and Equal Opportunities Inclusion, Social
Policy Aspects of Migration, Streamlining of Social
Policies Unit. http://ec.europa.eu/employment_
social/spsi/docs/social_inclusion/2008/study_
overindebtedness_en.pdf
Fudenberg, Drew, and David K. Levine. 2006.
A Dual-Self Model of Impulse Control. The
American Economic Review, 96(5): 144976.
Gonzalez, Adrian. 2008. Microfinance, Incentives
to Repay, and Overindebtedness. Evidence from
a Household Survey in Bolivia. Doctoral thesis,
Ohio State University, Ohio.
. 2008. How Many MFIs and Borrowers
Exist. Updated December 2008. MIX Data Brief
No. 5. Washington, D.C.: MIX. http://www.themix.
org/publications/microbanking-bulletin/2008/12/
how-many-mfis-and-borrowers-exist-updateddec-2008
. 2010. Is Microfinance Growing Too Fast?
MIX Data Brief No. 5. Washington, D.C.: MIX.
Grammling, Mattias. 2009. Cross-Borrowing
and Over-Indebtedness in Ghana: Empirical
Evidence from Microfinance Clientele and Small

41

Enterprises. Technical draft for discussion.


Frankfurt: Procredit Holding.

Constructing an Early Warning Index. Zurich:


Center for Microfinance, University of Zurich.

Gurin, Isabelle. 2006. Women and Money:


Lessons from Senegal. Development and Change,
37(3): 54970.
. 2010. Microfinance, Indebtedness and
Overindebtedness. Presentation slides.

Karlan, Dean, and Jonathan Zinman. 2007.


Expanding Credit Access: Using Randomized
Supply Decisions to Estimate the Impacts.
London: Center for Economic Policy Research.
http://www.cepr.org/Pubs/new-dps/dplist.
asp?dpno=6180

Gurin, I., M. Roesch, V. Subramanian, and


S. Kumar. 2011. The Social Meaning of OverIndebtedness and Creditworthiness in the Context
of Poor Rural South India Households (Tamil
Nadu). RUME Working Paper No. 2011-1. Paris:
RUME.

Kilborn, Jason. 2007. Behavioral Economics,


Overindebtedness & Comparative Consumer
Bankruptcy: Searching for Causes and Evaluating
Solutions. Emory Bankruptcy Developments
Journal, vol. 22: 1345. http://www.law.emory.
edu/fileadmin/journals/bdj/22/22.1/Kilborn.pdf

Gurin, Isabelle, Solene Morvant-Roux, and


Magdalena Villarreal. 2011. Over-indebtedness
and financial inclusion. http://www.rume-ruralmicrofinance.org/spip.php?rubrique8

Krishnaswamy, Karuna. 2007. Competition and


Multiple Borrowing in the Indian Microfinance
Sector. September. India: Institute for Financial
Management and Research. http://ifmr.ac.in/cmf/
publications/wp/2007/17_karuna-competition.pdf

Gul, Faruk, and Wolfgang Pesendorfer. 2004.


Self-Control and the Theory of Consumption.
Econometrica, 72(1)11958.
Hellmann, Thomas F., Kevin C. Murdock, and
Joseph E. Stiglitz. 2000. Liberalization, Moral
Hazard in Banking, and Prudential REgulation:
Are Capital Requirements Enough? The American
Economic Review, 90(1): 14765.
Hulme, David. 2007. Is Microdebt Good for Poor
People? A Note on the Dark Side of Microfinance.
In Thomas Dichter and Malcolm Harper, eds.,
Whats Wrong with Microfinance, Warwickshire:
Intermediate Technology Publications, 1922.
Jain, Sanjay, and Mansuri Ghazala. 2003. A
Little at a Time: The Use of Regularly Scheduled
Repayments in Microfinance Programs. Journal
of Development Economics, 72:25379. http://
www.microfinancegateway.org/gm/document1.9.24935/18121_A_little_at_a_time.pdf
Kappel, Vivien, Annette Krauss, and Laura Lontzek.
2010. Over-Indebtedness and Microfinance

. 2011. Some Insights into Overindebtedness: Fresh Data from India. CGAP
Microfinance Blog, 29 March. http://microfinance.
cgap.org/2011/03/29/some-insights-into-overindebtedness-fresh-data-from-india/#more-1876
Krishnaswamy, Karuna, and Alejandro Ponce.
2009. A preliminary analysis of mass defaults in
Karnataka, India. Presentation slides.
Laibson, David. 1997. Golden Eggs and
Hyperbolic Discounting. Quarterly Journal of
Economics, 112(2) 44377.
Laibson, David, Andrea Repetto, and Jeremy
Tobacman. 2003. A Debt Puzzle. In Philippe
Aghion, et al., eds. Knowledge, Information,
and Expectations in Modern Macroeconomics.
Princeton, N.J.: Princeton University Press, 22866.
Laschelles, David. 2008. Microfinance Banana
Skins 2008. London: Centre for the Study of
Financial Innovation. http://www.cgap.org/gm/
document-1.9.2956/MF_BananaSkins2008.pdf

42

Laschelles, David, and Sam Mendelson. 2011.


Microfinance Banana Skins 2011. London: Centre
for the Study of Financial Innovation. http://www.
cgap.org/gm/document-1.9.49643/Microfinance_
Banana_Skins_2011.pdf

Rinaldi, L., and A. Sanchez-Arellano. 2005.


Household Debt Sustainability. What Explains
Household Non-Performing Loans? An Empirical
Analysis. Working Paper No. 5. Frankfurt:
European Central Bank, 28 June.

Lupica, Lois R. 2009. The Consumer Debt Crisis


and the Reinforcement of Class Position. Loyola
University Chicago Law Journal, Vol. 40: 10154.

Robb, Cliff A. n.d. College Students and Credit


Card Use: The Effect of Personal Financial
Knowledge on Debt Behavior. Dissertation,
University of Missouri, Columbia.

Luttmer, Erzo F. P. 2005. Neighbors as Negatives:


Relative Earnings and Well-Being. Quarterly
Journal of Economics, 120(3): 9631002.
Martinez, Renso, and Scott Gaul. 2011. Measuring
Cross-Indebtedness in Microfinance: Evidence
from Latin America. CGAP Microfinance Blog, 8
March. http://microfinance.cgap.org/2011/03/08/
measuring-cross-indebtedness-in-microfinanceevidence-from-latin-america/
McIntosh, Craig, and Bruce Wydick. 2005.
Competition and Microfinance. Journal of
Development Economics, vol. 78: 27198.
. 2009. What Do Credit Bureaus Do?
Understanding Screening, Incentive, and Credit
Expansion Effects. Calif.: University of California.
MicroSave. 1999. Loan Portfolio Audit Toolkit.
India: MicroSave. http://microsave.org/toolkit/
loan-portfolio-audit-toolkit
Morvant-Roux, S. 2009. Accs au microcrdit
et continuit des dynamiques dendettement
au: Combiner lanthropologie conomique avec
lconomtrie. Revue Tiers Monde, 196: 10930.

Roodman, David. Forthcoming. Due Diligence:


An Impertinent Inquiry into Microfinance.
Washington, D.C.: Center for Global Development.
Rosenberg, Richard. 1999. Measuring Microcredit
Delinquency: Ratios Can Be Harmful to Your
Health. Occasional Paper 3. Washington, D.C.:
CGAP. http://www.cgap.org/p/site/c/template.
rc/1.9.2698/
. 2010. Does MicroCredit Really Help Poor
People? Focus Note 59. Washington, D.C.: CGAP.
http://www.cgap.org/gm/document-1.9.41443/
FN59.pdf
Rutherford, Stuart. 2000. The Poor and Their
Money. India: Oxford University Press.
Rutherford, Stuart, and Md Maniruzzaman, S. K.
Sinha, and Acnabin & Co. 2004. Grameen II at the
End of 2003: A Grounded View of How Grameens
New Initiative Is Progressing in the Villages. India:
MicroSave. http://www.microfinancegateway.org/
gm/document-1.9.25948/22083_file_Grameen_II_
Status_report_Rutherford.pdf

Mullainathan, Sendhil, and Sudha Krishnan. 2008.


Psychology and Economics: What It Mean for
Microfinance? Framing Note No. 3. New York:
Financial Access Initiative. http://www.iamfi.com/
documents/PsychologyandEconomics.pdf

Rutherford, Stuart, and Sukhwinder Singh Arora.


2009. The Poor and Their Money: Microfinance from
a Twenty-First Century Consumers Perspective.
United Kingdom: Practical Action Publishing.

Rhyne, Elisabeth. 2010. Perplexed about OverIndebtedness? Center for Financial Inclusion
Blog, 4 May. http://centerforfinancialinclusionblog.
wordpress.com/2010/05/04/perplexed-aboutoverindebtedness/

Schicks, Jessica. 2010. Microfinance OverIndebtedness: Understanding Its Drivers and


Challenging the Common Myths. Centre
Emile Bernheim Working Paper No.10/047.
Brussels: Solvay Brussels School of Economics

43

and Management. http://ideas.repec.org/p/sol/


wpaper/2013-64675.html

storage/documents/Client_Protection_in_the_
Group_Lending_Process__INDIA.pdf

. Forthcoming. From a Supply Gap to


a Demand Gap? The Risk and Consequences
of Over-Indebting the Underbanked. Under
review for Jean-Pierre Gueyie, Ronny Manos, and
Jacob Yaron, eds., Microfinance in Developing
and Developed Countries: Issues, Policies and
Performance Evaluation, Macmillan.

Soman, D., and A. Cheema. 2002. The Effect of


Credit on Spending Decisions: The Role of the
Credit Limit and Credibility. Marketing Science,
21(1): 3253.

Schoell, Christopher. 2010. Turkeys Credit Card


Industry: Swipe Wisely. Perspectives on Business
and Economics, vol. 28: 3343. http://www.lehigh.
edu/~incntr/publications/documents/CreditCard.
pdf
Sinha, Sanjay. 2010. The Fruits of Unbridled
Growth. Micro-Credit Ratings International Ltd
presentation at Deutsche Bank overindebtedness
rountable, November.

