Village Banking Dynamics Study: Evidence from Seven Programs

June 1999

by Judith Painter, Barbara MkNelly and the members of the SEEP Poverty Lending Working Group

INTRODUCTION Since John Hatch of FINCA first introduced village banking approximately twelve years ago much has been learned regarding the original model of growth and recommended financial delivery system. The flexibility and robustness of this lending methodology is apparent as village bank programs are now operating in over 70 countries, working with over 200 local and international organizations (Nelson et al., 1995). Expectations surrounding village bank programs have increased as they strive to reach broader and deeper into populations. While the emphasis on the very poor (i.e., womenusually motherswith credit needs of $300 or less) remains, long-range financial sustainability of program services has become a given objective of village bank practitioners. As a result, many programs are expanding the loan ceiling, focusing on urban areas and planning to reach thousands of clients within their first five years of operation. Also, while the original methodology emphasized a three-year process by which individual village banks would become autonomous and graduate, current village bank efforts, including all the programs featured in this paper, focus more on the overall financial sustainability of the program services and the institution. Despite the substantial adaptations programs make to suit their own needs and working environments, most village banking programs began with what is referred to here as the original Hatch model (Hatch, 1989). This model was the basis for planning expansion strategies, funding requirements and sustainability projections. Underlying the original Hatch model are the following assumptions: First-cycle entrants will remain active borrowers for nine four-month cycles (3 years). Village banks will grow in size as new members, encouraged by the example of others, join the program. Clients will save 20% of their current loan size each cycle. Loan sizes increase based on a formula of last loan plus accumulated savings. (See Box 1.) Based on this formula, clients, primarily women, will borrow a progression of loans starting at $50 dollars and reaching the $300 loan ceiling by the third year. The projected loan-size growth rate is 500% over a 28-month period. Box 1 Original Hatch Model Projections Loan Cycle (4-mo) 1 2 3 4 5 6 7 8 9
Village Banking Dynamics Study Evidence from Seven Programs

Loan Size per Borrower $50 $60 $82 $120 $182 $281 $300 $300 $300

Savings per Borrower per Cycle $10 $12 $16 $24 $37 $56 $60 $60 $60

Cumulative Savings per Borrower $10 $22 $38 $62 $99 $155 $215 $275 $335
2

The proliferation of village banking programs throughout the developing world, each with several years of experience, provides the opportunity to compare actual program information to the original Hatch model projections. This paper examines several core questions affecting loan portfolio growth using data collected from over 700 individual clients in 26 village banks from seven different programs. The intention here is not to assess the accuracy of the original model, but rather to learn from the experience and adaptations of a number of organizations implementing village bank programs since the model was introduced. Specifically, this study sought to improve the understanding of village banking dynamics and identify particularly influential factors affecting program performance over time. The primary question examined in the study is whether client loans grow or stagnate over time. This issue is important to village bank practitioners in terms of achieving their goals pertaining to positive impact as well as financial sustainability. Village bank programs are established with the ultimate goal of reducing poverty and raising the income and quality of life of its clientele. The ability of borrowers to work and repay progressively larger loans is a proxy indicator for the viability of loan activities and poverty alleviation at the client level. To accomplish these impact goals, programs recognize the need for financially sustainable operations. Financial sustainability of village bank programs presumes a declining cost per dollar loaned as members loans grow and as new clientele join existing banks. Therefore, if loan sizes are indeed stagnating, it is important to understand why. The questions below set forth specific areas of study in relation to loan-size growth: 1. What is the average loan size in practice versus the Hatch model and what are the key factors affecting loan-size growth? 2. Does program locationurban versus ruralaffect loan-size growth? 3. What is the actual savings rate in practice versus the Hatch model? What are the key factors affecting borrower savings and how does it affect average loan size? 4. What are the actual membership dynamics (drop-out, drop-in and client retention rates) as compared to the original Hatch model? What factors affect membership dynamics, and to what degree do they affect loan-size growth rates? 5. How does internal account lending affect average loan size (of external loans)? What factors affect the degree of internal account lending? Each of these dynamics are interrelated and all can potentially affect the loan-size growth rates of a given village banking program. It becomes imperative for village bank practitioners to have a thorough grasp of these dynamics if their program hopes to achieve financial sustainability. METHODOLOGY Five nonprofit organizations, all of whom are members of the Small Enterprise Education Network (SEEP) Poverty Lending Working Group, participated in the study: CARE, FINCA, Freedom from Hunger, Womens Opportunity Fund and World Relief. Each organization collected longitudinal information from a sample of village banks in one or two of their programs. In order to represent a diverse geographical area, data was collected from a total of 26 village banks operating in Bolivia, Burkina Faso, Colombia, Guatemala, Honduras (two programs, one rural and one urban) and Uganda. Still, the sampled programs represent only one or two sites of the participating agencies` larger program portfolios. In many cases, the program information is several years old and program policies have evolved and adapted since the
Village Banking Dynamics Study: Evidence from Seven Programs 3

reporting period. In recognition of this lag-time and to encourage openness and objectivity, programs are only identified in this paper by geographic areas rather than by implementing agency. The box below indicates membership numbers in each of the programs as of December 1998. This information is included to convey the scope each program has achieved since the data was originally collected in 1994-95. The original data set included records from approximately 700 individual members. Box 2: Number of Active Members by Program as of December 1998 Bolivia 12,863 Burkina Faso 5,098 Colombia 2,645 Guatemala 8,241 Honduras-urban 14,865 Honduras-rural 4,200 Uganda 17,228 Total for the seven programs: 65,140 Village bank records were sampled to provide information about member loans, savings, internal borrowing and participation for selected loan cycles (first, third, fifth and last completed loan cycle) using a protocol originally developed by Freedom from Hunger. By collecting information at the individual borrower level, insight into the intra-village bank dynamics is possible. By recording the names and program details for all borrowers, it was clear which of the original borrowers remained in or left the program, how many new members joined after the first loan cycle, and how many members left but later rejoined the program. This provided an historical record of individual client experience, participation, loan size, and how each of those factors relate to savings and internal loan borrowing, as well as how these change over time for the individual client. The protocol also included questions to be used by program staff to promote informal group discussion among clients about key factors influential to borrowing and program participation. In order to conduct a comparison of actual program dynamics to the original Hatch model, the village banks in the study needed to have reached a certain maturity (of 6 cycles). For this reason, programs were typically sampled for the study from among the organizations` oldest village bank groups. Still, at the time the information was collected the programs were still relatively young and in operation for only approximately two to four years. While all seven programs collected information at the first, third and fifth loan cycles, due to variation in village bank maturity the cycle number of the last completed cycle varied quite a bit (ranging from 6-10 cycles). Across all of the village banks sampled, the last completed loan cycle occurred on average approximately 35 months into the program. This average was calculated across village banks at each program level and then across the seven programs for a cumulative average. In addition, because programs had different loan-cycle lengths (4- or 6-month cycles), information is presented by number of months in the program at the end of the loan cycle, as well as the loan cycle number, for uniform comparison purposes. For example, cumulative averages across the seven programs at cycle 1 correspond to month 5 since some of the programs had a 4-month and
Village Banking Dynamics Study: Evidence from Seven Programs 4

others a 6-month first loan period.

