1

Microfinance and the Market

Preliminary draft for Journal of Economic Perspectives

--Not for citation or circulation--

Robert Cull World Bank

Asli Demirgüç-Kunt

World Bank

Jonathan Morduch New York University

and Financial Access Initiative

October 18, 2007 The views are those of the authors and not necessarily those of the World Bank or its affiliate institutions. Morduch is grateful for funding from the Bill and Melinda Gates Foundation through the Financial Access Initiative. The Mix Market provided data through an agreement between the World Bank Research Department and the Consultative Group to Assist the Poor. Confidentiality of institution-level data has been maintained. We thank Isabelle Barres, Joao Fonseca, and Peter Wall of the Microfinance Information Exchange (MIX) for their substantial efforts in assembling both the adjusted data and the qualitative information on microfinance institutions for us. We have benefited from conversations with Richard Rosenberg and Adrian Gonzalez. Varun Kshirsagar provided expert data analysis, and Aparna Dalal provided additional assistance. All views and any errors are ours only.

2

In April 2007 Banco Compartamos of Mexico held a public offering of their stock. Insiders sold 30 percent of their holdings for $407 million.1 The sale was over-subscribed by thirteen times, and Compartamos was suddenly worth $1.6 billion.2 The popularity of the offering was hardly surprising given Compartamos’s financial numbers: their return on equity had hit 57 percent in 2006.3 Two years before, the Economist magazine favorably compared Compartamos’s outsize profitability to Citigroup’s return on equity of 16 percent.4 A month before the offering, the Economist had written that “Compartamos may not be the biggest bank in Mexico, but it could be the most important.”5 Compartamos’s claim to importance stems from its clients—not from their elite status, but from the opposite. They are mainly poor women, taking loans to support tiny enterprises like neighborhood shops or tortilla-making businesses. The loans the women seek are small—typically hundreds of dollars rather than many thousands--and the bank requires no collateral. It is a version of “microfinance,” the idea associated with Muhammad Yunus and Grameen Bank of Bangladesh, winners of the 2006 Nobel Peace Prize. For Yunus, access to small loans is the key to unleashing the productive potential of hundreds of millions of poor people currently without access to banks or at the mercy of exploitative moneylenders. In Yunus’s vision, financial access can help customers build businesses that generate much-needed income and, with that, reduce poverty. The Grameen Bank experience showed that, with donor support, a large-scale bank could be built dedicated to serving very poor households--and the bank could do so without requiring that customers pledge assets as collateral. For the supporters of Compartamos, the public offering heralds a future in which microfinance routinely attracts investment from the private sector, freeing it from the ghetto of high-minded, donor-supported initiatives. As testimony to the power of profit, Compartamos’s supporters point to the institution’s aggressive expansion, fueled largely by retained earnings: between 2000 and 2007, Compartamos grew from 60,000 customers to over 600,000, making it by far the largest single “microlender” in Latin

1 News of the introductory public offering (IPO) was reported by Reuters, (Noel Randewich, April 20, 2007.) 2 Accion International, “Following Up: Replies to Remaining Compartamos IPO Webcast Questions”. http://www.accion.org/media_noteworthy.asp_Q_N_E_378. Accessed July 28, 2007. 3 Return on equity is an institution’s net income for a year divided by its total equity, giving a measure of how well the institution translates its net assets into profits and earnings growth. An important and clear analysis of the IPO is available in Richard Rosenberg, “CGAP Reflections on the Compartamos Initial Public Offering: A Case Study on Microfinance Interest Rates and Profits.” CGAP Focus Note 42. Washington, DC: CGAP, p. 14, fn 8. The numbers cited here are from Rosenberg; his original data are from the MixMarket (Microfinance Informatio Exchange). 4 The Economist (2005). “The Hidden Wealth of the Poor,” The Economist, November 5, 2005: 5. 5 The Economist (2007), “Economics Focus: Small Loans and Big Ambitions,” March 17, 2007.

3

America. Microlenders can and should compete shoulder-to-shoulder with mainstream commercial banks, they say, vying for billions of dollars on global capital markets.6 But Muhammad Yunus was not among those rejoicing: “I am shocked by the news about the Compartamos IPO,” he announced. “When socially responsible investors and the general public learn what is going on at Compartamos, there will very likely be a backlash against microfinance.”7 His ire was raised by Compartamos’s very high interest rates. Compartamos’s customers pay interest of over 100 percent per year on loans, once taxes are included. In 2005, nearly one quarter of the bank’s interest revenue went to pure profit, and it was this staggering profit rate that propelled the success of the IPO.8 For Yunus, the high interest rates and large profits were unconscionable, extracted from Mexico’s poorest citizens, women struggling to support families, often without the help of a husband at home. In Yunus’s depiction, Compartamos is nothing but a brute moneylender, the very beast that Grameen Bank was built to root out. An NGO leader in Latin America argued that Compartamos’s strategy is “socially, economically, and politically dangerous and should be morally condemned.”9 For Compartamos’s supporters, though, the profits allowed Compartamos to serve hundreds of thousands of poor customers who otherwise would have had even worse financial options.10 Would not serving them be a better moral outcome? Microfinance “has lost its innocence,” a Compartamos-supporter declared. “To mourn this loss of innocence would be wrong…To attract the money they need, [micro-lenders] have to play by the rules of the market. Those rules often have messy results.”11 Still, the prospects are alluring. The Compartamos IPO makes it possible to imagine funding microfinance globally at $30 billion per year, rather than today’s $2 billion.12 It is a hope that makes it possible to imagine serving over 1 billion low-income customers, rather than the 113 million counted in 2005.13

6 See, for example, Steven Funk, “Remarks by Steven Funk, Founder and Member of the Dignity Fund Board.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 7 Muhammad Yunus, “Remarks by Muhammad Yunus, Managing Director, Grameen Bank.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 8 The IPO also reportedly netted the two founders of Compartamos over $50 million each. (The analysis is from Chuck Waterfield, posted on the Devfinance listserve, and based on disclosure statements at the time of the IPO.) 9 Carmen Velasco, “Remarks by Carmen Velasco, Co-founder and Co-executive Director, Pro Mujer, Inc.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 10 For more on Compartamos, Grameen Bank, and their diverging paths, see Connie Bruck, “Millions for Millions,” The New Yorker, October 30, 2006. Rosenberg (2007) argues that Compartamos could have substantially reduced interest rates (and profit rates) and nonetheless expanded, but at a somewhat slower pace. 11 Damian von Stauffenberg, “Remarks by Damian von Stauffenberg, Executive Director, MicroRate.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 12 The $30 billion projection for the eventual scale for microfinance is from Steven Funk, “Remarks by Steven Funk, Founder and Member of the Dignity Fund Board.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 13 Data are from Sam Daley-Harris, State of the Microcredit Summit Campaign Report 2005. Washington, DC: Microcredit Summit Campaign.

