Association for Public Policy Analysis & Management

Truancy is a risky behavior that predicts many adverse outcomes. We develop a strategy to use administrative data to construct social networks based on students who are truant together. We simulate these networks to document that certain students systematically coordinate their absences in the observed data. We show these networks are meaningful by leveraging a parent-information intervention on student absences in which we find spillover effects from treated students onto peers in their network. Excluding these effects understates the intervention’s cost effectiveness by 43%. We develop an algorithm treatment-effect heterogeneity identified by machine-learning techniques to show there is potential to use these networks to target interventions more efficiently given a budget constraint.