THE RELATIONSHIP BETWEEN PROFITABILITY OF MICRO FINANCIAL INSTITUTIONS AND THEIR OUTREACH TO THE POOR PRE AND POST GLOBAL FINANCIAL CRISIS

(1)

THE RELATIONSHIP BETWEEN

PROFITABILITY OF MICRO FINANCIAL

INSTITUTIONS AND THEIR OUTREACH TO

THE POOR PRE AND POST GLOBAL

FINANCIAL CRISIS

Rick Jager

University of Groningen

Faculty of Economics and Business

Supervisor Drs. B. van Oostveen

June 2016

Abstract

This study investigates the relationship between profitability and outreach to the poor of MFIs worldwide using 10,860 from 2,229 different institutions between 2003 and 2014. Different measurements for profitability and outreach are used. It ascertains if there is a difference in this relation pre and post global financial crisis and if there are any differences between regions. It confirms the existing belief that the financial performance of MFIs hurts their outreach to the poor. Using two way fixed effect regression analyses it also shows that the outreach of MFIs has become larger after the global financial crisis. There are clear indications that there are differences between regions and especially Latin America and the Caribbean show some diverging results.

(2)

2

1. Introduction

One of the ways to reduce poverty worldwide is offering financial services via microfinance. Micro Financial Institutions (MFIs) are there to offer these financial services in underdeveloped regions around the world. Next to the goal of reducing poverty there are also financial interests at stake. Due to commercialization the focus of the MFIs has shifted more to the financial goals. Several studies have already looked into the relationship between profitability and the level of outreach, showing mixed results. By using a large dataset including data from before and after the global crisis this study investigates if there are any differences in this relationship before and after the crisis.

That the lack of financial services is a big problem worldwide is supported by a report of the World Bank’s Global Financial Inclusion Database (global FINDEX, 2014). They report that 2.5 billion adults are excluded from financial services worldwide from which 80 percent live under $2 per day. They state that three-quarters of the world’s poor lack access to a bank account because of poverty, costs, travel distances and the often demanding requirements for opening a bank account. It is estimated that the richest 20 percent of adults in developing countries are more than twice as likely to have a formal account compared to the rest of the population.Microfinance has been an important tool in reducing this poverty worldwide the last 40 years. Weiss & Montgomery (2005) find that microfinance is a good tool to serve the poor and has a positive effect on reducing poverty.

The importance of microfinance is also supported by the fact that the Nobel Prize for peace in 2006 was offered to Muhammad Yunus and his Graheem Bank. In the 1970’s, this economics professor of Bangladesh University, started making small loans to local villagers. He founded the Grameen bank in 1976 and officially authorized it in 1983 as an independent bank to start granting small loans to people in rural areas. This moment can be seen as the start of microfinance and Yunus is recognized as a visionary that spread the movement of microfinance globally.

(3)

3 Figure 1:

Number of borrowers per year worldwide

Source: The state of the Microcredit Campaign report, 2015

MFIs have three different operational objectives: 1) outreach to the poor, 2) to ensure their financial sustainability and 3) to have an impact on poverty reduction (Zeller & Johannsen, 2008). These goals can be contradicting as the first and the last are social goals and the second is a financial goal. Emphasis has shifted more to this financial goal due to the so-called commercialization of microfinance which is a major development in the microfinance sector. With the success of microfinance, commercial parties also became interested in participating due to the potential financial benefits. Opening new NGOs to provide microfinance to the poor is no longer driven purely by social impulses. More and more institutions are starting to engage with the traditional capital markets. This move to commercialization offers an opportunity to provide much more funding to the poor borrowers. A good example of this commercialization is the public stock offering of Banco Compartamos in Mexico. In April 2007, this MFI initiated a public stock offering that led to impressive returns for the original equity investors. The $6 million equity investments that launched the bank in 2000 turned out to be worth $2.2 billion in June 2007. The responses to this event were twofold. Some people argued that it was something positive as it proved that microfinance can be commercially viable and attract private equity independent from social motivations. Conversely people got furious because the high returns were a result of the high interest rates they raised to their borrowers. So, this event encouraged the discussion about what the purpose of MFIs really is.

The main contribution of this paper is that it is the first study to investigate the relationship between financial performance and outreach before and after the global financial crisis. Most of the existing research is done with data prior to the global financial crisis. There are indications (see literature

(4)

4 review) that the microfinance sector has suffered from the global financial crisis. Evidence is found that the sign of the relationship between outreach and profitability for both periods is negative but that the magnitude post financial crisis is much larger. Secondly, this paper contributes by making a clear distinction for the relationship between different regions worldwide. This is done because different economic circumstances can possibly lead to a different relationship among various regions. Dividing the sample into six different regional areas, this study observes no clear significant differences in the relationship between profitability and outreach among the regions. However, some differences in the control variables are noticeable and indicate that some factors have diverse influences between different regions.

The main question this study tries to answer is: “Does the profitability of MFIs have a negative impact on their outreach to the poor and is there a difference in this relationship pre and post global financial crisis?” This research also tries to see if there are differences in this relationship on a regional level worldwide.

The remainder of this paper is outlined as follows. Section two will give a brief overview of the relevant literature and explains the hypothesis development, section three describes the methodology used in this study, section four shows how the data is obtained and some descriptive statistics, section five present the results and finally section six concludes and offers some limitations of this study.

2. Literature review

This section will give an overview of the existing literature and is divided into three parts. First, the literature that deals with the commercialization of microfinance will be discussed, second, the literature covering the relationship between financial performance and outreach of MFIs will be reviewed, and finally, the papers that deal with the influence of the global financial crisis on Microfinance will be examined. From this literature review hypotheses will follow that help to answer the research question “Does the profitability of MFIs have a negative impact on their outreach to the poor and is there a difference in this relationship pre and post global financial crisis?” Commercialization

(5)

5 international commercial banks, have entered the industry in recent years. For example, a big international operating commercial bank like ING Group from the Netherlands entered the Microfinance sector in October 2004 by setting up their ING Microfinance Support group (Mix Market Service Providers). The commercialization influenced the microfinance sector and several studies investigate the effect this has on the level of outreach of MFIs.

Christen (2001) finds in his study that the increase in commercialization does lead to higher levels of profitability of MFIs in South America but he concludes that this increase in profitability has not led to mission drift. Cull, Kunt & Murdoch (2009) support the view of an increase of commercialization. They find that more and more organizations with a for-profit status (contrary to the not for-profit status that most traditional MFIs have) are entering the market of microfinance and that microfinance as a sector will continue to expand and become part of the financial mainstream.

Lensink (2011) provides two different views of the influence of commercialization on microfinance. First of all he argues that commercialization can be good for the outreach to the poor. The increasing focus on profitability is necessary to increase the funding to different MFIs. More funds mean that more people can be reached. A big advantage of commercial entities entering the market is that they can provide credit to the poor without being dependent on donor subsidies. On the other hand he argues that a potential disadvantage of commercialization is that traditional MFIs are confronted with increased competition in the market for micro loans. The danger is that the increase in competition can lead to a reduction in the scope to lending to the core poor. Next to this, it also has the potential of increasing the interest rates strongly, which raises the indebtedness of clients. Lensink’s (2011) conclusion is that if we really want to help the poor, we should probably give up the idea of financial sustainability and accept that there is nothing wrong with subsidizing MFIs.

