Juking the Stats

A recent report of retired New York City police officers warns that the NYPD is playing fast and loose with the numbers. Knowing when and where crimes occur is essential to good policing and Compstat, a system of crime-data analysis created in 1994, played a large role in bringing down crime in New York City. But ever since, numbers have ruled the NYPD’s roost. If crime numbers are not down, precinct commanding officers need numbers to show they’re doing something—something quantifiable.

In the police world, two statistical categories are important: Part I felony crimes reported to the FBI’s Uniform Crime Statistics and internal measures of “productivity,” namely arrests, citations, and summonses. There are ways to play with both. But perhaps surprisingly, the police department’s emphasis on the latter, the so-called productivity stats, is a much greater cause for concern.

Sergeants, lieutenants, captains and inspectors feel intense pressure to produce ever better stats. To some extent this can be good. Police are paid to work. But the pressure to produce more with less is as overwhelming as it is unrealistic. Mind you, the orders never come from above to just make numbers up, but when commanding officers talk about “productivity,” the rank-and-file hear “quotas.”

“I’d love it if I always had enough good C’s [criminal citations], but I need numbers,” one officer told me, “And if I don’t have enough stats and Compstat is coming up, I don’t care if they’re bullshit. I’ll take whatever the f— I can get!” In a world where “better stats” and “more stats” are synonymous, the tail is wagging the dog. And police are nothing if not creative in finding ways to please their bosses.

Officers know what they see on the streets. Any desk sergeant who reclassifies or “corrects” a report sends a terrible and destructive message. And these pressures have grown substantially in the past decade.

When a $2,000 stolen laptop model can be found on EBay for less than $1,000, a felony larceny might be reclassified as a misdemeanor and all but disappear from the stats. Or say a tourist reports a robbery but the police, knowing she’s on the next flight back to Germany, record her loss as lost property.

Of course statistical errors can run both ways. There’s a lot more false reporting of crime than the public realize, and police are certainly not fools. That German tourist may have simply wanted a police report to scam insurance money. Real life is not easily quantifiable, and trying to determine which bubble on a report best reflects reality leaves lots of room for honest interpretation.

For statistical errors, data are supposed to be small and random. But for crime data, we’ll generally settle for errors as long as they’re consistent. Given that the distinction between felony and misdemeanor is basically arbitrary anyway, it doesn’t really matter if ten percent of felonies are reclassified as misdemeanors as long as it’s done every year. After all, far more than ten percent of crimes are never even reported.

The problem with fudging crime numbers for political gain is that you can’t stop. You have to do it every year just to stay even. Eventually you’ll get promoted and transferred, if you’re lucky, leaving your more honest and naïve replacement to deal with surprisingly bad crime numbers.

Certainly some stats, like murders and car thefts, are more reliable than others. The former are hard to fudge and the latter are generally reported for insurance reasons. And by these measures, the drop in violent crime is impressively clear. Murders alone are down 70 percent from their 1994 peak and 11 percent in the last year alone. This is real. These numbers matter.

But too many measure of police “production” do little but produce internal stats and pad officers’ overtime pay. Take low-level marijuana possession arrests. In 1994 there were 3,141 of these in New York City. In 2008 they had exploded to 40,383! This 1,285 percent increase was not the result of a epidemic of marijuana possession but a simple change in police tactics.

To say these arrests caused the crime drop is absurd, akin to claiming that a parking-ticket blitz prevents traffic deaths. These arrests—at great taxpayer expense and motivated only by internal police pressure to produce “stats”—simply pad officers’ overtime pay while sending tens of thousands of mostly poor minority men through the criminal-justice system.

Messy as they may be, it’s hard to image a police world were numbers didn’t matter. What’s important is that these numbers aren’t produced for their own sake. Statistics need to stay focused on crime and not internal, malleable, and ultimately destructive measures of “productivity.” The hard-working men and women of the NYPD deserve as much.