Capital and Performance of Microfinance Institutions in Eastern Europe and Central Asia

Transcription

1 Capital and Performance of Microfinance Institutions in Eastern Europe and Central Asia Knar Khachatryan, Assistant Professor American University of Armenia, College of Business and Economics Affiliated researcher at the KTO research group, SKEMA Business School, France 40, avenue Baghramyan 0019 Yerevan, Armenia Tel: (+374) Valentina Hartarska, 1 Alumni Professor Department of Ag Economics and Rural Sociology; Department of Finance, Auburn University, USA hartavm@auburn.edu Aleksandr Grigoryan, Associate Professor American University of Armenia, College of Business and Economics Affiliate Fellow at CERGE-EI, Prague, Czech Republic aleksandr@aua.am 1 Corresponding author 1

2 Capital and Performance of Microfinance Institutions in Eastern Europe and Central Asia Abstract Recent trends in microfinance, such as commercialization and deposit mobilization, bring forward the importance of investigating the link between sources of funds and Microfinance Institutions (MFIs) performance. This paper estimates the joint impact of seven categories of capital on three dimensions of performance with seemingly unrelated regressions (SUR) method. We use panel data from MFIs operating in Eastern Europe and Central Asia during the period The results suggest that performance is influenced by the interest of the stakeholders behind the capital. Grants are associated with better depth of outreach. Concessional loans are useful in improving outreach without affecting financial results. Soft loans from social investors are related to lower ROA but improvement in outreach to poorer clientele. We find less clear evidence about the influence of savings on financial performance, but interpret the results to mean that savings should be encouraged to serve the needs of the poor as well as to lower the cost of capital. JEL classification: G21; O16 Keywords: Microfinance; capital structure; SUR; performance; efficiency; outreach 2

3 1. Introduction Microfinance emerged as the provision of financial services to clients outside the mainstream financial system. It has become visible in the past few decades and it is praised for its potential to be a profitable instrument for economic development. Microfinance Institutions (MFIs) provide small-scale financial services (loans and deposits) to marginalized clients and have the multiple objectives to be profitable and to alleviate poverty. MFI are either nonprofit or for-profit organizations. Their capital comes from stakeholders such as quasi-owners (large institutional donors) or charities, creditors, private investors, and more recently depositors. MFIs have widened their product range and in addition to microloans ($50 to several hundred or thousand $) now offer insurance, payment facilities as well as to collect deposits, which add to the diversity of their sources of capital. During the study period, over 10,000 microfinance programs operated worldwide reaching well over 100 million clients (Cull, Demirgüç-Kunt and Morduch, 2009) of which in Eastern Europe and Central Asia there were about 8 million borrowers and 10 million savers (CGAP, 2011). i, ii MFIs have financial and social performance goals. The financial goal, also called, sustainability, refers to the institutions financial viability and capacity to self-sustain its operations and or earn profits. The social performance goal, known as outreach, has two dimensions depth, measured by the average size of loan amount, and breadth, measured by the number of clients reached. The ability of MFIs to attract and use capital to maintain sustainable operations without eroding the focus on outreach is critical to the future growth and success of the microfinance industry. This paper studies how the capital structure of MFIs relates to their ability to meet both outreach and sustainability goals. The microfinance sector is considerably heterogeneous in terms of ownership structure (NGOs, NBFIs, credit unions, microfinance banks), institutional size and targeted clientele. The capital structure of MFIs is unique because part of the external financing is subsidized, for instance by donors, charities, socially responsible investors (Armendariz and Morduch, 2010). Recent trends add to this complexity. Through commercialization, nongovernmental organization (NGO)-MFIs transform into (regulated) institutions, moving away from donor-dependent, subsidized capital and attracting private investors, thus getting better access to external capital (Christen and Drake, 3

4 2002; Tchuigoua, 2015). MFIs transformation into deposit collecting institutions allow them to provide needed service to more poor clients as well as to lower the costs of capital (Hartarska et al, 2011, Delgado et al 2014, Malikov and Hartarska, 2017). Consequently, MFIs have numerous sources of capital that include funds from institutional investments (e.g. Microfinance Investment Funds), development agencies, individuals, foundations, NGOs, banks, international organizations, states as well as the newest group of depositors. These groups are likely to have differential impact on the various dimensions of MFI performance. Therefore, studying how observed capital structure is likely to affect the ability of MFIs to achieve their double bottom-line of outreach and sustainability is timely and important. Few previous studies estimate the impact of capital structure on MFIs social and financial performance. Previous work links capital inputs to MFIs performance through production or cost function analysis suggesting significant economies of scale and scope and thus underscores the benefits of access to capital (Cull et al., 2007; Caudill et al., 2009 and Hartarska et al., 2013). There is work evaluating how capital structure itself (proportion of external debt and of donations) relates to various firm characteristics. Tchuigoua (2015) finds that firm size and for-profit status relate to higher proportion of debt relative to donations. Caudill et al. (2009) find that MFIs that became more efficient in time relied less on subsidy and more on deposits. More profitable MFIs, and MFIs with better outreach we found to attract more international commercial debt (Mersland and Urgeghe, 2013). To our knowledge, only two studies evaluate directly how the capital structure (types of capital) affect MFIs performance. The closest to our work is Bogan (2011) who estimates the effects of capital structure on financial performance (operational and financial self-sufficiency) for a worldwide sample of MFIs. Her focus is on the impact of grants, which she argues may be endogenous, and thus instruments grant availability with the change in country GDP growth. However, her instrumental variable (IV) results are not qualitative different from the direct link, which weakens the case for endogenous selection, and suggest a direct link between capital structure on performance. Kyereboah-Coleman (2007) links performance to the capital structure of 52 MFIs in Ghana for the period, and finds that highly leveraged MFIs reach more clientele and enjoy scale economies. 4