4

Yunus’s vision can be messy too, and the fall-out from the Compartamos IPO offers an important lens on microfinance and its promise to reduce global poverty. We begin by describing microfinance’s triumphs. Most important, in the face of early pessimism, microfinance leaders successfully demonstrated that poor households can be reliable customers, paying back loans and saving in steady quantities. Increasingly, customers often buy insurance too. Moreover, the customers are willing to pay for quality financial services, an insight that enables microfinance institutions to cover most of their costs or even, as with Compartamos, to turn a handsome profit. The experiences--taking place in cities and villages in Latin America, Africa, and Asia--refute decades of assertions that the way to serve the poor is with massive subsidies. These accomplishments should be celebrated even if poverty reduction is not a primary goal. They pave the way for broadening access to finance for hundreds of millions, perhaps even billions, of low-income people who today lack ready access to formal financial services. Such access is not guaranteed to radically transform lives, increase economic growth, or contribute much to meeting the United Nations Millennium Development Goals (the first and best-known goal is cutting the worst of world poverty by half). Instead such access can do something more fundamental: it can expand households’ abilities to cope with emergencies, manage cash flows, and invest for the future – i.e., basic financial capabilities that most of us take for granted but that are especially valuable for low-income households seeking stability and the hope of a better future. When it comes to poverty reduction specifically, the picture is complicated, and the data reveal divides. We draw on a data set that includes most of the world’s leading microfinance institutions. Investors seeking pure profits would have little interest in most of the institutions we see that are serving poorer customers. These institutions focus on serving poor women and make small loans, but they only make modest profits or survive in part on subsidies, even if not massive ones. Yunus’s Grameen Bank is emblematic. On the other side are banks making considerably larger loans (and thus likely serving households that are less poor), serving fewer women as a fraction of customers, and typically earning higher profits. The hope for microfinance as a commercial success can be seen most naturally in this latter group. But the hope for microfinance as a social success—reaching the very poor and most disenfranchised—can be seen most naturally in the former. The big question, and one that remains fundamentally unresolved, is whether it is possible for the typical micro-lender to simultaneously be both a true commercial success and a maximally-effective social success. Some institutions approach the promise of the “win-win” scenario, but they are not typical in the data. The elusiveness of the win-win possibility does not mean that microfinance has no future. But the future will entail multiple paths. There is no reason in principle to have to choose between the paths represented by the profit-driven Banco Compartamos (or a less-expensive variant) and the socially-conscious Grameen Bank. As long as funding is available, a wide range of strategies is possible. The questions are: can commercial microfinance expand on its own terms? Can subsidies and other strategies be used elsewhere to make socially-driven microfinance institutions even more effective? And,

5

are those impacts apt to be large enough—and can causal links be proven rigorously—to ensure a continuing flow of donations? The logic of microfinance The earliest and greatest triumph of microfinance was the demonstration that poor households could be reliable bank customers. The received wisdom at the start of the 1970s held instead that substantial subsidies were necessary for institutions serving poor households. Government banks thus took on the task of serving the poor, focusing mainly on farmers. Driven by political imperatives rather than profit motives, the state-run banks charged low interest rates (well below what the market could bear) and collected loan repayments half-heartedly. The risks inherent in agricultural lending together with the misaligned incentives led to institutions that were costly, inefficient, and not particularly effective in reaching the poor. Beginning mainly in the 1980s, the early microfinance pioneers started by shifting the focus. They abandoned the farmers and instead turned to customers running “non-farm enterprises”—like handicrafts, livestock-raising, and petty trade. In most places men dominate farming decisions, but women tend to play larger roles in running household side-businesses, and women quickly became the main microfinance clients. By 2000, 95 percent of Grameen Bank’s customers were women, and similar percentages held for institutions outside of Bangladesh. Funding their businesses confers many advantages on the new bankers: the non-farm businesses tend to be less vulnerable to the vagaries of weather and crop prices, generate income on a fairly steady basis, and can initially increase scale quickly. It soon became clear that the women were repaying their loans to the micro-lenders far more reliably than their husbands had repaid the state banks. Microlenders boast repayment rates of 95 percent and higher, achieved without requiring that loans be secured with collateral. The high loan repayment rates were credited to new lending practices, especially “group lending” (also called “joint liability” lending). Customers were typically formed into small groups and required to guarantee each others’ loan repayments, aligning their incentives with those of the bank. The mechanisms sparked much interest, including interest from academic economists. But a broader set of mechanisms is now recognized as contributing to microfinance successes—especially the credible threat to deny defaulters’ access to future loans, with or without group contracts.14 Most borrowers in our data are still under group contracts, but a large share is not, and we expect trends to tilt toward the latter group. Getting loans repaid was a start, but the ante was soon raised. In the 1980s and 1990s, Washington-based policymakers began arguing that the new institutions should be fully profitable as well (or, in the prevailing code language, they should be “financially sustainable”). Compartamos is the extreme example. The logic begins with an argument that poor households can pay high interest rates—i.e., that, within reason, access to finance is more important than its price. Moneylenders, it is often pointed out, routinely 14 Work by economic theorists started with Stiglitz (1990). For other references, see Rai and Sjöström (2004) and Armendáriz and Morduch (2005).

