37822 FocusNote NO. 36 MAY 2006 COMMUNITY-MANAGED LOAN FUNDS: WHICH ONES WORK? Government and donor projects that deliver microfinance--that is, credit and other financial services for poor and low-income people--usually involve microfinance insti- tutions (MFIs) with professional staff.1 However, an increasing minority of microfinance The authors of this Focus Note projects rely instead on community-managed loan funds (CMLFs). In CMLFs, credit are Jessica Murray, consultant, to the members of a small group is managed by the members themselves, with no profes- CGAP, and Richard sional management or supervision of the approval, disbursement, and collection of loans. Rosenberg, senior advisor, CGAP. The authors thank These funds are referred to by a variety of names, including revolving funds, self- Hugh Allen, Jeffrey Ashe, managed village banks, accumulating savings and credit associations (ASCAs), and Tamara Cook, Malcolm community-based finance. Harper, Elizabeth Littlefield, Anne Ritchie, Stuart This Focus Note presents conclusions from a performance review of dozens of Rutherford, Ousa Sananikone, CMLF projects established or supported by donors and international nongovernment and especially Heather Clark organizations (NGOs) over the past 15 years. It turns out that success is strongly for their generous and useful advice. linked to the source of funding for the loans group members receive. Externally funded groups. When loans are financed by an early injection of © 2006, Consultative Group to external funds from donors or governments, CMLF projects appear to fail so Assist the Poor consistently that this model of microfinance support is never a prudent gamble. The Consultative Group to Savings-based groups. CMLFs are often successful when loans are financed by Assist the Poor (CGAP) is a members' own savings, and there is either no external funding, or such funding consortium of 33 development arrives in modest amounts after the group has a solid track record of lending and agencies that support microfi- recovering its own savings. nance. More information is available on the CGAP Web Self-help groups (SHGs).2 When groups start by collecting and then lending site (www.cgap.org). members' own savings, but subsequently receive large loans from a bank that is serious about collection, performance has been mixed so far. Of the three models, only the savings-based and the SHG models appear to be viable. In addition to the funding source, the other factor that seems to be a strong predictor of success is the quality of external support community groups receive. Such support is important on a continuing basis, not just at the inception of the groups. 1 In this Note, we use the term "donor" as shorthand for a wide range of development funding agencies, including bi- lateral government aid programs, multilateral development banks, and private philanthropic funders, such as foundations. 2 This Note discusses the SHG model as practiced in India. Many Indian SHGs are externally funded (by bank loans, not up-front donor or government infusions), and many are savings based. We treat them as a separate category because, de- spite some common characteristics, they tend to behave differently from the other CMLFs. It appears that a bank loan creates dynamics that are different from those created by an initial grant or loan from a donor or government program. Building financial systems for the poor This Note begins by describing the study's drawn from two other CGAP studies, as yet methodology, including sources and criteria for unpublished (Christen and Ivatury forthcoming evaluating CMLFs. Then it reviews the perform- and Prakash et al. forthcoming). ance of the three types of CMLFs, addresses whether CMLFs need long-term external support, Success Criteria and reviews the debate over the relative merits of CMLFs are often referred to as revolving funds, on community-managed and professionally managed the premise that their money will be lent out, approaches. It concludes with a brief summary of collected, and re-lent. The primary criterion we used the implications for development agencies that to judge the success of a CMLF program was support CMLFs. whether the funds did in fact revolve.3 In other words, were loans repaid well enough to maintain Methodology the fund for more than a few years? Other indicators of success were used, especially Sources in cases where loan collection was not meaning- We found an evaluation or some other implementa- fully reported. Programs where the majority of the tion report for 60 CMLF projects funded by 23 groups disbanded within a few years were judged agencies (13 bilateral or multilateral agencies and as failures. Evidence of "elite capture," where a few 10 NGOs) between 1990 and 2005. Where possi- powerful members expropriate for themselves the ble, we supplemented this documentary informa- resources that were meant to be at the service of tion by discussing the projects with knowledgeable the whole group, was a negative indication, development agency staff. Many of these projects though no project was graded as unsuccessful produced less concrete performance information solely for this reason. than one would hope. Only about half of the proj- The ideal criterion for measuring success would ects could be graded as a success or failure with a be demonstrable impact on the lives of group reasonable level of confidence, as shown in Table 1. members. Reliable measurement and attribution of the impact of financial services is surprisingly Table 1 Number and type of CMLF complex, expensive, and time consuming. projects evaluated Externally Savings- Credible impact studies are virtually never avail- funded based Total able for CMLF projects. However, there is a grow- projects projects Graded Projects 20 11 31 ing body of good impact studies for microcredit in Ungradable Projects 22 7 29 other settings, most of which find that access to Total 42 18 60 the financial service produces important welfare benefits for clients and their households The general pattern in microfinance is that there (Littlefield et al. 2003). The studies tend to find is a correlation between careful reporting and good that these benefits are associated with continuing results. In other words, projects that produce mean- access to services, not just one or two loans.4 In ingful performance information are more likely to be this sense, loan collection performance is strongly successful. If this pattern holds true for CMLFs, then the sample of projects we were able to grade may be 3Projects with repayment rates of 85 percent or lower were considered better performers on average than the universe of unsustainable and therefore unsuccessful for the purposes of this Note. CMLF projects as a whole. This is a very generous standard. If a portfolio of 6-month loans payable in monthly installments has an 85 percent repayment rate, it will lose We did not research the performance of Indian about half of its lending capital in a single year. Cf. Rosenberg (1999). SHGs. Rather, observations about SHGs are 4E.g., Dunn and Arbuckle (2001). 