Thursday, May 02, 2013
Statistical Shenanigans, NYPD Edition
The NYPD Commissioner, Ray Kelly, defended current stop-and-frisk practices today by noting two statistics:
1. 75% of all violent crimes are committed by African-Americans.
2. 53% of all those stopped are African-American.
The conclusion: "So really, African-Americans are being understopped in relation to the percentage of people being described as being the perpetrators of violent crime."
Put aside for a moment the dubious reliability of racial identification by eye-witnesses in general, and crime victims in particular. Evern putting that aside, and assuming this quote is not taken out of context or in any way mangled, Kelly compares apples and oranges in a very subtle, but very important, way.75% of violent crimes are reportedly by African-Americans, but 53% of all stops are of African-Americans.
That's not a fair comparison. Either compare stops for any reason to crimes of any type or compare stops for violent crime to violent crime incidents.
So think of it this way. Assume violent crime happens 1/10 as often as property crime (a roughly-fair approximation). Assume, too, that African-Americans are the perpetrators of 75% of violent crimes. And assume that there is one stop for each crime committed (a preposterous assumption, but keeps the example simple).
Perhaps the NYPD stops African-Americans in 95% of all violent crime cases. If they stop African-Americans in 48.8% of all property cases, then African-Americans will be stopped 53% of the time overall. But if African-Americans commit only 35% of all property crimes, then they are oversampled in both violent and property stops.
The comparison he makes is thus uninformative, since we have no data on the non-violent side of things.
And that is before we even start asking how stops map to crimes: it isn't that each single crime generates a single stop, especially given that many stops are unrelated to any underlying crime at all.
Now, it is true that African-Americans commit violent crimes at rates higher than their share of the population: about 50% of all murders (and thus about 50% of all murder victims, since murder, like most crime, is predominantly intraracial) are commited by African-Americans. But the comparison that Kelly is drawing here simply isn't the right one.
I'm not making any sort of argument about motive: maybe it was a malapropism, maybe it was a sincere effort, maybe it was an attempt to pull the wool over our eyes. But whatever it was, it wasn't a viable justification for the stops ratio.
Posted by John Pfaff on May 2, 2013 at 01:59 PM | Permalink
TrackBack URL for this entry:
Listed below are links to weblogs that reference Statistical Shenanigans, NYPD Edition:
I agree that the statistic doesn't make sense, but I 'm not sure I understand what it means to be stopped for a violent vs. non-violent crime given the Terry standard. Let's say someone is acting as a lookout for a conspiracy, and the officer sees the lookout, notes his suspicious conduct, and stops him. When the stop occurs, the officer may or may not learn about the underlying crime, and the underlying crime may turn out to be something violent (an assault), non-violent (stealing a car), or something that has both violent and non-violent aspects to it (say, drugs being sold by a gang). In which of those scenarios is the person stopped for a violent crime, and in which are they stopped in a non-violent crime?
Posted by: Orin Kerr | May 2, 2013 3:32:08 PM
I agree. I was roughly using it as an ex post definition: stops that result in arrests for violent crimes, or perhaps stops based on specific violent crimes ("we are looking for the man who mugged Sally, and we think you are he"). And I assume that each crime had one stop associated with it, and each stop was associated with a crime. (Going way, way back to HS math, I believe this is one-to-one and onto mapping.)
I think the real-world ambiguity of what is going on with a stop makes the comparison Kelley is drawing even harder to defend. If stops and crimes have a 1-1 and onto mapping, then there is a logical stops-to-crime comparison, but he isn't making the right one. If stops and crimes don't have a 1-1 and onto mapping--and they don't--then it isn't even clear what the right logical comparison is between stops and crime.
Posted by: John Pfaff | May 2, 2013 3:40:02 PM
I share Orin's uncertainty about what would constitute "stops for violent crime." That said, the New York policing strategy targets for aggressive stop-and-frisk policing statistical hot spots of violent crime. In that sense, the strategy probably involves about the best proxy one can produce for "stops for violent crime." And, of course, violent crime declined dramatically in New York in the wake of stop-and-frisk, with reductions in victimization disproportionately experienced among African Americans. Indeed, there is pretty impressive evidence to support the view that aggressive stop-and-frisk targeted at statistical hot spots of violent crime reduces violent crime: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1645436 Call it correllation if you will, but with so many African American lives at stake, we should not lightly condemn New York's policing strategy. If anything, New York may be a victim of its own success. Perhaps the NYPD was so encouraged by the data that it pushed stop-and-frisk past the point of diminishing returns. But, any strategy that focuses on statistical hot spots of violent crime will produce disproportionate minority stop rates, given the demographics of high-violent-crime neighborhoods.
