Content uploaded by Erwan Le Saout
Author content
All content in this area was uploaded by Erwan Le Saout on Feb 27, 2017
Content may be subject to copyright.
1
Modeling the profitability of microfinance institutions:
An application of the dynamic panel data methodology
*,**
*
Lâma DAHER, PhD student at Ecole de Management de la Sorbonne (PRISM)
**
Erwan LE SAOUT, professor of financial markets and risk management at Ecole de Management
de la Sorbonne (PRISM & LABEX REFI)
Abstract
The sector of microfinance is passing through a fundamental transformation that impinges on its
actors’ operations and performances. With the increasing scarcity of cheap funding and the
increasingly demanding investors, microfinance institutions are turning towards improving their
financial performance in order to grow and attract new non-free funding sources from external
financial markets. The aim of this paper is to investigate the drivers of financial performance from
within microfinance institutions and their macro-economic context. Cost inefficiency and
unanticipated inflation seem to be the most important risks that microfinance institutions have to
manage. Additionally, we suggest that profits are persistent from a period to another and that
microfinance institutions have quite low levels of competitiveness. Furthermore, we find that the
profitability of microfinance institutions is sensitive to global financial crunches, in reference to the
monetary crisis of 2007-2008. Empirically, a dynamic panel data analysis, using standard- and
system-GMM, is performed on a panel data set of 362 microfinance institutions from 73 countries
in 5 regions from 2005 to 2011.
Key words
microfinance institutions, profitability, profit persistence, determinants, internal factors, macro-
economic factors, macro-institutional factors, global financial crisis, dynamic panel data regression
JEL classification codes
G21, G23, O16
E-mail addresses
Lâma DAHER: lama.m.daher@gmail.com
Erwan LE SAOUT: erwan@lesaout.com
Postal address
Lâma DAHER
PRISM Sorbonne - Université Paris 1 Panthéon Sorbonne
17 rue de la Sorbonne
75231 Paris Cedex
France
2
Introduction
Poverty alleviation is the officially presumed engagement pursued by the microfinance industry
since its conception. Existent microfinance institutions (MFIs), as well as, new entrants are more
and more attracted by the potential profits that can be achieved out of microfinance activities. A
growing number of MFIs is shifting its main focus from, providing the poorer and most vulnerable
with microfinance services and products to help them improve their living, to maximizing their own
profitability through servicing less poor populations and trimming down their costs. This denotes
the so-called mission drift of microfinance (Christen, 2001; Christen and Drake, 2002; Mersland
and Strom, 2010; Armendaris and Szafarz, 2011).
Progressively, the focus of these institutions is laid on improving their financial performance with
the aim of attracting external sources of funding. Funds from private and public investors,
especially foreign capital investments, as well as, loans from commercial banks, are turned towards
revenue-generating investments. These new market sources of funding have a focus on profitability
(Ghosh and Van Tassel, 2011), which represents the cheapest source of capital – profits reinvested
into the MFI – required for an MFI to attract external capital. However, the rising
commercialization of MFIs raises new risks. These are mainly macro-economic risks caused by the
increasing interaction with formal financial markets, namely the foreign exchange rate risk and the
risk of financial markets slumps.
The present study has three objectives. First, we ascertain the determinants of MFIs’ financial
performance. Second, we examine the persistence of their profits over time. Third, we examine the
impact of the recent global monetary crisis on MFIs’ profits. The determinants of financial
performance comprise both firm-specific and country-specific factors. The former factors being
under the control of MFIs’ management, whereas the latter factors are related to their countries’
macro-economic and macro-institutional environment. Testing the persistence of MFIs’ profits
enables us to evaluate the speed of convergence of profits toward their long-run equilibrium levels,
and give evidence of the level of competitiveness within the microfinance sector. Evaluating the
impact of the crisis gives evidence on the degree of commercialization of MFIs and their reliance
on external sources of funding.
The paper is organized as follows. In the next section, we examine the existent literature on the
determinants of the financial performance of MFIs. Next, we present our data and our empirical
models and define our variables and assumptions. Then, we present and discuss the results of our
models and conclude.
1. Determinants of financial performance
Our analysis aims at studying whether and how various concepts can ascertain MFIs’ financial
performance. Among the researchers who attempted to determine the drivers of financial
performance in the microfinance sector, we have Nawaz (2010), Cull et al. (2011), Ahlin et al.
(2011) and Kar (2013). Through reviewing the existing empirical literature
1
, we intend to highlight
the main findings of existing studies and shed light on the very mixed results obtained. The
determinants of the financial performance are classified into internal factors and external factors.
1
Milana and Ashta (2012) and Daher and Le Saout (2013) have also surveyed but partially the literature related to the
financial performance of the microfinance industry.
3
1.1. Internal factors
Internal factors are defined as firm-based indicators. They can be sorted into five components:
financing structure, credit risk, efficiency, social outreach and firm size.
There is empirical evidence that financing structure influence the financial performance of MFIs.
Hartarska and Nadolnyak (2007) found that MFIs’ financial performance (measured by operational
self-sufficiency or OSS) is affected by the capital-to-assets ratio. Their results suggest that less
leveraged MFIs have better financial performance. This result is accredited by Ngendahayo (2008),
who found that leverage has a negative influence on OSS and on profitability (measured by return
on assets or ROA) of private MFIs only. Contrasting findings are attributed to Kar (2012), who
found a negative and significant effect of the capital-to-assets ratio on profit efficiency (measured
by return on equity or ROE). This implies that any increase in leverage would increase MFIs’ profit
efficiency.
The results of the literature are not homogenous regarding the impact of portfolio quality on
financial performance. Mersland and Strøm (2009), Kar (2012), and Nawaz (2010) found no
significant impact of the portfolio at risk over 30 days ratio on ROA. While, Kar (2013) found a
negative and significant relationship between the ratio of portfolio at risk over 30 days and financial
performance – measured by a composite factor obtained through factor analysis involving multiple
variables measuring profitability (ROA) and financial sustainability (OSS and self-sufficiency
index or SSI).
In terms of cost efficiency, Kar (2013) found that cost per dollar lent ratio has a negative influence
on financial performance. Whereas, Cull et al. (2007) claimed that the effect of costs on ROA
depends on the firm’s lending methodology.
Empirical findings regarding the two components of social outreach, i.e. depth of outreach and
breadth of outreach, are mixed. The impact of depth of outreach is measured through the variable
loan size. Mersland and Strøm (2009) found that ROA increases with average loan size, and
Ngendahayo (2008)’s results suggested a significant and positive impact of this variable on ROA
but only for private MFIs. Though, this result is inconsistent with Galema et al. (2012) who found
that average loan size impacts significantly and negatively ROA. Similarly, Nawaz (2010) found
that average loan size to GNI per capita has a significant and negative effect on ROA. This
relationship does not seem significant in other studies (Cull et al., 2007; Cull et al., 2011). Cull et
al. (2007) suggest that MFIs that make smaller loans are not necessarily less profitable. The
breadth of outreach is measured through the share of loan portfolio to assets. Nawaz (2010) found
no significant effect of gross loan portfolio to assets ratio on ROA. However, Cull et al. (2007)
found a significant and positive impact of the share of microcredit portfolio on ROA. Similarly, Kar
(2013) found a significant and positive impact of the size of the gross loan portfolio on the variable
indicating financial performance (obtained by factoring ROA, OSS and SSI) in his main data set,
but the results of its two other data sets revealed no significant impact.
Empirical findings regarding the impact of firm size on financial performance are mixed. The
number of studies that have not found any association between firm size and ROA (Ngendahayo,
2008; Hartarska, 2004) or ROE (Kar, 2012) is limited. Most of the studies demonstrate that MFI’s
size affects positively its ROA (Vanroose and D’Espallier, 2009; Mersland and Strøm, 2009; Cull et
al., 2007; Bassem, 2009; Cull et al., 2009).
4
1.2. External factors
External factors are those that are not under the control of firm management. They are specific to
the country where the MFI is located. Two groups of country-related indicators can be estimated:
macro-economic indicators and macro-institutional indicators. Additionally, we recall the results of
previous research on the effect of the global monetary crisis.
