The Lending Interest Rates in the Microfinance Sector: searching
for its determinants
Abstract
Using data of 1,299 microfinance institutions in 84 countries and following different approaches,
we find that the lending interest rate is determined by the funding cost, the loan size and the
efficiency level of microfinance institutions. Regarding competition, results are mixed. Is only
in Asia where a negative correlation between competition and lending interest rates can be
detected.
For other subsamples, we find that competition is more likely to be negatively
correlated with the size of loans.
Pablo Cotler
Universidad Iberoamericana, Mexico DF
pablo.cotler@uia.mx
JEL: G21, G28
Keywords: Africa, Latin America, Asia, lending interest rates, microfinance
0
I. Introduction
The correlation and causality among economic growth and financial development has
been thoroughly analyzed at the macroeconomic level (Levine, 2005). Further, studies such as
Beck, Demirguc-Kunt and Levine (2007) reveal the importance of greater financial leverage: it
affects in a disproportional manner the income of the poorest 20% of the population. Thus, the
development of financial institutions not only helps to reduce poverty, it may also help to scale
down inequality.
According to Beck, Demirguc-Kunt and Martinez Peria (2007), roughly 40 to 80 percent
of the populations in most developing countries lack access to formal sector banking services.
Thereby, the increased surge of microfinance institutions across less developed countries should
help cash starve entrepreneurs to unleash their productivity and raise their income above poverty
lines. However, measurable results of their impact are not easy to find. Papers by Pitt and
Khandker (1998), Morduch (1998), Banerjee and Duflo (2004), Alexander-Tedeschi and Karlan
(2006) and Hermes and Lensink (2010) show –among others- that there are several
methodological difficulties to correctly assess the impact of access to financial products. As a
result, the debate concerning the impact of microfinance on poverty reduction is far from settled.
Certainly this is puzzling given the steady growth of micro lending institutions across the
developing world.
Loans offered by microfinance institutions are not necessarily new to the population they
serve. Small entrepreneurs and poor families have typically being in contact with informal
lenders that offer short-term small-loans and make use of social and market sanctions to avoid
borrowers from defaulting (Aleem, 1990)1. If informal lenders exist, microfinance institutions
may have a positive impact on businesses profitability and on household wealth if they charge an
1
interest rate that is well below the cost of informal loans and/or if the loan size been offered is
sufficiently big to solve indivisibility problems.
But how high are interest rates in the
microfinance industry? According to the Microfinance Information Exchange database2, the
annual nominal lending interest rate charged by microfinance institutions during the period 20002008 was on average 42% in Africa and in Latin America and 35% in Asia. Since the annual
inflation rate was approximately 7% per year in all three continents, real interest rates were high.
Even though there is no worldwide cross section-time series data on interest rates charged by
individual moneylenders, rates charged by microfinance institutions are perceived –according to
conventional wisdom- to be well below those charged by neighborhood loan sharks.
Notwithstanding such perception, there is plenty of dissatisfaction among microfinance
practitioners regarding the high interest rates most microfinance institutions charge (see for
example, Rosenberg et al., 2009, Gonzalez, 2010 and the worldwide reactions following the
initial public offering of Banco Compartamos in 2007).
Since interest rates affect the probability that financing spurs growth, it would be useful to
know what factors cause these high interest rates. Do they closely follow the interest rates that
financial institutions pay for their funding? Or are the operating costs the main cause? Or
perhaps, these high interest rates are the result of a lack of competition in the credit markets. To
answer such questions, we used the information collected by the Microfinance Information
Exchange. This dataset includes information on financial income, the value of the loan portfolio,
average loan size, cost of funds, lending interest rates, operating costs, delinquency rates, number
of clients, profitability, etc., for 1,299 financial institutions located in 84 countries throughout
Africa, Asia and Latin America for the period 2000 to 2008.
As happens with any non-
mandatory database, self selection may bias our results. However, given the visibility of this
2
webpage (300,000 visits per month) and the funding opportunities it provides, it is very likely
that the most important microfinance institutions of all 84 countries are included in this database.
To fulfill our objective, we followed two approaches. Following Cull, Demirguc-Kunt
and Morduch 2006, we first examine the determinants of the lending interest rate by regressing
such variable against the cost of funds, the operational cost, the average loan size, the rate of
profitability, a proxy for competition, the type of financial institution (bank, cooperative, etc.)
that is offering such micro loans, and a set of dummies to control for time waves and geographic
location. This approach, however, may be misleading if managers of these institutions have a
profitability goal –regardless of whether it is distributed among owners- which enables them –for
example- to increase the loan portfolio and/or the number of borrowers, expand the number of
branches and/or the number of loan officers or simply to be financially self-sufficient. Whatever
their objective, a profitability target and the characteristics of their market niche are essential
when deciding an appropriate pricing policy and a loan size within the time-scale that they are
willing to offer. Taking this description into account, a second approach consisted on estimating
a system of equations that is simultaneously solved.
Regardless of the approach followed, three results stand out for their policy implication.
First a reduction in the funding cost leads to lower lending interest rates. However, 90% of all
microfinance institutions of our sample had an average cost of capital of 0% in real terms; thus,
there is not much room for further reductions. Second an increase in efficiency leads to lower
lending interest rates, being its impact increasing with the initial size of microfinance institutions.
Third, it appears that competition is more likely to have an impact on the loan size than on the
lending interest rate.
To show these results, the paper was divided into four additional sections. In the first one
3
we provide a literature review. The second one describes the database we consulted and the
methodology we used. Then in the third section, we present and discuss our findings. Finally in
the last section we conclude.
