47957 The New Moneylenders: Are the Poor Being Exploited by High Microcredit Interest Rates? O ver the past two decades, institutions that widespread agreement, within the industry at least, make microloans to low-income borrowers in PAPER that in most situations MFIs ought to pursue financial developing and transition economies have focused sustainability by being as efficient as they can and by increasingly on making their lending operations charging interest rates and fees high enough to cover financially sustainable by charging interest rates that the costs of their lending and other services.1,2 are high enough to cover all their costs. They argue that doing so will best ensure the permanence and Nevertheless, accepting the importance of financial expansion of the services they provide. Sustainable sustainability does not end the discussion of interest (i.e., profitable) microfinance providers can continue to rates, and where to draw the line is a complex serve their clients without needing ongoing infusions issue. An interest charge represents money taken of subsidies, and can fund exponential growth of out of clients' pockets, and it is unreasonable if it services for new clients by tapping commercial not only covers the costs of lending but also deposits sources, including deposits from the public. "excessive" profits into the pockets of an MFI's private owners. Even an interest rate that only covers The problem is that administrative costs are inevitably costs and includes no profit can still be unreasonable higher for tiny microlending than for normal bank if the costs are excessively high because of avoidable lending. For instance, lending $100,000 in 1,000 inefficiencies. loans of $100 each will obviously require a lot more in staff salaries than making a single loan of High microloan interest rates have been criticized $100,000. Consequently, interest rates in sustainable since the beginning of the modern microfinance microfinance institutions (MFIs) have to be substantially movement in the late 1970s. But the criticism has higher than the rates charged on normal bank loans. intensified in the past few years, and legislated interest rate caps are being discussed in a growing number As a result, MFIs that claim to be helping poor people of countries. Part of the reason for the increased nevertheless charge them interest rates that are concern about rates is simply that microfinance is considerably above the rates richer borrowers pay drawing ever more public attention, including political at banks. No wonder this seems wrong to observers attention. Another factor is that quite a few MFIs who do not understand, or do not agree with, the are now being transformed into private commercial argument that MFIs can usually serve their poor corporations. customers best by operating sustainably, rather than by generating losses that require constant infusions In the early years most MFIs were ownerless of undependable subsidies. not-for-profit associations, often referred to as nongovernmental organizations (NGOs). If an NGO In today's microfinance industry, there is still some generates a profit, the money normally stays in the OCCASIONAL debate about whether and when long-term subsidies institution and is used to fund additional services. might be justified in order to reach particularly But many NGO MFIs have eventually wanted to challenging groups of clients. But there is now add deposit-taking to their activities, because they No. 15 1 The term "microfinance" usually refers to the provision of financial services to poor and low-income clients who have little or no access to February 2009 conventional banks. The term is often used in a more specific sense, referring to institutions that use new techniques developed over the past 30 years to deliver microcredit--tiny loans--to informal microentrepreneurs. The range of services can include not only microcredit but also savings, insurance, and money transfers. This paper focuses on interest rates charged on microcredit; it does not address other microfinance services. Richard Rosenberg, 2 Among borrowers who have loans from nongovernmental organizations and private MFIs, a majority are served by financially sustainable Adrian Gonzalez, institutions (Gonzalez and Rosenberg 2006). and Sushma Narain 2 see savings services as valuable for their clients as excessive, and indeed about whether terms like and because capturing deposits allows them to this have any useful meaning, at least in the arena of fund expansion of their microlending. When NGOs for-profit microfinance. approach a government banking authority for a license to take deposits, they are usually required to In this paper, the authors are not using any reorganize their businesses into for-profit shareholder- theoretical framework or benchmark against which owned corporations. Once this happens, profits to measure what is excessive or not. We present can wind up in the pockets of private shareholders, available data, and then form our own admittedly inevitably raising the specter of such owners making intuitive judgment about the reasonableness of the extreme returns on their investment by charging general picture appearing from that data. Of course, abusive interest rates to poor borrowers who have readers will apply their own criteria or intuition to little bargaining power because their other credit the data in judging whether rates or profits strike options are limited. them as "abusive," "exploitative," "excessive," "unreasonable," etc. A firestorm of controversy erupted in April 2007 when shareholders of Compartamos, a Mexican MFI Some MFIs are charging their clients rates that seem with a banking license, sold a part of their shares hard to justify from a development perspective. in a public offering at an astonishingly high price, Rosenberg (2007) argues that this was the case at which made some of the individual sellers instant Compartamos, at least after it became able to fund millionaires. One important reason for the high price expansion of its services from other sources besides was that Compartamos was charging its clients very retained profits.3 But are these rare exceptions, or do high interest rates and making very high profits. they represent a pervasive problem in the industry? The annualized interest rate on loans was above 85 percent (not including a 15 percent tax paid by We approach the question from several perspectives. clients), producing an annual return of 55 percent on In the first section, we report on how high microcredit shareholders' equity (Rosenberg 2007). rates actually are around the world. Then we look at how those rates compare with the cost of other In fact, most MFIs charge interest rates well below forms of credit often available to low-income people, those that provoked controversy in the case of including consumer credit, credit unions, and informal Compartamos. But the story tapped into a deep well moneylenders. The section closes with a look at of concern about high microcredit interest rates and trends: are microcredit rates moving up or down? the trend toward commercialization of microfinance. In the second section we "deconstruct" interest This paper asks whether microcredit rates are rates by looking at what they fund. Mathematically, abusively high. Obviously there can be no one-size- an MFI's interest yield is equal to the sum of costs fits-all answer to this question, not only because there and profit on its loan portfolio. Most people would are huge variations in the interest rates and related agree that it is fair to criticize an MFI's interest rates circumstances of individual MFIs around the world, as unreasonable only if its profit or some controllable but also because there is no agreed standard for what element of its costs is unreasonable. In addition to is abusive. There is an intense dispute about how high profits, we analyze MFIs' cost of funds, loan loss interest rates and profits would have to be to qualify expenses, and operating (i.e., administrative) costs. 3 On the other hand, some defenders of Compartamos argue that its example--including its high profits--have benefited potential borrowers by supercharging the expansion of microcredit services in Mexico. 3 The third section briefly considers the question of sustainable MFIs that reported their results to MIX whether we can rely on competition to bring down for 2006.6 Why only sustainable MFIs? The reason interest rates and profits. is that much of the analysis in this paper depends on relationships between interest rates and costs. The final section summarizes the findings and our We usually exclude unsustainable MFIs because their conclusion that, despite occasional exceptions, MFI interest rates are not constrained by their costs--that interest rates generally seem quite reasonable and is, an unsustainable MFI can set its interest rates as that there is no evidence of any widespread pattern low as it wants no matter how high its costs are, of abuse. as long as some donor or government is willing to provide the subsidy necessary to cover the losses. A Note on MFI Data. Financial information on MFIs Not surprisingly, sustainable MFIs tend to charge is drawn from the databases of the Microfinance higher interest rates than unsustainable MFIs. The Information Exchange (MIX). Not all MFIs report to average interest yield (weighted by loan portfolio) MIX, but those that do (currently over 1,400 MFIs) for MFIs reporting to MIX in 2006 was 28.1 percent account for over 58 million borrowers worldwide in for sustainable MFIs, compared with 20.5 percent 98 countries.4 MIX maintains two different, though for unsustainable MFIs.7 Thus, if we had included overlapping, MFI databases. The MicroBanking unsustainable MFIs in our analysis, the interest rates Bulletin (MBB) dataset currently includes 890 reported in the paper would have been substantially institutions that report their data confidentially. lower. MBB adjusts the financial information of these MFIs to compensate for the effect of any subsidies they One might assume that sustainable (i.e., profitable) receive and, thus, tries to present a picture of what MFIs are typically for-profit commercial companies, the industry would look like if it had to pay market cost but this is not the case. In fact, almost two-thirds of for all of its resources. MBB organizes MFIs into "peer the 555 sustainable MFIs are NGOs, cooperatives, groups" of institutions with similar characteristics, and public banks, or other not-for-profit organizations. reports overall peer group performance. The other dataset is MIX Market (www.mixmarket.org), which When tracking recent year-to-year trend lines, we publicly reports individual performance of over 1,300 have had to use a smaller data set: all 175 sustainable MFIs, but does not adjust the reported data.5 Most MFIs who reported their data both for 2003 and 2006. MFIs are included in both datasets. These sustainable MFIs are a minority of the MFIs reporting to MIX, but they account for about half of all MFIs that report to MBB but not to MIX Market are the borrowers and microloan portfolio amounts that guaranteed that individually identifiable data about were reported by participating MFIs (both public and their institutions will not be disclosed publicly. Thus, private) each year. We believe this set is large enough we report only aggregate results, and we generally and representative enough to give a meaningful cannot discuss what is happening in particular MFIs. picture of recent industry trends at a worldwide level. For a snapshot of levels at a particular point in time, When analyzing revenues, costs, and profits, we the larger dataset--all 555 sustainable MFIs reporting usually use a large database that includes the 555 in 2006--is obviously more reliable.8 4 Our definition of "microfinance institution" here is somewhat circular: an institution that describes itself as providing "microfinance." Almost all of these institutions make use of the new microcredit methods that have been developed in the past 30 years. It is important to recognize that poor and low-income clients get financial services from many other institutions that usually do not describe themselves, and are not described by others, as doing microfinance. See Christen, Rosenberg, and Jayadeva (2004). 