Searching for the Devil in the Details: Learning about Development Program Design

Motivated by our experience in designing a particular social program, skill set signaling for new entrants to the labor market in Peru, we articulate the need for, and explore the empirical consequences of, alternative learning approaches to the design of development projects. Using a simulation, we demonstrate that even with only modest dimensioned design space and even modest ”ruggedness” of the outcome with respect to design a naive iterative approach of ”crawling the design space” dominates an RCT learning strategy. We suggest that the empirical results of RCTs to date are consistent with social programs having high dimensional design space and outcomes sensitive to design and hence project, program, and policy design must depend on more robust learning strategies than the attempt to directly apply results from ”systematic reviews” or move prematurely to an RCT.