FINDINGS
The study explores five major factors found to affect loan-size growth: Loan policies and practices Village bank location Savings policies Membership participation rates Internal account policies The first section presents information on the pattern of actual loan growth for the seven programs relative to the original Hatch model. Subsequent sections examine how policies, setting and membership dynamics were related to the loan growth rates in various programs.

LOAN GROWTH PATTERNS: What is the actual pattern of loan growth versus the Hatch model? On average, loan size across all seven programs did not stagnate but increased steadily, although at a rate lower than the original Hatch model projections. On average, the loan growth in Graph 1 shows a steady increase across the seven programs, although at a rate below the original model. For the first year, loan sizes were actually above the Hatch modelby the third cycle borrowers were at an average loan size of $104 as compared to the $82 originally projected. However, by the fifth loan cycle (22 months in the program), the $136 average loan size per borrower was only 75% of the projected average of $182. By the last completed loan cycle (35 months), the disparity between programs and the model grew even larger, with the actual average loan size being $173 as compared to $300, only 58% of the projected average loan size. Thus the per-borrower loan-size growth rate over approximately three years was 280% as compared to the more ambitious 500% rate of the original model. In general, however, loan-size stagnation is not evident across the seven village bank programs since loan sizes continue to climb albeit at a rate slower than projected by the original model.

Village Banking Dynamics Study: Evidence from Seven Programs

Gra ph 1: Ave ra ge P rogra m Loa n P e r Borrow e r for a ll S e ve n P rogra m s vs. Origina l Ha tch M ode l P roje ctions
$300 $300 Original Hatc h Model A v erage f or 7 Programs (all borrow ers ) $182 $150 $104 $100 $62 $50 $50 $82 $136 $173

$250

Average Loan ($US)

$200

$0 Firs t Loan (av g. 5 mos .)

Third Loan (av g. 14 mos .)

Fif th Loan (av g. 22 Las t Reported Cyc le mos .) (av g. 35 mos .)

Loa n Cycle (Ave ra ge M onth tha t the Cycle Ende d)

Diversity in Average Loan Sizes by Program There was tremendous variability in average program loan growth rates and patterns across the seven sampled programs.

The averages presented in Graph 1 mask the tremendous variability in loan-size growth across the seven programs. As seen in Graph 2, two of the seven programs actually experienced relatively rapid and high loan growth rates (Uganda and Colombia). One program, Guatemala, had high loan growth but over a longer four-year period; one program had modest but steady growth (Burkina Faso); and three had relatively flat loan-size patterns (Honduras-urban, Bolivia, and Honduras-rural).1 Only two of the seven programs did not exhibit increases in average loan sizes for the last completed loan cycle (Uganda and Honduras-urban). The Uganda program had a relatively rapid and steep increase in loan size and then dropped in the second year, whereas the Honduras-urban program had relatively good loan growth in its first two years but flattened out or stagnated during months 20-42.

The Bolivia and Honduras-rural programs show a much more dramatic increase in loan size when the total borrowing (external plus internal loans) is taken into account. (See Graph 6.) 6

Village Banking Dynamics Study: Evidence from Seven Programs

Graph 2: Average Loan Size by Program vs. Original Hatch Model

$325 $300 $275 $250 $225 $200 $251 Colombia $222 Uganda $300 Original Hatch Model $297 Guatemala

$US

$175 $150 $125 $100 $75 $50 $25 $0 12 18 20 24 27 28 30 32 33 34 38 42 48 4 6 $82 Bolivia $74 Honduras-rural $159 Burkina Faso $128 Hondurasurban

Months in the Program

KEY FACTORS INFLUENCING LOAN SIZE Program policies and implementation decisions regarding loan-size eligibility, location, internal accounts and saving-access all influence loan size, as do membership participation rates of sustained borrowers and late joiners. Village banking programs typically select a maximum initial loan size and a loan-size ceiling beyond which no individual client can borrow regardless of how long they have been in the program. Programs also have various formulas by which individual clients can become eligible for larger loans. The specifics of the loan policies as well as ancillary services such as internal loans and savings are factors which influence loan-size growth rates. In addition, the commercial development and opportunity in the program area can influence a borrowers ability to use larger loans. Finally, the membership dynamics of longer-term clients leaving the program and/or new members joining can also serve to dampen loan-size growth rates. In order to facilitate comparison of how program policies and context might influence loan growth rates, the seven programs in the study were divided into three groupings based on thirdyear average individual loan-size data from each village bank (see Table 1). The high-range category (with an average loan size of more than $200) consists of the Colombia and Uganda programs; the mid-range category (with an average loan close to $150) consists of Hondurasurban and Guatemala; and the low-range category (with an average loan of less than $100) consists of Burkina Faso, Bolivia and Honduras-rural. Programs are compared by initial and maximum loan-size policies, program context, loan-growth policies and internal account loan policies.

Village Banking Dynamics Study: Evidence from Seven Programs

Table 1.Average Per Borrower Categories for Third Year (24-36 months) Program Program Maximum Highest Loan Ceiling Location Initial Loan Actual Loan Policy High-Range - Average Loan Size More than $200 Colombia - $251 urban $80 $488 $500 at 24 mos. at 24 mos. Uganda - $222 urban $75 $1830 part of time no at 28 mos. at 20 mos. ceiling; changed to $600 max. Mid-Range - Average Loan Size Close to $150 Guatemala rural $75 $1413 up to $1,000 $139 at 30 mos. at 54 mos. Honduras - urban urban $60 $439 $300 + max. Estimated $141at at 20 mos. 31 mos. Low-Range - Average Loan Size Less than $100 Burkina Faso 50% rural $80 $352 $370 max. $92 at 27 mos. 50% urban at 29 mos. Bolivia $82 rural $50 $202 $300 max. at 33 mos. at 32 mos. Honduras - rural rural $50 $250 $300 max. $74 at 32 mos. at 40 mos.

Internal Loan Activity low low

medium none

none high high

The following sections will discuss how the program location, loan ceiling policy and internal loan activity relate to loan size. Loan-Size Policies Initial Loan Size If consistently applied, the initial loan-size ceilings reduce average loan size, especially for programs with a large influx of new members joining after the first loan cycle. In five of the seven programs, initial loan sizes are higher than the $50 projected by the original model. The range of initial loan sizes span from $50-$80 across the programs. Therefore, it is not surprising that most programs equal or exceed original model projections in the early cycles. Also influential to loan growth rates was whether the initial loan-size policy was applied to firsttime borrowers joining after the first loan cycle. For the Guatemala program35% of the new entrants in the last reported cycle received loans greater than $100 even when the maximum initial loan was set at $75. The practice of allowing initial loans above $100 raises the overall average in the later months for these two programs. In contrast, the Uganda program experienced a sharp drop in average loan size in the last completed loan cycle in part due to a large influx of new borrowers who primarily received initial loans which reflected the initial loan ceiling of $75. Thus, if consistently applied, the initial loan-size policy can reduce the overall average loan size
Village Banking Dynamics Study: Evidence from Seven Programs 8