II. Literature Review
When analyzing lending interest rates charged to small entrepreneurs, the first issue that
comes into mind is whether they follow a pattern that is consistent with how much competition
financial institutions encounter in such markets. While the structure-conduct-performance theory
suggests that greater competition among lending institutions should bring interest rates down, the
informational problems that surround credit market transactions may weaken such argument.
Furthermore, some authors predict the existence of a negative correlation between market power
and interest rates. For example, Peterson and Rajan (1995) found that lending institutions that
wield greater market power are those with enough resources to invest in relationship lending.
Thus, as market power increases, the likelihood that smaller firms will be granted loans is greater
and therefore interest rates should decline. With a different argument, Marquez (2002) and
McIntosh and Wydick (2005) arrive to the same conclusion: as competition among financial
institutions increase, default risks may follow a similar path and so do interest rates. However,
other authors have an opposite view.
For example, Boot and Thakor (2000) claim that a
relationship orientation helps to partially protect the financial institution from competition.
Thereby, higher competition may induce financial firms to reallocate resources towards more
relationship lending and therefore smaller firms may face a reduction in the lending interest rates.
Thus, two conflicting hypotheses may be found regarding the effects that an increased
competition in credit markets for small firms has on the lending interest rate.
4
But the structure-conduct-performance framework may be misleading if market
contestability is a relevant feature of credit markets for small firms.
However if potential
borrowers lack formal documentation to certify their income and expenditure flow, it is very
likely that financial institutions may need to develop special techniques to assess the risk profile
of these potential borrowers. In this scenario, an entrance threat is not necessarily sustainable and
could therefore call into question the issue of market contestability and its effects on interest
rates. Taking into account all these factors, the interaction between competition and the lending
interest rate charged to small entrepreneurs may be ambiguous.
Whether the correlation between market structure and interest rate is positive, negative or
null, it is a question to be solved empirically. However, a review of the literature also shows
conflicting results. On the one hand the results found by Boot and Thakor (2000) and Ongena
and Smith (2001), support the traditional structure-conduct-performance hypothesis. On the
other hand, the findings of Petersen and Rajan (1995) and Zarutskie (2003) support the
alternative hypothesis.
Furthermore, the results depend on the methodology and on the
characteristics of the database. For example, Carbó, Rodríguez and Udell (2006) show that the
sign of the correlation is sensitive to how market power is assessed. If market power is defined
by the Lerner index, their results support the conventional theory: greater market power implies
higher interest rates. However, if market power is defined by concentration indexes, their results
are the opposite and the conventional theory is discarded.
But interest rates are not only determined by real or potential competition; the
characteristics of borrowers and lenders also matter. For the microfinance sector, Rosemberg,
Gonzalez and Narain (2009) and Gonzalez (2010), suggest that tiny loans with very low default
rates require high administrative expenses that may not be offset by economies of scale. Further,
5
even though microfinance institutions have -on average- a higher profitability rate than
commercial banks, these same authors claim that the search for returns is not an important driver
of interest rates. While such hypotheses may be appealing, these authors do not explain how they
arrived to such conclusions: there is no information regarding the econometric method being used
nor what other explanatory variables were considered nor any information regarding the
statistical significance of their results.
Cull, Demirguc-Kunt and Morduch (2006) examine the determinants of profitability,
portfolio at risk and loan size in the microfinance sector without taking into account how much
competition lenders face because the typical proxies for such variable have endogeneity problems
and do not measure how intensive is competition. Using the Microfinance Information Exchange
database for the period 1999-2002, they find –among other results- that the lending interest rates
and the capital costs affect the profitability of financial institutions. Their conclusions however
may need to be re-examined since they are obtained through the use of least squares estimations
in which it is implicitly assumed that the estimated coefficients are independent of the size of the
institutions and that there is no simultaneity on the decisions taken by managers of microfinance
institutions regarding interest rates, loan size and profitability.
III. Data and Methodology
Obtaining financial information of those institutions involved in microfinance is not easy
because in most countries there is no financial authority that supervises the recollection of such
data. Furthermore, the absence of organized market supervision means that these entities could
freely decide how to measure the variables describing their different sources of income and
expenditure.
Finally, even if there were an informal consensus on how to measure these
6
variables, that would not necessarily ensure that the information is reliable since it is very likely
that accounting deficiencies might exist.
To solve these problems we used the information collected by the Microfinance
Information Exchange (Mix). Members of this network report their financial results to managers
of this organization which in turn make sure that the definition and methodology used to define
each variable is unique across institutions. Thanks to this, we have annual information for the
period 2000 to 20083 concerning financial income, the value of the loan portfolio, average loan
size, cost of funds, lending interest rates, operating costs, delinquency rates, the loan loss
provision rate, number of clients, profitability, etc., for 1,299 financial institutions located in 84
countries throughout Africa, Asia and Latin America (see table 1).
Given the time span
considered and the number of years that these institutions have been reporting their data to the
Mix, we have an unbalanced cross section-time series panel data of 4,718 observations.
INSERT Table 1
Even though the database could have a self-selection bias, it is worth using it for several
reasons. First of all, it is a conceptually homogeneous database: each variable has the same
meaning for each institution. Second, very few micro-finance institutions are willing to share
their inter-temporal experiences, so having such a panel of data may help understand the
dynamics of lending interest rates. Finally, even if the panel of data is not representative of all
microfinance institutions, collectively however is very likely that they serve a very large fraction
of microfinance customers worldwide.
Thus, we consider that this work may be an important
7
step in the empirical literature on the management of microfinance institutions in developing
countries.
