The Washington PostDemocracy Dies in Darkness

How a controversial study found that police are more likely to shoot whites, not blacks

Diamond Reynolds cries outside the governor's residence Thursday in St. Paul, Minn. Reynolds's boyfriend, Philando Castile, was shot and killed after a traffic stop by police last Wednesday. (Jim Mone/AP)

Roland Fryer Jr. never cared much for the cops. When he was growing up, his family dealt crack in Daytona Beach, Fla., and while Fryer was on his way to becoming a celebrated economist at Harvard University, many of his cousins and closest friends were serving mandatory sentences in prison. During his childhood, encounters with police were fraught with danger.

"As a kid, I didn't like the police at all," Fryer said. "I grew up on one side of the story."

Fryer said that after the death of Michael Brown in Ferguson, Mo., and other recent cases in which police killed unarmed black civilians, Fryer felt he had to know more. "My emotions are flaring," he recalled. He spent a year not only gathering data with teams of research assistants, but also riding along with officers and completing several days of law-enforcement training himself.

The economist published his findings this week in a draft paper that is already causing controversy.

On the one hand, the study shows that, nationwide, black and Hispanic civilians are indeed more likely to be manhandled, handcuffed or beaten by the police — even if they are compliant and law-abiding. Fryer writes that prejudice in law enforcement is real and harmful in many ways, causing cynicism and disillusionment especially among boys of color.

When it comes to police killings, though, Fryer has painstakingly reviewed evidence from Houston that suggests the police there are not racially biased in how they use lethal force. This is a surprising finding that challenges a widespread view that the police disproportionately shoot black suspects. In Houston, at least, that might not be the case.

Some economists have criticized the report. They argue that Fryer has not overcome one of the fundamental challenges of studying bias in police shootings: To detect if police fire on suspects in racially prejudiced ways, researchers must examine comparable situations involving white and black suspects. Compared with previous research, Fryer's study collects much more detail on individual shootings, but some say it is not enough to ensure his comparisons are fair.

It's unclear what this research implies about policing in other places. What is true in Houston — a large, multicultural city with a diverse police force — might not be true elsewhere. Despite its narrow purview, the study still offers one of the first comprehensive and rigorous examinations of some of the most detailed data that exists on police shootings.

'It speaks to dignity'

In his career so far, Fryer has achieved notoriety that is unusual for an economist by coming up with clever ways to answer uncomfortable questions — for example, collecting data that shows that black children with a 4.0 grade-point average have fewer black friends than those with a 3.5, which wasn't true for white children with white friends. Last year, Fryer, 39, became the first African American to win the prestigious John Bates Clark Medal, which is awarded annually to the top U.S. economist under the age of 40.

Meet the economist who grew up among crack dealers and won a ‘mini-Nobel’ for his research on race

In his latest paper, Fryer partly focuses on violent encounters in which officers didn't fire a gun. He uses two sources of data that show police are more likely to use higher levels of force on minorities.

First, using records collected by the New York Police Department from its stop-and-frisk program, he finds that officers were about 53 percent more likely to use force on black and Hispanic civilians than against white civilians. The police were more likely to push and shove minorities, handcuff them, use batons and pepper spray them.

How civilians behaved could not completely account for the large disparity. Where officers found contraband while frisking someone, they were 11 percent more likely to use force if the civilian was black than if they were white. Even in cases where the officer described the civilian as perfectly compliant with questioning and instructions, officers were 21 percent more likely to use force against black civilians than against white civilians.

Second, Fryer looked at a federal survey of civilians nationwide about their contacts with law enforcement. That data suggested that black and Hispanic people were about three times more likely than white people to be treated with force by an officer.

These findings on the use of nonfatal force are consistent with a white paper also published this month by researchers at the University of California at Los Angeles. The authors likewise found that officers were more likely to use a Taser on or manhandle black civilians, even accounting for the difference in the rates at which black suspects are arrested in connection with violent crimes.

Such incidents "happen thousands of times a day. They’re so much more frequent than the shootings," Fryer said. "It’s important. I think it speaks to dignity."

Startling data on police shootings

The data from New York and the federal survey did not contain enough information about killings by police for Fryer to draw conclusion about the lethal use of force. He collected records on police shootings in 10 major police departments between 2000 and 2015. The data showed that officers reported shooting before being attacked more frequently in encounters with white suspects, while officers said that they were attacked first more frequently in encounters with black suspects.

