Document Type


Publication Date



Equilibrium models of racial discrimination in law enforcement encounters suggest that in the absence of racial discrimination, the proportion of searches yielding evidence of illegal activity (the hit rate) will be equal across races. Searches that disproportionately target one racial group, resulting in a relatively low hit rate, are inefficient and suggest bias. An unbiased officer who is seeking to maximize her hit rate would reduce the number of unproductive stops toward a group with the lower hit rate. An unbiased policing regime would generate no differences in hit rates between groups.

We use this framework to test for racial discrimination in pedestrian stops with data from the contentious “Stop, Question and Frisk” (SQF) program of the New York City Police Department (NYPD). SQF produced nearly five million citizen stops from 2004–2012. The stops are regulated by both Terry (federal) and DeBour (New York) case law on reasonable suspicion. Stops are well-documented, including a structured format for reporting the indicia of reasonable suspicion that motivated the stop. We exploit these data to examine the Floyd court’s claim. We decompose stops on the basis of suspicion, as reported by officers at the time of the stop. We conduct five tests to assess whether racial discrimination characterizes SQF stops: the allocation of officers relative to crime and population in specific areas, the decision to sanction conditional on a stop, the decision to arrest or issue a summons conditional on the decision to sanction, the efficiency of stops in seizing contraband including weapons, and updating processes by officers in their search activity. In each test, we include the reasonable suspicion rationale that officers indicated as the basis of the stop. We find consistent evidence of disparities in police responses to Black, Hispanic, and Black Hispanic civilians, and significant differences by race in the use of specific indicia of reasonable suspicion that motivate stops. The higher error rates for specific indicia of suspicion suggest that rather than individualized bases of suspicion, officers may be activating stereotypes and archetypes to articulate suspicion and justify street seizures.


Law | Law and Race | Law Enforcement and Corrections