The first macro-economic indicator is overall economic development. Empirical findings
regarding its impact on financial performance are mixed. For instance, Imai et al. (2012)’s results
show that countries with higher gross domestic product (or GDP) per capita have higher return on
asset, suggesting a significant and positive impact of economy size on MFIs' profitability. Yet,
Bassem (2009), and Hartarska and Nadolnyak (2007) found no significant evidence that GDP per
capita has an impact on either profitability (ROA) or financial sustainability (measured by
operational self-sufficiency or OSS). Nonetheless, Vanroose and D’Espallier (2009) found a
significant and negative impact of gross national income (or GNI) per capita on ROA.
Empirical findings regarding the effect of the second macro-economic indicator, i.e. inflation, on
profitability are also mixed. For instance, Vanroose and D’Espallier (2009) found that MFIs are less
profitable where inflation is high, suggesting that MFIs benefit from stability of the formal financial
system. This result is consistent with Bassem (2009) and Cull et al. (2011) who found a significant
and negative impact of inflation on ROA and OSS. Cull et al. (2009)’s results suggest that the
impact of inflation on financial self-sufficiency (FSS) is significant and negative, whereas the
impact on profitability (ROA) is not significant. Yet, Ahlin et al. (2011)’s findings suggest that
there is no significant impact of inflation on OSS. Conversely, Hartarska and Nadolnyak (2007)
observed a significant and positive impact of inflation on OSS.
Regarding the macro-institutional factors, mitigated results are reported for the financial
sustainability indicator (OSS). Crabb (2008) found that MFIs operate primarily in countries with a
relatively low degree of overall economic freedom and that various economic policy factors are
important to financial performance, such as government intervention that may lower financial
sustainability. Also, Hartarska and Nadolnyak (2007) found no evidence of a significant impact of
economic freedom on OSS. Similarly, empirical results have not yet revealed evidence of a
significant impact of the freedom from corruption indicator on OSS (Ahlin et al., 2011). Moreover,
Hartarska and Nadolnyak (2007) found no evidence of significant impact of the property rights
protection index on OSS.
Lastly, the long held notion that microfinance, as an asset class, is largely uncorrelated to global
financial markets, and that it is an exception to the inherent instability of the traditional financial
system, can no longer be endorsed. A few studies discussed that microfinance is not isolated from
global financial turmoil as it seemed to be (Littlefield and Kneiding, 2009; Dokulilova et al.,
2009; Kruijff and Hartenstein; 2013). This may be caused by several interdependent factors
including: the rise of competition among MFIs (Assefa et al., 2013), the shift of their objectives
from exclusively helping the poor get rid of poverty to maximizing profits, and the new trend of
raising funds from domestic and international capital markets, commercial banks, as well as
microfinance investment vehicles to foster credit growth (El-Zoghbi et al., 2011). Gonzalez (2011)
pointed out at the fact that the impact of financial crises on MFIs and their clients depends on
several factors including: the macro-economic environment, the level of integration of the country
to the global economy, cost and funding structures for the MFI, and the ability of management to
deal with crises. Dokulilova et al. (2009) predicted “the full effect of the worldwide crisis and its
triple shocks – economic contraction, currency depreciation and scarcity of credit, to fully show its
5
power in emerging countries in the second half of year 2009”. Empirically, Di Bella (2011) found a
significant correlation between the performance of MFIs and the domestic and international
financial and economic conditions when adding the years 2008 and 2009 to the analysis. Lensink
(2011) found that MFIs face significant negative shifts in performance indicators related to
profitability, growth, and portfolio quality as a consequence of the financial crisis. Wagner and
Winkler (2012) found that the global financial crisis had a strong negative impact on credit growth
of MFIs. Lützenkirchen et al. (2012) reported that following the crisis, asset growth slowed
markedly, portfolio risk rose, and profitability declined.
2. Modeling financial performance
Many benefits come from using panel data sets (Baltagi, 2005). Compared to pooled data sets –
pure cross-section or pure time-series, panel data sets include much larger data, and allow
constructing and testing more sophisticated behavioral models with less restrictive assumptions.
With panel data sets, data are more informative, estimates are more reliable, collinearity is lower
among variables, and efficiency is greater with more degrees of freedom. Additionally, panel data
sets allow controlling for individual heterogeneity, which are unobserved individual-specific effects
that have been omitted in pooled data. Most importantly for our study, panel data sets are better able
to study complex issues of dynamic behavior, in particular to understand the dynamics of
adjustment.
2.1. Data
Data are collected from four information sources. The Mix Market
1
provides our sample with
financial ratios and other firm-specific data. Macro-economic indices are obtained from the World
2
Development Indicators (WDI) database and Bloomberg. Macro-institutional data are acquired
from The Heritage
3
Foundation.
The hard core of our sample composed of firm-specific time-variant information is put up in five
steps. Firstly, we choose only MFIs delivering information yearly between 2005 and 2011.
Secondly, we keep exclusively MFIs that have observations for ROA variable without disruption
between 2005 and 2011 – 7 consecutive years. At this stage, our data set involves 857 MFIs. Next,
we select those MFIs that publish audited financial statements every year, and those which, in
addition to the latter, have their rating or due diligence reports published yearly. In other words, we
retain 372 MFIs ranked four and five diamonds, respectively
4
. Furthermore, we exclude those MFIs
that do not hold a defined legal status. For instance, the two MFIs holding the legal status Other
were removed. Lastly, some MFIs having approved the abovementioned selection conditions
incorporate observations that lie an abnormal distance from the rest of observations. We investigate
these points carefully before considering their possible elimination from our dataset. For 8 out of
370 retained MFIs, some extreme values were identified on a short time period, ranging from a
single year to a maximum of two consecutive years – no tangible tendency was detected.
Finally, we obtain a panel data set – combined time-series and cross section data – composed of 362
MFIs from 73 countries between 2005 and 2011. This is so far the largest data set in terms of
sample size and length of the study period, featuring the latest financial crisis era. Our sample is
defined as a short balanced panel data set. In other words, it has a large number of entities but few
time periods – this is common in microeconomic data (Wooldridge, 2002) – and the time periods
for which we have data are the same for all entities.
6
2.2. Empirical model and tests of hypotheses
This study analyzes the dynamics of profitability in the global microfinance industry. The study of
the dynamic performance-competition mechanism is new, even among the banking industry
research (see, Berger et al., 2000; Agostino et al., 2005; Goddard et al., 2010 and 2011). We
estimate a dynamic panel data model to control for profit persistence, while investigating the
determinants of the profitability of MFIs. Testing the persistence of MFIs’ profits enables us to
evaluate the speed of convergence of profits toward their long-run equilibrium levels, and give
evidence of the degree of competition within the microfinance sector. Additionally, we tested for
heterogeneity across years in order to inspect the impact of the international financial crisis on
MFIs’ profitability.
2.2.1. Dynamic panel data models
The multivariate dynamic panel data regression model expresses profitability as a linear function of
one lag of the profitability variable, time-variant firm-specific factors, and time-variant country-
specific factors. We estimate two model equations: the first one controls for individual-specific
effects only (one-way error component model) and the second one controls for both individual- and
time-specific effects (two-way error component model).
One-way error component model equation:
Two-way error component model equation:
where,
represents the financial performance or profitability of MFI located in country at
period ; is the regression constant, the mean of the remaining unobserved heterogeneity;
is the one-period lagged dependent variable used as a right hand-side regressor; is the coefficient
of the lagged variable representing the speed of adjustment to the equilibrium; is the coefficients
vector of the regressors;
is a vector of internal factors of MFI at period . These
characteristics vary across MFIs and time;
is a vector of external factors in country at
period . These indicators vary across countries and time;
is the unobserved individual-specific
effects, which represents heterogeneity across MFIs;
is the unobservable time-specific effects,
which represents heterogeneity over time and are introduced in the model via the year dummies.
Time dummies are exclusive for the two-way model. In the one-way model,
is equal to zero;
is the idiosyncratic error term;
is the remainder disturbances stochastic.
The basic hypotheses of the one-way model are: the individual-specific effect
is independent of
the error term and the exogenous regressors, which are also independent of each other for all
and; the disturbance includes the idiosyncratic errors and the individual-specific effect, i.e.,
; and the effects and the errors are independent identically distribute for all and ,
i.e.,
!""#$%&
'
(
) and
!""#$%&
*
(
).