6

charge (annualized) interest rates of over 100 percent per year, so charging anything lower must be a benefit.15 The next part of the argument holds that subsidies were at the root of problems in state banks, and that, even in nongovernmental institutions, ongoing subsidization weakens incentives for innovation and cost-cutting. The final part of the argument holds that subsidies are not available in the quantities necessary to fuel the growing sector, so, no matter how you feel about other arguments, there is no practical alternative to pursuing profitability. Raise interest rates, institutions were thus instructed. Use subsidies sparingly, and only in the start-up phase. Earn ample profits. Attract private investors. Expand without limit to serve more of the under-served. Here is where the messiness comes in. While Yunus also decries the corrosive role of subsidies (arguing that they weaken the incentives to build strong financial relationships), Yunus nevertheless looks to “socially-responsible investors” for funding. Those investors support microfinance in pursuit of a “double bottom line”: they seek a financial return, but not the maximum possible return. In compensation, they also seek a social return, knowledge that their investment is reducing poverty or creating good in some other way. Social investors, whether individuals or institutions, are turning to microfinance in a big way: in 2006, investors put $2 billion into microfinance, much of it in pursuit of a double bottom-line.16 Social investors range from international financial institutions like the World Bank’s International Finance Corporation to major mutual fund families like TIAA-CREF, in addition to individuals investing $100 or so (at zero financial return) through internet-based sites like Kiva.org. But even if called “investors,” ultimately they also provide subsidies (equal to the size of the investment multiplied by the difference between the microlenders’ cost of capital if obtained through the market and the financial return, if any, taken by the social investor). For microfinance to continue expanding on these terms, institutions will need to maintain access to a stream of subsidized funds—and that will depend on the ability to prove the institutions’ social worth relative to other social interventions. The evidence below shows that subsidized institutions look different from others (in ways that are consistent with their having greater outreach to the poor), but better evidence is needed to strongly make the case. The future will likely continue along varied paths. To the extent that the institutions seek commercial funding, they will have to continue improving their competitiveness and profitability. And, to the extent that institutions seek continued subsidies, they will have to start taking impacts (and rigorous impact evaluations) more seriously, a process which is picking up steam.

15 This is a common argument, but it risks comparing apples with oranges. Moneylender loans are often taken for short periods (under a month), while microfinance loans are typically held for months. The standard Grameen Bank loan had a year term. 16 CGAP Press Release: “Threefold increase in microfinance investment funds signals boom in foreign capital investments, CGAP.” Consultative Group to Assist the Poor, Washington, DC. http://www.cgap.org/portal/binary/com.epicentric.contentmanagement.servlet.ContentDeliveryServlet/News/press_release_07-02-07.html.

7

The MixMarket data set The essay draws on a data set that includes most of the world’s leading microfinance institutions.17 The data set is relatively large, covering 346 institutions with nearly 18 million active microfinance borrowers and a combined total of $25.3 billion in assets (in purchasing power parity terms). Most of the clients (about 10 million) are in the top 20 largest institutions. This fact alone shows how the microfinance world has segmented into some very large organizations (with over one million customers) alongside many smaller, community-based organizations with membership just in the thousands.18 The data were collected by the Microfinance Information Exchange (the MIX), a not-for-profit private organization that aims to promote information exchange in the microfinance industry. The database contains one observation per institution from 2002 to 2004 (we choose the most recent data). These data, collected for publication in the MicroBanking Bulletin (MBB), have been adjusted to help ensure comparability across institutions. A critical strength of the data set is that the numbers are adjusted to show the roles of both explicit and implicit subsidies—and, to the extent possible, to bring them into conformity with international accounting standards. (There are no international standards now, and Grameen Bank, for example, has for decades claimed profitability when in fact its earnings have seldom covered its full costs.) The adjustments include an inflation adjustment, a reclassification of some 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. The data thus allow the clearest evidence on profitability and subsidy, as well as giving reliable data on portfolio at risk, cost structures, and financial leverage. The institutions have been selected based in large part on the quality and extent of their data, and participation is voluntary. The data set is thus not representative of all microfinance institutions, and the sample is skewed toward institutions that have stressed financial objectives and profitability. They do, however, collectively serve a large fraction of microfinance customers worldwide. More important, the data set focuses on the institutions that are best-positioned to meet the promise of microfinance – to reduce poverty and create sustainable banks. While the data set lacks direct measures of outreach to the poor, it includes proxies that include average loan size, percentage of women served, and rural location. These 17 The data set is not in the public domain (to protect the confidential financial information of participants), but the World Bank Research Department have access through a negotiated agreement with the MixMarket (Microfinance Information Exchange). Information on the MIX is available at their website www.themix.org. The MFI data include outreach and impact data, financial data, audited financial statements, and general information. 18 In the larger data set of Gonzalez and Rosenberg (2006), 91 percent of the 1565 institutions they analyze in 2003-4 institutions are small, collectively serving just a quarter of the borrowers. The other three-quarters are served by just 145 institutions.

8

indicators turn out to be highly correlated and, at a broad level, help to distinguish between institutions serving the poorest customers versus those that have focused up-market (to mainly poor and low-income, but not the poorest, customers). The limits of the data set are addressed in part through comparisons with the parallel work of Gonzalez and Rosenberg (2006). They also analyze the Mix Market data, but they merge it with two much larger data sets. The two other data sets have information on a wider range of institutions, forming a total of 2600 institutions world-wide (serving 94 million borrowers), but the data are lower quality. For the most part, the comparison makes our conclusions even sharper. Profitability Figure 1 sets the scene with a plot relating profitability and the extent of non-commercial funding. The measure of profitability is the financial self-sufficiency ratio, a measure of an institution’s ability to generate sufficient revenue to cover its costs.19 Values below one indicate that it is not doing so. The horizontal axis gives the “non-commercial funding ratio.” The ratio is zero if all funds come from either commercial borrowing or deposit-taking. The ratio is 1 if the institution draws funds from neither source, instead relying on donations, borrowing at below-market interest rates (i.e., subsidized loans) or equity.20 The gently downward sloping line shows a link between lower profitability and greater reliance on non-commercial funding. This is unsurprising since unprofitable institutions turn to subsidies to stay afloat, while more profitable institutions have an easier time raising commercial capital. More important is the scatter plot of data points, each representing a microfinance institution. Many points are above the threshold for profitability, and many are on the left of the graph, indicating low reliance on soft funds. This is the hope of commercial microfinance. But note too that an ample number fall below the threshold and to the right. In regressions below, we show that the distinctions persist even after controlling for age, location, and financial structure. [Figure 1 about here] The bottom row of Table 2 shows that, of the 315 institutions with data on profits, 57 percent were profitable, an impressive achievement given the challenges of the undertaking. Moreover, since profitable institutions tend to serve more customers, 87 percent of all borrowers were served by profit-making institutions.

