The standard message from advocates is that savings group programs reach poorer people than do credit-led microfinance institutions (MFIs). In post # 78, I found that MFIs in general don’t do a great job of including the poor. Even the poor-oriented MFIs seem to struggle to include a higher percentage of poor people in their clientele than exist in the local population. Not a high bar. What is the evidence that savings group programs can do better?
Amounts of Loans and Savings on Deposit
In Savings Groups at the Frontier, Susan Johnson and Silvia Storchi review the recent evidence of poverty outreach by savings group programs (Chapter 2, pp. 44–49). They start with the standard comparison of loan size and savings on deposit, both of which are much smaller for savings group members than for MFI clients. I have long argued that loan size is as much a function of the provider’s policy and the availability of alternative providers as it is reflective of the borrower’s poverty status—poor people seldom take large loans, but better-off people often take small loans, especially when they have no alternative.
I’m much more impressed by the big difference in amounts of savings on deposit, which is an important asset-based measure of “wealth.” Considering the small, if any, return on savings deposits offered by formal financial institutions compared to the substantial return offered by savings groups, we should expect people to put a more substantial portion of their total savings in a savings group than in an MFI. So, the difference in ownership of this major component of “wealth” must be even greater than shown by Johnson and Storchi; that is, this difference probably represents a major difference in poverty status of the average MFI client vs. the average savings group member.
But does this evidence indicate that savings group programs reach people too poor to be reached by MFIs? No, it doesn’t. It may just indicate that only poorer people join savings groups while poor and non-poor alike patronize deposit-taking MFIs and other formal financial institutions, resulting in much higher average deposits per MFI client.
People Who Cannot or Will Not Use MFI Services
To demonstrate that savings groups serve people who cannot use MFI services, we must have evidence that substantial numbers of poor people in savings groups are not also clients of MFIs and other formal providers for one of three reasons:
there are no local MFIs (as in rural areas);
they cannot meet the MFI requirements (such as fees to open an account and do transactions); or
there are social barriers (such as not understanding the language in which business is done at the MFI).
There is much evidence in Savings Groups at the Frontier that all three reasons are at play in preventing the poorer population from using MFI services, but we need more systematic, even if qualitative, evidence of exclusion of savings group members from using MFIs.
What if the poorer population chooses to join savings groups and to avoid MFIs just because savings groups offer them a better deal, as I argued from the evidence in Theme Two? The better deal would be in the form of net benefit compared to costs, both explicit (like dividends on their savings deposits) and implicit (like greater trust). There is much evidence in Savings Groups at the Frontier to corroborate that savings group members prefer savings groups to MFIs, but do poorer people prefer savings groups more than better-off people do? The evidence to answer this more complex question is harder to come by.
“Poverty Gap” between Savings Group Membership and the Local Population
Let’s return to the initial question: Can savings group programs include a higher percentage of poor people in their membership than exist in the local population? This is the “poverty gap” question explored in post # 78 regarding MFI clients. Here we have better evidence already at hand for savings groups.
Johnson and Storchi cite a study in Cambodia, which found that 13 percent of savings group members fall under the national poverty line compared to 22 percent of non-members living in the same area and 30 percent of the national population. Not a promising example. By contrast in Guatemala, application of the Progress out of Poverty Index (PPI) measurement tool found that 67 percent of new members were likely below the national poverty line, compared to 72 percent of non-members living in the same area but only 40 percent of the national population. The savings group program had targeted poorer areas of the country.
Similar geographic targeting in Kenya resulted in 64 percent of members below the national poverty line compared to 38 percent of the general population. In contrast in Uganda, where there was little geographic targeting of poorer areas, only 15 percent of members were below the national poverty line compared to 19 percent of the national population. Similarly in Tanzania, 27 percent of both members and the national population were below the poverty line, which is better than most MFIs can do.
In Rwanda, targeting specifically sought to include marginalized and excluded (arguably poorer) segments of the population, with the result that 63 percent of members were below the national poverty line, compared to 54 percent of the general population.
How Savings Groups Draw Members from their Communities
As part of the recently completed impact evaluation of the Saving for Change program in Mali, Innovations for Poverty Action (IPA) and the University of Arizona’s Bureau for Applied Research in Anthropology (BARA) examined how the group membership reflects the distribution of wealth and poverty in the local community. The researchers conclude that the Saving for Change program has clearly succeeded at outreach to hard-to-reach people; the evaluation was done in villages which were rarely even visited by MFIs. But does Saving for Change reach the poorer households of those villages? The research was a mixed-methods design, in which BARA generated most of the qualitative evidence and IPA generated most of the quantitative evidence. On the qualitative side, here is a quote from p. 100 of the full report:
Results from interviews with poor households show that women from these households are enthusiastic members of Saving for Change groups, even when finding funds for the weekly contributions may prove difficult to procure. One main reason for this is the high social pressure to become part of a Saving for Change group and share the experience with other women in their family and wider social networks. Despite such personal difficulties faced by the poorest women (particularly widows or women whose husbands have migrated and created a deficit of household labor), BARA research did not identify any formal and informal exclusionary mechanisms that prevent poor women from joining Saving for Change, which points to the program’s accessibility to the most vulnerable groups in a village.
