The mobile money revolution has begun to substantially increase financial access around the world. Mobile money accounts allow people previously excluded from the formal financial sector to access savings accounts, make payments to merchants, and make person-to-person transfers, among other services. Research has shown that mobile money access can have important effects on key outcomes such as improved resilience and food security. Marginalized communities are seeing benefits; for instance, by catalyzing market integration, increases in mobile money use in areas surrounding refugee camps can potentially help better integrate refugees with the host communities.
However, some groups may still be less likely to enroll in or use mobile money services. In areas studied by the Strengthening Host and Refugee Populations in Ethiopia (SHARPE) market systems development program, women in the host community and refugees in general both tend to enroll in and use (conditional on enrollment) mobile money with a lower frequency than men in the host community. To test ways to increase enrollment and use among target groups, IFPRI and Dadimos Development Consulting collaborated with SHARPE and Shabelle Bank to design two randomized trials to test methods of increasing enrollment and use: A referral study to promote enrollment and an incentives study among enrolled customers to promote use. This post outlines the results, which were mixed.
For the referral study, we randomly selected from among 800 high use customers in the SHARPE targeted areas, stratified by gender, refugee status, and location (we call them “community referrers”). The randomly selected groups were offered three types of rewards if they referred new customers: A low reward or bonus, a high bonus for referring women and a lower bonus for referring men, and a high bonus; we also selected a control group whose members could make referrals but would not receive a bonus for doing so. We hoped to test whether the different bonus levels would have different effects on enrollment.
The incentives study design was quite different. In this experiment, we took a sample of customers who had enrolled in HelloCash (Shabelle Bank’s mobile money product) but had not yet made a transaction, again from the refugee hosting areas. We randomized them into three groups: First, a group of customers received 25 birr (about $0.50) with a text message telling them to use it however they wanted. A second group received a text telling them they would receive 25 birr if they made a transaction within three weeks. We compared these two groups to a control group.
Referral study results
The referral study worked quite differently than we expected. Only 55 of the 800 community referrers attempted to make a referral at all, or 7%; only 41 of them made successful referrals. However, those 41 CRs made at least 2,406 successful referrals in total. And there was one extremely active referrer: She registered 1,354 unique individuals herself during the course of the study—more than half the total. So the RCT results were substantially influenced by one extremely active individual. While the participation rate was disappointing from an academic perspective, Shabelle Bank was satisfied with it, as the referral study doubled the number of customers enrolling during the two primary months it ran. However, we also note that the gender and refugee composition of referred customers was nearly the same as those enrolling through other means—meaning it did not make a noticeable difference reaching women and refugees.
Qualitative work accompanying the trial suggests that some community referrers were not appropriately targeted—they were too busy to try to make referrals, or the opportunity cost for them was too high. Because of the one super-active individual referrer, it is hard to say anything about the size of incentives at this point. So there are two obvious ways to make potential use of the results. First, one could repeat the exercise using a clearer screening device for individuals, attempting to flag people with high incomes or constrained time allocations as inappropriate for referral incentives. Perhaps some screening could help increase the proportion of individuals participating. A second strategy would be to open up the process, advertising to allow any customers to make referrals for small incentives. Since Shabelle Bank has about a million customers in all of the Somali region, even a 0.1% participation rate in a referral scheme could lead to thousands of additional new customers each month.
Incentive study results
For the second trial, the unconditional transfer led to a greater likelihood of transactions: Recipients were 9.3 percentage points more likely to make a transaction during the study period than non-recipients. Men receiving the unconditional transfer were 14 percentage points more likely to make a transaction, while women were 6.8 percentage points more likely. We find no significant increase in likelihood among refugees, however. The conditional transfer was somewhat less successful; it only increased the number of transactions made among women, and we found no effects on the gross value of transactions. Though the results do suggest some additional transactions can be catalyzed by incentives, particularly among women, it was clearly not successful among refugees, and the magnitude of effects suggest it would be a costly way to catalyze use among a larger sample.
Summary
While both trials were successful in one way, they were unsuccessful in others—the referral study did not increase the proportion of women and refugees enrolling, and the incentives trial only led to a higher proportion of women using the system and not refugees. It is not clear that the referral system could be adjusted to lead to more women or refugees enrolling, though one could restrict referrals to among refugees (either restricting referrers to be refugees, or only paying bonuses for enrolling refugees). From the perspective of the incentives pilot, there are clear adjustments that could be made for further research. For example, the number of transactions needed for conditional incentives could be reduced to one or two; a second text could be sent to unconditional recipients giving ideas of how to use the money. Regardless of the method, more work is needed to learn about strategies to integrate refugees into host communities.
Alan de Brauw is a Senior Research Fellow with IFPRI’s Markets, Trade and Institutions Division.
