The Trade-Off Between Sustainability And Outreach: The Experience Of Commercial Microfinance Institutions

Transcription

1 The Trade-Off Between Sustainability And Outreach: The Experience Of Commercial Microfinance Institutions Henry Francis Millson Haverford College Department of Economics Advisor: Shannon Mudd Spring 2013 Abstract The paper examines the recent trend of commercialization in the Microfinance Industry in a cross-country analysis. More specifically, the paper seeks to analyze the effects of commercialization on the sustainability and outreach of individual Microfinance Institutions (MFIs). The author uses data on 854 MFIs in 102 developing countries from the Microfinance Information exchange (MIX). The author employs fixed effects and OLS econometrics techniques, with the later accounting for endogeneity, to estimate the effects of commercialization. Ultimately, the paper finds that commercialization increases MFI sustainability and decreases outreach. The latter provides evidence of mission drift. Ultimately, if MFIs aim to alleviate poverty and provide financial services to the poorest, then commercialization may not be the solution.

2 Acknowledgments The author expresses his gratitude to the Haverford College Economics department, especially Professor Mudd and Professor Jilani, for their suggestions and support in writing this thesis. The author is also grateful to Norm Medeiros and the Haverford and Bryn Mawr College libraries for their help in researching the topic. The author also thanks the Microfinance Information exchange (MIX) for access to their datasets for the paper as well as their assistance in understanding the dataset. The author also wishes to thank his parents, Linda and Rory, his family, and Coach Bathory and the entire Men s Lacrosse team at Haverford College. Their unfaltering support throughout his entire college career invaluably helped him in the writing of this thesis. Finally, the author wishes to thank his Lord and Savior, Jesus Christ, for guidance, as all things are possible through Him.

3 Table of Contents 1 Introduction 2 Literature Review 3 Methodology 4 Data Table 3 MFI Lending Type by Region Figure 1 Profitability, Yield, and Expense Ratio by Lending Type Table 1 Summary Statistics Table 2 Correlations 5 Results i. Fixed Effects Table 4 Sustainability Regressions Table 5 Outreach Regressions Figure 2 FSS Distribution Table 6 Summary Statistics for MFIs with FSS >1 Table 7 Sustainability Regressions for MFIs with FSS <1 Figure 3 Trade-Off between ROA and Interest Rates Table 8 Sustainability Regressions for MFIs with FSS <1 Table 9 Outreach Regressions for MFIs with FSS <1 Table 10 Sustainability Regressions for MFIs with FSS >1 Table 11 Outreach Regressions for MFIs with FSS >1 ii. Latin America iii Commercialization Levels Table Commercialization Measures Affect Sustainability Table Commercialization Measures Affect Outreach 6 Conclusions 7 References 8 Appendix Variable Definitions Table 12 Latin American Sustainability Regressions Table 13 Latin American Outreach Regressions Financial Statement Adjustments and their Effects

4

5 Introduction Historically, the poor have lacked access to finance due both to assumptions about their creditworthiness and a high cost of coverage when loans are small. Microfinance institutions (MFIs) aim to bank the unbanked with innovative low costs financial services, especially microcredit, to the poor. 1 Unlike development banks of the past, though, MFIs take a market approach to lending by charging market-based interest rates and aim to serve the poor on a sustainable basis. MFIs use innovative lending techniques to make uncollateralized loans and still boast high repayment rates. Their anecdotal success led to a boom in funding, as well as growth in the number of clients. Yet, it is estimated that MFIs only serve about 15% of the target market of 3 billion potential clients (CGAPa, 2004). How can MFIs meet this unmet demand? Some suggest that tapping commercial sources of funds and deposits, i.e. becoming commercial MFIs, is the only way MFIs will be able to meet the unsatisfied demand with high quality services (CGAPb, 2004). Commercial MFIs utilize private, unsubsidized sources of funding to raise capital in order to grow and serve more clients. In becoming commercialized, these institutions now have a double bottom line : a financial objective of sustainability to satisfy their funders in addition to their social objective of outreach (CGAPa, 2004). Originally, most MFIs rely on donations and subsidies for funding. Because these sources of funds are 1 The intuition is that access to finance can alleviate poverty by smoothing consumption (Kaboski et al, 2005), increasing educational expenditures (Chowdury, 2011) and encouraging productive activities, such as business investment or housing/vehicle repairs (Kaboski et al, 2012). 1

6 vulnerable to shocks, many MFIs chose to attract private sources of funds. 2 These private sources of funding include: deposits, loans from commercial banks, bond issuances, other short-term financial liabilities and IPOs. Because they are the most prevalent, only the first two sources are considered in this paper because. By accessing deposits and loans from commercial banks, commercial MFIs can grow and avoid a liquidity crunch by tapping commercial sources of funding (Berger et al, 2006). The emphasis on profitability by depositors and lenders can push these MFIs towards an increased focus on efficiency to control costs and the innovation of new products and technologies to stay competitive. Sustainability should also be important to MFIs, especially, if they hope to help their clients out of poverty because then their longevity as an organization matters. The commercialization of microfinance has raised concerns regarding mission drift arising from the tension of the double bottom line. Serving the poorest clients has the highest average costs, which negatively impacts sustainability and profitability. Thus, if the financial objective outweighs the social objective for the MFI management then outreach will likely suffer. In addition, MFIs may move up market (Swanson, 2008) with their more successful clients. This shift away from poorer potential clients would again indicate mission drift. Bateman (2009) points out the enormous benefits for those promoting commercialization, but not for the poor on the receiving end. Since 1997, commercial financing has increased substantially, both in magnitude and as a percentage of total funding (Garmaise et al, 2010). Some within the industry have also called for the future of the microfinance industry to be driven by commercial 2 A reliance on subsidies donations are thought of as subsidies is unsustainable (Cull et al, 2007) and shocks can cause liquidity problems (Hamada, 2009). 2