5 In all previous work, however, the effect of capital structure on outreach is estimated independently from the effect on self-sufficiency (Tchuigoua, 2015; Hartarska et al., 2013; Caudill et al., 2009; Bogan, 2011; Cull et al., 2007; Kyereboah-Coleman, 2007). At the same time, the literature provides evidence for a trade-off between the outreach and the sustainability dimensions of MFIs performance, suggesting that financial success may come at the expense of serving fewer and less poor clients mission drift. Several studies confirm the existence of the mission drift (Cull et al., 2007 & 2009; Augsburg and Fouillet, 2010; Nawaz, 2010, Armendariz and Szafarz, 2011; Hermes, Lensink and Meesters, 2011, Hartarska et al., 2013), while some suggest that financial sustainability and social outreach complement and reinforce each other (Gonzalez and Rosenberg, 2006; Schicks, 2007). Thus, we addresses the concern by evaluating the simultaneous effect of capital structure on sustainability and two outreach dimensions of MFIs in Central and Eastern Europe and Central Asia (ECA) during the period Our focus in on the ECA region for two reasons. First, we were able to find detailed data on various sources of external capital only for this region and period. For example, while previous studies only use donation, in addition to donation, we have data on two other types of subsidy non-market interest rates loans and soft loans by social investors. Next, we believe that the diversity of age and size of MFIs in the sector and our sample provides opportunity to evaluate the broader impact of various sources of funds on different types of MFIs. This relates to the different needs of an MFIs along their lifecycle less reliance on donor grants and soft loans and more on equity financing especially for mature and regulated MFIs seeking to improve outreach (Fehr and Hishigsuren, 2006; Farrington and Abrams, 2002). The diversity of the level of economic development of the countries in our dataset not an issue because Booth et al. (2001) demonstrate that the capital structure choices are affected by the same variables independent of the level of country economic development. In particular, we estimate the marginal impact of several sources of funds - subsidized loans, bank loans and loans by social investors, as well as deposits, grants and own funding on three dimensions of performance by employing a panel seemingly unrelated regression (SUR). We use new data from MFIs operating in 24 countries of the ECA region obtained from a grass-root network Microfinance Centre for Central & Eastern Europe and the New Independent States and covering the five-year period between 2005 and

6 The main findings are as follows: MFIs performance is influenced by the interest of the stakeholders behind the capital. Grants are associated with better depth of outreach and better financial performance. Loans at below market interest rate (concessional loans) and soft loans by microfinance investment funds are associated with better outreach to poorer clients without affecting financial results. Relative to the use of own equity, current level of deposits do not affect performance but previous year savings are associated with higher ROA. We interpret this to mean that savings should be encouraged to serve the needs of the poor and possibly as a way to lower the cost of capital. The rest of the paper is organized as follows. Section two develops the empirical framework, section three describes the empirical model; section four describes the data, section five discusses the results, and the last section concludes. 2. Conceptual Framework Microfinance institutions are similar to banks but also to NGOs. The literature linking banks (financial or risk) performance to capital structure is different from that for firms because lending institutions differ significantly from corporate firms in their revenue generation mechanism, leverage ratio as well as the fact that they are regulated. Further, their specific risk management objectives also influence the capital structure and, in turn, the financial results of lending institutions (Cebenoyan and Strahan, 2004). At the same time, little can be learned from the NGO literature because NGOs are not lending institutions. Therefore, to study the joint impact of capital structure on MFI performance we develop several hypothesis based on the finding of the microfinance literature which describe various aspects of MFI capital structure and performance. The earlier microfinance literature focused on the role of subsidies in observing that MFIs used various types of implicit or explicit subsidies (typically grants and in-kind payments for personnel and physical infrastructure). There was a longstanding focus on subsidy dependence, defined as the inverse of self-sustainability, which is achieved when the return on equity, net of any subsidy received, equals or exceeds the opportunity cost of the equity funds (Yaron, 1992). MFIs were working toward financial sustainability by ensuring that clients repay their loans on 6

