Written by a black economist/stats/data prof at Harvard. Read the bolded part.
"I have led two starkly different lives—that of a Southern black boy who grew up without a mother and knows what it’s like to swallow the bitter pill of police brutality, and that of an economics nerd who believes in the power of data to inform effective policy.
I set out on a mission to quantify racial differences in police use of force.
There are large racial differences in police use of nonlethal force. My research team analyzed nearly five million police encounters from New York City. We found that when police reported the incidents, they were 53% more likely to use physical force on a black civilian than a white one. In a separate, nationally representative dataset asking civilians about their experiences with police, we found the use of physical force on blacks to be 350% as likely. This is true of every level of nonlethal force, from officers putting their hands on civilians to striking them with batons. We controlled for every variable available in myriad ways. That reduced the racial disparities by 66%, but blacks were still significantly more likely to endure police force.
• We didn’t find racial differences in officer-involved shootings. Our data come from localities in California, Colorado, Florida, Texas and Washington state and contain accounts of 1,399 police shootings at civilians between 2000 and 2015. In addition, from Houston only in those same years, we had reports describing situations in which gunfire might have been justified by department guidelines but the cops didn’t shoot. This is a key piece of data that popular online databases don’t include.
No matter how we analyzed the data, we found no racial differences in shootings overall, in any city in particular, or in any subset of the data. I have grappled with these results for years as I witnessed videos of unmistakable police brutality against black men. How can the data tell a story so different from what we see with our eyes?
Our analysis tells us what happens on average. It isn’t average when a police officer casually kneels on someone’s neck for 8 minutes and 46 seconds. Are there racial differences in the most extreme forms of police violence? The Southern boy in me says yes; the economist says we don’t know.
Several scholars have rightly pointed out that these data all begin with an interaction, and suggested that racist policing manifests itself in more interactions between blacks and the police. The impact of this hypothesis in our shootings data seems minimal. The results on police shootings are statistically the same across all call types—ranging from officer-initiated contact with a suspicious person (where racism in whom to police is likely paramount) to a 911 call of a homicide in progress (where interaction with the potential suspect is more likely independent of race).
Are the data nationally representative? We don’t know. But at least two other studies, both published in 2016—by Phillip Atiba Goff et al. and Ted R. Miller et al.—have since found the same using different data. Moreover, when we use our data to calculate the descriptive statistics used in popular databases such as the Washington Post’s, we find a higher percentage of black civilians among unarmed men killed by the police than they do. Those statistics, however, cannot address the fundamental question: When a shooting might be justified by department standards, are police more likely actually to shoot if the civilian is black? Only our data can answer this question, because it contains information on situations in which a shooting might meet departmental standards but didn’t happen. The answer appears to be no."
https://www.wsj.com/articles/what-the-data-say-about-police-11592845959?mod=searchresults&page=1&pos=7