Deconstructing data on police deaths

Okay, here's a strange one. The headline in the New York Times Monday read: "Even as Violent Crime Falls, Killing of Officers Rises," reporting "a 25 percent increase" in 2011 in police officer homicides "and a 75 percent increase from 2008. Startling data, huh? But the truth is more complex. While police officer deaths did rise in 2011, they've also dramatically declined so far in 2012, as Radley Balko pointed out over the weekend. Here are the data for 2012 so far compared to last year via the National Law Enforcement Officers Memorial Fund:

Law Enforcement Officer Fatalities
Preliminary 2012 Numbers
April 11, 2012

	2012	2011	% Change
Total Fatalities	31	59	-47%
Firearms-related	11	26	-58%
Traffic-related	12	19	-37%
Other Causes	8	14	-43%

Why wasn't THAT the news hook? Such preliminary data were good enough to hype last year. IMO the answer is that sensationalistic bad news draws more readers than stories about positive trends. Journalists have a natural schadenfreude that encourages them to frame stories in the most alarming ways, but good news travels slowly, even during the 24-hour news cycle.

In this case, the Times also cherrypicked the data a bit. Dating the 75% increase from 2008 obscures the fact that on-the-job officer deaths were much higher in 2007, when 189 officers died on the job compared to 141 in 2008. This graph depicting the long-term pattern on officer deaths shows the recent "trend" hyped by the Times is really more a regression to the mean:

Police Officer Deaths in the Line of Duty, 1961-2011

So the tendency analyzed in the Times is a bit of a construct. Reporters could just as easily have said the total number of police officer deaths declined 8.5% compared to 2007, but would Times editors have considered that news "fit to print"?

In addition, limiting their stat to police homicides ignores most on-the-job police deaths, which much more frequently happen because of accidents, often in traffic. In 2010, for example, 153 officers died on the job nationwide, but the FBI data on which the NY Times based its analysis counted 56 officers "feloniously killed" that year. So when calculating the increase in the Times story, most on-the-job police deaths weren't counted.

Anytime you're analyzing statistics involving such small numbers, special care is warranted. With more than 700,000 sworn officers in the United States, these small fluctuations are not necessarily statistically significant. Notably absent from the Times story was any analysis by a statistician on whether these short-term data fluctuations are meaningful. Instead, the article is filled with speculation about the reasons for a trend that may or may not exist.

Being a police officer is not remotely the most dangerous job out there, nor even the most dangerous government work. Excluding soldiers who die in combat, garbage collectors rank highest among government workers in the likelihood they'll be killed on the job. The people picking up your trash put their lives on the line every day and are more likely not to make it home at night than their brethren in blue. But one suspects we won't any time soon see a New York Times headline memorializing their sacrifice.