5 In the rest of the paper, the term "MIX MFIs" includes those MFIs that report to MBB, MIX Market, or both. 6 MFIs were classified as sustainable if their adjusted return on assets, or their unadjusted return when adjusted return was not available, was positive. 7 Includes all countries with three or more MFIs reporting. 8 The paper is based on MIX data through 2006, updated as of April 2008. MIX data for 2007 became available in October 2008, too late for inclusion in this paper. 4 In presenting data from these two sets, we do not How High Are Microcredit use simple averages or distributions, because doing Interest Rates? Where so would distort the picture of the industry by giving Are They Moving? as much weight to a tiny MFI with 1,000 clients as to a huge one with a million clients. Instead, we give 2006 Interest Rate Levels more weight to larger MFIs. Usually we weight by the size of the MFIs' gross loan portfolio (GLP), because Most MFIs charge interest rates far below those most of the income, costs, and profits are analyzed that have provoked controversy in the case of as percentages of loan portfolio. Occasionally we also Compartamos. In 2006, the most recent year weight by numbers of clients.9 available, the median interest income for sustainable MFIs in MIX, weighted by GLP, was 26.4 percent of In looking at interest rates, we use interest yield on loans outstanding.11 Figure 1 shows the distribution of GLP, which is the total amount of cash borrowers interest yields, worldwide and by region, compared pay the MFI during a period for interest and loan with Compartamos' interest yield. The regional fees divided by the average outstanding GLP over breakdown consists of East Asia and the Pacific the same period. As a measure of what clients are (EAP), Eastern Europe and Central Asia (EECA), actually paying, this is far more meaningful than an Latin America and the Caribbean (LAC), Middle East MFI's stated interest rate.10 and North Africa (MENA), South Asia (SA), and Sub- Saharan Africa (SSA). Finally, readers may occasionally notice a seeming disparity in data--for example, interest yields, costs, The Compartamos case was striking and prompted a or profits may be somewhat different in different lot of discussion. But it is a serious mistake to assume graphs or tables. There are several reasons for these that Compartamos' interest rates were typical of differences: the industry, or even of a substantial part of the industry. In fact, less than 1 percent of MFI borrowers · Point-in-time data are usually based on all 555 worldwide were paying rates as high as Compartamos sustainable MFIs reporting to MIX for 2006, while was charging. trend-line data are based on a different set--the 175 sustainable MFIs that reported for both 2003 Some MFIs require borrowers to make compulsory and 2006. deposits before they can receive a loan; borrowers · Some figures present average values, which will typically must maintain these deposits during the life differ from the median shown in a graph reporting of the loan. The interest rates borrowers receive on the distribution of individual values. these deposits are well below the rates borrowers pay · The ratios being reported may have different on their loans. The effect of such deposit requirements denominators. For instance, administrative costs is to reduce the net additional cash borrowers realize as a percentage of interest earnings will look from their loans and, thus, to increase the effective much higher than the same administrative costs cost of the loan to them. About one-third of the expressed as a percentage of GLP. sustainable MFIs reporting to MIX for 2006 required 9 Weighted averages will be familiar to most readers, but perhaps not weighted distributions. The concept is that all MFIs are spread out along a line, in order of their interest yield (for example). In a normal unweighted distribution, each MFI occupies the same amount of space along the line, regardless of the MFI's size. In a weighted distribution, each MFI occupies a space proportional to the size of its loan portfolio (for instance). Once this line is assembled, the median is a point halfway along its distance, and the 90th percentile (for instance) is a point one- tenth below the high end of the line. 10 Some MFIs offer multiple loan products that may entail varying levels of cost and profit. The interest yield calculated by MIX in such cases represents combined interest and fee income divided by the combined portfolio. 11 When the interest yields of the individual MFIs are weighted by number of borrowers, the median is 28.7 percent. The average yield weighted by GLP is 28.4 percent. 5 Figure 1: Interest Income as Percentage of Gross Loan Portfolio, 2006 90 80 70 60 Portfolio 50 Loan 40 oss Gr 30 of %20 10 0 WORLD SSA EAP ECA LAC MENA South Compartamos Asia Note: 555 sustainable MFIs reporting to MIX, distribution weighted by gross loan portfolio. The thick horizontal bars represent medians; the top and bottom of the white boxes represent the 75th and 25th percentiles, respectively; and the high and low short bars represent the 95th and 5th percentiles, respectively. such savings deposits, and on average these MFIs Microcredit Rates vs. Other are smaller than the ones that do not use compulsory Small-Loan Rates savings. Data limitations prevent us from calculating the additional cost due to compulsory savings in How expensive is microcredit compared with other individual institutions. credit available to poor and low-income borrowers? Answering this question poses data challenges. From It is important to keep the interest and fees paid to MIX we have good country-by-country information on the MFI in context: they are only part of borrowers' interest rates for a large set of MFIs (using the most total loan costs. Transaction costs can be substantial, recently reported year, as of March 2008). But it has including for instance the time borrowers have to been much more difficult to assemble information spend away from their businesses, their transportation on rates for consumer credit or credit unions; expenses, and the negative impact of delays in the challenges of determining rates for informal receiving loan funds. Because interest charges can moneylenders are even greater. be quantified easily, they tend to receive much more attention than borrowers' transaction costs. In fact, We found small amounts of published country data these transaction costs often represent a greater and supplemented them by canvassing our contacts expense for the borrower than the interest being in various countries.12 We cannot guarantee the charged on the loan. Sometimes borrowers with other accuracy of each piece of information we were given. credit options are willing to pay a higher interest And in some cases, it is hard to be precise about rate to an MFI because the MFI loan entails lower the effective annual rate based on that information. transaction costs (Adams, Graham, and von Pischke However, we think we have enough to start drawing 1984; Meyer and Cuevas 1992; Robinson 2001; Tran a rough general picture, at least with respect to 1998; and Cuevas 1989). consumer credit and informal credit rates. 12 The data permitted comparison of microlending rates to at least one form of alternative credit in 36 countries: Cambodia, Indonesia, the Philippines, and Thailand (EAP); Armenia, Bosnia and Herzegovina, Kyrgyzstan, Mongolia, Romania, Serbia/Montenegro, Tajikistan, and Uzbekistan (EECA); Bolivia, Colombia, Ecuador, Mexico, Peru, and Nicaragua (LAC); Morocco, Jordan, and Egypt (MENA); Nepal, Bangladesh, Sri Lanka, India, and Pakistan (SA); and Cameroon, Congo, Ghana, Ethiopia, Kenya, Madagascar, Senegal, Tanzania, Togo, and Zambia (SSA). 6 Table 1: MFI Rates vs. Consumer Lending Rates Lower than credit card/ Close to credit card/ Higher than credit card/ consumer interest rates consumer rates consumer rates Bangladesh Armenia Cameroon Bolivia Bosnia and Herzegovina Morocco Cambodia Colombia Peru Congo Ecuador Senegal Ethiopia Egypt Tanzania India Ghana Uzbekistan Indonesia Madagascar Zambia Jordan Mexico Kenya Romania Kyrgyzstan Tajikistan Mongolia Nepal Nicaragua Pakistan Philippines Serbia/Montenegro Sri Lanka Togo Thailand Note: Latest available year; MIX data for sustainable MFIs, CGAP research for consumer lending rates. Consumer credit. Table 1 compares loan rates in involve more frequent transactions than microcredit, sustainable MFIs with consumer credit rates in 36 which could offset their cost advantage in terms countries for which we had reports. Consumer credit of borrower evaluation. Installment financing of includes credit cards, installment loans for furniture or merchandise would not seem more transaction- appliances, and other similar loan products. MFI rates intensive than microcredit, and these "hire-purchase" appear lower than consumer credit rates in a majority lenders usually have collateral and the means to sell of these countries; they are higher than consumer it when repossessed. Notwithstanding these factors, credit rates in less than one-fifth of the countries.13 the general picture suggests that microcredit rates are on average lower than consumer rates. In consumer lending, borrowers are typically salaried, so assessing creditworthiness usually can be done Informal credit. We found 34 reports on rates for automatically, using computerized scoring algorithms. informal lending (mainly unregistered moneylenders By contrast, microlending tends to require a more and pawnshops) in 21 countries and the West African labor-intensive relationship between loan officer and Monetary Union. In all of those countries except client. In light of this higher cost, one might expect Ghana, microcredit rates were lower--usually far microcredit rates to average higher than consumer lower--than informal rates. The median informal rate credit rates, rather than lower, as appears to be the reported was 10­25 percent per month. Rates of 5­20 actual case. On the other hand, credit cards may percent per day were reported in five countries. 