Editor's note: The authors have retracted this study. See their full statement here.

While fatal shootings of black Americans by white police officers have garnered national headlines and punctuated political debates, new research from the University of Maryland and Michigan State University suggests that officers' race doesn't play a role in how likely they are to shoot nonwhite community members.

Instead, the strongest predictive factor of the race of people fatally shot by police officers was violent crime rates in counties where the shootings took place. In places where whites committed more violent crime, a person fatally shot by police was more likely to be white; the same relationship held true for blacks and Hispanics as well.

The findings, featured Monday in The Proceedings of the National Academy of Sciences, don’t suggest law enforcement is free of racial bias, and can’t be used to judge the propriety of individual shootings, said David Johnson, a postdoctoral researcher in the Lab for Applied Social Science Research and lead author of the paper.

“It’s not our goal to argue that there are no racial disparities in all policing outcomes,” he said. “However, our data do not support the idea that, at the national level, white officers are more responsible for fatal shootings of minority civilians.”

The research team created the first comprehensive database of fatal officer-involved shootings—including officer race, a factor commonly left out. Because federal databases are incomplete, they used more comprehensive lists compiled by The Washington Post and The Guardian as a starting point.

The research team spent roughly 1,500 hours contacting police departments and scouring department websites, case reports, legal documents and news stories to gather information on officers involved in more than 900 fatal shootings nationwide in 2015, the first year in which these news organizations collected data.

Another finding: A higher proportion of nonwhite police officers in an fatal shooting incident meant the shooting victim was more likely to be nonwhite as well. The researchers found this relationship was due to county demographics—cities with larger populations of nonwhite civilians also have a higher proportion of nonwhite officers—rather than biases on the part of nonwhite officers.

Thus, while programs to increase diversity in police hiring can increase public trust in law enforcement, the researchers said, they are unlikely to affect racial disparities in fatal shootings.

The higher per capita number of nonwhite individuals fatally shot by police points to broader issues that contribute to racial disparities in crime rates, the researchers concluded.

“If we want to reduce the rates at which people from minority racial groups are shot by police, we need to address differences in crime rates between races,” Johnson said. “That involves considering what causes those differences, like racial disparities in wealth, unemployment and education.

“I’m not saying that’s easy. We asked a difficult question, and the answer ended up being difficult.”

Johnson and colleagues hope to expand their research to non-fatal shootings and other forms of police force. However, they say before definitive research can take place, the federal government needs more comprehensive data on police use of force. The FBI took a step in that direction this year, launching the National Use of Force Data Collection, designed to gather more detailed information about use of force incidents.

While it’s a start, Johnson said, data will only be collected from agencies that voluntarily report it.

“We have better information on how many people die of shark bites every year than we do people who are fatally shot by police,” Johnson said.