WaPo won the Pulitzer Prize in the category of National Reporting in 2016 for their ongoing collation and reporting on people who were shot and killed by police during the year in the entire nation. The 2017 database is here, listing, as of today, 737 dead. Andrew Sullivan in New York magazine uses this database to analyze the current uproar concerning NFL players kneeling during the national anthem:
The Post has indeed found that there’s a strikingly consistent number of fatal police shootings each year: close to 1,000 people of all races. But that figure includes the armed and the unarmed. Fatal police shootings of the unarmed — the issue Kaepernick and Reid cite — are far fewer. In the first six months of this year, for example, the Post found a total of 27 fatal shootings of unarmed people, of which black men constituted seven. Yes, you read that right: seven. There are 22 million black men in America. If an African-American man is not armed, the chance that he will be killed by the police in any recent year is 0.00006 percent. If a black man is carrying a weapon, the chance is 0.00075. One is too many, but it seems to me important to get the scale of this right. Our perceptions are not reality.
Since Andrew is not citing how many black men are armed, I frankly don’t have a lot of faith in this math. I should think the proper math for calculating the probability of an unarmed black man being killed in an encounter with police might be something along the lines of
Pb * (1 – Gb) * Sb
Where Pb is the probability of a black man encountering the police, Gb is the probability that the black man is armed (and does it have to be a gun?), and Sb is the probably that the police will shoot an unarmed black man. Decorate with some solid statistical numbers and do a bit of algebra, and I’d feel more confident.
His note concerning some work done by a Cornell Ph. D. student, based on data from the Police-Public Contact Survey, is more interesting:
It’s a big survey — around 150,000 people, including 16,000 African-Americans. And it provides one answer (although not definitive) to some obvious questions. First off, are black men in America disproportionately likely to have contact with the police? Surprisingly, no. In the survey years that Lamoine looked at, 20.7 percent of white men say they interacted at least once with a cop, compared with 17.5 percent of black men. The data also separates out those with multiple encounters. According to Lemoine, black men (1.5 percent) are indeed more likely than whites (1.2 percent) to have more than three contacts with police per year — but it’s not a huge difference.
On the key measure of use of force by the cops, however, black men with at least one encounter with cops are more than twice as likely to report the use of force as whites (one percent versus 0.4 percent). That’s the nub of it. “Force,” by the way, includes a verbal threat of it, as well as restraining, or subduing. If you restrict it to physical violence, the data is worse: Of men who have had at least one encounter with the police in a given year, 0.9 percent of white men reported the use of violence, compared with 3.4 percent of black men. (For force likely to cause physical injury, i.e. extreme force, however, the ratio is actually better: 0.39 percent for white men compared with 0.46 percent of black men.)
What do we make of this data? I think it shows the following: that police violence against black men, very broadly defined, is twice as common as against white men, and narrowly defined as physical force, three times as common, but that there’s no racial difference in police violence that might lead to physical harm, and all such violence is rare. (Recall that the 3.4 percent of black men who experience violence at the hands of the police are 3.4 percent of the 17.5 percent of those who have at least one encounter with the cops, i.e., 0.5 percent of all black men.)
Of course, there’s always some question about data collection, and here I’d have to ask if this is self-reported data or not. Self-reported data is always subjective, and therefore always a trifle dubious. Still, it’s interesting how it turns out.
I am a bit disappointed that that Andrew limits himself to violent encounters. After all, these incidents are not isolated, but are part of a spectrum and collection of every day behaviors. We’ve often heard of profiling – does this show up in the data? And what sort of effect do unsupportable traffic stops and harassment have on the black community?