Stearns, K. 1991. The Hidden Beast: Delinquency


in Microenterprise Credit Programs. Discussion
Paper Series no. 5. Washington, D.C.: ACCION
International.
Swadhar FinServe Pvt. Ltd. 2011. Swadhaars
Policy on Collection of Overdues from Delinquent
Clients and Other Recovery Policies. http://
www.smartcampaign.org/storage/documents/
SwadhaarFINAL_v2.pdf
van Aardt, Carel, and Abigail Moshoeu. 2009. A
Consumer Financial Vulnerability (CFV) Index for
South Africa Pretoria: University of South Africa,
Bureau of Market Research.

SMART Campaign. 2011. Collections Guidelines


for Financial Service Providers. http://www.
smartcampaign.org/storage/documents/Tools_
and_Resources/Collections_Guidelines_FINAL.pdf

Vandone, Daniela. 2009. Consumer Credit in


Europe: Risks and Opportunities of a Dynamic
Industry. Heidelberg: Physica.

SMART Campaign. n.d. Getting Started


Questionnaire: Client Protection Self Assessment
for Microfinance Institutions. http://www.
smartcampaign.org/tools-a-resources/2/41

Vogelgesang, U. 2003. Microfinance in Times


of Crisis: The Effects of Competition, Rising
Indebtedness, and Economic Crisis on Repayment
Behavior. World Development, 31(12): 2085114.

SMART Campaign. n.d. Smart Lending: Client


Protection in the Gramneen-Style Group Lending
Process (India). http://www.smartcampaign.org/

White, Michelle. 2007. Bankruptcy Reform and


Credit Cards. Journal of Economic Perspectives,
21/4 Fall: 17599.

No. 19
September 2011

Please share this


Occasional Paper
with your
colleagues or request
extra copies of this
paper or others in
this series.
CGAP welcomes
your comments on
this paper.
All CGAP publications
are available on the
CGAP Web site at
www.cgap.org.
CGAP
1818 H Street, NW
MSN P3-300
Washington, DC
20433 USA
Tel: 202-473-9594
Fax: 202-522-3744
Email:
cgap@worldbank.org
CGAP, 2011

The authors of this Occasional Paper are Jessica Schicks and Richard
Rosenberg. Schicks is a doctoral candidate at the Centre for European
Research in Microfinance at Solvay Brussels School of Management,
Centre Emile Bernheim, Universit Libre de Bruxelles. She is currently
on educational leave from an international strategy consulting firm.
Rosenberg is a senior adviser at CGAP. Sections 2 and 3 draw on
Schicks (2010). The authors are grateful for extensive research and
analysis by Jonna Bickel and Abigayle Seidel, as well as material on

credit reporting from Xavier Reille, Tim Lyman, and Gregory Chen;
on redress mechanisms from Kate McKee; and on financial capability
from Margaret Miller. The paper has been improved by review and
comments from Gabriel Davel, Gregory Chen, Robert Christen, Daryl
Collins, Tilman Ehrbeck, Marek Hudon, Kate McKee, Alexia Latortue,
David Porteous, Marguerite Robinson, David Roodman, Stuart
Rutherford, Jeanette Thomas, and Jacob Yaron. Of course, this does
not imply that they agree with everything in the paper.

The suggested citation for this Occasional Paper is as follows:


Schicks, Jessica, and Richard Rosenberg. 2011. Too Much Microcredit? A Survey of the Evidence on Over-Indebtedness? Occasional Paper
19. Washington, D.C.: CGAP, September

Center for Financial Inclusion


Publication No. 15

Over-Indebtedness of
Microborrowers in
Ghana

Over-Indebtedness of
Microborrowers in Ghana
An Empirical Study from a
Customer Protection Perspective

Jessica Schicks
CERMi (Centre for European Research in Microfinance)
Solvay Brussels School of Economics and Manangement (SBS-EM)
Universit Libre de Bruxelles (ULB)

November 2011

Contents
Preface

iv

Acknowledgments

Part I. Empirical Research on Over-Indebtedness in Ghana

Protecting customers against over-indebtedness


The microfinance market in urban Ghana
How to learn about over-indebtedness from a customer perspective

1
2
3

Part II. The Sacrifices of Microborrowers

Microborrowers in Ghana
Over-indebtedness, coping strategies, and sacrifices
The causes of over-indebtedness

6
7
12

Part III. Outlook for the Industry

16

What have we learned about over-indebtedness?


What are the prospects for microfinance in urban Ghana?
What are the policy implications for the wider microfinance sector?

16
17
17

References

19

Appendix I: The Client Protection Principles

20

Appendix II: List of Borrower Sacrifices

21

Figures
Figure 1. Measuring Over-Indebtedness by Customer Protection Standards
Figure 2. A Detailed Split of Loan Uses by Microborrowers
Figure 3. The Prevalence of Repayment Struggles among Microborrowers in Accra, Ghana
Figure 4. The Over-Indebtedness of Microborrowers in Ghana
Figure 5. The Acceptability and Frequency of Borrower Sacrifices
Figure 6. Potential Causes of Over-Indebtedness

5
7
8
9
10
14

Preface
Providing access to nance for those excluded from the formal nancial system was and still is the mission
of micronance. What has changed, however, are the main challenges on the way to success. In the past, the
nancial sustainability of the service was the main hurdle to be cleared. Today, responsibility in service provision claims more of our attention.
This is because something has happened that seemed virtually impossible only a few years back. In some parts
of the world where micronance services have expanded rapidly, the problem is no longer too little access to
nance, but too much. A phenomenon that has previously been known exclusively in industrialized countries
has reached the developing world: over-indebtedness.
Over-indebtedness has to be taken very seriously, particularly since in many countries where micronance
operates, customer protection and social safety nets are not well developed. Accordingly, a customer with an
unsustainable debt burden cannot le for insolvency and expect ofcial procedures to pave the way to a fresh
start. Few debt counseling agencies exist, and in most places there is no social security system to provide a
subsistence income for the family if a micronance client fails. In a nutshell, when a micronance client becomes over-indebted, she is on her own.
This is why micronance institutions carry all the more responsibility to protect their customers. These institutions have a social mission to assist their customers in improving their lives through access to nancial servicesprovided in a responsible way. To give guidance on how to live up to the standard of responsible nance,
the Smart Campaign (www.smartcampaign.org) developed the Client Protection Principles with widespread
industry participation. Prevention of over-indebtedness is one of these principles. Undoubtedly, most leading
MFIs are rmly committed to this principle, as evidenced by the fact that the overwhelming majority of the
MIX 500 largest MFIs have endorsed the Smart Campaign. However, commitment is not sufcient to put a
principle into practice. This is especially true when there are large knowledge gaps, as is the case with overindebtedness. This paper makes an important contribution to deepening knowledge on this elusive topic.
It is surprisingly difcult to agree on a rm denition of over-indebtedness, and just as difcult to determine
the point at which a specic individual becomes over-indebted. Financial service providers generally identify
over-indebtedness with chronic delinquency. They see debt stress as a clients inability to make regular loan
repayments, and move into action only when overdue collections or restructuring becomes necessary. In some
respects, this approach resembles the curative rather than preventive approach to medicine. More recently in
highly competitive markets, multiple loans from several different providers have become a concern.
In order to be effective at preventing over-indebtedness, it is essential to know more about the clients situation. This is exactly what the research presented here is aiming at, choosing the urban micronance market in
Ghana as an example. This study works from a client perspective, dening over-indebtedness in terms of the
frequency and severity of sacrices clients make to repay debts. Ghana was chosen as a market that has not
experienced an over-indebtedness crisis, but that has a number of signicant micronance institutions addressing the same clients. The study reveals that while the incidence of multiple lending and delinquency does not
indicate a market in crisis, there are nevertheless many clients in good standing who make serious and frequent
sacrices in order to repay debt. There is thus a gap between the providers and the clients in the perception and
experience of over-indebtedness. This gap challenges providers to take greater care in observing the situation
of clients, while avoiding the pitfall of restricting access to nance (and therefore pulling back from their mission). This is not an easy challenge.

iv

Even though these results are restricted to a single country and certainly do not answer all our questions, they
offer new insights from the clients perspective that must be incorporated into our understanding of overindebtedness. These insights will serve as valuable inputs to improve MFIs efforts, in Ghana and in similar
contexts, to make the prevention of over-indebtedness work in practice.

Elisabeth Rhyne
Center for Financial Inclusion at ACCION International

Eva Terberger
Independent FC Evaluation Unit
KfW Entwicklungsbank

Acknowledgments
Many organizations and individuals cooperated in making this study possible. The author would like to thank
the Independent Financial Cooperation Evaluation Unit of KfW Entwicklungsbank and the Smart Campaign for
their cooperation and nancial support. This paper has beneted from their contribution to research design as
well as from comments by Elisabeth Rhyne, Managing Director of the Center for Financial Inclusion, and Eva
Terberger, Gunhild Berg, Ron Weber, and Jan Schrader from KfWs Financial Cooperation Evaluation Unit.
KfWs Africa Department, especially the Competence Center for Financial Sector Development, also played
an important facilitating role. Equally important, Rich Rosenberg (CGAP) and Prof. Marek Hudon (Center for
European Research in Micronance) provided valuable input to earlier drafts. We gratefully acknowledge the
Marie Christine Adam Foundation and the German National Merit Foundation for their funding.
Special thanks go to our partner MFIs in Ghana who supported us in sampling their clients for the interviews:
ProCredit Ghana, Opportunity International Ghana, Sinapi Aba Trust, EB-ACCION, and Advans Ghana. We
also thank the National Bank of Ghana, the Ghanaian credit bureau XDS data, and our research assistants from
MEL Consulting. This project would not have been possible without them. Finally, we are extremely grateful to the Ghanaian borrowers who shared their sorrows and aspirations with us and gave us insight into their
nancial situations. Many have done so in the belief that telling us about their struggles would improve the
experiences of microborrowers in the future, and we share this hope with them.

An empirical study from a customer protection perspective

Part I. Empirical Research on Over-Indebtedness in Ghana


Since its beginnings, the micronance industry has aimed to provide access to nancial services for underserved micro and small entrepreneurs and other low-income households excluded from the formal nancial
market. The main challenge lay in expanding outreach and providing access to as many clients as possible.
While there was always some awareness that for single customers credit may not turn out well, there was
generally no concern about too much credit, but rather about too little. While expanding to reach more
and more low-income developing country populations, the sector has moved away from its original charity
approach and its focus on microenterprise lending to provide a broad range of nancial services. The majority
of micronance institutions aim to combine commercial viability with a social mission, and some institutions
even follow a purely commercial approach.
Recognizing that the growth and commercialization of micronance require conscious efforts to keep and
strengthen the industrys focus on serving and beneting above all its vulnerable client group, the Smart Campaign represents a global effort of micronance leaders to protect micronance customers. It has developed a
living set of Client Protection Principles1 and implementation guidelines and is in a constant dialogue with the
industrys leading institutions. Putting clients rst, the Smart Campaign aims to help the micronance industry
remain both socially focused and nancially sound.