As is usually done, the lending interest rate is measured as the portfolio yield: all interest
and fee revenue from loans divided by the average gross loan portfolio. Thus, is a weighted
average of the interest rate actually received by the financial institution. Using such definition,
Table 2 shows that the median interest rate charged in Asia is lower –relative to what is being
charged in America and Africa. Notwithstanding such result, microfinance entities in Asia earn a
similar rate of return. Surely default rates, operating and funding costs as well as the size of
financial firms may help explain such outcome.
INSERT Tables 2 and 3
Following Martinez and Mody (2003) three sets of variables might explain the pattern
followed by the lending interest rate. First we have the capital cost and the operational cost per
peso lent of the financial entities. If the latter is turned upside down, we could use it as a proxy
of how efficient these entities are4. However the “quality of the product” needs to be considered.
To explain this better consider two examples. First, imagine two financial entities that report the
same average operational cost but have different default rates. In this scenario, the institution
with the lowest non-payment rate should be considered more efficient. Bearing this in mind,
efficiency may be better measured by adjusting the average operational cost by the portfolios’
default rate. Now let us consider two financial entities whose average operational cost and
default rates are similar, but report different average loan size. Taking into consideration that
smaller loans carry out higher operational costs, the entity with the smaller loan size should be
8
considered more efficient. For these reasons, a better proxy for efficiency will be achieved if the
operational cost per peso lent is adjusted by the portfolio default rate and by the relative size of
loans granted by the institution1.
The second set of variables includes those describing the size and nature of the financial
institutions, the earnings they make and the number of years they have been operating. As an
indicator of the financial firms’ size we used the value of their financial assets and loan portfolio,
both measured in real terms and expressed in logarithms. Regarding the nature of these financial
entities, six dummies were considered, each one representing the following type of institution:
non-bank financial institution, non-government organization, rural banks, credit union or
cooperative, banks, and any other type. Further, the average loan size, the loan loss provision
rate and a proxy for their market niche were also included. For the latter, denoted as outreach, we
used the average loan size divided by gross domestic product per capita.
As explained before, it is difficult to measure how much competition financial entities
face. Further, given the characteristics of our dataset we would need to have a cross section-time
series panel data set for such variable. Since the vast majority of the institutions that comprise
our sample do not operate at the national level and usually compete with informal moneylenders,
the information required to construct such a variable would be tremendous and to our knowledge
such dataset does not exist.
Notwithstanding these problems, we followed the traditional
approach (see for example Kai, 2009) and considered the fraction of the adult population who are
borrowers of microfinance institutions as a proxy for competition. Finally, to describe the
1
Denoting efficiency as effic, operational cost per peso lent as (cop), default rate as (def), and the relative size of
loans of institution i in country j as (Lij)/(Lj), it follows that the efficiency of institution i in country j equals:
(1/cop)(1-def)( (Lj)/(Lij).
9
economic environment we included a legal rights index and a credit information index, variables
that are available in the World Bank dataset Doing Business.
INSERT Table 4
With the use of this data, first we followed the standard approach and examined the
determinants of the lending interest with the use of one equation. In this scenario our hypothesis
may be described by the following functional relationship:
Iloan = F(Ifund, effic, roa, avgloan, competition), whereas: I1>0, I2<0, I3>0, I4<0, I5<0 ….(1)
Where Iloan describes the lending rate of interest, Ifund measures how much they pay for
their funds, effic is the efficiency of the institution to deliver its loans, roa is the return on assets
and avgloan is the average loan size. Next to this functional form we describe the sign of the
partial derivative we expect to find. Thus for example, we expect to find a positive correlation
between the lending interest rate (Iloan) and the cost of funds (Ifund), being such hypothesis
denoted by: I1>0. Further, we posit a negative correlation between the lending rate and our proxy
for efficiency because as financial firms become more efficient they may be able to reduce their
lending rate and still reach the same profit rate. However, if these financial firms wish to
increase profits they will surely increase their lending rate, thus, I3>0. Regarding the average loan
size, two possible explanations may be provided to explain why we posit a negative correlation.
First, if financial firms wish to finance their variable costs, then a bigger loan size could be
correlated with smaller interest rates. On the other hand, if bigger loans were received by more
experienced borrowers, then credit risk would decline and so could interest rates: I4<0.
10
Regarding competition, we follow the standard theory and assume that it has a negative
correlation with the dependent variable. Further some control variables were included. Among
them, we have dummies capturing the different organizational structure of the financial entities
and other dummies that consider differences across countries and years.
However, many of the variables that could explain the behavior followed by the lending
interest rate are endogenous. In particular, the profitability rate and the size of loans are two
variables which the financial institution seeks to influence. Thus, managers of these organizations
may have a profitability goal that allows them –for example- to enlarge the portfolio and/or
increase the number of borrowers and/or create more branches and/or recruit more loan officers,
etc. The profit target and the characteristics of the market niche, results in an optimal pricing
policy comprising a loan size within a time-scale.
Accordingly we believe that the loan
transaction may be described in three steps. First, the financial firm decides how much to charge
and what the optimal loan size must be in order to reach its profitability goal. Once known the
value of the lending interest rate and the average loan size the financial institution offers, a
potential customer decides whether s/he wants to request a loan. Taking into account the credit
history of the potential borrower and its income-expenditure stream, the financial institution
builds a risk profile of the individual. With this at hand, they decide where to lend or not. While
for first time customers the typical microfinance institution does not allow any kind of
negotiation related to the loan size, for repeated customers some sort of negotiation is possible
but is the financial institution that –taking into account its goals- decides the loan size. Taking
this description into account it could be more reasonable to estimate simultaneously the following
three equations:
11
2a. Iloan = F(Ifund , effic, roa, avgloan, competition…); I1>0, I2<0, I3>0, I4<0, I5<0.