of a program with a large influx of new members who have smaller loans. Maximum Loan Size Programs with average loan sizes closer to the amounts predicted by the original Hatch model had approximately 30% of their clients taking loans greater than $300. Program loan ceilings in some cases were non-existent or were as high as $1,000 per borrower. Programs with the lowest average loans consistently applied a $300 loan ceiling. The maximum loan size or loan ceiling policy was even more influential to the patterns of loansize growth in the sampled programs. While the model assumes all loans constitute poverty lending levels of $300 or less, the three programs with the highest average loan sizesUganda, Colombia and Guatemalahad loan ceiling policies well above this amount. Table 2 shows the loan ceiling policy and the percentage of clients in each program receiving loans above $300 in the fifth and last reported cycles. It is interesting to note that in three (Uganda, Guatemala and Honduras-urban) of the seven programs, borrowers were allowed to take loans larger than the programs stated loan ceiling. The relatively sharp climb in loan size during the first 20 months of the Uganda program is related to the fact that initially the program applied no loan ceiling. Consequently, some very high individual loans of up to $1,800 were made. The later introduction of a maximum loan size of $600 and a major decline in retention of original borrowers between the fifth and last reported cycle contributed to a decline in the average loan size after the 20th month.2 Still, it is important to note that approximately 30% of the borrowers in the fifth and last completed loan cycle took loans in excess of $300. The Guatemala program applied a loan ceiling of $1,000 per borrower. Although loans rarely reached that size, there were three cases where the loans were above or near $1,000. Similar to the Uganda program, by the fifth loan cycle, close to a third of the loans were more than $300. However, Table 1 shows that unlike Uganda, only 7% of Guatemala program loans were above $300 in the last reported cycle. One factor that may account for this is the fact that the last reported cycle of Guatemala occurred much later. The last cycle represented over four years of program participation and by that time most of the borrowers who took large loans in the fifth loan cycle either borrowed less or had left the program by the last report cycle. Like Uganda and Guatemala, Colombias high loan ceiling of $500 resulted in a higher average loan size. Graph 2 illustrates the Colombia programs rapid loan-size growth in a single loan cycle. The average loan size jumped from $147 at 20 months to $251 at 24 months. A single village bank explains much of this growth with almost half of the loans over $300. The urban program area, high loan ceilings and client ability to save fueled this rapid loan growth.

The Uganda program also changed its loan eligibility formula from the original models last loan plus total savings to first loan plus total savings in an effort to better control loan amounts following a period of repayment problems.

Village Banking Dynamics Study: Evidence from Seven Programs

Three programs had a policy of limiting loans to a maximum of $300. However, it seems this policy was not consistently applied in the Honduras-urban program since some individuals borrowed as much as $439. The two programs that did consistently apply the $300 povertylending ceilingHonduras-rural and Boliviaalso had the most modest program growth rates. However, as will be discussed in the section on internal borrowing, very active borrowing from the internal account also dampened loan growth rates for both these programs. Table 2: Loan-Ceiling Policies and Loan Sizes by Program Largest % Borrowing Program Loan Ceiling Individual More Than Policy Loan Made $300 in Cycle 5 (avg. 22 mos.) High-Range Average Loan Size Uganda Colombia
Initially no loan ceiling, later ceiling of $600 $500 $1,830 at 20 mos. $488 at 24 mos. $1,413 at 54 mos. (above ceiling) $439 at 20 mos. $352 at 32 mos. $202 at 32 mos. $250 at 40 mos. 31%

% Borrowing More Than $300 in the Last Reported Cycle (avg. 35 mos.)

29%

0%

14%

Mid-Range Average Loan Size Guatemala Honduras urban

$1,000 28% 7%

$300

9%

10%

Low-Range Average Loan Size Burkina Faso Bolivia Honduras rural

$370 $300 $300 6% 0% 0% 6% 0% 0%

Loan-ceiling policies are partly a reflection of the economic demand and activity of the program area. However, the fact that only two programs have a poverty-lending ceiling of $300 begs the question of whether a certain portion of the loan portfolio must be allocated to non-poverty lending in order to achieve the original loan growth rate projections.

Village Banking Dynamics Study: Evidence from Seven Programs

10

Rural vs. Urban Location and Average Loan-Size Growth The urban programs had an average loan size of $205 at 28 months as compared to only $98 at 32 months for the three rural programs. An urbanization trend is apparent in village banking worldwide as more programs move to urban and peri-urban centers to facilitate fast portfolio growth and expansion. As might be expected, the level of commercial economic activity in the program area appears to impact individual loan size. All of the village banks in the lowest average loan-size category by year three were in rural areas. Rural banks in Burkina Faso had an average loan size of $92 by 27 months, Bolivia had $82 at 33 months and Honduras had $74 at 32 months. The majority of village banks in the sample operate in urban or peri-urban areas3 and are located in capital cities, district headquarters or towns. The urban programs (Colombia, Honduras-urban and Uganda) had an average loan size of $205 at 28 months as compared to only $98 at 32 months for the three rural programs (Bolivia, Guatemala and Honduras-rural). The dampening effect that less vibrant economic activity can have on loan size is even more evident among the rural banks in Bolivia and Honduras; those in less commercially developed areas have average loan amounts of two-and-a-half times less than those in more developed areas (MkNelly and Stack 1996). This disparity was evident in the Honduras-rural program where none of the four village banks sampled were located in a town. For example, in a relatively small community of only approximately 100 households located along a rarely traveled and difficult-topass dirt road, the average loan size of a village bank in the seventh cycle was only $38, whereas in a relatively larger agricultural community with several shops located relatively closer to more active commercial market centers, the average loan size in the eighth cycle was $113. An urbanization trend is apparent in village banking worldwide as more programs move to urban and peri-urban centers to facilitate fast portfolio growth and expansion (Nelson, et al. 1996). The programs in the study sample mirror this trend, where approximately half of the sampled village banks were located in urban or peri-urban areas. While lending in such locations is associated with faster loan growth rates, this tendency has implications to program outreach and servicing of the poorest rural households who have even less relative access to formal financial services. Locating banks in areas with more economic or commercial activity results in faster loan-growth rates and higher average loans. Rurally oriented programs face the double challenge of working with clientele more costly to reach and who may have lower or inconsistent credit needs. Of course, one of the original goals of village banking programs was to provide rural households with improved access to credit. An overemphasis on the need for rapid loan-growth rates and financial sustainability are likely to shift the focus away from the less profitable rural operations to areas with more active economic development. Still, practitioners realize there is a limit to how remote and commercially undeveloped an area can be and still support a village banking program. Borrowers are unlikely to productively and successfully use a working capital loan, particularly for non-farming activities which earn steady returns, unless there is some degree of market activity and commercial development in their area. SAVINGS:
3

Urban or peri-urban is defined here as in or near a town center with a concentrated population of 5,000 or more. 11

Village Banking Dynamics Study: Evidence from Seven Programs

What is the savings rate in practice versus the Hatch model? Like average loan-size, savings rates fall short of the original Hatch model by the third year. Graph 3 illustrates that savings rates, like loan rates, fall short of the original Hatch model by the third year. Although in the first 22 months the savings rate closely follows or exceeds the model, by month 35 the average per-borrower savings is $139 as compared to the projected $335. Of course, since village bank programs no longer aim to graduate village banks after three years, there is less of an imperative to require borrowers to save $300 to replace their ninth and final program loan. Still, most village bank programs link borrowers loan-size eligibility to the amount of savings they have on deposit with their village bank. For this reason, it is possible that clients ability or inability to save could act as a brake on loan-size growth rates. Graph 4 shows that per-borrower savings amounts vary considerably across the seven programs.
Graph 3: Average Program Savings for all Seven Programs (total 26 village banks) Relative to Original Hatch Model
$350 $300 $250 $335

$US

$200 $150 $99 $100 $56 $50 $38 $0 $88 $139 Hatch Model Average for 7 Programs

Third Loan (avg. 14 mos.)