Assefa, Hermes & Meesters (2013) look specifically into this increasing competition in the microfinance industry and find that since 2000, competition has increased. However, they find little evidence that MFIs face lower outreach when faced with more intense competition.

Financial performance and outreach

Several studies have already investigated the relationship between the financial performance of MFIs and their outreach to the poor by using different variables, different methodologies, and coming to diverse conclusions.

(6)

6 sustainability they argue that the social goals of MFIs get compromised. When MFIs transform into more formalized banking institutions and thereby focus more on financial sustainability this can have negative consequences for the poor because of a lower outreach. Hermes, Lensink & Meesters (2011) research 435 MFIs worldwide with a total of 1318 observations between 1997 and 2007. They investigate the relationship between efficiency and outreach. They measure outreach in two different ways, 1) as the average loan balance per borrower/GNI per capita and 2) as the percentage of female borrowers. Their research finds that efficiency is negatively related to both measurements of outreach. They also state that there is an increasing emphasis on sustainability and efficiency of MFIs and this emphasis may be at the cost of their outreach. Olivares-Polanco (2005) comes to similar conclusions when investigating 28 MFIs between 1999 and 2001 in Latin America. In his research he focuses on different commercialization effects that may affect outreach. For measuring outreach he uses the average outstanding loan and by running simple regression he finds that the return on assets(ROA) of an MFI has a negative effect on the outreach. He also looks at the influence of the maturity of the institution by using the years that an MFI is in operation as a control variable. The conclusion of this study is that the older the institution, the lower the average loan size. This means that the longer an MFI is in operation the lower their outreach is to the poor. Daher & Le Saout (2015) confirm the negative relationship investigating 372 MFIs worldwide between 2005 and 2011. An interesting finding of their study is that the recent global financial crisis has a negative impact on the profitability of MFIs. Another study that confirms the negative relationship between outreach and financial performance is that of Kipesha & Zhang (2013). They investigate 47 MFIs in East Africa between 2008 and 2011. They use an unbalanced panel data set and explore the trade-off between different measures of outreach (average loan balance per GNP per capita, number of active borrowers and percentage of female borrowers) and different measures of financial performance (OSS, cost per borrower, debt to equity ratio, portfolio yield and ROA). They find that ROA has a negative influence on the outreach measurements but they observe a positive relationship between OSS and outreach. Adding to this finding they conclude that it is possible for MFIs to focus on financial sustainability while still reaching out to the poor. However, this focus should be controlled at certain levels above which institutions generate profit at the expense of the outreach to the poor.

The hypothesis that follows is:

H: The size of the profitability of an MFI is negatively related to its outreach

(7)

1999-7 2002. In their paper they try to answer two different questions: “Is there a trade-off between depth of outreach to the poor and the pursuit of profitability?” and “Has Mission drift occurred?” To measure the financial performance of the MFIs they use the Financial Self Sufficiency (FSS) ratio, Operational Self Sufficiency (OSS) ratio and the Return on Assets (ROA). To indicate outreach they use average loan size per borrower/GDP per capita, average loan size per borrower/GDP per capita of the poorest 20% of the population and percentage of female borrowers. To control for other aspects they also add several control variables in their analysis including the legal status of the MFI, the region in which it operates, its maturity, and its size. They did not find a clear relationship between profitability and outreach. MFIs that have a higher outreach are not less profitable on average. This conclusion is supported by Quayes (2012) who analyses 702 MFIs in 83 different countries worldwide in 2006. In his paper he investigates the relationship between financial sustainability and the depth of outreach by using the FSS as the independent variable and the average loan size per borrower as the dependent variable. He finds that the level of FSS has no impact on the depth of outreach of MFIs. However, when controlling for the disclosure level of the MFI the paper shows that firms with a low-disclosure level have a trade-off between outreach and financial sustainability and that for high-disclosure firms financial sustainability has a positive impact on the level of outreach. In his paper Quayes (2012) also makes a distinction between for-profit and not for-profit firms as defined by their legal status. Analyzing these two groups he concludes that not for-profit firms have a better outreach but poorer financial performance compared to for-profit firms. Kar (2013) also investigates the mission drift hypothesis of microfinance by defining it as a trade-off between profitability and depth of outreach. In his study he investigates 409 MFIs during the period 2003-2008 in 71 countries worldwide. Just like Cull, Kunt & Murdoch (2007) he defines outreach as the average loan balance per borrower/GNI per capita and as a second measurement he uses the percentage of female borrowers to specify the depth of outreach. To indicate the financial performance of MFIs Kar (2013) uses three different variables: ROA, FSS and the yield of the loan portfolio. He concludes that there is no significant evidence for a negative relationship between profitability and depth of outreach of MFIs. Although none of these studies find a significant relationship between outreach and profitability they did find several other variables that influence the level of outreach. Size, age and MFIs with a for-profit status seem to have a negative impact on the level of outreach and these factors will be taken into account in this study.

The following hypotheses follow from this literature:

(8)

8 H: An MFI with a for-profit status has a lower outreach than an MFI with a not for-profit status. Conflicting with both results that are described above (a significant negative relationship and no significant relationship between outreach and profitability) are the findings of Nurmakhanova, Kretzschmar & Fedhila (2015). Investigating 450 MFIs worldwide between 2006 and 2008 they conclude that focusing on financial performance does not hurt the depth of outreach and that a win-win proposition is even possible. However, for financial performance they only used OSS as a measurement which could limit the reliability of their conclusions.

The mixed results that were found in previous studies show either a significant negative relationship or a non-significant one. The only study that found a positive relationship (Nurmakhanova, Kretzschmar & Fedhila, 2015) is questionable for their reliability because they used a short time period and only limited indicators for financial performance.

Global financial crisis

The global financial crisis also has had its impact on the microfinance industry. The studies described above are all conducted with data samples from before the global financial crisis. Dokulilova, Janda & Zetek (2009) explored the financial sustainability during the financial crisis. They argue that in past crises MFIs did not really suffer because they were able to survive these macroeconomic crises due to their close links with micro-foundations of the economy. However, they did suffer from the most recent global financial crisis and the consequences were most adverse for MFIs operating in Eastern Europe and Central Asia. Di Bella (2011) confirms these findings and concludes that MFIs financial performance nowadays is correlated to changes in international capital markets. Daher & LeSaout (2015) also confirm these findings and conclude that the recent global financial crisis had a negative impact on the profitability of MFIs.

Lensink (2011) states that, due to the recent commercialization as described earlier, MFIs are much more vulnerable to international macroeconomic developments. By using regression analysis he showed that the profitability of MFIs decreased in the fourth quarter of 2008 and in the first quarter of 2009 which was right after the start of the global financial crisis. Because most of the literature that deals with the influence of commercialization uses data from before 2009, Lensink (2011) states that the considered effects should be studied again with data from after the global financial crisis.

