Does a cassava research-for-development program have impact at the farm level? Evidence from the Democratic Republic of Congo

This paper evaluates the impact of a cassava research-for-development program on farm level outcomes. The program was implemented in the Democratic Republic of Congo from 2001 to 2009. We apply propensity score matching, Rosenbaum bounds on treatment effects, Altonji et al. method of selection on observables and unobservables and endogenous switching regression to farm survey data collected during the 2009 cropping season. We use these methods to test whether the R4D program has a statistically significant effect on outcomes of interest and if these are not driven by selection on unobservables. Using propensity score matching, we find statistically significant positive effects on household participation in cassava markets, adoption of improved varieties and crop management practices and household food adequacy; and no statistically significant effects on yields and profits. The results show that bias due to selection on unobservables is not severe enough to invalidate the impact estimates. Bias may still be a problem that is present in the analysis. But there is evidence that it is not substantial. Although the program does not have a statistically significant positive effect on yields and profits, the significant program effects on market participation, variety adoption, and food adequacy merit further promotion of the program since these positive outcomes tend to be pre-conditions for realizing long-term yield and profit benefits.