7 MFIs. The best-practice model that emerged within the international development community envisaged the provision of microfinance through market-driven, independent, commercially oriented, and financially self-sustaining MFIs (Harper, 2011). Whether this commercialization trend has had an impact on the industry s sustainability and outreach is an important question to address. The paper is interested in examining whether higher levels of commercialization cause MFIs to become more sustainable or lend to more poor people (more outreach). This motivates the author s research question: Do Commercial Microfinance Institutions achieve either higher sustainability or more outreach than those MFI s who rely on less commercial funding? Three terms are central to the research question: commercialization, sustainability, and outreach. This paper defines those terms differently than the previous literature so clear definitions of those three terms are important. 3 Within the industry, commercialization has several different meanings and proxies. Previous papers have used the presence of an international investor, a Board of Directors, or whether the MFI is regulated as measures of its commercial orientation. Access to a database with newly expanded information on MFI funding sources allows for alternative, specific definitions and measurements that are connected to commercial, private sector funding. For this paper, commercialization is defined in two ways: (1) the percentage of total MFI funding that is commercial, i.e. banks loans and deposits, and (2) the portion of the loan portfolio financed using commercial funds. 3 For a more thorough explanation, please refer to the Methodology section or the Variable Definitions table in the Appendix. 3

8 The definition of sustainability within the industry and previous literature is more consistent. For this paper, the main measure of sustainability is financial self-sufficiency, which measures the ability of the MFI to cover operating costs with operating revenues. The paper also uses profitability as a proxy for sustainability as profitability is a signal for sustainability. Profitability is measured as the MFI s ability to generate a return on its assets (ROA) and equity (ROE). Sustainability and profitability are the financial objectives of each MFI and usually of direct concern to outside investors. These investors expect efficient use of their resources and potentially a financial return. The measurement of outreach is a little more complex than for commercialization or sustainability. Outreach is often used to indicate something about a MFIs clients, for example, its clients poverty level. Most commonly, the industry uses average loan sizes and the gender of the client as both are linked to poverty levels. Following this convention, the paper uses average loan sizes adjusted to each country s average income level and the percentage of female clients to measure outreach. The papers also uses the number of active borrowers as an outreach measure to see if commercialization allows MFIs to reach more potential clients, regardless of poverty level. The paper proceeds as follows: Section II provides an overview of the literature on commercialization and its effects on sustainability and outreach, Section III describes the methodology, Section IV describes the dataset, Section V describes the results, and Section VI describes the conclusions. Literature Review Financial sustainability is becoming a more important objective for MFIs as the industry matures. It indicates, by definition, the MFI s ability to cover both operating and 4

9 financing costs with revenue generated from the existing loan portfolio (Hermes et al, 2011). Donations and subsidized loans have historically been the main source of funds. Recently, though, both to achieve growth and in response to donor pressure for sustainability, MFIs have turned to commercial funding, i.e. private capital markets (Mersland et al, 2011). The main advantage of this commercialization of microfinance is access to those markets in order to grow. Currently, the industry requires immense funding to meet the unmet demand. Swanson (2008) estimates that the industry only serves $17 billion out of the $250 billion estimated microcredit market. The push towards sustainability has shifted the industry view of profitability, which is generally anathema to the non-profit sector. Profitability is now widely viewed as a signal of cost efficiency, portfolio quality, and sustainability. Increasingly, MFIs are transforming into for-profit organizations, which formally adds the goal of profitability to their objectives (Tchakoute-Tchuigoua, 2010). The relationship between interest rates and profitability is key to understand how commercialization impacts sustainability. Eisenhardt (1989) evaluates the concept of agency theory where principals (MFIs) and agents (clients) view risk differently. The situation worsens with informational asymmetry inherent in credit markets. Essentially, high interest rates can screen out safe borrowers or incentivize agents to invest in riskier projects. An MFI s loan portfolio is the firm s largest asset so agency theory is an important aspect to understand the impacts on sustainability and outreach. Interest rates affect agency theory in two main ways: first, interest rates affect delinquency rates; and second, interest rates affect financial sustainability levels (Kar, 2011). Increasing interest rates will positively impact profitability, up to a point, and hurt outreach, but if rates are 5

10 raised too high this may reduce demand, client quality, and thus profits (Kar, 2011). The relationship between profitability and interest rates is, therefore, not expected to be linear as there will be a turning point when rates are too high and profitability decreases. 4 The empirical literature shows that commercialization positively affects MFI sustainability, thus helping MFIs achieve their financial objective. Access to commercial debt is related to strong financial performance and a high level of professionalization (Mersland et al, 2011). 5 Commercialization is also essentially transforming MFIs into commercial banks with a social mission (Kar, 2011). Does this transformation indicate profitability? Cull et al (2007) find that most of the 124 MFIs in their sample are profitable after some accounting adjustments with many others approaching profitability. It should be noted that the age and size of MFIs are significantly positively linked to sustainability (Cull et al, 2007). Commercial MFIs successfully manage their return on assets (profitability) by maintaining portfolio quality, increasing efficiency and profitability levels, and accessing commercial funding (Berger et al, 2006). An important distinction in the literature on outreach is the difference between the breadth and depth of outreach as commercialization s effects may not be homogenous. Schreiner (2002) defines the breadth of outreach as the number of clients the MFI reaches, while the depth of outreach is the client s poverty level, which is more difficult to measure. To measure depth, common proxies are gender, location, and loan size as those measures are significantly correlated with poverty levels. Poorer clients tend to be 4 Kar (2011) and Cull et al (2007) find evidence of an inverted U shaped relationship suggesting a turning point and agency problems if rates are too high 5 Mersland et al (2011) define a high level of professionalization as the adoption of a Board of Directors and a stronger management team. They find these to increase efficiency and create more formal institutions. 6

