Deaths involving police have been greatly undercounted in the United States, and African American people die in such encounters at 3.5 times the rate of whites, according to a new analysis by public health researchers.
In an article published Thursday in the medical journal The Lancet, researchers found that deaths from police violence between 1980 and 2018 were misclassified by 55.5% in the U.S. National Vital Statistics System, which tracks information from death certificates.
"For most causes of death, the death certificate filled out by a physician is sort of the gold standard," says Chris Murray of the Institute for Health Metrics and Evaluation at the University of Washington, who is one of the study's authors. But he says that in this area, the certificates seem to fall short. "There is a pretty systematic underrecording of police violence deaths. "
That realization isn't entirely new. After the 2014 shooting of Michael Brown in Ferguson, Mo., news organizations started to keep their own tallies of police-related deaths, which turned out to be higher than the government's numbers.
What Murray and his co-authors have done, though, is measure the discrepancy between independent tallies and the government data, and project it back in time.
"We've used those relationships of what fraction get underreported to go back and infer, for example, in the 1980s, what was the likely number of police violence deaths," Murray says.
The researchers based their inferences on numbers from three open-source databases: Fatal Encounters, Mapping Police Violence and The Guardian's The Counted, which they compared with the data from the death certificates.
They calculate that the death certificates misclassified the cause of death on more than 17,000 such deaths since 1980.
"If it's legit, it's pretty cool how they can take existing data from a short time frame and work backwards," says Justin Nix, associate professor of criminology at the University of Nebraska.
But as a criminologist who studies shootings by police, Nix has reservations about the underlying data.
"My concerns with this paper are the same as many that use these crowd-sourced databases," he says. He has documented cases where the databases count, for example, domestic violence by off-duty officers as police killings.
"I'm not saying we don't need to track that in these sorts of databases, but I'm just saying that all police killings are not created equally," he says.
"I think there's definitely issues around exactly the criteria used," says the IHME's Murray. "I think that's an important question, given that we're looking at multiple sources. [But] I don't think it's really influencing the time-trend we're seeing. In other words, the numbers are going up, regardless."
The study shows the death rate in these encounters dropping in the 1980s, then generally rising again since about 2000.
The article also highlights the disparity in the mortality rate for African-Americans, which it says is 3.5 times higher than that of whites.
The article suggests the disparity is caused by "systemic racism in policing," but it doesn't specify how that happens. Specifically, it doesn't address whether police are more likely to use lethal force against African-Americans or whether nonpolicing factors lead African-Americans to have more encounters with police.
Murray says this analysis doesn't answer that.
"I don't think from a scientific point of view, we have enough information here to parse out how much of this is, you know, basic differences in where people live, what sort of disadvantage they have, versus the actual specific actions of the police," he says.
But as a public health expert, Murray says the more we know about these deaths, the easier it will be to find policy solutions.
"It's the old saw: You manage what you measure. And so we've got to do a better job of tracking in what's actually happening," he says.