In a prior post (# 49), I promised a closer examination of the unprecedented research reported by Jeffrey A. Flory on savings accounts offered by mobile units of the Opportunity International Bank in rural Malawi. This study is primarily concerned with the indirect effects of formal savings accounts on consumption-smoothing of non-users of the savings accounts—in particular, increase of inter-household financial assistance received by the most vulnerable households and positive impact on food security and other welfare indicators for these households during the pre-harvest hungry season. In Flory’s words: “In qualitative interviews in rural areas of central Malawi, formal-savers report the top reasons people ask them for cash help are for medical expenses and sickness-related issues, to buy food, or to pay for funeral expenses.”
Here is a brief summary of results and implications from Flory’s abstract:
Using a panel dataset of over 2,000 households collected during a rapid expansion of formal savings services in Central Malawi, this paper shows that experimentally boosting use of formal savings in rural areas sharply increases inter-household transfers during peak periods of hunger. The impact on transfer receipts is strongest among the poorest households, a de facto financial services-ineligible group, among whom the effects are also linked to significant changes in welfare. The strong impacts of formal savings expansion on non service-users suggests that formal finance can have much greater immediate-term effects than would be suggested by focusing exclusively on impacts experienced by service-users. The findings also highlight the sensitivity of traditional safety nets and welfare outcomes among the highly vulnerable in villages to expansion of formal financial markets.
In my view, this research is exceptionally well-designed and analyzed (I’ve asked folks at the Financial Access Initiative [FAI] for their assessment, which I hope will be forthcoming). I now excerpt and paraphrase from Flory’s report.
In late 2007, OIBM (not named by Flory) began expanding formal savings and credit access to rural areas of the three largest districts of central Malawi through a mobile van-bank, which traveled along paved roads and had six different stops at local trading centers. The data consist of a two-year household panel which spans the initial phases of access expansion. The baseline data was collected over February–April of 2008, during the pre-harvest “hungry” season, when household resources are often stretched thin and food-stocks for some are running low. This was prior to any measurable use of the bank’s services in these areas. The second round was collected over the same period in 2010, after an extended information campaign designed to encourage use of the bank’s services.
Flory refers to this study as a “natural experiment,” but it seems to me a straightforward experimental intervention, in which the “treatment” was an information campaign at the community level to promote the use of the mobile van’s banking services. The communities were paired with others of similar size and distance from the main highway and then randomly assigned to get or not get the information campaign, which consisted of paid extension agents entering and operating in the same way that agricultural and health promoters were already entering and operating in these communities. These agents visited periodically on foot and bicycle to bring informational materials on the bank’s services, to talk with community members, and to leave posters and other promotional materials in each community assigned to the marketing treatment. Distances from the mobile bank stop ranged from 0 to 14 kilometers.
Perhaps the most important innovation in this research design is that households were classified into seven levels of vulnerability to hunger and low welfare outcomes, using baseline (2008) variables on food-security status, assets (proxied by cell phone possession), education (proxied by literacy level), distance from major roadways and trading centers (proxied by distance from the van stop locations), and gender of household head. The primary indicator was the household’s 2008 food-security status. The survey included a slightly modified version of the USAID Household Food Insecurity Access Scale for Measurement of Food Access—very similar to Freedom from Hunger’s Food Security Survey tool (both are derived from a USDA tool developed for use with U.S. populations). Food-insecurity scores were generated by examining the frequency with which each of seven possible food-insecurity conditions occurred in the 30 days preceding the interview. The survey was conducted during the pre-harvest “hungry” season, so these scores reflect conditions during the most intense period of vulnerability to low food-intake. In the most vulnerable category, households were classified as “severely food-insecure” in 2008, were located three or more kilometers from the van stop, did not have a cell phone, and either no member was literate in Chichewa or the household head was female.
The information intervention had a significant effect on the proportion of previous non-savers that adopted formal savings (from 9.3% at baseline to 12.4% after the intervention). Both the magnitude and significance of this effect increased with distance more than 3 km from the van stop (8.6% at baseline to 12.3%), confirming that information on services is more effective in more remote locations. This result makes sense, because the households close to the van stops would learn from several other sources about the opportunity to open a formal savings account.
