Measuring externalities in program evaluation

Impact evaluations of development programmes usually focus on a comparison of participants with a control group. However, if the programme generates externalities for non-participants such an approach will capture only part of the programme’s impact. Based on a unique large-scale quantitative survey we estimate the direct as well as the spillover effects of a women’s empowerment programme in rural India on child immunization and school enrolment. The survey covers both participants and non-participants living in programme villages, as well as respondents in control villages where the programme is not yet active. We account for participation selection bias using instrumental variables. The control villages allow us to test the exclusion restriction and provide us with an effective control group to analyze programme impact. We find both direct effects and significant spillovers on non-participants. The impact of interventions might be substantially underestimated if such external effects were not taken into account.