2 linked to client impact, because groups who "hawks" and poverty-focused "doves," that lend- lose their capital through default can't provide ing where default is likely is not the preferred tool continuing services. for any development purpose. One sometimes hears an argument that loan collection should not always be used as the criterion Three Types of CMLFs of CMLF success. The contention is that the main objective of some of these projects is not financial In his classic study The Poor and Their Money, Stuart Rutherford (2000) distinguishes between service delivery per se but rather the transfer of "provider" and "promoter" approaches to poor capital into poor communities, an objective that can people's finance. In the provider model, a special- be achieved even if loans are not repaid and groups ized, professional financial institution delivers fall apart. This reasoning has been used in several loans, deposit facilities, and other retail financial CMLF projects in East Asia, following the 1997 services to its clients. In the promoter model, com- financial crisis. Donors wanted to transfer money to munities are taught how to organize themselves so poor communities during the crisis and set up that community members can offer such services to CMLF groups without worrying too much about each other. repayment of the loans or the long-term survival of Rutherford argues that most poor people prefer the groups. We do not find this rationale convincing, the provider model, if it is available, because it though we admit that our objections to it are based tends to be more reliable and require less organiza- on anecdotal experience rather than specific evidence tional effort and risk on the part of customers.5 found in the documents we reviewed for this Note. Middle-class customers in rich countries are no dif- If a project proposal were to state candidly ferent. They normally would rather not have to that loans within the groups were likely to experience high default, and that the groups concern themselves with the organization and would be likely to collapse as a result, the management of the firms that provide them with project probably would never be approved in goods and services. the first place. Most donor support for microfinance goes to Social capital within the community is more providers: formal, specialized, professional micro- likely to be hurt than helped when a few finance institutions (MFIs). However, there are members are able to capture the benefits limits to the outreach of these MFIs. In particular, intended for the whole group, by the simple expedient of defaulting. These situations can they may find it difficult to serve remote or sparsely create friction, resentment, and mistrust populated areas, where the combination of low among individuals, and can reduce the com- client density and high transport and communica- munity's confidence in its ability to work tions costs can make it impossible for them to oper- together for shared goals. ate viable branches. Some assert that when a community gets used In the face of such constraints, a substantial to defaulting on poorly managed loans, it minority of microfinance interventions take a becomes harder for a competent MFI to do business in that community. decentralized approach, forming small community- managed groups that lend to their members and Redistribution of capital into poor communi- ties is a thoroughly worthwhile objective, but often mobilize members' savings as well. CMLFs there are many other ways to do this besides tend to have 5­40 members. Groups are usually setting up local loan funds. 5While noting this preference, Rutherford (2000) acknowledges that There is wide consensus in the microfinance self-managed ROSCAs, savings clubs, and savings-led CMLFs play a use- community, including both financially oriented ful role in many communities. 3 organized for the specific purpose of the revolving Advantages and Drawbacks of CMLFs fund, though sometimes pre-existing groups are In remote, rural communities that are not served by used. The nonprofessional and sometimes illiterate MFIs or the formal banking sector, the CMLF members both own and manage their fund. They model is sometimes the only feasible method of collect the savings (if any), decide on individual expanding basic financial intermediation beyond loans and loan terms, disburse the loans, and traditional informal mechanisms, such as family handle collections, all without authoritative super- lending or moneylenders. Internal transaction costs vision by an outside MFI and its professional staff. tend to be lower for CMLFs than for MFIs or other There is no way to develop an accurate estimate formal financial institutions, including banks. of the amount of donor funding that goes to Unpaid members of a CMLF perform functions CMLFs, but the volume of such funding is large. that an MFI would have to pay professional staff to Recent CGAP evaluations of microcredit projects carry out. As a locally self-contained operation that in the World Bank and the United Nations needs no building or regular office, a CMLF avoids Development Programme (UNDP) found that most of the infrastructure, transport, and commu- about 30 percent of those projects used a CMLF nications costs incurred by an MFI branch office. model. CGAP's very rough global estimate of Even when taking into account the cost of group funding agency flows for microfinance is about promotion, training, and monitoring, cost per bor- rower tends to be much lower for a CMLF than for US$ 800 million per year. If other agencies use an MFI (Ashe 2002, Allen 2006, Christen and CMLFs as heavily as the World Bank and UNDP Ivatury forthcoming, and Prakash et al. forthcom- use them, then CMLF funding would be well over ing). However, the fact that the group is not paying $200 million every year.6 professional staff to perform certain functions does Donor-supported CMLFs use a wide variety of not mean that those functions are cost-free. Instead approaches, and it is hard to separate them into of paying someone to manage their fund, the mem- completely watertight categories with no ambigu- bers have to contribute their own time and negoti- ity at the edges. We found that the most powerful ating energy, and assume risks they would prefer to framework for analysis was to group the projects avoid, especially the risk of losing their savings when based