Experimental Design
o Partnering with a local microfinance institution (MFI), the research team will use the MFI’s existing client data to select 800 of their top digital financial service (DFS) users to invite into the study. These individuals were selected to ensure equal sample sizes of refugees vs. host community members, of men vs. women, and within the two areas where the program operates.
o After these 800 current users agree to participate in the study and complete a baseline phone survey, they will be encouraged to make referrals of others in their community who they think would be interested in enrolling in free DFS with the MFI. Two options for making these referrals will be set up by the research team (a digital option and a call center), with referred individuals’ contact information passed to the MFI for follow-up.
o The 800 current users will be randomly assigned to one of four groups. All four groups will be contacted, thanked for being among best customers, and offered the option to refer others who the MFI will then facilitate enrollment for. The treatment groups will additionally be offered different rewards for successfully referred enrollees, where reward amounts will depend on the gender of each newly enrolled referral. In particular, 200 current users will be assigned to each group:
- Control Group: No rewards for referrals
- Treatment Reward Group 1: 25 Birr for men, 25 Birr for women
- Treatment Reward Group 2: 25 Birr for men, 50 Birr for women
- Treatment Reward Group 3: 50 Birr for men, 50 Birr for women
o The above listed randomization across treatment and control groups will be stratified by gender, refugee vs. host status, region, and baseline service usage.
o Over the first two weeks, all study participants will be texted messages that remind them of how to make referrals; in the case of treatment groups, messages will also remind them of the rewards they will receive for successfully referred enrollees.
o Follow-up data collection will be conducted using a phone survey, with the full set of original users included in the baseline. In addition, a random sample of new enrollees – both those who were registered through the referral system as well as others who registered on their own from the same regions over the same time period– will be chosen to provide household-level information in a single interview conducted at the same time.