Between 2015 and 2018, New York City adopted “neighborhood policing,” an expansive policy to encourage interactions between police officers and community members. Among other changes, the initiative established hundreds of new “neighborhood-coordination” officers and gave “steady-sector” officers time away from 911 response to dedicate to resident interactions. This study evaluates the initiative’s effects on crime, complaints of police misconduct, racial disparities, and arrests. Using monthly data on New York City’s 76 police precincts between 2006 and 2019, we estimate the policy’s causal effect using high-dimensional time series models. This approach accounts for the policy’s staggered adoption, addresses potential correlation among outcomes and between precincts, and controls for unobserved precinct characteristics. We find neighborhood policing reduced misdemeanor and proactive arrests, especially in higher-poverty precincts outside of Manhattan, though it did not change the racial disparities of such arrests. The policy did not affect crime. It briefly increased complaints against police.

Objectives: New York City implemented an expansive community policing program, called “neighborhood policing,” between 2015 and 2018. This study examines the initiative and evaluates its effects on crime, arrests, and racial disparities.Methods: We gather monthly data on New York City’s 76 residential police precincts between 2006 and 2019. We then use a novel method to estimate the causal effect of neighborhood policing that couples high-dimensional time series models with machine learning methods. The approach accounts for the policy’s staggered adoption by different precincts at different times and addresses the potential correlation among outcomes across precincts. In the context of all precincts eventually adopting the new policy, our approach controls for all unobserved precinct characteristics and allows for the examination of heterogeneous treatment effects across different communities. Further, the method is empirically validated using realistic simulation studies of the data before the initiation of neighborhood policing to assess the robustness of the approach to untestable assumptions.Findings: We find neighborhood policing did not impact crime. It did reduce misdemeanor arrests and proactive arrests, especially in high-unemployment precincts outside of Manhattan. The policy reduced the racial disparity in proactive arrests in high-unemployment areas and increased it in low-unemployment areas.Conclusions: Though New York’s leaders promoted neighborhood policing as a method to reduce crime, our findings echo other literature showing community policing interventions do not impact crime rates, at least in the short term. City policymakers who want to reduce low-level arrests might pursue community policing, as it decreased such arrests.