Association for Public Policy Analysis & Management

The rise in broken windows policing has led to a substantial increase in arrest activity in the last twenty years, with US police departments now making over 10 million arrests annually. This surge in activity has been partly motivated by the argument that punishing juveniles for low-level offenses deters them from future criminal careers. However, relatively little research has provided guidance about the impacts of arrests on the future criminal behavior of arrested individuals. In our study, we use unique data on a sample of arrestees and a comparison group of individuals who are suspected of a crime but who were not arrested for the years 2014 to 2018 in Dallas, Texas. Our data collection is the first of its kind, linking a criminal incident from the initial 911 call to responding officers, any potential suspects, any arrests made, jail time by arrestees, and eventual criminal court outcomes. We leverage two methodological approaches to answer this question. First, we estimate the impact of an arrest on future recidivism using an event study design in our sample of criminal suspects. Second, we instrument for arrest outcomes conditional on being suspected in an incident, using the estimated arrest propensity for the officers that responds to the incident. Because we see the responding officers to all 911 calls regardless of whether a suspect is reported, we are able to correct for the possible selection of individuals into the suspects sample. Our results suggest that arrests lead to a short-run reduction in criminal activity but no long-run impact. This effect is specific to arrestees who are also jailed, suggesting the impact is due to incapacitation rather than a behavioral response. While our analysis is preliminary, we plan to focus on how these responses vary across arrestee ages and crime categories.