11 female, rural, and take out smaller loans (Schreiner, 2002). 6 Loan size is the most commonly used proxy for the depth of outreach. The relationship between commercialization and interests rates is also important in understanding how commercialization impacts outreach. The concern is that MFIs may need to pass higher interest expenses from debt financing on to clients as higher interest rates on loans (Hamada, 2009). This decreases outreach because poorer clients would be unable to afford new loans and the MFI would then begin to serve relatively richer clients. Studies have shown alternate ways to control costs for example, increasing the borrower per loan officer ratio so that MFIs do not need to turn away from high operating costs on tiny, unsecured loans to the poor (Kar, 2011). The presence of a Board of Directors seeking to maximize shareholder benefit (Armenádriz de Aghion et al, 2005) may push MFI management towards sustainability and providing a financial return for investors. Focusing on sustainability and efficiency may go at the cost of lending to the poor. 7 Operational costs are very high for lending to the poor especially the poorest and the rural poor so sustainability and outreach may be conflicting (Hermes et al, 2011). Most MFIs primary mission is to serve the poor and alleviate poverty through access to financial services, but the transformation of MFIs after commercialization may push MFIs away from the poorest. Empirically, there is evidence that efficiency and outreach are contradictory. Larger loans are associated with lower average costs (Cull et al, 2007) so MFIs 6 Schreiner (2002) also discusses the length of outreach, which is an estimate of the future capacity of MFIs to extend loans with profits as a good estimate. This is indirectly measured in the sustainability regressions. 7 To measure efficiency, papers typically use either cost per borrower (Hermes et al (2011), Lafourcade et al (2005)) or borrowers per loan officer (D Espallier (2011)). 7

12 emphasizing cost reductions and efficiency might begin to serve the near poor, indicating mission drift. Hermes et al (2011) finds that MFIs with lower average loan balances, or more focus on female borrowers, are less efficient. 8 Even though women are better credit risks, their loans are often smaller, which negatively impacts efficiency (D Espallier, 2011). These results suggest that MFIs that emphasize outreach are less efficient so if investors push for efficiency then there are legitimate concerns about mission drift. The literature suggests that sustainability and outreach may be contradictory, so commercial MFIs may have to choose between the two. Using the portion of female borrowers as a proxy for the depth of outreach, commercial MFIs have lower outreach, since the emphasis on female borrowers negatively impacts sustainability (Mersland, 2011). However, using the average loan size as a proxy for the depth of outreach, commercialization positively impacts outreach by decreasing average loan sizes (Kar, 2011; Mersland, 2011; & Cull et al, 2007). Thus, the effects of commercialization on the depth of outreach are unclear because studies using different proxies have opposite findings. Few studies have used the number of active borrowers as a proxy so little is know about how the breadth of outreach is affected. Another presumed benefit of commercialization is stronger management and governance. Schreiner (2002) predicts that innovation and sustainability will flourish subsequent to commercialization because donors do not reward innovation and provide weaker incentives for profitability. While this paper has no measure of innovation, it 8 The paper uses cost per borrower to measure efficiency and uses the log of average loan balances (ALB) as the outreach indicator. The paper then finds that ALB negatively and significantly affects efficiency. The results are consistent when using percentage of female borrowers as the outreach indicator. 8

13 could be a channel through which commercialization affects sustainability and profitability. It is important to note that in the analysis of this paper, additional aspects of microfinance beyond microcredit are not considered. Any conclusions reached here may not be applicable to other services offered by MFIs. Also, the paper does not analyze the use of loans or over-indebtedness, which are important for understanding the benefits of microfinance (CGAP, 2011). These limitations will be important for future research. The following analysis adds to the literature in two ways. First, this paper uses funding data only recently available from the Microfinance Information exchange, also known as the MIX (refer to the Data section). This data allows for a new way to analyze the effect of commercialization by allowing for measures of both the extent and change in commercial funding sources of MFIs. Second, the author expands the examination of the impacts of commercialization on outreach by considering not only the depth but also the breadth of outreach, specifically, the number of active borrowers. Expanding the analysis to these effects is important to understand the growth of the industry and its attempt to meet unmet demand. Methodology The model investigates how commercialization affects sustainability including profitability and outreach in a cross-country analysis. The baseline specification is: Y it = α + β 1 X it + η 2 C it + γ 3 Z it + δ 4 V it + MFI FE + TIME FE + ε i The dependent variable Y is a measure of sustainability or outreach; X is a set of commercialization variables measuring funding sources; C is a set of variables measuring firm costs; Z is a set of MFI-specific controls; and V is a set of country dummies for 9

14 controls. To estimate the baseline specification, the author employs a fixed effects model with both MFI and time fixed effects. There are two issues with the basic formulation of the model reverse causality and outliers that are discussed later. β, η, γ, and δ are the parameters for each set and ε is the usual error term. The index i represents a single MFIlevel observation and t indicates the year. For the dependent variables, measures of sustainability and outreach, (Y), each has a key proxy and then alternate proxies for robustness checks. The impacts of commercialization on sustainability and outreach are measured in separate sets of regressions. The proxies for sustainability are 9 : Financial self-sufficiency (FSS) measures the MFI s ability to cover direct operating costs, with a cost of capital adjustment, using operating revenue, a value of 0.9 means the MFI is able to cover 90% of its costs, it is possible for FSS to exceed a value of Return on assets (ROA) measures the MFI s ability to generate profits on their assets, which is the traditional measure of profitability. Return on equity (ROE) measures the MFI s ability to generate profits on their equity, which is the preferred measure of profitability in the industry (Berger et al, 2006). It should be noted that the last two proxies are measures of profitability, but that is certainly a signal for sustainability. Profits raise concerns regarding mission drift as MFIs might shift toward larger loan sizes and relatively richer clients or charge higher 9 For an explanation on variable definitions or calculations, please refer to the Variable Definitions table in the Appendix. 10 The paper also uses operational self-sufficiency (OSS) as a proxy for sustainability, but the results are not significantly different from using FSS so they are not reported. 10