2b. Roa = G(Iloan , Ifund, effic, reserves, outreach, …); roa1>0, roa2<0, roa3>0, roa4 <0 and
roa5 with an ambiguous sign.
2c. Avgloan = H(effic, age, competition, Iloan …) ; Avgloan1<0, Avgloan2>0, Avgloan3 <0 and
Avgloan4 <0.
In this system, equation (2a) is similar to equation (1). With regard to the profitability
rate, equation (2b) suggests that it will be explained by the lending interest rate, the cost of funds,
the efficiency of microfinance institutions, their loan loss ratio and its market niche. As shown
by equation (2b), profitability will raise when the lending interest rates and/or efficiency
increases or when the cost of funds declines. Since holding loan loss reserves imply an expected
loss and an opportunity foregone, we expect a negative correlation between this variable and the
profitability rate. Finally, with regard to outreach, the correlation could go either way. Assuming
poorer people receive smaller loans, we could expect a positive correlation between outreach and
the profitability ratio (roa) because the incentives to default among this population are smaller
since they want to avoid informal moneylenders. However, if smaller loans imply higher average
operational costs, then outreach could be negatively correlated with the profitability rate. Finally
equation (2c) describes the behavior followed by the average loan size.
As microfinance
institutions become more efficient, they could offer smaller loans: Avgloan 1<0. Regarding age
(that works as a proxy for experience of the financial institution), it may affect both the supply
and the demand for bigger loans.
On the one hand, when microfinance institutions start
operations they usually offer loans of small amounts because they do not have much capital or
experience and debtors tend to be people without credit history. However, if the supply of loans
12
has dynamic incentives (i.e., the services offered by the institution increases as the debtor builds
his credit history), is very likely that the loan size will increase through time. On the other hand,
if loans have a positive impact on wealth, is possible to assume that the demand for bigger loans
will rise. Under these assumptions, we expect to find a positive correlation between the average
loan size and age (Avgloan2>0). Regarding competition we follow the credit card interest rate
literature and assume that competition leads microfinance institutions to search for new markets.
However the sign of the correlation could be ambiguous since it depends on whether they start
lending to relatively non-poor people (Avgloan3>0) or moving to poorer neighborhoods and offer
smaller loans (Avgloan3<0). Given their technology and stated mission, we will assume that the
latter effect is more likely to dominate. Finally, if a higher lending rate is being charged we
could expect a smaller loan size being demanded (Avgloan4<0).
IV. Results
1. Individual Estimates
Our first approach consists on estimating equation (1). As Table 5 shows, the lending
interest rate (Iloan) follows a behavior in all three continents that is consistent with our
hypothesis: the dependent variable responds positively to changes in the cost of funds (Ifund) and
to the return of assets (roa), and negatively to the proxy for efficiency (effic)5 and to the average
loan size (AvgLoan). With regard to competition, we find that it only has an impact in Asian
markets.
Insert Table 5
13
Given the heterogeneity in the size of the financial institutions that comprise our sample,
next we analyzed whether the parameters reported in Table 5 are sensitive to the initial size of
these financial institutions6. Since a learning curve regarding the appropriate use of techniques to
mitigate information asymmetries may exist, age could also be a factor that may lead to
heterogeneous impacts. For this purpose we added -to the set of explanatory variables used in
Table 5- new variables that are created by multiplying such set by the initial value of each
institution’s asset. As results in Table 6 show, the impact of each independent variable over the
lending rate will vary according to the initial size of microfinance institutions and its age.
Notwithstanding such heterogeneity, on average, the sign of the estimated parameters of all
independent variables are consistent with our hypothesis stated in equation (2a): the cost of funds,
efficiency, the quest for profits and the average loan size help explain the behavior followed by
the lending interest rate. Furthermore, once interactions are considered, we find –for the overall
sample- that the impact of efficiency on the lending interest rates is higher the bigger the initial
size of microfinance institutions.
Insert Table 6
2. Simultaneous Estimates
However, results reported in Tables 5 and 6 may be misleading if reality is better
described by a system of equations in which interest rates, loan size and profitability are jointly
determined. From informal interviews we learned that managers of microfinance institutions
14
have a profitability goal which enables them to increase the loan portfolio and/or the number of
borrowers and/or expand the number of branches and/or the number of loan officers, etc.
Whatever their objective, profitability and characteristics of their market niche are essential for
them to adopt an appropriate pricing policy and a loan size within the time-scale.
Thus we believe that the loan transaction may be described in several steps. First, the
financial firm decides how much to charge and what should be the optimal loan size in order to
reach its profitability goal. Once known the value of the lending interest rate and the average
loan size the financial institution offers, a potential customer decides whether s/he wants to
request a loan. Taking into account the credit history of the potential borrower and its incomeexpenditure stream, the financial institution builds a risk profile of the individual. With this at
hand, they decide where to lend or not. While for first time customers the typical microfinance
institution does not allow any kind of negotiation related to the loan size, for repeated customers
some sort of negotiation is possible but is the financial institution who –taking into account its
goals- decide the loan size. So, the financial institution sets the price and negotiates the size of
the loan within a range so that it may achieve the profitability goal drawn at the beginning.
To consider such approach, we estimated a system of three equations that are
simultaneously solved. For such purpose we used a simultaneous equation estimation (three
stage least squares regression) that considers the existence of three endogenous variables (the
lending interest rate, the average size of loans and the profitability on assets) that are jointly
estimated. As explained before, our joint hypothesis may be described by the following set:
2a. Iloan = F(Ifund , effic, roa, avgloan, competition….) where: I1>0, I2<0, I3>0, I4<0, I5<0.
