Minding the Gap : Aid Effectiveness, Project Ratings and Contextualization

This paper applies novel techniques to long-standing questions of aid effectiveness. It first replicates findings that donor finance is discernibly but weakly associated with sector outcomes in recipient countries. It then shows robustly that donors' own ratings of project success provide limited information on the contribution of those projects to development outcomes. By training a machine learning model on World Bank projects, the paper shows instead that the strongest predictor of these projects' contribution to outcomes is their degree of adaptation to country context, and the largest differences between ratings and actual impact occur in large projects in institutionally weak settings