15 interest rates to ensure profitability. Interest rates are included as a control in the regressions. It is expected that raising interest rates would only increase profitability up to a point, though, because if rates are raised too high then adverse selection might affect the portfolio quality and decrease profitability (Kar, 2011). The proxies for outreach are: Average loan balance per borrower to GNI per capita (AVG LOAN): industry preferred measure for outreach as loan size is significantly correlated with the borrower s poverty level (Schreiner, 2002), which is adjusted to the country s income level for easier comparison in a cross-country analysis. Portion of female borrowers (FEMALE): females traditionally lack access to finance so MFIs reaching more women are assumed to reach more of the poorest potential clients (D Espallier et al, 2011), important to use portion as it controls for differences in the number of borrowers for each different MFI. Number of active borrowers (NAB): assumes that if the MFI is reaching more people then it is increasing its outreach. NAB is the weakest proxy for outreach as it relies on the assumption that increasing the clientele base reaches more poor people. There is a reasonable alternate hypothesis that MFIs could increase NAB while moving up-market and serving richer clients, thus decreasing outreach. As noted earlier, AVG LOAN and FEMALE measure the depth of outreach while NAB measures the breadth of outreach. From the outreach analysis, the author will be able to comment on how commercialization affects both the breadth and depth of outreach. The main dependent variables (X) measure commercialization by examining an MFI s reliance on deposits and commercial debt. The two variants differ only in the 11

16 denominator used: commercial funding liabilities ratio (CFLR) and commercial funding utilization ratio (COM). CFLR = (Deposits + Debt) / Total Liabilities For CFLR, as the ratio approaches one it indicates a more commercial MFI, as a larger percentage of the total MFI funding is commercial funding. COM = (Deposits + Debt) / Gross Loan Portfolio For COM, higher values correspond to a more commercial MFI, as a larger portion of their loans are financed using commercial funding sources. The author expects CFLR and COM to, on average, positively impact sustainability, but negatively impact outreach. Differences between the coefficients on CFLR and COM could suggest that MFIs are using the commercial funds to fund other activities and thus not necessarily to extend more loans. Individual variables for borrowings (DEBT), deposits (DEP), and equity (EQ) are also included to estimate how the different funding sources impact sustainability and outreach. To avoid multi-collinearity, the commercialization measures are measured in three separate regressions. 11 A dummy variable is also created for whether the MFI takes deposits (TAKEDEP) to examine possible governance issues. A limitation for the DEBT variable, and therefore also CFLR and COM, is that the MIX does not entirely differentiate between loans at subsidized or commercial interest rates. The MIX does, however, remove some subsidized loans from the DEBT measure, reclassifying them into equity or other applicable categories. 12 While the extent of subsidized debt remaining is 11 One regression uses CFLR as the commercialization measure, another regressions uses COM as the commercialization measure, and the final regression uses DEBT and DEP as the commercialization measures. EQ is included in all three regressions. 12 The MIX adjusts the data to ensure comparability across institutions and accounting practices. All data used by this paper is adjusted data from the MIX. For a further 12

17 unknown, it is expected to be relatively small using the adjusted data; therefore, DEBT and DEP are expected to be reasonably good proxies for measuring commercialization. The set of variables measuring firm costs (C) are important controls because firms looking to maximize profits and improve sustainability reduce their costs, instead of relying on donations and subsidies to cover the shortfall. To measure firm costs, the author includes: operating expense divided by total assets (OPEX), labor costs (LABOR) measured by personnel expenses divided by total assets, and cost per borrower (CPB) in all the regressions. These are the measures of firm costs used in the previous literature. The author expects higher costs to, on average, negatively impact sustainability, but positively impact outreach. The author includes various other controls in the baseline specification. First, for the MFI-specific controls (Z), the author uses the logarithm of total assets for size (SIZE), the logarithm of the gross loan portfolio (LLP), the real gross portfolio yield for interest rates (RATE) following Cull et al (2007), a yield squared variable (RATESQ) to check for non-linearity following Kar (2011), portfolio-at-risk at 30 days to measure portfolio quality (PAR30), organizational type (PROFIT) following Tchakoute-Tchuigoua (2010), and the age of the MFI in years (AGE). The predominant lending type of each MFI solidarity groups (SOLID), individual liability (INDIV), or village banking (VILLAGE) is also included as a dummy variable. 13 Second, for the country dummy explanation of the adjusted data, please refer to the Data section and the Appendix for the list of adjustments. 13 MFIs use different predominate lending methodologies. Solidarity groups follow the Grameen model and the Sixteen Decisions (Holcombe, 1995). Group liability means each group is liable for repayment. Individual liability means each individual borrower is liable for repayment. Village banking are self help groups that differ from solidarity groups because the groups are expected to transform into their own mini-banks. 13

18 variables (V), a binary variable is created for each country in which a MFI in the sample is located. Previous studies (Kar (2011), Hermes et al (2011), Cull et al (2007), Hartarksa et al (2007)) use regional dummies instead of country dummies. This paper initially used regional dummies. However, the author considers country dummies as more explanatory so the paper employs country, instead of regional, dummies in all regressions. Lastly, there are two main concerns regarding the baseline specification: endogeneity concerns regarding reverse causality, especially in the sustainability regressions, and outliers. The former means that the causality may be that more sustainable MFIs receive commercial funding, not that more commercial funding makes MFIs more sustainable. This violates the Gauss-Markov assumptions (Wooldridge, 2009). Fixed effects, while dealing with some unobservables, will not entirely solve the endogeneity issue. Therefore, the author employs an OLS regression with robust standard errors that determines how the 2008 level of each commercialization measure affects the percentage change of each dependent variable in the time frame 2008 to The OLS regression does not suffer from reverse causality because lagging variables is an accepted ad hoc solution to deal with endogeneity (Wooldridge, 2009). The coefficient will show whether a higher level of CFLR, for example, affects the direction and speed of change in the dependent variable of interest, for example FSS. To deal with outliers, the author reports some trimming in the models where outliers beyond a 14