15
2b. roa = G(Iloan , Ifund, effic, provisions, outreach,…) where roa1>0, roa2<0, roa3>0; roa4<0
and roa5 with an ambiguous sign.
2c. Avgloan = H(effic, age, competition, Iloan …) where Avgloan1>0, Avgloan2>0, Avgloan3 <0
and Avgloan4 <0.
Insert Table 7
As Table 7 show, once interest rates, loan size and profitability are assumed to be jointly
determined only two of the parameters of equation (2a) have the expected sign across all
subsamples and are consistent with our prior and with results reported in table 5.
Thus, a
positive correlation with the cost of funds and a negative correlation with efficiency are found
regardless of the subsample considered. Further, similar to the results reported in Table 5, we
only detect a correlation between competition and the lending interest rate for Asian markets,
being such correlation negative as conventional theory would suggest. In all other markets, we
could not find any correlation from a statistically point of view.
With regard to the other two dependent variables -return on assets and average loan sizeour results are consistent with the priors stated on equations (2b) and (2c). Three results are
worth mentioning. First, taking as granted that microfinance institutions were truly trying to
reach poorer people, we were expecting to find a negative correlation between efficiency and
average loan size. However, with the exception of Asian markets where no correlation is found,
we find the opposite: an increase in efficiency leads to higher average loan size. Since we also
find that greater efficiency leads to lower lending rates, it is possible that a higher efficiency may
be the result of having a more appropriate lending technology that enables microfinance
16
institutions to pick better customers. In so doing, they are able to offer lower lending rates,
higher loan sizes and as a result of their improved technology, a greater profitability. Thus, such
correlation does not imply necessarily a mission drift.
Second, as explained before the correlation between profitability and outreach (average
loan size divided by gross domestic product per capita) could have had a positive or negative
sign. Assuming smaller loans go to poorer people, a negative correlation may be expected if
default rates among this population are smaller. However, if smaller loans imply higher average
operational costs, then outreach could be positively correlated with the profitability rate. Our
results suggest that the latter argument only holds for the African subsample.
Third, competition could have an impact on the lending interest rate and on the loan size.
However, our results suggest that it is only in Asian markets where we can detect a negative
impact in both variables. For other markets, results are mixed: in Latin America, competition
only has an effect on the loan size and in Africa it has no effect -from a statistically point of view.
Following our results of tables 5 to 7, a reduction in the funding cost could help reduce
the lending interest rate that microfinance institutions charged.
However, 90% of all
microfinance institutions of our sample had an average cost of capital of 0% in real terms; thus,
there is not much room for further reductions. Regarding competition, economic theory suggests
that a stronger competition in credit markets could help reduce the lending interest rate.
However, in markets where reputation and loyalty matters, it may well happen that an increase in
competition, leads in the short run to changes in the loan size but not in interest rates. Our results
suggest that: an increase in competition leads to a reduction in the lending interest rate in Asia
and a reduction in the average loan size in Latin America.
Finally, with respect to efficiency, all our results suggest that an increase in such variable
17
would lead to lower lending interest rates and a higher profitability. However, how exactly does
efficiency increase?
Technology use, management quality and hard work surely matters.
However, not only are these dimensions difficult to measure but also they are not exogenous
since they may well depend –among other things- on how much capital are financial firms willing
to invest. Taking into account the distinctive features of microfinance lending technologies,
efficiency however is not only the result of a better technology or management quality;
geographical characteristics of markets in which microfinance institutions work may also have an
effect on the operational costs of these institutions and thereby on how efficient they are.
To include some characteristics of the environment that may shape the value of the
financial firms’ efficiency, a biodiversity index7 -that captures the variability of a country’s
territory in terms of height and climate- could be considered. A higher value of this index may
signal a greater geographical heterogeneity and may imply that –on average- is more costly to
reach customers. Further, if economies of scale in lending were to exist, population density may
also have an effect on the operational costs of microfinance institutions. In this regard, data at
the national level may help explain why –as figure 1 suggests- Asian microfinance institutions are
most efficient: the biodiversity index takes an average value of 20.57 in America, 11.87 in Asia
and 4.75 in Africa; and that population density in 2007 was 85 people per square kilometer in
Asia, 44 in Africa and 28 in America. However, since countries are not homogenous, to find the
potential of economic policy to increase efficiency (and thereby reduce lending interest rates) we
would need to have data on biodiversity and population for all local markets where this
microfinance institutions work. This is a quest for future work.
18
V. Conclusions
One of the reasons that gave rise to this paper was the claim that the high interest rate that
microfinance institutions were charging was reducing the impact of their financing. Taking this
as granted, our objective was to find how these rates could be reduced. Through various
empirical methods, three policies -usually mentioned by policymakers- were indirectly tested.
The first one is reducing the funding rate. Consistent with such recommendation, all of our
estimations suggest a non-negative correlation between such rate and the lending interest rate.
However, since almost 90% of all microfinance institutions of our sample had an average funding
cost of 0% in real terms, there is not much room for market-driven reductions. Further, economic
theory and economic history show that government intervention to reduce this price is likely to be
a short-lived policy in lieu of the distortions it creates.
Fostering competition is always claimed to be the best public policy to reduce the lending
interest rate. However, in markets where reputation and loyalty matters, it may well happen that
an increase in competition leads in the short run to changes in the loan size but not in the lending
interest rate. Further, testing such policy is not easy since it is difficult to find a proxy for 1,299
microfinance institutions in 84 countries that may properly measure the intensity of competition.