This leads to the next hypothesis:

(9)

9 Last of all this paper takes into account regional differences. No studies yet clearly identify differences in the relationship between outreach and profitability of MFIs between regions, but this paper will investigate the relationship separately for six different regions. Namely, Africa (AF), South Asia & the Pacific (SAP), Eastern Europe & Central Asia (ECA), Latin America & the Caribbean (LAC), Middle East & North Africa (MENA) and South Asia (SA). This is done because different economic circumstances can possibly lead to a different relationship among various regions.

H: The relationship between outreach and profitability of MFIs is different between different regions.

3. Methodology

In order to check the hypothesis a longitudinal dataset is used. To analyze the panel data three different models are set up. The first model uses pooled OLS to analyze the data, the second one includes entity fixed effects and the last model is a two way fixed effect model.

The first model is the pooled OLS model. This model will estimate the relationship between the independent variable profitability and the dependent variable outreach. The pooled OLS model combines the data over time and entity dimension into a single regression and does not take into account the effect of specific entities and time. Several control factors are included. The associated equation reads as:

OUTREACHi,t=α+β₁PROFITABILITY_i,t+β₂AGE_i,t+β₃SIZE_i,t+ β₄PROFIT_i,t + β₅MACROECONOMIC_i,t+ε_i,t (1)

Where i is an index for the MFI, t is a time index, OUTREACH is a measurement that indicates the level of outreach to the poor, PROFITABILITY is a measurement that indicates the financial performance of an MFI, the coefficient AGE measures the maturity, SIZE is a measurement for the size, PROFIT indicates if an MFI has a profit status or not and MACROECONOMIC is a proxy that measures the macroeconomic factors for the region an MFI operates in. ε represents the standard error.

The second model that is used is an entity fixed effect model. This model allows analysis of the relationship between profitability and outreach while excluding the variation that is explained within an entity. So in using this model specific characteristics of individual entities that may bias the results of the regression are controlled for. The associated equation reads as:

OUTREACHi,t=α+β1PROFITABILITYi,t+β2AGEi,t+β3SIZEi,t+ β4MACROECONOMICi,t + ci+εi,t (2)

(10)

10 profit status of an entity is a time invariant variable and is therefore excluded in this model. The corresponding symbols have the same meaning as described above for model 1 and ci represents the

entity fixed effect.

The third model is the two way fixed effect model. This model is used to control for both entity and time specific effects that could possibly bias the results. The corresponding equation reads as:

OUTREACHi,t=α+β1PROFITABILITYi,t+β2AGEi,t+β3SIZEi,t+β4MACROECONOMICi,t+ut+ci+εi,t (3)

The corresponding symbols have the same meaning as described in models 1 and 2. ut is added and

represents the time-fixed effect. This is added into the equation to analyze the influence of time on outreach and will lead to year specific coefficients. As mentioned before the interest of this paper is to investigate the influence of the global financial crisis on the relationship between outreach and profitability of MFIs. In order to do so the year 2008 is been taken as the year the crisis took place and the coefficients of the other years will be compared to this base year.

All three models will first be applied to the whole dataset. After this the dataset is divided into six subsamples based on the region in which the MFIs operate. The six regions that will be analyzed are Africa (AF), East Asia and the Pacific (EAP), Eastern Europe and Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and North Africa (MENA), and South Asia (SA). Models 1 and 2 will then be applied to these six subsamples separately. This way it is possible to see if there are any differences in the hypothesized relationships between different regions.

4. Data

(11)

11 duplicates which were deleted and finally five observations were deleted because these were considered as outliers of the main independent variable ROA (See appendix figure A.1). The second source is the World data Bank. The data from the World data Bank is used to indicate the macroeconomic factors from the six different regions.

Variables

Dependent Variable: the dependent variable that will be investigated in this paper is outreach to the poor. This variable can be measured and defined in different ways according to existing literature. Rosenberg (2009) states that the Average Outstanding Loan Balance (AOLB) is a proper indicator for outreach to the poor. The AOLB measures the gross amount of loans outstanding divided by the number of active clients. To make this indicator more comparable between different countries the AOLB is divided by GNI per capita. Hermes, Lensink & Meesters (2011) suggest two different measurements of outreach. The first indicator they use is the same as defined by Rosenberg (2009) and the second one is the percentage of female borrowers. Kipesha and Zhang (2013) suggest three different measurements of outreach: The AOLB per GNI per capita, the percentage of female borrowers, and the total number of active borrowers. This study will use both the AOLB per GNI per capita and the percentage of female borrowers to measure the outreach to the poor. These indicators are chosen because most of the existing literature uses these variables and they are accepted as good measurements of outreach.

(12)

12 topic of this paper. The Microfinance Financial Reporting Standard (The SEEP Network, 2010) state that ROA is the best way to determine an MFIs financial performance because measurements of financial sustainability (OSS and FSS) become less helpful as measures of financial performance when exceeding the breakeven point of one (respectively, MFIs being operationally self-sufficient and financially self-sufficient). The data provided by Mixmarket.org adjust for the effects of subsidies so the variables are measured as they would be on an unsubsidized basis. Both the ROA and OSS will be used in this study to indicate financial performance.

The following part will give a short definition of the variables used and an overview of them is given in table 1.

OUTREACH is measured in two different ways. 1) The natural logarithm of the average loan balance per borrower over GNI per capita measured in US dollars. 2) The percentage of total borrowers that is female.

PROFITABILITY is also measured in two different ways. 1) The ROA. 2) The OSS. A definition of both these measurements is given in table 1.

AGE in the dataset can get three different values. “New” when the MFI has been in business between 1 to 4 years, “Young” if it has been in business between 5 to 8 years and “Mature” if it has already been in business for more than 8 years. In this model AGE is a dummy variable which takes the value of one when it is mature and zero otherwise.

SIZE can also get three different values: “Small”, “Medium” or “Large”. This size measurement is based on the gross loan portfolio in US dollars and is corrected for the region in which the MFI operates to reflect differences in income levels across regions. SIZE, like AGE, is a dummy variable which takes the value of one if it is large and zero otherwise.

PROFIT is used as a dummy variable and represents the legal status of an MFI. It takes the value of one if it has a for-profit status and zero if it has a not for-profit status.

(13)

13

Source: Mixmarket and World Data Bank

Table 2 gives an overview of the observations per year per region. It is noticeable that most of the observations come from the region of Latin America and the Caribbean and are between 2008 and 2011. The fact that most of the observations come from that region is not surprising as the reporting of MFIs in Latin America and the microfinance sector there are best developed and most present in the world.