7
The basic hypotheses of the two-way model are: the individual- and time-specific effects are
independent of the error term and the exogenous regressors, which are also independent of each
other for all and; the disturbance includes the idiosyncratic errors and the individual- and time-
specific effects, i.e.,
; and the effects and the errors are independent identically
distribute for all and , i.e.,
!""#$%&
'
(
),
!""#$%&
+
(
) and
!""#$%&
*
(
).
2.2.2. Estimators and tests of hypotheses
By construction,
is correlated with the unobserved individual-level effect
. In order to
produce consistent estimators with fixed ,, the solution is to first-difference the model equation and
then construct estimators based on moment equations using lagged levels of the dependent variable
with first-differenced errors. The moment conditions formed by assuming that particular lagged
levels of the dependent variable are orthogonal to the differenced disturbances are known as
generalized method of moment or GMM-type conditions. We used two estimation techniques: first-
difference generalized method of moments and system generalized method of moments. Developed
by Arellano and Bond (1991), the first-difference generalized method of moments, or standard
GMM, estimator forms moment conditions using lagged-levels of the dependent variable and
strictly exogenous variables with first-differences of the disturbances. Arellano-Bover (1995) and
Blundell-Bond (1998) suggested that if the autoregressive process is too persistent, the lagged-
levels are weak instruments. So, they developed the System generalized method of moments, or
System-GMM, estimator using additional moment conditions, in which lagged differences of the
dependent variable are orthogonal to levels of the disturbances.
Both estimations can be made through one step or two steps. The one-step estimator assumes
homoscedastic or identically distributed errors, while the two-step estimator uses the first-step
errors to build heteroskedasticity-consistent standard errors. According to Arellano and Hahn
(2007), the two-step estimator is asymptotically more efficient even in the presence of
homoscedastic error terms. Subsequently, the instrumentation can be made transparent through a
two-stage least square (2SLS) procedure, where in the first step the regressors are regressed on the
instrument, and in the second step, the resulting estimates are used as explanatory variables in the
equation of original interest. Additionally, using two-step robust GMM-type estimators produces
models robust to heteroskedasticity and gives more efficient estimates for the model and unbiased
standard errors estimates. Finally, both estimators are based on the assumption of no serial
correlation.
The consistency of the standard GMM and the system GMM estimators relies on two hypotheses.
First, we test for the over-identifying restrictions using the Sargan test. The null hypothesis of the
Sargan test is that the model and the over-identifying conditions are valid. Rejecting this null
hypothesis implies that the model or the instruments need to be reexamined, unless the rejection is
due to heteroskedasticity in the data generating process. Second, we test for serial correlation using
Arellano-Bond (1991)’s test for first-order and second-order autocorrelation in the first-differenced
disturbances. The null hypothesis is that there is no autocorrelation in the first-differenced error at
order one and at order two. Though, when the idiosyncratic errors are independently and identically
distributed, i.e.,
!""#$%&
*
(
), the first-differenced errors are first-order serially correlated or
AR(1).
In general, the results of the GMM estimator give a value of convergence speed, or
, between zero
and one at confidence level of 95%. Any shock to profits will persist before converging to their
long-run equilibrium level, i.e., their industry average, after a certain time. The speed at which
profits revert to normal is slower as
is closer to one. A convergence speed closer to one implies
8
persistence of profits, which is a sign of lower competitiveness degrees among MFIs. Whereas, a
value equal to zero indicates zero persistence of profits and implies that competition is sufficiently
fierce that an abnormal profit earned in a certain year does not persist at all into the following year
(Goddard et al., 2004). And, a value close to zero implies a rapid adjustment of profits and
abnormal profits that fade away over time. This suggests that competition is less fierce because
barriers to entry or other impediments to competition are effective to some extent (Goddard et al.,
2004).
2.3. Variables and assumptions
Profitability indicates the extent to which a business generates earnings from its assets as compared
to its expenses and other relevant costs incurred during a specific period of time. We use the ratio of
return on assets to gauge profitability. The return on assets ratio, abbreviated by ROA, also known
as the return on investment, shows how well an institution used its assets to generate returns. In
other words, it measures the efficiency with which the management has used its resources to obtain
income. It is calculated as the net operating income, less taxes, divided by the average total assets
(The Mix Market, 2014; The Seep Network, 2010).
A large number of studies evaluated MFIs’ financial performance expressed by the ability to
survive over time, using financial sustainability indicators, namely the ratio of operational self-
sufficiency (OSS), the ratio of financial self-sufficiency (FSS) and the self-sufficiency index (SSI).
However, the use of profitability measures rather than financial sustainability measures is
recommended in the guide on Microfinance Financial Reporting Standards (The Seep Network,
2010 page 2) in assessing the financial performance of MFIs. The guide justified the rejection of
OSS and FSS to the fact that they become less helpful as measures of financial performance once an
MFI exceed the breakeven point, i.e., 100% sustainability. Thus, the return on asset ratio is better
suited to analyze an established MFI’s financial performance.
We use the variable one-year lagged profitability (ROA
-1
) to evaluate the persistence of profits
over time. The coefficient relative to this variable in the dynamic model indicates the speed of
convergence of profits to their long-run equilibrium levels. We assume that previous profitability
values have a positive impact on the current profitability values. We also predict the speed of
convergence not be extremely fast, since competitiveness levels are not generally high enough in
the microfinance sector.
We select our regressors mainly after examining the existing literature relating the determinants of
financial performance expressed as a function of internal factors and external factors. Table 1
summarizes the assumptions discussed in the present section.
2.3.1. Internal factors
Internal factors are MFI-specific variables that are within the control of the management of the
institutions and vary across years.
Financing structure can be assessed using the variable capital-to-assets ratio. It is defined as the
ratio of equity capital to gross total assets, and is used as a proxy for equity financing. Capital-to-
assets ratio is considered as a standard inverse measure of debt-to-equity ratio or financial leverage.
An MFI should prevent itself from excessively leveraging its own funds to finance its portfolio. An
excessive leverage would increase its risk profile. Accordingly, capital-to-assets ratio is
hypothesized to have a positive impact on profitability.
9
Credit risk can be measured by the variable portfolio at risk past 30 days to gross loan portfolio
ratio. Abbreviated by PAR30, it gives the percentage of loans outstanding that are at substantial risk
of default. An increase in credit risk default is associated with a deterioration of loan portfolio
quality and further impairments and losses to the firm. We therefore assume that the impact of this
factor on profitability is negative.
Efficiency can be estimated using an indicator of cost efficiency represented by the variable cost
per dollar lent. It is defined as the ratio of operational expenses to gross loan portfolio. A declining
trend of this indicator means that the MFI’s efficiency is improving. Accordingly, we assume a
negative impact of cost per dollar lent on profitability.
Social outreach covers two aspects. Depth of outreach refers to the quality of outreach to the poor
and can be measured by the ratio average loan balance per borrower to GNI per capita. This is a
proxy of loan size: the smaller the loan size, the greater is the depth of outreach. Breadth of
outreach refers to the degree of coverage or the scale of outreach to the poor. It can be assessed by
the ratio of gross loan portfolio to total assets. It is indicator of the focus of an MFI on lending to
microfinance clients: the larger the loan portfolio, the greater is the breadth of outreach.
Nonetheless, we cannot clearly predict the impact of these aspects on profitability
Firm size is assessed by the natural logarithm of total assets. This variable controls for
diseconomies or economies of scale, as well as, effects of differences in technology, investment
opportunities and diversification among MFIs. We assume that the impact of firm size on
profitability is positive.
2.3.2. External factors
As argued by Ahlin et al. (2011), “any comparison that does not take into account the macro-
economic and macro-institutional environment, if these are found to non-negligibly predict MFI
performance, is incomplete. Accounting for context allows a clearer picture of institutional success
and failure to emerge.” We select a number of macro-economic and macro-institutional factors that
might affect MFIs’ profitability.
Overall economic development is primarily measured by gross national income per capita
converted to USD using the World Bank Atlas method. This indicator, abbreviated by GNIPC, is
assessed as the sum of incomes of all resident producers of an economy in a given period divided
by the midyear population. It is used as a broad measure of average living standards and economic
well being of a country’s population. We assume that MFIs would yield larger profits in countries
where the population have comparably low living standards or in economically less developed
countries where the need for microfinance services is higher. Therefore, we predict a negative effect
of GNIPC on MFIs’ profitability.