Protecting customers against over-indebtedness


One of the most urgent customer protection principles is to prevent over-indebtedness. If over-indebtedness were
left to spread, it would represent a serious risk to the impact of micronance on borrowers lives; on the nancial
sustainability of micronance institutions (MFIs); and on the industrys reputation with governments, donors,
and investors. The 2011 Micronance Banana Skins report on micronance risks has ranked highest three risks
that are all closely related to the challenge of over-indebtedness: credit risk, reputation risk, and competition.2
An urgent need for research. The recognition of over-indebtedness as an industry priority and the willingness

of MFIs and investors to invest in protection against over-indebtedness is an important step toward client protection in micronance. However, for efforts to be effective, the industry needs a sound understanding of the
phenomenon; of the meaning of over-indebtedness from a perspective of micronance clients; of the prevalence of the phenomenon outside of crisis markets such as Bosnia, Nicaragua, and certain regions in India;
and of the mechanisms that may put clients at risk. To evaluate which prevention measures are most helpful
among the typical suggestions such as introducing credit bureaus, conducting literacy campaigns, or tightening
lending standards, we need to gain a sound understanding of what over-indebtedness is about for micronance
customers on the ground.
This paper is based on a study that the author has conducted as part of her PhD research.3 The research project
was designed and implemented with the support of the Independent Evaluation Department of the German
development bank KfW Entwicklungsbank and of the Smart Campaign, hosted by the Center for Financial
Inclusion at ACCION. It aims to address the most urgent questions about over-indebtedness from the viewpoint of the clients of micronance. Instead of working with the usual risk management indicators of overindebtedness, it uses an over-indebtedness denition based on the subjective experiences of microborrowers
with their loans.
The purpose of this paper is to inform decision makers who are promoting nancial inclusion in developing
countries about the over-indebtedness challenge and to support their efforts to protect poor borrowers. It aims
1. See Appendix 1 for the list of Client Protection Principles. After focusing on microcredit initially, they have been revised as of July
2011, to include all nancial products.
2. Lascelles and Mendelson (2011).
3. Watch the authors upcoming academic publications for more detail on over-indebtedness denitions, causes, and predictors.

Over-Indebtedness of Microborrowers in Ghana

to enrich the public debate with direct insights from


micronance clients. The paper:


of debt struggles in a
micronance market that, while starting to experience competition, is far from the exceptional
levels of debt found in recent crisis countries.
Suggests what the
may be and refutes some common assumptions
of the current over-indebtedness debate.
Offers guidance to the industry in how to think
about the phenomenon of over-indebtedness from
a customer protection perspective and how to
in other markets.
Identies
.

'

&

"

"

&

&

'

Pinpoints the
in repaying their loans.

Learning reaches beyond Ghana. While the results are

of course specic to the urban Ghanaian micronance


market where this study was conducted, we believe
that they can inform the over-indebtedness debate on
a much broader level. It being the rst empirical study
that analyzes the client perspective on over-indebtedness in such detail, the audience for this paper includes
MFIs in all countries that have reached the minimum
level of nancial inclusion and competition at which
over-indebtedness may emerge, investors in micronance, governments and regulators, as well as consultants and advisors in nancial inclusion. Given its
product-specic ndings, its contribution in canvassing client voices and its focus on a rather average
micronance market that is still not in crisis, the paper
may be of interest to MFIs in younger micronance
markets as well. It can contribute to product development that will improve the borrowing experience of
micronance customers anywhere in the world.

The microfinance market in urban Ghana


Ghana has developed an active micronance market
over the past years, serving 358,717 borrowers at the
end of 2009 with a gross loan portfolio of $US131.2
million.4 The MFIs in Ghana have started feeling signs
4. www.mixmarket.org. The site provides self-reports of limited
liability but represents the best available data source for this
high-level sector overview. Figures may somewhat underestimate the scope of micronance in Ghana, as not all institutions

of competition, especially in the most common urban


markets such as the center of Accra. Loan ofcers
report an increase in repayment difculties. In 2009,
two of Ghanas MFIs conducted a small study on multiple borrowing and found reasons for concern that
borrowers might start getting too indebted and combining loans from several lenders at the same time.5
Also, relying on several highly preliminary indicators
of potential over-indebtedness risks in micronance
markets, a study by the University of Zurichs Center
for Micronance describes Ghana as a market with a
medium to slightly higher risk of over-indebtedness.6
No signs of crisis. However, the penetration of micro-

nance in Ghanas low-income population is still rather


low: Of its working-age population below the poverty
line, only 9 percent have microloans. In Kenya, this
gure is at 14 percent, in Ecuador at 21 percent, and
in Mongolia at 51 percent.7 The Center for Micronance study mentioned above nds that Ghana has the
lowest micronance market penetration rate of all 12
countries in its sample. The FINSCOPE survey that
was recently conducted to measure nancial access
in Ghana nds that 44 percent of Ghanaian adults are
unbanked. They access neither formal nor informal
nancial services. In the urban markets of Accra, this
gure still amounts to 30 percent, leaving room for
further growth of the micronance sector.
Furthermore, Ghanas main MFIs stick to a rather
careful lending methodology. We have found no evidence of them deliberately poaching clients of competitors or pushing customers to take more or larger
loans than they may need. Instead, the MFIs in this
study routinely restrict loan sizes, rarely disburse the
full amount a customer applies for, and emphasize
detailed evaluations of repayment capacity. They
tend to limit their lending to business purposes, educating their borrowers to be careful with their loan
use and ensure that they will earn the returns to repay their debt. Borrowers also report strong messages from MFIs to be careful about multiple borroware reporting to the MIX Market. All data are reported as of
2009 unless indicated otherwise.
5. Grammling (2009).
6. Kappel et al. (2010).
7. These are high-level estimates based on data from the MIX
Market and CIA World Fact Book, counting micronance coverage only. They would probably be slightly lower if reduced for
borrowers with more than one loan.

ing. At the same time, multiple borrowing is simply


not an option for many: While interviewers tried to
motivate clients to tell the truth about their number
of loans, they often met with a lack of understanding:
A

u
d

l
f
p

d
t

m
t

g
w

k
u

B
D

C
V
F

h
B

P
P
V

b
C

g
C

P
B

8
p

B
P

C
f
G

E
a
Q

S
e
C

a
H

G
I

`
b

X
Q

Y
C

G
X

b
T

C
F

F
B

B
C

T
V
Q

S
E

(Schicks, 2010). The term


refers to a
certain permanence of problems over time to avoid
counting borrowers with one-off repayment problems as over-indebted (see below). To avoid imposing our own cultural views on clients in Ghana,
the only ones who can determine at what point
sacrices get
are borrowers themselves. Client protection is about preventing clients
from getting hurt, and suffering is a very subjective
r

D
F

C
R

ing with a focus on protecting clients, this research

A sacrifice approach to over-indebtedness. Work-

The over-indebtedness debate being relatively new


to the promoters of nancial inclusion, there still is
a lot of ambiguity about the concept. Before we are
able to understand the causes and consequences of
over-indebtedness in more detail, it is thus necessary
to agree on a measurement to identify which borrowers are over-indebted.

How to learn about over-indebtedness from


a customer perspective

G
f

As a result, this study works with this denition:

At the same time, it is good news that in the given


market segment the results do not portray micronance in Ghana as seriously overheated. With this research we are able to challenge some of the original
concerns about emerging lender overlap in Ghana, at
least among the most professionally managed lending
institutions. Most of their borrowers are not currently
going through extreme debt levels, vicious cycles of
borrowing from multiple sources, and high levels of
default. Nevertheless, we will show that there is a
need for improved customer protection measures to
avoid that microborrowers struggle with their loans.

In fact, from a client perspective, microborrowers


who manage to repay only at unacceptably high costs
such as going hungry, selling essential household
assets, or taking their children out of school should
count as over-indebted. While strong repayment incentives are an important success factor of the micronance methodology, from a client protection
point of view, there are limits to the sacrices borrowers should have to make. In developed countries,
insolvency regulations include a minimum existence
level, but most microborrowers do not benet from
such protection or live below a minimum existence
level to start with.

Learning from a normal market. This research analyzes a micronance market that is not yet suffering
from an explicit over-indebtedness crisis. The market in Ghana has reached a certain level of development, but one that may count as rather usual in
the micronance industry. Instead of highlighting
the downsides of extreme cases, the project informs
readers about the common experiences of microborrowers with their debt under ordinary circumstances.
It emphasizes that risks are an inherent feature in every borrowing and lending activity and that the micronance methodology can still improve and adapt
increasingly well to these risks.

clients replied.

project uses a client-focused denition of over-indebtedness, rather than common proxies like debtservice ratios or default/delinquency. It considers
debt-service ratios too imprecise and inexible to
take the individual circumstances of each borrower
into account. While some families regularly spare
30 percent of their income for debt repayments, for
others this can be a serious struggle. Default as an
approach for measuring over-indebtedness is equally inappropriate to our purposes: It reects only the
last stages of over-indebtedness, whereas problems
have almost always started much earlier. Default
assesses the consequences of over-indebtedness
from a risk management perspective but does not
take into account the struggles of microborrowers
who are often going to great lengths to pay back
their loans.8

8. See Schicks and Rosenberg (2011) for a discussion of overindebtedness denitions and Schicks (2010) for a more academic perspective.