Furthermore, since most financial entities do not work at the national level such proxy would
need to be able to measure the intensity of competition in local markets through time. We did not
have the necessary information to build such a good proxy. Instead we used a variable commonly
used in the literature and found that while in Asia an increase in competition leads to a reduction
in the lending interest rates, in America it leads to a reduction in the average loan size. While not
being the panacea to reduce lending rates, results shown here suggest that fostering competition
should continue to be a public policy objective.
19
The third mechanism to reduce the lending rate consists on policies that may help increase
the efficiency of these financial institutions. All our results suggest that a higher efficiency leads
to lower lending interest rates. To increase efficiency, technology use and management quality
need to be considered. Further, since the impact of efficiency on lending rates is increasing with
the initial size of microfinance institutions, government also need to consider the possibility of
fostering mergers and acquisitions in this sector. However, how much could public policy help
increase efficiency? Given the lending technology that most microfinance institutions use, an
answer to such question will depend on how difficult is for loan officers to reach their target
clients. Thus, geographic characteristics of the territories and the spacial distribution of their
clients should matter.
In this regard, data at the national level may help explain why
microfinance institutions are more efficient in Asia and less in Africa. But of course, countries
are not homogenous and thereby the quest is to find data that could describe these features at the
micro level.
If we had such data, we could estimate the natural rate of efficiency of each
microfinance institution and thereby learn the potential of public policy to increase efficiency and
help to reduce the lending interest rate.
20
Bibliography
Aleem, I. (1990), “Imperfect Information, Screening and the Costs of Informal Lending: A Study
of a Rural Credit Market in Pakistan. The World Bank Economic Review, Vol. 4(3), pp. 329-349
Alexander-Tedeschi, Gwendolyn and Dean Karlan. 2006. “Microfinance Impact: Bias from
Dropouts.” Unpublished manuscript, Department of Economics, Yale University.
Banerjee, Abhijit and Esther Duflo. 2004. “Do Firms want to Borrow More? Testing Credit
Constraints Using a Directed Lending Program.” Unpublished manuscript, Department of
Economics, MIT.
Beck, T., A. Demirguc-Kunt and R. Levine (2007), “Finance Inequality and the Poor”. Journal of
Economic Growth, vol. 12(1), pp. 27-49.
Boot, A. and A.Thakor (2000), “Can relationship banking survive competition?” Journal of
Finance, Vol. 55: 679-713.
Calem, P. and L. Mester, “Consumer Behavior and the Stickiness of Credit-Card Interest Rates”.
American Economic Review, Vol 85(5): 1327-1336.
Carbó, S. F. Rodriguez and G. Udell (2006), “Bank Market Power and SME Financing
Constraints”. Working Paper 237/2006 Funcas.
Cull, R., A. Demirguc-Kunt and J. Morduch (2006), “Financial performance and Outreach: A
Global Analysis of Leading Microbanks”. World Bank Policy Research Working Paper 3827.
Gonzalez, A. (2007), “Efficiency Drivers of Microfinance Institutions (MFIs): The case of
Operating Costs”. MicroBanking Bulletin, Issue 15.
Gonzalez, A. (2010), “Analyzing Microcredit Interest Rates”.
www.themix.org
Mix Data Brief No.4,
Kai, H. (2009), “Competition and Wide Outreach of Microfinance Institutions”.
Personal RePEc Archive 17143.
Munich
Levine, R. (2005), “Finance and Growth: Theory and Evidence” Handbook of Economic Growth.
P. Aghion y S. Durlauf (eds.) Northe-holland Elsevier Publishers.
Lewis, J. (2008), “Microloan Sharks”,
http://www.ssireview.org/articles/entry/939/
Stanford
Social
Innovation
Review.
Marquez, R. (2002), “Competition Adverse Selection, and Information Dispersion in the Banking
Industry”. The Review of Financial Studies, Vol. 15: 901-926.
21
Martinez, S. and A. Mody (2003), “How Foreign Participation and Market Concentration Impact
Bank Spreads: Evidence from Latin America”. Journal of Money, Credit and Banking, Vol. 36
(3): 511-537.
McIntosh, C. and Wydick, B. (2005) “Competition and microfinance” Journal of Development
Economics 78, 271-298.
Morduch, Jonathan. 1998. “Does Microfinance Really help the Poor? New evidence from
Flagship Programs in Bangladesh”. Unpublished manuscript, Department of Economics Harvard
University.
Ongena, S. and D. Smith (2001), “The duration of bank relationships”. Journal of Financial
Economics, Vol. 61: 449-475.
Petersen, M. and R. Rajan (1995), “The Effect of Credit Market Competition on Lending
Relationships”. Quarterly Journal of Economics, Vol. 110: 407-433.
Pitt, M. and S. Khandker (1998), “The Impact of Group-Based Credit Programs on Poor
Households in Bangladesh: Does the Gender of Participants Matter? Journal of Political
Economy. Vol. 106(5): 958-996.
Rosemberg, R., A. Gonzalez and S. Narain (2009), “The new Moneylenders: are the poor being
exploited by high Microcredit Interest Rates”. CGAP Occasional Paper No. 15.
Zarutskie, R. (2003), “Does bank competition affect how much firms can borrow? New evidence
from the U.S,” en Proceedings of the 39th annual conference on Bank Structure and
Competition, Federal Reserve Bank of Chicago, pp. 121–136.