Inflation specifies the rate of price change in the economy as a whole. The inflation rate is
measured by the consumer price index, which reflects the annual percentage change in the cost to
the average consumer of acquiring a basket of goods and services that may be fixed or changed at
specified intervals, such as yearly (Laspeyres formula). On the one hand, higher inflation could
indicate higher incomes from operations, i.e. higher nominal interests received on the microcredit
portfolio, and therefore higher MFIs profitability. On the other hand, higher inflation could indicate
higher financing costs, i.e. higher interests paid on borrowings, and therefore lower MFIs
profitability. We investigate the impact of mis-anticipating inflation, i.e. anticipating an inflation
10
rate equal to the one year-previous inflation rate, on MFIs’ current profitability using the variable
one-year lagged inflation. In other words, we study whether the MFI has fully anticipated the
inflation rate, which means that it can appropriately adjust its interest rates in order to raise incomes
faster than costs, and thus realize higher profits
2
. Accordingly, we assume that the impact of the
one-year lagged inflation rate on profitability is negative.
We incorporate a new variable, the annual variation of period average official exchange rate in local
currency units relative to the US dollar, to capture the foreign exchange rate risk. It is the first
time that this variable is surveyed amongst the determinants of MFIs’ profitability. In general, MFIs
get funding mainly in form of subsidized grants, loans and equity capital from international
organizations and commercial lenders. These funds are usually acquired in foreign currency, such
as Euro or US dollar – approximately 70% of MFI cross-border borrowing is denominated in hard
currency (Reille and Forster, 2008). Though, MFIs lend money and are paid back in domestic
currency of the country in which they are established and/or operating. Some of them also collect
deposits in local currency. This results in a currency mismatch in MFIs’ balance sheets, where
liabilities are in hard currency and assets in local currency (see Featherston et al., 2006). A
devaluation of the local currency against the hard currency, such as US dollars, would raise the cost
of funds for MFIs since more units of the local currency are required to pay the interest on hard
currency loans. This expands the gap between the cost of funding and the profit from lending. If this
occurs, it would increase the risk of default of MFIs and eventually render their profitability.
Accordingly, foreign exchange rate risk is expected to have a negative effect on profitability.
We use two macro-institutional factors, which are proxies for the quality of the institutions in a
specific country. Property rights protection is linked to the quality of the investment climate in
the country, since it refers to the ability of individuals to accumulate private property secured by
clear laws that are fully enforced by the country. Financial freedom is a measure of banking
efficiency, as well as, an indicator of independence from government control and interference in the
financial sector. We assume that both variables have a positive influence on profitability of MFIs.
And finally, we decide to include year dummies as time period indicators to evaluate the
hypothetical influence of the recent global financial crisis, of late 2007 - early 2008, on the
profitability of MFIs. We assume that the crisis has effectively had an impact on the microfinance
sector and that the years following 2007 would much likely have had a negative effect on MFI
profitability. The influence of the remaining years on profitability is uncertain.
-./0112345607
3. Empirical results
Relatively limited are the studies that measured financial performance using profitability indicators.
Besides, some results are conducted on specific groups of MFIs, such as for-profit MFIs, or MFIs
located in a particular country or region, and thus cannot be generalized. Also, econometric
methodologies applied in most of the studies are restrictive. Some have examined the factors
influencing profitability through one-period cross-sectional ordinary least square regressions and
others through time-series pooled ordinary least square regressions, without taking into
consideration the potential variations due to individual-specific effects and to time-specific effects.
Actually, a limited number of microfinance studies have researched the impact of the last global
2
Also, for the low-income population, when inflation rate is not correctly anticipated, it leads to a decline in their
purchasing power, which could engender an augmentation of the credit default risk due to difficulties in paying back
their loans.
11
financial crisis on MFIs’ profitability. Moreover, the level of competitiveness among MFIs is not
largely explored. All these drawbacks are tackled in the present study.
3.1. Descriptive analysis
Univariate descriptive statistics of the outcome variable and explanatory variables are featured in
Table 2. Descriptive analysis reveals some valuable information regarding the distribution of the
data. For instance, the tables exhibit relatively low average ratio of return on assets (3%) for the
whole sample. However, this shows that the MFIs are profitable in average, as they yield a positive
average return. The average capital-to-assets ratio is 31%. This reveals a good financing structure of
the MFIs, as in average they hold more than 8% of capital against their asset (Basel I requirement).
Surprisingly, the average portfolio at risk past 30 days ratio is only 6%. This is a relatively a low
level of credit risk. Like the average PAR30, most of the obtained figures are not much appealing
because they do not take into consideration the heterogeneity between MFIs and across years.
-./0112348607
The split by region (see Figure 1) uncovers some disparities. In terms of mean, the lowest
profitability is recorded for Sub Saharan Africa; where the MFIs might lack capital, competent
staff, efficient management and information systems, free and developed financial markets, non-
corrupted institutions, stable macro-economic and political environment, etc. The highest average
return on asset ratio is recorded for Latin America and the Caribbean. This region comprises the
most developed microfinance activity. Middle East and North Africa has the highest capital-to-asset
ratio, where more than 74% of the firms are Non-governmental organizations and 26% are non-
bank financial institutions. In this region, the MFIs have easy access to external financial sources,
as the Middle Eastern and North African financial markets are very liquid and the share of foreign
investments is important in the region despite unrest (Agence France Presse, 2013). South Asia has
the lowest average loan balance per borrower to GNI per capita and the highest cost per dollar lent.
The MFIs from this region are operating in the most populated countries in the world (see Table 3),
where the percentage of population living in poverty is of the highest. The South Asian MFIs,
therefore, are forced to grant small loans and thus face high operating costs. Eastern Europe and
Central Asia has obviously the highest average loan balance per borrower to GNI per capita where
34% of the MFIs are banks and 41% are non-bank financial institutions; they also have the lowest
portfolio-at-risk past 30 days. The highest levels of share of gross loan portfolio to asset ratio and
natural logarithm of total asset are seen in Latin America and the Caribbean and Eastern Europe and
Central Asia, where the microfinance industry is well developed.
-./0112349607
The break up by year (see Figure 2) gives information on the evolution of MFIs’ performance. As
for the average return on asset ratio, between 2005 and 2007 the progression was positive, then it
decreased dramatically between 2007 and 2009, before gradually and slowly augmenting after
2009. In line with this progression, the average capital-to-asset ratio followed a downturn during the
whole period. While, average portfolio at risk past 30 days ratio had a dramatic rise in 2008 and
continued increasing till 2010. Cost per dollar lent fell in 2005 then increased slightly after 2007.
Average loan balance per borrower to GNI per capita dropped dramatically in 2007 then recovered
slowly from 2009 to 2010 before dropping again in 2011. Average share of gross loan portfolio to
asset ratio increased from 2005 to 2008, decreased severely from 2008 to 2009, and then started to
rise again after 2009.
12
-./01:;0/5<8607
Descriptive analyses conducted in this section allow us to realize the interdependence of indicators,
and points out at the potential effects of the recent global financial crisis on MFIs performance
indicators. However, descriptive statistics do not allow us to make conclusions. Therefore, we
conduct inferential statistics to determine accurately which variables are influencing profitability,
from within MFIs themselves and their environment.
3.2. Empirical analysis
The present section displays and discusses the empirical results related to the questions that this
paper aims at answering. First, we estimate the persistence of MFIs profits over time and we
explore the determinants of profitability from within the MFIs and their country-level context.
Second, we investigate the effect of the recent global financial crisis on the profitability of MFIs,
while verifying the robustness of our early results to the inclusion of time dummies.
3.2.1. Profit persistence and profitability drivers
We estimate the multivariate dynamic individual-specific effects panel data regression model using
both standard GMM and system GMM estimators
5
. Tables 4 and 5
6
report a stable convergence
speed across specifications of almost 52% and a significant positive effect of the one-year lagged
profitability on current profitability values. This is evidence that profits persist over time and that
the speed of returning to their normal levels is relatively medium to slow. This result also suggests
that the global microfinance industry still have a long path to achieve high competitiveness levels.