Over-Indebtedness of Microborrowers in Ghana

experience. To truly understand how clients feel


and to what extent they may suffer, we therefore
rely on their subjective judgement.9

this study result from weighted analysis and are thus


representative of the research population.
Our detailed interview guide covered:

Our measurement in practice. To put this denition

The most innovative part of these interviewsand


the core of our analysisis the table of borrower
sacrices. Letting borrowers brainstorm rst and
then checking for additional items with the help of
a predened list, the interviewers obtained information about the struggles borrowers experience to
repay their loans and about all the sacrices they
make in relation to repayments. The list of sacrices
appears in Appendix 2. Respondents weighted each
individual sacrice by its frequency (
and
by its acceptability (
As a result,
someone might report that cutting down on his
food was totally acceptable, such as when it simply implied substituting cheaper food for meat. Another borrower might report that reducing his food
was not really acceptable, given that he was cutting
down to only one meal per day and staying hungry
most of the time.
Q

9. The appropriateness of over-indebtedness denitions depends


on their purpose. In spite of its value in a research context, a
subjective measurement is, for example, not suitable for juridical
procedures such as insolvency regulations.
10. We dened as microborrowers all MFI customers with
active personal loans below 5,000 Ghana Cedis (GH); rough
exchange rate 1 GH = 0.7 USD. For most of our partner
institutions, all loans in their portfolio are below 5,000 GH.
More than half of all loans in the sample are below 1,000 GH
and nearly all are below 2,000 GH. Interviews were usually
conducted at the respondents home or workplace.
11. The oversampling aimed at a sufciently large number of
observations from clients with serious repayment difculties,
given their low share of the population and an expected response
rate below that of the average borrower. With some MFIs, oversampling of group customers implied oversampling delinquent
groups rather than individuals. For the analysis, to ensure a
representative sample, all respondents were assigned individual
case weights according to their lending institution, delinquency
status, and lending methodology. There were no disparities in
gender to correct for.

Sociodemographic and economic information


about the borrowers and their households
Information about all their outstanding loans,
formal or informal
The detail of all sacrices the borrower experienced in the past year to repay loans
A test of nancial literacy
An experiment to test risk aversion
General questions about the experience of borrowers with MFIs.

From all active microborrowers in the Accra branches of our ve partner institutions, we drew a random
sample of micronance customers and contacted
them with an independent team of researchers for
anonymous interviews at a site of their convenience.10
We oversampled delinquent customers according to
the MFIs management information system (MIS)
data. For purposes of evaluation, we corrected for
this bias as well as the response rate of different
groups with sample weights.11 All data reported in

of over-indebtedness into practice, we conducted


531 structured questionnaires and 10 qualitative interviews with microborrowers in Accra, the capital
of Ghana and the heart of its micronance industry. The project was enabled by the participation
of Ghanas credit bureau XDS as well as ve of
Ghanas leading micronance institutions: ProCredit Ghana, Opportunity International Ghana, Sinapi
Aba Trust, EB-ACCION, and Advans Ghana. Together our partner MFIs make up nearly half of the
Ghanaian micronance market in terms of numbers
of customers (45 percent in 2008, 44 percent in
2009) and in terms of gross loan portfolio (46 percent in 2008, 43 percent in 2009) as reported on the
MIX Market.

Following the above denition, borrowers are overindebted if they struggle to repay their loans on time
and structurally make unacceptable sacrices. Sacri12. Once in past year, 13 times in past year, > 3 times but
not often, or Frequently in past year. For a respondent cutting
down on food at several points for a week at a time, instead of
every individual day, each week would count as one occurrence.
In this paper sacrices experiences >3 times count as repeated
or frequent sacrices.
13. Easily acceptable, Only just acceptable, Not really acceptable, or Not acceptable. In this paper we summarize the
rst two categories under acceptable and the latter two under
unacceptable.

Figure 1. Measuring Over-Indebtedness by Customer Protection Standards

Source: Schicks (2010).


a. No repeated experience required for unacceptable sacrifices of suffering an asset seizure, taking a new loan to repay, or selling/
pawning assets.
b. Either >3 unacceptable sacrifices, or 1 unacceptable sacrifice made > 3 times.

ces count as structural when they have been made


repeatedly, more than three times. Typical one-off
sacrices such as a seizure of assets, taking a new
loan to pay off an old one, and selling or pawning
ones assets to repay count as over-indebtedness
triggers even if the respondent experienced them
only once. In these cases, a one-off occurrence is
a sign of structural repayment problems.14 Figure
1 visualizes this measurement in a graphical funnel
entered by all 531 respondents. At each lter, those
borrowers who do not meet the respective criteria
for over-indebtedness drop out. In the end, only
those respondents on the right side of the funnel that
fulll all over-indebtedness criteria are counted as
over-indebted.

14. These sacrices either have long-term effects (e.g., help to


repay several installments) or are only triggered by long-term
problems (e.g., a seizure after 90 days delinquency). Only
unacceptable occurrences of loan recycling and therefore severe
sacrices act as a trigger, excluding, for example, simple loan
juggling for liquidity management purposes. As is true for all
sacrices, the severity of the loan recycling criterion may differ
in other research contexts/countries; our respondents in Ghana
try to avoid it at all means. In any case, this exception does not
substantially impact results.

A word of caution. Before we describe our ndings

in the next section, we would like to point out a potential source of misunderstanding: Our respondents
have indicated that they perceive their sacrices as
related to loan repayments. Also, 86 percent of them
state that they had suffered no or fewer sacrices
before borrowing. Nevertheless, our analysis does
not track causality. We are not saying and cannot tell
from our data that the microloans caused the borrowers struggles. Increasing nancial difculties
might also have been the reason for some to start
borrowing. Similarly, difculties in the course of the
loan term are not necessarily related to the overall
impact a loan has on borrowers lives in the long run.
In fact, the vast majority of our interviewees, even if
struggling, stated that they do not regret the amount
they borrowed, and many of them wish they could
have borrowed more.15

15. Part of this phenomenon may be due to a number of biases


that we will discuss in the next section. Nevertheless, we should
take the borrowers judgement seriously and not be too quick
with customer protection efforts that are in fact patronizing and
not to the borrowers benet, maybe overly reducing nancial
access again.

Over-Indebtedness of Microborrowers in Ghana

Part II. The Sacrifices of Microborrowers


The next section will report the voices of microborrowers in Ghana, providing unprecedented insights into the
experiences of borrowers struggling with their loans. We will rst describe typical Ghanaian borrowers as we
found them in our sample. We will then report which sacrices borrowers are making, what are the most common coping strategies to keep up regular repayments, and what experiences borrowers most suffer from. Those
borrowers that exceed the threshold for sacrices dened above are considered over-indebted. We will discuss
how these struggles relate to the borrowers perception of their loans. At the end of the section, the paper will
suggest some rst indications of what the causes of over-indebtedness might be.

Microborrowers in Ghana
The microborrowers in our Ghanaian sample can generally be considered typical micronance customers: 72
percent of our respondents are female, most of them married. The most frequent ethnicities are those predominant in Accra, especially the Akan (65 percent), the Ewe (14 percent), and the Ga (12 percent). Nearly without
exception, our microborrowers main source of income is self-employment. The majority of businesses (82
percent) are active in trading, either importing goods from abroad or buying wholesale and selling in small
convenience stores or on the markets. Services such as hairdressing and manufacturing account for only a
small share of microbusinesses. Only seven borrowers (1.3 percent) have permanent employment as salaried
workers.
Poor but not extremely poor. The sample conrms that, while living vulnerable lives on low incomes, the typi-

cal micronance clients are not extremely poor. On average, a respondent household lives on 500-650 GH per
month, about US$400. With ~ve members per household, this is clearly above the local poverty line (World
Bank, 2011) or the international poverty line of US$2 per person per day. The median of personal assets of all
households in our sample of microborrowers amounts to ~6,000 GH. More than half of the loans (59 percent)
are less than 1,000 GH at the time of disbursement and most of them (87 percent) are below 2,000 GH.
Careful and experienced borrowers. Once borrowers have access to an MFI, they tend to borrow repeatedly,

many following the traditional model of loan cycles in group lending or coming back for individual loans
demonstrating that they value the services of MFIs. In our sample, 68 percent of borrowers have previous
experience with borrowing from (semi-)formal institutions, 41 percent having borrowed for several years in a
row. Nevertheless, the group of borrowers who have had only a single semi-formal loan so far (32 percent) is
a relevant portion of the sample.
The sample is about evenly split between group (48 percent) and individual (52 percent) loans. However,
practices differ between lending institutions. NGOs and lenders with NGO roots give more group loans, while
some of the more commercial MFIs focus almost exclusively on individual lending. In terms of loan use, the
emphasis on enterprise loans is very strong; 96 percent of loans are used at least partly for business purposes.16
Repeating the messages of lenders, many borrowers claim that they would never do otherwise, as consumption loan uses make repayments far too difcult. They tend to be quite curious but rather incredulous about
interviewers potentially thinking otherwise. Nevertheless, 27 percent of all borrowers do admit thatbesides
investing in their businessthey also used part of their loan for a personal or household use. Of these, education (37 percent) and day-to-day consumption (22 percent) are by far the most common loan uses. They are
followed by expenses for housing or land (10 percent), emergencies (9 percent), special consumption such as
a mobile phone (9 percent), and the acquisition of durables for the household (6 percent).

16. This is approximate information only, as it does not account for the fungibility of money, borrowers investing less of their household
resources in their business when they get the loan and thereby implicitly cross-subsidizing their consumption. Also, many borrowers
report not investing the full loan amount but holding a substantial part of it back for the rst repayments, without perceiving or reporting
this as using their loans for consumption purposes.

Figure 2. A Detailed Split of Loan Uses by Microborrowers


Most common
secondary loan uses

Business
only 68.6%

27.2%

Mainly business
with a secondary
loan use

4.2%
Mainly
household

Borrowers are either very careful about taking multiple loans or simply constrained in terms of access:
The maximum number of loans reported by any borrower at the same time is three. Only 8 percent of
borrowers take loans from more than one lender in
parallel, of which less than 1 percent borrow from
three lenders at the same time. While self-reports
may understate difcult-to-admit phenomena such
as multiple borrowing, delinquency, and sacrices,
we do not nd evidence of borrowers lying about
their loans, at least with regard to formal borrowing:
Their information is consistent with what we nd in
the MIS of the participating MFIs. Instead, the low
level of multiple borrowing may to a large extent
be due to the sound lending decisions of the participating institutions and may thus not apply to other
MFIs in the same lending market.
Helpful loans. While this study is not an impact
study, we asked borrowers to provide a subjective
assessment of their returns on investment. Out of all
borrowers that invested their loans in their business,
43 percent state that their earnings increased signicantly and on a regular basis due to the investment;
40 percent claim increases in income that were not
sufcient or not stable enough to cover repayments,
at least over the period of the loan. The remaining 17
percent of respondents did not experience a permanent increase in earnings as a result of their loan.