7 time, generate enough interest revenue as well as by controlling costs to guarantee efficient use of resources (Crombrugghe, Tenikue and Sureda, 2008). iii There is somewhat mixed evidence one the impact of subsidies and donor funds on MFI performance because different aspects of performance are evaluated, with datasets capture from various regions and for various time period during which MFIs have moved away from relying on direct subsidies into relying on investment funds. The literature follows these trends. Hudon and Traca (2006) find that subsidy intensity is associated with a lower financial sustainability because MFIs receiving more subsidies tend to focus on the poorest and serving this segment of the population is more expensive. Hartarska and Mersland (2012) found that donors service on the MFIs board presumably to defend their social investment was not associated with more outreach efficient institutions. Mori and Mersland (2014) find that donors representation is associated with better results. It is possible that these differences come from a possible trade-of between various dimensions of MFI performance. Hartarska (2005) shows that donors in general prefer better outreach to better sustainability. D'Espallier, Hudon, and Szafarz (2013) find that lack of subsidy worsened social performance and MFIs in different regions deal differently with the lack of subsidies. For example, unsubsidized MFIs in ECA chose to target less poor clients. In a study similar to objectives of the present study Bogan (2011) utilizes panel data on MFIs in Africa, East Asia, Eastern Europe, Latin America, the Middle East and South Asia for the years 2003 and 2006 and finds that increased use of grants, rather than own capital by large MFIs decreases operational self-sufficiency in larger firms. The hypothesis that we test is that both outreach and sustainability are affected by the MFIs use of donor funds or grants. The literature also suggest tradeoff between subsidies (grants) use and the use of savings. Caudill et al 2009 and 2013 find that more efficient MFIs rely less on grants and are more likely to offer savings. Cozarenco, Hudon, and Szafarz (2016) also found tradeoff between savings and donation in that credit only MFIs received more subsidies than savings and loan MFIs and suggested that subsidies may crowd out micro savings. Evaluating the impact on outreach, Hartarska and Nadolnyak (2007) found that better breadth of outreach (measured by the number of borrowers) is associated with higher levels of deposits. Thus, another hypothesis that we test is whether the proportion of savings affects outreach and sustainability. 7

8 The microfinance literature has less information on how capital raised through commercial loans affect performance. Hartarska and Mersland (2012) found that higher proportion of creditors representing their organizations on the MFI boards is associated with more efficient MFIs, while Mori and Mersland (2014) found that creditors on the board are associated with better outreach. Evaluating the link between capita structure and performance of a group of Ghanian MFIs for 10 year period, Kyereboah-Coleman (2007) finds that highly leveraged MFIs reach more clientele and enjoy scale economies. Thus, we test the hypothesis that higher leverage affects sustainability and outreach of MFIs. To evaluate how MFIs use of various sources of capital affects their ability to serve the poor and cover their costs we develop an empirical specification and make several assumptions. We assume that, at least in short run, capital structure is exogenous, and focus on the fact that MFIs use available funds to achieve their objectives by offering a choice of products and services designed to serve the target clientele using funds from various sources. Since some of the capital may come with special preferences from the lender (investor), owner, etc., it is likely that the use of that source may come at the expense of one dimension of performance, as variety of microfinance studies suggest. The use of own capital may give MFIs freedom to maximize both dimensions of performance, while use of borrowed capital and quasi-ownership investments may direct the performance toward different aspects. Empirical Specification Unlike previous work evaluating the impact of capital structure on performance with a single or IV regression, we use the seemingly unrelated regressions method along the line of work on efficiency in MFI and capital structure (Hartarska et al., 2012; Hartarska et al., 2013, Tchuigoua, 2015). SUR allows estimation of the simultaneous impact of capital structure on several dependent variables measuring multiple aspects of MFI performance. Previous work specifies one performance measure as a function of the same or similar independent variables (for instance, Bogan, 2011). These studies assume that the three dimensions of performance measures and the regressions errors are 8

9 uncorrelated, so we can infer impact on the multiple objectives of the MFIs (e.g. depth and breadth of outreach as well as financial sustainability) by examining these independent regression results. This work assumes, that as MFIs strive to reach many poor clients, improvements in breadth of outreach (number of clients served) is unrelated to the depth of outreach (the poverty level of clients) and to MFI s financial sustainability ability to cover costs. These are strong assumptions and because serving more and poorer borrowers is costly and there is evidence on tradeoff between outreach and sustainability (Cull et al., 2009; Hermes, Lensink and Meesters, 2011). Thus we use SUR method to study the simultaneous impact of capital structure on the three dimensions of performance. Within the basic linear SUR model, y it is the dependent variable, x it = (1, x it,1, x it,2,, x it,ki 1 ), is a K i - vector of explanatory variables for observational unit of i and ε it is an unobservable error term, where the double index it denotes the t th observation of the i th equation in the system. 2 A SUR model is a system of linear regression equations:.. y 1t = 1 x 1t + ε 1t y Nt = N x Nt + ε Nt where i = 1,, N and t = 1,, T. If we denote L = K K N and stack each observation t, we obtain Y t = [y 1t,, y Nt ], X t = diag(x 1t, x 2t,, x Nt ), a block-diagonal matrix with x 1t,, x Nt on its diagonal, E t = [ε 1t,, ε Nt ], = [ 1,, N ]. Then, (1) Y t = X t b + E t. The joint SUR estimator is a generalized best linear unbiased estimators and with a normality assumption for the error terms, maximum likelihood and diffuse prior Bayesian estimators (e.g., Geweke, 2003; Greene, 2003; Judge et al., 1985; Meng and Rubin, 1996). In equation (1), Y is the profitability and outreach indicator for the i th MFIs, X is a matrix of exogenous MFI-level and country-level control variables, and E i is the error term. 9