13 Some of the consumer credit rates reported to us may be lower than the true effective cost of the loans once payment timing and fees-- especially annual fees--are factored in. The rates used for MFIs are actual interest yield on portfolio, which is a more reliable index of actual cash cost to the client. None of the rates we analyzed includes clients' transaction costs, such as time or travel. 7 Both informal credit and microcredit are generally Table 2: MFI Rates vs. Credit Union Rates uncollateralized and are used by lower income Regular credit union Specialized credit union borrowers. But there are big differences, the most loan products: microcredit products: important of which from the customer's perspective Bolivia = Bolivia = is that informal loans are usually available very quickly. It is commonly assumed that the astronomical rates Colombia + Colombia ­ found in informal lending reflect moneylenders' Ecuador + Ecuador =, ­ exploitation of poor borrowers' lack of options and Ethiopia + Peru = weak bargaining power. However, a body of research suggests this characterization is only occasionally India + Philippines = accurate. Usually, the biggest driver of high informal Kenya = rates is the high cost of such lending.14 Mexico + Credit Unions. We were able to compare MFI rates Nepal = with credit union rates in only 10 countries. The Peru + average MFI rate in these countries was usually higher Philippines + than the rates reported to us for regular credit union loans (Table 2). However, in five cases where credit Notes: + MFI rates generally higher than credit union rates. unions also have a special product aimed at lower ­ MFI rates generally lower than credit union rates. = MFI and credit union rates generally in same range. end microcredit clients, MFI rates have tended to be about the same or lower. It is difficult to draw conclusions from these that have been recipients of technical assistance from comparisons with credit unions, for reasons that the World Council of Credit Unions, and that may not go beyond the small sample size. MFIs often claim be typical of credit unions in their country. that they are reaching poorer customers than credit unions are, though there is little statistical evidence Credit unions have been serving some substantial about whether this is the case.15 Other dimensions of number of poor and low-income clients for many the comparison are ambiguous as well--for instance, years. It would be very useful to have research that loan officers in many MFIs spend much of their time gives a clearer picture of how they stack up against in neighborhoods bringing services to clients where MFIs in terms of outreach, loan products, costs, and they live and work, whereas this is much less common interest charges. in credit unions. Nor is it clear what to make of the comparison between MFI rates and the rates credit Are MFI Rates Climbing or Dropping? unions charge for their special microcredit programs, because many of these credit union programs are Microcredit techniques and the institutions that aimed at a low-end village banking clientele, which employ them are relatively new, having sprung up may in fact be poorer than the average clientele of the over the last 20­30 years. The industry is still in a country's MFIs. Finally, most of our credit union data nascent state in most countries. Standard theory come from relatively small groups of credit unions would lead us to expect reductions of costs (driven 14 See, for example, Robinson (2001). 15 Grace (2007) reports 2006 data from 11 Ecuadoran credit unions that are affiliated with the World Council of Credit Unions. Average per capita income of members was in the bottom quartile of the national distribution. A third of the loans were identified as microenterprise loans. Loan balances below $1,000 accounted for only 7.7 percent of the outstanding loan portfolio, but 36.1 percent of the number of active borrowers. 8 Figure 2: Evolution of Bolivian Microcredit Rates, 1992­2007 70 60 MFIs 50 Rate est 40 Inter 30 fective Ef 20 Banks 10 0 Dec-92 Jun-93 Dec-93 Jun-94 Dec-94 Jun-95 Dec-95 Jun-96 Dec-96 Jun-97 Dec-97 Jun-98 Dec-98 Jun-99 Dec-99 Jun-00 Dec-00 Jun-01 Dec-01 Jun-02 Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Year Sources: Gonzalez-Vega, Claudio, and Villafani-Ibarnegaray (2007) and Banking Superintendency data. by learning curve and economies of scale) and prices 2003 and 2006. Figure 3 shows that the interest yield (driven by competition). Have such reductions in fact on their combined loan portfolio has been dropping happened? Bolivia is a frequently cited case where the quite fast over the period everywhere except for South answer is clearly yes: from 1992 to 2007, microcredit Asia, averaging 2.3 percentage points each year.16 interest rates dropped 43 percentage points, from about 60 percent to about 17 percent (Figure 2). A separate analysis for 2000­2005, using a larger During the same period, bank rates dropped only dataset of MFIs, but a somewhat less rigorous about 12 percentage points. methodology, showed even steeper declines--3.4 percent per year. During the same period, loan rates But is the Bolivia case representative? In Bangladesh, for commercial banks in developing and transition another microcredit pioneer, interest rates have not economies fell by about 0.8 percent per year.17 dropped substantially (though rates there started out quite low). What has been the worldwide pattern in Later in this paper we look more closely at the question recent years? of whether competition is driving the decline in rates. But regardless of how it is explained, the substantial To answer these questions, we used data from the set downward trend in rates is certainly encouraging. of 175 sustainable MFIs that reported to MIX in both 16 If interest yield is calculated individually for each MFI and then weighted by number of borrowers, the 2003 worldwide average was 33.0 percent, dropping an average of 1.2 percent a year to 29.3 percent in 2006. 17 Calculated from International Financial Statistics (IMF). 9 Figure 3: Interest Yield Trends, 2003­2006 50 44.3 40 36.0 35.1 34.5 Portfolio 32.6 31.7 30 28.2 28.0 25.5 24.8 26.1 26.2 23.9 Loan 20.9 20 oss Gr of 10 % 0 WORLD Africa EAP ECA LAC MENA South Asia ­2.3% ­1.4% ­3.9% ­1.5% ­0.9% ­3.7% +/­0% 2003 2006 Avg. change per year Note: Total Interest and Fee Income/Average Total GLP: average of 175 sustainable MIX MFIs, weighted by loan portfolio. The Components of whether each of these components that are funded Microcredit Interest Rates from interest income is at a reasonable level. Four main components are reflected in an MFI's Figure 4 shows the relative importance of each of interest rate: cost of funds, loan loss expenses, those elements, expressed as a percentage of after- operating expenses, and profits. In this section, we tax income. Worldwide, operating expenses consume break out these components and discuss how they more than half of income, followed by funding costs, may affect interest rate trends. profits, and loan losses. MFIs use their interest income to cover costs, and Cost of Funds the difference between income and costs is profit (or loss). A simplified version of the relevant formula is The relatively high price MFIs have to pay for money they borrow contributes substantially to the interest Income from loans = Cost of funds + Loan loss they charge borrowers. As of 2006, total funding expense + Operating expense + Profit18,19 cost (interest expense) for 554 sustainable MFIs was equivalent to 8.3 percent of their total average GLP Lowering interest rates would require lowering one for the year (Figure 5). of the four components on the right side of the equation. If we want to judge whether interest rates Figure 6 shows no downward trend in cost of funds are reasonable, the most direct approach is to look at (here expressed as the ratio of interest expense to loan portfolio) for MFIs worldwide. 18 "Operating expense" is the term used by MIX to describe personnel and administrative costs, such as salaries, depreciation, maintenance, etc. 19 The full formula is Income from loans + Other income = Cost of funds + Loan loss expense + Operating expense + Tax +Profit We want to look at costs and profit as percentages of loan portfolio, but taxes and other income do not relate directly to the portfolio. In addition, the current MIX reporting structure does not allow us to derive taxes and other income as separate items: we can calculate them only as a single net figure (i.e., other income ­ taxes). To solve these problems for the purposes of Figure 6, below, we have netted out taxes against other income on the left side of the equation, leaving us with the formula Income from loans + (Other income ­ Taxes) = Cost of funds + Loan loss expense + Operating expense + Profit 10 Figure 4: Costs and Profits as Percentage of Income, 2006 100 80 60 Income of 40 % 20 0 WORLD Africa EAP ECA LAC MENA South Asia Op. Expense Cost of Funds Loan Losses Profit Note: 554 sustainable MFIs reporting to MIX for 2006 weighted by GLP. The Unit Desa system of Bank Rakyat Indonesia is excluded from this calculation because its reported cost of funds reflects artificial pricing arrangements between Unit Desa and the rest of the bank. The income on which the percentages are based is interest yield plus other income minus taxes. Figure 5: Average Cost of Funds (Interest Expense) as Percentage of GLP, 2006 10 8.3 8.6 8.5 8.6 7.7 8 6.1 Portfolio 6 5.1 Loan 4 oss Gr of 2 % 0 WORLD Africa EAP ECA LAC MENA South Asia Note: 554 sustainable MFIs (without BRI) reporting to MIX, weighted by GLP. Compared with commercial banks, MFIs tend to be 2006, compared with about 3 percent for commercial less leveraged--that is, less of their portfolio and banks in the same countries.20 other assets is funded by liabilities on which they have to pay interest. Figure 7 compares MFIs' interest Are MFI managers prejudicing their borrowers by expense with their liabilities rather than with their incurring unnecessarily high funding costs? In general, loan portfolio, and shows that their borrowings have it would seem unfair to criticize MFI managers much been relatively expensive--averaging 5.1 percent in on this score, because even though they can control 20 Commercial bank computation by Christoph Kneiding based on the most recent BankScope data for countries that have MFIs reporting to MIX, weighted by liabilities. 