Fifth Loan (avg. 22 mos.)

Last Reported Cycle (avg. 35 mos.)

Loan Cycle (Average Month that the Cycle Ended)

Village Banking Dynamics Study: Evidence from Seven Programs

12

Graph 4: Average Amount of Savings by Program

$350 $335 Original Hatch Model

$300

$250

$252 Guatemala

$US

$200 $179 Uganda $150 $155 Bolivia $163 Honduras-urban

$100

$99 Colombia $70 Honduras-rural $57 Burkina Faso

$50

$0 12 18 20 24 27 28 30 32 33 34 36 38 42 48

Months in the Program

Table 3 groups programs by high, medium and low average savings in year three. The Uganda program had the highest average savings per borrower ($179) which at 28 months exceeded the amount projected by the original Hatch model. It is interesting to note that two of the programs in the highest savings groupingBolivia and Honduras-urbanwere categorized as low- to midrange in terms of average loan size. The Bolivia program, in particular, had a high savings-toloan ratio of 189%. Both the urban and rural village banks in Honduras also had relatively high savings-to-loan ratios approaching 100%.

Village Banking Dynamics Study: Evidence from Seven Programs

13

TABLE 3: Third-Year Savings Amounts, Ratios and Policies by Program

HighestRange Savings $130 or greater Uganda $179 at 28 mos. Savings/Loan Requirements Access Formula Return

81%

20% required

resign

Bolivia $155 at 33 mos. Hondurasurban $130 at 31 mos. Mid-Range Savings close to $100 Guatemala $116 at 30 mos. Colombia $99 at 24 mos. Low-Range savings less than $100 Honduras-rural $70 at 32 mos. Burkina Faso $33 at 27 mos.

189% 99%

$0.25-$1/wk 20% encouraged

resign emergency resign

Was last loan plus total savings changed to first loan plus total savings last plus total savings last plus total savings

dividend

dividend no

83%

$1.5/wk

discouraged withdrawal resign

last loan plus total savings last loan plus total savings

dividend

39%

10% required

dividend

96% 36%

$0.25-$1/wk 10% required

resign emergency resign

last loan plus total savings last loan plus total savings

no no

In the original model, the savings-to-loan ratio was projected to be nearly one-to-one, or 100%, and by the eighth cycle with savings at $275 and the loan capped off at $300. While most of the programs had relatively high savings-to-loan ratios (above 80%), those with the smallest average loan sizeHonduras-rural, Honduras-urban and Boliviaare among the highest at 96%, 99% and 189%, respectively. These ratios indicate that many members can reach the point where they could capitalize themselves and graduate from the program. Yet many clients continue to borrow because they value building their savings while at the same time having access to credit. The policy of linking loan-size eligibility to a borrowers savings does not seem to limit loan-size growth rates. The three programs with the relatively lowest average loan sizesBolivia, Honduras and Burkina Fasoeach had savings-to-loan ratios that indicate borrowers would have been eligible for even larger loans than they actually took.

Factors Affecting Savings Rates among Programs and Savings Effect on Average Loan Size
Village Banking Dynamics Study: Evidence from Seven Programs 14

Savings requirements and policies linking loan-size eligibility to savings do not hinder loan growth directly or even show close correlation to savings rates. However, restricted access to savings may indirectly affect loan growth by causing member resignation. Dividend payments based in part on borrowers savings and liberal access to internal loans is related to very high savings-to-loan ratios even when savings requirements are relatively small. By the third year, the savings-to-loan ratio of all seven programs was higher than the 20% required by the original Hatch model. However, the correlation between average loan size and average savings is not as direct as might be expected. For example, the Colombia program had relatively high loan sizes but only mid-range savings. And the Bolivia program had the lowest average loan size but relatively high average savings. Program policies pertaining to savings requirements, access and returns are factors that are likely to affect savings amounts and ratios. All of the programs linked borrowers loan size eligibility to their savings. Most applied the original Hatch formula that allows clients next loan to be up to the amount of their last loan plus their accumulated savings. The only exception is the Uganda program which after some time revised its loan eligibility policy to allow loans only equal to clients accumulated savings plus the amount of their first rather than their last program loan. This requirement might in part explain their relatively higher savings-to-loan ratio. The programs applied a range of savings requirements from 10% to 20% of the amount borrowed. Not surprisingly, the programs with a 20% requirement (Uganda and Hondurasurban) had a higher savings-to-loan ratio than those requiring 10% of the outstanding loan (Columbia and Burkina Faso). However, the program in Bolivia only required members to save the local equivalent of $0.25-$1.00 per week (which over time works out to be less than the 20% required by many of the other programs), yet it had the highest savings rates. For the programs included in the study, the savings incentives inherent in program policies were more influential to savings rates than were the stated savings requirements. Two of the programs with relatively high savings rates, Bolivia and Honduras-rural, had very active internal accounts which served as strong financial incentives to save. During each loan cycle, virtually all of the member savings were being lent at interest rates higher than the program loan rate. Therefore, in addition to providing access to supplementary loans with more flexible terms, active internal accounts provided a very tangible economic incentive to savers in the form of dividend payments. Village bank members were paid a dividend at the end of the loan cycle based partly on the amount of their savings. Over time, these dividend payments could become quite significant and members saw them as offsetting the interest they paid on their program loans. Liberal internal account access, along with guaranteed returns through dividends, culminated in high savings-to-loan ratios in those programs which offered internal loans. Access to savings for all programs except Guatemala was limited to cases of emergency or member resignation. Burkina Faso program managers specifically mentioned that the policy of not allowing clients access to their savings except upon resignation seemed to encourage people to leave the program in later cycles. This issue of access played an indirect role on average loan amounts because members had to resign in order to use their savings. High savings-to-loan requirements might also indirectly dampen loan and savings amounts if over time these policies
Village Banking Dynamics Study: Evidence from Seven Programs 15

