Experimental Design
Our unit of randomization is the livestock cooperative, where each cooperative encompasses multiple villages and hundreds or even thousands of members. We randomized in private using Stata. The randomization assigns 116 cooperatives—the current universe of HPN goat-marketing cooperatives with a working CAVE—to be equally split between four treatment arms: control, dCM, dMCE, and dCM + dMCE. Each cooperative has between one and four CAVES, all of which will be included in the sample. This design is motivated by the need to compare impacts on both goat farmers, CAVEs, and cooperatives, and by the possibility of either (i) pooling the two treatment arms together or (ii) estimating their separate effects relative to the control group.
For cooperative members, outcomes of interest will be measured at baseline and in two follow-up rounds: immediately after the peak goat marketing season in 2024 and following the peak goat marketing season in 2025. For CAVEs, outcomes will be measured at baseline and at four follow-up rounds. To estimate impacts on cooperative level outcomes, HPN will collect monthly data from cooperative officers, including three rounds of baseline data and 17 follow-up rounds. As with any study, our proposed work faces threats to internal validity.