Education
Day-to-day consumption
Housing expenses or land
Emergencies
Special consumption (e.g., mobile)
Household durables (e.g., fridge)
Give money to someone else
Social purpose (e.g., wedding)
Pay off a loan (own or for someone)
Other

%
37
22
10
9
9
6
3
2
1
2

benets of borrowing were absolutely obvious and


repayments sufciently easy to consider our concerns
about the distress of borrowers rather strange.

Over-indebtedness, coping strategies, and


sacrifices
Taking a loan always comes at a certain risk. Especially for the typical micronance borrower who
lives on a low and volatile income, repaying a loan
is not necessarily easy. It implies regularly assembling the cash for repayments and managing this
cash demand among many other competing needs
for money.
In our sample in Ghana, 26 percent of all respondents nd it easy to repay their loans. Some do not
make sacrices at all (17 percent of total sample);
others make only minor sacrices that do not give
them an overall sense of struggling. However,
many borrowers experience repayments as a challenge. About one-third of borrowers are struggling
to repay at certain occasions, 26 percent struggle
regularly over the course of the loan but not all the
time, and 17 percent permanently struggle with (almost) every single installment. Figure 3 displays
the prevalence of repayment struggles among our
respondents.
High tolerance for sacrifice. The tolerance of Gha-

On the whole, while we were looking for the many


challenges of debt repayment in this study, it was
comforting to see how many borrowers were surprised
at our questions. For quite a signicant number, the

naian microborrowers for sacricing for their loan


repayments is rather high. Ghanaians have a strong
sense of obligation and, for many, making every
possible effort to keep their repayment records clean
Over-Indebtedness of Microborrowers in Ghana

Figure 3. The Prevalence of Repayment Struggles among Microborrowers in Accra, Ghana

Not struggling

26%
31%

Struggling rarely
Struggling regularly

26%
17%

Struggling (almost) always

(Figure 4) They struggle to repay their loans on time,


and they repeatedly make unacceptable sacrices.

customer protection denition of over-indebtedness


to our sample population in Ghana, we nd that
over-indebtedness from the clients point of view is
a matter for concern. While delinquency levels are
still acceptable among our partner MFIs and multiple
borrowing is hardly prevalent in the sample, many
customers struggle with their repayments

Also, as in many countries, speaking about personal


nancial hardships is difcult in the Ghanaian cultural context. Our interviewers made every effort to
create an atmosphere of trust, guarantee absolute
anonymity to the respondents, and show a personal
interest in their experiences. As a result, many borrowers opened up and told us their complex personal
stories of indebtedness. Nevertheless, we believe
that borrowers rather understated their sacrices on
the whole, hesitating to admit that sacrices were
unacceptable to them.

Concerns about over-indebtedness. If we apply our

is a question of honor. Also, borrowers consider


sacrices acceptable because choosing this specic
hardship is still better than the consequences of default (e.g., assets seized, shop closed, and creditworthiness lost). Some borrowers state that pretty much
anything would be acceptable for them, once they
have incurred the obligation to repay a loan. We met
borrowers who go hungry or take their children out of
school and still do not want to complain. Ninety-two
percent of our respondents identify with the claim I
do everything I can to repay on time, prioritizing
loan repayments above most other cash needs. Only
7 percent admit that they sometimes prioritize other
urgent expenses over a loan repayment, and only 1
percent admit sometimes paying late even if they do
have moneyusually in cases where they travel and
cannot come to group meetings or MFI premises on
the exact day of payment. One respondent explained
her willingness to sacrice as follows:

The Ghanaian micronance market does not appear


overheated or in crisis. Still, this level of borrower
struggles is worrisome from a perspective of customer protection. The micronance industry will
need to nd ways to address these client concerns.
This is all the more true as not all but some of these
struggling borrowers are likely to slip into delinquency at some point: Among those that are not yet
over-indebted according to the sacrice-based denition, very few borrowers fear that they wont be
able to keep up their payments at their current level
of sacrices. Among the over-indebted, in contrast,
8 percent were already convinced at the time of their
interview that their level of efforts was not sustainable and admitted that they wouldnt be able to meet
their future loan obligations. Protecting customers
from getting into the kind of repayment difculties
that cause over-indebtedness in terms of sacrices is
therefore also good risk management.
Coping strategies. Of course, sacricing starts much

earlier than at the level we call over-indebtedness.


The next paragraphs will examine the sacrice ex-

Figure 4.The Over-Indebtedness of Microborrowers in Ghana

a. No repeated experience required for unacceptable sacrifices of suffering an asset seizure, taking a new loan to repay, or selling/
pawning assets.
b. Either >3 unacceptable sacrifices, or 1 unacceptable sacrifice made > 3 times.

periences of Ghanaian borrowers in more detail.


Figure 5 provides an overview of how many times
the borrowers in our sample reported each of the
common sacrices. For each sacrice, it details the
percentage of borrowers who made this sacrice and
considered it unacceptable as well as the percentage
who made the respective sacrice frequently over
the course of one year. The graphical representation
of how many borrowers made which sacrice and to
what extent they considered their efforts acceptable
shows that, as a collective, microborrowers behave
according to rational expectations: They make those
sacrices rst that, on average, are easiest to accept,
and resort to harder measures only when they have
no other option.
When repaying a loan gets challenging, the typical
coping strategies of microborrowers in Accra are
to work harder in their businesses (61 percent of
all borrowers), to postpone other expenses (54 percent), andif availableto deplete existing savings
(34 percent). These are the types of efforts that most
borrowers consider acceptable: 68 percent of those
who had to work harder considered their efforts acceptable, as did 67 percent of those who postponed
important expenses and 63 percent of those who depleted their savings. Therefore, they choose to meet
their challenges with these less painful measures

rst and apply them rather frequently as regular coping strategies. Out of the borrowers who increased
their work load, 72 percent did so more than three
times in a year. For those who postponed expenses,
60 percent did so more than three times. If depleting savings is a less frequent strategy (28 percent),
this is most likely because after the rst occurrence
there are simply no savings left and it is difcult for
most microborrowers to build them up again in better times.
If these sacrices are classied as coping strategies,
that is because they are common reactions by borrowers to debt problems and not necessarily a sign
of over-indebtedness. Most borrowers easily accept that they have to make an extra effort for the
purpose of their loan. Some may have even made a
conscious choice of working harder in their business
when they get the loan, hoping for their efforts to
bear fruit. These cases are not yet a reason to worry
about over-indebtedness.
Of course, the spectrum of subjective experiences is
broad, and in some cases these same coping strategies reach the extent where we would call them serious sacrices. For those borrowers, for example,
who consider working harder unacceptable, this
does not imply that they are lazy: Some borrowers

Over-Indebtedness of Microborrowers in Ghana

reported that they had started working day and night,


hardly nding time to sleep and nding no time at all
to take care of their children. Working at night can
come at enhanced personal danger. Others reported
that they could no longer go to church and participate in their local communities, because they had
to continue working even on Sundays. They had to
continue working when they were seriously ill, simply because they could not afford losing their daily
income. In these cases, a useful coping strategy may
have been overextended under the pressure to keep
up repayment performance, and a legitimate effort
has turned into a serious sacrice.
Serious sacrifices. The more painful sacrices are,

the more borrowers try to avoid them. When the


easier coping strategies are no longer sufcient to
meet repayment deadlines, many borrowers resort to
cutting back on food. For those who do sacrice on
food, this usually becomes a repeated experience.17
Twenty-eight percent of all borrowers who have to
cut down on eating consider the sacrice acceptable,
but most of them perceive it as a real hardship.
17. For 63 percent of those who sacrice food, it occurs during
more than three periods throughout the year.

Another common step for Ghanaian microborrowers


who can no longer cope with their debt is to ask their
friends and family for help. It is clearly a measure
that most borrowers prefer to avoid, being similarly
difcult to accept as going hungry. It seems that the
feeling of dependence on others, the obligation to
return favors in the future, and the shame of admitting that one cannot cope alone are strong impediments to asking others for support. The barriers to
discussing nancial matters with others, as well as to
admitting that one is in debt, might pose other hurdles. Of those who do ask others for help, only 21
percent resort to this option more than three times in
a yearthe availability of support is probably limited, and when asking too often a borrowers personal
network risks getting overextended.
Sometimes, when there are no other options left, borrowers resort to taking a new loan elsewhere to repay
an old one, selling or pawning some of their household
or business assets, or taking children out of school because they cannot afford the school fees or need their
children as a workforce too urgently. Only 10-20 percent of the microborrowers experiencing such a situation still consider their sacrices acceptable. This is
the stage where some borrowers experience serious

Figure 5. The Acceptability and Frequency of Borrower Sacrifices


NUMBERS (%) OF
BORROWERS MAKING
EACH SACRIFICEa
325 (61%)
240 (54%)
179 (34%)
96 (18%)
67 (13%)
51 (10%)
26 (5%)
20 (4%)
20 (4%)
15 (3%)
14 (3%)
4 (1%)

SACRIFICES
Work more than usual
Postpone important expenses
Deplete savings
Reduce food quantity/quality
Use family/friends support
Suffer psychological stress
Reduce education
Borrow anew to repay
Sell or pawn assets
Feel threatened/harassed
Suffer from shame or insults
Seizure of assets

a. Out of all borrowers in the sample.


b. Out of the borrowers who made each respective sacrifice.