Table 3 gives an overview of the distribution of the different categorical variables between the different regions. Looking at the profit status it can be observed that in EAP, ECA and MENA the number of MFIs with a for-profit status are just as high as the number of MFIs with a not for-profit status. This is not the case for the other regions where there are noticeably more for-profit MFIs than institutions with a not for-profit status. This is especially true for the MENA region where the share of for-profit MFIs is around 85%. Taking a look at the size of MFIs it can be observed that this is quite evenly distributed (Small 39%, Medium 26% and Large 35%). However, between regions these

Table 1 : Description variables

Variable Explanation Source

LNO (Natural logarithm of outreach)

Outreach = Average loan balance per borrower/GNI per capita Mixmarket Female (Percentage of

female borrowers)

Number of active borrowers who are women/number of active borrowers

Mixmarket ROA

(Return on assets)

Financial Revenue - (Financial Expense + Impairment Loss + Operating Expense)-Taxes

Mixmarket OSS (Operational

self-sufficiency)

Financial Revenue / (Financial Expense + Impairment Loss + Operating Expense)

Mixmarket Age Dummy variable. 1 if MFI is 8 years or older and 0 otherwise Mixmarket Size Dummy variable. 1 if large and 0 otherwise Mixmarket Profit Dummy variable. 1 if the MFI has a profit status and 0

otherwise

Mixmarket lnGDP The natural logarithm of Gross Domestic Product of the region

the MFI operates in

World Data Bank Inflation The inflation of the region the MFI operates in World Data

Bank GDP growth The growth of Gross Domestic Product compared to the year

before of the region the MFI operates in

(14)

14 Table 2:

Observations per year per region

This table gives an overview of the observations per year per region. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia.

Year AF EAP ECA LAC MENA SA

2003 154 43 83 87 21 80 2004 138 79 149 157 30 127 2005 163 109 165 207 36 183 2006 204 120 200 264 40 174 2007 191 136 210 291 46 153 2008 187 142 260 336 52 157 2009 156 116 205 347 63 195 2010 158 140 195 357 60 214 2011 161 157 161 348 53 220 2012 129 112 134 293 25 173 2013 113 77 111 278 20 149 2014 112 111 103 279 26 135

differences are larger. In the MENA, SA and LAC region most of the MFIs are defined as large whereas in Africa half of the MFIs are defined as small. Most of the MFIs (67%) are less than four years in operation and are labelled “new”. The dominance of new MFIs is the case for all the different regions and is particular strong for the LAC region where 82% of the MFIs are new.

Table 3 :

Observations per region per categorical variable

This table gives an overview of the observations per region per categorical variable. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia.

Variable AF EAP ECA LAC MENA SA Total

Profit status Profit 1,078 582 976 1,989 401 1,134 6,160 Non-profit 705 587 973 1,237 67 769 4,338 Size Large 478 331 763 1,216 207 763 3,758 Medium 460 377 456 859 157 535 2,844 Small 928 634 756 1,169 108 662 4,257 Age New 1,025 878 991 2,639 314 1,324 7,171 Young 330 126 383 180 48 234 1,301 Mature 495 193 598 414 106 388 2,194

(15)

15 female, age, size and profit. This is due to the fact that these observations are not included in the database of MixMarket and follows from the fact that some MFIs do not report these numbers.

Table 4:

Summary statistics of variables

This table gives the mean, standard deviation, minimum and maximum and number of observations per variable that are used in the model. The numbers are rounded to three decimals.

Variable Mean Stdev Min Max Observations

LNO -1.177 1.299 -9.210 6.099 10,860 Female 0.651 0.268 0.000 1.000 9,480 ROA 0.007 0.121 -1.850 0.973 10,860 OSS 1.176 0.679 -1.276 36.627 10,860 Age 0.672 0.469 0.000 1.000 10,666 Size 0.346 0.476 0.000 1.000 10,859 Profit 0.413 0.492 0.000 1.000 10,502 lnGDP 8.022 0.857 6.297 9.144 10,860 Inflation 0.058 0.022 0.021 0.120 10,860 GDP growth 0.053 0.030 -0.046 0.122 10,860

Source: author’s own calculations based on the data of Mixmarket.org and World Data Bank.

5. Results

This section provides the results of the analysis and is divided into five different parts. The first section gives the results of model 1 applied to the complete dataset, the second part of this chapter provides the results of model 2 applied to the complete dataset, thirdly the results of model 3 applied to the complete dataset are shown, after this the results of model 1 applied to the six regional subsamples are provided, and lastly the results of model 2 applied to the six regional subsamples of the dataset are presented.

(16)

16 profitability (ROA and OSS). Remember that the measurement of outreach is defined as the average loan balance per borrower over GNI per capita which means that the higher this variable is the lower the outreach is (the lower the average loan balance per borrower the higher the outreach). A positive coefficient in this model when using LNO as the measurement of outreach would imply that the relationship between the independent variable and level of outreach is negative. This is the case for both profitability measurements ROA (β=0.837) and OSS (β=0.116). The coefficients are highly significant as well. This result holds when the control variables and robust standard errors are added. Age seems to have a positive effect on the level of outreach (β=-0.056), size has a negative effect on

Table 5:

Results model 1 complete dataset

This table shows the main results of the OLS model. The first dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO) (Used in columns 1 to 4) and the second dependent variable is the percentage of female borrowers (Female) (Used in columns 5 to 8). Column 1 (column 5) estimates the relationship between outreach and profitability using the return on assets (ROA) as the main independent variable. Column 2 (column 6) takes the control variables into account and adds robust standard errors to account for heteroscedasticity. Columns 3 and 4 (columns 7 and 8) do the same thing as column 1 (column 5) and column 2 (column 6) respectively only taking the operating self-sufficiency (OSS) as the main independent variable. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

LNO Female (1) (2) (3) (4) (5) (6) (7) (8) ROA 0.837*** (0.102) 0.608*** (0.112) -0.050** (0.022) 0.025 (0.024) OSS 0.116*** (0.018) 0.123*** (0.038) -0.132*** (0.004) -0.009** (0.004) Age -0.056* (0.029) -0.043 (0.029) 0.029*** (0.006) 0.029*** (0.006) Size 0.467*** (0.027) 0.472*** (0.026) -0.040*** (0.006) -0.039*** (0.006) Profit 0.251*** (0.025) 0.256*** (0.025) -0.048*** (0.006) -0.048*** (0.006) lnGDP -0.082*** (0.017) -0.079*** (0.017) -0.065*** (0.004) -0.064*** (0.004) Inflation 0.185 (0.583) 0.127 (0.583) 0.068 (0.135) 0.076 (0.135) GDP growth -2.896*** (0.446) -2.880*** (0.448) 1.174*** (0.101) 1.192*** (0.101) Constant -1.182*** (0.012) -0.611*** (0.156) -1.313*** (0.025) -0.783*** (0.156) 0.652*** (0.003) 1.121*** (0.035) 0.667*** (0.006) 1.120*** (0.035) Number of observations 10,860 10,406 10,860 10,406 9,480 9,135 9,480 9,135 Adjusted R-squared 0.006 0.057 0.004 0.058 0.001 0.104 0.001 0.104 Robust standard errors

yes yes yes yes

Source: author’s own calculations based on the data of Mixmarket.org and World Data Bank. Results are statistically significant for

(17)

17 the outreach (β=0.467) and the profit status (β =0.251) has a negative effect as well when using ROA as the measurement of profitability with robust standard errors. However, when using OSS as the measurement of profitability the coefficient age is not significant anymore although the sign stays the same. The findings are not completely in line with the expectations based on the hypothesis. The influence of ROA, size and profit status are negative on the level of outreach as expected. However, it was also expected that age would have a negative influence on the level of outreach and this model shows the contrary when using ROA to indicate the profitability. When using Female as the indicator of the level of outreach slightly different results are obtained. First of all it is important to emphasize that contrary to the LNO measurement a positive sign indicates a positive relationship between the corresponding variable and the level of outreach (the higher the percentage of female borrowers the higher the level of outreach). Columns 5 and 7 show that for both ROA and OSS there is a significant negative relationship (β=-0.050 and β=-0.132). However when adding the control variables and robust standard errors the sign of ROA becomes positive (β=0.025) but also not significant. The OSS (β=-0.009) stays positive and significant although the magnitude does become much smaller. For the control variables age, size and profit the relationship to the outreach using Female as the indicator is the same as using LNO as the measurement of outreach. Positive for age (β=0.029 and β=0.029) and negative for size (β=-0.040 and β=-0.039) and profit (β=-0.048 and β=-0.048).

