The New York Police Department is on track to do 700,000 stop-and-frisks this year, the most in its history. Twelve percent of these stops meet the NYPD’s current criteria for success: An arrest or summons. Put another way, 88% of these stop-and-frisks yield no such result. At least some of those “negative results” from stop-and-frisks have value (e.g., I decided not to put my gun in my belt before I went to settle an argument with another drug dealer because I was afraid a police officer would stop-and-frisk me, and that in fact happened). But some portion of the negative results truly are a dead weight loss, not just in police time but in how they make law-abiding people angry at the police and feed racial hostility more broadly throughout the city.
Modern policing, including in New York City, is all about data monitoring, targets and performance management. Is there a performance management strategy that would increase the benefits and reduce the costs of stop-and-frisks?
Currently, police are rewarded when they make a stop-and-frisk that meets their success criteria (arrest or summons) but are not punished when they make an unnecessary stop-and-frisk. That creates an incentive to do as many stop-and-frisks as possible with a fairly low threshold for suspicion. A better performance management system would instead reward police for the proportion of their stop-and-frisks that were successful
The first challenge for such a system would be deciding what would count as a “successful stop-and-frisk”, because you don’t want it to be something that the officer’s judgment can influence on the spot, like an arrest (one of the current success criteria). Sometimes, the very best thing a cop can do when for example s/he finds a few cans of spray paint on a tagging teenager, both in terms of resource allocation and in terms of improving relationships between cops and citizenry, is to take the contraband and say “I am letting this one pass as a favor to you – go and sin no more.” If the standard of justifiability were the number of arrests per stop-and-frisk, cops would be tempted to make arrests in situations like that when they shouldn’t. A more objective success standard with less perverse potential would be something like calling a stop-and-frisk successful if the suspect turned out to have an outstanding warrant or was carrying an unlicensed firearm.
The second challenge for performance management is figuring out what the right proportion of success should be. As mentioned, good, crime-preventing policing will generate at least some stop-and-frisks in which the person for example chose to leave their unlicensed gun at home, so the performance standard can’t be 100%. In performance management, you don’t have to know the precise numerical goal to know which direction you want to move current practice. If the proportion of stop-and-frisks that were successful by the criteria I have proposed is currently say, 10%, police leaders could simply say that individual officers and precincts which beat that standard would be rewarded, which would move the ball in the proper direction. In itself, this would be a system change in that stop-and-frisks would move in officers’ minds from being an unlimited resource to a budget line item…do I want to spend one of my valuable stop-and-frisks on this guy when I really don’t have very strong suspicions?
The third challenge is to avoid gaming that could occur through under-reporting of the denominator. That is, a group of officers could agree among themselves that they will increase their success rate by doing as many stop-and-frisks as ever, but not reporting the unsuccessful ones as often. One way to get around this problem is to do community audits in the neighborhoods where police work. You would need an independent assessment unit to conduct surveys of people and ask whether they had been stop-and-frisked in the past 30 days (or whatever) as well as whether they had seen someone else stop-and-frisked. The assessor would declare shenanigans when police reports of stop-and-frisk rates were wildly variant from community reports. Mark Kleiman suggested to me that technology could also help, for example videotaping stop-and-frisks much as police already often videotape traffic stops.
Setting up such a performance management system would be a lot of work for the police and would require political leadership, but I actually think the police would benefit as much as would the citizenry at large. Imagine you are a police officer setting off for your morning patrol in two neighborhoods that are similar except that in one, when your colleagues stop-and-frisk a suspect it’s usually successful and in the other it usually isn’t. In which neighborhood would you feel safer? In which would you expect more cooperation from witnesses? And in which would you feel like part of the community versus a distrusted outsider?