10

% OF BORROWERS
FINDING SACRIFICES
UNACCEPTABLEb
32%
33%
38%
72%
73%
80%
80%
85%
90%
100%
100%
100%

% OF BORROWERS
MAKING SACRIFICES
FREQUENTLY (>3 TIMES)b
72%
60%
28%
63%
21%
53%
53%
50%
40%
44%
23%
33%

psychological stress from the pressure and the fear of


not being able to repay their loans. Unfortunately, 19
percent of our respondents have experienced one or
more of these tough sacrices over the course of one
year. Here is what one borrower said in describing her
psychological stress:

Finally, the indebtedness experiences that no borrower is willing to accept are being threatened or harassed, suffering from shame and insults, and losing
their assets in a seizure. This shows how personal
honor and integrity can be more important than nancial hardships and material sacrices. Nevertheless, 5 percent of the borrowers in our sample were
not able to avoid this level of sacrice, some even
experiencing these repeatedly. As an inherent risk
to any lending activity, some customers will always
end up unable to repay despite their best efforts. As
much as collections have to be strict, our ndings
underline the importance of treating those borrowers
with dignity.

reason might be that for many borrowers these resources are simply not available or, as good tools of
liquidity management, have already been sacriced
at an earlier stage of debt struggles. As a result, those
borrowers are more likely to slip into arrears.
Independently of their status with regard to overindebtedness and delinquency, we nd that female
borrowers differ from their male peers in several
ways: Men resort to external solutions more often,
taking new loans to repay old ones and selling or
pawning assets. This might partly be due to their
easier access to alternative loans and assets. In contrast, women have a higher likelihood of reducing
their food intake. They are probably the ones that
are responsible for the familys meals and in difcult
times try to keep up their childrens eating habits by
compensating for problems with their own nutrition.
For all other sacrices, wherever we have sufcient
observations to judge, we do not nd a difference
between the experiences of men and women. Equally, in spite of a tendency for less delinquency and
more sacrices in groups, there are no signicant
differences in delinquency and the total amount of
sacrices between group customers and individual borrowers. However, group customers deplete
their savings more often, rely more on the support
of friends or family, and suffer more psychological
stress than individual customers. These differences
may result from peer pressure.
Neverending optimism. As we have indicated earli-

Experiences differ among borrowers. Comparing

the sacrice situation for borrowers who have been


delinquent at any point during their current loan to
those who have managed to always repay on time,
we nd that that those delinquents suffer from sacrices much more frequently.18 This is in line with
our customer protection approach to understanding
over-indebtedness that considers (nonfraudulent)
delinquency as a late stage of over-indebtedness.
Depleting their savings and relying on the help of
friends and family are the only sacrices that delinquent borrowers do not make signicantly more
often than those with a clean repayment record. The
18. For borrowing anew to repay, selling or pawning assets,
asset seizures, and suffering insults or harassement, we do not
have sufcient observations to conrm this relationship statistically.

er, our respondents value their access to loans, even


in spite of their sacrices. Many of them are hoping
to borrow again and still wish that MFIs would give
them larger loans. Only 4 percent of borrowers admit that they regret the amount of debt they took. We
believe that the reasons for this seemingly inconsistent picture are threefold.
On the one hand, this is the point where personal
embarrassment comes most strongly into play when
the research methodology is based on interviews.
Admitting to suffering serious sacrices may be difcult, but at least in showing that they are doing all
they can to live up to their obligations, respondents
can keep up their self-respect. It is in admitting that
they have made a mistake and borrowed too much
that borrowers are really embarrassed. Instead of
Over-Indebtedness of Microborrowers in Ghana

11

admitting their regret about an error of their own


judgment, they prefer to rationalize that they were
simply unlucky this time and that their decision was
justied as they could not foresee their bad luck.
On the other hand, Ghanaians seem to live with a
strong sense of optimism; probably over-optimism
in some cases. Many borrowers refuse to answer our
backward-looking question regarding their current
loan but focus instead on their determination to try
again and be more successful with their next loan. The
fact that a loan has been a negative experience does
not spoil their fundamental belief that more investment will help in the long run. They even argue that
now that their troubles have worsened over the rst
loan, they need a larger loan even more urgently.19
Both motivations lead to similar arguments that attribute the repayment problems to a factor other than
the decision to borrow: Taking a loan was correct,
but

The investment went wrong or started paying


off too late compared to the installments
An emergency made repayments difcult
The interest was too high or installments too frequent
The amount disbursed was too low for the required investment or was disbursed too late.

Based on these arguments, borrowers do not regret


their loans as such but they regret the specic problem they attribute their struggles to. One woman told
us that she wont manage to repay and never wants to
borrow againbut instead of regret about her loan
she still indicated the wish to have borrowed more,
hoping that with more credit her business would
have been more successful.
Finally, borrowers lives are shaped by a constant
need for cashthey simply value their creditworthiness and access to credit so highly that they prefer to
keep up the borrowing relationship to the MFI even
when borrowing experiences can be more painful
19. This is in line with insights from behavioral economics that
humans often do not make fully rational decisions, inter alia,
when deciding to borrow (see, e.g., Banerjee and Duo [2007]
on microborrowers nancial lives). To a certain extent it represents a call for increased nancial literacy.

12

than immediate loan impact may justify. Borrowers


may still be better off with access to nance than
without. Some Ghanaians seem to perceive a loan
as something so precious that they would take it
without reection, no matter what the cost. This is
a strong message not to overreact to over-indebtedness: Customer protection should aim to reduce borrower sacrices. But customer protection measures
that overly reduce poor peoples access to loans may
not always be the right response. Instead, a redesign
of loan products seems to be called for.

The causes of over-indebtedness


While lives in poverty are a continuous struggle and
we will have to accept that repaying loans is not always easy for the target clientele of micronance, we
are concerned to nd one-third of borrowers struggling repeatedly and heavily to repay their loans.
At the same time, borrowers clearly indicate their
continued need for access to credit. Simply reducing
microlending therefore may not be to the average
clients benet.
To develop appropriate customer protection mechanisms, the industry will need to understand much
better what is causing the borrowers difculties
and what the contribution (as well as alleviating effect) of micronance is. The answers may well be
country and context specic. This section develops
some rst hypotheses of what the causes of overindebtedness are in Ghana and which borrowers are
particularly at risk.
Some groups experience more over-indebtedness
than others. Looking at the distribution of over-in-

debtedness across different subgroups of our population, without paying attention to statistical signicance just yet, we nd an above-average share of 44
percent over-indebtedness among minority ethnicities.20 Potentially, ethnic outsiders face additional
economic challenges in terms of business opportunities and safety network. There also is slightly more
over-indebtedness among borrowers living without a
marital partner (34 percent) than among married bor20. Minority borrowers indicated their ethnic background as
Gurma or Other. The majority ethnic groups (Akan, DagboneDagomba, Ewe, and Ga) have an over-indebtedness share of 29
percent.

rowers (28 percent). Health problems (27 percent),


including severe health problems (2 percent), are
more common among the over-indebted than among
those who are managing their debt without serious
sacrices (22 and 1 percent, respectively). Thirty-six
percent of the borrowers with an educational level at
or below primary school are over-indebted, versus
only 28 percent of borrowers with middle school
or higher education. Similarly, over-indebtedness is
higher among people who borrow from several lenders in parallel (44 percent versus 29 percent among
single borrowers) and among rst-time debtors who
lack previous borrowing experience (35 percent vs.
28 percent among experienced borrowers).
However, the ndings above do not stand the test
of statistical signicance, mostly because the subgroups where over-indebtedness is more prevalent
are too small to allow for reliable econometric results. Some of them might only be coincidental differences in our sample. In the next paragraph, we
use simple statistical tests to analyze which factors
are related to over-indebtedness and nd that, in particular, economic and loan-related factors, but also
some personal characteristics, are correlated with
over-indebtedness.
Several economic and personal factors are related
to over-indebtedness. Of all our potential over

indebtedness causes, the factor with the strongest


statistically signicant relationship to over-indebtedness is a low return on an investment loan,22 followed closely by partial or total loan use for nonproductive purposes. This is not surprising: It is harder
to repay a loan when it does not cause any or only
low investment returns (Gonzalez, 2008). Similarly,
we nd a comparably strong relationship for adverse
shocks to a borrowers nancial situation, especially
sudden drops in income and, for a lack of assets,
most importantly savings that could serve as a buffer
21. Using contingency analysis with Chi Square and Cramers V
for our main over-indebtedness measurement and with Kendalls
Tau-c for an alternative approach where we have broken the
measurement down into four categories of severity. Upcoming
academic publications of the author will provide more detailed
econometric analysis. Correlations do not prove causality.
22. Self-reported investment returns by borrowers in three
categories (earnings increased signicantly and regularly; the
increase was not sufcient or not stable enough to cover repayments; no permanent increase in earnings from the investment).

for difcult times. Living on a low income comes


up as a relevant factor in only one of the tests and
volatility of income not at all.
In addition to these economic factors and hardships
that may well be typical causes for making borrowers struggle with their loan repayments, we nd that
borrowers personal nancial literacy is related to
their over-indebtedness risk. The lower their score
on our nancial literacy test, especially in the section
with debt-specic questions, the higher the share of
borrowers that is over-indebted. This is in line with
ndings from Godquin (2004) that access to literacy
services can improve repayment rates. Finally, we
conducted an experiment on the risk preferences of
our respondents.23 The resulting risk aversion score
is correlated to our main over-indebtedness measurement, but does not provide consistent ndings
about the direction of the relationship. It seems that
at least extreme levels of risk aversion are rather
counterproductive and related to higher over-indebtedness risks. This might be due to overly restrained
business decisions but could also be a question of
reverse causality, with debt problems reducing the
willingness to take risk.
In spite of the importance of nancial literacy and
previous ndings that delinquency goes down
with the number of past loans a borrower has had
(Schreiner, 2004), in Ghana a lack of borrower experience does not seem to increase over-indebtedness risks. Also, our ndings challenge the assumption that over-indebtedness corresponds to multiple
borrowing (McIntosh and Wydick, 2005; Roesch
and Hlis, 2007; Reille, 2009; Venkata and Veena
Yamini, 2010). At rst sight there is a higher share
of over-indebtedness among cross-borrowers, but at
least at the low level of multiple borrowing in our
sample, taking more loans is not statistically correlated to over-indebtedness. Figure 6 provides an
overview of factors related to over-indebtedness in
a contingency analysis. Cramers V and Kendalls
Tau-c indicate the strength of the relationship that is
relevant in all cases except total assets, though each
23. Borrowers could choose from two bags of marbles with
higher/lower chances of getting lower/higher returns. The probabilities and payoffs varied over ve rounds. It is not guaranteed
that such a game can measure actual risk behavior of borrowers
in their businesses.