The level of competitiveness is linked to the level of competition among MFIs. The effect of
competition is however unclear. As argued by Assefa et al. (2013), it could go both. On the one
hand, competition may result in improved and new financial product designs, better customer
services, lower costs and lower interest rates, which may help overcome the adverse effect of
competition without risking growth of the microfinance sector. On the other hand, competition may
lead to low screening and lending standards and increased information asymmetry, resulting in
multiple borrowing, heavy debt burdens, low repayment rates and poor portfolio quality.
The results of estimated models show that capital financing has a robust and significant positive
impact on MFIs’ profitability. The coefficients of the capital-to-assets ratio remain high across
specifications. This result is consistent with Hartarska and Nadolnyak (2007) and Ngendahayo
(2008). So, better-capitalized MFIs or less leveraged MFIs are more profitable. This demonstrates
that better-capitalized MFIs would be more efficient and better immunized against unanticipated
operational expenses, because they do not need to adjust their mission in order to get additional
capital. Furthermore, as they are not heavily indebted, they would face lower costs of funding.
In terms of credit risk, results show a robust and significant negative impact of the credit default
risk indicator, i.e., portfolio-at-risk over 30 days to assets ratio on MFIs’ profitability. This effect is
maintained across specifications. It confirms our assumption that a bad credit portfolio quality is
associated with lower profitability, and suggests that MFIs have to manage well the credit default
risk in order to maintain high profitability levels. This result is consistent with Kar (2013).
A robust and significant negative impact on MFIs’ profitability is also reported for the cost
efficiency indicator, i.e., cost per dollar lent. This effect is maintained across specifications with the
highest t-statistics. This constitutes evidence that inefficient cost management, translated by an
13
increase in operational costs, constitutes the most important internal risk that would lower MFIs’
profitability. This result supports Kar (2013)’s results on cost inefficiency.
Regarding the proxies for depth of outreach, unlike Mersland and Strom (2009) and Ngendahayo
(2008) who found a positive effect of loan size, and Galema et al. (2012) and Nawaz (2010) who
obtained a negative effect of loan size, our results suggest no evidence of a significant impact of the
average loan balance per borrower to GNI per capita on MFIs’ profitability. We cannot thus claim
that MFIs with larger loans per borrower have better profitability, which would refer to a decline in
depth of outreach, since offering larger loans means servicing less vulnerable poor clients.
In terms of breadth of outreach, the coefficients of the variable share of microcredit portfolio to
assets are only significant and positive in the basic model (specification number 0). Thus, we cannot
assert that expanding the breadth of outreach by focusing more on providing micro-credits to the
poor improves MFIs profitability. This result is inconsistent with Cull et al. (2007) and Kar (2013),
who suggested that the financial performance of MFIs improves as they scale up, all else being
equal, as larger loan portfolios reduce delivery costs.
We find a significant positive impact on MFIs’ profitability of firm size, measured by logTA.
Unlike Kar (2013) who reported that all large firms are not profitable and all small firms are not
unprofitable, our results suggest that larger MFIs have higher profits as they would benefit from
higher economies of scale. This result supports those found earlier by Hartarska and Nadolnyak
(2007), Mersland and Strom (2009), Cull et al. (2007, 2009) and Bassem (2009).
In terms of economic development, we find that gross national income per capita has robust and
significant negative influence on MFIs’ profitability. This means that MFIs’ profitability is better in
less developed countries, where the living standards are lower and the client-base for microfinance
services is larger. Additionally to being a proxy for economic development, GNIPC can be
interpreted as an indicator of economy size. Subsequently, our results reveal that in larger
economies, or bigger countries in economical terms, MFIs’ profitability is lower compared to
smaller economies. This result confirms the findings by Vanroose and D’Espallier (2009) on ROA,
while Hartraska and Nadolnyak (2007) have found no evidence of a significant effect of per capita
income on OSS.
The variable of lagged inflation rate, which embodies inflation expectations, has a significant
negative impact on MFIs’ profitability. This result indicates that MFIs that do not adequately
anticipate an increase in consumption prices index, i.e. inflation, are less disposed to balancing their
prices, precisely increase their interest rates faster than the increase in salaries and other inputs, and
are confronted to an increase in expenses, which damage their profitability. Moreover, the t-
statistics of lagged inflation are the highest among macro-economic factors. This suggests that
globally MFIs need to develop sufficient safeguards in order to perform successfully in highly
inflationary environments. We recall that Vanroose and D’Espallier (2009, ROA and OSS), Bassem
(2009, OSS) and Cull et al. (2011, OSS) found also a negative effect of inflation on financial
performance. While Hartarska and Nadolnyak (2007) reported a positive effect of inflation on OSS,
and Cull et al. (2009) and Ahlin et al. (2011) found no evidence of a relationship between inflation
rate and ROA nor OSS respectively.
As predicted previously, foreign exchange risk, assessed by the variable annual variation of local
currency unit against USD, has significant negative impact on MFIs’ profitability. This result is
robust across specifications. It suggests that a depreciation of the domestic currency against the hard
currency is associated with higher costs of funding for the MFIs, causing a reduction of their
14
profitability. This also suggests that, due to the increasing commercialization of MFIs and their
growing reliance to funding from international investors, foreign exchange rate fluctuations
constitute an important external risk for MFIs profitability, especially in presence of inappropriate
asset liability management.
We argued that in countries with advanced institution quality, MFIs benefit from healthier
institutional environment and relationships with third parties, particularly government
administrations. This would prevent them from loosing time and money, and improve their
profitability. Though our assumptions regarding the influence of macro-institutional factors on
MFIs’ profitability are not confirmed in the dynamic model analysis. Like Hartarska and Nadolnyak
(2007) and Ahlin et al. (2011), we do not find evidence of a significant effect on profitability of
either the rule of law measured by the property rights protection index, or the market openness
measured by the index of financial freedom.
-./011234=607
-./011234>607
3.2.2. Impact of the recent international monetary crisis
We perform two-way individual- and time-specific dynamic panel data estimations in order to
control for time effects
7
. The results of our model suggest that the year 2009 has a significant
negative effect on profitability, while the year 2006 has a positive impact on ROA (See Table 6).
This confirms our predictions, and those of Dokulilova et al. (2009). Hence, we can argue that the
commercialization of MFIs has made them more sensitive to international financial crises. The
resulting depreciation of MFIs profitability could be linked to the following macro-economic
effects that emerged with the crisis:
the prominent global liquidity contraction raised the cost and availability of funding to MFIs,
especially the non-deposit taking ones, as the money from both domestic and international
banks has become more scarce and expensive, and investors have become more risk averse;
rising foreign exchange rate fluctuations, caused by local currency depreciations, made it more
difficult for MFIs to contract new debts from domestic and international commercial banks, and
to refinance their existing debts especially when they lack appropriate asset liability
management;
higher financial costs due to inflation and liquidity crunch, and higher operating costs due to
inflation on salaries, have raised interest rates for micro-borrowers and reduced demand for new
loans;
increasing food and energy prices and the decreasing purchasing power have incited micro-
borrowers to withdraw their savings to finance consumption, and in some cases, made them fail
to repay their loans;
diminishing remittances from abroad used by micro-borrowers to repay their loans have caused
decline in MFIs repayment rates and reduced demand for new loans;
growing financial problems of MFIs caused the loss of confidence that they could be able to
provide other loans and thus discouraged micro-borrowers to repay their existing loans.
These consequences of the recent crisis on MFIs’ performance shed light on the important role of
local sources of funding, namely local deposits and loans from local commercial lenders.
15
Furthermore, considering profitability drivers, we obtain the same significant coefficients as
previously, except for the variables variation of foreign exchange rates and for the gross national
income per capita (see Table 6), which become insignificant. As discussed beforehand, this could
be justified by the fact that the impact of these macro-economic factors is now captured by the time-
specific effects.
-./011234?607
Conclusion
The main objective of this paper is to investigate the financial performance of MFIs while going
beyond the existing literature in five ways. The first advantage is the appraisal of profitability rather
than financial sustainability, which reflects better the financial performance of established MFIs.