19 threshold level are trimmed. The lending type variables are a complete set of dummy variables so INDIV is excluded. Therefore, the empirical, FE model 14 is: Y it = α + β 1 (Commercialization measure or funding source variable + EQ it + LIAB it ) + η 2 (OPEX it + LABOR it + CPB it ) + γ 3 (SIZE it + LLP it + RATE it + RATE 2 it + PAR30 it + AGE it + PROFIT it + SOLID it + VILLAGE it ) + δ 3 (Country Dummies it ) + MFI FE + TIME FE + u i Where Y it measures either how commercialization affects: sustainability using FSS, ROA, and ROE as proxies; or outreach using AVG LOAN, FEMALE, and NAB as proxies. Data The empirical analysis uses data collected by the MIX Market ( a NGO that aims to promote information exchange and transparency in the microfinance industry. To date, the MIX contains the best publicly available cross-country data on individual MFI s balance sheets and financial indicators. MFIs also report outreach and administrative measures to the MIX, which are accessible through their database. Importantly, the MIX uses a diamond ranking system to rate each individual MFI s reports based on data completeness where 1 is the least complete and 5 is the most complete. It is important to note that MFIs self report to the MIX. For the empirical analysis, the dataset consists of annual data on 854 MFIs in 102 developing countries collected from the MIX database. The data contains one 14 The OLS regression have a similar syntax except all independent variables are held at their 2008 levels. The dependent variables are averaged over the four-year period and thus become a variable measuring the change in the variable of interest. 15

20 observation per institution for each year from 2008 to Only 4 or 5 diamond ranked MFIs are chosen. There is no qualitative difference between the two rankings other than the numbers of years of reported data. 15 The MIX adjusts the data to ensure comparability across institutions and accounting practices. All data used by this paper is the adjusted data from the MIX. The adjustments include an inflation adjustment, a reclassification of certain long term liabilities as equity, an adjustment for the cost of subsidized funding, an adjustment for current year cash donations to cover operating expenses, an in-kind subsidy adjustment for donated goods and services, loan loss reserve and provisioning adjustments, some adjustments for write-offs, and the reversal of any interest income accrued on non-performing loans. 16 Further, MFIs with no assets or liabilities, incorrectly entered ROE or FSS, or other improperly reported values are removed. 17 As mentioned, the MFIs are selected based on the quality and extent of their data. Therefore, the data set is not necessarily representative of all microfinance institutions, nor is it a random sample. Summary statistics are presented in Table 1. In the sample, 36% of MFIs are located in Latin America, 21% in South Asia, 14% in East Asia & the Pacific, 10% in the Middle East and North Africa, 11% in Eastern Europe, and 8% in Sub-Saharan Africa. Referring to Table 3, 30% of MFIs use the solidarity group lending methodology, 59% 15 5 diamond MFIs report data for all four sample years and 4 diamond MFIs typically report data for three of four years sample years, but some report for all four years. 16 Adjustments are summarized in the Appendix observations are deleted because Assets < 50, 12 observations are deleted because GLP < 50, 6 observations are deleted because LIAB < 50, 61 observations are deleted because EQ < 0, 2 observations are deleted because ROE < -50 (extreme outliers), 214 observations are deleted because FSS > 50, 109 observations are deleted because CFLR > 1 (accounting error), and 5 observations are deleted because COM > 6. 16

21 use individual liability lending methodology, and 11% use village banking lending methodology. Regarding organizational type, 44% are for-profit institutions. Interestingly, over three quarters of MFIs in the sample are profitable when using ROA as a proxy with a median level of (or 2.07%). Referring to Figure 1, MFIs with individual liability are, on average, the most profitable while those that practice village banking are the least profitable while charging the highest interest rates. This is consistent with Cull et al (2007), who similarly found that a majority of MFIs in the sample were profitable. Consequently, most MFIs in the sample are also sustainable 17

22 in terms of FSS with a median FSS of Consistent with industry standards, MFIs in the sample lend predominantly to women with a median portion of female borrowers of There are MFIs in the sample, though, that lend predominantly to men. The key independent variables, the sets on commercialization and firm cost, have substantial variation. Also, the MFIs in the sample, on average, have strong portfolios with high repayment average PAR30 of Table 2 presents the results from a correlation at the 10% confidence level of the dependent variables and control variables of interest. 18 Consistent with D Espallier et al (2011), FSS is negatively correlated with FEMALE suggesting that MFIs that lend to more women are less sustainable. The individual commercialization variables, DEBT and DEP, are positively correlated with ROA and ROE, but not FSS, which suggest a positive relationship between commercial funding sources and profitability, but not necessarily sustainability. Interestingly, though, COM is negatively correlated with the sustainability proxies, which makes the relationship between commercialization and sustainability less clear. For the breadth of outreach proxies, AVG LOAN is positively correlated with CFLR, COM, and DEP, while negatively correlated with DEBT. FEMALE is positively correlated with the same variables. These opposing signs suggest that commercial funding sources may cause mission drift where MFIs are serving richer clients by giving out larger loans, but not necessarily serving fewer women. NAB is positively correlated with CFLR, DEP, and DEBT, which suggest that MFIs with access 18 According to Kennedy (2008), correlations above 0.8 detect collinearity among explicative variables. From the table, correlations among explicative variables are below that threshold. Remember, though, to avoid multi-collinearity, the commercialization measures and additional funding source variables are estimated separately. Thus, multicollinearity does not appear to be an issue. 18

23 to capital markets reach more people, but not necessarily the poorest that are likely to use smaller loans. 19

24 20

25 21

26 Results Fixed Effects In each table, the columns represent different specifications and are grouped by the dependent variable. In the sustainability regressions, columns (1)-(3) use financial selfsustainability, return on assets, and return on equity as the dependent variable, respectively. In the outreach regressions, columns (1)-(3) use average loan sizes, percentage of female borrowers, and the number of active borrowers as the dependent variable, respectively. The columns are also broken up according to the different sets of independent variables from the Methodology section, with the commercialization variables (the independent variables of interest) at the top. Table 4 presents the results for the sustainability regressions estimated using the fixed effects model on the entire sample. First, considering (4.1) with FSS as the dependent variable, the results surprisingly show that commercialization only significantly affects FSS using the commercial funding utilization ratio (COM) as the measure. When COM increases by 100 basis points, FSS actually decreases about 8 basis points from a median level of This suggests that MFIs that finance more of their loans using commercial funds are becoming less sustainable. This relationship holds when considering (4.2) with ROA as the dependent variable. Again, when COM increases by 100 basis points, ROA decreases about 2 basis points from a median level of The negative effects of commercialization on sustainability hold when considering (4.3) with ROE as the dependent variable. When COM increases by 100 basis points, ROE decreases about 24 basis points from a median level of These results are unexpected, as the hypothesized effect is that sustainability benefits from commercialization. One interesting result from Table 4 is the sign on the LABOR 22