on the source of the funding for loans neighbors default on loans. within the community groups. Speaking in broad averages, CMLF members tend In externally funded groups, loans are financed to be poorer than MFI customers, at least on the predominantly by capital that a donor or govern- plausible assumption that rural location and the size ment agency has injected early in the life of the of loan or savings balances correlate with poverty to group. Savings-based groups fund their loans some extent. Because of their relative cost structures, mainly with members' deposits. In the SHG CMLFs can often handle transaction sizes that are model, most groups begin by collecting and lend- below what is feasible in more formal, professional- ing out their own savings. Later in the process, ized institutions. CMLFs have relatively few mem- some but not all of the SHGs secure an external bers, and the members usually know each other well, loan from a bank, which tends to be considerably 6This estimate does not include lending to Indian SHGs by government larger than the amount members have saved. banks. Banks are typically serious about collecting the 7To keep this research exercise within manageable bounds, we did not loans they have made to SHGs.7 include community-managed credit unions in its scope, though we rec- ognize that there can be significant overlap between credit union and CMLF methodology. 4 so these groups can sometimes be more flexible and savings products, or payment and cash transfer about adjusting the timing of repayment to accom- services) that can be available when microfinance is modate poor borrowers' unstable cash flows. offered by an institution with a banking license. Commercial banks and MFIs often find it impos- sible to operate in regions experiencing conflict or Externally funded CMLFs. Except in India, most post-conflict rebuilding. In contrast, CMLFs can of the CMLFs supported by donors and govern- continue to work in settings of political or economic ments receive a substantial infusion of capital at, or crisis, when most other financial services collapse. soon after, the formation of the group. The capital CMLFs have no institutional headquarters and no can come in the form of a grant to the group, or a branches that can be robbed or destroyed in times of loan on very subsidized terms. Some of these conflict or insurgency. The money is often spread groups do not collect members' savings at all, so out among members, lowering the risk of wholesale that the loans to group members are entirely theft. Even groups that cannot continue during a financed by external funds. In other programs, crisis are often able to convene again quickly when groups also collect and lend out members' deposits, an immediate threat subsides. but these are small in relation to the external capi- CMLFs are sometimes used to reach members tal. In either case, loans are financed mainly by for nonfinancial activities. Group meetings may money that the members themselves have not pro- include education about health, human rights, or vided. These CMLFs are initiated and organized other social topics. Groups sometimes take on more often by the funder than by the community, development projects, such as establishing schools, and members' main motivation to join them is building water pumps, or constructing roads. based on wanting access to the external funds. Finally, and very importantly, CMLFs can provide Externally funded CMLFs almost always fail, a service that many MFIs cannot offer: savings. mainly because of high rates of default. Of the 20 Because these groups are largely informal, they usu- externally funded CMLFs in this study that pro- ally do not have to comply with regulatory require- duced enough information to support a judgment ments that prevent many MFIs--at least about performance, only one was successful. 9, 10 temporarily--from accepting voluntary deposits.8 Why do externally funded CMLFs fail (while sav- Savings is a service that is highly valued by the poor. ings-based CMLFs often succeed)? There are several Many of them are even willing to pay for a safe place to save their money. In addition, savings-based 8In many countries, MFIs without a license from banking authorities can CMLFs can generate dividends for their members as require clients to make savings deposits in order to receive loans. These savings are lent out and repaid with interest. obligatory deposits are best thought of as a cash collateral feature of the loan contract, rather than as a true savings service. The major limitations of the CMLF approach 9 The single successful externally funded CMLF we found was a 1993 are linked to the absence of professional manage- World Bank project in Albania. The community groups formed by this ment. It is easier for a formal, professionally run project maintained very good loan collection. However, these groups had a considerably higher level of professional support than was the case in MFI to institutionalize good recordkeeping, care- most other projects. Decisions and responsibilities about collection and ful follow-up on loan repayments, and sound support were in the hands of members, but a professional loan officer financial management. It is no surprise that paid by the funding apex organization was an ex officio member of each group's loan committee and participated in all meetings with voice but CMLFs tend to be less stable than MFIs, though no vote. The World Bank project manager described their role as "promi- some CMLF models can achieve stability when nent" and "influential" but not "authoritative." they receive adequate external support. 10 For each project whose success we graded, we also assigned a confi- dence level to the grade. Of the 20 grades for externally funded projects, Another limitation is that CMLFs cannot pro- one was characterized as low confidence, four as medium confidence, and vide the range of services (including varied loan 15 (including the single successful project) as high confidence. 