Experimental Design
The program takes place in two different sites, namely refugee settlements Kyaka II and Kyangwali, in both refugee and host villages. Refugee and host villages are split into smaller clusters of the target size of 25 households . This division is done to create a number of clusters that can provide sufficient statistical precision for estimation of impacts.
In the first stage, clusters will be randomized into control clusters and treatment clusters. Within control clusters, there will be eligible and non-eligible households. Within treatment clusters, there again will be eligible and non-eligible households. Among eligible households, there will be another layer of randomization into the standard program (T1), the block grant (T2) and small asset (T3). Note that while the first randomization is on the cluster level, the second randomization is on the individual level. Within the control cluster, neither eligible nor non-eligible households will receive any graduation intervention.
Overview of the treatment arms:
Standard program (T1): Participants in T1 will receive the standard graduation program, which includes 160€ in consumption support and an individual grant of 220€.
Block grant (T2): In T2, participants will receive 160€ in consumption support. Additionally, they will receive a small individual grant (e.g. 70€) and another bigger grant (e.g. 170€) per person in the form of VSLA shares. Per group with 25 members that means 4,250€ in total. VSLA members then choose how to allocate the loanable funds. The purpose of the loanable funds is to provide larger loans to the VSLA members, for example, to set up their own business. The idea is that only high performing VSLA members with promising business ideas are granted these larger loans from the loanable funds. During the operating cycle, the loanable funds are only accessible to those VSLA members who were granted a loan, approved by the other VSLA members. All the other VSLA members who didn’t take out a loan during the operating cycle don’t have access to the loanable funds. At the end of the project (cohort 1), the full amount of the shares (i.e. 170€) in addition to the interest will be shared out amongst all the VSLA members.
Small asset grant (T3): In T3, participants will receive 160€ in consumption support and 70€ asset grant.
Non-eligibles in the treatment clusters (TN): Finally, non-eligible households (TN) in the treatment clusters will not be exposed to any graduation interventions but might benefit from MSD.
Overview of control arms:
Eligibles in the control clusters (CE) and non-eligibles in the control clusters (CN): Eligible households in the control clusters will not be exposed to any graduation interventions or MSD intervention. Hence, treatment as usual (TAU) doesn’t involve the exposure to any intervention under the SMILES project. Likewise, non-eligible households in the control clusters won’t receive any intervention.
The cross-randomization of Graduation intervention and MSD intervention was executed at the cluster level. IPA randomly assigned half of the Treatment, Control, and Cohort 2 clusters to MSD and half to no-MSD. Adjustments were made to the stratification variable used previously in the first-stage randomization of the Graduation intervention. The objective was to guarantee that within each stratum, and Treatment, Control and Cohort 2, there were a minimum of two clusters available for randomization into MSD and no-MSD arms.
In Cohort 2, eight geographic strata contained only one cluster each. In those cases, the strata with a solitary cluster were either merged with another stratum or incorporated in a cluster from another stratum. This adjustment to the stratification variable, achieved through the manual merging of geostrata, was carried out based on the implementing partner's regional expertise. The merging process focused on combining geographical areas that were close to each other.
The balancing variables used in the Graduation-MSD cross-randomization include variables such as the number of households, household head characteristics (female or child head), scorecard score, and poverty wealth rank. In total, there are 93 clusters randomized into the MSD arm across Treatment (46) and Control (47). 143 groups and 3,790 participants were assigned to receive an MSD intervention in addition to the Graduation program.
MSD intervention
To assess the extent to which additional MSD interventions enhance the impact of the Graduation program and to explore the complementarities between the Graduation and MSD interventions, IPA implemented a cross-randomization of the Graduation and MSD interventions at the cluster level. Specifically, half of the Treatment and Control clusters were randomized into either MSD or no-MSD arms. This resulted in 46 Treatment clusters and 47 Control clusters being assigned to the MSD arms, which include 3,904 and 3,897 Graduation-eligible households, respectively. Additionally, 49 Treatment clusters and 45 Control clusters were assigned to the no-MSD arms, corresponding to 3,982 and 3,561 Graduation-eligible households, respectively.