27 coefficients in (4.2). When LABOR increases by 100 basis points, ROA increases about 22 basis points. MFIs with higher labor costs, on average, are in fact more profitable. This result could suggest that MFIs are raising wages to incentivize current personnel, attract new personal, or hire more people that in turn increase profitability. This paper can only speculate regarding the positive, significant coefficient on LABOR. Further research will be necessary to determine the causality. 23

28 Table 5 presents the results for the outreach regressions. First, considering (5.1) with average loan sizes adjusted to each country s level of average income (GNIPC), increased utilization of commercial funding, COM, positively and significantly impacts average loan sizes. A 100 basis point increase in COM corresponds to a 13 basis point increase in average loan sizes from a medial level of This means that MFIs that 24

29 finance more loans with commercial debt, on average, are making larger loans, suggesting that those MFIs are moving away from an emphasis on poorer clients. Consistently, a 100 basis point increase in DEP corresponds to a 54 basis point increase in average loan sizes. Interestingly, the coefficient on RATE in (5.1) is negative and significant a 100 basis point increase in interest rates corresponds to a 32 basis point decrease in average loan sizes. Thus, MFIs, on average, are charging higher interest rates on smaller loans i.e. on poorer clients. Second, considering (5.2) with the percentage of female borrowers as the dependent variable, increased commercial funding (CFLR) negatively and significantly reduces the emphasis on females. A 100 basis point increase in CFLR corresponds to a 3 basis point decrease in the percentage of female borrowers. Therefore, due to commercialization MFIs may make larger loans and reach less female clients, suggesting diminished outreach. The results from Table 4 suggest that commercialization negatively impacts sustainability, which seems counterintuitive. Table 1 shows that the mean and median FSS are both greater than 1 in fact, over 75% of the sample have FSS greater than 1. Figure 2 shows the distribution of FSS within the sample and it confirms that most MFIs are already self-sufficient. Therefore, commercialization may not have much room to improve sustainability if most MFIs in the sample are already self-sufficient and sustainable. To determine whether the effect of commercialization differs depending on whether FSS is above or below 1, the sample is split and separate regressions are run. Table 6 presents the summary statistics for MFIs with FSS less than 1 and Table 7 presents the summary statistics for MFIs with FSS greater than 1. Some differences are immediately evident. MFIs with FSS less than 1: lend to more women, have lower 25

30 average loan sizes, have less clients, are less profitable, receive less commercial funding, but utilize more of that funding for loan generation, have higher costs per borrower, and have riskier loan portfolios than MFIs with FSS greater than 1. 26

31 27

32 28

33 Table 8 presents the results for the sustainability regression for MFIs with financial self-sufficiency less than 1. First, considering (8.1) with FSS as the dependent variable, the results show that commercialization does in fact increase FSS for those MFIs in the sample that were not all ready self-sufficient. All three sets of commercialization measures have positive coefficients, but only the coefficient on COM 29

34 is significant. A 100 basis point increase in CFLR, COM, DEBT, or DEP corresponds to a 7 basis point, 7 basis point, 7 basis point, or 20 basis point increase in FSS from a median level of , respectively. Second, considering (8.2) with ROA as the dependent variable, the results are positive, but insignificant for the effects of commercialization on profitability. The results also show the same interesting coefficient on LABOR with a positive and significant impact on ROA. A 100 basis point increase in LABOR corresponds to 22 basis point increase in ROA from a median level of Again, MFIs could be paying higher wages or recruiting more personnel. This effect is decreasing, though, as shown by the negative, insignificant coefficient on the RATESQ term. Figure 3 shows the relationship between interest rates and predicted values of ROA. There is slight inverted U-shape similar to Kar (2011). Interestingly, interest rates never approach the turning point where profitability begins decreasing, which is quite high at 406%, according to the data. 30

35 Table 9 presents the results for the outreach regressions for MFIs with FSS less than 1. First, considering (9.1) with AVGLOAN as the dependent variable, the effects of commercialization are insignificant. Increasing the MFIs equity does significantly decrease average loan sizes, although the effect is marginal. This suggests that increased equity funding does allow for MFIs to keep loan sizes small and lend to poor clients. Second, 31

36 Second, considering (9.2) with FEMALE as the dependent variable, MFIs that utilize more commercial funding in loan financing lend to less women a 100 basis point increase in COM corresponds to 3 basis point decrease in the percentage of female clients from a median level of This significant, negative result suggests that the depth of outreach decreases due to commercialization. Interestingly, PAR30 is significant and negative in column (2), which suggests that MFIs that lend to more women have better repayment, consistent with D Espallier (2011). Third, considering (9.3) with the number of active borrowers as the dependent variable, the effects of commercialization are positive, but insignificant. It seems MFIs reach more clients with increased funding, but without statistical significance the paper cannot comment on the effect of commercialization. Table 10 presents the results for the sustainability regressions for MFIs with FSS greater than 1. First, considering (10.1) with FSS as the dependent variable, the commercialization measures are insignificant and the coefficients oscillate between positive and negative. Thus, for MFIs that are already financially self-sufficient, increased commercialization has little effect in terms of increasing sustainability. Interestingly, the RATE term is negative and significant a 100 basis point increase in interest rates reduces FSS about 35 basis points from a median level of This result seems counterintuitive. One possible explanation is that MFIs with FSS greater than 1 lend to relatively richer clients that may have credit alternatives. Therefore, charging higher interest rates could screen out the safer borrowers as they could find alternative sources for loans. Second, considering (10.2) with ROA as the dependent variable, the commercialization measures are negative, but insignificant. Again, the 32

37 coefficient on LABOR is positive and significant while the coefficient on OPEX is negative and significant. This suggests that increasing costs decrease profitability unless those increased costs are labor related. The RATESQ term from (10.2) is positive and significant, suggesting the effect of charging higher interest rates on profitability is 33