5 reasons, the main one being members' perceptions distorted incentives and jeopardized the success of about what Richard Montgomery (1995) describes the project: as "hot" and "cold" money. Capital generated CARE's Kupfuma Ishungu program in through local savings feels hot because it comes Zimbabwe experienced false expectations and from one's neighbors. Defaulting on loans that are distorted savings behavior as a result of prom- ising external credit while promoting the proj- savings based feels like stealing from neighbors, ect. After watching two thirds of the groups with the result that borrowers are more likely to disband after receiving their loan, CARE take repayment seriously. stopped injecting capital into the groups. Cold money is outsiders' money, usually from Since then, new group start-ups have been donors or governments. Cold money is often treated nearly three times as large as the original pro- gram of 270 groups, and an additional 1,462 with less respect. If there is little negative reinforce- groups were created through a savings-led sis- ment preventing people from defaulting, such as the ter program (Allen 2002). fear of losing collateral, then repayment of cold The Mexican government's 1997 Rural money may not be a priority, even if the money will Development in Marginal Areas program helped go back into a revolving fund for other members of farmers' groups start revolving loan funds using the community to use. When a member defaults on capital provided by the government. After find- a loan funded by cold money, the other members are ing that the funds were not being recovered, the government received support from the World no worse off than they were before the fund was Bank to develop a savings-led alternative. established. Furthermore, people in many places are Members were taught to save their own funds conditioned by past experience to assume that repay- and lend them to other members. The savings- ment expectations are low for any money received led project proved to be successful, with a repay- from donors or the government. These core incen- ment rate of nearly 100 percent (Zapata 2002). tive problems are exacerbated when systems and Savings-based CMLFs controls are weak. Not surprisingly, loan funds that are administered by nonprofessional (and sometimes Communities all over the world can and do form loan illiterate) community members are more likely than funds based on their own savings without any exter- professional MFIs to have disorganized recordkeep- nal support. The most common type is the rotating ing, poor follow-up on loan collection, and weak savings and credit association (ROSCA). ROSCA financial management. members meet regularly, the number of meetings in In externally funded groups, members are likely a cycle usually being equal to the number of mem- to experience savings not as a valuable service but bers. At every meeting, all members deposit an rather as a hurdle they have to clear in order to get agreed contribution. Each member in turn receives loans. This kind of environment is seldom effec- the total amount collected at one meeting. tive at encouraging good savings habits. ROSCAs require a high trust level, because a Many experienced microfinance practitioners member who receives her distribution late in the have expressed skepticism about externally funded cycle has to depend on the continued contributions CMLFs for years. This study fully confirms their of members who have already taken their payout. view. The track record of externally funded groups However, ROSCAs are relatively easy to administer, is so poor that funders should simply abandon them because transactions are completely standardized and as a vehicle for poor people's finance. We found two no money needs to be kept safe between meetings. examples of CMLF projects that started out The ROSCA model propagates itself by spontaneous providing external capital to groups, and then replication, and presents little scope for external sup- stopped after realizing that external funding port. ROSCAs do not revolve; groups are meant to 6 disband after every member has received their allo- Table 2 Illustrative loan repayment rates in cation. For these reasons, ROSCAs are not consid- externally funded and savings-based CMLFs Country Loan Repayment Rate (%) ered CMLFs for the purposes of this study. They are, Externally funded CMLFs however, a much simpler form of community- Albania 99 managed finance, whose principles form the founda- Indonesia 45 tion of many CMLF models. Indonesia 77 Accumulating savings and credit associations Indonesia 50 (ASCAs) do not distribute all the money at each Kyrgyz Republic 85 Lao PDR 16­60 meeting, and are able to offer more flexible loan and Lao PDR 68 savings options. Some members join because they Malawi 40 want to save, while others are mainly interested in Mexico 80 borrowing, although every member must save regu- Rwanda 55­83 larly. ASCAs, the most basic type of savings- Zanzibar 50 led CMLF, pose greater risk than ROSCAs and Savings-based CMLFs are more complicated to manage. While they use Nepal 96­100 Niger 100 no external capital, they are more likely candidates Mexico 100 than ROSCAs for outside administrative and tech- Syria 99.7 nical support. This means that donors can play a Uganda 95­98 useful role, for instance by promoting and organ- Note: Stable microlending usually requires repayment rates higher izing groups, training members how to operate than 95 percent, often much higher, depending on the length of the loans. For a portfolio of 3-month loans payable weekly, a 95 percent the fund, and helping set-up appropriate record- repayment rate entails an annual loss of 37 percent of the lending capital. For an explanation of this surprising result, see Rosenberg keeping systems. (1999). This analysis assumes that the repayment rate is being measured conservatively: cash received during some period divided Thus far, donors have used the savings-based CMLF by cash due during the same period. Few of the reports were spe- model much less often than the externally funded cific about how repayment rates were calculated. A 95 percent re- payment rate using a less conservative definition would be associ- model. This may be because, although savings- ated with even greater actual losses. based CMLFs mobilize local funds and allow Our research found enough performance infor- members to use those funds more productively, this mation to grade 11 savings-based CMLF projects. model does not involve transfers of capital into poor Surprisingly, every project in this sample appeared communities, which may be a donor's principal to be successful.12 objective. Also, public funding agencies and their The basic reason why this model performed better staffs often have strong incentives to move large than the externally funded model was discussed in the amounts of money, but savings-based CMLFs previous section. When members, rather than out- (where the donor is financing only support func- siders, provide the funds that are being lent, the tions) cannot channel as much money as externally incentives favor more careful management of the funded CMLFs can (where the donor is financing money and better repayment of loans. Table 2 shows most or all of the lending capital). A further factor repayment rates reported by graded CMLF projects.13 is that some donors think of microfinance exclu- sively as a support for enterprises, not as a multipur- 11 We cannot confirm or contradict this assertion. 