Rural Business Development Services 2007-2011

The challenge of this and all impact evaluation efforts is to identify a control group that is identical to the treatment group in every way except that they have not benefitted from the intervention under evaluation.the evaluation team worked with the RBD implementation team to identify all geographic clusters that would eventually be observed RBD services. The evaluation team then selected a subset of these clusters for random assignment to either early or late treatment status. This strategy not only created a temporary conventional control group, it also randomized the duration of time in the program, a feature that will prove vital in the continuous treatment estimates presented below.In late treatment clusters, services were not initiated until approximately 18 months later, in early 2009 at the time of the midline survey. Because clusters were randomly allocated to early and late treatment conditions, we can anticipate that on average the late treatment group should function as a valid control group, identical to the early group in every way except early receipt of RBD services. The economic status of the late group in 2009 should thus be a good predictor of what the status of the early group would have been in the absence of RBD services. Both early and late treatment clusters were then surveyed again near the end of the program in 2011. Once the random assignment of early and late clusters was made, the impact evaluation team created a roster of all eligible producers in these clusters, and then randomly selected a sample of 1600 households split between early and late areas. These 1600 households were then invited to participate in the impact study, and completed a baseline survey in late 2007, just as the RBD project was beginning in the early treatment clusters. Within these clusters, 64% of the eligible households chose to participate in the RBD project. A second-round survey was applied to all 1600 households in the first quarter of 2009, just as the RBD project was rolled out in the late treatment area. While it was not clear at baseline which of the eligible households in the late treatment areas would choose to participate in the project, those households made their participation decision around the time of the second-round survey. Similar to the early treatment clusters, 57% of eligible households in late treatment clusters elected to participate. Because the timing of the surveys and project rollout allow determination of farmer type in both early and late treatment areas (participants versus non-participants), the impact evaluation has the opportunity to study impacts on both eligible households as well as impacts on participating or complier households. The evaluation here will primarily focus on the complier households as we are interested in the impact of the program on the types of self-selecting individuals who adopt it.