11 Figure 6: Trend in Funding Costs (Interest Expense as Percentage of GLP), 2003­2006 10 9.4 8.9 8.4 8.0 8.2 8 7.5 7.6 6.8 6.5 5.9 6 4.8 4.4 Liabilities 4 3.3 3.3 of % 2 0 WORLD Africa EAP ECA LAC MENA South Asia 0.2% 0.4% 0.3% 0.6% ­0.2% 0.5% 0.6% 2003 2006 Avg. change per year Note: MIX data for 174 sustainable MFIs reporting in both years. BRI has been excluded because its reported cost of funds is artificial. Figure 7: MFI Interest Expense as Percentage of Liabilities, 2006 6 5.5 5.7 5.1 5 4.8 5.0 4 3.1 3 Liabilities 2.6 of % 2 1 0 WORLD Africa EAP ECA LAC MENA South Asia Note: 554 sustainable MIX MFIs, weighted by liabilities their own operations, they usually have little control Loan Loss Expenses over their rate of borrowing/funding costs. Most of them get funding where they can find it and tend Loan losses due to borrower default have relatively to be price takers rather than price makers when little effect on MFI interest rates, for the simple reason it comes to the interest rates they pay. Increasing that such losses are quite low in most MFIs, especially reliance on deposit funding will lower costs over the in large ones that account for most of the market. longer term, as regulators authorize more MFIs to Figure 8 shows the global and regional medians for take savings. However, this option is unavailable to loan losses, weighted by portfolio. managers of MFIs if their country does not have the enabling regulation, or if their business is not yet solid As a point of reference, the general rule of thumb in enough to meet the hurdle for depository licensing. microcredit is that annual loan losses of more than 12 Figure 8. Global and Regional Loan Losses as Percentage of GLP, Averages for 2006 6 4.9 5 Portfolio 4 3.0 Loan 3 2.3 oss 2 Gr 1.5 1.5 1.2 of 1.0 % 1 0 WORLD Africa EAP ECA LAC MENA South Asia Note: Data from 555 sustainable MFIs reporting to MIX for 2006, weighted by gross loan portfolio. about 5 percent tend to become unsustainable. loan loss rates can go too low. An MFI that has no Above that level, loan collection must be improved loan losses at all is probably being too risk-averse quickly and substantially or it will spin out of control. in its selection of borrowers, which hurts not only MFIs usually have delinquency and default rates well the expansion of poor people's access to finance below those of commercial banks in their countries. but also the MFI's own profitability. If there is any (Interestingly, emerging evidence also suggests that widespread abuse in microcredit interest rates today, MFIs are more stable than banks when it comes to it certainly cannot be traced back to excessive loan the effect of general economic stress on their loan loss expense. collection [Gonzalez 2007].) Operating Expenses (Efficiency) The high average loan loss rate for Africa (4.9 percent) is driven by a few outliers.21 [Compared to other topics in this paper, analysis of whether MFI operating expenses are "reasonable" is Figure 9 shows 2003­2006 trends in loan losses. relatively complex. The following is a brief summary. Regional trends vary, but loan losses measured Readers who want a more detailed discussion, globally are down slightly. Further improvement including data and sources, should refer to the might be possible, but there seems to be little room Appendix at the back of this paper, and then return for an improvement big enough to have a substantial to the section on profit beginning on page 15.] effect on the interest rates clients have to pay. Indeed, 21 The 2006 distribution of loan losses, weighted by GLP, was Region 5% 25% Median 75% 95% Africa 0.3 0.7 2.3 3.9 22.4 EAP 0.1 0.9 0.9 0.9 1.1 ECA 0.3 0.8 1.6 2.0 2.7 LAC 0.8 1.2 2.2 3.6 10.0 MENA 0.0 0.2 1.8 1.8 2.0 SA 0.0 0.4 1.7 2.4 2.7 World 0.2 0.9 1.6 2.7 5.7 13 Figure 9: Annual Loan Losses as Percentage of GLP, 2003­2006 4 3.7 3.5 3.4 3.1 3 Portfolio 2.5 2.2 2.3 1.9 1.9 Loan 2 1.6 1.5 1.5 oss 1.5 1.3 1.1 Gr 1 0.9 of 0.5 0.3 % 0 WORLD Africa EAP ECA LAC MENA South Asia 2003 2006 Note: 175 sustainable MIX MFIs. The majority of MFIs' interest income goes to pay operations, including many factors that aren't part operating costs (salaries and other administrative of any public database. Nevertheless, there are a costs), which are about 60 percent of total MFI costs. number of considerations that shed light on issues The worldwide median operating expense was 11.4 surrounding operating costs. percent of GLP in 2006. Ninety percent of the values lie between 7.9 percent and 33.7 percent. The range Effect of small loan sizes. The principal justification is considerably narrower if Africa is excluded. offered for high microcredit interest rates is the claim that administrative costs are inevitably higher when Are MFIs reasonably efficient, or do their operating placing a given amount in many tiny loans rather expenses include substantial waste? Unfortunately, than in a few big loans. This claim sounds plausible one cannot calculate an answer from available in theory and is confirmed by the data. Regression performance statistics. MFIs vary widely in a range analysis shows a strong inverse relationship between of factors that affect operating costs, including not loan size and operating expense, even after screening only loan sizes but also age, scale, client location and out the effects of other variables, such as age, scale, density, type of loans provided (e.g., group versus productivity, legal status, savings mobilization, region, individual lending), client stability, communication macroeconomic environment, and some proxies for and transport infrastructure, salary levels, and rural physical infrastructure. versus urban location. MIX categorizes MFIs into various peer groups for rough comparison purposes, Economies of scale. Even though small loans cost but each peer group contains so wide a range of more to administer than big ones, we might look circumstances that one cannot confidently judge an for those costs to be lowered by economies of MFI's efficiency just by comparing its indicators with scale as MFIs grow larger. Some observers express those of its peer group, let alone judge whether the disappointment that the growth in size of MFIs has peer group as a whole is reasonably efficient. not improved efficiency as much as they expected: shouldn't an MFI that can spread its fixed operating Currently at least, the only reliable way to tell whether costs over a million borrowers have a much lower an MFI's operating costs are appropriate is to conduct cost per loan, and be able to charge a lower interest an on-the-ground study of its individual situation and rate, than an MFI with only a few thousand clients? 14 Regression analysis produces a surprising answer: We do not have a good statistical proxy for the MFIs appear to capture most of their scale benefits competitiveness of an MFI's market, so we cannot by the time they reach about 2,000 clients. They gain quantify how much of this age effect can be attributed relatively little from scale economies after that very to competition. However, it's likely that most of the early point. This is probably because microcredit is effect is due to the learning curve, because relatively so labor intensive: salaries make up the majority of few MFI markets are competitive yet. most MFIs' operating expenses, and fixed costs are relatively low compared with variable costs. MFIs that Loan size and mission drift. Since smaller loans are are still small enough to reap major economies of associated with higher operating expenses, an MFI scale account for only a tiny percentage of microcredit could reduce its operating expense ratio by simply loans and customers. In short, economies of scale making larger loans. If an MFI whose borrowers cannot do much to offset the added expense that all have $100 loans suddenly lets these borrowers comes from making very small microcredit loans. double their loan size, or finds new borrowers who want larger loans, the MFI does not need to double Trend of operating costs. Analysis of the set of 175 its staff or other administrative expenses. Is the sustainable MFIs that reported to MIX for both 2003 downward trend in MFI costs nothing more than the and 2006 reveals good news. Worldwide, the ratio of result of increasing loan sizes, which might reflect a operating expense to loan portfolio declined about 1 movement toward serving a richer clientele?22 percentage point per year, from 15.6 percent in 2003 to 12.7 percent in 2006. This pattern held for all regions Up to this point we have analyzed operating expense except South Asia, where operating costs were already divided by loan portfolio. Of two MFIs with equally quite low in 2003. What is causing this improvement in competent and efficient management, the one efficiency, and can we expect it to continue? with larger loans would automatically look better using this measure. We can avoid this distortion Learning curve and competition. One driver-- by looking at operating expense divided by the perhaps the most important driver--of improved number of borrowers, thus removing loan size from MFI efficiency seems to be the increasing age of the calculation. This "cost per borrower" shows the large MFIs as the industry develops. As institutions same downward trend from 2003 to 2006. What we mature, they learn their business better and are able are seeing is real efficiency improvement, not just a to control costs more effectively. Regression analysis statistical result produced by larger loans. shows that the age of an MFI is strongly associated with lower operating costs, even after separating Still, these analyses don't tell us how much avoidable out the effects of loan size, scale, and other relevant "fat" is built into MFI operating expenses. Because variables. Not surprisingly, the effect weakens over most microfinance markets are so young and because time, but it continues for quite a while. Operating most are not yet competitive, it is unrealistic to expect cost as a percentage of loan portfolio tends to drop MFIs in those markets to be operating at the most by 2­8 percentage points for each of the first six efficient levels possible. Immature industries always years, 1­2 points for each of the next five years, and have some level of correctable inefficiency, and it is less than 1 point for each year thereafter. not easy to think of a reason why microfinance should be any different. We know of no evidence suggesting 22 Some observers interpret growth in average loan sizes as a sign of "mission drift" away from poor clients, but it is far from clear how often this is valid. The link between loan size and client poverty is only a very rough one at best. Most MFIs have a sequential ladder of loan sizes for clients, and the very small loans at the beginning of that ladder often reflect the MFI's risk management policy rather than the actual needs or repayment capacity of the borrowers. When borrowers move into later and larger loans, or when the MFI relaxes its size limits on initial loans, the MFI's average loan size will climb even if there has been no change in the kind of client it is serving. And even where the MFI is adding better off clients (i.e., small business operators) as it grows, that does not necessarily mean restricting service for poorer customers. A much more reliable way to judge mission drift is to look at the character of the villages, towns, and neighborhoods where the MFI is opening its new branches. 