Using Network Data to Measure Social Returns and Improve Targeting of Crime-Reduction Interventions
Randomized controlled trials (RCTs) have been influential in shaping policy to address the stark racial and income disparities in criminal justice involvement. Yet crime-prevention experiments typically ignore the possibility of peer spillovers, which could bias treatment effect estimates in either direction. We propose combining four existing RCTs in Chicago (N > 12,000) with multiple administrative measures of social networks to estimate how changes in individual criminal behavior spread through local populations. In addition to quantifying how spillovers change the net effects of these interventions, we will test for heterogeneity in peer effects to determine which targeting strategies would be most effective in maximizing the social impact of an intervention. The results will improve our understanding of a set of influential experiments (and potentially many other RCTs), expand our knowledge of how people affect each other's criminal decision-making, and provide guidance to policymakers about how to leverage peer effects to maximize future program impacts.