The sampling frame for the survey were 150 farmer groups that expressed interest in the extension training from the project. This made up 2000 households, the remaining 500 were selected from Village Development Committees and Wards that lied outside of the command area of RJK to act as the external control group. The 500 households were selected from the census of the command area conducted by the project at the beginning of phase 1 of RJK project.
At baseline we gave the survey team all 40 early and 40 late treatment groups to be surveyed. The team was instructed to survey all households that were members of the selected farmer group, households with multiple people enrolled in the same farmer group were only interviewed once. Then enumerators were provided with the list of all control groups and 500 households outside of the command area, regardless of farmer group status. It was observed that control farmer groups were largely inactive, and some could not be reached at baseline.
At endline the survey team was able to reach 90% of baseline respondents, however the control group outside of the RJK command area experienced higher attrition than other groups. For the regressions analysis we will combine the late treatment and control inside RJK groups, because they were both offered the training after the baseline survey took place. The “Other” groups in the above table are households that lived in same wards as control group households, but were not included in the original randomization, and thus they will be excluded from the analysis. The control outside RJK group will be the main counter-factual groups, because they received neither the irrigation improvement nor extension training from RJK project.