(18)

18 Table 7 shows the results of model 3 applied to the complete dataset. By using a two way fixed effect model it is possible to analyze the influence of time on the level of outreach. It takes the variance that is due to time out of the other control variables and makes independent coefficients that explain the change in the level of outreach due to time. First of all the results show that there is no significant relationship between outreach and financial performance when using OSS as the independent variable. For ROA there is a significant sign for both the measurements LNO (β=0.287) and Female (β=0.059). Looking specifically at the coefficients related to time it can be observed that only in columns 1 and 2 they are significant for all years. Columns 3 and 4 only show significant signs for the year 2008. Looking at column 1 (column 2) a trend can clearly be observed that in the years

Table 6:

This table shows the main results of the entity fixed effect model. The first dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO) (Used in columns 1 to 4) and the second dependent variable is the percentage of female borrowers (Female) (Used in columns 5 to 8). Column 1 (column 5) estimates the relationship between outreach and profitability using the return on assets (ROA) as the main independent variable. Column 2 (column 6) takes the control variables into account and adds robust standard errors to account for heteroscedasticity. Columns 3 and 4 (columns 7 and 8) do the same thing as column 1 (column 5) and column 2 (column 6) respectively only taking the operating self-sufficiency (OSS) as the main independent variable. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

LNO Female (1) (2) (3) (4) (5) (6) (7) (8) ROA 0.320*** (0.051) 0.300*** (0.062) 0.055*** (0.014) 0.060*** (0.017) OSS 0.012 (0.008) 0.011 (0.010) 0.004** (0.003) 0.004 (0.004) Age -0.052*** (0.017) -0.052*** (0.017) 0.001 (0.005) 0.001 (0.005) Size 0.121*** (0.020) 0.125*** (0.020) -0.008 (0.005) -0.007 (0.005) lnGDP -0.033* (0.020) -0.029 (0.020) -0.022*** (0.006) -0.021*** (0.006) Inflation -1.069*** (0.207) -1.071*** (0.209) 0.165*** (0.062) 0.162*** (0.062) GDP growth 1.143*** (0.188) 1.177*** (0.189) 0.169*** (0.049) 0.175*** (0.049) Constant -1.179*** (0.0506) -0.918*** (0.158) -1.192*** (0.010) -0.966*** (0.158) 0.651*** (0.001) 0.808*** (0.047) 0.646*** (0.003) 0.796*** (0.047) Number of observations 10,860 10,666 10,860 10,666 9,480 9,352 9,480 9,352 Number of MFIs 2,229 2,123 2,229 2,123 2,105 2,020 2,105 2,020 Adjusted R-squared 0.910 0.909 0.910 0.909 0.866 0.868 0.866 0.867 Robust standard errors yes yes yes yes

(19)

Table 7:

This table shows the main results of the two way fixed effect model. The first dependent variable is the natural

logarithm of the average loan balance per borrower over GNI per capita (LNO) (Used in columns 1 and 2) and the second dependent variable is the percentage of female borrowers (Female) (Used in columns 3 and 4). Column 1 (column 3) estimates the relation between outreach and profitability using return on assets (ROA) as the independent variable including control variables, entity and time fixed effect and using robust standard errors to account for heteroscedasticity .Column 2 (column 4) estimates the same but uses the operating self-sufficiency (OSS) as the independent variable. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

LNO Female (1) (2) (3) (4) ROA 0.287*** (0.063) 0.059*** (0.017) OSS 0.010 (0.010) 0.004 (0.004) Age -0.037** (0.017) -0.037** (0.017) 0.002 (0.005) 0.002 (0.005) Size 0.124*** (0.020) 0.127*** (0.020) -0.007 (0.005) -0.007 (0.005) lnGDP 0.267*** (0.061) 0.277*** (0.061) -0.001 (0.019) 0.001 (0.019) Inflation -0.599 (0.484) -0.570 (0.486) 0.544*** (0.142) 0.543*** (0.141) GDP growth 1.561*** (0.305) 1.582*** (0.306) 0.094 (0.080) 0.098 (0.080) 2004 -0.135*** (0.032) -0.136*** (0.032) 0.003 (0.011) 0.002 (0.011) 2005 -0.213*** (0.035) -0.216*** (0.035) -0.007 (0.012) -0.007 (0.012) 2006 -0.215** (0.039) -0.218*** (0.039) 0.002 (0.013) 0.001 (0.013) 2007 -0.238*** (0.047) -0.242*** (0.047) -0.013 (0.016) -0.014 (0.016) 2008 -0.333*** (0.061) -0.340*** (0.061) -0.039** (0.020) -0.040** (0.020) 2009 -0.267*** (0.053) -0.273*** (0.053) -0.006 (0.018) -0.006 (0.018) 2010 -0.415*** (0.061) -0.422*** (0.061) -0.013 (0.020) -0.014 (0.020) 2011 -0.399*** (0.070) -0.406*** (0.070) -0.019 (0.023) 0.020 (0.023) 2012 -0.370*** (0.069) -0.377*** (0.069) -0.015 (0.023) -0.016 (0.023) 2013 -0.343*** (0.071) -0.349*** (0.071) -0.019 (0.023) -0.020 (0.023) 2014 -0.409*** (0.071) -0.415*** (0.071) -0.023 (0.024) -0.024 (0.024) Constant -3.091*** (0.449) -3.182*** (0.448) 0.634*** (0.139) 0.615*** (0.139) Number of observations 10,666 10,666 9,352 9,352 Number of MFIs 2,123 2,123 2,020 2,020 Adjusted R-squared 0.910 0.910 0.868 0.868

(20)

after 2007 the coefficient gets a higher magnitude. In the year 2004 the coefficient β =-0.135 (β=-0.136), it rises to β =-0.238 (β=-0.242) in 2007 and after the global financial crisis (taking 2008 as the year when the crisis took place) the coefficient rises to β=-0.415 (β=-0.422). In 2010 and in 2014 it has a coefficient of β=-0.409 (β=-0.415). All these coefficients are strongly significant. So, a clear declining trend is visible from 2004 to 2008 with an especially strong decline between 2007 and 2008. In the years after the global financial crisis (except in 2009) it declines even further. The results show that the sign of the coefficient stays the same for the year prior and after the crisis. However, there is a clear indication that the magnitude of this sign becomes much larger in the years after the crisis which indicates that the level of outreach of MFIs is higher in the years after the crisis (keep in mind, the lower the variable LNO the higher the level of outreach). These results do not apply when taking Female as the measurement for outreach.