The second improvement is the inclusion of external factors that were not tested so far in similar
models, namely the annual variation of official exchange rates and the financial freedom index,
which take into account the impact of the country-specific context of the MFIs following the
development of the sector and its increasing reliance on domestic and international sources of
funding. Thirdly, given the opportunity to use a time-series cross-sectional data set, we jump at
employing an econometric approach that consists of exploiting panel data econometrics. On the one
hand, the multiple observations on each unit provide superior estimates as compared to cross-
sectional models of association, and thus can be used to inform policy on better drivers of MFIs
profitability. On the other hand, the detection of a change in performance over time ascertains the
occurrence of events, such as financial crises. Accordingly, the fourth pioneering point is the
incorporation of time-specific effects to capture the impact of the recent financial crisis on MFIs’
profitability. Our study period covers years pre- and post- financial crisis. And, the fifth point is the
evaluation of the persistence of profits over time, using a dynamic panel data model, to ascertain the
level of competitiveness of the global microfinance industry.
In order to ascertain the determinants of MFIs’ profitability, we used panel econometrics on a data
set composed of 362 MFIs in 73 countries between 2005 and 2011. The study provides new
evidence on the persistence of profits over time, the impact of internal and macro-economic factors,
as well as, the recent international crisis, on the profitability of MFIs. Our results suggest that MFIs’
profits are persistent from a period to another, and that globally MFIs have quite low levels of
competitiveness. Additionally, we find evidence that better-capitalized MFIs, MFIs with lower
default credit risk, and cost efficient MFIs are more profitable. Furthermore, we show that
profitability is enhanced in less developed economies, with higher inflation and that a deterioration
of local currency renders profitability. Our results also suggest that due to the increasing
commercialization of MFIs and the increasing reliance on external funding, MFIs are being more
sensitive to fluctuations in global financial markets. In particular, the recent global monetary crisis
has driven down in 2009 MFIs’ profits.
16
References
Agostino, M., Leonida, L. & Triveri, F. (2005). Profits Persistence and Ownership: Evidence from
the Italian Banking Sector. Applied Economics, 37(14):1615-1621.
Ahlin, C., Lin, J., & Maio, M. (2011). Where does microfinance flourish? Microfinance institution
performance in macroeconomic context. Journal of Development Economics, 95:2, 105–120.
doi:10.1016/j.jdeveco.2010.04.004
Armendáriz, B. & Szafarz A. (2011), On mission drift in microfinance institutions, in B.
Armendariz and M. Labie (Eds), The Handbook of Microfinance, London-Singapore: World
Scientific Publishing, 341-366.
Arellano, M. & Bond S. (1991). Some Tests of Specification for Panel Data: Monte Carlo Evidence
and an Application to Employment Equations. Review of Economic Studies 58, 277–297.
Arellano, M. & Bover O. (1995). Another Look at the Instrumental Variable Estimation of Error
Component Models. Journal of Econometrics 68, 29–51.
Arellano, M., & Hahn, J. (2007). Understanding bias in nonlinear panel models: Some recent
developments. Econometric Society Monographs, 43, 381.
Assefa, E., Hermes, N., & Meesters, A. (2013). Competition and the performance of microfinance
institutions. Applied Financial Economics, 23(9), 767-782.
Baltagi, B., Econometric Analysis of Panel Data, 3rd Edition, Wiley, 2005.
Bassem, B. S. (2009). Governance and performance of microfinance institutions in Mediterranean
countries. Journal of Business Economics and Management, 10(1), 31–43. DOI:10.3846/1611-
1699.2009.10.31-43
Berger, Allen N., Seth D. Bonime, Daniel M. Covitz, & Diana Hancock (2000). Why Are Bank
Profits So Persistent? The Roles of Product Market Competition, Information Opacity and Regional
Macroeconomic Shocks. Journal of Banking and Finance 24, 1203–1235.
Blundell, R. & Bond, S. (1998). Initial Conditions and Moment Restrictions in Dynamic Panel Data
Models. Journal of Econometrics, 87:115-143.
Christen, R. P., (2001). Commercialization and mission drift: the transformation of microfinance in
Latin America. CGAP: Occasional Paper No. 5.
Christen, R. & Drake D. (2002). Commercialization. The new reality of microfinance. In D. Drake
and E. Rhyne (Eds.), The Commercialization of Microfinance. Balancing Business and
Development, Chapter 1, pp. 2-22. Bloomfield: Kumarian Press.
Crabb, P. R. (2008). Economic Freedom and the Success of Microfinance Institutions. Journal of
Development Entrepreneurship. 13 (2): 205–19.
17
Cull, R., Demirgüç-kunt, A., & Morduch, J. (2007). Financial performance and outreach: a global
analysis of leading microbanks. Economic Journal, Royal Economic Society, vol. 117(517), pp 107-
133, 02.
Cull, R., Demirgüç-kunt, A., & Morduch, J. (2009). Does Regulatory Supervision Curtail
Microfinance Profitability and Outreach ? Policy Research Working Paper Series. 4948, The World
Bank.
Cull, R., Demirgüç-kunt, A., & Morduch, J. (2011). Does Microfinance Regulation Curtail
Profitability and Outreach ? World Development, Elsevier, vol. 39(6), pages 949-965, June.
Daher L., & Le Saout E. (2013). Microfinance and Financial Performance. Strategic Change;
22: 31–45. DOI: 10.1002/jsc.1920.
Di Bella, G. (2011). Impact of the Global Financial Crisis on Microfinance and Policy Implications
(No. 11-175). International Monetary Fund.
Dokulilova, L., Janda, K., & Zetek, P. (2009). Sustainability of microfinance institutions in
financial crisis. European Financial and Accounting Journal, 2: 7–33.
El-Zoghbi, M., G.hwiler, B. & Lauer, K. (2011) Cross-border Funding of Microfinance, CGAP
Focus Note 70, Washington, D.C.
Featherston, S., Littlefield, E., & Mwangi, P. (2006). Foreign Exchange Rate Risk in Microfinance:
What is it and how can it be managed? CGAP Focus Note, 31, 1–16. CGAP.
Galema, R., Lensink, R., & Mersland, R. (2012). Do Powerful CEOs Determine Microfinance
Performance? Journal of Management Studies, 49(4), 718–742. doi:10.1111/j.1467-
6486.2012.01046.x
Ghosh, S., & Van Tassel, E. (2011). Microfinance and competition for external funding. Economics
Letters, 112(2), 168–170. doi:10.1016/j.econlet.2011.03.037
Goddard, J., Liu, H., Molyneux, P. & Wilson J.O.S. (2004). The Profitability of European Banks: A
Cross Sectional and Dynamic Panel Analysis. The Manchester School, 72:363-381.
Goddard, J., Liu, H. & Molyneux, P. (2010). Do Bank Profits Converge?. European Financial
Management, doi:10.1111/j.1468-06X.2010.00578.x.
Goddard, J., Liu, H., Molyneux, P. & Wilson J.O.S. (2011). “The Persistence of Bank Profit”
Journal of Banking and Finance 35:2881-2890.
Gonzalez, A. (2011). Lessons for Strengthening Microfinance Institutions through Financial Crises,
Fluctuations in Food and Fuel Prices, and Other Major Risks. Micro Banking Bulletin.
Hartarska, V., & Nadolnyak, D. (2007). Do regulated microfinance institutions achieve better
sustainability and outreach? Cross-country evidence. Applied Economics, 39(10), 1207–1222.
doi:10.1080/00036840500461840
18
Imai, K. S., Gaiha, R., Thapa, G., Annim, S. K., & Gupta, A. (2012). Financial Performance of
Microfinance Institutions-A Macroeconomic and Institutional Perspective. Discussion Paper Series
DP2012-04, Research Institute for Economics & Business Administration, Kobe University.
Kar, A. K. (2012). Does capital and financing structure have any relevance to the performance of
microfinance institutions?, International Review of Applied Economics, Taylor & Francis Journals,
vol. 26(3), 329-348, March.
Kar, A. K. (2013). Mission drift in microfinance: are the concerns really worrying? Recent cross-
country results, International Review of Applied Economics, Taylor & Francis Journals, vol. 27(1),
44-60, January.
Kruijff, D. & Hartenstein, S. (2013). Microfinance and the Global Financial Crisis: A Call for
Basel. IFC.
Lensink, R. (2011). Recent Developments in Microfinance and the Impact of the Financial Crisis.
Ethical perspectives-Katholieke Universiteit Leuven, 18(4), 569-590.