19 The financial self-sufficiency ratio is adjusted financial revenue divided by the sum of adjusted financial expenses, adjusted net loan loss provision expenses, and adjusted operating expenses. It indicates the institution’s ability to operate without ongoing subsidy, including soft loans and grants. The definition is from MicroBanking Bulletin (2005), p. 57. 20 The “non-commercial funding ratio” is defined as (donations + non-commercial borrowing + equity) divided by total funds. Here, donations are defined as: donated equity from prior years + donations to subsidize financial services + an in-kind subsidy adjustment. Equity is the sum of paid-in capital, reserves, and other equity accounts; it does not include retained earnings or net income. Commercial borrowing refers to borrowing at commercial interest rates (though in practice it can be hard to determine where the market would set those rates). Non-commercial borrowing, in parallel, is borrowing at concessional interest rates (with the same caveat as above). Total funds are the sum of donations, equity, deposits (both savings and time deposits), commercial borrowing, and non-commercial borrowing.

9

Given that this is a self-selected sample of leading institutions, we also look to evidence from the larger data set of Gonzalez and Rosenberg (2006). There, profit-making institutions are again much larger than others. But they find that only 44 percent of borrowers are served by profit-making institutions (in their data, profits are self-reported, so this is likely an upper bound). The average is dragged down by some large and very unprofitable government banks. When focusing on private institutions and NGOs, about 60 percent of borrowers are served by (self-described) profitable institutions. The top panel of Table 2 disaggregates the data by institutional type. Banks are, not surprisingly, more likely to be profitable than others (73 percent of institutions are profitable), and nongovernmental institutions less profitable (54 percent). Unlike the banks, the NGOs do not all aim for profitability, and the most commercially-successful NGOs may choose to transform into banks. The evidence is thus partly driven by the choice of social mission and partly by the choice of institutional type. These patterns suggest differences that will emerge in finer detail below. The second panel divides by lending method. Individual lending refers to traditional bilateral contracting between the bank and customer. Solidarity group lending refers to group contracts, and the “village bank” approach captures a participatory lending method also based around group responsibility for loan repayments. In general, the village banks reach the most costly-to-reach and poorest customers; the solidarity group lenders also pursue poorer households, and the individual lenders tend to go “up market.” The profitability figures echo this pattern, with the village banks being least profitable (43 percent of institutions), the solidarity group lenders slightly more profitable (55 percent), and the individual lenders most profitable (68 percent). Table 2 maps institutional types into lending methods, showing that nearly two-thirds of banks lend through individual methods, while three-quarters of NGOs lend through one of the two group-based methods. Similarly, three-quarters of the village banks are either NGOs or “non-bank financial institutions,” a regulatory category that allows NGOs to be more closely regulated in return for greater capability, like deposit-taking. Differentiation Table 2 begins to show the differences between institutions huddled under the microfinance umbrella. The pie charts in Figure 2 show how the difference play out on a wider set of dimensions. The top left chart shows the composition of our data set of leading institutions. About three-quarters of the institutions are either NGOs or non-bank financial institutions, and about 10 percent are banks. But the banks are large in scale, comprising over half of all assets (converted into purchasing power equivalents to yield $25.3 billion in total assets). NGOs, in contrast, make up 45 percent of the institutions but can claim just 21 percent of the total assets. [Figure 2 about here]

10

The story shifts when measuring scale in terms of borrowers rather than assets. NGOs can claim about half of the 18 million customers here, with banks claiming one quarter. Donors at large aid agencies have pushed hard to encourage the commercialization of microfinance, and have often diminished the place of NGOs in the microfinance landscape. But the evidence here suggests that they still matter in a big way. That impression is reinforced in the data of Gonzalez and Rosenberg (2006) which shows that NGOs served one quarter of the 94 million borrowers seen in 2004, with self-help groups serving another 30 percent. (The self-help groups are an Indian variant of microfinance and are often organized by NGOs.) Banks and licensed non-bank financial institutions served just 17 percent of all borrowers. Government institutions—often inefficient and substantially-subsidized--over-shadowed the banks by serving 30 percent of all coverage. In terms of borrowers, the greatest outreach at this juncture is thus not from commercial institutions but from others. Trends in outreach will likely shift toward banks as they grow and spread, but the evidence suggests that the banks tend to complement the NGOs rather than substitute for them. That assertion (and, with the available data, it can only be that), gains further support in Figure 2. The pieces of evidence on assets and borrowers combine to give an important result. Since the biggest component of an institution’s assets is generally its loan portfolio, the charts can only be reconciled if banks are making much larger loans per borrower than NGOs. Having the capacity to use (and repay) larger loans is usually assumed to be associated with higher socio-economic status, and to the extent that this is so, the evidence suggests that NGOs are focused more sharply on poorer households than are the banks. Gonzalez and Rosenberg (2006) again give helpful corroborating evidence. In their data, institutions are asked to self-report on the percentage of poor borrowers among customers. They are also asked to self-report on the percentage of small loans they make (specified as loans under $300). A simple univariate regression yields that a 10 percentage point increase in the fraction of small loans is associated with a 9 percentage point increase in the self-reported fraction of poor borrowers served. The link between smaller loans and greater outreach to the poor is tight, though self-reporting bias could explain some of the correlation. The lack of sharp data on the poverty levels of customers is a major hindrance here, and the evidence lags far behind the loftiest rhetoric on the potential for microfinance to reduce poverty. In particular, debate persists about whether, outside of Asia, microfinance can make a major dent in populations living on under $1 per day, or make a dent in national economic growth rates. And, most fundamentally, debate persists on whether there are important trade-offs between pursuing profit and reaching the poorest customers. The data here suggests that the trade-offs are real, but the evidence comes from proxy indicators rather than direct evidence. The bottom two pie charts in Figure 2 provide further evidence of differentiation. The left chart focuses on gender: while NGOs serve 51 percent of all customers, they serve 74 percent of women. Banks, in contrast, serve 25 percent of all customers but just 6