encourage clients to leave the program. As explored in greater detail in the participation section of this paper, the Guatemala program-allowed savings access in later cycles and had a relatively smaller dropout rate of original borrowers. The economic environment of the program can also be very important as is seen in the case of Burkina Faso. There, the low average amount saved was partly due to the large devaluation of local currency that took place during the study period. In local currency, the amount of savings had increased by 77% between the third and fifth cycles, and 67% between the fifth and last reported cycles (Graber, 1997). However, because local currency amounts were all converted to dollar values at contemporary exchange rates, the average dollar amount saved appeared relatively modest and was among the lowest of the seven programs. INTERNAL ACCOUNT POLICIES: Liberal internal lending policies did detract from external account loan growth in three programs. This phenomenon was most evident for programs when external loan repayment installments were included as part of the internal fund. How does internal account lending affect the average loan size (of the external account)? Practitioners approach internal account lending with both excitement and trepidation. On one hand, it offers an additional loan pool with flexible terms that is member-managed and funded, thereby enhancing opportunities for income-generation, business skills, solidarity and empowerment. On the other hand, it has proven to be full of financial management pitfalls and periodic fraud, requiring extensive promoter time without compensation to the program because interest income earned on internal loans is disbursed as dividends or transfers to group equity rather than accruing to the program. The internal account is commonly comprised of savings, interest income earned on the lending of savings, and fees and fines collected by the group. For some programs, the internal account also includes external account loan payments that stay in the village bank until the end of the loan cycle when they are repaid to the implementing organization. Including external loan repayment installments as part of the internal fund dramatically increases the amount of capital available for internal lending. This mixing of external payments with internal lending during the cycle is part of the original village-banking model. However, this practice is no longer very common since programs usually require repayment installments to be deposited in institutions outside the village bank (either retained by the program or held by local financial institutions). Internal account loans were originally conceived as a source of complementary short-term loans which could serve consumption and emergency needs or supplement external loan amounts for those borrowers whose working capital needs exceeded the amount allowed by the program. In this scenario, the internal account does not compete with the external account. The same is true when the internal account is lent mainly to individuals outside the village bank. The original long-range purpose of the internal account supplanting the external account through graduation is no longer followed. Therefore, programs are carefully looking at their internal account policies and how they will affect the program goals of poverty alleviation and financial sustainability.
Village Banking Dynamics Study: Evidence from Seven Programs 16

Graph 5 shows that internal account lending relative to external borrowing was only significant in three programsBolivia, Honduras-rural and Guatemala. The Bolivia and Honduras-rural programs had relatively low average loan sizes but relatively large internal loans per borrower. Among the programs surveyed, unrestricted access to internal account loans seems to have detracted from external account activity, lowering external account loan size (MkNelly, Stack, 1996). By the end of the third year, Bolivias average internal loan was $129 compared to the external loan of $82. Because members had ready access to internal as well as external loans, a true picture of borrower demands for and use of loan capital requires examination of the total amount of loans borrowers took (internal plus external loans).
Graph 5: Internal and External Loans by Borrower
$525
$156

$450

Internal
$375

External

$300

$US

$11

$7 $7 $60 $129

$225
$29 $4 $270 $86 $51 $251

$69 $296 $222 $151

$150

$75

$141

$147 $68 $68

$74

$82

$-

Guatemala Colombia 18 mos. 20 mos.

Uganda 20 mos.

Bolivia 20 mos.

Honduras 20 mos.

Colombia 24 mos.

Uganda 28 mos.

Guatemala Honduras 30 mos. 32 mos.

Bolivia 33 mos.

Guatemala 48 mos.

M onths in the Program

Comparing the average total loans (external plus internal) by program to the original Hatch model reveals an even more pronounced pattern of loan growth rate over time. In Graph 1 and Table 1, the external loan growth for the Bolivia and Honduras-rural programs was relatively modest and flat, but the total loan amounts, $211 and $165 respectively, are much closer to the original model projections. The Guatemala program also had relatively active internal borrowing and a total loan that was 40% greater than the average program loan. In Graph 6, only one of the seven programs (Honduras-urban) exhibits stagnant loan growth.

Village Banking Dynamics Study: Evidence from Seven Programs

17

Graph 6: Total Loan Size (External + Internal) by Program vs. Original Hatch Model
$500 $450 $400 $350 $300 $300 Original Hatch Model $258 Colombia $229 Uganda $211 Bolivia $143 Honduras-rural $159 Burkina Faso $128 Honduras-urban $452 Guatemala

$US

$250 $200 $150 $100 $50 $0 12 18 20

24

27

28

30

32

33

34

38

42

48

Months in the Program

Table 4 groups the programs with active internal lending into high-, middle- and low-range groups on the basis of internal-to-external loan ratios. Table 4 Internal Loan Activity for Third Year and Policies (24-36 months)
Program Average Internal Loan $129 Average Program Loan $82 Average Total Loan $211 Internal Loan as a % of External Loan 157% Internal Loan Policies
Unrestricted internal loans including repayment installments Borrowers prefer lump sum rather than weekly repayment Borrowers prefer receiving dividend from internal interest rather than pay interest to program Majority of internal loans made to village bank members Unrestricted internal loans including repayment installments of principal and interest Majority of internal loans made to village bank members Lump sum rather than weekly repayment Dividends to all savers Unrestricted internal loans only made from members savings Internal lending to members and nonmembers Lump sum rather than weekly repayment Dividends disbursed to all savers No internal lending before cycle 3 Never above 5% of external loan amount and village bank must maintain 30% cash reserve High performance standards in external account required first; cap at 25% of total per loan All internal borrowers must save with village bank Dividends disbursed to all savers Program limits amount of internal lending that is allowed Very little internal lending to members. Focus on making internal loans to non-members Weekly repayment of internal loan required Dividends disbursed to all savers

High-Range - Internal Loan More than 100% of External Loan

Bolivia at 33 mos.

Mid-Range - Internal Loan from 40-75% of External Loan

Honduras -rural at 32 mos. Guatemala at 30 mos. $69 $74 $143 93%

$60

$151

$211

40%

Low-Range - Internal Loan below 5% of External Loan

Uganda at 20 months (data not available for 28 mos.) Colombia at 24 mos. $11 $270 $281 4%