Littlefield, E. & Kneiding, C. (2009). The global financial crisis and its impact on microfinance.
CGAP Occasional Paper.
Lützenkirchen, C., Weistroffer, C., Speyer, B., & AG, D. B. (2012). Microfinance in evolution. An
industry between crisis and advancement,„Deutsche Bank Research “, September, 13.
Mersland, R. & Øystein Strøm, R. (2009). Performance and governance in microfinance
institutions. Journal of Banking & Finance, 33(4), 662–669. doi:10.1016/j.jbankfin.2008.11.009
Mersland, R. & Øystein Strøm, R. (2010). Microfinance mission drift?. World Development, 38,
28-36.
Milana C., & Ashta A. (2012). Developing microfinance: A survey of the literature. Strategic
Change; 21: 299–330. DOI: 10.1002/jsc.1911.
Nawaz, A. (2010). Issues in Subsidies and Sustainability of Microfinance: An Empirical
Investigation. Working Papers CEB 10-010.RS, Universite Libre de Bruxelles.
Ngendahayo, E. (2008). Statut juridique, gouvernance et performance des institutions de
microcrédit : Une étude empirique sur les données du MIX.
Reille, X., & Forster, S. (2008). Investissements étrangers en microfinance: Conjuguer objectifs
sociaux et financiers. CGAP, (44).
Tchakoute-Tchuigoua, H. (2010). Is there a difference in performance by the legal status of
microfinance institutions? The Quarterly Review of Economics and Finance, 50(4), 436–442.
doi:10.1016/j.qref.2010.07.003
The Seep Network (2010). Pocket Guide to the Microfinance financial reporting measuring
financial performance of microfinance institutions.
19
Vanroose, A., & D’Espallier, B. (2009). Microfinance and Financial Sector Development. Working
Papers CEB 09-040.RS, Universite Libre de Bruxelles.
Wagner, C., & Winkler, A. (2013). The vulnerability of microfinance to financial turmoil–evidence
from the global financial crisis. World Development, 51, 71-90.
Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data, MIT Press.
20
Tables
Table 1 Formulation of assumptions
Factors Proxies Relationship
Lagged dependent variable
Profit persistence One-year lagged Return on assets: ROA
-
1
Positive
Internal factors
Financing structure Equity financing: CAR Positive
Credit risk Default risk: PAR30 Negative
Inefficiency Cost-inefficiency: CPDL Negative
Social outreach Depth of outreach: LBBGC Indefinite
Breadth of outreach: GLPA Indefinite
Firm size Total assets: logTA Positive
External factors
Overall economic development Economic strength: GNIPC Negative
Inflation expectations One-year lagged inflation rates: INFL
-
1
Negative
Depreciation of local currency Foreign exchange rate variation: FXVAR Negative
Quality of institutions Rule of law and market openness: PR, FF Positive
Time dummies
Financial crisis Before 2007 Indefinite
After 2007 Negative
21
Table 2 Univariate descriptive statistics (whole sample)
Variable N Mean Median St. Dev. Min. Max.
Return on assets 2534
0.03
0.03
0.06
-0.52
0.36
Capital to assets ratio 2534
0.31
0.23
0.23
-1.20
1.00
Portfolio at risk over 30 days ratio 2388
0.06
0.03
0.09
0.00
1.00
Cost per dollar lent 2531
22.14
17.12
22.79
0.85
670.18
Loan balance per borrower to GNI per capita 2491
0.63
0.29
1.28
0.02
33.93
Gross loan portfolio to assets 2532
0.78
0.81
0.15
0.03
1.54
Log Total assets 2534
16.65
16.51
1.82
10.28
22.36
GNI per capita 2498
2553.74
1890.00
2058.57
100.00
12250.00
One-year lagged inflation 2483
0.07
0.06
0.05
-0.10
0.53
Variation of foreign exchange rate 2482
0.00
0.00
0.07
-0.28
0.66
Property rights 2421
33.38
30.00
11.05
0.00
90.00
Financial freedom 2421
49.27
50.00
14.75
10.00
90.00
22
Table 3 Number of MFIs by region and country
Region Country (Number of MFIs) Total
East Asia and the Pacific Cambodia (12), China (2), East Timor (1), Indonesia (4), Papua New
Guinea (1), Philippines (23), Samoa (1), Thailand (1) and Vietnam
(4) 49
Eastern Europe and Central Asia
Albania (3), Armenia (5), Azerbaijan (7), Bosnia and Herzegovina
(11), Bulgaria (3), Georgia (4), Kazakhstan (4), Kosovo (6),
Kyrgyzstan (3), Macedonia (4), Moldova (1), Mongolia (2), Romania
(3), Russia (2), Serbia (3), Tajikistan (6) and Ukraine (1)
68
Latin America and the Caribbean
Argentina (1), Bolivia (14), Brazil (6), Chile (1), Colombia (6), Costa
Rica (5), Dominican Republic (3), Ecuador (19), El Salvador (7),
Guatemala (7), Haiti (2), Honduras (10), Mexico (4), Nicaragua (10),
Panama (2), Paraguay (5), Peru (29) and Venezuela (1)
132
Middle East and North Africa Egypt (8), Iraq (1), Jordan (4), Lebanon (2), Morocco (1), Palestine
(4), Tunisia (1) and Yemen (2) 23
South Asia Afghanistan (1), Bangladesh (15), India (26), Nepal (11), Pakistan (6)
and Sri Lanka (2) 61
Sub Saharan Africa Benin (2), Burkina Faso (1), Burundi (1), Congo (1), Ghana (2),
Kenya (6), Malawi (1), Mozambique (2), Niger (1), Nigeria (2),
Senegal (2), South Africa (2), Tanzania (1), Togo (1) and Uganda (4) 29
Total 362
23
Table 4 Estimation results for the dynamic individual-specific effects model
using standard GMM estimator
ROA
0
1
2
3
L1. .5214277***
(8.74)
.5238241***
(10.04)
.5439977***
(9.56)
.5173531***
(9.87)
CAR .0956222***
(2.99)
.0991903***
(3.09)
.0969537**
(2.58)
.1086677***
(3.35)
PAR30 -.1225789***
(-2.92)
-.1177736***
(-3.29)
-.1365274***
(-2.85)
-.1200275***
(-3.29)
CPDL -.1368661***
(-5.55)
-.1518621***
(-4.15)
-.1275477***
(-6.54)
-.1522114***
(-4.83)
LBBGC .1625502
(0.64)
-.0285982
(-0.10)
.0900344
(0.32)
-.0525113
(-0.17)
GLPA .0441795**
(2.24)
.02449
(1.27)
.0363688
(1.64)
.0270281
(1.40)
logTA 1.013625**
(2.35)
1.192334**
(2.17)
1.129822**
(2.41)
1.340168**
(2.43)
FXVAR
-.045199**
(-2.42)
-.0357308**
(-2.02)
INFL-1
-.121734***
(-5.05)
-.1315359***
(-5.61)
GNIPC
-.0006068**
(-2.13)
-.000673**
(-2.37)
PR
.0499066*
(1.66)
.0393927
(1.42)
FF
.0178938
(0.73)
.0149098
(0.65)
Intercept -18.42727**
(-2.28)
-17.15647*
(-1.80)
-22.4187**
(-2.41)
-21.835**
(-2.22)
Number of
instruments 21
25
24
27
Number of
Observations 1632
1567
1556
1528
Number of MFIs 361
354
349
346
Wald chi2 160.84***
200.51***
157.20***
197.02***
Sargan test (not
robust) 32.8593***
19.46921
32.23715***
17.01652
Arellano-
Bond test
AR(1) -4.851***
-4.8813***
-5.7489***
-5.8473***
AR(2) -.79482
-.31674
-1.22
-1.3193
Robust estimations were performed using standard GMM estimator (xtabond). The t-statistics
are in parentheses and significance at the 10%, 5%, and 1% level is denoted by *, ** and ***
respectively.
Sargan test Ho: over identifying restrictions are valid. Arellano-Bond test Ho: no
autocorrelation. Significance at the 10%, 5%, and 1% level is denoted by *, ** and ***
respectively.