11

percent of female customers. The identification of microfinance with serving women comes through the NGOs, not the traditional commercial financial institutions. (Note that only 290 of the 346 institutions report on their coverage of women, and NGOs are more likely to report, which is telling in itself.) The bottom right chart completes the funding picture. We count $2.6 billion in subsidized funds fueling the 346 institutions. Of this, the NGOs take a share that is disproportionate, in terms of the number of customers reached and, especially, in terms of their assets. Banks absorb subsidies too, but in much smaller relative quantities. The evidence we turn to next explores costs, subsidies, profits, and outreach in greater detail. Table 3 maintains the focus on NGOs and banks. For each group, the range of experience is captured with data at the 25th percentile, median, and 75th percentile. The first row of column (2) indicates that for over half of NGOs at least 85 percent of customers are female. The top quarter of NGOs exclusively serve women. Banks serve women, but in lower numbers. For slightly less than half of institutions, men make up the majority of customers. The banks tend to be more profitable, and perhaps this explains the difference here, but column 4 breaks out the median for profitable NGOs as a comparison. By and large, their data are much closer to that of other NGOs than that of banks. The second row of Table 3 shows how loan sizes vary. For comparability, average loan sizes are divided by the income of households at the 20th percentile of the income distribution in the given country. The ratio is 48 for the median NGO, but over 4 times that for the median bank (ratio = 224). Again, profitable NGOs (ratio = 60) are much closer to other NGOs than to banks. At the 75th percentile of the bank sample, the loan size measure reaches 510, suggesting a very different scale of lending. The sample of banks is slightly older than the NGOs (on average by about one year), but it is a modest difference. Less modestly, their size is far bigger. Average assets (in purchasing power terms) are $30.4 million for the median bank, but just 6 million for the median NGO. The top quarter of banks have assets greater than $244 million, while the top quarter of NGOs has assets above $12.2 million only. The banks also have many more borrowers per institution; a comparison of the median NGO vs. bank yields a ratio of roughly 1:3 in the number of active borrowers. The larger loan sizes translate into lower costs per dollar lent, as seen in the sixth column. The median bank spends 12 cents per dollar of loans outstanding in operating costs, while the median NGO spends 26 cents. Figure 3 shows this relationship graphically for the entire sample. The average operating cost per borrower in Table 3 is in line with this finding: the comparison at the medians for NGO vs. banks is $156 to $299. The NGOs are keeping costs down (in part by giving lower quality services), but it is not enough to compensate for the smaller loan sizes. Being able to spread costs across larger loans helps the banks. [Figure 3 about here]

12

A recent study from Indonesia shows how a major commercial bank incorporates this fact into their decision-making (Johnston and Morduch 2007). Bank Rakyat Indonesia sets a break-even loan size, below which they estimate that the bank would lose money on the loan. The study considers households under the poverty line that are otherwise deemed creditworthy by BRI staff; of these potential customers, 42 percent desired loans under BRI’s threshold. The response for NGOs facing high costs is to raise interest rates—not to the triple digits charged by Compartamos, but to levels much higher than banks. The interest rate premium in the eighth row is the excess of the interest rate over the risk-free cost of capital prevailing in the formal financial sector. At the median, NGOs charge 18 percentage points in interest (annualized) over and above the risk-free rate. Banks, at the median, charge just 8 percent as a premium. The highest costs of borrowing in microfinance are not being charged by the banks, the institutions most focused on profits. The highest costs are being charged by the institutions most focused on social missions – and their cost structures explain why. Figure 4 shows the relationship for the whole sample, with a clear positive relationship between costs and interest rates. Figure 5 shows that it is operating costs (rather than capital costs and loan loss provisions) that drive the differences in total costs – where, here, we show the relationships by lending method. Figure 6 puts the pieces together by showing the average real yields (a measure of average interest rates) for the different types of institution. Banks have a relatively compressed distribution of interest rates and they are comparatively low. NGOs, on the other hand, have a wider distribution and a higher mean. [Figures 4, 5, and 6 about here.] The lowest tier of NGOs is doing poorly in terms of profitability (though that may not be their primary concern). At the median, though, the NGOs do respectably in financial terms, thanks to the relatively high interest rates they charge. Figure 7 shows this graphically; the positive link between larger loan sizes and greater financial self-sufficiency is weaker than the link between larger loan sizes and smaller average costs in Figure 3. Table 3 shows that the median NGO is profitable (bear in mind, once more, that this is a selected sample of leading institutions), and the median profitable NGO has a financial self-sufficiency ratio of 1.14 relative to 1.04 for the median bank. Profitability is a hard-won accomplishment given that the presumption had long been that meaningfully serving the poor can only be done with subsidy, a presumption consistent with mainstream economic theory (e.g., Stiglitz and Weiss, 1981). Still, the profit levels are modest in a comparative sense. We started by noting Compartamos’s outsize return on equity above 50 percent and Citigroup’s 2004 return on equity of 16 percent. Here, though, the median return on equity for NGOs is 3 percent and, for banks, 10 percent. Mere profitability will not be enough to attract funds from investors optimizing their portfolios to maximize returns – though the returns may be enough to tempt social investors.