10 The dependent variables capture all aspects of MFI performance - financial sustainability (how profitable is the MFI) and outreach itself with two dimensions depth of outreach or how poor the clients are relative to the general population, and the number of poor clients (breadth). Specifically, we estimate: (2) Financial performance i = a 0 + a j X j + a k MFI k + a l M l + i j k l (3) Breadth of outreach i = 0 + j X j + k MFI k + l M l + i j k l (4) Depth of outreach i = γ 0 + γ j X j + γ k MFI k + γ l M l + υ i j k l where X represents capital structure variables, MFI represents firm characteristic variables, and M represents country-level macroeconomic indicators. A detailed description of all of the variables used can be found in Appendix A. Financial performance is measured by the return-on-assets ratio (ROA) 3. We account for the breadth of outreach, or how many clients (borrowers) the MFIs reach by the natural logarithm of the total number of active borrowers (lnab). 5 We account for the poverty level of clients by using a measure of the depth of outreach. It shows whether a MFI addresses the needs of the poorest or targets better-off clients (see Quayes, 2012). Since MFIs are expected to lend to poor borrowers, their ability to reach more poor people is measured by depth of outreach. It is the use the ratio of the total average loan balance per borrower to the GNI per capita (abb). Adjusting the average loan size by GNI per capita normalizes the variable for different income levels found in different countries, thereby controlling for cross-country differences. The vector of capital structure variables in X is of most interest in our analysis. We measure capital structure by the percentage of capital (scaled by total assets) coming from each specific source of funds, which are represented as percentage of total assets in our study. We have categories of equity, grants, deposits, retained earnings, and debt where loans are further disaggregated into concessional loans, bank loans, and other commercial funding ( that is both private and institutional social investor funding in the region such as BlueOrchard, Oikocredit, IFC etc.). 6 10

11 The vector of MFI variables control related to MFI specific internal characteristics, such as organizational types, age, gender focus and risk characteristics. Controlling for age, for example, captures the fact that older, more experienced MFIs are more efficient (e.g., see Caudill et al., 2009). Gonzalez-Vega et al. (1996) point out several possible benefits of the passage of time on microfinance performance increase: improved lending technology, accumulated information on clientele, acquired reputation and connections with international networks, which will ease access to capital funding. We use New, Young and Mature dummies age dummies and expect that age is positively linked to MFI profitability, and use Mature as the base category. Total assets and gross loan portfolio, both adjusted for inflation, are used as measures of MFI size and focus on lending respectively. The size effect may be an indicator of larger MFIs being more cost-effective. The empirical evidence shows that the larger size leads to a possible cost savings due to the advantages afforded by potential economies of scale, as well as potential scope economies between deposits and loans (Hartarska, Shen and Mersland, 2013). We also control for MFIs focus on gender by including the percentage of female borrowers since lending to women is associated with lending to poorer borrowers. For example, women may be riskier borrowers because of their limited repayment capacity (Hermes et al., 2011; D Espallier, et al., 2011). On the other hand, since women living in developing regions often face restricted opportunities for accessing financial services they will be more inclined to exhibit higher repayment rates in order to continue to be further financed (Hartarska, Nadolnyak and Shen, 2012; Hartarska, Nadolnyak and Mersland, 2014, Van Tassel, 2004). We control for asset quality and risk taking with the standard non-performing loan ratio of loans overdue more than 30 days. Lower asset quality (e.g. higher nonperforming loan ratio) requires more resources to manage the higher risk (Hartarska, Nadolnyak and Shen, 2012) and makes outreach and sustainability harder to achieve. The last group of independent variables represented by the vector of M variables includes country-level macroeconomic indicators. Existing empirical evidence shows that external factors related to a country s macroeconomic environment, level of financial development, population density, etc. affect significantly the MFIs efficiency, and need to be controlled for. For instance, lending to rural borrowers, which in the ECA region are 11