38 increasing, which is not consistent previous results or Kar (2011). Table 11 presents the results for the outreach regressions for MFIs with FSS greater than 1. First, considering (11.1) with AVGLOAN as the dependent variable, increased utilization of commercial funding, COM, positively and significantly impacts average loan sizes. A 100 basis point increase in COM corresponds to a 23 basis point increase in average loan sizes from a medial level of This means that MFIs that 34

39 utilize more commercial funding to finance their loan portfolios, on average, make larger loans. Consistently, a 100 basis point increase in DEP corresponds to a 68 basis point increase in average loan sizes. One possible reason for this is that deposit accounts for the poor are extremely costly to the MFI so they may need to lend to relatively richer clients, and thus reduce costs, in order to compensate for the cost of providing savings to the poor. Second, considering (11.2) with FEMALE as the dependent variable, the commercialization measures again negatively and significantly impact the targeting of female clients. A 100 basis point increase in CFLR, DEBT, or DEP corresponds to a 5 basis point, 5 basis point, or 6 basis point decrease in FEMALE, respectively, from the median level of Thus self-sufficient MFIs lend to less women, on average, when receiving more commercial funding. These two results suggest that increased levels of commercialization, even after full FSS is achieved, negatively impacts the breadth of outreach. Third, considering (11.3) with NAB as the dependent variable, commercialization also negatively and significantly impacts the breadth of outreach. Results Latin America Although the author includes country dummies for controls, there are cultural factors that could influence the results. For example, regional factors, such as shared culture and history, could affect the preferred lending methodology or organizational type of the MFI. Therefore, the author re-runs his models using data from institutions within Latin America the region with the most MFIs in the sample. 19 The results are presented in the Appendix. Table A-12 presents the results from the Latin American sustainability regressions. The results are consistent with Table 4 where the commercialization 19 Consistent with Hull et al (2007) who also find their results are robust to potential unobservable cultural factors after re-running their regressions for Latin America. 35

40 measures negatively, and significantly, impact self-sufficiency and profitability. Again, a majority of MFIs in Latin American are already sustainable so the differential impacts of commercialization by sustainability level might be hidden. Table A-13 presents results from the Latin American outreach regressions. The results are consistent with Table 5 where the commercialization measures positively, and significantly, impact average loan sizes, suggesting mission drift. The results from the previous tables seem robust to cultural factors. Results 2008 Commercialization Measures As mentioned earlier, the above fixed effects regressions may suffer from endogeneity. Therefore, the author employs an OLS model holding all variables at their 2008 levels except for the dependent variables of interest. For those, the author creates the average percentage change of the dependent variables, for example FSS, over the four year time period. The variables in the tables are thus labeled as change in the dependent variable of interest, for example CFSS. The OLS model uses robust standard errors and country dummies again that are not reported. The author posits that this OLS model does not suffer from reverse causality so it satisfies the Gauss Markov assumptions. As standard OLS struggles with outliers, the author also reports trimmed regressions in each table. Table 14 presents the results for the regressions to determine how the 2008 commercialization measures affect MFI sustainability. First, considering (14.1) with the percentage change in FSS as the dependent variable, the results are consistent with Table 8 from the fixed effects regressions. A 100 basis point increase in COM corresponds to a 36

41 37

42 6 basis point, or 4 basis point with 1% trimming, increase in CFSS. This suggests that more commercial firms in 2008 are more sustainable in the subsequent years. Interestingly, the positive, significant coefficient on LABOR shows that MFIs that have higher labor costs become more sustainable. This is consistent with previous findings where paying higher wages or hiring more people benefits sustainability. Higher wages could attract more competent workers or incentivize current workers to work harder and become more productive. Second, considering (14.2) with the percentage change in ROA, the results suggest that more commercial firms in 2008 are subsequently less profitable, but only in the trimmed regressions. Higher levels of commercialization tend to, on average, make MFIs more sustainable in subsequent years. Increased interest expenses might negatively impact profitability. Table 15 presents the results for the regressions to determine how the 2008 commercialization measures affect MFI outreach. First, considering (15.1) with the percentage change in average loan sizes as the dependent variable, the commercialization measures significantly increase average loan sizes. The trimmed and untrimmed coefficients are similar when both are significant. For example, a 100 basis point increase in COM leads to a 6 basis point or 5 basis point increase in average loan sizes in the years after Second, considering (15.2) with the percentage change in female clients as the dependent variable, commercialization reduces the emphasis on females in the trimmed regressions. A 100 basis point increase in COM leads to a 3 basis point reduction in the percentage of female clients. These two results suggest that MFI s are decreasing their breadth of outreach due to commercialization. One explanation is the higher cost of lending to women as shown by the significant negative coefficient on CPB. 38

43 39

44 Conclusions At the outset of the paper, the author posed the research question: Do Commercial Microfinance Institutions achieve either higher sustainability or more outreach than those MFI s who rely on less commercial funding? Embedded within that question is an examination of the double bottom line within each MFI as the management juggles the social and financial objectives. In regards to sustainability, the effects of commercialization depend on the initial sustainability level of the MFI. Using the entire sample, the results in (4.1) show that increased commercialization negatively affects sustainability. This seemed counterintuitive. However, as noted earlier, most MFIs in the sample are profitable and sustainable. Splitting the sample uncovered more nuanced effects of commercialization. For MFIs that were working towards self-sufficiency (8.1), commercialization increases sustainability. The results from the OLS specification that accounted for endogeneity (15.1) showed again that increased commercialization increases MFI sustainability. Another result consistent across the specifications is that higher labor costs actually increased sustainability. This suggests MFIs are paying higher wages to recruit new talent or incentivize better performance from current personnel or are hiring more people. As expected, interest rates positively affected sustainability levels, but only to a certain point for MFIs that were not yet sustainable. In regards to outreach, the effects of commercialization were surprisingly more robust and consistent for the outreach regressions. Concerning average loan sizes, the results from the entire sample (5.1) suggest that commercialization increases average loan sizes. This indicates that MFIs are loaning to less poor clients. These results are 40

