Document Type


Publication Date




In this paper, we exploit a policy experiment in the New York Police Department (NYPD) to test for bias in police stops. The NYPD launched Operation Impact in 2003 to change the scale of officer deployments. High crime areas were designated as “impact zones” and saturated with recent police academy graduates. These officers were encouraged to stop, question, and frisk (SQF) crime suspects as part of the NYPD’s overall crime-reduction strategy (MacDonald, Fagan, and Geller 2016). We focus on the expansion of impact zones in Brooklyn and Queens in July 2007. We use geographic data on the boundaries of the impact zones and the specific locations of recorded SQF encounters to test for racial bias in the outcomes from police stops. We use a difference-in-difference (D-D) framework that exploits time and place varying sources of variation in police incentives to stop criminal suspects. We combine the D-D identification with a doubly robust estimator to assure that similarly situated stops are compared in areas before and after impact zones were formed. If the police are not discriminating based on race of crime suspects, then changes in stop outcomes in areas affected by the impact-zone program should be proportional across racial groups relative to unaffected areas.


Law | Law and Race | Law Enforcement and Corrections


Copyright © 2019 by the American Economic Association.