Statistical Analysis of “Stop and Frisk”

+ Brandon Fuller

Given the ongoing debate about the NYPD’s “stop and frisk” program, I’m surprised that this recent working paper (gated) by Decio Coviello and Nicola Persico has not received more attention. The authors go beyond numbers about the racial incidence of stops, asking interesting questions about differences in the likelihood that a stop leads to arrest in New York City’s various police precincts.

The legal challenge to the NYPD’s use of the “stop and frisk” tactic is entering its second week. Police officers use the tactic to stop and question people they suspect of unlawful behavior and frisk people they suspect are carrying weapons. At the heart of the legal challenge are two claims: first, that the NYPD’s stops are motivated by race (rather than reasonable suspicion of crime), and second, that the tactic violates the right against unreasonable search.

Coviello and Persico’s work may shed light on the question of whether the NYPD’s stop, question, and frisk program is racially biased. The paper does not address the issue of whether the tactic itself is legal, but it does attempt to identify racial bias on the part of the NYPD officers making the stops.

The authors analyze NYPD stop, question, and frisk reports (UF-250 forms) for African American and white pedestrians between 2003-2011. Of the roughly 2.6 million stops they considered, 6 percent resulted in arrests. Most arrests fell into one of the following categories: possession of a weapon (27 percent); robbery (16 percent); rent gouging (12 percent); grand larceny auto (9.3 percent); burglary (8.7 percent). 

84 percent of the stops involved African American pedestrians, 16 percent white. This reflects the broader truth that:

New York City’s stop-and-frisk program disproportionally impacts minorities. The New York Civil Liberties Union makes this point forcefully by documenting that, in 2011, 52.9 percent of stops were of blacks, 33.7 percent were of Latinos, while whites accounted for only 9.3 percent of the stops.

As the authors point out:

This disparate impact is unfortunate, but should not be surprising if we believe that crime and therefore policing are disproportionally concentrated in minority-rich neighbourhoods.

Coviello and Persico also point out that disparate impact does not necessarily reflect intent to discriminate. The goal of their analysis is therefore to go beyond disparate impact—to find evidence that the police discriminate on the basis of race.

They begin by doing a “hit rate” analysis of stops. Of the white people stopped, they find that 1 in 15 were arrested. Of the African American people stopped, 1 in 17 were arrested. In other words, African American pedestrians were less likely to be arrested than whites conditional on being stopped. On first pass, one could take the disparity in hit rates to indicate police bias against African American people. If stopping an African American pedestrian is less likely to result in an arrest, unbiased officers would scale back stops of African American pedestrians and scale up stops of white pedestrians until the hit rates are equivalent.

The hit rate analysis, however, is not the end of the statistical story. Under the NYPD’s stop and frisk program, a police officer can only stop a person within the geographical bounds of his or her precinct. To demonstrate discrimination, the authors therefore need to find evidence of disparities in hit rates within precincts rather than across them. To see why:

…suppose hypothetically that of all the blacks and whites stopped in the Bronx 3% were arrested, and 6% of the blacks and whites stopped in the Financial District were arrested. If we aggregated the data from the two precincts we would mistakenly conclude that the police officers making the stops are biased against blacks, because in the aggregate sample most blacks are searched in the Bronx and have a 3% arrest rate, much lower than whites, most of which are searched in the Financial District. Thus the hit rate test carried out without controlling for precincts would be potentially biased, or more precisely, uninformative about the racial bias exhibited by police officers within each precinct.

After accounting for the fact that different precincts have different rates of arrest conditional on search, the authors find that African Americans are no longer less likely to be arrested conditional on being searched. (Recall that a lower hit rate would be suggestive of bias because officers would be stopping African Americans even though such stops are less likely to result in arrest.) Their results therefore fail to support the hypothesis that the NYPD’s stop and frisk program is biased against African Americans relative to whites.

It’s worth pointing out that the failure to find institutional bias in the stop and frisk program cannot be interpreted as evidence that there aren’t any racist officers in the NYPD. Nor does this result rule out in any way the possibility that the stops of the four plaintiffs in the ongoing federal case were racially motivated. Racist officers will almost certainly find their way into the department from time to time and the city has to have a continual strategy for weeding them out.

Coviello and Persico’s paper nevertheless seems like it can inform the city’s ongoing debate about the future of the NYPD’s stop, question, and frisk tactic.

Tile image by shadey_shades.

Back to top
see comments ()