15 Figure 10: Returns on Average Assets and Equity--MFIs vs. Banks 20 17.7 18 16 MFIs 14 13.0 Banks 12 10 cent 8 Per 6 4 2.1 2 1.4 0 ROAA ROAE Note: Asset-weighted averages of all MIX MFIs with 2006 data available, except BRI Unit Desa, whose reported equity is artificially low because of its relationship to the larger bank. Most MFIs in the sample report to MBB, where their returns are adjusted to compensate for the effects of subsidy. Bank data by Christoph Kneiding, based on the most recent year available in Bankscope, for countries that have MFIs reporting to MIX. that MFIs in general are out of line with the normal little competition so far. Standard economic theory evolution of efficiency for businesses in immature predicts that profits will be higher in such markets markets. And the pronounced downward trend in than in more developed markets where competition costs is highly encouraging. constrains prices. Figure 10 compares MFI profitability with bank profitability, measured by both return on Profit assets and return on equity. When profit is measured against assets, it does seem to average higher for Of the four components of microcredit interest rates, MFIs than it does for banks, most of which face more profit is the one that is most obviously subject to competition than the MFIs in their countries do. management control. It is also the most controversial. Some observers are uncomfortable with the notion of But compared with MFIs, banks can leverage their private parties making any profit from microlending, capital structure more: that is, they fund more of which they view as a service to poor people, and not their assets with other people's money--deposits and as a business opportunity. Others accept the idea of other borrowings--rather than with their own equity. private profits in microlending, but are concerned that As a result MFIs, despite their higher return on assets, MFIs will exploit the weak bargaining position of their do considerably less well than banks in producing borrowers to extract abusive levels of profit. Still others returns for their owners. Return on equity averages think that high profits in the early stages are a positive about 5 percent lower for MFIs than for banks. (Note good, because high returns will attract more investment that we are including all MIX MFIs here, not just the and more rapid outreach of services to people who need profitable ones, because profitability is the variable them, and because they are confident that competition being examined, and MFIs are being compared will eventually moderate those profits. against all banks, including the unprofitable ones.)23 How high are MFI profits? In most countries, Figure 11 gives the distribution of return on equity the microcredit market is still immature, with low for all MIX MFIs. Overall, the returns are moderate, penetration of the potential clientele by MFIs and at least by commercial standards. However, some of 23 Including the unprofitable MFIs here does not affect the results as much as one might think. Over half of MIX MFIs are unprofitable, but they account for only about a fifth of the world loan portfolio or world assets, which reduces their influence in an average or distribution that is weighted by those variables. 16 Figure 11: Distribution of MFI Returns on Average Equity, 2006 150 125 100 earY 75 50 during 25 0 Equity ­25 verage A ­50 of ­75 % ­100 ­125 WORLD SSA EAP ECA LAC MENA South Asia Note: All MIX MFIs with 2006 data available, except BRI Unit Desa, weighted by GLP. The thick horizontal bars represent medians; the top and bottom of the white boxes represent the 75th and 25th percentiles, respectively; and the high and low short bars represent the 95th and 5th percentiles, respectively. the profits look quite high in the upper percentiles. (i.e., at or above the 90th percentile, which is not Further analysis of these most profitable MFIs sheds shown in the figure) produced 2006 returns on equity some light. above 35 percent. Over two-thirds of the MFIs with these high returns were not-for-profit organizations. To begin with, the top 5 percent of the distribution Absent illegal manipulation, net earnings of NGOs in South Asia seems to be extremely profitable.24 do not go into private pockets, but remain in the However, the very high figure shown there for return organizations to fund further expansion of financial on equity is driven by a few Indian MFIs with very or other services for their target clientele. One of odd balance sheets: almost no equity and therefore the very-high-profit MFIs is a large government extremely high leverage. As the denominator of operation. Profits of government MFIs are available the ratio (equity) gets close to zero, the calculation to expand outreach or to fund other government produces very high numbers, even if profits in relation priorities. Such profits would be captured by private to loan portfolio or assets are moderate. The return pockets only in cases of corruption, and we know on assets at the 95th percentile in South Asia is 11.2 of no reason to think that microfinance-generated percent, which is similar to the comparable figure for profits would be more subject to corruption than any other regions. other government revenues. Turning from South Asia back to the worldwide The remaining third of these most profitable MFIs distribution summarized in Figure 11, we find that the are organized as for-profit business corporations. most profitable 10 percent of worldwide loan portfolio Most of them started out as not-for-profit NGOs and 24 The 127 percent figure here represents, not the most profitable 5 percent of the MFIs in South Asia, but rather the most profitable 5 percent of the loan portfolios in South Asia. The large losses in the lower percentiles for EAP is driven by a single huge government MFI--the Vietnam Bank for Social Policies. 17 transformed to qualify for a deposit-taking license, the cachet of microfinance, but have no intention of or for some other reason. In such cases, the NGO investing in any MFI whose risk-return profile is not usually retains some significant portion of the shares, competitive from a purely commercial standpoint. and the profits accruing to those shares stay in the An unpublished 2008 CGAP study found that the NGO to fund its work. Finally, some significant part of composition of MIV funding has been changing in the shares in the for-profit MFIs is owned by private recent years. The majority of MIV money is now individuals or companies. Profits on those privately coming from investors at or near the fully commercial owned shares--and only those profits--make their end of the spectrum. MIV investments are heavily way sooner or later into private pockets. Again, the concentrated in the largest MFIs. Three quarters high profits accruing to private owners of MFIs in of MIV money goes into loans to MFIs, which are, the top 5 or 10 percent of the profit distribution are of course, less risky than the other quarter that is exceptional rather than typical in the microfinance invested in equity. industry. But they are not insignificant, and they do raise a question of appropriateness. What kind of profits are MIVs making? Assembling data is not easy, but the CGAP study was able to Some observers think it is immoral for private parties analyze returns for a set of MIVs that were organized to take a profit, or anything but a minimal profit, in 2002. The average annual return reported for debt out of services to the poor (a view that has been investments has been 6.3 percent. The reported return expressed by Nobel laureate Muhammad Yunus). on riskier equity investments has been 12.5 percent. To them, those high profits going to private parties Both figures reflect gross returns; they do not include are abusive by definition, as are most of the lower the effect of the funds' administrative costs, which profits reported in our data, at least to the extent that tend to range from 2 to 6 percent, mostly toward the they're being captured privately. lower end. These returns are respectable but far from spectacular, especially considering the relatively brief Other observers think that commercial investment, track record of MFIs in most countries, and the fact which requires returns at least as high as those found that most MIV investments face substantial country in ordinary banks, will enable microfinance services to risk, including currency and political risks. reach a greater number of poor clients faster. From this perspective, the appropriateness of high-end Are MFI profits rising or falling? Figure 12 presents profits in this minority of cases can be addressed only weighted average after tax profits (interest yield + by investigating individual circumstances, including other income ­ expenses ­ taxes) the same way that most importantly the risk investors took on when they earlier figures have shown funding costs, loan losses, put in their capital without knowing what the results and operating costs--namely, as a percentage of loan would be. CGAP has published such a review for portfolio for the 175 sustainable MFIs that reported Compartamos (Rosenberg 2007). their performance to MIX Market/MBB for both 2003 and 2006. There has been a substantial downward In the past few years there has been a proliferation trend--0.6 percentage points (about one-tenth of of international investment funds (microfinance their value) each year. investment vehicles--MIVs) that lend to MFIs or invest in their shares. The objectives of MIV investors cover a What is the impact of profits on interest rates? It spectrum. At one end are public development finance is axiomatic that if