About . . . . . . Classes . . . . . . Books . . . . . . Vita . . . . . . . Links. . . . . . Blog

by Peter Moskos

September 9, 2016

It Depends What Your Definition of "Rose" Is

What's your definition of "rose"? Leaving aside the one that "would smell as sweet," mine is "goes up." I ask because the headline in the New York Times says: "Murder Rates Rose in a Quarter of the Nation’s 100 Largest Cities."

When you say homicides "rose" in 25 of 100 cities, you might think it didn't rise in 75 cities.
Since crime rates fluctuate from year to year, we used a statistical technique to determine places where we can definitely say rates were rising.
A "statistical technique"? What might that be? Praytell, Times, praytell. My advanced quantitative methods are rusty (and were never good), but I'd love to know what you do.

If the data were presented in some kind of useful fashion (they're not) you'd see something similar to Prof. Richard Rosenfeld's solid research (pdf link). Rosenfeld looked at 56 cities and found an increase in 40.

You can't really tell from the Times' crappy graphic, but in the 70 cities "Where murder rates...," I count 27 cities where murder rates "fell slightly." Combined with the 5 with significant decreases, that leaves 68 of the nation's 100 largest cities where the murder rate -- what's the word? -- rose. (Which is consistent with Rosenfeld's 40 of 56.) [It's worth mentioning that those who write the story don't write the headlines. In the story it's clear that murders rose "significantly" in 25 cities.]

Now of course as a PhD, I'm supposed to use five-dollar words when fifty-cent words will do:
Cities are obviously heterogeneous. There is tremendous variation across the largest cities in basic features such as demographic composition, the concentration of poverty, and segregation that relate to city-level differences in rates of violence.
O.K.

As an academic, I'm expected to endorse platitudes like:
There is no consensus on what caused the recent spike.
And
Many crime experts warn against reading too much into recent statistics.
And I should urge restraint, lest we get carried away with caring about murder. (My fear: restraint will lead to a right-wing law-and-order backlash). Also, apparently, I'm not supposed to worry about murder until more murder is up in every damn city in America. Nor should I worry about homicide because it's been worse in the past. (An interesting argument, I note, should one apply it to poverty, racism, lead, infectious disease, or police-involved shootings. But I digress.)

In terms of numbers (in what I would call burying the lede):
Nationwide, nearly 6,700 homicides were reported in the 100 largest cities in 2015, about 950 more than the year before.
That's a 16.5 percent increase. In one year? That, my friends, is huge. Now the nationwide percentage increase will almost certainly be smaller, but the last time there was even a double-digit percentage increase in homicide was 1968. That last time the homicide numbers increased by more than 1,000 was 1991.

Back in January, based on less data, I guessed that 2015 would see about 1,500 more murders than 2014. Gosh, am I a swamy? No, just somebody who can remove the ideological blinders long enough to use a calculator. I even offered an open $100 bet to anybody who said, "We don't know if homicides are up." Nobody put their money where their mouth was. Odd. It's like they didn't even believe what they were saying.

If we focused on the carnage instead of arguing about reality and methodology, you see, we'd have to consider the why? And then, perhaps, we'd notice that increased violence isn't really linked to any change in poverty or gun laws or even legitimacy. Perhaps we'd take note, as have Professor Rosenfeld and myself, that the cities where violence is most up are the cities where police have been, to put it mildly, in the news (or even charged criminally for no good reason). Perhaps crime is up because police are doing exactly what we're asking them to do: be less proactive and have fewer interactions with the public.

3 comments:

Thos Wallace said...

Its virtually impossible to explain away Chicago. The Times 'expert' cites tearing down the worst, high rise projects and arresting major gang leaders. These 'improvements' having unexpected consequences.

Meanwhile, we have not only detailed data of a very sharp, indisputable increase in homicides, but also reasons. Those reasons being articulated by police rank and file, per blog 2nd City Cop and also police superintendent Eddie Johnson
:
"The violence right now in Chicago is completely unacceptable. We know who's driving it. We have a strategic subject list. That list documents people. It's a computer software program that documents who will most likely be victims of gun violence or the perpetrators of gun violence.

"This past Memorial Day weekend we had 66 people shot," he added. "Eighty percent of those individuals were on this list that I'm talking about. Eighty percent. One hundred percent of the identified offenders from those shootings were on that list. One hundred percent."

It isn't 3 million people running amok. It is a couple thousand. There is now zero tolerance for bad outcomes from aggressive policing in Austin and Englewood.

Thos Wallace said...

Johnson is reerring to the Strategic Subject List.

http://www.nytimes.com/2016/05/24/us/armed-with-data-chicago-police-try-to-predict-who-may-shoot-or-be-shot.html?_r=0

Someone on the list has a personal, custom 'homicide rate' of 15,000 per 100,000.

To be able to identify future homicide 'subjects' so precisely is a strong proof of concept. In fact, if homicides are concentrated in such a small population, it calls into question any use of classical statistical techniques. Independent and identically distributed random variables, no?

From the description, the list is based mostly on data from police encounters -- and doesn't draw on most 'big data' types of analytics.

Adrian said...

@Thos, I believe the Strategic Subjects list is based solely on the subject's criminal history (looks like mainly type and tempo) and the criminal history of their associates.

"The Strategic Subjects List (SSL) ... is generated based on empirical data that lists attributes of a person's criminal record, including the record of violence among criminal associates, the degree to which his criminal activities are on the rise, and the types and intensity of criminal history."

http://directives.chicagopolice.org/directives/data/a7a57bf0-1456faf9-bfa14-570a-a2deebf33c56ae59.html