Regional differences

(21)

21 Table 8

Model 1 per region using LNO and ROA

This table shows the results of the OLS model applied to each region specific. The dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO). The main independent variable is the return on assets (ROA). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

AF SAP ECA LAC MENA SA

ROA 0.503** (0.206) 1.275*** (0.306) -0.367 (0.306) 0.883*** (0.175) -0.336 (0.311) -0.629*** (0.148) Age 0.159*** (0.060) 0.120 (0.084) -0.017 (0.056) 0.538*** (0.061) -0.307*** (0.094) 0.101** (0.045) Size 0.705*** (0.064) 0.351*** (0.080) 0.280*** (0.052) 0.960*** (0.042) 0.367*** (0.083) 0.129*** (0.039) Profit 0.142** (0.57) 1.170*** (0.065) 0.618*** (0.051) -0.224*** (0.045) 1.285*** (0.138) 0.004 (0.040) lnGDP -0.287*** (0.095) -0.067 (0.082) -0.561*** (0.096) -0.454*** (0.072) -0.185 (0.177) -0.371*** (0.066) Inflation -2.205 (1.693) -2.785* (1.479) 0.559 (1.200) -2.001 (1.249) 0.124 (1.349) 2.378* (1.344) GDP growth -0.245 (1.878) 0.852 (2.294) 1.762** (0.817) 0.745 (1.000) -0.107 (2.106) 1.680 (1.550) Constant 1.344* (0.749) -1.727** (0.779) 3.816*** (0.811) 1.954*** (0.646) -0.232 (1.473) 0.360 (0.464) Number of observations 1,782 1,099 1,945 3,219 465 1,896 R-squared 0.093 0.270 0.135 0.205 0.262 0.029

Source: author’s own calculations based on the data of Mixmarket.org and World Data Bank. Results are statistically significant for the *10%, **5% and ***1% levels.

(22)

22 Table 9

Model 1 per region using Female and OSS

This table shows the results of the OLS model applied to each region specific. The dependent variable is the percentage of female borrowers (Female). The main independent variable used is operational self-sufficiency (OSS). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

AF EAP ECA LAC MENA SA

OSS 0.002 (0.010) -0.045 (0.028) 0.003 (0.010) -0.029*** (0.009) -0.008** (0.004) 0.103*** (0.020) Age -0.037*** (0.014) -0.117*** (0.020) -0.010 (0.012) -0.008 (0.012) 0.130*** (0.032) 0.021 (0.013) Size -0.091*** (0.016) 0.068*** (0.020) -0.056*** (0.013) -0.112*** (0.008) -0.084*** (0.026) -0.022* (0.013) Profit -0.061*** (0.014) -0.231*** (0.018) -0.050*** (0.011) 0.021** (0.008) -0.178*** (0.044) -0.033*** (0.013) lnGDP 0.004 (0.024) 0.021 (0.022) -0.046** (0.021) -0.015 (0.015) -0.153*** (0.058) 0.068*** (0.022) Inflation -0.534 (0.405) -0.858* (0.492) 0.125 (0.241) -0.185 (0.239) 1.007** (0.493) -0.152 (0.437) GDP growth -0.730 (0.479) 0.959 (0.633) 0.196 (0.171) 0.323* (0.195) -0.201 (0.689) -0.319 (0.471) Constant 0.709*** (0.189) 0.778*** (0.212) 0.906*** (0.180) 0.822*** (0.135) 1.808*** (0.479) 0.322** (0.161) Number of observations 1,505 840 1,750 2,821 435 1,784 R-squared 0.049 0.216 0.051 0.072 0.117 0.046

Table 10 shows the results of applying model 2 to the six different regions using LNO as the measurement for outreach and ROA as the measurement for profitability. Model 2 allows controlling for the MFI specific effects. The results indicate that when controlling for these entity specific effects only the regions AF and LAC still show a significant influence of ROA on the level of outreach (β=0.605 and β=0.349). The constant of the model is not significant for AF and MENA and the results from these regions can therefore not be interpreted. Age is only significant for LAC and SA (β=0.079 and β=-0.100), implying that in the LAC region the age of an MFI is negatively related to the level of outreach and in the SA region it is positively related to the level of outreach. Looking at size, model 2 shows a strongly positive significant sign for AF, SAP, LAC and SA (β=0.198, β=0.105, β=0.155, β=0.196) and hints that the size of an MFI has the same influence on the level of outreach between different regions.

(23)

23 When accounting for entity fixed effects all regions (but MENA) show a positive sign of the OSS in relation to female. This indicates that for all regions there is a positive relationship between the level of profitability and the level of outreach. The magnitudes are all small and are only significant for EAP (β=0.007), LAC (β=0.009) and SA (β=0.023). The variable age is not significant for all regions but EAP (β=0.040). Size is negatively significant for AF (β=-0.046) and SA (β=-0.018) and positively significant for MENA (β=0.046). However, most results in table 11 are not significant.

Table 10

Model 2 per region using LNO and ROA

This table shows the results of the entity fixed effect model applied to each region. The dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO). The main independent variable is the return on assets (ROA). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately including entity fixed effects. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

ROA 0.605*** (0.129) 0.187 (0.193) 0.130 (0.192) 0.349** (0.150) -0.008 (0.173) 0.085 (0.082) Age -0.069 (0.048) 0.058 (0.046) 0.001 (0.040) 0.079** (0.032) -0.091 (0.070) -0.100*** (0.033) Size 0.198*** (0.057) 0.105** (0.051) -0.015 (0.052) 0.155*** (0.042) -0.051 (0.066) 0.196*** (0.028) lnGDP -0.094 (0.072) 0.145*** (0.048) -0.310*** (0.067) 0.031 (0.032) -0.186* (0.110) -0.152*** (0.048) Inflation -0.872 (0.753) -1.339*** (0.493) 0.602 (0.715) -2.240*** (0.351) 0.463 (0.469) -1.542** (0.597) GDP growth -0.711 (0.889) 2.418** (0.944) 0.958** (0.459) 0.765*** (0.255) 0.068 (1.01) -0.164 (0.644) Constant 0.216 (0.674) -2.898*** (0.432) 2.113*** (0.550) -1.742*** (0.266) -0.035 (0.899) -0.630*** (0.321) Number of observations 1,850 1,197 1,972 3,233 468 1,946 Number of MFIs 499 255 413 505 69 382 Adjusted R-squared 0.885 0.904 0.826 0.935 0.876 0.868

(24)

24

6. Conclusions

By using a large panel data set of 10,860 observations using several regression models, this study investigates the relationship between profitability and outreach of MFIs. This paper takes a specific look at differences between this relationship on a regional level and if the relationship has been changed after the global financial crisis.

The results found in this study confirm the existing theory about the negative impact of profitability on the level of outreach of MFIs and the influence of different control variables on this relationship. However, this paper does provide some additional results that have not been addressed before and also a few issues that are contradicting with existing literature.