Over-Indebtedness of Microborrowers in Ghana

13

Figure 6. Potential Causes of Over-Indebtedness


CRAMERS V
Income
Income volatility
Assets
Savings
Adverse shocks (esp. to income)
Borrowing experience
Multiple/Cross-borrowing
(Partial) Nonproductive loan use
ROI (if only invested)c
Financial (esp. debt) literacy
Risk attitude

0.170***

n/a
0.156**
n/ab
n/a
n/a
0.173***
0.211***
0.180***
0.191***

KENDALLS TAU-ca

0.056*
0.111***
0.199***

0.161***
0.102**
0.126***

RELATIONSHIP
CONFIRMED
()

()

a. Tau-c based on an ordinal over-indebtedness definition for robustness check.


b. Chi Square invalid due to lack of observations in > 20% of contingency table. However, highly significant also in other robustness
checks.
c. Relationship stronger (always 1% significance) for all investment loans. This analysis avoids collinearity with loan use excluding
even partial nonproductive loan use.

of the factors is obviously just one inuence factor


among several others. It is probable that nonliquid
assets are of little help and liquid savings make a
bigger difference to borrowers struggles. One star
denotes a 10 percent statistically signicant level,
two stars 5 percent, and three stars 1 percent.
Lending institutions may also play a role. In addition
to the inuence that borrowers and their circumstances have on over-indebtedness risks, the MFIs
and their products and policies may also play a role.
Indeed, many borrowers complain that their repayment difculties are due to product features such as
the high interest rates on microloans, loan amounts
being too small, or grace periods being too short for
investments to start paying off.

While most borrowers consider their MFIs treatment as fair, their terms and conditions as transparent, and their evaluation of repayment capacity as
fair and sound, borrower opinions are split when it
comes to specic product features. Many borrowers state that MFIs disburse loans too late for their
investment opportunities (53 percent), offer tooshort maturities (51 percent), insist on too-frequent
installments (47 percent, mainly those that are on
weekly instalment schedules),and do not provide
14

fair rescheduling options for borrowers in honest


difculties (58 percent). These are clear messages
to product developers in MFIs, not only from a customer protection, but also from a customer satisfaction point of view. Nevertheless, this is not a clear
call for product changes in favor of those who currently complain: Even if 51 percent of borrowers
consider maturities too short, 49 percent do not want
longer maturities that would increase their interest
charge on a given loan amount and would delay their
access to a follow-up loan. It is therefore rather a call
for more exibility. The standardized micronance
product offer does not match every borrowers cash
ows and makes repayment more difcult for many
of them than is actually necessary.24
Furthermore, borrower sacrices are often made
out of a constant fear of the consequences of delinquency, especially of tough collection practices.
Once borrowers reach the stage of arrears, harsh
collection practices make their over-indebtedness
experiences worse. Augsburg and Fouillet (2010)
describe allegations of overly harsh collection prac24. Flexibility requires careful experimentation to keep up
strong repayment incentives and needs to be weighed against the
increase in operating costs that may result from the additional
complexity.

a second seizure when he had just started rebuilding


his shop. When he nally repaid his dues, the MFI
returned his assets only partially and in a deteriorated condition.

25. For measures that are specic to the group lending methodology, percentages apply to group borrowers only. They may
understate borrower expectations as not all respondents were
familiar with the effects of delinquency in group lending.

Ghanaian borrowers25 expect that delinquents will


experience pressure from their fellow group members (64 percent), being subjected to public blame
(40 percent), group meetings getting extended for
hours (35 percent), and nally and most prominently, their assets being seized (72 percent). While
collections need to be strict and a seizure of assets
is an appropriate step in case of default on a collateralized loan contract, there seems to be widespread
confusion in Ghana: Some loan ofcers seem to use
threats of asset seizures as a tool to pressure borrowers who lack the nancial literacy to properly
understand their rights and obligations. According
to client interviews, they sometimes pretend to be
able to seize assets much earlier than 90 days after
missed payments or to be able to seize any belongings of the borrowers. One woman reported that her
MFI had seized her main business assets and personal kitchen equipment although she had not been
aware of her loan contract including any collateral.
Another borrower said that the MFI had seized all
of his merchandise, exceeding the 1,800 GH outstanding value of his loan and still coming back for

tices in the Indian context. In our sample, 49 percent indicate that loan ofcers threaten borrowers
or use abusive words, although only 24 percent say
that MFIs are generally impolite in the collections
process. This seeming contradiction probably results
from the general politeness of loan ofcers except
in especially difcult repayment situations. Borrowers fears may be linked more to hearsay and the experience of others than their own. After all, only 3
percent of borrowers listed threats or harassment as
their own sacrices in the past year. Also, some borrowers do not classify threats as impolite but consider them necessary loan ofcer behavior.

Our ndings indicate a clear call for exibility of


micronance products, for fairness in MFI policies,
and for improving clients understanding of contract
terms and nancial matters in general. Striking the
right balance between setting strong incentives and
accommodating customer needs will require careful experimentation with extending and developing
the microlending methodology in the future. More
research should be done on the relationship of product features and other lender characteristics such as
growth rates and protability to over-indebtedness.
It will inform the development of solutions to the
over-indebtedness challenge so that remedies can
address the most important levers.

Over-Indebtedness of Microborrowers in Ghana

15

Part III. Outlook for the Industry


What have we learned about over-indebtedness?
This paper has offered insights into the repayment experiences of 531 microborrowers in Accra, Ghana. It has
revealed the sacrices that many borrowers are making to repay their loans and used a sacrice-based denition of over-indebtedness to assess the extent to which borrowers are over-indebted from a client protection
perspective.
In fact, we learned that only 26 percent of our respondents manage to repay their loans without struggle and
make no or only minor sacrices. The other 74 percent do struggle, and for 17 percent of all borrowers struggle
is a permanent companion for (almost) every installment they make. At the same time, we found that borrowers in Ghana have a rather high tolerance for such sacrices. While personal experiences and struggles are
of course subjective, many borrowers express a strong will to meet their repayment obligations, keep a clean
credit history, and invest whatever it takes to maintain their honor and creditworthiness.
When repayments get challenging, the average borrower rst resorts to easily acceptable coping strategies such
as working harder, postponing important expenses, and depleting savings if there are any. These reactions to
repayment difculties are very common and not necessarily a reason for concern. When the coping strategies
above are not sufcient for a borrower to meet the loan obligations, sacrices become more serious. Borrowers
cut down on food, try to nd family or friends who can help them out at least temporarily, and sometimes take
their children out of school. They only partially consider this level of sacrice acceptable. With increasing debt
problems, borrowers suffer from psychological stress, get threatened or harassed, and suffer from shame and
insults. In the nal stages of repayment problems, borrowers resort to taking new loans to pay off old ones or
selling or pawning their personal and business assets. The ultimate experience of debt problems when a borrower actually defaults is of course losing ones belongings in a seizure of assets. These experiences are less
frequent, but also more severe, and are hardly ever acceptable to microborrowers.
Applying our sacrice-based denition of over-indebtedness to our population sample from Accra, we found
that about 30 percent of all borrowers sacrice to the extent that they fulll all our over-indebtedness criteria.
In our sample of strong lending institutions, delinquency and multiple borrowing are low, but one-third of
borrowers struggling that seriously to keep up their loan repayments is reason for concern. We do not argue
that microloans are necessarily what caused these sacrices, but we do call for attention to the difculties
that repaying loans imply for MFI customers. At the same time, the paper found strong optimism of most microborrowers about taking future loans. This nding represents a warning to the micronance community not
to overreact: While it should certainly address the challenges of loan repayment in its client protection efforts,
borrowers value their access to microloans very highly and do not want to lose it over protective measures.
In looking for the potential causes of over-indebtedness, we found that borrower sacrices seem related mainly
to the economic challenges of failed business investments, loan use that does not earn returns, adverse shocks
that reduce the borrowers income, and a lack of assets, especially savings, to serve as a buffer against delinquency. Even if shocks cannot be fully anticipated, these items require lenders to focus on sound due diligence.
A lack of nancial literacy, and particularly debt literacy, among borrowers also drives over-indebtedness risk.
Interestingly, this does not apply to being a rst-time borrower, having a volatile income, or to taking loans
from more than one institution at the same time.
With regard to lender inuences, it seems that, besides standard complaints about interest rates, borrowers
consider inexible product features a main reason for their sacrices. Installment schedules dont always t
the borrowers cash ows, and a strict application of the zero-tolerance policy can prevent rescheduling where
it would actually be appropriate. Also, disbursements may be too late for productive investments. Of course,
customers in repayment difculties frequently complain about the high level of interest on microloans. When
16

borrowers reach the stage of delinquency, collections practices may in some instances be too tough
and aggravate the over-indebtedness experiences of
microborrowers.
Among the target group of micronance that is living
on low and volatile incomes, difculties in regularly
assembling the cash for loan repayments are not surprising. For a long time, these customer experiences
havent earned much attention, seeming less relevant
in the face of a strong belief in the positive impact
of micronance on borrowers lives. Nevertheless,
cash demands are a challenge for the poor, and credit
is always a risk as well as an opportunity. In the context of the industrys current focus on responsible
nance and client protection, it needs to pay attention to the hardships that some microborrowers go
through and try to reduce them. At the same time,
given the value even struggling borrowers place on
further access to loans, simply stopping or reducing
microcredit does not seem an appropriate reaction
in the Ghanaian context. While some measures may
involve smaller loan sizes or recognizing that not
everybody needs a loan, other measures may rather
be about improving credit products to make repayment easier.

What are the prospects for microfinance in


urban Ghana?
There have recently been concerns about over-indebtedness risks in Ghana, one study raising concerns about multiple borrowing, another applying
high-level early-warning indicators to the market
and concluding that over-indebtedness risks in Ghana are not yet extreme but on a medium to high
level given market characteristics. Also, local MFIs
are aware of increasing competition and borrowers
struggling to repay their loans.
On the other hand, neither the penetration of the
Ghanaian market with microloans or with access to
nance in general, nor the competitive behavior of
our partner MFIs, represents a reason for concern. In
fact, the lending policies of the main Ghanaian MFIs
are rather conservative, and many MFI customers
are still credit-constrained in terms of both volume
and access to alternative loan sources. There are no
excessive levels of delinquency, nor did we nd an

issue with multiple borrowing, at least among the


markets top institutions.
Nevertheless, this paper has shown that many borrowers are indeed facing severe challenges in repaying their loans; 30 percent of them struggle so
much with their repayments that we call them overindebted from a customer protection perspective.
These ndings represent a call to Ghanaian MFIs to
focus more on customer satisfaction and borrower
experiences, meeting their clients nancing needs
more precisely. This will probably imply rethinking
features of their product portfolio. At the same time,
there seems to be a need for improved client literacy
and for more transparency about the rights of borrowers in the collection process.
The Ghanaian micronance industry is not in a stage
of crisis. Nor does it seem very close to a wave of
delinquency among its well-managed MFIs. Nevertheless, the repayment struggles that borrowers have
reported in this survey may be a sign of increasing delinquency problems. And the situation in other market segments that do not benet from the top MFIs
skills to identify the best borrowers may well be of
more concern. It is therefore a good time for Ghanaian MFIs to take action, from both a motivation of
customer protection and of risk management.

