A Problematic Map of Stop-and-Frisk
One of the biggest stories in America's biggest city this year has been the public debate about the NYPD's "stop-and-frisk" tactic. The general idea is that officers can stop someone if there's a reasonable suspicion the person will commit a crime, and can frisk the person if they suspect the presence of a weapon. City authorities say stop-and-frisk reduces illegal guns and crime. Opponents say it intrudes on civil liberties and leads to racial profiling.
A central point in the debate is the startling disparity between the high number of stop-and-frisk incidents that occur and the low number of illegal guns that are recovered. But this point can be interpreted two ways. Police officials say it proves the system is working (i.e. stop-and-frisks have deterred people from carrying weapons), while civil liberties advocates say it proves the system isn't (i.e. stop-and-frisks are excessive and ineffectual).
In an effort to clarify the debate, WNYC recently published a map linking the prevalence of stop-and-frisks with the location of gun recoveries. On the surface the map reveals a non-existent connection between the two: there's very little overlap between the map's stop-and-frisk "hot spots" and the places where officers found a gun. WNYC used the cartographic portrait to conclude that "police aren't finding guns where they're looking the hardest."
As it stands the map lends support to those who feel that most stop-and-frisks are conducted irresponsibly. (Fast Company reported on it with the headline: "How Dumb Are Stop-And-Frisk Tactics?") But the map contains a number of flaws that may cloud the debate it's attempting to clarify, according to an analysis performed by Steven Romalewski, the mapping director of the Center for Urban Research at CUNY.
"When it got posted, so many people on Twitter and on blogs picked it up — not just re-tweeting but saying it's an amazing visualization that emphasizes the problem of stop-and-frisk," says Romaleswski. "Separate from whether it refutes or supports stop-and-frisk, it's not reflective of comparing these two data sets. ... As a map that's striking, it worked. But I don't think it helped, necessarily, to understand the nuances and the issues about stop-and-frisk.
At his blog, Spatiality, Romalewski examines the WNYC map at length and points out several shortcomings. The complete post is highly recommended, but here we'll focus on the biggest problem described by Romalewski: the map's poor definition of a stop-and-frisk "hot spot."
The map divides the city into Census blocks and highlights stop-and-frisk hot spots, where NYPD officers engaged in at least 300 stop-and-frisks. But the complete set of data shows no "natural break" at 300, writes Romalewski, and using that figure seems unnecessarily narrow. The areas of New York with 300 or more stops account for less than .7 percent of all blocks in the city only less than 20 percent of all stop-and-frisk incidents.
"There's no rationale for it," says Romalewski of the WNYC threshold. "It's arbitrary, as opposed to something that's grounded in the data."
Romalewski re-evaluated the data and concluded that 38 stop-and-frisks per block was a more natural "hot spot" threshold than 300. Areas with 38 or more stops account for 10 percent of city blocks and 70 percent of all stop-and-frisk incidents. Using this new "hot spot," Romalewski created another map that shows a far stronger relationship between stops and gun recoveries than the one published by WNYC. Here's a comparison of the two maps, with stop-and-frisks in purple and pink, and gun recoveries in green:
Just by using a different "hot spot" measurement, Romalewski devised a stop-and-frisk map that essentially supports the opposite conclusion about the program's effectiveness. He writes (original emphasis):
The pink-to-hot pink blocks in the second map account for 433 recovered guns, or 56% of the total in 2011.
That's not to call stop-and-frisk legitimate. Romalewski notes in his post that he's "not defending" the practice (and for that matter neither are we). His point, as an impartial cartographer, is simply that there were so few gun recoveries (779) and so many stop-and-frisks (more than 685,000) that it's tough to map them together in a meaningful way. Basic statistics make the case clear enough: according to a recent federal court decision on stop-and-frisk, for every 69 stops a police officer made on the basis of a suspected weapon from 2004 to 2009, only a single gun was found.
"You don't need a map to make that point," says Romalewski, who closes his post with some cautionary words:
No question that a massive number of stop and frisks have been taking place in the last few years with very few resulting in gun recovery. But simply mapping the two data sets without accounting for underlying data patterns, temporal trends, and actual hot spots rather than artificial block boundaries risks jumping to conclusions that may be unwarranted. When you’re dealing with an issue as serious as individual civil rights and public safety, a simplified approach may not be enough.
Top image: Eduardo Munoz/Reuters