12 We assigned a high confidence level to seven of these grades and a low pose household financial management tool. Many confidence level to four. believe that savings-based groups cannot mobilize 13 The graded projects not shown in Table 2 used other performance in- loans that are large enough to create or develop dicators, such as timeliness of loan payments, or the net return members earn on their savings. For example, participants in CARE's MMD pro- microenterprises.11 Finally, there is a common mis- gram in Niger earned about 76 percent on their deposits annually, after perception that the poor do not or cannot save. accounting for loan losses (Allen 2002). 7 The table highlights the general conclusion that Self-help Groups and Bank Linkages savings-based CMLFs are more successful than As mentioned earlier, we did not review individual externally funded groups at collecting their loans, SHG programs for this Note. The description of maintaining their capital, and thus being able to SHGs and their performance is drawn mainly from keep the fund revolving. other sources, including two forthcoming CGAP When a donor is supporting a savings-based papers (Christen and Ivatury and Prakash et al.). CMLF, it does not fund the loans to the groups' When donors inject capital into CMLFs, their members. The donor funds auxiliary functions, main reason is a belief that the members can man- such as promotion, organization, training, or age, and benefit from, loans that are bigger than bookkeeping assistance. This nonfinancial support what the members' savings alone could fund. In has often proved critical to success. When such many cases this belief is true. But a better option nonfinancial support is the donor's only role, we exists. Once a group's credit needs outweigh their suspect that the donor is more likely to pay close local resources, donors can help link the group attention to these functions and ensure that they with commercial banks or other formal financial are performed effectively. service providers. The individual members usually The list of savings-based CMLF programs for cannot use the bank because their balances are too which we found evaluations is not a long one-- small, or they have no collateral for loans, or the only 11 projects. But every one of these appeared bank branch is too far away. But once the group to be successful. Of course, this does not mean has built some savings assets and has a track record there are no failures: projects with good results of managing their internal lending, its members are more likely to be documented than projects can sometimes get access to the bank as a single with bad results. But the consistent success of the collective client, which lowers the bank's transac- projects we were able to grade is striking, to say tion costs in dealing with them. the least. In India, SHGs serve many more people than One interesting example is the Women's conventional MFIs do. The model is being picked Empowerment Program (WEP), funded by up elsewhere, but the vast majority of SHGs are USAID and implemented by PACT in Nepal. In still in India. In this model, an NGO, government only one year, this savings-led CMLF project pro- agency, or bank promotes the formation of the vided training and support to 130,000 women in groups and gives them a greater or lesser degree of 6,500 groups. The program focuses on literacy support services. The groups collect members' sav- and savings, with 96 percent of groups reporting ings and lend them out. Some of the groups--per- perfect repayment rates on loans. Since joining haps about a third--continue to function with no WEP, many of the participants started businesses, capital beyond their savings, but the majority of improved their literacy, and experienced greater them eventually move on to borrow from a bank. decisionmaking authority in the home. About 800 The bank linkage seems to work best if it is delayed new groups were spontaneously replicated until the groups have gone through several through word of mouth, and a market developed preparatory stages, illustrated in Figure 1. for second-hand training materials. Such replica- The first stage is building social capital. The tion increases outreach without requiring addi- SHG needs leadership, trust among members, and tional donor funds, and offers a tangible training on group management, collection meth- demonstration of the participants' belief that the ods, recordkeeping, and other topics. CMLFs are making their lives better (Ashe and The second stage is building internal capital Parrot 2001). through savings. Regular deposits test and 8 Figure 1 The building blocks of successful CMLF bank linkage After building social and Linkage to formal credit financial capital, development services can provide permanent agencies can provide advisory access to financial services, but services, training on record- it should be introduced only keeping, etc., which help the after groups have built up their group manage and lend out its own resources and have savings. successfully lent and repaid Bank Linkage funds from their own savings. Development agencies first need to ensure groups have a solid Financial Intermediation Groups should social foundation by build their own encouraging local capital through leadership development regular savings. Building Financial Capital through group formation and training. Building Social Capital demonstrate members' ability to make loan pay- rules. After some years of experience, a few of the ments at a later point. banks have come to view SHGs as having commer- The third stage is financial intermediation, cial potential, and have exceeded their priority- when the group members lend out their savings sector lending quotas. internally and collect the loans with interest. Even though reliable information on most SHG External support is important at this stage, programs is hard to come by, experience to date because recordkeeping becomes complex and indicates that SHGs can be a viable model, if enforcement of rules more challenging. implementation is competent. After groups have enough experience success- Use of the SHG model is certainly no guarantee fully repaying loans from their own savings, they of success. APMAS, a respected and experienced can handle a bank loan relatively safely, as long as SHG support institution that has evaluated hun- they do not commit to payments that are larger dreds of SHGs, estimates that a majority of Indian than they can handle. SHGs are of poor quality (Christen and Ivatury The crucial question, of course, is the bank's forthcoming). But a number of SHG programs, willingness to lend to the group. In India, the ini- including some of the largest, appear to be doing tial reason that almost