12 borrowers without permanent employment and regular income or liquid assets, might be associated with higher risk to MFIs (see Sheremenko, Escalante, and Florkowski, 2012). This aspect is controlled for by including the percentage of rural population to total population. Similarly, we include the level of unemployment in the country because an increase in the unemployment level could lead to a further increase of the risk associated with the loan default in a country where the MFI is located. A measure of the agricultural value added as percentage of GDP is also added to control for the fact that borrowers engaged in agricultural production may be more reliable since they have fewer alternative sources of funds and have history of employment, income, and marketable asset ownership. GDP per capita and GDP growth are other important indicators of a country s macroeconomic context, which could affect borrowers purchasing power and could be associated with their risk of default. Finally, the private credit bureaus coverage is important in terms of credit evaluation and portfolio management by MFIs. The existence of credit registers can reduce the extent of asymmetric information by making a borrower s credit history available to MFIs. The higher coverage can be associated with decrease in lending to high risk individuals, with poor repayment histories, defaults or bankruptcies. In order to test if the errors across equations in the SUR model are contemporaneously correlated, we run the Breusch-Pagan specification test of independent errors typically used for SUR models. iv The null hypothesis is no contemporaneous correlation of the error term. Thus, a rejection of the null will indicate that SUR is the more appropriate method to study the impact of capital structure on performance in MFIs. For a robustness check we also offer alternative specification where we lag the independent variables one period to avoid contemporaneous correlation between sources of funds and our dependent variables. Data The data comes from a grass-root network called the Microfinance Centre for Central & Eastern Europe and the New Independent States (MFC for CEE and NIS). The data covers MFIs from 24 countries from the ECA region and during the five-year period from 2005 to , 8 The time period and the number of observations are limited because the data collection was discontinued in The special characteristic of our data is that it contains 12

13 information for several dimensions of the subsidized funding and not only grants, which is what previous research has used. While we have data on grants we also have data on loans at below market interest rates (subsidized loans), loans from social investors (another type of subsidy which way have both interest rate as well as term adjustments) that are offered with preferential terms relative to the standard commercial loans. The capital structure data are merged with additional financial statement data self-reported by individual MFIs to the Microfinance Information Exchange (MIX, online 2012). Credit unions (CUs) which are the largest group of MFIs in the region (about 8,000 according to the Micro banking Bulletin, 2011) are not included in the sample because of smaller sizes and tendency to lend to members and to larger businesses. Consequently, the countries, which have only CUs functioning as MFIs, are eliminated from the sample. 9 The data on country specific socio-economic characteristics come from World Development Indicators (WDI). All dollar-value figures in the dataset are 2010 dollar based on U.S. CPI. Summary statistics are presented in Table 1. It shows the average values of the performance measurements, capital structure, MFIs characteristics and macro-environment factors used to estimate the model. The data reveal that for the 310 annual observations of MFIs, the average ROA is about 3 percent. The average number of active borrowers is 11,188 per MFI and varies from only 31 in the smallest to 108,103 for the largest MFI. For the MFIs in the sample, the average loan balance per borrower/gni per capita is percent and it varies from 3.15 to more than The capital components as percent of total assets in general range from 0 to 100 percent. For example, equity funding as percent of total assets is on average about percent. Grants as percent of total assets are on average 8.09 percent and range from 0 to 78 percent (due to the rounding error). Savings as percent of total assets, as compared to other capital structure components in the sample, are the smallest with on average of 2.09 percent. Among 310 MFIs, only 40 have positive savings, which is fewer but still in line with the number for MFIs worldwide, and the average share for those with deposits is percent. The average of retained earnings as percent of total assets is 9.97 percent and it varies from 0 percent to percent. 13

14 The average of long-term debt (all three subgroups of loans together) as a percent of total assets equals to percent, which is the largest average as compared to the other components of capital and ranges from 0 to 100 percent. Using subgroups, the average of concessional loans makes is 6.65 percent, the average of funding from commercial bank loans as percent of total assets is The largest category of debt is that to social investors at about percent on average, and ranges from 0 to 97 percent. This last group of soft debt is typically not accounted for previous capital structure studies nor is the value of concessional loans Insert Table Results First, the results of the Breusch-Pagan test show that the null for independent errors of the regressions is rejected in favor of the alternative, confirming that need to use SUR, rather than independent regressions. Thus, the SUR model results reinforce the view that capital structure components affect differentially various aspects of MFIs performance. We run the SUR model as a type of a fixed effects model. In order to identify the model as such within the SUR, which does not separate the fixed-effects component, we directly embed firm dummies in the SUR structure. The least square dummy variable (LSDV) estimator is identical to the fixed effects estimator after we control for firm fixed effects in the SUR model. The capital structure elements are grouped into five categories: a) Shareholder equity; b) Grants, c) Retained earnings d) Deposits; e) Loans. The loans/debt category is disaggregated into several subcategories because some of the loans are given at subsidized interest rates and previous work has missed these details. These loans categories are loans at subsidized interest rates or Concessional loans; standard Bank loans, social investment loans (which refer to socially responsible investment ). The base group against which the categories will be compared is equity. We provide estimates from two specifications one with loans as one category, and second with loans disaggregated by type of credit to see what the impact of subsidized interest rate and of social lenders might be. These groups 14