MFIs were to shrink their profits, institutions whose core objective is development, and they could charge their clients lower interest rates. who typically are willing to accept higher risks or But how much lower? Figure 13 looks at net MFI lower returns than a commercial investor would. At profits (including profits from other activities besides the other end are commercial actors who may like lending) in relation to the interest they collect from 18 Figure 12: After-Tax Profit as Percentage of Average GLP, 2003­2006 12 11.4 10.3 9.8 oss 10 Gr 8 7.7 6.3 6.3 6.0 5.9 Portfolio 6 5.25.0 5.0 5.1 average 4.0 4 of Loan 2.9 % 2 0 WORLD Africa EAP ECA LAC MENA South Asia ­0.6% ­0.1% ­0.5% ­0.4% ­0.3% ­1.7% +/­0% 2003 2006 Avg. change per year Note: GLP-weighted average of 175 sustainable MIX MFIs. borrowers. It tells us how much an MFI could reduce on investments), compared with 7.5 percent for the the interest it charges if it completely eliminated other three quartiles. Lowering interest rates enough all profit. This is, of course, a drastic and unrealistic to eliminate all profit would mean that these MFIs scenario. The MFI would be forgoing not only returns would be subsidizing losses on their loan portfolio to its owners but also growth in equity capital to with the net income from other activities. expand the business. Competition Figure 13 shows that an MFI at the median could reduce its interest rate by 17 percent of the interest People who are enthusiastic about the rate (not 17 percent of the loan amount). In other commercialization of microfinance have sometimes words, completely eliminating all profit would reduce taken it as an article of faith that markets will eventually the median MFI's interest rate by only about one- approach saturation, at which point competition will sixth, an effect that is smaller than many people might put downward pressure on interest rates, forcing MFIs expect. By way of comparison, the hypothetical interest to lower their profits and become more efficient. But reduction by eliminating all profit is less than the drop it is not guaranteed that such effects will always occur. in rates that actually happened from 2003 to 2006. If For instance, credit card rates in the oversaturated all sustainable MFIs swore off profits tomorrow, they U.S. market have proved stickier than other credit would still have to charge interest rates that might look rates, probably due in considerable part to the abusive to those who don't understand the high costs substantial time investment required for a customer to that tiny lending inevitably entails. search for and switch to a new card with better terms (Calem and Mester 1995 and Ausubel 1991). Also, Cutting out profit would have more substantial lenders might compete for customers by increasing effects at the upper percentiles, especially the top their advertising, or enhancing service quality, instead 5 percent, where interest rates could be shaved by of lowering their interest rates (Bertrand et al. 2005 almost two-thirds. One of the things going on here is and Wright and Alamgir 2004). that MFIs above the 75th percentile get a lot more of their income from sources other than lending. Over It is widely thought that microloan customers are 20 percent of their income comes from nonlending not very sensitive to interest rate changes. If true, activities (mainly other financial services and returns this would reduce competition's downward effect on 19 Figure 13: MFI Net Income as indeed declined steeply in Bolivia, but considerably Percentage of Interest Yield, 2006 less in Bangladesh. Porteous reported that the rate reductions that had occurred there resulted more 70 from political pressure than from competition. ieldY 60 Interest rates, profits, and operating costs started out est much lower in Bangladesh than was typical elsewhere 50 in the world, probably mainly due to the attitudes of 40 pioneers there toward the business. Thus, there was less room for reductions than in some other places. Income/Inter Nevertheless, some large Bangladesh MFIs are now est 30 generating profits that are high enough to suggest Inter 20 room for further interest rate reduction.25 Net otalT10 It is unclear why competition isn't squeezing profits in Bangladesh more seriously. One possibility is 0 that many or even most borrowers with more than WORLD one option may want to borrow more than any single MFI will lend them, and so they take multiple Note: 555 sustainable MIX MFIs, weighted by loan portfolio. The thick horizontal bar represents medians; the top and bottom of the loans wherever they can get them, even if one of white box represent the 75th and 25th percentiles, respectively; and the high and low short bars represent the 95th and 5th percentiles, the providers is more expensive than another. MFI respectively. managers in Bangladesh report high levels of multiple indebtedness. Another theoretical explanation would rates, profits, and costs. But two recent studies have be implicit collusion among the few MFIs that occupy found considerable price sensitivity among customers most of the market. of a Bangladesh microlender and a South African consumer finance provider (Dehejia, Montgomery, Blaine Stephens (2007) reviewed 2003­2005 trends in and Morduch 2005 and Karlan and Zinman 2007). four competitive microcredit markets--Bolivia, Bosnia, Three other CGAP-commissioned studies are testing Morocco, and Peru--and found interest rates and interest rate sensitivity in other markets. The Gates operating expenses had dropped in parallel each year Foundation is also investing in such research. in all four markets. As the market in Cambodia has become more competitive, interest rates there have David Porteous (2006) has analyzed microcredit dropped by about a half between 2000 and 2007, competition and its effects in Boliva, Uganda, and according to Eric Duflos, a senior microfinance analyst Bangladesh. At the time of his study (2005), interest from CGAP.26 A forthcoming MIX Benchmarking rates had not yet declined much in Uganda, but report for Mexico, where competition has set in quite Porteous judged that the microcredit industry was recently, shows modest declines in interest rates and still in a consolidation phase there and thus that profits during 2007. Knowledgeable observers tell classical competition theory would not yet predict us that microfinance is becoming competitive in price competition. Markets in Bolivia and Bangladesh important markets in India and that it has resulted in were more highly saturated. Interest rates had downward pressure on interest rates. 25 Return on average equity for 2006 was 26.1 percent at ASA, 23.3 percent at BRAC, and 22.2 percent at Grameen. Return on average assets was 14.4 percent for ASA, 6.9 percent for BRAC, and 2.4 percent for the more heavily leveraged Grameen (MIX Market data). ASA and BRAC are launching major efforts to move into other countries, and we speculate that they are probably using their profits to finance this major expansion of services. 26 According to unweighted MIX data, median MFI interest rates in Cambodia dropped from 42 percent in 2003 to 32 percent in 2006. 20 Taken together, these data points suggest that · MFI rates were typically higher than credit union rates competition may not inevitably produce lower interest in the 10 countries for which we found data. In the rates, profits, and operating expenses in all markets, cases where the credit unions offered a specialized but that such effects do appear to be happening in microcredit product, their interest charges tended most of the markets now regarded as competitive. to be the same as, or higher than, prevailing MFI Present data suggest an optimistic picture, but it is rates. However, it is hard to make much of this still too soon for any robust prediction about how information, not only because the sample size is so universal the lower-interest-rates-through-competition small, but also because we know so little about the scenario will be. Perhaps the more relevant fact is comparability of customers and products. the substantial worldwide decline in interest rates, administrative costs, and profits that we observed Cost of Funds earlier in the paper. All this is distinctly positive for borrowers, who may not care very much whether we · MFIs have to pay more than banks pay when they attribute it to competition or some other factor. leverage their equity with liabilities, and their cost of funds as a percentage of loan portfolio showed Summary and Conclusions no sign of dropping 2003­2006. But MFI managers don't usually have much control over these costs, The question we have tried to address is whether in the medium term at least. microcredit borrowers are being abused by unreasonably high interest rates. Here are what Loan Losses we see as the highlights of the evidence we have assembled. · MFI interest rates are not being inflated by unreasonable loan losses. In fact, default rates are Level of Interest Rates very low--about 1.9 percent in 2006. · Using the best data available, the median interest Administrative Expenses rate for sustainable (i.e., profitable) MFIs was about 26 percent in 2006. The 85 percent interest rates · Tiny loans require higher administrative expenses, that drew so much attention to the Mexican MFI which are not substantially offset by economies Compartamos are truly exceptional, rather than of scale. On the other hand, the learning curve of representative of the industry. MFIs as they age produces substantial reductions. · MFI interest rates have been declining by 2.3 · Administrative costs are the largest single percentage points a year since 2003, much faster contributor to interest rates, but they have been than bank rates. declining by 1 percentage point per year since 2003. This decline appears to be a true improvement Comparison with Other Rates in the cost of serving each borrower, not just the Paid by Low-Income Borrowers result of expanding loan sizes. · We have no statistical way to quantify how much · MFI rates are significantly lower than consumer and avoidable fat remains to be trimmed from MFI credit card rates in most of the 36 countries for which operating costs. Given the finding that the level of we had rate indications, and significantly higher these costs is strongly related to the age of the MFI, than those rates in only a fifth of the countries. it would be unrealistic not to expect substantial · Based on 34 reports from 21 countries, MFI rates inefficiency at a time when most MFIs are relatively were almost always lower--usually vastly lower-- young, and when most national microfinance than rates charged by informal lenders. markets are immature and noncompetitive. We are 21 unaware of any evidence to suggest that MFIs in to politicians and