Program Evaluation and Spillover Effects

This note defines what spillover effects are, why it is important to measure them, and how to design a field experiment that will enable researchers to measure the average effects of the treatment in the presence of spillover effects on subjects both eligible and ineligible for the program. In addition, it discusses how to use nonexperimental methods for estimating spillover effects when the experimental design is not a viable option. Several practical examples are provided to show how spillover effects can be estimated. Evaluations that account for spillover effects should be designed in such a way that they explain both the cause of these effects and who is affected by them. Failure to have such an evaluation design can result in wrong policy recommendations and in the neglect of important mechanisms through which the program operates. To estimate the direct and indirect effect of a program, one has to use control groups that are not affected by the program either directly or indirectly. This often means selecting the control groups from different geographic units (e.g. the village or school). In order to understand the mechanisms that cause spillover effects one has to think about competing explanations and collect data on relevant outcomes. In many cases, unveiling the mechanisms behind the spillover effects results in a better understanding of how the program works in general.