15 represent an indirect subsidy that is typically not included in the donation and subsidy categories in previous capital structure studies. We specify a model with contemporaneous explanatory variables assuming that cap structure as well as a model with lagged explanatory variables to alleviate possible contemporaneous endogeneity issue. However, when we use lagged independent variables we lose about 1/3 of the observations and end up with only 200 observations, so we only use these results to support our main conclusions. Table 2 (Model 1a) and Table 3 (Model 1b) show main results 10. We find that relative to the use of equity ratio, one percent point increase in the ratio of grants to total assets is associated with almost 0.05 point higher ROA and equivalent improvement in the depth of outreach in our two specifications, and statistically significant at the 10 percent level only. Thus, we partially reject the hypothesis on the link of grants to financial sustainability. We do seem to observe a tradeoff between the link of grants to financial sustainability and depth of outreach. The results show that higher level of grants is associated with reaching poorer borrowers (negative and significant at the 5 percent level coefficient) suggesting that grants helped MFIs to serve poorer borrowers consistent with the dual mission of microfinance. Previous studies results that grants are less used by MFIs in the ECA region and MFIs without grants have chosen to target less poor borrowers are consistent with our finding (Cozarenco, Hudon, and Szfaraz 2016). The positive relation of grants to financial sustainability is not consistent with Bogan (2011) who found a negative link for large MFIs. We also examine the additional variables measuring (the indirect) subsidy impact on financial sustainability through the categories of loans (at subsidized interest and loans by social investors). The estimates show that the loans are negatively related to financial sustainability, and positively related to the depth of outreach. One percent change is associated with 0.04 percent point decrease in profitability of MFIs and 0.7 percent increase in reaching more and poorer clients (column 1 in Table 2). Since most of the loans are by socially oriented investors or offered at concessional interest rate, we could argue that we also find that indirect subsidies are associated with worse off financial results. Table 3 presents the disaggregate impact of loans. It shows that relative to a unit of equity to assets, a unit increase in soft loans by social investors to assets is associated with lower ROA and 0.71 improvement in the 15

16 depth of outreach measure (negative statistically significant coefficient). The results suggest that poorer clients benefit from a different indirect type of subsidy related to soft loans. In addition, a percent increase in loans at subsidized interest rate is associated with 0.96 improvement (negative coefficient means poorer borrowers are reached) in the depth of outreach measure. These result are in line with Hudon and Traca (2006) and Caudill et al (2009) suggesting that if social lenders are helping MFIs to reach target clientele, it may be at the expense of financial sustainability. Therefore, our results seem to support the idea for a mission drift or at least tradeoff between outreach and sustainability. The results show no impact of direct subsidy grants - and of indirect subsidy soft loans on the breadth of outreach of MFIs. Loans from commercial banks have been an important source of capital for MFIs since the industry opened up to the commercial loan market. Our results show that one unit increase in the use of funding from commercial banks as compared to the equity leads to a decrease in the breadth of outreach or about 1 percent decrease in the number of borrowers reached without affecting MFI profitability and depth of outreach (Column 2, Table 3). This result partially supports the hypothesis that commercial lending may negatively affect MFIs outreach confirming that commercial lenders focus on the financial bottom line. It is in line with Hoque et al. (2011), who state that increased use of commercial sources of capital tends to decrease outreach. Compared to equity to total assets, retained earnings to total assets have a positive association with profitability (0.3 point) without affecting the social performance dimensions. This result may be explained by the fact that current-period retained earnings are strongly correlated with ROA as they are used to reconcile financial statements. Better insight on the role of retained earnings may come from our specifications with lagged independent variables (but only 2/3 of the sample observations). Similarly, current period deposits relative to equity are not associated with change in any aspects of the performance of MFIs. The impact of other controls is largely as expected. We find that size matters and a percent increase in total assets is associated with 7 percent higher ROA and 40 percent more borrowers. Higher focus on lending, measured by MFI s gross loan portfolio affects performance differently. An additional percent increase in gross loan portfolio entails 0.27 percent lower ROA and about 0.50 percent increase in the number of borrowers. These results are 16

17 consistent with results showing scale economies in terms of outreach in ECA region e,g, Hartarska et al., (2013). Similarly, consistent with the results in the same paper, finding that the quality of portfolio affects financial results, we find that a percentage increase in the portfolio at risk is associated with 0.3 percent lower profitability of MFIs and is linked to more outreach to poorer borrowers. The results also show no differences in performance among MFIs of different age at least compared to the base group of mature MFIs or those established for more than eight years. Focus on women as measured by the percentage of women clients is associated with better profitability and outreach indicators. One percent increase in the number of female borrowers is associated with 0.07 points increase in the ROA and with 2 percent increase in the clients reached. Most of the macroeconomic variables do not seem to be associated with performance with a few exceptions. More rural countries have MFIs with better outreach to poorer customers because one percent increase in rural population is associated with 0.33 points decrease in the average loan balance per borrower/gni per capita. Similarly, higher GDP per capita is associated with better depth of outreach as the change in magnitude of the annual GDP per capita growth is close to its average value. This means that a country with one percent higher growth rate ceteris paribus has MFIs sector reaching significantly more poor clients, with 68 percent decrease in the mean value of the average loan balance per borrower Insert Table Insert Table