22
Table 1
The database
Continent
# MFIs
#Countries
# obs.
Average number of years
operating in 2008
Africa
294
32
1,096
11.44
America
358
20
1,431
15.06
Asia
647
32
2,191
10.42
Total
1,299
84
4,718
12.43
Table 2
Distribution of some key Data during the period 2000-2008
Africa
America
Asia
Real
ROA
Loan as
Real
ROA
Loan as
Real
ROA
Loan as
interest
(%)
% of
interest
(%)
% of
interest
(%)
% of
GDPpc
rate (%)
GDPpc
rate (%)
rate (%)
GDPpc
1%
- 4.1
- 57.6
2.8
-11.2
- 40.6
1.9
- 5.2
- 48.1
3.0
10%
10.9
- 15.9
12.3
13.1
- 4.7
5.0
9.1
- 6.5
8.2
25%
18.4
- 4.8
25.5
21.3
0.6
12.2
15.5
0.0
13.8
50%
28.2
0.8
56.6
30.9
3.0
29.3
23.3
2.5
25.4
75%
45.1
4.3
117.4
44.4
6.2
58.2
34.6
6.0
69.7
90%
64.9
8.9
252.4
63.2
10.2
103.5
50.6
10.9
164.6
99%
145.8
21.5
692.0
103.0
19.1
312.2
88.0
25.9
1,125.0
23
Table 3
Distribution of some key data during the period 2000-2008
Africa
America
Asia
Total
Funding
Operating
Total
Funding
Operating
Total
Funding
Operating
Assets
Interest
Cost per
Assets
Interest
Cost per
Assets
Interest
Cost per
(1982-84
Rate
dollar lent
(1982-84
Rate
dollar lent
(1982-84
Rate
dollar lent
dollars)
(%)
dollars)
(%)
dollars)
(%)
1%
174
0.0
0.046
862
0.0
0.049
227
0.0
0.024
10%
1,427
0.1
0.118
4,588
0.9
0.107
1,318
0.1
0.066
25%
4,341
0.8
0.172
10,181
3.1
0.141
4,379
2.1
0.115
50%
12,364
2.0
0.290
34,640
5.4
0.218
13,847
4.4
0.175
75%
54,464
3.6
0.484
135,092
7.6
0.363
45,355
7.3
0.288
90%
176, 319
6.2
0.771
464,817
9.9
0.573
162,716
10.4
0.473
99%
962,372
14.6
2.065
1,927,736
16.6
1.242
1,916,596
21.8
1.158
24
Table 4
Descriptive statistics of main variables
Nominal Lending
interest rate
Funding cost
Efficiency (in logs)
Return on assets
Average loan size
(in logs)
Competition
Initial assets (in logs)
Age of the financial
institution
Provisions as a % of
assets
Biodiversity
Population Density
Mean
Maximum
Minimum
St. Deviation
Africa
America
0.42
0.42
2.58
1.50
0.01
0.02
0.27
0.21
Asia
0.35
2.01
0.03
0.18
Africa
America
0.03
0.06
0.22
0.71
0
0
0.03
0.04
Asia
0.05
0.56
0
0.05
Africa
America
1.23
1.47
3.95
5.08
-1.73
-0.72
0.77
0.70
Asia
1.73
7.01
-1.20
0.79
Africa
America
-0.02
0.02
0.83
0.53
-0.97
-0.89
0.14
0.11
Asia
0.02
0.73
-1.80
0.12
Africa
America
0.19
1.15
3.39
3.88
-2.91
-2.12
1.06
0.93
Asia
0.40
4.62
-5.28
1.39
Africa
America
0.015
0.057
0.046
0.149
0.0002
0.0004
0.012
0.0.46
Asia
0.036
0.238
0.00002
0.036
Africa
America
8.69
9.68
13.51
14.66
0.90
4.01
1.86
1.78
Asia
8.52
15.47
0.62
1.91
Africa
America
9.43
14.07
58.00
51.00
0
0
6.90
9.47
Asia
11.01
52.00
0
8.98
Africa
America
Asia
2.1
2.4
1.4
83.9
33.3
55.4
0
0
0
3.9
3.1
2.9
Africa
America
5.94
27.00
29.22
100.00
0.13
0.89
7.00
24.01
Asia
19.20
80.96
0.17
22.37
Africa
America
82.95
65.96
394.03
358.36
2.59
8.95
67.28
79.64
Asia
287.62
1229.16
1.70
352.89
25
Table 5
Dependent variable: Lending Interest Rates
Africa
America
Asia
3 continents
Ifund
1.26***
1.11***
0.87***
1.04***
Effic
- 0.24***
- 0.17***
- 0.16***
- 0.18***
Roa
0.76***
0.71***
0.73***
0.72***
- 0.03***
- 0.04***
- 0.03***
0.007
-0.14*
-0.09
Avgloan
Competition
within
between
overall
Rsquare
N
- 0.02**
0.04
0.34
0.66
0.65
1066
0.45
0.76
0.74
1420
0.44
0.54
0.54
2082
0.39
0.64
0.63
4568
Notes:
1.-* p<.1; ** p<.05; *** p<.01
2.- Estimation Method: Generalized Least squares. Taking into account the results provided by the Hausman test, all estimations
were done with fixed effects.
3.- Dummies for each year were not included for African and Asian countries since a joint test rejected the need for them; for
America and the world wide sample, the omitted year was 2000. We included a constant and five dummies to account for the
type of institutions: NGO, rural bank, credit union or cooperative, bank, non- bank financial institution. The category “other” was
the omitted category. Finally, we also included variables to describe the statue of the country’s legal rights and credit
information. The parameters of these dummies are not reported here.
