Racial Differences in Police Use of Force

In an NBER working paper released today, Roland Fryer, Jr. uses the NYPD Stop, Question and Frisk database, the Public Police Contact Survey,  to conduct "An Empirical Analysis of Racial Differences in Police Use of Force." The paper also uses data collected by Fryer and students coded from police reports in Houstin, Austin, Dallas, Los Angeles, and several parts of Florida. The paper is worth reading in its entirety, and is also the subject of a New York Times article, which summarizes the main findings more thoroughly than I will do here.

Fryer estimates odds ratios to measure racial disparities in various types of outcomes. An odds ratio of 1 would mean that whites and blacks faced the same odds, while an odds ratio of greater than 1 for blacks would mean that blacks were more likely than whites to receive that outcome. These odds ratios can be estimated with or without controlling for other variables. One outcome of interest is whether the police used any amount of force at the time of interaction.. Panel A of the figure below shows the odds ratio by hour of the day. The point estimate is always above 1, and the 95% confidence interval is almost always above 1, meaning blacks are more likely to have force used against them than whites (and so are Hispanics). This disparity increases during daytime hours, with point estimates nearing 1.4 around 10 a.m.

Panel B shows that the average use of force against both blacks and whites peaks at around 4 a.m. and is lowest around 8 a.m. The racial gap is present at all hours, but largest in the morning and early afternoon.

Fryer builds a model to help interpret whether the disparities evident in the data represent "statistical" or "taste-based" discrimination. Statistical discrimination would result if police used race as a signal for likelihood of compliance of likelihood of having a weapon, whereas taste-based discrimination would be ingrained in officers' preferences. The data are inconsistent with solely statistical discrimination: "the marginal returns to compliant behavior are the same for blacks and whites, but the average return to compliance is lower for blacks – suggestive of a taste-based, rather than statistical, discrimination."

Fryer notes that his paper enters "treacherous terrain" including, but not limited, to data reliability. The oversimplifications and cold calculations that necessarily accompany economic models  never tell the whole story, but can nonetheless promote useful debate. For example, since Fryer finds racial disparities in police use of violence but not shootings, he notes that "To date, very few police departments across the country either collect data on lower level uses of force or explicitly punish officers for misuse of these tactics...Many arguments about police reform fall victim to the 'my life versus theirs, us versus them' mantra. Holding officers accountable for the misuse of hands or pushing individuals to the ground is not likely a life or death situation and, as such, may be more amenable to policy change."