18 Robustness Checks Other contemporaneous regressions results In addition to the specifications in Tables 2 and 3, we run the SUR model with retained earnings as the omitted category. Compared to the retained earnings, a unit increase in all four capital shares is associated with lower ROA, without affecting the social performance dimensions. The only funding is that more bank loans to total assets is associated with a decrease in the number of borrowers (lower breadth of outreach) but the economic impact is relatively small; in particular, one percent increase in the ratio of bank loans to total assets is associated with about 2 percent decrease in the number of borrowers served. Next, we estimate how the use of various categories of capital structure compares to the use of deposits and re-estimating the model where deposits are the omitted base category. We do that because there is an ongoing debate whether the MFIs should and could mobilize deposits, and whether country regulations should be in favor of this transformation. In our sample of 310 MFIs there are about 1/5 deposits takers, which is a little less that other datasets utilized in the microfinance studies worldwide. The results indicate that relative to a unit of deposits, a unit increase in own retained earnings to assets ratio leads to 0.3 higher ROA without affecting on social outreach. We also evaluate if there may be differences in capital use in MFIs taking deposits. We do that by interacting deposits with all capital structure variables. The objective of this test is to see if there is a difference between the impact of the capital structure on performance in deposit taking MFIs versus lending-only MFIs. This is essentially an alternative to running separate regressions by deposit taking and loans only MFIs. We estimate specifications with added deposits times each capital structure variables. We first keeping equity as the base we add deposits times retained earnings (when equity is the base), deposits times grants, deposits times loans (and similarly interact with each of the loan categories for the second subset of specification). Next, keeping retained earnings as the base we add deposits times equity, deposits times grants, deposits times loans (and in a second set of specs its subgroups). The results (available on request) demonstrate that there is no difference in deposit taking versus lending only MFIs. This means that the current volume of deposits is not related to MFI financial and social performance (in line with Rossel- Cambier, 2012). 18

19 Lagged independent variables To address concerns about endogeneity of capital structure and performance we estimate the same models as in Table 2 and 3 but with lagged explanatory variables. This approach has shortcomings and tradeoffs that is why we are using it only as robustness check. The main issue comes from the fact that MFIs products and services are typically short term, thus the impact of a lending decision as well as the target of the lending should be seen immediately in the current period outcome. Thus, except for the category of retained earnings, we should not have contemporaneous correlations of performance and the explanatory variables including capital structure. It is also important to note that lagging the explanatory variables significantly decreases the sample size to about 200 annual observations. Nevertheless, we find and discuss several suggestive results. First, we observe that higher proportion of previous year retained earnings is associated with better breadth of outreach with one percent increase of this ratio associated with 2.6 percent increase in the number of borrowers served and no impact on the other dimensions of performance (column 2 in Tables 4 and in Table 5). Consistent with the results from Table 3 Column 3, we find that the indirect subsidy imbedded in loans at concessional interest rate is associated with improvement of the depth of outreach allowing MFIs to reach poorer borrower. The most interesting result from this specification is that previous year deposit is associated with significant improvement in ROA with one percent higher deposits raising ROA with 0.4 percent. This is in line with studies showing economies of scope from collecting savings as opposed to lending-only MFIs (Hartarska et al 2010, 2011, Delgado 201, Malikov and Hartarska, 2017) Conclusions Recent developments in the microfinance industry, such as commercialization and deposit taking, bring attention to institutions use of capital and the link to MFIs performance. The debate on whether there are tradeoffs between MFI outreach and profitability and a mission drift away from reaching many and poorer borrowers as MFIs are 19

20 becoming more commercially oriented is on-going and empirical results are mixed. The empirical literature about the capital structure of microfinance institutions is scare and growing. We contribute to the literature by focusing on the link between several dimensions of MFI performance (financial sustainability, depth and breadth of outreach) and several sources of capital. We use new panel data from MFIs operating in the ECA region during the period. Rather than using a single equation regression analysis, we use a system of equations approach the seemingly unrelated regressions method - to estimate the joint impact of seven different types of capital on the three aspects of performance. Moreover, we take advantage of data uniqueness where for the first time loans offered at a subsidized rate are separated from other subsidies such as donation so that we can see how subsidizing credit to MFIs themselves affects their outreach and sustainability. Previous work only uses one category of loans and does not account for subsidized interest rates to the MFIs, nor does it evaluate the role of other soft loans provided by social investors. This is important because our data shows that such loans amount to about 90 percent of all loans. The results suggest that in most cases the type of capital used is associated with the performance preferences of the stakeholder it represents, consistent with previous literature (Hartarska and Mersland, 2012). Relative to equity, use of grants allows MFIs to improve efficiency and depth of outreach. However, with increased commercialization, the role for grants is becoming limited, and grant funding is already a very small share in the capital structure of MFIs in ECA. Subsidized loans (both concessional and socially oriented microfinance investment), on the other hand, remain a very important source of capital. Concessional loans are positively associated with MFI s outreach without affecting financial results. Thus, we can argue that concessional loans allow poorer clients to be served, consistent with Hudon and Traca (2006). Relative to a unit of equity, a unit increase in social investors loans (another type of soft loan) entails decrease in MFI profitability but improved social performance. Our finding is also consistent with the literature on commercial loans being associated with a mission drift, because use of more commercial banks loans is associated with fewer borrowers served without affecting MFI profitability and depth of outreach. 20