However, the information intervention did not affect the uptake of savings accounts in the most vulnerable households. Flory argues that these households were “essentially ineligible to take advantage of increased formal savings access,” due to barriers such as fees, distance and formality. On the other hand, in looking for indirect effects, the study examined the effects of local formal savings rates on receipts of assistance, particularly by the most vulnerable households. Information on inter-household transfers was collected during the pre-harvest hungry season, the time of year when household resources are most restricted and any transfers received are likely to have the highest positive marginal impacts. It is also the time during which requests for assistance are arguably most abundant. The researchers gathered data on cash gifts of 50 kwacha (about $.30) or more, received over a 90-day recall period preceding the interview. The vast majority of gifts were from within the local community.
Flory first reports the overall difference in receipts of assistance between intervention and non-intervention areas. While 20.8% of all households in the non-intervention areas received a cash gift in the last 90 days, 30.6% of those in the intervention areas received one—a difference of almost 50%. In addition, while 7.4% of all households in the non-intervention areas received more than one cash gift, 12.0% of all those in the intervention areas received multiple cash gifts—a difference of 62%. Both differences are highly statistically significant (p<.001).
Then Flory shows the difference in receipts of cash gifts depends quite heavily on household vulnerability level. The difference among the least vulnerable groups attenuates substantially. There is a remarkably consistent pattern of increasing difference as indicators of vulnerability increase. The amount by which the percentage of households receiving gifts is higher in intervention than non-intervention areas is only 4.3 percentage points for category A households (not significant) and 8.5% for category B (not significant). The difference grows to 10.4 percentage points for category C households, 10.6 for category D, 10.9 for category E, 11.5 for category F, and 17.8 percentage points more for Category G, all of which are highly significant (at the .01 level or higher). For category G, the change was from 9.9% of households receiving cash gifts at baseline to 27.7% after the intervention, an impressive 180% increase among the most vulnerable households.
Flory goes on to show that the average value of cash gifts (which declines as the vulnerability of the household increases) was not affected by the intervention, and there was no increase or decrease of non-cash, in-kind assistance. However, there was a substantial increase of the proportion of highly vulnerable households receiving cash loans from friends or relatives (remarkably similar in scale to the estimated effect on the proportion receiving cash gifts), but not for the less-vulnerable households. Thus, the total value of inter-household assistance to the most vulnerable households seems to have been increased by the intervention, which suggests possible welfare improvements among the highly vulnerable.
The study provides evidence for improvements in three different welfare indicators: two food-security indicators and one simple health indicator. Highly vulnerable households in treated villages are 11.8 to 16.3 percent more likely than comparable households in control villages to exit the worst food-security category (“severely insecure”) to enter one of the three less-severe categories. They also experience a 1.3 to 1.4 point reduction in a continuous food-insecurity score relative to the highly vulnerable in control villages. This represents a 10–12% improvement in food-security over baseline values. In addition, highly vulnerable households in treated villages were 12 to 17.4 percent less likely than those in control villages to report any members of the household as recently unwell.
We seldom see such dramatically positive direct, much less indirect, welfare impacts of microfinance or any other type of large-scale intervention, particularly for the poorest, most vulnerable people in the local population. Clearly in this case, there was a fortunate match between a simple information intervention tied to a basic savings-account offer and the needs and functions of a traditional social safety net providing mutual assistance among households of varied wealth and vulnerability status. As Flory cautions, “It is not clear that introducing formal savings will always have a positive effect. Differing local customs, cultural norms, and levels of social cohesion may cause strong negative effects in other settings. A deeper understanding of the underlying causal mechanisms and the drivers which may lead to positive or negative impacts on local social safety nets is an important avenue for further research.”
What is clear, from Portfolios of the Poor and similar financial diary studies, is that such informal webs of mutual assistance among households are common in poor communities. The power of microfinance for poverty alleviation is demonstrated when it augments and enhances these informal systems to support the resilience of households in the face of financial setbacks and even severe shocks.