On the quantitative side, this is from pp. 42–43 of the full report (note: “adopters” are savings group members):
Though adopters are more likely to come from wealthier households than non-adopters, this does not mean that the program is only reaching the better off. First, these villages are on average very poor. Second, the size of the difference in participation between those better off and those worse off is not large. … we used per capita food consumption to capture those who are better off (upper 33% of the village), those who are in the middle (between 33% and 66% of the distribution) and the worst off (bottom 33%). Per capita food consumption is a standard way of capturing poverty, as food consumption in many developing country contexts constitutes 80% or more of poor households’ total budget share. Participation is highest among the top tercile: 42% of women in the top tercile in the treatment group participate in Saving for Change. This contrasts with a 33% likelihood of participating in Saving for Change among women in the bottom tercile. While this 9 percentage-point gap is considerable, it does not suggest that the poorest women are excluded from Saving for Change since a significant percentage of them do join.
This analysis focuses on food consumption. I should note that IPA’s research also found that, in aggregate, members and non‐members did not differ in whether or not they had savings, in their scores on the food‐security indices, or in their poverty (PPI) scores. Continuing on p. 43:
In summary, at the time the program started, several key differences were identified between adopters and non-adopters. Adopters tended to be older women from larger, wealthier households. Additionally, adopters had higher initial social capital and [were] more likely to participate in decision-making at the village and household level. … Adopters were also more financially active in terms of savings and borrowing. Therefore, our baseline data indicates that prior to any changes caused by the Saving for Change program, adopters were, on average, more empowered than non-adopters along several dimensions. Though members do tend to be slightly better off than non-members, we do not intend to say that Saving for Change does not reach the poorest of the poor. The differences in participation rates across the wealth distribution are not large (despite being statistically significant), a finding [that] highlights a strength of the Saving for Change program.
This analysis is quite similar to the work of Nteziyaremye and MkNelly for Credit with Education offered by credit unions in Mali (see post #79), with strikingly similar results. Both savings groups and credit groups in southern Mali draw from all levels of the “wealth pyramid” in the villages, with slight bias toward the better-off. More important than wealth of her household are the social capital and empowerment characteristics of the woman who joins vs. the one who does not join, at least the first group formed in the village. This corroborates the conclusion of Theme Five, that social capital is an important precursor for group formation and success. This also corroborates the conclusion in post #79, that a high percentage of members of self-selected groups will come from poor households only if the program, whether it is credit groups or savings groups (both Credit with Education and Saving for Change rely on self-selection of group members), is placed in communities where poor households constitute a high percentage of the total households. Program participation reflects the distribution of wealth and poverty in the whole community.
Self-Selection vs. Targeting
This finding may apply only to areas, like southern Mali, with high percentages of poor households. Johnson and Storchi (pp. 47–48 of Savings Groups at the Frontier) note that where poverty rates are low, as in Ecuador and Cambodia, the proportion of poor households in savings groups may be low, even with targeting of rural areas. They suggest a general conclusion:
Where self-selection is the main recruitment mechanism and poverty rates are low, poorer people are less likely to be early adopters and more likely to be marginalized by the majority. Where poverty rates are high, the poor are more likely to be included as savings groups form. Geographic targeting toward communities with relatively high poverty incidence is likely to make inclusion of poor people easier. Where poverty incidence is lower or inequality is greater, more specific targeting of the poor will likely be required to ensure their inclusion.
The phrase “where … inequality is greater” is an important caution regarding the program design decision to depend on self-selection for group membership recruitment or to actively target recruitment to the poorer or otherwise marginalized households of each community. Even in relatively homogeneous Malian villages, there are ethnic differences that affect group membership. Specifically, the Saving for Change evaluation in Mali found that participation among the Bobo women is significantly lower than in the other ethnic groups, probably due to language differences (facilitating agents being primarily Bambara-speakers) and the Bobo being a marginalized ethnic group.
While Freedom from Hunger should find satisfaction in the good poverty outreach of Credit with Education and Saving for Change in countries like Mali, the organization’s mandate to focus on households so poor they are chronically hungry may require a different member-recruitment policy and process in other countries with lower poverty rates and more unequal distribution of wealth and power within the villages.