45 consistent after the data split (11.1) and in the OLS specification that accounts for endogeneity (15.1). This provides evidence of mission drift where MFIs are changing their client composition, potentially due to a preference for sustainability. Concerning the percentage of female clients, the results were again consistent across the regressions. Commercialization decreased the emphasis on female borrowers in the entire sample (5.2), in the data split (9.2, 11.2), and in the OLS specification (15.2). Therefore, the increased commercialization decreased the breadth of outreach for MFIs within the sample. MFIs, on average, are increasing loan sizes and lending to fewer women. The results in (5.1) also show that MFIs, on average, charge lower interest rates on larger loans and charge higher interest rates when targeting more female clients. The latter could be evidence of MFIs passing the higher costs from interest payments onto clients. Commercialization does positively impact the microfinance industry through sustainability gains, but this does not offset the negative impacts of diminished outreach and mission drift. This suggests that the financial objective of sustainability outweighs the social objective of outreach for MFIs in the sample. There are significant policy implications for an industry supposedly concerned with helping the poorest. MFI management need to weigh the benefits of increased funding against the cost of moving away from those whom they aim to serve. Also, if poverty alleviation and eradication are actually goals then it seems we need to reconsider commercialization as a solution to solve issues within the microfinance industry. Commercialization does not necessarily diminish outreach, but MFI management need to be clear with investors that the objectives in the double bottom line are equally important. 41

46 References Armenádriz de Aghion, Beatriz, and Jonathan Morduch. The Economics of Microfinance. Cambridge: MIT (2005). Bateman, Milford. "Commercialization: The Death of Microfinance." Why Doesn't Microfinance Work? The Destructive Rise of Local Neoliberalism. London: Zed (2009) Berger, Marguerite, Maria Otero, and Gabriel Schore. "Pioneers in the Commercialization of Microfinance: The Significance and Future of Upgraded Microfinance Institutions." Inside View of Latin American Microfinance. Washington: Inter-American Development Bank (2006): CGAP. Financial Inclusion with a Double Bottom Line: Implications for the Future of Microfinance. CGAP Occasional Paper 8 (2004a). CGAP. Foreign Investment in Microfinance: Debt and Equity from Quasi-Commercial Investors. CGAP Focus Note 25 (2004b). CGAP. Foreign Capital Investment in Microfinance: Reassessing Financial and Social Returns (2011). Chowdhury, S. S., & Chowdhury, S. A. Microfinance and Women Empowerment: A panel data analysis using evidence from rural Bangladesh. International Journal of Economics and Finance, 3(5) (2011): Cull, Robert, Asli Demirgüç-Kunt, and Jonathan Morduch. "Financial Performance and Outreach: A Global Analysis of Leading Microbanks." The Economic Journal 117 (2007): F D'Espallier, Bert, Isabelle Guérin, and Roy Mersland. "Women and Repayment in Microfinance: A Global Analysis." World Development 39.5 (2011): Eisenhardt, Kathleen M. Agency Theory: An Assessment and Review. The Academy of Management Review 14.1 (1989): Garmaise, Mark J., and Gabriel Natividad. "Information, the Cost of Credit, and Operational Efficiency: An Empirical Study of Microfinance." The Review of Financial Studies 23.6 (2012): Hamada, Miki. "Commercialization of Microfinance in Indonesia: The Shortage of Funds and the Linkage Program." The Developing Economies 48.1 (2010): Hamilton, Lawrence C. Statistics with Stata. 10th ed. Belmont, CA: Brooks/Cole of Cengage Learning (2009). Harper, Malcolm. "The Commercialization of Microfinance: Resolution or Extension of Poverty?" Confronting Microfinance - Undermining Sustainable Development. Ed. Milford Bateman. Sterling, VA: Kumarian, Hartarska, Valentina, and Denis Nadolnyak. "Do Regulated Microfinance Institutions Achieve Better Sustainability and Outreach? Cross-country Evidence." Applied Economics 39 (2007): Hermes, Niels, Robert Lensink, and Aljar Meesters. "Outreach and Efficiency of Microfinance Institutions." World Development 39.9 (2011): Holcombe, Susan. Managing to Empower: The Grameen Bank's Experience of Poverty Alleviation. Zed Books: London (1995). Kaboski, J. P., & Townsend, R. M. The Impact of Credit on Village Economies. American Economic Journal: Applied Economics, 4(2) (2012): Kaboski, J. P., & Townsend, R. M. Policies and impact: An Analysis of Village-level 42

47 Microfinance Institutions. Journal of the European Economic Association, 3(1) (2005): Kar, Ashim Kumar. "Microfinance Institutions: A Cross-Country Empirical Investigation of Outreach and Sustainability." Journal of Small Business and Entrepreneurship 24.3 (2011): Kennedy, Peter. A Guide to Econometrics. 8th ed. Cambridge: MIT (2008). Lafourcade, Anne-Lucie, Jennifer Isern, Patricia Mwangi, and Matthew Brown. "Overview of the Outreach and Financial Performance of Microfinance Institutions in Africa." MIX Bulletin April (2005). Mersland, Roy, and Ludovic Urgeghe. "Performance and International Investments in Microfinance Institutions." CEB Working Paper 11/054 (2011): Mersland, Roy, and R Øystein Strøm. "Performance and Trade-Offs in Microfinance Organizations - Does Ownership Matter?" Journal of International Development 20.4 (2008): Schreiner, Mark. "Aspects of Outreach: A Framework for Discussion of the Social Benefits of Microfinance." Journal of International Development 14 (2002): Swanson, Brad. "The Role of International Capital Markets in Microfinance." Microfinance: Emerging Trends and Challenges. Ed. Suresh Sundaresan. Cheltenham: Edward Elgar, (2008): Tchakoute-Tchuigoua, Hubert. "Is There a Difference in Performance by the Legal Status of Microfinance Institutions?" The Quarterly Review of Economics and Finance 50 (2010): Wooldridge, Jeffrey M. Introductory Econometrics - A Modern Approach. 4th ed. Mason, OH: South-Western of Cengage Learning (2009). 43

48 Appendix 44

49 45

50 46

51 47