13

[Figure 7 about here.] As a consequence, subsidies enter the funding equation. Subsidy per borrower, again in purchasing power parity equivalents, was $139 for the median NGO, reaching $659 for the NGO at the 75th percentile. The median bank, on the other hand, took no subsidy. The final row shows that the median bank relied mainly on commercial funding and deposits. The median NGO, in contrast, turned to non-commercial borrowing and donations with far greater frequency. Table 4 gives the full picture on funding, showing banks and NGOs at extremes in terms of financial structures, with other institutions falling in between. For the NGOs, 39 percent of funding came from donations, with another 16 percent coming from non-commercial (soft) loans. For the banks, the two categories contributed just 3 percent to total funding. In contrast, commercial borrowing and deposits combined to give 84 percent of total funds. In sum, we see a rough split between two types of institutions. The first cluster is more likely to involve an individual lending method, larger loans, fewer women customers, lower costs per dollar lent, higher costs per borrowers, and greater profitability. The second cluster is more likely to employ a group lending method, give smaller loans, serve more women, employ subsidies more heavily, face higher costs per dollar lent, and be less profitable. The looming exceptions come from South Asia, where the cost of hiring staff of requisite quality tends to be lower than elsewhere, and thus allows more favorable pricing and profitability. In criticizing the Compartamos IPO, Muhammad Yunus declared that “A true microcredit organization must keep its interest rate as close to the cost-of-funds as possible…My own experience has convinced me that microcredit interest rates can be comfortably under the cost of funds plus ten percent, or plus fifteen percent at the most.”21 Table 3 and Figure 5 show that most NGOs charge more than Yunus’s desired upper range. And the cost data suggest that, if they charged less, the NGOs would require even more subsidy or would go out of business—unless the NGOs could somehow find untapped ways to cut costs. The logic of subsidy The question around the “win-win” possibility takes the ambitions of microfinance insiders on their own terms. But it not clear that we should. Public economics provides a different frame, one stemming from the utilitarian principles of Jeremy Bentham and John Stuart Mill. The focus is on simple costs and benefits: do the costs of subsidizing microfinance generate large social benefits? Does microfinance reach the poor in large numbers? Does it create meaningful changes in customers’ lives? Does it compare well to alternative interventions? The track record so far is encouraging in terms of scale. The most commonly-cited count yields over 100 million microfinance customers at the end of 2005.22 Multiply that by 5 to approximate the total number of people affected through family members’ access, and the total comes to about half a billion poor people worldwide. Evidence so far, though, suggests that in most places, especially outside of

21 Muhammad Yunus, “Remarks by Muhammad Yunus, Managing Director, Grameen Bank.” Microcredit Summit E-News, Volume 5, No. 1, July 2007. 22 Sam Daley-Harris, State of the Microcredit Summit Campaign Report 2005. Washington, DC: Microcredit Summit Campaign.

14

South Asia, typical customers are seldom among the world’s 1 billion “very poor” people, defined as living on under $1 per day per person; typical customers are nevertheless disadvantaged and operate largely in the informal economy.23 Perhaps the “very poor” are the wrong target group (generally too limited in skills and marketing ability, too vulnerable to economic shocks), but the case is not strongly established. No matter the group, there is a strong prima facie case for expecting substantial economic and social impacts when eliminating credit rationing, and the presumption is backed up by ample anecdotal evidence and evidence on high returns to capital for poor businesses—e.g., McKenzie and Woodruff 2006 for data from Mexico. But rigorous empirical evidence based on credible control-treatment evaluations remains scant. With the institutional data, we can not determine whether the subsidies are welfare-enhancing or not. Theory suggests that subsidies can increase both efficiency and welfare by helping to fix missing markets, but, under other conditions, subsidies can reduce efficiency by weakening incentives for diligence and innovation, a circumstance commonly attributed to subsidies used by large state-run banks. Here, there is no clear evidence that subsidy reduces efficiency, but we do see that subsidized institutions look different from others. Table 5 gives regression results from a series of equations that control for regional unobserved factors, and country-level economic variables. Our main interest is in the role of the “non-commercial funding ratio” described earlier. We have not corrected for possible endogeneity biases (e.g., institutions with higher costs are likely to turn more often to non-commercial funding, driving reverse causation). Instead, we investigate whether basic relationships, of the sort described earlier, are robust to a broad range of controls. The first column shows that institutions that rely more heavily on subsidized resources also make substantially smaller loans, an indicator of their focus on poorer households. The second column shows that their operating costs per unit lent are also substantially higher. This second result could simply result from small loans driving up unit costs, or it could be a symptom of subsidization breeding inefficiency, or it could be that the NGOs using more subsidized resources work with costlier-to-reach households. The third column shows that institutions that rely more heavily on subsidized resources face higher operating costs per borrower, with a coefficient that is large but noisily measured. As we saw earlier, institutions raise interest rates to compensate for high costs, so the net effect on profitability is ambiguous. But the final column of Table 5 shows that raising interest rates is insufficient to cover all costs: use of subsidy is strongly associated with lower profit rates. The future The lack of financial access for poor and low-income households continues to be a concern for policymakers globally, and it is increasingly presented as a profit opportunity

23 The “$1 per day” poverty lines are short-hand. In practice, the international poverty lines are set using purchasing power parity (PPP) adjusted exchange rates, anchored so that it the line approximates the purchasing power of $1.08, set in 1993. The lines are then adjusted for local cost of living increases in later years.

