The Dubious Math Behind Stop and Frisk


Chart courtesy of "Stop Question And Frisk Police Practices in New York."

Yesterday Ray Kelly took to the pages of the Wall Street Journal to defend NYPD's Stop and Frisk tactics and its indiscriminate spying on Muslim communities:

Since 2002, the New York Police Department has taken tens of thousands of weapons off the street through proactive policing strategies. The effect this has had on the murder rate is staggering. In the 11 years before Mayor Michael Bloomberg took office, there were 13,212 murders in New York City. During the 11 years of his administration, there have been 5,849. That's 7,383 lives saved--and if history is a guide, they are largely the lives of young men of color.

So far this year, murders are down 29% from the 50-year low achieved in 2012, and we've seen the fewest shootings in two decades.

To critics, none of this seems to much matter. Sidestepping the fact that these policies work, they continue to allege that massive numbers of minorities are stopped and questioned by police for no reason other than their race.

As one of Ray Kelly's critics, and a citizen of New York, I will say that the declining murder rate matters a great deal. But the question before us isn't "Do we want the murder rate reduced?" The question is "Is Stop and Frisk a moral and effective policy?" We could also start punishing all murderers with public torture and beheading. That too might reduce the murder rate. Or perhaps the murder rate might fall for less conspicuous reasons, and those who endorsed public beheadings can loudly claim the credit anyway. At least we'd have correlation. Presently that is more than you can say for Stop and Frisk. Kelly rightly points out that the murder rate in our great city is falling. But for some reason he neglects to mention that Stop and Frisk numbers are falling too.

Perhaps there is some relationship between the long drop in homicides and Stop and Frisk, but Ray Kelly has never furnished such actual proof. Understanding why crime rises and falls has bedeviled social scientists for decades, so it's not surprising that Kelly would have trouble offering hard evidence. But we can certainly examine Ray Kelly's claim that Stop and Frisk is responsible for large numbers of weapons coming off the street.

During roughly half of all stops in 2008 (54.40% or 293,934 stops), officers reported frisking the suspect. Officers are legally authorized to pat down the outer clothing of a suspect in order to determine if the person is carrying a weapon. As shown in Figure 6, a very small percentage (1.24%) of total stops resulted in the discovery of a weapon of any kind (gun, knife, or other type of weapon). A slightly higher percentage (1.70%) resulted in the discovery of some other kind of contraband. Contraband is any item that is against the law to possess, including illegal drugs.

Given Ray Kelly's claims about saving black and brown lives, it's worth seeing how these numbers correlate to race:

In terms of recovering weapons and other contraband, stops of Whites yielded a slightly greater share, proportionally, of contraband other than weapons (1.98% versus 1.75%). The difference in the recovery of knives and weapons other than guns is greater among Whites as well (1.46% compared to 1.06%). In terms of recovering guns, the situation is reversed: proportionally, stops of Blacks and Hispanics were slightly more likely than stops of Whites to result in the recovery of a gun (0.17% versus 0.07%), but this difference is extremely small - 0.10%.

Finally, we should look at how the seizure of guns correlates to an increase in Stops:

While the total number of stops annually has climbed to more than half a million in just a few years (up from 160,851 in 2003), the number of illegal guns discovered during stops has remained relatively steady and modest in comparison. As Figure 8A shows, the number of guns recovered over this six-year period ranges from a low of 627 (2003) to a high of 824 (2008), averaging 703. It should be noted that over this same period, the number of stops more than tripled, meaning the yield of guns per stop has declined considerably (see Figure 8B).

Any serious proponent of Stop and Frisk must grapple with the fact that gun recoveries during Stops are vanishingly small, that they are vanishingly small regardless of race, and that there is little, if any, correlation between a rise in Stops and a rise in gun seizure.

The deeper and more poignant charge is not simply that Stop and Frisk is a bad tool for recovering guns, but that it amounts to systemic discrimination against black and brown communities. Ray Kelly frequently faults his opponents for measuring the demographics of Stop and Frisk against the demographics of the city. Kelly asserts that in a city where much of the violent crime is committed by black and brown males, it is logical that they would constitute the majority of the stops.

I agree with Kelly that it is not particularly telling to look at census data and extrapolate. It would be much more telling if we could somehow control for the actual commission of crime and then see if there was any bias in Stop and Frisk.

In the period for which we had data, the NYPD's records indicate that they were stopping blacks and Hispanics more often than whites, in comparison to both the populations of these groups and the best estimates of the rate of crimes committed by each group. After controlling for precincts, this pattern still holds. More specifically, for violent crimes and weapons offenses, blacks and Hispanics are stopped about twice as often as whites. In contrast, for the less common stops for property and drug crimes, whites and Hispanics are stopped more often than blacks, in comparison to the arrest rate for each ethnic group.

That was the conclusion of Columbia professor of Law and Public Health Jeffrey Fagan in 2007. Perhaps, since then, Ray Kelly has managed to craft a bias-less policy of Stop and Frisk:

NYPD stops are significantly more frequent for Black and Hispanic citizens than for White citizens, after adjusting stop rates for the precinct crime rates, the racial composition, and other social and economic factors predictive of police activity. These disparities are consistent across a set of alternative tests and assumptions.

That is from Fagan's 2010 study. It's important to understand that this data is widely available to the public. So when you hear Ray Kelly say something like this...

"It makes no sense to use census data, because half the people you stop would be women." should understand that he is not telling bold truths, he is confronting the weakest arguments he can find.

Kelly offers some apparent sympathy, conceding that it is "understandable that someone who has done nothing wrong will be angry if he is stopped." But that category of people stopped who've "done nothing wrong" and are understandably angry are not a small minority, but a large majority of the people being stopped and frisked:

Arrest rates take place in less than six percent of all stops, a "hit rate" that is lower than the rates of arrest and seizures in random check points observed in other court tests of claims similar to the claims in this case.

I am not totally opposed to policies in which individuals surrender some of their rights for the betterment of the whole. The entire State is premised on such a surrendering. But at every stop that surrendering should be questioned and interrogated, to see if it actually will produce the benefits which it claims. In the case of Stop and Frisk you have a policy bearing no evidence of decreasing violence, and bearing great evidence of increasing tension between the police and the community they claim to serve. It is a policy which regularly results in the usage of physical force, but rarely results in the actual recovery of guns. But don't take my word for it. Take Ray Kelly's:

"A large reservoir of good will was under construction when I left the Police Department in 1994,'' Mr. Kelly said. ''It was called community policing. But it was quickly abandoned for tough-sounding rhetoric and dubious stop-and-frisk tactics that sowed new seeds of community mistrust.

That was 13 years ago. Times have changed. The evidence has not.