the public, neither of whom general are out of line with the normal evolution of usually understand the high costs that tiny lending efficiency for businesses in such markets. inevitably entails. Profits Competition · MFIs on average have higher returns on assets · One cannot assume that competition will always than commercial banks do, but MFIs produce lower interest rates. Interest rates appear to have considerably lower returns on equity for their dropped in the markets where microcredit has owners. The average return on MFI owners' equity already become competitive, except for Bangladesh. in 2006 was moderate--12.3 percent, compared But it is still too early to make any robust prediction with 17.7 percent for banks. The very high profits about how universal the lower-interest-rates- that have drawn so much attention to Compartamos through-competition scenario will be. are outliers, not at all typical of the industry. · Whatever the role that competition plays, the · At the same time, the most profitable 10 percent important fact is that interest rates, profits, of the worldwide microcredit portfolio produced and administrative costs have shown a marked returns on equity above 34 percent in 2006, a level downward trend in recent years. that is no doubt high enough to raise concerns about appropriateness for some observers. Much How all this information is put together is up to each of this profit is captured by NGOs and never reader. We approach the issue from a development reaches private pockets. But some of it does go to perspective, where the main concern is not financial private investors. A judgment about whether such results but rather client benefit--including, of course, profits are "abusive" would depend not only on those future clients who will get access to financial the observer's standard for what is a reasonable services as new investment expands the outreach profit but also on investigation of individual MFI of MFIs. A few MFIs have charged their borrowers circumstances, including the risk levels faced by interest rates that may be considerably higher than investors when they committed their funds. what would make sense from this perspective. Indeed, · The burgeoning volume of money passing through it would be astonishing if this were not the case, international microfinance investment funds is given the diversity of the industry and the scarcity of coming mainly from investors who are not willing competitive markets. to accept higher risks or lower returns for the sake of social objectives. Yet the profits generated by The real question is whether unreasonable MFI these funds seem unimpressive so far. lending rates are more than occasional exceptions. · Profits of sustainable MFIs, measured as a We do not find evidence suggesting any widespread percentage of loan portfolio, have been dropping pattern of borrower exploitation by abusive MFI by about one-tenth (0.6 percentage points) per interest rates. We do find strong empirical support year since 2003. for the proposition that operating costs are much · Profits are not a predominant driver of interest higher for tiny microloans than for normal bank loans, rates. For the median MFI, the extreme and so sustainable interest rates for microloans have to be unrealistic scenario of complete elimination of significantly higher than normal bank interest rates. all profit would cause its interest rate to drop by We are encouraged by the rapid decline in interest only about one-sixth. Such an interest reduction rates, operating costs, and profits in recent years, would not be insignificant, but it would still leave and we would expect this trend to continue in the microcredit rates at levels that might look abusive medium-term future. 22 Appendix. Operating Expenses mobilization, region, macroeconomic environment, and some proxies for physical infrastructure. Operating expenses constitute the majority of MFI costs. As shown in Figure A-1, the worldwide median It is clear that loan administration will cost MFIs more for operating expense ratio (OER--salaries and other than it costs banks that make much larger loans. But administrative costs divided by GLP) was 11.4 percent is the actual level of operating costs for most MFIs for 2006.27 the inevitable result of loan size, or does it reflect unreasonable inefficiency? One cannot calculate an Effect of small loan sizes. The principal justification answer to this question from available performance offered for high microcredit interest rates is the claim statistics. MFIs vary widely in a range of factors that that administrative costs are inevitably higher when affect operating costs, including not only loan sizes placing a given amount in many tiny loans rather than but also age, scale, client location and density, type a few big loans. This claim sounds plausible enough of loans provided (e.g., group versus individual in theory, and Figure A-2 suggests that it stands up lending), client stability, communication and transport empirically. This figure shows the results of a regression infrastructure, salary levels, and rural versus urban analysis testing the relationship between average location. MIX categorizes MFIs into various peer loan size and OER, both measured as percentage groups for rough comparison purposes, but each of gross national income per capita (GNIPC). The peer group contains so wide a range of circumstances regression screened out the effect of other variables, that one cannot confidently judge an MFI's efficiency such as age, scale, productivity, legal status, savings just by comparing its indicators with those of its peer Figure A-1: Distribution of Operating Expense Ratio, 2006 90 80 70 60 Portfolio 50 Loan 40 oss Gr of 30 % 20 10 0 WORLD SSA EAP ECA LAC MENA South Asia Note: 555 Sustainable MIX MFIs, weighted by GLP. The thick horizontal bars represent medians; the top and bottom of the white boxes represent the 75th and 25th percentiles, respectively; and the high and low short bars represent the 95th and 5th percentiles, respectively. 27 This paper's version of the operating expense ratio (operating expense/GLP) is different from the one used by MIX (operating expense/total assets). 23 Figure A-2: Predicted Change in Operating Expense Ratio, by Loan Size 70 60 50 40 OER 30 20 10 0 0 20 40 60 80 100 120 140 160 Loan size/GNIPC Note: Regression results based on 1,144 MIX Market and MBB MFIs reporting data for 1999­2006. The percentage scale at the left of the figure measures relative change, but does not correspond to absolute operating cost levels (Gonzalez Forthcoming). group, let alone judge whether the peer group as a for those costs to be lowered by economies of whole is reasonably efficient. scale as MFIs grow larger. Some observers express disappointment that the growth in size of MFIs has The only reliable way to tell whether an MFI's not improved efficiency as much as they expected: operating costs are appropriate is to conduct an shouldn't an MFI that can spread its fixed operating on-the-ground study of its individual situation and costs over a million borrowers have a much lower operations, including many factors that aren't part cost per loan, and be able to charge a lower interest of any public database. Nevertheless, there are a rate, than an MFI with only a few thousand clients? number of considerations that shed light on issues The regression result in Figure A-3 is surprising. MFIs surrounding operating costs. appear to capture most of their scale benefits by the time they reach about 2,000 clients; they appear Economies of scale. Even though small loans cost to gain relatively little from scale economies after more to administer than big ones, we might look that very early point. This is probably because Figure A-3: Predicted Change in Operating Expense/GLP Ratio, by Scale 70 65.0 60 55.2 53.6 52.2 53.0 50 40 OER 30 20 10 0 1st (0­2k) 2nd (2­5k) 3rd (5­10K) 4th (10­30K) 5th (30K +) Quintiles defined by number of borrowers Note: Regression results based on 1,144 MIX/MBB MFIs reporting data for 1999­2006 (Gonzalez Forthcoming). The percentages shown to the left of the graph show relative costs but do not correspond to actual cost levels. 24 microcredit is so labor intensive: salaries make up effectively. Regression analysis (Figure A-5) shows the majority of most MFIs' operating expenses, that the age of an MFI is strongly associated with and fixed costs are relatively low compared with lower operating costs, even after separating out the variable costs. MFIs that are still small enough to effects of loan size, scale, and other relevant variables. reap major economies of scale account for only a Not surprisingly, the effect weakens over time, but it tiny percentage of microcredit loans and customers. continues for quite a while. The operating expense In short, economies of scale cannot do much to ratio tends to drop by 2­8 percentage points for offset the added expense that comes from making each of the first six years, 1­2 points for each of the very small microcredit loans. next five years, and less than 1 point for each year thereafter. Trend of operating costs. Figure A-4 shows good news. MFI efficiency has improved substantially--i.e., We do not have a good statistical proxy for the operating costs have declined--in recent years in all competitiveness of an MFI's market, so we cannot regions except South Asia, where operating costs quantify how much of this age effect can be attributed were already quite low. to competition. However, it's likely that most of the effect is due to the learning curve, because relatively Why are operating cost ratios dropping: learning few MFI markets are competitive yet. curve, loan size, or competition? The decline in operating costs is a major contributor to the decline Loan size and mission drift. Since smaller loans are in interest rates that borrowers are paying. What is associated with higher operating expenses, an MFI causing this improvement in efficiency, and can we could reduce its operating expense ratio by simply expect it to continue? making larger loans. If an MFI whose borrowers all have $100 loans suddenly lets these borrowers Learning curve. One driver--perhaps the most double their loan size, or finds new borrowers who important driver--of improved MFI efficiency seems want larger loans, the MFI does not need to double to be the increasing age of large MFIs as the industry its staff or other administrative expenses. Are the develops. As institutions mature, they learn their improvements in efficiency shown in Figure A-4 business better and are able to control costs more nothing more than the result of increasing loan sizes, Figure A-4: Operating Costs as Percentage of GLP, 2003­2006 25 21.9 23.2 oss 20 18.8 18.5 Gr 15.6 16.7 15.9 15 14.0 13.8 12.7 11.1 11.311.9 Portfolio 10 9.0 average of Loan 5 % 0 WORLD Africa EAP ECA LAC MENA South Asia ­1.0% ­1.0% ­1.7% ­1.6% ­0.6% ­3.1% 0.2% 2003 2006 Avg. change per year Note: 175 sustainable MFIs reporting to MIX for both 2003 and 2006. 