Sunday, June 14, 2015

Police Killings and Racial Bias

The Guardian newspaper, which has been keeping count, announced Wednesday that fatal shootings by police in the U.S. in 2015 had reached 500, and that's only through the end of May. That seems like a shockingly large number.

The article points out that a disproportionate number of African-Americans were killed, using statistics that I've put into this table:

Police killings US population
White 49.6% 64.2%
Black 28.2% 13.2%
Hisp. 14.8% 17.1%

Yes, clearly, a disproportionate number of African-Americans are killed. But is that evidence of discrimination by police? After all, there are more arrests of African-Americans, and if there are more arrests, then surely we would expect more gunplay:

Police killings  Arrests
White 49.6% 68.9%
Black 28.2% 28.3%
Hisp. 14.8% 16.6%

The numbers in the right column add to more than 100, because "Hispanic" is not considered a separate racial category in government statistics. But if we assume that 90-95% of Hispanics are white, and reduce the totals for "white" and "black" in the second column accordingly, then the number of shootings seems roughly in line with the number of arrests.

But that just pushes the question back a step. Why are more African-Americans arrested? Is it because they commit more crimes, or is the statistic itself a sign of police bias? There were complaints in New York City, for example, that blacks were arrested more than whites for marijuana possession, not because they were more likely to smoke marijuana, but because of discriminatory use of "stop-and-frisk" searches.

This, however, doesn't seem to be true for arrests in general. African-Americans make up 30% of all drug arrests, and 26% of all vandalism arrests, but 52% of all murder arrests. It seems unlikely that discriminatory enforcement is more common in murder cases than in drug cases.

So far, the facts don't seem to bear out a claim of police bias. But wait... here's an odd thing.

More than one in every five people killed so far in 2015 – 108, or 21.6% – were unarmed. A significant disparity in the proportion of black and white people killed who were unarmed, which was reported last week by the Guardian, has since narrowed slightly. While 30.5% of white people killed were unarmed, 16.1% of black people killed had no weapon.

Apparently, whites were a lot more likely than blacks to be unarmed when they were killed. Does that mean whites were more likely to be killed when they were unarmed? That would be surprising.

I decided to take look at The Guardian's data set. The data set is actually quite admirable. In addition to demographic information, it has a brief narrative and often a photo for each case. And it will provide counts using various filters, including race of victim and how they were  armed.

The first thing I discovered is that they got the numbers reversed: it's actually 30% of blacks and 16% of whites who were unarmed when they were killed. [Update: The webpage linked above now notes and corrects this error.] Playing just a little more with the data, I find that The Guardian has kind of missed the big story here. Rather than presenting the data as in the first table above, they should have presented this:

Police killings:
Armed people Unarmed people
White 53.5% 36.7%
Black 26.9% 40.4%
Hisp. 15.3% 15.6%

Let's recap. Blacks are disproportionately likely to be arrested. Probably that's mostly because they commit more crimes. While racism undoubtedly is part of the ultimate explanation for that, racism on the part of the police doesn't seem to explain why blacks are arrested more. And if you're armed, your chances of being killed appear about the same for blacks and whites.

But if you're unarmed, the chance of being killed is much greater for blacks than for whites. That does start to look like racism. It needn't mean that the police hate blacks; it could mean they're more likely to perceive blacks as threatening. But of course, that's not much consolation to the familes.

"Hands up, don't shoot" indeed.


  1. Excellent probing of the Guardian data.
    Staggering, and revelatory, that it takes a UK newspaper to gather the data about American gun violence. I think this reveals the real origin of the problem of gun violence: for whatever reason, Americans are unconcerned about gun violence, to the point that even the Brits are more concerned about it that we are. Indeed a part of us actually likes gun violence.

  2. Don't you have "unarmed" and "armed" switched in your last table?

  3. Don't you have "unarmed" and "armed" switched in your last table?