This study uses a multi-arm cluster randomized controlled trial (RCT) to examine the relative and combined impacts of the enhanced self-employment support component (i.e., enhanced livelihood) and a wage employment component (i.e., public works), and cross-randomizes the mode of payment (digital vs. manual) to understand the impact of digital e-payment.
The study considers villages corresponding to regions in the domains of Mainland and Zanzibar, Tanzania with the following sampling frame:
Mainland - 1553 villages from 41 PAAs in 19 regions
Zanzibar - 190 villages from 11 PAAs in 5 regions
1. In the first stage, 32 PAAs were randomly selected out of 52 PAAs in the sampling frame.
2. In the second stage, one village was randomly selected per ward (if there were multiple villages in each ward in the sampling frame).
The final selected sample was set at 434 villages from 14 regions in Mainland and 3 regions in Zanzibar. These villages were then divided into two groups: a) those that already received some support of switching to e-payment (226 villages); and b) those which haven’t switched e-payments (208 villages). Only those villages that have not been encouraged to use the e-payment cash transfer delivery option by the implementing agency yet, are planned to be part of the cross-randomization to assess the comparison between manual payments versus e-payments.
Villages will be randomly assigned into the following treatment arms (please note that all 434 villages will receive the basic livelihoods package as well as the cash transfer component): a) enhanced livelihoods, b) public works, c) public works and enhanced livelihoods, d) controls. For the e-payment evaluation (208 villages), villages will be further randomized into cash payments and digital payments (e-payments).
For the sampling of households, all households in the selected villages will be defined into the following types:
• Type 1: those nominated by communities AND interviewed by TASAF for PMT AND above the selection threshold (i.e., beneficiaries);
• Type 2: those nominated by communities AND interviewed by TASAF for PMT BUT below the selection threshold;
• Type 3: those NOT nominated by communities AND NOT interviewed by TASAF.
Type 1 households are potential PSSN beneficiaries that can be identified in both control and treatment groups, depending on how their villages will be randomly assigned to the different groups. Type 2 and Type 3 households are not PSSN2 beneficiaries, but we compare the characteristics of households between Type 1, 2, and 3 to assess the accuracy of the TASAF targeting rule. The number of households to interview is 29 households per village (i.e., 17 HHs for Type 1, 8 HHs for Type 2, and 4 HHs for Type 3).
Overall, there are two types of analysis that this study will perform:
- Analysis I - IE analysis of PSSN programs on beneficiaries: For this analysis, the sample of households will be randomly drawn from the households classified as “poor” in the PMT data (type 1). Therefore, there is no need to do a listing of households for this analysis. Since the PMT data is at the village level, the analysis will be done at the village level as well.
- Analysis II - Targeting exercise and spillover analysis: To test the validity of the targeting methods, we will have to compare poverty characteristics of type 1, type 2, and type 3 households. And the sample of households for each type needs to be drawn from a sampling frame that covers the same area. For example, comparing type 1 households that are drawn from the entire village with type 3 households that are drawn from only one sub-village is not valid. Similarly, if we want to measure how the impacts of PSSN programs are different between beneficiaries (i.e., type 1 households) and non-beneficiaries (i.e., type 2 and type 3 households) then type 1-3 households need to be sampled from the same area to ensure comparability. Therefore, for targeting exercise and spillover analysis, we will rely on the listing exercise to create a sampling frame from which we can randomly draw a sample of all three types (i.e., 4 type 1 households, 4 type 2 households, and 4 type 3 households).