Chapman University School of Law
Posted by: Larry Rosenthal | May 2, 2013 3:50:15 PM
John: imagine research finds that African-Americans commit at least 53% of all crimes in NYC. Would that change anything for your analysis? I rather suspect that the 53% of the overall crime rate is a realistic number, given that total crimes rates are heavily influenced by drug crimes and related property offenses. So, is your objection merely that the police chief mentioned the wrong statistic, even though your preferred statistic would point to the same result?
Posted by: curious | May 3, 2013 11:50:19 PM
Looking at any correlation between the rate of criminal conduct and the rate of stop and frisk seams to miss the point. If an individual had committed a robbery in the last day, week, or month, the act of stopping that individual won't help. Stop and frisk will only be a "success" from the stand point of police when and individual has an outstanding warrant or material that provides probable cause for arrest, which in practice will mean controlled substances, firearms, or some other sort of contraband. If someone had committed a robbery or even a violent crime and is not carrying such proof of the crime on their person, stop and frisk will not give the police a ground for arresting the individual. Therefore, stop and frisk is largely irrelevant to a hold swath of crime, perhaps with exception of the moments immediately following the incident.
Posted by: just a law student | May 4, 2013 11:24:07 PM
Just a law student:
Stop and frisk, if done in volume, makes it risky to carry guns and drugs in public places. The perception of increased risk associated with carrying guns and drugs reduces the rate at which they are carried in public, and that turns out to be a terrifically effective way to reduce violent crime. Read the article cited above.
Posted by: Larry Rosenthal | May 5, 2013 2:02:33 AM
No body likes Kelley, but lets get real about NYC. The quibling about percebntages means nothing, because Blacks in NYC are disproportionately involved in other crimes as well. NYC is heavily Black and Hispanic. Even so, these two groups in NYC are involved in disproportionate to their number of violent crimes, property crimes, drug dealing crimes. Scream all you like but thems the facts. They are going to get stopped and frisked more than other groups, simply because larger numbers of them are carrying concealed weapons and drugs for sale. It's only blind liberals in NYC who can't deal with those facts. Given the number of Hispanic offenders it's possible that they are not being subjected to stop and frisk in great enough numbers, but really it's only libs who are desperate to return to pre stop and frisk policing and enormously higher crime rates. If only libs were the victims here it might be fitting, but that is not the case, and for the welfare of all, the current policing works just fine.
Posted by: leeada47 | May 5, 2013 8:09:15 AM
Larry wrote that "Stop and frisk, if done in volume, makes it risky to carry guns and drugs in public places. The perception of increased risk associated with carrying guns and drugs reduces the rate at which they are carried in public, and that turns out to be a terrifically effective way to reduce violent crime." It also seems to be a terrifically effective way to make an entire group of law-abiding citizens feel victimized for where they live or for their skin color.
Even if stop and frisk is primarily responsible for the overall drop in violent crime in NYC, it also appears to encourage police officers to stop some wholly innocent people (often repeatedly over the course of months and years) and touch ("frisk") them. If we are going to trade off security for everyone at the expense of forcible, if temporary, detentions, it seems that this is a reasonable conversation to have.
Posted by: Matthew Bruckner | May 5, 2013 9:04:11 AM
Professor Bruckner is, of course, quite right that this important issue merits discussion and debate. I do worry, however, that those of us who can afford to live in low-crime communities bring their own biases to bear on the "tradeoff" to which he alludes. In 1991, when the NYPD began to reform its policing strategy, the homicide victimization rate in New York per 100,000 population was 58 for Blacks, 44 for Hispanics, and 8 for whites. By 1998, it had declined to 17 for Blacks, 8 for Hispanics, and 4 for Whites. In other words, the vast majority of lives saved were Black and Hispanic. Beyond the lives lost, there is a considerable sociological literature about the impact that fear of violent crime has on populations that experience high rates of violent crime. When those costs are tallied, the costs associated with stop-and-frisk are put in a perspective that those of us who do not have to live with the risk of violent victimization as part of our daily lives may not fully appreciate.