$7

$251

$258

3%

Village Banking Dynamics Study: Evidence from Seven Programs

18

Key Factors Affecting the Degree of Internal Account Lending Aspects of the internal account which borrowers find attractive is instructive to village banking practitioners. In those programs where internal and external loan terms differ, clients preferred the lump sum rather than weekly repayment schedule. The requirements of weekly repayment and weekly meetings over time act as disincentives to program participation and increased borrowing from the external account. The major factor affecting the degree of internal account lending is certainly the program policies, which dictate borrower access to internal loans. There is considerable range in the internal loan policies of the surveyed programs. Two programsHonduras-urban and Burkina Fasoallowed no internal account lending. At the other extreme, the Bolivia and Hondurasrural programs allowed unrestricted lending of the internal account, which included repayment installments on the external loans. The Guatemala program also allowed unrestricted internal lending of member savings and group funds but not external repayment installments. The Colombia and Uganda programs had more restrictive and less flexible internal loan policies. The relatively high internal loan amounts seen in the Bolivia and Honduras-rural programs can also be explained by the fact that only with these two programs did the external loan payments flow through the internal account. Because financial institutions are lacking in rural Bolivia and rural Honduras, deposit-taking institutions are not readily available and village banks there disburse all internal account funds at each meeting as a form of safekeeping. With the other programs in Colombia and Uganda, external account repayment installments were paid directly to the program on a regular basis rather than held by the village bank members. In the case of Guatemala, external account was collected in one balloon payment at the end of the cycle, but no external payments flowed through the internal account during the cycle. This blending of external payments into the internal account is the main distinguishing factor behind the high internal account rate of Honduras-rural and Bolivia. The programs with more active internal lending also offered terms different from external loans which borrowers found attractive. The Bolivia, Honduras-rural and Guatemala programs all allowed borrowers to repay their internal loans (principal plus interest) in a lump sum rather than regular weekly installments. So, even though the interest rate for internal loans is typically higher than that for external loans, the effective interest is likely to be lower. Borrowers also appreciated the fact that the interest they paid on internal loans accrued to the village bank itself and was paid out to members as dividends. In some cases, borrowers also felt that the approval process or discussion of loan feasibility was less rigorous for internal loans, giving them greater flexibility. Internal loan activity in Colombia and Uganda is minimal with negligible impact on external account loan size, with the exception of one bank in Uganda. Both programs have restrictive internal account loan policies that make borrowing less attractive to members and limit the availability of internal funds. Colombia allows only 50% of group savings to be lent. Uganda allows lending only after the third loan cycle and requires high performance standards before internal account lending can begin. The program also requires each village bank to hold a cash reserve equal to 30% of the outstanding loan. Internal account loan amounts are restricted in
Village Banking Dynamics Study: Evidence from Seven Programs 19

Uganda in that any one person can borrow no more than 25% of total funds available. Colombias repayment schedule mirrors the external account loan schedule and requires weekly repayment. Uganda and Colombia require all loans to be paid off by the end of the cycle. In addition, Colombia targets non-members for internal account loans, whereas Bolivia gives members priority. Uganda also lends to non-members. However, both Uganda and Colombia disburse dividends to savers. The sum of these policies decreases flexibility and therefore the attractiveness of internal account loans to clients. PARTICIPATION RATES: Sustained borrowers make up only 65% of the membership by the end of the first year, approximately 50% by the end of the second year and only 35% by the end of the third year. Inconsistent borrowing and late joiners dampen loan-size growth. The average loan size of sustained borrowers was approximately 25% higher than the loan-size averages for the general village bank membership at the end of the third year. What are the membership dynamics in practice as compared to the Hatch model? The retention rate of sustained borrowers, those who consistently borrowed from the first to the last reported cycle, has a critical impact on the average loan size of a village bank. Graph 7 plots the average loan size of sustained borrowers relative to that of all borrowers as well as to the original Hatch model projections. By the end of year three, the sustained borrowers average loan size was $218 as compared to $173 for all borrowers and was 73% of the model projection as compared to only 60% for all borrowers.
Graph 7: Average Program Loan for All Borrowers and for Sustained Borrowers from the First Loan Cycle Relative to Original Hatch Model (all 7 programs)
$300
$300

Average Loan ($US)

Original Hatch Model

$225
$182

$218 $173 Average for 7 programs (all borrowers) Average for 7 program (sustained from first cycle)

$150
$108 $62 $62 $50 $104 $82

$158 $136

$75

$0 First Loan (avg. 5 )

Third Loan (avg. 14 mos.)

Fifth Loan (avg. 22 mos.)

Last Reported Cycle (avg. 35 mos.)

Loan Cycle (Average Month that the Cycle Ended)

The drop-out of longer-term consistent borrowers and the drop-in of new borrowers have the same effecta reduction in average loan size. Members who borrowed and saved for five
Village Banking Dynamics Study: Evidence from Seven Programs 20

consecutive cycles had an average loan size of $158 or higher (see Graph 7). Losing these members represents a loss of a loan size that was achieved over two years. New members joining after the first loan cycle also reduce the average loan size. First-time borrowers commonly start at the initial maximum loan size which averages $62. Thus, a program with a low resignation rate but a high late joiner or drop-in rate will also experience a dampening of average loan-size growth. While expansion is in general a positive program impact, it does initially result in a lowered loan-growth rate. Looking across all seven programs, membership turnover (drop-out and drop-in) reduced the average loan size by about 24% by the end of the third year. The Hatch model projects membership starts at approximately 30 and increases to no more than 50 after the first few cycles (Hatch, 1989). While it is expected that new members will join in later cycles, the model does not account for the financial implications of original members leaving the program. Graph 8 illustrates this impact by comparing active borrowers to sustained borrowers on a per-bank level. Average membership per bank remains around 28 while the number of members who joined in the first cycle decreases to 10, or about one-third of the original membership by month 35. The most dramatic exit of first-cycle borrowers occurs in the village banks first year. Graph 8: Average Number of Active Borrowers per Village Bank vs. Number of Sustained Borrowers from the First CycleAverage for the Seven Programs
35 Average Number per Village Bank 30 25 20 15 10 5 0 First Loan Third Loan Fifth Loan Last Reported (avg. 5 mos.) (avg. 14 mos.) (avg. 22 mos.) Cycle (avg. 35 mos.) Loan Cycle (Average Month Cycle Ended) 18 15 10 27 28 29 28 Sustained Borrow ers (from 1st cycle) Active Borrow ers

By the end of the first year, 65% of the first-cycle borrowers were still active in their village bank and had consistently borrowed in each of the first three loan cycles (see Graph 9). Near the end of year two, 53% of the borrowers had consistently borrowed since the first loan cycle. This number dropped to 35% by the end of year three. It is not possible to say whether these represent relatively high or low rates of sustained borrowing as compared to other credit methodologies since little empirical evidence is available to make such a comparison. In any case, these
Village Banking Dynamics Study: Evidence from Seven Programs 21

membership dynamics run counter to the original models assumption that clients will demand steadily increasing loans every four to six months over a three-year period.

Village Banking Dynamics Study: Evidence from Seven Programs

22

The largest decrease in sustained borrowers takes place within the first 14 months of joining the program. One theory behind this is that the first three cycles are a period of weeding out as clients come to better understand the requirements of the system and the premium placed on ontime repayment. Members with other options for acquiring working capital, perhaps the wealthier, may decide to resign completely or suspend loans temporarily. These causes of resignation are addressed more fully in a later section of this paper. Despite this lack of consistent clientele, overall program loan growth rates do show steady increases for most of the programs. Also evident in the data were those individuals who enter and exit the program numerous times for various reasons. Some of these individuals had joined as of the first loan cycle, left in a later cycle and then rejoined. Others joined after the first loan cycle and followed a similar pattern of leaving and returning in later cycles. These intermittent borrowers represented only 10% of the entire sample and therefore did not have a significant effect on the overall trends. However, this dynamic of suspending borrowing for one or more cycles at a time does have programmatic implications. Practitioners should be aware of this element in planning projections, and program loan policies should accommodate inconsistent borrowing while not penalizing the member.