all of the banks made these quite well. Prakash, et al., (forthcoming) and loans was a government mandate. Under "priority Christen and Ivatury (forthcoming) describe a sector" lending rules, banks in India--most of half-dozen large SHG programs that are keeping which are government owned--have long been loan default at very low levels and doing a good to required to allocate a certain percentage of their excellent job of collecting enough interest income assets to loans for poor, rural, or otherwise disad- to cover all the operating costs involved, including vantaged target groups (Prakash, et al. forthcom- the costs of external support. ing). The banks may not have had great What distinguishes successful SHG programs confidence in the SHGs, but they saw SHGs as from the rest? The clearest pattern that seems to be preferable to the even riskier target groups that emerging is that success tends to correlate with the were their alternatives under the priority-sector quality of external nonfinancial support for the 9 groups, including standardized products and MFIs or SHG programs from growing rapidly and norms, training, help with member acquisition dominating the field. and retention, bookkeeping and administration, Indian SHGs aren't the only CMLFs linking and in some cases direct authoritative supervision groups with banks after successful financial inter- of the group's operations. Likewise, careful phas- mediation from member savings. CLASSE-B in ing is important, as described earlier. Rwanda, a project funded by IFAD and imple- Most SHGs receive external capital from banks. mented by CARE International, organizes and Why, then, are some SHG programs able to oper- trains CMLF groups of 15 to 30 members. Groups ate more successfully than the other externally are trained to mobilize savings and make loans to funded CMLFs described earlier? Three factors their members. Once they have finished the 8- may account for the difference: month training period and have shown a satisfac- Compared to a typical CMLF that is financed tory repayment history (most groups have a 100 by an early one-time injection from a donor percent repayment rate so far on their internal or government agency, the SHG funding loans), they are invited to submit proposals for structure creates stronger incentives for bank loans. More than 50 percent of the submitted responsible lending and borrowing. SHG projects were returned to the groups to revise and members know that the external funds come from a regular bank, which they assume is reformulate before approval was considered. As of serious about collecting its loans. They expect June 2005, the repayment rate on bank loans is that the bank will continue to provide 100 percent (Vita 2005). future--and perhaps larger--loans as long as Although SHG replications do exist outside of the group collects its own loans and repays the bank responsibly. They know that they India, they are not nearly as prevalent in countries may lose their own savings if the bank loan is where banks do not have government-imposed not repaid. social lending targets. SHGs can work without As mentioned earlier, SHGs usually begin by a bank linkage, though without the linkage there collecting and lending out their own savings, is little to distinguish them from other savings- sometimes for an extended period, before based CMLFs. they receive a bank loan. This tends to pro- duce more disciplined groups. Continuing External Support It appears that groups in the better SHG pro- grams receive external nonfinancial support Although CMLFs are not themselves profession- and guidance that is stronger than what is typ- ically found in the externally funded CMLFs ally managed, they seem to do better when they discussed earlier. get external support from professionals. Does that support need to be permanent, or can the individ- India's SHG model is relatively young. Before ual groups be expected eventually to continue suc- making final conclusions about the model, one cessful operations entirely on their own? It is hard would want better information about the majority to answer this question solely on the basis of the of the programs, and more experience with long- evaluations we could identify for this study, term performance. But the weight of present evi- because most of the programs evaluated were fairly dence suggests that it is a viable model that young, so that few of the evaluations were able to deserves support, expansion, and refinement. look at long-term experience. However, the over- True, most Indian SHGs are probably weak. The all history of donor- and government-supported same is true of the majority of the world's individ- community finance strongly suggests that perma- ual MFIs, but this does not prevent the better nent external support structures are needed. 10 When new community finance models are devel- between the two approaches, should development oped, there has been a tendency to underestimate agencies have any general preference for one or the the degree of continuing external services commu- other? As mentioned earlier, Rutherford argues nity groups would need. For instance, when that, all other things being equal, most poor people FINCA International first designed its "village would rather get financial services from a profes- banking" approach two decades ago, it hoped that sional "provider" than be told by a "promoter" how its support to each group would be limited to ini- to manage such services for themselves. Many prac- tial formation and training, and that groups could titioners concur, but others do not. Which model spin off into fully independent operation after a few should be preferred when both are feasible is a con- three-month loan cycles. However, groups experi- troversial question whose answer would require evi- enced a high rate of collapse after external support dence far beyond the scope of the CMLF ended, so FINCA altered its approach. Now, evaluations we reviewed for this Note. All we can do FINCA stays heavily involved with all of its groups, here is to outline some of the issues and arguments. to the extent of providing not just support but The evidence so far suggests that the cash costs authoritative external management. of administration tend to be lower--often much Another example can be found in credit unions lower--on average for CMLFs than for MFIs, even and other forms of cooperative finance, where, after factoring in costs of promotion and external after decades of experience, prominent promoters support. However, this analysis does not include and technical advisors are practically unanimous in the noncash transaction costs for CMLF members the view that the groups do better with continu- who are performing some management