Table 11

Model 2 per region using Female and OSS

This table shows the results of the entity fixed effect model applied to each region. The dependent variable is the percentage of female borrowers (Female). The main independent variable used is operational self-sufficiency (OSS). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately including entity fixed effects. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

OSS 0.006 (0.010) 0.007* (0.004) 0.006 (0.006) 0.009** (0.004) -0.003 (0.002) 0.023*** (0.008) Age -0.001 (0.015) 0.040** (0.017) -0.002 (0.010) 0.004 (0.007) 0.027 (0.024) 0.000 (0.010) Size -0.046*** (0.015) 0.027 (0.020) -0.009 (0.011) -0.008 (0.007) 0.046* (0.027) -0.018* (0.009) lnGDP -0.031 (0.022) -0.038** (0.016) -0.055*** (0.014) -0.030*** (0.009) -0.079* (0.042) 0.054*** (0.016) Inflation -0.032 (0.240) -0.103 (0.236) 0.331** (0.150) 0.041 (0.106) 0.155 (0.196) -0.189 (0.182) GDP growth 0.075 (0.268) 0.439 (0.304) 0.063 (0.086) 0.234*** (0.077) -0.245 (0.359) -0.059 (0.201) Constant 0.818*** (0.152) 0.982*** (0.135) 0.927*** (0.115) 0.866*** (0.081) 1.226*** (0.340) 0.493*** (0.109) Number of observations 1,553 927 1,773 2,831 437 1,831 Number of MFIs 465 227 395 482 69 382 Adjusted R-squared 0.772 0.835 0.801 0.86 0.830 0.856

(25)

25 Using different measurements of outreach and profitability, this paper found that the level of ROA has a negative influence on the LNO of MFIs. However, this finding is not confirmed when using the percentage of female borrowers as a level of outreach. This is in contrast with the expectations as both outreach measurements are considered to be valid according to existing literature. Although it is generally accepted that women constitute a majority of the poorest microfinance clients (Armendáriz & Morduch, 2010) it is possible that both outreach measurements do not estimate the same thing. This can be observed by looking at the pair wise correlation matrix (appendix table A.5) where the correlation between LNO and female is only -0.53. This is fairly strongly correlated but far from perfect (remember that the indicators measure outreach in the opposite way, so a negative sign in the correlation means that they are positively correlated). This finding indicates that more profitable MFIs grant higher loans on average but they also reach more women as borrowers. A possible explanation is that female borrowers are more profitable for MFIs (Armendáriz & Morduch, 2010). However, existing theory does not support this view and further research should look into this.

The findings presented in this paper also indicate that the level of outreach of MFIs has changed after the global financial crisis. It can be observed that the level of outreach has become larger in the years after 2008 compared to the years before. A certain reason for this increase in outreach cannot be given but it could possibly be related to the fact that in the last fifteen years the microfinance sector has become more and more interconnected with the international capital markets. This has to do with the increased commercialization. Because these commercial institutions have a for-profit status (which have a lower outreach as found by the research of this paper), it could be that these commercial parties have become less involved in microfinance after the global financial crisis. So if that is the case, relatively more nonprofit organizations contributed to granting loans to the poor which could be an explanation for the after crisis increase in outreach. This is just a possible explanation and further research should focus on the reasons of this growth in outreach.

(26)

26 literature. Strong conclusions about the influence of the maturity at the regional level cannot be made as the different models give different output. The most interesting is that when controlling for entity fixed effects maturity does have a negative effect on the level of outreach in the LAC region. This could be because this region has relatively high numbers of new MFIs (shorter than 4 years) in business (82%) which makes the dummy variable more appropriate for this region.

Size and having a for-profit status both have a negative influence on the level of outreach as expected. The influence of size does not differ between regions. Having a for-profit status has a negative influence on outreach in all regions except the LAC region which shows a positive influence. The positive influence of the for-profit status on the level of outreach in LAC is surprising but a reason could be that the microfinance sector is very well developed and large in this region. How this explains the opposite effect remains an empirical question that should be explored in future research.

This study also has some limitations. First of all the data of this study had to deal with heteroscedasticity which means that the standard errors of the model used do not have a constant variance. By making use of robust standard errors this problem was partially solved. However, the reason for this heteroscedasticity could have to do with the fact that there is an unobserved confounding variable. Future research should try to indicate this potentially existing confounding variable and use instrument variables to analyze the relationship again.

(27)

27

References

Armendariz, B. and Morduch, J., 2005. The economics of microfinance. The MIT press, Cambridge MA.

Assefa, E., Hermes, N., & Meesters, A., 2013. Competition and the Performance of Microfinance Institutions. Applied Financial Economics, 23(9), 767-782

Christen, R., 2001. Commercialization and Mission Drift: The Transformation of Microfinance in Latin America. Consultative Group to Assist the Poor (CGAP), Washington DC.

Cull, R., Demirguς-Kunt, A. & Morduch, J. (2007). Financial Performance and Outreach: A Global Analysis of Leading Microbanks. The Economic Journal, 117, F107-F133

Cull, R., Demirguς-Kunt, A. & Morduch, J. (2009). Microfinance Meets the Market. The Journal of Economic Perspectives, 23(1),167-192

Daher, L. & Le Saout, E., 2015. The Determinants of the Financial Performance of Microfinance Institutions: Impact of the Global Financial Crisis. Strategic Change: Briefings in

Entrepreneurial Finance 24, 131-148

Di Bella, G., 2011. The Impact of the Global Financial Crisis on Microfinance and Policy Implications. IMF Working Paper 11/175, Washington DC

Dokulilova, L., Janda, K. & Zetek, P., 2009. Sustainability of Microfinance Institutions in Financial Crisis. European Financial and Accounting Journal, 4(2), 7-33

Global FINDEX database, 2014. The World Bank

Hermes, N. & Lensink, R. (2011). Microfinance: Its Impact, Outreach, and Sustainability. World Development 39 (6), 875-881

Hermes, N., Lensink, R. & Meesters, A. (2011). Outreach and Efficiency of Microfinance Institutions. World Development 39 (6), 938-948

Kar, A. K., 2013. Mission Drift in Microfinance: are the concerns really worrying? Recent Cross-country results. International Review of Applied Economics 27 (1), 44-60

Kipesha, E.F. & Zhang, X. (2013) Sustainability, Profitability and Outreach Tradeoffs: Evidence from Microfinance Institutions in East Africa. European Journal of Business and Management, 5 (8), 136-148

Lensink, R. (2011). Recent Developments of Microfinance and the Impact of the Financial crisis. Ethical Perspectives 18 (4), 569-590

(28)

28 Mix Market Service Providers. Retrieved from http://www.mixmarket.org/service-providers/ing

(09-05-2016)

Nurmakhanova, M., Kretzschmar, G. & Fedhila, H. (2015). Trade-off between Financial Sustainability and Outreach of Microfinance Institutions. Eurasian Economic Review, 5, 231-250

Olivares-Polanco, F. (2005) Commercializing Microfinance and Deepening Outreach? Journal of Microfinance, 7 (2), 47-69

Quayes, S. (2012). Depth of Outreach and Financial Sustainability of Microfinance Institutions. Applied Economics, 44, 3421-3433

Rosenberg, R., (2009). Measuring results of microfinance institutions: Minimum Indicators that Donors and Investors Should track. Washington DC: The World Bank.

