Experimental Design
Due to implementation capacity constraints, the entrepreneurship support programme will be delivered in two phases over a 5-year period, and is expected to reach more than 600 villages, across the three regions of Diffa, Tillabery, and Tahoua. During the first phase, only half of those villages are expected to receive the program interventions, while the others will be reached roughly two years later. We will leverage this phased-in approach and rely on a 2-level randomization to test the impact of the programme on social and economic outcomes for individuals and communities in the study region. As a first step, we will randomly allocate villages in the study area to phase 1 (treatment) villages and phase 2 (control) villages.
For the second randomization we will take advantage of the fact that, due to limited resources and logistic constraints, even within treatment villages not every eligible individual will be able to receive the program (the expected capacity is to cover about 30% of the community). As per program guidelines, eligible individuals will have to be poor but able to meet basic consumption needs. This will be determined through a Proxy Means Test (PMT), based on census data that will be collected before the beginning of the program. The Census is being carried out by CERISES-CSF, a local survey firm, as part of the broader PARCA project beginning in September 2020. Within each village (treatment and control) individuals will then be classified into three groups: very poor, poor, and non-poor. The target population will be the middle group of poor individuals. A public lottery will be used in treatment villages to identify, among the eligible individuals, those that will be offered the opportunity to join the programme.
Overall, this design will allow us to identify three groups of individuals: program recipients; spillover individuals (i.e. non-recipients in treatment villages); and pure control individuals. The inclusion of spillover individuals will allow us to detect the impact of the program on non-beneficiaries within treatment villages thus providing insight into the general equilibrium effects. More specifically, by comparing the different groups, we will be able to evaluate the impact of the intervention at three different levels: impact on the village as a whole (which includes both community-level outcomes and average impact on individual-level outcomes); impact on individual program recipients; and impact on non-recipients within treatment villages (spillovers).
We will conduct two rounds of surveys, at baseline and end-line, and rely on three survey tools:
1. Village/community survey: a relatively short survey will be conducted with the village chiefs to collect information related to the community.
2. Individual/Household survey: a detailed survey will be administered to a randomly selected sample of eligible individuals within each village.
3. Market survey: we will also carry out a short market survey that will register local prices on a selected group of common agricultural and non-agricultural commodities.
The specific context for our study will allow us to investigate additional policy-relevant questions. We will take advantage of a rich baseline survey and the presence of both displaced and host individuals among the target population to explore heterogeneous effects by status (displaced vs local), by initial level of social capital within the community, and by psychological attributes. Finally, we will also collect detailed data on community-level economic and social integration to evaluate the impact of the program on this dimension.