functions, ing external support. This pattern holds not just and perhaps assuming some risks, that they would for tiny savings and loan cooperatives in poor not be burdened with if they were MFI clients. On countries, but also for large and highly sophisti- the other hand, MFI clients incur transaction costs, cated credit unions in rich countries that may not too, and there are certainly cases where dealing need institutional development support but do with the group in one's own village is easier than need external regulation and supervision. dealing with the MFI branch in the next village. So funding agencies, governments, and NGOs In most countries, professionally managed MFIs that assist CMLF programs need to plan for the struc- that do not have a deposit-taking license from finan- tures that will provide the needed long-term support. cial authorities cannot provide voluntary savings Of course, continued external support for groups services to their clients.14 Informal CMLFs do not does not necessarily mean continued presence of face this constraint, so they can provide both savings international donor or promotion agencies: member- and loans. But MFIs that do have deposit-taking owned federations or other domestic support struc- licenses already account for the majority of MFI tures will be the normal permanent arrangement. clients worldwide, and the percentage of clients served by unlicensed MFIs will continue to shrink. CMLFs versus Professionally Licensed MFIs can offer services that are unavail- Managed Finance able in a CMLF, such as larger and longer-term loans, long-term savings instruments, or payment CMLFs can reach some locations and clients that and cash transfer services. would be impractical for a professionally managed MFI or credit union to serve. But there are plenty 14Unlicensed MFIs often collect obligatory deposits as a condition to making loans. These mandatory arrangements should be thought of as a of situations where either of the two models could cash collateral requirement for the loan, rather than as a savings service be viable. In such cases where there is a choice to help clients manage their liquidity. 11 Some people favor CMLFs because they are dem- SHGs, most of which have bank linkages, ocratically governed, and thus empower members to have shown mixed performance, but results obtained by the bigger and better programs take more control over their own financial lives. At the suggest that the model itself is effective when same time, other observers argue that client owner- it is implemented competently. ship and governance tends to hurt rather than help CMLF projects need to do a much better management of financial services, especially where job of reporting performance, not least of all there is a one-person-one-vote rule. They point out, because reporting performance tends to for instance, that governance of large, efficient credit improve performance. CMLFs cannot be bur- dened with elaborate recordkeeping, but it is unions is usually not democratic in practice. critical that projects report at least (1) out- The core of the discussion should probably con- reach--numbers of clients and groups; (2) cern the longer-term stability of the two models. loan repayment, using industry-standard Proponents of formal institutions acknowledge measures, and (3) group survival.15 Other indicators that have been used in good CMLF that professional management is costly, but argue programs include financial return on mem- that without such management, community bers' savings, reasons for group disbandment, groups will have a hard time preventing losses of frequency of staff visits, source and amount of members' savings and maintaining continued members' outside borrowings, and members' perceptions about quality-of-life changes as a access to services. The only way to resolve this result of the fund. question will be better research into the long-term Instead of injecting loan capital into CMLFs, performance of community-managed models, funders should use their resources to provide viewed over decades rather than years. support services for the groups. The cases re- Finally, it is important not to let this discussion viewed for this study as well as decades of experi- create a false dichotomy. Professional and nonprofes- ence with other community finance models indi- cate that CMLFs need competent, continuing sional finance can and do coexist in many settings. external support for a range of functions, includ- Clients often use both of them simultaneously to ing promotion, organization, training, bookkeep- deal with different financial needs. ing, networking, liquidity management, and per- formance monitoring. In some cases, the groups Conclusions for Development Agencies do better when they are subject to some degree of control by external management. There will continue to be questions about whether to CMLFs can provide savings and loan services for use a community-managed approach to finance in a millions of poor people, including many who are given situation. But once one has decided to develop beyond the practical reach of formal, professional a community-managed program, there seem to be MFIs. Development agencies should, and no doubt some straightforward lessons about sound practice. will, continue to support CMLFs. As the lessons Externally funded CMLFs practically never from past experience are better documented, more work, because they have to swim against the widely understood, and appropriately reflected in stream of the natural incentives of group project designs, we can expect a strong improve- members. The odds of success are so low that ment in the overall effectiveness of these projects. development agencies should abandon them completely and rely on the other two models 15 Measurement of microcredit repayment is a notorious minefield. when they want to do CMLFs. Some commonly used indicators cloud more than they clarify, and there is little consistency in terminology or calculation methods. For CMLFs Savings-based CMLFs that use no external that receive external loans, repayment reporting should include both the capital perform surprisingly well, at least external loans to the group and internal loans to group members. For a based on the sample of eleven that we were guide to meaningful repayment reporting, see Rosenberg (1999) and able to analyze in this review. Bruett (2006). 12 References Adler, Matthias. 2001. "Village Banks in Mali: A Successful Project of Self-Help Promotion." D+C Development and Cooperation, no. 1:18­20. Allen, Hugh. 2002. "CARE International's Village Savings & Loan Programmes in Africa: Micro Finance for the Rural Poor that Works." Niger: CARE Publication. ------. 