Posted by: Larry Rosenthal | May 5, 2013 4:55:58 PM
Are you sure you are being fair Larry?
As I understand it, few experts agree on the reasons why NYC (and most of the country) have enjoyed a substantial decrease in the number of violent crimes. Your comparison of NYC's crime statistics in 1991 and 1998 mirrors an overall decrease in violent crime in this country and has not been, to my mind, proven to be related--at all--to "stop and frisk", CompStat, the "broken windows" theory of policing, NYPD leadership, etc. Of course, I'm no expert. I'm merely concerned that your invocation of these statistics leads one to believe that it is the NYPD's reformed policing strategy that led to the drop in violent crime in NYC.
Moreover, as I understand the matter, the NYPD's "stop and frisk" policy results in an arrest in approximately 10-15% of cases. In other words, 85-90% of cases could be construed to involve the wrongful detention of a person because of the place that they live or that they "fit the description" of the suspected criminal, which is often a young, non-white male.
Nevertheless, if community groups representing the folks who are disproportionately affected by the NYPD's "stop and frisk" policy want to support it, I--who live in a relatively low-crime community--would be glad to listen.
Posted by: Matthew Bruckner | May 6, 2013 9:21:10 AM
In fact, during the 1990s, New York experienced substantially larger reductions in crime than other large cities, even though it entered that decade at the median for large cities. My own article cites a fairly large body of empirical work that establishes a connection between stop-and-frisk and reductions in crime. If you prefer, consider the view of the eminent criminologist Frank Zimring in his latest book, THE CITY THAT BECAME SAFE: NEW YORK’S LESSONS FOR URBAN CRIME AND ITS CONTROL. As Professor Zimring explains, it is difficult to study the relationship between any particular policing tactic and crime rates, but the data makes it pretty plain that there is not likely to be any explanation for New York's striking crime decline that is unrelated to policing tactics. Beyond that, one of the few interventions that has shown demonstrable success in reducing violent crime is stop-and-frisk policing targeted at statistical hot spots of crime. See, e.g., COMM. TO IMPROVE RESEARCH INFO. & DATA ON FIREARMS, NAT’L RESEARCH COUNCIL, FIREARMS AND VIOLENCE: A CRITICAL REVIEW 230–35 (Charles F. Wellford, John V. Pepper & Carol V. Petrie eds., 2005); COMM. TO REVIEW RESEARCH ON POLICE POL’Y & PRACTICES, NAT’L RESEARCH COUNCIL, FAIRNESS AND EFFECTIVENESS IN POLICING: THE EVIDENCE 235–40 (Wesley Skogan & Kathleen Frydl eds., 2004).
As for hit rates, stop-and-frisk policing utilizes the reasonable suspicion standard of Terry v. Ohio, which "falls considerably short of satisfying the preponderance of the evidence standard." United States v. Arvizu, 534 U.S. 266, 274 (2002). Accordingly, it "accepts the risk that officers may stop innocent people." Illinois v. Wardlow, 528 U.S. 119, 126 (2000). Thus, hit rates at 10 percent or so may be just about right.
I don't want to pretend that the case for stop-and-frisk is airtight. Like all debates in criminology, it is plagued by methodological problems. But if there is even a substantial case to be made that stop-and-frisk has made an important contribution to the decline in violent victimization in New York, then we should surely be hesitant to truncate our commitment to this tactic.
Posted by: Larry Rosenthal | May 6, 2013 6:42:53 PM
As between you and I, you are the criminal law expert.
While I agree that "if there is even a substantial case to be made that stop-and-frisk has made an important contribution to the decline in violent victimization in New York, then we should surely be hesitant to truncate our commitment to this tactic," I am hesitant to endorse any policy for my city that intentionally stops innocent people 90% of the time in order to catch criminals the other 10%. Lots of policy may reduce violent crime that reasonable people would find inappropriate.
Maybe many of those people being stopped, frisked and not arrested are willing to trade off reduced personal freedom for them for greater security for all of us (themselves included). Any discussion about the effectiveness of a policing strategy would be incomplete to my (admittedly non-expert) mind without taking into consideration the collateral effects on those people we stop, frisk and release.
Posted by: Matthew Bruckner | May 6, 2013 9:30:21 PM