24
Table 5 Estimation results for the dynamic individual-specific effects model
using system GMM estimator
ROA
0
1
2
3
L1. .5287881***
(9.81)
.5130957***
(10.44)
.5392829***
(11.11)
.505603***
(10.56)
CAR .0811497***
(3.05)
.0785256**
(2.59)
.0760249***
(2.60)
.0823895***
(2.77)
PAR30 -.1105683***
(-2.82)
-.1107473***
(-3.11)
-.1362217***
(-2.97)
-.1159829***
(-3.19)
CPDL -.1230045***
(-4.71)
-.139377***
(-3.95)
-.1185193***
(-5.69)
-.1452535***
(-4.70)
LBBGC .2317013
(0.92)
.017788
(0.06)
.1695997
(0.62)
.0075164
(0.03)
GLPA .033579*
(1.73)
.019838
(1.03)
.0252307
(1.21)
.0198487
(1.04)
logTA .9636038***
(2.64)
.9299299*
(1.91)
1.04329**
(2.55)
1.045808**
(2.10)
FXVAR
-.0463103**
(-2.53)
-.0373747**
(-2. 20)
INFL-1
-.1235454***
(-5.28)
-.1324043***
(-5.83)
GNIPC
-.0005322**
(-2.21)
-.0005897**
(-2.37)
PR
.0576648*
(1.90)
.0370363
(1.34)
FF
.0181665
(0.74)
.0178765
(0.81)
Intercept -16.81022*
(-2.49)
-12.2603
(-1.39)
-20.03111**
(-2.50)
-16.02364*
(-1.75)
Number of
instruments 27
30
29
32
Number of
Observations 2036
1965
1946
1916
Number of MFIs 362
357
350
349
Wald chi2 224.02***
285.52***
223.94***
284.04***
Sargan test (not
robust) 38.97211***
22.06245
37.85121***
20.25659
Arellano-
Bond test
AR(1)
-4.9923***
-5.0809***
-6.2728***
-6.2269***
AR(2) -.80765
-.39395
-1.2663
-1.3879
Robust estimations were performed using system GMM estimator (xtdpdsys). The t-statistics are
in parentheses and significance at the 10%, 5%, and 1% level is denoted by *, ** and ***
respectively.
Sargan test Ho: over identifying restrictions are valid. Arellano-Bond test Ho: no autocorrelation. Significance at the
10%, 5%, and 1% level is denoted by *, ** and *** respectively.
25
Table 6 Estimation results for the dynamic one-way and two-way individual- and
time-specific effects models using standard GMM and system GMM estimators
ROA Standard GMM System GMM
L1. .5173531***
(9.87)
.5010052***
(8.62)
.505603***
(10.56)
.4902359***
(10.08)
CAR .1086677***
(3.35)
.0967871***
(3.02)
.0823895***
(2.77)
.0848719***
(2.88)
PAR30 -.1200275***
(-3.29)
-.1277652***
(-3.18)
-.1159829***
(-3.19)
-.1217172***
(-3.05)
CPDL -.1522114***
(-4.83)
-.1567947***
(-4.88)
-.1452535***
(-4.70)
-.1472883***
(-4.94)
LBBGC -.0525113
(-0.17)
-.1085986
(-0.38)
.0075164
(0.03)
-.0622414
(-0.23)
GLPA .0270281
(1.40)
.0300292
(1.52)
.0198487
(1.04)
.0274622
(1.40)
logTA 1.340168**
(2.43)
1.897428***
(2.73)
1.045808**
(2.10)
1.712916***
(2.69)
FXVAR -.0357308**
(-2.02)
-.0198173
(-0.84)
-.0373747**
(-2.20)
-.0193714
(-0.82)
INFL-1 -.1315359***
(-5.61)
-.0948444***
(-3.05)
-.1324043***
(-5.83)
-.0963153***
(-3.32)
GNIPC -.000673**
(-2.37)
-.0001848
(-0.59)
-.0005897**
(-2.37)
-.0001021
(-0.36)
PR .0393927
(1.42)
.0476459*
(1.65)
.0370363
(1.34)
.0432675
(1.54)
FF .0149098
(0.65)
-.0122878
(-0.51)
.0178765
(0.81)
-.0118226
(-0.50)
Intercept -21.835**
(-2.22)
-31.07517**
(-2.51)
-16.02364*
(-1.75)
-27.71516**
(-2.41)
YR2
1.114487***
(3.40)
1.097956***
(3.79)
YR4
-.5412786
(-1.63)
-.4849141
(-1.56)
YR5
-.9812647*
(-1.73)
-.9704114*
(-1.82)
YR6
-.7035021
(-1.13)
-.7327742
(-1.40)
YR7
-.9088596
(-1.22)
-.923559
(-1.53)
Number of
instruments 27
32
32
37
Number of
Observations 1528
1528
1916
1916
Number of MFIs 346
346
349
349
Wald chi2 197.02***
229.06***
284.04***
329.58***
Sargan test (not
robust) 17.01652
19.72653
20.25659
21.1498
Arellano-
Bond test
AR(1) -5.8473***
-5.7601***
-6.0035***
-5.9318***
AR(2) -1.3193
-1.3682
-1.385
-1.4246
Robust estimations were performed using standard GMM and system GMM estimators (xtabond
and xtdpdsys). The t-statistics are in parentheses and significance at the 10%, 5%, and 1% level
is denoted by *, ** and *** respectively.
Sargan test Ho: over identifying restrictions are valid. Arellano-Bond test Ho: no autocorrelation. Significance at the
10%, 5%, and 1% level is denoted by *, ** and *** respectively.
26
Figures
Figure 1 Average profitability by region
Figure 2 Average profitability by year
2,93 2,77
3,55
4,07
1,47
0,96
East Asia and
the Pacific
Eastern
Europe and
Central Asia
Latin America
and the
Caribbean
Middle East
and North
Africa
South Asia Sub Saharan
Africa
ROA
3,1 3,26 3,46
3,16
1,9
2,3 2,38
2005 2006 2007 2008 2009 2010 2011
ROA
27
1
The Mix Market references a database of nearly 2000 MFIs around the world. To date, it is the primary source for
microfinance data and analysis, where MFIs directly publish their financial and organizational information on the online
platform.
2
The World Development Indicators database includes a wide range of national, regional and worldwide estimates of
global development indicators compiled from officially recognized international sources and produced by the World
Bank.
3
The Heritage Foundation publishes an annual index that measures the degree of economic freedom for 158 countries.
The index grades nations on ten factors of economic freedom. The statistics are provided from trustworthy
organizations like The World Bank, the International Monetary Fund and the Economist Intelligence Unit.
4
The Mix Market have set forth a rating system, called diamonds system, to indicate the transparency level of an MFI
and supporting information for all data. According to their website, a higher number of diamonds means a more
transparent MFI and more reliable data. Ahlin et al. (2011), Hartarska and Nadolnyak (2007), and Tchakoute-
Tchuigoua (2010) made this same choice and assume that there is no qualitative difference between four and five
diamonds ranked MFIs, “except that those with a rank of five have at least 3 years of financial statements, while those
with rank four have less than 3 years” (Hartarska and Nadolnyak, 2007, page 1212).
5
We check for the consistency of our estimators. We find that the two-step Sargan test for over-identification does not
reject the null, which means that the models and over-identifying restrictions are correctly specified. Also, we find that
the serial correlation test’s null is rejected and that there is indeed evidence of AR(1) in the first-differenced errors. But,
the null that there is no second order serial correlation or AR(2) is not rejected, which means that there is no significant
evidence of serial correlation in the first-differenced errors at order two. This implies that the used moment conditions
are valid.
6
Tables 4 and 5 report results from four distinct model specifications using the variable return on assets as a proxy for
financial performance. First, we estimate a multiple regression equation with internal factors only (noted basic equation
or model specification number 0). Next, we add consecutively to the basic equation the macro-economic and the macro-
institutional factors. The aim of comparing the results of these models is to check, on the one hand, the robustness of
MFI-specific variables to the inclusion of macro-economic and macro-institutional factors, and on the other hand, the
robustness of the macro-economic factors to the inclusion of the macro-institutional indicators and vice-versa. This way
we avoid the risk of mis-specifying the functional form of the association. The specification number 3 gives our
preferred model.
7
After testing for the time-specific effects with @
A
B
(
C
D
E
% at confidence level of 95%, we drop the
null hypothesis and keep year dummies.