2006. "Village Savings and Loan Associations." Pending in Small Enterprise Development (March). Ashe, Jeffrey. 2002. "A Symposium on Savings-Led Microfinance and the Rural Poor." Journal of Microfinance 4 (2): 129. Ashe, Jeffrey, and Lisa Parrot. October 2001. "Impact Evaluation on PACT's Women's Empowerment Program in Nepal: A Savings and Lit- eracy Led Alternative to Financial Institution Building." Independent evaluation financed by Freedom from Hunger and the SEEP Net- work. Bruett, Tillman. March 2006. "Measuring Performance of Microfinance Institutions: A Framework for Reporting, Analysis, and Monitoring." Washington D.C.: SEEP Network. Christen, Robert Peck, and Gautam Ivatury. Forthcoming. "A Systematic View of the SHG Bank-Linkage System: Four Sustainable Models." A Report for the State Bank of India. Washington D.C.: CGAP. Dunn, Elizabeth, and J. Gordon Arbuckle Jr. 2001. "The Impacts of Microcredit: A Case Study from Peru." USAID Office of Microenter- prise Development. Washington D.C.: USAID AIMS Project. Grant, William, and Hugh Allen. 2002. "CARE's Mata Masu Dubara (Women on the Move) Program in Niger: Successful Financial Inter- mediation in the Rural Sahel." Journal of Microfinance 4 (2). Littlefield, Elizabeth, Jonathan Morduch, and Syed Hashemi. 2003. "Is Microfinance an Effective Strategy to Reach the Millennium De- velopment Goals?" Focus Note 24 (January). Washington, D.C.: CGAP. Matthews, Brett, and Ahsan Ali. 2002. "Ashrai: A Savings-Led Model for Fighting Poverty and Discrimination." Journal of Microfinance 4 (2). Montgomery, Richard. 1995. "Disciplining or Protecting the Poor? Avoiding the Social Costs of Peer Pressure in Solidarity Group Micro- Credit Schemes." Papers in International Development No. 12. Swansea: Centre for Development Studies, University of Wales. Prakash, L. B., Anuradha Pillai, Syed Hashemi, and Jennifer Isern. Forthcoming. "Do India's Self-Help Groups Provide Value for Money?" Washington, D.C.: CGAP. Rosenberg, Richard. 1999. "Measuring Microcredit Delinquency: Ratios Can Be Harmful to Your Health." Occasional Paper 3 (June). Washington, D.C.: CGAP. Rutherford, Stuart. 2000. The Poor and Their Money. Oxford: Oxford University Press. Seibel, Hans Dieter. 2003. "Community Development at Jabal al-Hoss II: Promoting Sustainable Livelihoods and Eradicating Poverty through a Sustainable Network of Village Development Funds." Consultancy report for UNDP and the Syrian Arab Republic. Seibel, Hans Dieter, and Dave Harishkumar. 2002. "Linking Banks and Self-Help Groups: Social or Commercial Banking? The Experience of India." Rome: NABARD, Mumbay & IFAD. Vita, Massimo. 2005. "Review and Monitoring of CLASSE-B Project- Rwanda." Project evaluation prepared by Microfinanza SRL. Wilson, Kim. 2002. "The New Microfinance: An Essay on the Self-Help Group Movement in India." Journal of Microfinance 4 (2): 217­245. Zapata, Gabriala. 2002. "Community Savings Funds: Providing Access to Basic Financial Services in Marginalized Rural Areas of Mexico." Journal of Microfinance 4, (2): 163­187. 13 Annex I. List of Projects Reviewed Externally Financed (graded projects) Location Agency Project Name Albania World Bank Rural Development Project Bangladesh SDC (Swiss) Ashrai (portion with external credit) Cambodia UNICEF Seth Koma Program Ghana FAO and Dutch Government People's Participation Program Global FAO-sponsored evaluation Small Farmer Groups (with external credit) Indonesia World Bank Bengkula Regional Development Project Indonesia World Bank Integrated Swamps Indonesia World Bank Kecamatan Development Project Indonesia World Bank Nusa Tenggara Area Development Project Indonesia World Bank Urban Poverty Project Kyrgyz Rep. World Bank Rural Finance Project Lao PDR SIDA Lao Swedish Forestry Programme Lao PDR UNDP/UNCDF Small Scale Irrigation Schemes in Oudomxay and Luang Namtha Malawi CARE International Village Savings and Loans Program (portion with external credit) Rural Development in Marginal Areas Program (precursor to Mexico Mexican Government Community Savings Funds) Mexico World Bank Rural Development in Marginal Areas Project Nepal GTZ and ADBN Small Farmer Cooperatives Rwanda World Bank Community Reintegration Zanzibar CARE International JOSACA Zimbabwe CARE International Kupfuma Ishunga (portion with external credit) Savings Based (graded projects) Location Agency Project Name Bangladesh SDC (Swiss) Ashrai (portion without external credit) Eritrea USAID Community-Managed Savings and Credit Associations Global FAO-sponsored evaluation Small Farmer Groups (without external credit) Malawi CARE International Village Savings and Loans Program (portion without external credit) Mexico World Bank Community Savings Fund (portion without external credit) Mozambique CARE International Ophavela Nepal USAID/PACT Women's Empowerment Program Niger USAID/CARE International MMD Program (portion without external credit) Syria UNDP Rural Community Development at Jabal al Hoss II Uganda DFID Financial Sector Deepening Project (FSDU) Zimbabwe CARE International Kupfuma Ishunga (portion without external credit) 14 Ungraded Projects (both externally funded and savings based) Location Agency Project Name Albania UNDP Village Development Fund for Income Generating Activities Cambodia PACT Worth Program Ethiopia World Bank Women's Development Initiatives Project Global IFAD Revolving Funds Honduras UNDP/IFAD National Programme of Local Development India CARE International Community Managed Revolving Loan Funds--CASHE Program Support to the Implementation of the Indonesia Indonesia UNDP Community Recovery Programme Jordan European Commission Social Development Project/ Development and Employment Fund Lao PDR ADB Community-Managed Livelihood Improvement Project Lao PDR FIAM (Thai NGO) Women in Development Lao PDR German Agro Action Community-Based Rural Development to Reserve Watershed Project Lao PDR GTZ Rural Development Project Mennonite Central Committee Lao PDR Village Development Committee Credit Funds (MCC) Lao PDR Oxfam Solidarity Cattle Banks Lao PDR Quaker Services Revolving Loan Funds Lao PDR World Concern Village Revolving Funds Lao PDR ZOA Village Credit Associations Madagascar World Bank Rural Development Support Project Mali CARE International MJT (Musow Ka Jigiya Ton) Nepal UNDP Participatory Development Project Niger World Bank Agro-Pastoral Export Promotion Project Panama UNDP Desarrollo Rural Sostenible en el Darién Rwanda World Bank Agricultural and Rural Market Development Project Senegal Oxfam America Oxfam America Self-Help Group Model Senegal UNDP Programme Elargi de Lutte Contre la Pauvrete Uganda CARE International JENGA Zimbabwe, Mali, Oxfam Banking on the Poor Cambodia 15 FocusNote No. 36 Please feel free to share this Focus Note 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