What to make of this pattern was not clear. Without more detail about the individual cases, Fryer could not determine whether there was racial bias in their decision to shoot.

So the economist took a closer look at incidents in Houston, the city that provided the greatest amount of detail. In addition to shootings, he considered cases in which an officer reported that the suspect reacted violently or resisted arrest, but the officer did not shoot. In these potentially violent encounters with suspects, there was no disparity in how likely officers from the Houston Police Department were to shoot black suspects and white suspects.

Here's how Fryer came to his conclusions. He began with the data on all 507 officer-involved shootings that happened in Houston the last 15 years. About 52 percent of those shootings involved a black suspect, while 14 percent involved a non-black, non-Hispanic suspect.

Since Houston is 24 percent black, the fact that more than half of the police shootings involved black people might seem like a sign of racially biased policing. Yet it is also possible that Houston police more frequently encounter black residents in dangerous situations.

To account for this possibility, Fryer analyzed encounters with police in which the suspect was arrested on a charge that indicates a potentially dangerous situation. He analyzed arrests in which officers accused the suspect of one of the following charges:

  • Aggravated assault on a peace officer
  • Attempted capital murder of a peace officer
  • Resisting arrest
  • Evading arrest
  • Interfering in an arrest

About 58 percent of such arrests involved black suspects, while about 12 percent of these arrests involved non-black, non-Hispanic suspects. Those statistics are pretty similar to the statistics from the officer-involved shootings. So, while it’s true that more than half of police shootings involved a black suspect — from the perspective of the Houston police, more than half of the dangerous situations they encountered also involved black suspects.

Aren’t more white people than black people killed by police? Yes, but no.

This approach is not perfect, of course, and Fryer readily points out that his research has limitations.

Police officers have a lot of leeway in deciding whether to book people on certain charges. For instance, in some cities, nearly any minor pretext is enough for an officer to arrest a suspect on suspicion of assaulting a police officer. In New York, police are twice as likely to accuse suspects in misdemeanor drug cases of resisting arrest if the suspects are black.

Fryer's analysis assumes that the people charged with those five crimes posed similar levels of danger to police. But if officers are more likely to accuse peaceful black civilians of resisting or evading arrest or attacking a cop, then encounters involving black civilians would be less dangerous than the arrest charge makes them appear. An ideal study would look at how police officers respond to identical situations involving black and white suspects, but of course, that's impossible.

Fryer attempted to correct for this problem by reading the officers’ statements about each arrest, which gave him more information about the danger level of each arrest. This required thousands of hours for his team to pore over police records, but it allowed him to control for hundreds of factors, including whether the suspect attacked, how many officers were at the scene, whether the suspect brandished a weapon and so on. He also noted if the officer was responding to a report of a violent crime. Even after taking all these factors into account, Fryer did not find any major differences in how officers treated black and white suspects. 

This approach has shortcomings too; it relies on police write-ups, which only tell one side of the story and might not be completely accurate. There are many well-documented cases of perjury by law enforcement, and studies in the laboratory show that police officers view black people as being more criminal and more dangerous than white people.

"At the end of the day, biases . . . will still be possible," Fryer said. "We can't randomly assign race, and we don't have the perfect social science experiment here."

In the coming months, these and other parts of the paper will be debated among academics as the study undergoes peer review.

A Washington Post database of fatal police shootings nationwide has yielded a widely cited statistic: Black people are 2.5 times more likely to be shot dead by police. Looking specifically at the 93 fatal shootings of unarmed civilians recorded by The Post, one scholarly analysis concluded those civilians were significantly more likely to be black than white, even after taking into account the level of violent crime in the neighborhood of the killing.

Fryer's study goes a step further, looking at a broad range of details of individual cases in attempt to determine the level of danger the officer might have confronted. Fryer, however, still does not have the answers he wants, he said.

"To make progress, we have to elevate the discussion here," Fryer said, adding that he welcomes critiques of his study.

"If folks want to ask hard questions about . . . whether Houston is representative, and what if we used Chicago — God, man!" Fryer said. "That's where we need to be in these discussions."

More from Wonkblog:

More police officers die on the job in states with more guns

Why white people see black boys as older, bigger and guiltier than they are

Police are safer under Obama than they have been in decades