25 Figure A-5: Predicted Change in Operating Expense/GLP, by Age of MFI 70 60 50 40 OER 30 20 10 0 0 2 4 6 8 10 12 14 16 18 20 Age of MFI Note: Regression results based on 1,144 MIX Market and MBB MFIs reporting data for 1999­2006 (Gonzalez Forthcoming). The percentage scale at the left of the figure measures predicted relative change, but does not correspond to actual operating cost levels. which might reflect a movement toward serving a that the efficiency gains are not simply the result richer clientele?28 of larger loan sizes, but that they reflect true improvement in the cost of serving each client. The Figure A-4 reported the most common measure same conclusion can be drawn from the regression in of efficiency: Operating expense ratio, which is Figure A-5, which shows that the learning curve drives administrative cost divided by the amount of the loan a strong efficiency improvement even after screening portfolio. This operating expense ratio automatically out the impact of loan size. makes MFIs with larger loan sizes look better. Will cost per borrower tell us which managers are A more useful measure for our immediate purpose efficient and which are wasteful? Measuring normalized is administrative cost divided by the number of cost per borrower, rather than cost per dollar lent, borrowers. Normalizing the resulting "cost per filters out the effect of loan size and, to a lesser extent, borrower" by expressing it as a percentage of per country differences in labor and other input costs. But capita national income produces an indicator that this measure does not filter out the effect of many is better at describing MFIs from different countries other variables that impinge on an MFI's cost, such without automatically giving an advantage to those as geographical density of clients, transport and that make larger loans, or that have lower labor costs communications infrastructure, or the flexibility and in their country. sophistication of the products offered to clients. Figure A-6, which shows the 2003­2006 efficiency Also, MFIs report their total administrative expenses. trend using normalized cost per borrower, suggests They are seldom able to separate costs associated with 28 Some observers interpret growth in average loan sizes as a sign of "mission drift" away from poor clients, but it is far from clear how often this is valid. The link between loan size and client poverty is only a very rough one at best. Most MFIs have a sequential ladder of loan sizes for clients, and the very small loans at the beginning of that ladder often reflect the MFI's risk management policy rather than the actual needs or repayment capacity of the borrowers. When borrowers move into later and larger loans, or when the MFI relaxes its size limits on initial loans, the MFI's average loan size will climb even if there has been no change in the kind of client it is serving. And even where the MFI is adding better off clients (for instance small business operators) as it grows, that does not necessarily mean restricting service for poorer customers. A much more reliable way to judge mission drift is to look at the character of the villages, towns, and neighborhoods where the MFI is opening its new branches. 26 Figure A-6: Operating Expenses Per Borrower as Percentage of Per Capita Gross National Income 60 49.8 50 GNI 40 37.4 37.9 capita 30 25.7 per 20 15.4 12.5 14.0 of % 10 7.8 9.4 5.9 4.2 2.7 2.2 1.9 0 WORLD Africa EAP ECA LAC MENA South Asia ­1.0% ­4.1% ­0.6% ­4.1% ­1.5% ­0.5% ­0.1% 2003 2006 Avg. change per year Note: 175 sustainable MFIs reporting to MIX for both years, weighted by GLP. lending from costs associated with other products, abusively high today. But the data have permitted us like savings. So if an MFI adds savings services, its to make some relevant observations. The plausible reported cost per borrower will rise, but this doesn't argument that costs need to be higher when smaller signal deterioration in efficiency. loans are being delivered is supported by regression analysis of a large set of MFIs. MFI operating cost Finally, at a certain level, larger loan sizes do translate levels have dropped substantially in recent years, into larger administrative expenses per borrower--for and we have seen that this represents real operating instance, small business loans need more sophisticated improvements, not just the effect of increasing loan and expensive analysis than microloans, so one would sizes. Regression analysis suggests that much of this not expect to see cost per borrower as low in Eastern improvement reflects the learning curve as MFIs gain Europe as it is in South Asia. As a result of all these experience. factors, one cannot automatically conclude, for example, that managers of East Asian MFIs are more This still doesn't tell us how much avoidable "fat" efficient than managers in Latin America, just because is built into MFI operating expenses. Because most the former have a lower cost per borrower. microfinance markets are so young and because most are not yet competitive, it would be unrealistic to The various factors that impinge on cost per borrower expect MFIs in those markets to be operating at the affect individual MFIs differently. However, they stay most efficient levels possible. Immature industries more or less the same for a given MFI from one always have some level of correctable inefficiency. year to the next, so the overall decline in cost per We know of no evidence suggesting that MFIs in borrower shown in Figure A-6 very probably reflects general are out of line with the normal evolution of true efficiency improvement. efficiency for businesses in immature markets. Finally, the year-to-year cost of operating expenses shows an We know of no statistical approach that can directly encouraging decline, however we measure it. measure whether MFI administrative costs are 27 References Gonzalez-Vega, Claudio, and Villafani-Ibarnegaray. 2007. "Las Microfinanzas en la Profundización del Adams, Dale, Douglas Graham, and J. D. Von Pischke. Sistema Financiero. El Caso de Bolivia." El Trimestre 1984. Undermining Rural Development with Cheap Economico, Vol. 74, No. 293. Credit. Westview. Grace, David. 2007. "Middle Is Not a Four-Letter Ausubel, Lawrence. 1991. "The Failure of Competition Word." MicroBanking Bulletin, No. 14 (Spring): 3. in Credit Card Markets." American Economic Review, Karlan, Dean, and Jonathan Zinman. 2007. Observing vol. 81 (1). Unobservables: Identifying Information Asymmetries Bertrand, Marianne, et al. 2005. What's Psychology with a Consumer Credit Field Experiment. Yale Worth? A Field Experiment in the Consumer Credit University Economic Growth Center Working Paper Market. Yale University Economic Growth Center. no. 911. New Haven, Conn.: Yale University Press. New Haven, Conn.: Yale University Press. Meyer, Richard L., and Carlos E. Cuevas. 1992. Calem, Paul S., and Loretta J. Mester. 1995. "Reduction of Transaction Costs of Financial "Consumer Behavior and the Stickiness of Credit- Intermediation: Theory and Innovations." In Card Interest Rates." American Economic Review, Savinqs and Credit for Development. Report of the vol. 85 (5): 1327­36. International Conference on Savings and Credit for Development, Klarskovgard, Denmark, 28­31 May Christen, Robert, Richard Rosenberg, and Veena 1990. New York: United Nations, 285­317. Jayadeva. 2004. "Financial Institutions with a `Double Bottom Line': Implications for the Future Porteous, David. 2006. "Competition and Microcredit of Microfinance." Occasional Paper 8. Washington, Interest Rates." Focus Note 33. Washington, D.C.: D.C.: CGAP, July. CGAP. Cuevas, Carlos E. 1989. "Loan Transactions and Robinson, Marguerite. 2001. The Microfinance Borrowing Costs in Developing Countries." American Revolution, vol. 1. Washington, D.C.: World Bank. Journal of Agricultural Economics 71: 1356. Rosenberg, Richard. 2007. "CGAP Reflections on the Dehejia,Rajeev, Heather Montgomery, and Jonathan Compartamos Initial Public Offering: A Case Study on Morduch. 2005. Do Interest Rates Matter? Credit Microfinance Interest Rates and Profits." Focus Note demand in the Dhaka slums. Financial Access 42. Washington, D.C.: CGAP, June. Initiative. Stephens, Blaine. 2007. "Commercialization Continues Gonzalez, Adrian. 2007. "Resilience of Microfinance Apace." MicroBanking Bulletin (Spring): 35. to Local Macroeconomic Events: An Econometric Tran Dat. 1998. "Borrower Transaction Cost, Credit Analysis of MFIs Asset Quality. MIX Discussion Paper Rationing, and Segmented Market: A Study in the No. 1. Washington, D.C.: MIX, July. Rural Credit Market in Vietnam." Canberra, Australia: ------. Forthcoming. "Efficiency Drivers of Australia National University, National Center for Microfinance Institutions (MFIs): The Case of Development Studies. Operating Expenses." MIX Discussion Paper No. 2. Wright, David, and Dewan Alamgir. 2004. "Microcredit Washington, D.C.: MIX, forthcoming. Interest Rate in Bangladesh: Capping vs. Competition." Gonzalez, Adrian, and Richard Rosenberg. 2006. Unpublished. http://www.microfinancegateway.org/ "The State of Microcredit--Outreach, Profitability, content/article/detail/23259. and Poverty Findings from a Database of 2600 Microfinance Institutions. Presentation at World Bank Conference, "Access to Finance," May 30. No. 15 February 2009 Please share this Occasional Paper with your colleagues or request extra copies of this paper or others in this series. CGAP welcomes your comments on this paper. All CGAP publications are available on the CGAP Web site at www.cgap.org. CGAP 1818 H Street, NW MSN P3-300 Washington, DC 20433 USA Tel: 202-473-9594 Fax: 202-522-3744 Email: cgap@worldbank.org © CGAP, 2009 The authors of this Occasional Paper are Richard Rosenberg, a unpublished paper by Brigit Helms, Xavier Reille, Samer Badawi, senior consultant for CGAP; Adrian Gonzalez, lead researcher and Adrian Gonzalez and are grateful for useful comments from at the Microfinance Information eXchange; and Sushma Narain, Blaine Stephens, Elizabeth Littlefield, Ignacio Mas, and Jeanette an independent consultant. (Views expressed in this paper are Thomas, as well as analysis of BankScope data by Christoph not necessarily those of MIX.) The authors have made use of an Kneiding and patient and skillful editing by Anna Nunan. The following is a suggested citation for this Occasional Paper: Rosenberg, Richard, Adrian Gonzalez, and Sushma Narain. 2009. "The New Moneylenders: Are the Poor Being Exploited by High Microcredit Interest Rates?" Occasional Paper 15. Washington, D.C.: CGAP.