15

for commercial banks. Microfinance will no doubt continue to expand and become part of the financial mainstream. Experience so far, though, suggests that the profile of commercial banks in low-income communities looks very different from that of NGOs. The latter focus more heavily on the poor and on women, and they lend at smaller scales. The focus here has been largely on lending, but banks are also very much in the deposit-taking business, and they are turning increasingly to insurance. Deposits can offer a reasonable cost of funds, as well as giving customers secure ways to accumulate. The challenge for microfinance institutions is that the transactions costs of handling small deposits can be high, and without innovation to lower those costs, they are unlikely to seek the business of poorer potential depositors. The scenario parallels that of credit, seen in the tables and figures above—and it holds for insurance too. Reducing transactions thus costs becomes a major goal. At the same time, it should not be assumed that bargain-basement cost-minimization is always the right strategy, even when working in poor communities. We see that customers are willing to pay for quality financial services—as the Compartamos example shows in the extreme. The optimal outcome is to develop products with a mix of reliability, flexibility, and structure that works for customers and that can be delivered at reasonable prices (but not necessarily the very cheapest prices). Muhammad Yunus’s dictum to set interest rates no higher than the cost of capital plus 15 percent would likely worsen outcomes and access for many of the poorest households, especially outside of South Asia—though that could change with future innovations. New technologies are looming that can help. Banking through mobile telephones is taking off in the Philippines, South Africa, and Kenya. Mobile-banking can help cut costs and increase the quality of services by taking advantage of the spread of mobile phones, even in poor communities. New automated teller machines and debit cards are also being developed and implemented. As these technologies spread, the basic idea of microfinance will likely expand too. In the course of rethinking, it will be helpful to go back to the very start. The original idea of “micro-credit” focused sharply on funding small, capital-starved businesses, but several decades of experience has shown that the demand for loans extends beyond customers running businesses. And even customers with small businesses often seek loans for other needs, like paying for school fees or coping with health emergencies. A brighter future would offer new loan products for general purposes, new savings products, and better to reduce risks. That future will also entail the idea that access to basic financial capabilities is important in its own right, even if it does not lead to wide-scale poverty reduction or economic growth. Moreover, the problem of lacking access to finance extends well beyond households below poverty lines, and commercial possibilities appear greater when moving up-market. The evidence here shows that, when it comes to poverty reduction, institutions driven only by the impetus of the market are unlikely to focus on the poorest households and communities. The evidence also suggests that there will remain important roles for social

16

entrepreneurs and social investors with strong social missions—a notion that is undiminished by recognition of the importance of strengthening markets.

17

References Armendáriz de Aghion, Beatriz and Jonathan Morduch (2005). The Economics of

Microfinance. Cambridge, MA: MIT Press. Besley, Timothy (1995). “Savings, Credit, and Insurance,” Handbook of Development

Economics, Volume 3A, Chapter 36. Amsterdam: North-Holland. Conning, Jonathan (1999). “Outreach, sustainability and leverage in monitored and peer-

monitored lending,” Journal of Development Economics 60: 51–77. Cull, Robert, Asli Demirgüç-Kunt, and Jonathan Morduch (2007). “Microfinance

Performance: A Global Analysis” Economic Journal, February. Dehejia, Rajeev, Heather Montgomery, and Jonathan Morduch, (2005). “Do interest

rates matter? Evidence from the Dhaka slums.” Columbia University, Department of Economics and NYU Wagner School working paper.

Gonzalez, Adrian and Richard Rosenberg (2006). “The State of Microfinance—

Outreach, Profitability, and Poverty (Findings from a database of 2600 microfinance institutions.” Presentation at World Bank Conference on Access to Finance, May 30, 2006.

Johnston, Don and Jonathan Morduch (2007). “The Unbanked: Evidence from

Indonesia.” Presentation at World Bank Economic Review Symposium, March 15-16.

Karlan, Dean and Zinman, Jonathan (2005). “Observing Unobservables: Identifying

Information Asymmetries with a Consumer Credit Field Experiment,” Princeton University, Woodrow Wilson School, draft.

McKenzie, David and Christopher Woodruff (2006). “Do Entry Costs Provide an

Empirical Basis for Poverty Traps? Evidence from Mexican Microenterprises,” Economic Development and Cultural Change, October.

Microbanking Bulletin (2005). “Trend Lines,” Issue 10, No. 5, March. Rai, Ashok and Tomas Sjöström (2004). “Is Grameen lending effcient? Repayment

incentives and insurance in village economies,” Review of Economic Studies 71 (1), January: 217-34.

Robinson, Marguerite (2001). The Microfinance Revolution: Sustainable Banking for the

Poor. Washington, DC: The World Bank. Rutherford, Stuart (2000), The poor and their money. New Delhi: Oxford.

18

Rutherford, Stuart (2006), “Uses and users of MFI loans in Bangladesh,” MicroSave Briefing Notes on Grameen II, Number 7. [Available at www.microsave.org.]

Stiglitz, Joseph (1990). “Peer Monitoring and Credit Markets” World Bank Economic

Review 4 (3). 351-366. Stiglitz, Joseph, and Andrew Weiss (1981). “Credit Markets in Credit with Imperfect

Information”, American Economic Review 71: 393 – 410.

19

.51

1.5

Fina

ncia

l Sel

f - S

uffic

ienc

y

0 .2 .4 .6 .8 1Non-Commercial Funding Ratio

Figure 1. Profitability versus non-commercial funding. Profitability is measured by the financial sustainability ratio. The financial self-sufficiency ratio is adjusted financial revenue divided by the sum of adjusted financial expenses, adjusted net loan loss provision expenses, and adjusted operating expenses. It indicates the institution’s ability to operate without ongoing subsidy, including soft loans and grants. The definition is from MicroBanking Bulletin (2005), p. 57.

20

Table 1. Profitability of institutions and borrowers. Source: MicroBanking Bulletin data set. Profitability is defined by a financial sustainability ratio above 1. NBFI = non-bank financial institution. NGO = non-governmental organization. Credit unions include credit cooperatives.

Institutions Active Borrowers

Number in sample

Percent Profitable

Number (millions)

Percent

served by profitable

institutions Institution type Bank 30 73%

4.1 92% Credit union 30 53% 0.5 57% NBFI 94 60% 2.6 75% NGO 148 54% 8.9 91% Lending method Individual 105 68%

7.2 95% Solidarity group 157 55% 7.4 85% Village bank

53 43% 1.6 67%

Total

315 57%

16.1 87%

21

Table 2. Lending method by institution type (percent). 346 Observations. NBFI = non-bank financial institution. NGO = non-governmental organization. Credit unions include credit cooperatives. Numbers in parentheses are percentages by column. 16 rural banks and one non-classified institution are omitted.