What are the policy implications for the


wider microfinance sector?
The fundamental learning from this paper for the
micronance sector is that we need to pay more attention to the experiences of microborrowers in repaying their loans. A continued demand for loans
and strong repayment statistics do not guarantee that
customers are well and that they are sufciently protected from suffering.26 In spite of the large demand
gap in micronance and of the potential benets of
microloans to the poor, the downsides of repayment
difculties represent an inherent risk to every debt.
Any responsible lending institution, but especially
those with a social mission, should pay attention to
this downside risk and take appropriate customer
26. See Schicks and Rosenberg (2011) for the challenges in
drawing conclusions about over-indebtedness from MFIs
repayment statistics.

Over-Indebtedness of Microborrowers in Ghana

17

protection measures to reduce the potential for overindebtednessnot only in terms of portfolio quality
but also in terms of borrower sacrices.
This paper highlights the experiences of microborrowers with their loans and aims to raise awareness
of their repayment struggles. At the same time, it
contributes to the micronance industrys understanding of what over-indebtedness means from
the perspective of clients and how we can assess it
empirically. While getting reliable information from
customers about their sacrices represents both a
challenge and a large effort, this paper suggests an
innovative way to identify over-indebtedness risks
from the viewpoint of micronance clients. We will
need many more research projects of this type to get
a more comprehensive understanding of borrower
experiences, also in other countries and cultural contexts. In the long run, we will need to identify proxies of over-indebtedness that are easier to track for
MFIs and regulators.
There are also policy implications from this research
for the development of customer protection mechanisms and especially for preventive measures against
over-indebtedness. The ndings are of course limited to the Ghanaian context, but are likely to be of
relevance to other markets as well: Even signicant
levels of over-indebtedness in a micronance market
do not automatically imply that there should be no
more microlending. In fact, microborrowers in Accra continue to express a strong interest in borrowing
in spite of their struggles. Looking at the causes of
over-indebtedness can inform us about better reactions to the over-indebtedness challenge. Both borrower sacrices and delinquency may result not only
from too much access to credit but alternatively from

18

access to inappropriate loan products. In addition to


sound due diligence, making products more exible
and tailoring loan disbursements, grace periods, and
installment schedules to the borrowers investment
cash ows have the potential to signicantly reduce
over-indebtedness risks. It can improve both the
experience of microborrowers with their loans and
their repayment performance. Further research will
be necessary to test this hypothesis and understand
the implications and feasibility of product exibility
in more detail. The same applies to the promotion of
savings products that may reduce the need for loans
or the risk of repayment difculties.
Finally, there is a message in our ndings about loans
used for consumption. Unsurprisingly, we nd the
commonly assumed relationship between using a
loan for consumption purposes and experiencing repayment difculties conrmed. In light of the recent
trend toward seeing the benets of micronance in
terms of consumption smoothing rather than productive investment (Collins et al., 2009), the relationship
between over-indebtedness and borrowing for consumption doesnt necessarily mean there shouldnt be
consumption loans, especially for emergencies. Also,
Ghana provides a good example of how preventing
consumption loan use is not a practical option. Money is fungible, and borrowers will always use parts of
their loans for nonbusiness purposes or will reduce
their households investment in their business as a
consequence of having the loan. Nevertheless, it is
not surprising that paying back without investment
returns is more difcult and we should keep these
challenges in mind. We will need to improve our understanding of how the risks and benets of borrowing balance and how we can best support borrowers
in managing their nances and their lives in poverty.

References

9(3) 327355.

Lascelles, D., and S. Mendelson. 2011. Micronance


Banana Skins 2011, The CSFI Survey of Micronance
Risk: Losing Its Fairy Dust. Centre for the Study of
Financial Innovation (CSFI) No. 99. Kent, UK.

Banerjee, A.V., and E. Duo. 2007. The Economic


Lives of the Poor.
21(1) 141167.

McIntosh, C., and B. Wydick. 2005. Competition and


Micronance.
78(2) 271298.

Augsburg, B., and C. Fouillet. 2010. Prot Empowerment: The Micronance Institutions Mission Drift.

Schicks, J. 2010. Micronance Over-Indebtedness:


Understanding its Drivers and Challenging the Common Myths. Centre for European Research in Micronance, Universit Libre de Bruxelles CEB Working
Paper No. 10/048.
Schicks, J., and R. Rosenberg. 2011. Too Much Microcredit? A Survey of the Evidence on Over-Indebtedness. CGAP Occasional Paper 19. Washington DC.
Schreiner, M. 2004. Scoring Arrears at a Microlender
in Bolivia.
6(2) 6588.

Kappel, V., A. Krauss, and L. Lontzek. 2010. Over-Indebtedness and MicronanceConstructing an Early
Warning Index. Center for Micronance, University
of Zurich, responsAbility, Council of Micronance
Equity Funds, Triodos Investment Management
Zurich.

Grammling, M.2009. Cross-borrowing and Overindebtedness in Ghana: Empirical Evidence from


Micronance Clientele and Small Enterprises. Technical draft for discussion. ProCredit Holding Frankfurt
am Main.

Gonzalez, A. 2008. Micronance, Incentives to Repay, and Overindebtedness: Evidence from a Household Survey in Bolivia. Doctoral thesis. Ohio State
University, Ohio.

Roesch, M., and O. Hlis. 2007. La micronance:


outil de gestion du risque ou de mise en danger
par surendettement? [Micronance: A tool for risk
management or for the creation of over-indebtedness
risks].
44 199-140.

Godquin, M. 2004. Micronance Repayment Performance in Bangladesh: How to Improve the Alloca32(11)
tion of Loans by MFIs.
19091926.

Reille, X. 2009. The Rise, Fall, and Recovery of the


Micronance Sector in Morocco. CGAP Brief. Washington, DC.

Collins, D., J. Morduch, S. Rutherford, and O. Ruthven. 2009.


Princeton, NJ: Princeton
University Press.

Venkata, N. A., and A. Veena Yamini. 2010. Why Do


Micronance Clients Take Multiple Loans?. MicroSave India Focus Note No. 33.
World Bank. 2011. Tackling Poverty in Northern
Ghana. Report No. 53991-GH. Washington, DC.

Over-Indebtedness of Microborrowers in Ghana

19

Appendix I: The Client Protection Principles

Appropriate product design and delivery

Providers will take adequate care to design products and delivery channels in such a way that they do not
cause clients harm. Products and delivery channels will be designed with client characteristics taken into
account.

Prevention of over-indebtedness

Providers will take adequate care in all phases of their credit process to determine that clients have the
capacity to repay without becoming over-indebted. In addition, providers will implement and monitor internal systems that support prevention of over-indebtedness and will foster efforts to improve market-level
credit risk management (such as credit information sharing).

Transparency

Providers will communicate clear, sufcient, and timely information in a manner and language clients can
understand so that clients can make informed decisions. The need for transparent information on pricing,
terms, and conditions of products is highlighted.

Responsible pricing

Pricing, terms, and conditions will be set in a way that is affordable to clients while allowing for nancial
institutions to be sustainable. Providers will strive to provide positive real returns on deposits.

Fair and respectful treatment of clients

Financial service providers and their agents will treat their clients fairly and respectfully. They will not
discriminate. Providers will ensure adequate safeguards to detect and correct corruption as well as aggressive or abusive treatment by their staff and agents, particularly during the loan sales and debt collection
processes.

Privacy of client data

The privacy of individual client data will be respected in accordance with the laws and regulations of individual jurisdictions. Such data will only be used for the purposes specied at the time the information is
collected or as permitted by law, unless otherwise agreed with the client.

Mechanisms for complaint resolution

Providers will have in place timely and responsive mechanisms for complaints and problem resolution
for their clients and will use these mechanisms both to resolve individual problems and to improve their
products and services.
(www.smartcampaign.org)

20

Appendix II: List of Borrower Sacrifices


Interviewers asked each respondent about the following list of sacrices:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

Reduce food quantity/quality


Reduce education
Work more than usual
Postpone important expenses
Deplete your nancial savings
Borrow anew to repay
Sell or pawn assets
Seizure of assets
Use family/friends support to repay
Suffer from shame or insults
Feel threatened/harassed by peers/family/loan ofcer
Suffer psychological stress yourself or in your marriage
Other

Respondents ranked the acceptability and frequency of each sacrice on a scale from 1 to 4.

Acceptability:
Easily acceptable, Only just acceptable, Not really acceptable, Not acceptable at all.
Frequency:
Once in past year, 1-3 times in past year, > 3 times but not often, Frequently in past year.

Over-Indebtedness of Microborrowers in Ghana

21

THE CENTER FOR FINANCIAL INCLUSION at ACCION (CFI) was launched


in 2008 to help bring about the conditions to achieve full financial inclusion
around the world. Constructing a financial inclusion sector that reaches everyone with quality services will require the combined efforts of many actors. CFI contributes to full inclusion by collaborating with sector participants
to tackle challenges beyond the scope of any one actor, using a toolkit that
moves from thought leadership to action.
www.centerforfinancialinclusion.org

The Over-Indebtedness of Microfinance Customers


An Analysis from the Customer Protection Perspective

DISSERTATION

Presented in Partial Fulfillment of the Requirements for


The PhD in Economics and Management

By
Jessica Schicks, MPhil, Dipl.Oec

At the
Centre for European Research in Microfinance (CERMi)
Solvay Brussels School of Economics and Management
Universit libre de Bruxelles

Members of the Jury:


Prof. James Copestake, Centre for Development Studies, University of Bath
Prof. Isabelle Gurin, IRD, Universit Paris I Sorbonne and French Institute of Pondicherry
Prof. Marek Hudon, Supervisor, CERMi, Universit libre de Bruxelles
Prof. Marc Labie, CERMi, Universit de Mons, Universit libre de Bruxelles
Prof. Ariane Szafarz, Secretary of the Jury, CERMi, Universit libre de Bruxelles

Academic Year 2012/2013


UNIVERSITE LIBRE DE BRUXELLES
- 285 -