Policy Evaluation in the Presence of Spatial Externalities: Reassessing the Progresa Program

This paper considers whether and how the evaluation results of a large-scale social policy in rural Mexico are sensitive to the local concentration of program beneficiaries. Combining evaluation survey data with geo-referenced information on the areas covered by the intervention, we find evidence of positive spillovers across villages on school participation decisions. Those effects occur exclusively between households entitled to receive the program, and they tend to vanish among less adjacent villages and at higher degrees of program density. Analyzing treatment impacts at the level of sufficiently extended geographical areas seems crucial to account for such spatial externalities.