Individual

Solidarity group

Village bank

Bank 64 (19) 33 (7) 3 (10) Credit union 47 (14) 50 (10) 3 (10) NBFI 36 (31) 50 (30) 14 (30) NGO 23 (31) 52 (41) 25 (45)

Total

34 50 17

22

Bank10.4%

Credit Union10.4%

NGO44.5%

NBFI29.8%

Other0.3%

Rural Bank4.6%

Bank54.9%

Credit Union4.3%

NGO20.7%

NBFI18.7%

Rural Bank1.3%

Institutions (346 total)

Assets

($25.3 billion, purchasing power parity adjusted; 276 observations)

NGO50.8%

Credit Union5.7%

NBFI17.5% Bank

25.3%

Other0.1%

Rural Bank0.7%

Individual Lending46.2%

Solidarity Group Lending44.1%

Village Bank Lending9.7%

Borrowers (17.6 Million total; 346 observations)

NGO74.1%

Bank5.9%Rural Bank

0.6%

0.1%

Credit Union3.7%

NBFI15.6%

Bank18.4%

Credit Union2.6%

NGO60.7%

NBFI18.3%

Rural Bank0.1%

Female Borrowers (12.1 million total; 290 observations)

Subsidized funds

($2.6 billion, purchasing power parity adjusted; 281 observations)

Figure 2. Distribution of microfinance institutions by institutional type

23

Table 3. Non-governmental organizations versus banks. Return on equity is adjusted net income divided by total equity. Subsidy per borrower numbers are donations from prior years plus donations to subsidize financial services plus an in-kind subsidy adjustment plus an adjustment for subsidies to the cost of funds.

Non-governmental organizations Banks

25th

pctile Median 75th

pctile Median if profitable

25th pctile Median

75th pctile

(1) (2) (3) (4) (5) (6) (7) Percent women 64 85 1 86 23 52 58 Average loan size/income at 20th percentile (%)

27 48 135 60 110 224 510

Age 5 9 13 9.5 6 10.5 12.5 Assets ($PPP million) 1.6 6.0 30.2 12.2 4.2 30.4 244 Active borrowers 3,063 7,416 23,010 11,115 1,889 20,262 60,678 Operating cost/loan value (%) 15 26 38 21 7 12 21 Operating cost/active borrower 84 156 309 157 118 299 515

Interest rate premium (%) 11 18 28 20 4 8 16

Financial self-sufficiency ratio 0.78 1.02 1.17 1.14 0.99 1.04 1.15 Return on equity (%) -11 3.3 13.8 11.3 1.6 10 22.9 Return on assets (%) -5.9 0.7 4.7 4.1 -.1 1.3 3.2 Subsidy/borrower (PPP$) 72 139 659 199 0 0 136 Non-commercial funding ratio

.31 .74 1 .53 0 .11 .22

24

Table 4. Shares of total funding by institutional type. Means [standard deviations in brackets]

Shares of total funding

InstitutionType Donations

Non-Commercial Borrowing Equity

Commer-cial

Borrowing Deposits

Median non-

commer-cial

funding ratio

0.02 0.01 0.13 0.13 0.71 0.11 Bank

(24 obs) [0.09] [0.037] [0.16] [0.19] [0.30]

0.01 0.005 0.11 0.09 0.78 0.12 Rural Bank (13 obs) [0.04] [0.01] [0.05] [0.12] [0.15]

0.11 0.03 0.16 0.06 0.64 0.21

Credit Union

(30 obs) [0.22] [0.11] [0.15] [0.10] [0.29]

0.23 0.11 0.18 0.28 0.21 0.45 NBFI (88 obs) [0.30] [0.20] [0.24] [0.30] [0.29]

0.39 0.16 0.08 0.26 0.10 0.74 NGO

(134 obs) [0.34] [0.25] [0.20] [0.29] [0.18]

0.26 0.11 0.13 0.23 0.27 0.43 Total (289 obs) [0.33] [0.21] [0.20] [0.27] [0.34]

25

0.2

.4.6

.8O

pera

ting

Expe

nses

/ G

ross

l Loa

n Pr

otfo

lio

0 2 4 6 8 10Avg. Loan Size/Income(20th pctile)

Figure 3. Average costs per dollar lent fall as loans get larger. X-axis gives the average loan size as a fraction of the average income of households at the 20th percentile of the national income distribution.

26

-.20

.2.4

.6.8

Prem

ium

0 .2 .4 .6 .8 1 Adj.Operating Expenses / Gross Loan Portfolio

Figure 4. Interest rates rise with costs. The “premium” is the excess of the average real yield over the cost of capital as given by IMF International Financial Statistics.

27

0 .1 .2 .3 .4% Gross Loan Portfolio

Village Bank Lending

Solidarity Group Lending

Individual Lending

All variables are means

% Operating Expense % Cost of Funds % Loan Loss Provisions

Figure 5. The composition of costs. Data are given as a fraction of the average gross loan portfolio.

28

0 50 100Real Yield (percent)

Rural Bank

Non-Bank Financial Intermediary

Non Governmental Organization (NGO)

Credit Union/cooperative

Bank

Figure 6. Interest rates on loans, by institutional type. Average real yields are total cash financial revenue from lending divided by the average gross loan portfolio, adjusted for inflation. They give a measure of average interest rates for the institution. The edges of the boxes give the 25th and 75th percentile of the variable. The line in the middle of the box gives the median. The “whiskers” that extend give the 5th (left whisker) and 95th percentiles (right whisker).

29

.51

1.5

Fina

ncia

l Sel

f - S

uffic

ienc

y

0 2 4 6 8 10Avg. Loan Size/Income(20th pctile)

Figure 7. Average loan size has a weak relationship with profitability

30

Average

Loan Size / Income at

20th percentile

Oper-ating

Expenses / Gross

Loan Portfolio

Operating Expenses (PPP$)/

Borrowers

Financial Self-

Sufficiency Ratio

(1) (2) (3) (4)

-1.03** 0.09** 53.9 -0.24*** Non-Commercial Funding Ratio [0.42] [0.04] [121.0] [0.07]

0.30*** -0.01** 47.5** 0.03*** Log Assets (PPP$) [0.08] [0.005] [22.2] [0.01]

1.27 0.20 206 1.03 Mean of dependent

variable Observations 225 239 240 230

R-squared 0.271 0.294 0.243 0.285 Table 5. The associations between non-commercial funding, outreach, and costs. Controls for (log) age of institution, inflation, country-level governance, GDP growth, 5 region variables, lending method (village banking and solidarity group lending). Standard errors in brackets.