Fifteen years ago I was contacted by Jeff Fagan, a professor of public health at Columbia University, about a project he was working on with the New York State Attorney General’s office, studying the rate of police stops of whites, blacks, and Hispanics in New York City. Police records revealed that ethnic minorities were being stopped at a much greater rate than whites. Defenders of the system argued that the patterns of stops represented a reasonable set of choices given the information available to police. As Howard Safir, the police commissioner at the time, stated, “The racial/ethnic distribution of the subjects of ‘stop and frisk’ reports reflects the demographics of known violent crime suspects as reported by crime victims. Similarly, the demographics of arrestees in violent crimes also correspond with the demographics of known violent crime suspects.” This seemed implausible, though, given that minorities were stopped much more often than whites, even if the baseline being used was not population but rather their rates of arrests in the previous year.
One suggestion was that the data on ethnic breakdowns represented differential treatment at the neighborhood rather than the individual level: many parts of the city with dense populations of ethnic minorities also have high crime rates. To address this concern we performed a series of multi-level analyses predicting the rate of police stops for different ethnic groups for different categories of crime, controlling for precincts and previous year’s arrest rates. The disparity between whites and others remained large, indicating that the police were disproportionately stopping minorities even within local areas.
As we wrote at the time, these statistical patterns do not necessarily imply active discrimination on the part of the police. One possible explanation is that members of different ethnic groups are more or less likely to be hanging out on the street, where the stops are taking place. One might also argue that disproportionate stops are not such a problem because a stop, and even a frisk, is just a way for police to get information; indeed, the overwhelming majority of stops do not lead to immediate arrests. Yet another explanation, familiar to scholars who study stereotyping, has to do with the macro effects of micro-decisions. If, as indeed is the case, that a randomly-chosen African-American man is more likely to have a criminal record, compared to a white man, then in settings with little or ambiguous information it would be logical for a police officer to go with that background information. But all these choices, reasonable as they might be for individual cases, would have a negative aggregate impact on ethnic minority groups in the city.
When our report with these findings was released in 1999, I was surprised that the representatives of the city government and the police department did not justify their policies based on any of these just-mentioned arguments. Instead they questioned various aspects of our analysis and argued that the police had not stopped minorities at disproportionate rates. This line of reasoning may be appropriate — the data at hand are messy and can lend themselves to different conclusions — but it was unexpected to me to see the defenders of the stop-and-frisk policy directly confront a statistical report performed by outside experts, rather than taking what would seem to me to be the easier tactic of accepting the statistical evidence and simply reframing it as resulting from acceptable behavior.