Weiss, J. & Montgomery, H., 2005. Great Expectations: Microfinance and Poverty Reduction in Asia and Latin America. Oxford Development studies, 33(3&4), 391-416

Zeller, M. & Johannsen, J. (2008). Is There a Difference in Poverty Outreach by Type of Microfinance Institution? Country Studies from Asia and Latin America. Savings and Development, 32 (3), 227-269

(29)

29

Appendix

(30)

30 Table A.1

Model 1 per region using LNO and OSS

This table shows the results of the OLS model applied to each region specific. The dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO). The main independent variable is the operational self-sufficiency (OSS). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

OSS 0.000 (0.028) 0.294*** (0.087) 0.049 (0.052) 0.250*** (0.063) 0.048** (0.024) -0.029 (0.035) Age 0.184*** (0.060) 0.160* (0.082) -0.008 (0.056) 0.559*** (0.060) -0.313*** (0.093) 0.073 (0.045) Size 0.727*** (0.063) 0.359*** (0.078) 0.292*** (0.053) 0.962*** (0.042) 0.339*** (0.081) 0.117*** (0.039) Profit 0.153*** (0.057) 1.159*** (0.065) 0.615*** (0.051) -0.215*** (0.045) 1.312*** (0.139) -0.001 (0.041) lnGDP -0.286*** (0.096) -0.056 (0.081) -0.565*** (0.096) -0.447*** (0.072) -0.191 (0.177) -0.378*** (0.066) Inflation -2.182 (1.695) -2.777** (1.430) 0.717 (1.199) -2.074* (1.247) -0.297 (1.324) 2.392* (1.351) GDP growth -0.082 (1.885) 0.739 (2.305) 1.600** (0.819) 0.699 (0.999) -0.247 (2.998) 1.690 (1.559) Constant 1.285* (0.752) -2.174*** (0.778) 3.767*** (0.813) 1.601** (0.662) -0.220 (1.476) 0.473 (0.470) Number of observations 1,782 1,099 1,945 3,219 465 1,896 R-squared 0.090 0.289 0.134 0.205 0.027 0.020

(31)

31 Table A.2

Model 2 per region using LNO and OSS

This table shows the results of the entity fixed effect model applied to each region. The dependent variable is the natural logarithm of the average loan balance per borrower over GNI per capita (LNO). The main independent variable is the operational self-sufficiency (OSS). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately including entity fixed effects. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

OSS 0.051 (0.046) -0.001 (0.014) 0.046* (0.027) 0.030 (0.023) -0.005 (0.004) -0.021 (0.031) Age -0.069 (0.048) 0.062 (0.046) 0.002 (0.040) 0.079** (0.032) -0.090 (0.071) -0.100*** (0.033) Size 0.214*** (0.058) 0.104** (0.050) -0.013 (0.052) 0.155*** (0.042) -0.049 (0.066) 0.201*** (0.028) lnGDP -0.068 (0.072) 0.147*** (0.048) -0.309*** (0.067) 0.031 (0.032) -0.187* (0.110) -0.149*** (0.048) Inflation -0.992 (0.762) -1.366*** (0.496) 0.599 (0.714) -2.220*** (0.354) 0.516 (0.476) -1.552*** (0.597) GDP growth -0.718 (0.896) 2.401** (0.948) 0.929** (0.468) 0.787*** (0.256) 0.075 (1.01) -0.160 (0.649) Constant -0.033 (0.510) -2.914*** (0.429) 2.045*** (0.544) -1.776*** (0.270) -0.021 (0.902) -0.627** (0.314) Number of observations 1,850 1,197 1,972 3,233 468 1,946 Number of MFIs 499 255 413 505 69 382 Adjusted R-squared 0.883 0.904 0.826 0.935 0.876 0.868

(32)

32 Table A.3

Model 1 per region using Female and ROA

This table shows the results of the OLS model applied to each region specific. The dependent variable is the percentage of female borrowers (Female). The main independent variable used is return on assets (ROA) This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

ROA -0.109** (0.043) -0.170*** (0.062) 0.193*** (0.068) -0.012 (0.041) 0.143 (0.122) 0.389*** (0.066) Age -0.031** (0.014) -0.110*** (0.020) -0.008 (0.012) -0.011 (0.012) 0.129*** (0.032) 0.013 (0.013) Size -0.086*** (0.016) 0.068*** (0.021) -0.054*** (0.013) -0.113*** (0.008) -0.094*** (0.026) -0.021* (0.013) Profit -0.059*** (0.014) -0.234*** (0.018) -0.517*** (0.011) 0.021** (0.008) -0.170*** (0.043) -0.037*** (0.013) lnGDP 0.004 (0.024) 0.024 (0.022) -0.047** (0.021) -0.013 (0.015) -0.153*** (0.058) 0.062*** (0.022) Inflation -0.519 (0.404) -0.800 (0.493) 0.147 (0.242) -0.187 (0.240) 0.932* (0.488) -0.107 (0.437) GDP growth -0.675 (0.478) 1.010 (0.635) 0.157 (0.170) 0.300 (0.195) -0.230 (0.069) -0.229 (0.470) Constant 0.697*** (0.189) 0.690*** (0.216) 0.906*** (0.179) 0.781*** (0.134) 1.802*** (0.480) 0.480*** (0.158) Number of observations 1,505 840 1,750 2,821 435 1,784 R-squared 0.052 0.21 0.058 0.068 0.117 0.054

(33)

33 Table A.4

Model 2 per region using Female and ROA

This table shows the results of the entity fixed effect model applied to each region. The dependent variable is the percentage of female borrowers (Female). The main independent variable used is the return on assets (ROA). This table shows the estimations of the relationship between outreach and profitability including control variables and robust standard errors for each region separately including entity fixed effects. AF=Africa, EAP=East Asia and the Pacific, ECA=Eastern Europe and Central Asia, LAC= Latin America and the Caribbean, MENA=Middle East and North Africa, SA=South Asia. The standard deviation is shown between brackets under each coefficient. The numbers are rounded to three decimals.

ROA 0.030 (0.044) 0.002 (0.031) 0.036 (0.034) 0.045 (0.033) 0.196*** (0.065) 0.085** (0.033) Age -0.001 (0.015) 0.040** (0.017) -0.002 (0.010) 0.004 (0.007) 0.031 (0.024) 0.000 (0.010) Size -0.046*** (0.015) 0.028 (0.020) -0.009 (0.011) -0.008 (0.007) 0.036 (0.026) -0.018** (0.009) lnGDP -0.032 (0.022) -0.039** (0.016) -0.055*** (0.014) -0.030*** (0.009) -0.079* (0.041) 0.052*** (0.016) Inflation -0.022 (0.239) -0.112 (0.236) 0.330** (0.150) 0.040 (0.106) 0.141 (0.192) -0.183 (0.189) GDP growth 0.078 (0.268) 0.432 (0.304) 0.065 (0.086) 0.235*** (0.077) -0.227 (0.356) -0.041 (0.210) Constant 0.834*** (0.155) 0.994*** (0.136) 0.936*** (0.115) 0.880*** (0.080) 1.219*** (0.333) 0.528*** (0.107) Number of observations 1,553 927 1,773 2,831 437 1,831 Number of MFIs 465 227 395 482 69 382 Adjusted R-squared 0.772 0.845 0.801 0.860 0.834 0.856

Table A.5

Pair wise Correlation matrix