21 Results are less clear about the role of deposits as a source of capital. While current level of deposits are not linked to performance, previous year deposits are associated with better financial results. The result seems to supports the idea that savings can be a way to serve the poor and possibly lower the cost of capital. Since the data is for the study period includes the financial crisis of 2008 and a year later, future work should analyze larger dataset and longer period dataset, perhaps with more regions included, to evaluate how the capital structure affects the multiple dimensions of MFI performance. 21

22 References Armendariz, B. and Szafarz, A. (2011), On Mission Drift in Microfinance Institutions. In B. Armendariz, B and Labie, M. ed. (2011), The Handbook of Microfinance, Toh Tuck Link, Singapore: World Scientific Publishing Co. Pte. Ltd, Armendariz, B. and Morduch, J. (2010), The economics of Microfinance, 2nd edn. MIT Press, Cambridge, MA. Armendariz, B. and Morduch, J. (2005), The Economics of Microfinance, MIT Press, Cambridge, MA. Augsburg, B., and Fouillet, C. (2010), Profit Empowerment: The Microfinance Institution's Mission Drift, Perspectives on Global Development & Technology 9 (3/4), Bogan, V.L., (2011), Capital Structure and Sustainability: An Empirical Study of Microfinance Institutions, Review of Economics and Statistics, DOI /REST_a_ Booth, L., Aivazian, V. and Demirguc-Kunt, A. (2001), Capital Structures in Developing Countries, The Journal of Finance 56(1), Caudill, S., Gropper, D., Hartarska V., (2009), Which microfinance institutions are becoming more cost-efficient with time? Evidence from a mixture model, Journal of Money, Credit, and Banking 41(4), Cebenoyan, A.S and Strahan, Ph., (2004), Risk Management, Capital Structure and Lending at Banks, Journal of Banking and Finance 28, CGAP and MIX (2011), MIX Microfinance World: Eastern Europe and Central Asia Microfinance Analysis and Benchmarking Report Christen, R.P. and Drake, D. (2002), Commercialization. The new reality of microfinance. In D. Drake and E. Rhyne, editors, The Commercialization of Microfinance. Balancing Business and Development, chapter 1, Kumarian Press, Bloomfield.Cozarenco, A., Hudon, M. and Szafarz, A. (2016), What Type of Microfinance Institutions Supply Savings Products? Economics Letters 140, Crombrugghe, A. d., Tenikue, M., and Sureda, J. (2008), Performance analysis for a sample of microfinance institutions in India, Annals of Public and Cooperative Economics 79(2), Cull, R., Demirguc-Kunt, A., Morduch J. (2009), Microfinance meets the market, Journal of Economic Perspectives 23(1), Cull, R., Demirguc-Kunt, A., Morduch J. (2007), Financial performance and outreach a global analysis of leading MFIs, The Economic Journal 117, F107-F133. Delgado M., Parmeter, C., Hartarska, V. and Mersland, R. (2014), Should All Microfinance Institutions Mobilize Microsavings? Evidence from Economics of Scope, Empirical Economics 48, DOI /s D'Espallier, B., Hudon, M. and Szafarz, A. (2013), Unsubsidized Microfinance Institutions, Economics Letters 120(2), D'Espallier, B., Guérin, I., and Mersland, R. (2011), Women and Repayment in Microfinance: A Global Analysis. World Development, 39(5), Farrington, T. and Abrams, J. (2002), The Evolving Capital Structure of Microfinance Institutions, Micro-Enterprise Development Review, Washington D.C., (Inter-American Development Bank Working Paper). Fehr, D. and Hishigsuren, G. (2006), Raising capital for micro finance: sources of funding and opportunities for equity financing, Journal of Developmental Entrepreneurship 11(2), Geweke, J. (2003), Contemporary Bayesian Econometrics and Statistics, Hoboken, NJ: Wiley. Gonzalez, A., and Rosenberg R. (2006), The State of Microcredit: Outreach, Profitability and Poverty, Microfinance Information Exchange, Inc. (MIX) and The Consultative Group to Assist the Poor (CGAP), Washington DC. Gonzalez-Vega, C., Schreiner, M., Meyer, R.L., Rodriguez, J. and Navajas, S. (1996), BANCOSOL: The Challenge of Growth for Microfinance Organizations, Rural Finance Program, Department of Agricultural Economics, The Ohio State University, Economics and Sociology, Occasional Paper No Greene, W. (2003), Econometric Analysis, 5th ed. New Jersey, Prentice Hall. Hartarska V., D. Nadolnyak and T. MacAdams*, (2013) Microfinance and Microenterprises Financing Constraints in Eastern Europe and Central Asia, pp in Microfinance in Developed and Developing Countries: Issues Policies and Evaluation, Gueyie, Jean-Pierre, and Jacob Yaron 22