Considering Unintended Consequences: Evidence From Recent Evaluations of U.S. Foreign Assistance Programs

Abstract

This article examines factors that could have influenced whether evaluations of U.S. government–funded foreign assistance programs completed in 2015 had considered unintended consequences. Logit regression models indicate that the odds of considering unintended consequences were increased when all or most of seven standard data collection methods had been used, and there were some agency effects; however, no associations were found with evaluation type (impact vs. performance), timing, or whether evaluations were conducted by external evaluators. These results suggest that some proportion of the evaluations that did not consider unintended consequences would have considered such consequences if they had employed more data collection methods. Moreover, few of the evaluations that considered unintended consequences provided many details about their investigative methodologies or the consequences themselves, which raises further concerns that users might not have received information that would have helped for learning or accountability purposes.