26
Table 6
Dependent variable: Lending Interest Rates
With heterogeneous impacts
Africa
America
Asia
3 continents
Ifund
1.78
1.07***
0.68**
0.86***
Ifund * initial assets
- 0.04
0.01
0.01
0.02
Effic
0.05
- 0.38***
- 0.02
- 0.08***
Effic * age
- 0.01*
0.01***
- 0.01***
- 0.0003
- 0.04***
0.02***
- 0.02***
- 0.01***
0.003**
- 0.001***
0.002***
- 0.006***
0.002***
- 0.006***
- 0.002***
Effic* initial assets
Effic*initial assets*age
Initial assets*age
0.0007***
Roa
0.43
0.19
0.51***
0.49***
Roa*initial assets
0.04
0.070**
0.02
0.03**
Age
0.05**
- 0.02***
0.05***
0.01***
- 0.01
- 0.02***
- 0.05***
- 0.03***
- 0.02
- 0.20**
- 0.11
Avgloan
Competition
R-sq
0.59
within
between
overall
0.37
0.35
0.34
0.47
0.63
0.61
0.46
0.31
0.36
0.40
0.51
0.52
N
1053
1410
2053
4516
Notes:
1.-* p<.1; ** p<.05; *** p<.01
2.- Estimation Method: Generalized Least squares. In all regressions a constant was included. Taking into account the results
provided by the Hausman test, all estimations were done with fixed effects.
3.- Dummies for each year were not included for African and Asian countries since a joint test rejected the need for them; for
America and the world wide sample, the omitted year was 2000. We included a constant and five dummies to account for the
type of institutions: NGO, rural bank, credit union or cooperative, bank, non- bank financial institution. The category “other”
was the omitted category. Finally, we also included variables to describe the statue of the country’s legal rights and credit
information. The parameters of these dummies are not reported here.
27
Table 7
Simultaneous Estimations
Dependent Variable
Africa
America
Asia
3 continents
Ifund
0.71***
1.30***
0.42***
0.86***
Effic
- 0.21***
- 0.10**
- 0.07***
- 0.12***
Roa
- 0.11
- 0.89***
Iloan
Avgloan
Competition
0.09***
- 1.27
0.10***
- 0.20**
- 0.02*
- 0.08***
- 0.03***
0.39
- 0.43***
- 0.56***
Roa
Iloan
0.59***
1.03***
0.45***
0.77***
Ifund
- 0.79***
- 0.95***
- 0.50***
- 0.76***
Effic
0.22***
0.25***
0.11***
0.19***
- 0.000***
- 0.000***
- 0.000
- 0.89***
- 1.13***
- 1.03***
- 0.95***
Effic
0.92***
0.96***
0.0004
0.60***
Age
0.01***
0.004***
0.008**
0.01***
Outreach
Provisions
0.0001***
Avgloan
Competition
Iloan
N
- 5.65
2,73***
1,053
- 2.91**
- 2,26**
- 1.60**
0.11
- 5.32***
- 0.52
1,419
2,049
4,511
Notes:
1.-* p<.1; ** p<.05; *** p<.01
2.- Estimation Method: Three stage least squares.
3.- We included a constant, a dummy for each year and for each country. The omitted categories were the year 2000, Benin (Africa),
Argentina (America), and Afghanistan (Asia). Similar to the last table, we included in the equation that had the lending interest rate as
dependent variable, five dummies to account for the type of institutions and two variables to describe the statue of the country’s legal
rights and credit information.
4.- For the 3 continents estimation we included a constant, a dummy for each year, a dummy for each country, being Afghanistan the
omitted category and a dummy for each continent with Africa as the omitted one.
28
Figure 1
Average Efficiency by Size Distribution of Microfinance Institutions
4
3.5
3
2.5
2
1.5
1
0.5
0
-0.5
1%
5%
10%
25%
50%
75%
90%
95%
99%
-1
Africa
America
Asia
29
Notes
1
However, these lenders usually lacked specialization in credit transactions, operated in small local markets and
were unable to reach economies of scale. Thus, they were unlikely to become a cheap source of funding. Leaders in
the microfinance industry learned how informal lenders operated and thanks to the support of international
organizations and governments, were able to build institutions with governance schemes that help to minimize
principal-agent problems and used technologies that allow them to achieve economies of scope. As default rates
declined and group lending and sequential loans were developed, the number of microfinance institutions grew.
Adding to these improvements, the introduction of credit bureaus and the build-up of an appropriate regulatory
framework has certainly help to develop –in many countries- a very dynamic microfinance industry.
2
The Microfinance Information Exchange is considered the most important source for objective and unbiased
microfinance data and analysis. They provide unparalleled access to operational, financial and social performance
information on more than 1,900 MFIs covering 92 million borrowers globally. Once collected, all the data is review
for its coherence and consistency, and reclassified to comply with International Financial Reporting Standards
(IFRS).
3
Even though we had data for 2009, we decided not to include it because of the singularity of the economic and
financial crisis that started in the second half of 2008.
4
There is a caveat to such proxy. If an institution has put in place a growth strategy it may well happen that in the
short run their costs may be growing faster than their lending operations. This of course would not mean that they
are less efficient than others.
5
We tried with both proxies for efficiency and found that results improved when we used as proxy the inverse of
operational cost without adjustments. The correlation between both proxies within each continent was about 0.6.
6
The log of initial assets had the following distribution:
Percentiles Africa America Asia
4.489
5.136
4.437
1%
6.169
7.627
5.961
10%
7.540
8.516
7.397
25%
8.696
9.679
8.589
50%
9.868 10.781 9.690
75%
11.178 12.031 10.853
90%
12.460 13.691 13.761
99%
7
This index is a weighted average of terrestrial (80%) and marine (20%) characteristics of each country. The
terrestrial features capture the distribution of species and threats to them, and the variety of climates and ecological
factors in their territory. The source of this data is: http://data.worldbank.org/indicator/ER.BDV.TOTL.XQ
30