Abstract:
In practice, social and development interventions are often targeted at groups or individuals with the largest expected benefits. In such cases, treatment effects are usually affected by selection on unobservable factors. We show that modeling the process of selective intervention placement allows us to correct for this and identify the Average Treatment Effect using observational panel data. We illustrate the estimation method using simulated data, as well as, data on a large-scale sanitation intervention in Mozambique. Our results provide a useful tool for the assessment of targeted policy interventions, and inform decisions on their continuation or replication.