This week we’re all thinking a lot about the role of police in minority communities, and specifically the question of whether suspects are more likely to be killed if they, like George Floyd, are black. I agree with pretty much everyone, by the way, that Floyd was murdered and justice should be done.

I’ve been following the research on this for years, and I wanted to provide a brief overview of some of the work that has been published. My current sense is that the vast majority of police killings are justified, but that at least some communities have a serious problem, both in terms of race and in terms of cops’ being too eager to shoot in general.

Here are the statistics, facts, and studies I find most important:

You Might Like

• In most years, about 25 percent of people killed by American police are black. This is much higher than blacks’ share of the overall population (13 percent), though it’s lower than blacks’ share of some other relevant comparison groups, including murderers and cop killers. (Cops don’t shoot people at random — if they’re following the law, they fire only when they reasonably believe a violent criminal poses an immediate threat to life or limb — so racial differences in crime rates are unfortunately important to consider here.) One study tried “benchmarking” blacks’ share of those killed by police against numerous measures of crime, and found no anti-black disparity in nearly all of these comparisons.

• It is hard to tell which specific shootings are unjustified, save for the handful each year that result in thorough investigations and convictions, so it’s hard to say what the racial balance is for those. No, one can’t simply focus on “unarmed” suspects, because an unarmed person can still pose a grave threat to a police officer. And yes, unjustified shootings happen to white people sometimes too.

• A few years ago, Roland Fryer published a landmark study that analyzed policing incidents in great detail, finding that, after accounting for how suspects behaved, there was no racial difference in lethal force (though blacks and Hispanics were more likely to suffer non-lethal force). However, much of this analysis relied on data from a single city, Houston, and the suspects’ behavior was taken from reports the police themselves provided.

• Several studies have checked to see if black cops, who presumably are less likely to hold anti-black racist beliefs, are less likely to be involved when black suspects are killed. The difficulty of this method is that different parts of the country have different black population shares, and places with bigger black populations will have more black suspects and more black cops, causing a meaningless correlation between the two that falsely implies black cops are more likely to shoot black suspects. After accounting for the demographics of the area where each killing occurs, these studies find no difference between black and white cops. Importantly, however, black cops might be assigned to black neighborhoods even within jurisdictions, which would throw this method off.

• A study earlier this year used a good approach to get around this issue, looking at an unnamed city where the nearest cop is assigned to each call, with no discretion involved. It found that, relative to a black cop working the same beat and shift, a white cop was a whopping five times as likely to discharge his gun when sent to a neighborhood that was 80 percent black or more. As I noted at the time, though, there are limitations to this study as well: It’s just one city (though there’s a separate analysis of force against Hispanics in another), and there are possible explanations for the findings besides racist wrongful killings, including white cops’ being less able than black cops to deescalate tense situations in black neighborhoods.

• There’s forthcoming research in the pipeline suggesting that when police unionize, killings escalate, with the killings concentrated among minority suspects. This work is still tentative, but you can read a Twitter thread from one of the authors here.