Association for Public Policy Analysis & Management

Despite a rise in international development project evaluations, there is little evidence that their findings influence the design of later projects. Methods: I use data from 445 World Bank health and education projects from between 2000 and 2010 to identify whether the performance of past projects predict the outcomes of later, closely-related (sharing a sector and sub-sector classification) projects. For each project, I calculate the average rating of all closely-related projects launched in the five years preceding its own approval. I use ordinary least squares regressions with project- and country-level characteristics as control variables to identify whether the previous five-year average has any significant correlation with the later projects’ performance. Results: The five-year average performance of past, closely-related projects is not significantly associated with project outcomes (p-value of 0.845). Rather, the main effects came from project approval year, various project supervision and quality control characteristics, primary school enrollment rates in the country and year of the project, and other project- and country-specific characteristics. Conclusion: This finding implies inadequate learning from evaluation results, as learning ought to result in improved performance over time, with every additional point of improved past performance being associated with improved later project performance. I recommend better communication between the World Bank’s International Evaluation Group and project design teams, improved data collection, and further research.