The Statistics of Stop-and-Frisk
Browsing the recent ruling finding New York's stop-and-frisk policy unconstitutional, I found this argument particularly interesting:
I reject the testimony of the City’s experts that the race of crime suspects is the appropriate benchmark for measuring racial bias in stops. The City and its highest officials believe that blacks and Hispanics should be stopped at the same rate as their proportion of the local criminal suspect population. But this reasoning is flawed because the stopped population is overwhelmingly innocent — not criminal. There is no basis for assuming that an innocent population shares the same characteristics as the criminal suspect population in the same area. Instead, I conclude that the benchmark used by plaintiffs’ expert — a combination of local population demographics and local crime rates (to account for police deployment) is the most sensible.
This really gets to the heart of the racial-bias argument, and the use of the indefinite article -- an innocent population -- is crucial here. Obviously, this was not a cross section of the innocent population in these neighborhoods, but a population of people who caught the eye of the police. Is it true that such a population should racially resemble the overall innocent population rather than the criminal population, unless officers are illegally using race itself as a factor?
Not necessarily. Police departments use various factors in trying to locate criminals -- e.g., targeting known criminal hangouts within a neighborhood, focusing on time periods when crimes are likely to happen, and learning to spot suspicious behavior. Even if we restrict the discussion to people who are in fact innocent (about 12 percent of stops resulted in arrests or summonses), there is no reason to assume that the folks who legitimately meet these criteria will be a cross sample of the neighborhood's population along any demographic lines. In fact, almost nothing in American life breaks down neatly this way.
There are a few more damning numbers in the ruling and an expert report provided by the plaintiffs, however. For example, some advanced statistical analyses indicate that an area's minority population predicts the number of stops there even after accounting for crime rates.
Further, if police are targeting minorities for stop-and-frisk above and beyond their likelihood of being involved in a crime, we might expect searches of minorities to be less likely to uncover concrete evidence of criminal activity. We do see this to some extent: Whites were carrying weapons 1.9 percent of the time; the number for blacks was 1.1, Hispanics, 1.3. For "contraband other than weapons," the numbers are 2.3 percent for whites, 1.8 percent for blacks, and 1.7 percent for Hispanics.
However, Mayor Bloomberg has said he's specifically targeting guns, and on that front the trend runs in the opposite direction: 0.16 percent of stopped blacks were carrying guns, as compared with 0.07 percent of whites and 0.09 percent of Hispanics. And the trend for arrests and summonses is less pronounced, though we should bear in mind that the decision to take these actions is partly subjective:
More coverage of stop-and-frisk -- Heather Mac Donald for, and Justin Peters against -- in the morning update today.
Robert VerBruggen is editor of RealClearPolicy. Twitter: @RAVerBruggen