Graph 9: Percentage of First-Cycle Borrowers Who Are Still Active Average for All Seven Programs
Percent of 1st Cycle Borrowers
120% 100% 100% 80% 65% 60% 40% 20% 0% First Loan (avg. 5 mos.) Third Loan (avg. 14 mos.) Fifth Loan (avg. 22 mos.) Last Completed Loan Cycle (avg. 35 mos.) 53% 35% Average Across 7 programs

Loan Cycle (Average Month that the Cycle Ended)

Members Who Join After the First Cycle The influx of new members joining after the first cycle is equal to or greater than the number of borrowers leaving. The large influx of new members in later cycles should signal program success, although one outcome is a reduced average loan portfolio. The number of clients who enter in later cycles may be as significant to loan-size growth as the number of those who leave. On average, village banks across all of the programs grew in size.
Village Banking Dynamics Study: Evidence from Seven Programs 23

This phenomenon of late joiners is consistent with Hatchs original theory that those more riskaverse (usually the very poor) will join after having the opportunity to observe the program for some time. The risk takers (usually those with more income) tend to join first, but also leave quicker if the requirements of the village bank are too taxing on their time. Discussions with late joiners in the Honduras-rural program revealed that some had not heard about the program. Some did not join at the time because they had young children, and others were afraid to join because they did not fully understand how the program worked or they were hesitant to assume a loan. Graph 10 shows a rapidly increasing percentage of active borrowers who joined after the first cycle. Decreasing sustainers and increasing drop-ins (members who join after the first cycle), yielded a generally consistent membership level of 28 as seen in Graph 8.

Graph 10: Percent of Active Borrowers who Joined After the First Loan CycleAll Seven Programs
100% 80% 71% 54% 38%

Percentage

60% 40% 20% 0% Third Loan (avg. 14 mos.)

Fifth Loan (avg. 22 mos.)

Last Completed Loan Cycle (avg. 35 mos.)

Loan Cycle (Average Month that the Cycle Ended)

Sustained Participation and Average Loan Size Among Individual Programs If client drop-out were a major cause for slow loan-growth rates, programs with high rates of sustained borrowing should have relatively high average loan sizes. However, this relationship was not consistently found among the programs included in the study. The data does, though, suggest why some programs may have higher sustained membership. The program with the highest sustained borrower rate (Guatemala) had relatively flexible program policieshigh loan ceilings, liberal access to internal loans and lump-sum rather than weekly repaymentthat may have been sufficient incentive to keep borrowers active in the program. Programs in Colombia, Guatemala and Burkina Faso had the highest retention rate (40% or higher) of first-cycle borrowers in the third year of operation. These programs represented all three different average loan-size categories: high (Colombia), medium (Guatemala) and low (Burkina Faso).4 Those with the lowest retention rate, Honduras-rural and Bolivia, show higher savings rates and relatively low program loans but mid-range total loan amounts. Table 5 below
4

As mentioned earlier, the devaluation of the FCFA in Burkina deflates the dollar average loan size. 24

Village Banking Dynamics Study: Evidence from Seven Programs

compares sustainer rates to average savings and loan size, internal loan size and site location for all seven programs.
Table 5 Sustained Participation Rates Compared by Program in the Third Year Country Program Program Location Percent Sustainers from 1st Loan Cycle 48% at mo. 38 67% at mo. 30 48% at mo. 24 37% at mo. 28 36% - estd at mo. 31 30% at mo. 32 17% at mo. 33 Average Savings Average Internal Loan Average Program Loan

Burkina Faso Guatemala Colombia Uganda Honduras Urban Honduras Rural Bolivia

rural & urban rural urban urban urban rural rural

$33 $116 $99 $179 $130 $70 $155

N/A $60 $7 N/A N/A $69 $129

$92 $139 $251 $222 $141 $74 $82

Guatemala had the highest retention rate of sustained borrowers at month 30 (67%). Its relatively high loan ceilings, liberal access to internal account lending and more flexible repayment requirements may have been sufficient incentive to keep borrowers active in the program. The Guatemala program allowed borrowers to repay program loans in lump sum payments at the end of the loan cycle rather than weekly. After the fifth loan cycle, required meetings are biweekly rather than weekly. In addition, refresher training is given periodically in Guatemala regarding bank regulations that may also contribute to higher sustained participation. In Burkina Faso, the programs initial policy of graduating village banks after the ninth loan cycle may have contributed to a relatively high drop-out toward the end of year three. Burkina Fasos two banks in the ninth cycle have the lowest sustainer ratesonly three and five members consistently borrowed from the first loan cycle until the last. In addition, overall membership for these village banks dropped by 43%. The village banks in their seventh and eighth cycles had 12 and 27 sustained members from the first to last completed cycle. Their overall active membership increased in one bank and decreased by only 9% in the other. This pattern suggests that clients expected the village bank would graduate, and therefore resignation increased as banks reached their ninth and anticipated last cycle. Colombias retention rate was close to the average at 53% in month 22. Since later data is not available, it is difficult to tell if this trend would have continued into the third year. Program managers attribute the retention rate to a lack of competition in microfinance services and member reluctance to withdraw savings from the group.

Village Banking Dynamics Study: Evidence from Seven Programs

25

Rural versus urban site location does not show a consistent pattern in terms of retention rates. Rural and relatively remote communities characterized the program with the lowest retention rate, Bolivia. However, the retention rate of the Honduras-rural program is similar to that of the Honduras-urban and the Uganda program which also primarily lends in towns or communities near towns. However, those village banks in rural Honduras operating in relatively larger and less remote communities had higher average retention rates (above 60%). The degree of internal account lending or the payment of dividends also did not predict retention rates since programs with relatively active internal funds had among the highest and lowest retention rates. Causes of Resignation Resignations are attributed to factors both internal and external to program policy and practice. External factors include migration, illness and inconsistent economic activity. Early-cycle resignation is more tied to lack of compliance with group regulations, whereas late-cycle resignation is linked with inconvenience of weekly meetings and limited savings access. Several of the programs participating in the study conducted informal discussion groups with staff and clients about the major reasons why clients leave the village bank programs.5 Based on this feedback, the following factors were the most common causes of resignation.

Expulsion by fellow members for delinquency or default: This group includes members who did not fully appreciate the regulations when they joined and consequently resigned early. This may be one reason behind the rapid reduction in sustained membership (35%) between the first and third cycles. Seasonality, migration, or poor market or economic activity: Inconsistent commercial or economic activity was cited in four out of seven programs causing resignation or, more often, intermittent borrowing. Dissatisfaction with weekly payments: This was cited in four out of seven programs as a deterrent to continued participation. Not only did borrowers see weekly meetings and payments as time-consuming, but they reduced the amount of time borrowers had to use the loan as working capital and increased the real interest rate of loans. Illness: Health problems often caused members to leave the program or to temporarily halt borrowing.

Other reasons cited less frequently included; clients wanted access to their savings, small loan sizes, repayment problems, and group guarantee requirements. Many of the causes behind resignation are outside program control and influence. Seasonality and migration, for example, are largely linked to site selection, where more rural programs will see a higher degree of intermittent borrowing. Borrowers have changing life circumstances or may become ill or pregnant and want to stop borrowing for a period of time. Still, program policies can influence borrower incentives to remain or leave the program. Looking at general
Feedback from programs was not uniformly gathered on this issue. In addition, the discussions primarily were conducted with current clients rather than members who had left the program. Village Banking Dynamics Study: Evidence from Seven Programs 26
5

trends on an individual program level, a pattern for the timing of resignations seems to occur either early or in later cycles. Policies regarding earlier orientation and training are likely to have a greater effect on the early drop-outs, while greater programmatic flexibility will affect the latercycle member withdrawals. Early Cycle Resignations Honduras-rural: 48% of the first-cycle members resigned by 3rd cycle Bolivia: 67% of the first-cycle members resigned by 3rd cycle Third-Year Resignations Uganda: 20% of the first-cycle members resigned over last 6 months reported Honduras-urban: 20% of the first-cycle members resigned over last 22 months reported Burkina: 26% of the first-cycle members resigned over last 18 months reported Strategies that may increase loan size by increasing the permanence of original members include: better orientation and follow-up on regulations; reducing the frequency of meetings and repayment installment; being more flexible with inconsistent borrowing; and increasing savings access. According to program staff, high early resignation in Honduras-rural and Bolivia is largely caused by poor orientation of group regulations before the village banks were formed and the subsequent expulsion by the group of those not willing to comply. Since the banks chosen for this study were among the first established by the programs, the respective staffs did not have the depth of understanding or experience for introducing the program and clarifying membership roles and responsibilities. As programs gain experience, they are able to pass it on to the new village banks based on more complete training and orientation sessions. Improved up-front orientations and follow-up training on village bank regulations may lower early-cycle resignations. Later-cycle resignations in Honduras-urban and Burkina Faso were related to issues of savings access and weekly payment problems, weekly payment schedules, and increasing savings access. The Honduras-urban program managers cite inability to pay back loans and dissatisfaction with weekly meetings as the catalyst for most resignations. Burkina Fasos large reduction in membership was linked to limited savings access and the expectation of village bank graduation. Uganda experienced considerable drop-out when program regulations were tightened (such as the introduction of a loan ceiling) to encourage better repayment. As loan sizes increase, weekly meetings and repayments become more of a disincentive for membership as clients want more time to work with the funds within a loan cycle. Similarly, as savings amounts increase, so does the incentive to leave a program when the cost of the inaccessible funds versus other options is weighed. Programs may want to consider allowing more flexible repayment and meeting schedules and increased access to savings, particularly in the later cycles, once banks are more mature and demonstrate stability. Also, programs need to have flexible membership rules that allow individuals to sit out or save only for certain loan cycles. This flexibility is likely to be particularly important in rural areas where work patterns are more seasonal.

Village Banking Dynamics Study: Evidence from Seven Programs

27

CONCLUSIONS The experience of these seven programs indicates that village banking clients are able to absorb loans of increasing size. In general, average loan sizes, especially for sustained borrowers, do not stagnate at least for the first two to four years. When trends in average total loans (external plus internal loans) are considered, six of the seven programs exhibit patterns of variable but steady growth. Although loan growth rates were slower than the 500% projected by the original Hatch model, average loan sizes do increase on average by 280% over an approximately three-year period. Without comparable empirical data, it is not possible to say whether this represents relatively high or low growth rates relative to other credit methodologies. The variability in loan growth rates for the seven programs provides useful insight into the effect of various policies, services, membership dynamics and context on loan growth rates. The Guatemala village banks demonstrated the most pronounced loan growth rates, in part, perhaps, due to their relatively longer program experienceon average approximately four years. But this good growth is also likely explained by the programs relatively flexible policieslonger loan period, less frequent repayment installments, access to internal funds and relatively high loan ceilingsand its subsequently high rates of sustained client participation. The Uganda program starts off in its first two years with very rapidly escalating loan growth rates, in part because no loan ceilings are applied and some borrowers take relatively large loans. However, this period of rapid growth is followed by a decline in loan size when more restrictive policies such as loan ceilings were put in place to better control loan-size growth and repayment problems. These problems and new policies were associated with major membership turnover (high rates of sustained borrowers leaving and large numbers of new borrowers joining) which dampens average loan size. The Bolivia and Honduras-rural programs are examples of the dampening effect unrestricted internal borrowing can have on external loan growth rates. This phenomenon also provides an important lesson that village bank clients are willing to pay an even higher interest rate for loan terms they find more attractive, e.g. lump-sum rather than weekly repayment installments. The village banks from the Burkina Faso program illustrate how the anticipation of graduation with its cessation of program services can lead to relatively high rates of program exit. Their relatively modest loan growth pattern also reflects the impact of the larger macro-economic climate and the effects of a major currency devaluation on dollar values.

The fact that poverty-lending programs usually cater to low-income women, often mothers, indicates that the clientele profile does not lend itself to rapid loan growth at the proposed rate of the original model. Instead this profile shows steady but more modest loan increases. This reality is a challenge for NGOs and financial institutions that have made this type of client their market niche. Timeframes for expected self sufficiency may be skewed given client demand for loans. Although an examination of each programs financial sustainability performance on the
Village Banking Dynamics Study: Evidence from Seven Programs 28

relationship to loan-size growth rates is outside the scope of this study, it should be considered for further review. A recent study by Gloria Almeyda for the Inter-American Development Bank touches on this issue and its implications for the client and institution: If large loan sizes result from institutions gradually recognizing that they cannot cover their costs given their current loan-size composition, then this has significant implications for microentreprenuers. Poorer women microentrepreneurs could be without a financial product fitted to their demand characteristics if NGOs adjust their lending methodologies ... away from micro loans toward more remunerative loans. It is important, then, that these NGOs and others continue to improve the methodologies and techniques which allow them to expand both the number and the total volume of loans in the lowest ranges (Almeyda, 1997). Small loans serving the very poor, especially women in rural areas was the original motivation behind the classic Hatch village bank model. Over a decade later, the practice shows that for the most part this incentive still dominates the field. As the pressures for sustainability increase, so will the challenges for practitioners to continue to evolve and innovate village bank programs to serve this clientele with programs that are financially and institutionally sustainable. Implementation strategies that allow for greater flexibility over time and that consider well client incentives to remain or exit the program will play an essential role in the continued successful evolution of the village banking methodology.

Village Banking Dynamics Study: Evidence from Seven Programs

29

BIBLIOGRAPHY

Almeyda, Gloria, Money Matters, Inter-American Development Bank, 1996. Graber, Ken, Analysis of Internal Account Data, presented to SEEP Poverty Lending Working Group, 1997. Hatch, John and Mimi Hatch, FINCA Village Banking Manual, FINCA, 1989. MkNelly, Barbara and Kathleen Stack, Loan-size Growth and Sustainability in Village Banking Programs, Small Enterprise Development, 9 (2): 4-16, 1998. Nelson, Candace, Barbara MkNelly, Kathleen Stack and Lawrence Yanovitch, Village Banking, the State of the Practice, UNIFEM/SEEP, 1996. Pugh, Rosemary, Analysis of Internal Account Data, presented to SEEP Poverty Lending Working Group, 1997.

Village Banking Dynamics Study Evidence from Seven Programs

30