As Congress and local governments across the United States consider police reform, one important question is why police use excessive force, even after scores of protests across the country against police brutality following the death of George Floyd and other people of color.

A common answer is that officers are afraid of being injured or killed because their jobs are dangerous, but data show that being a police officer is relatively safe. Law enforcement doesn’t crack the top-15 list of most dangerous jobs in America, according to the Bureau of Labor Statistics.

The fear police officers have of civilians, especially people of color, is a risk exaggerated and amplified through training and storytelling within the law enforcement community, according to crime and policing research by Brittany Arsiniega, an assistant professor in the Department of Politics and International Affairs at Furman University.

“A lot of research shows that what people think is dangerous and what they believe poses the greatest risk of harm is determined by your membership in various social groups,” Arsiniega says. This is a phenomenon called “social amplification.” Police may believe they are at a great risk of civilian assault, even when empirical data say otherwise.

Research also shows that law enforcement officers, along with society generally, believe that people with black and brown skin pose an even greater threat.

“Police live in the same society as the rest of us, where people still equate blackness with criminality,” Arsiniega says. And, she says, the way policing is carried out is systemically racist. She points to traffic stops that disproportionately affect black Americans three to one.

“Even if officers didn’t go into the force being racist, even if they have the best intentions, in order to be seen as doing the job well by their peers and superiors they are often engaging in racist practices.”

Mixing systemic racism and socially amplified fear of assault becomes deadly because police are armed, and they have “legal justification to kill people and be legally excused if they’re protecting their lives or lives of a third party,” Arsiniega says.

However, in more than 90 percent of cases in which an officer kills a civilian – about 1,000 each year in America – officers claim to be protecting themselves, not a third party.

Neither do the majority of injuries to on-duty officers come from civilians.

For her dissertation at the University of California at Berkeley, Arsiniega studied police officer injury data from all causes in Savannah, Georgia, between 2010 and 2018. Only 11 percent of all officer injuries were the result of civilian assaults.

She also studied assaults on Atlanta Police Department officers from 2015 to 2018 using reports written by officers themselves. Despite having a violent crime rate roughly 2.5-times the national average, not a single officer was shot or stabbed in the years of her sample. The most serious injuries caused by civilians were broken bones in the hand and cuts and scrapes on officers’ hands and knees while arresting suspects.

Meanwhile, the use of force by the police during encounters with civilians was often disproportionate to the offense. In one report, a man believed to be homeless threw a plate of food at an officer, which constituted assault. The officer used pepper spray, but when the man continued toward the officer she shot him.

The social amplification of fear begins in training, Arsiniega says, when police recruits are shown horrific videos of officers being murdered and are made to listen to tapes of officers gasping for their last breaths. “Cops are drilled with the lesson that you never know when it’s going to happen, that it could be you, so you always have to be on edge,” she says.

Police departments also nurture an attitude of “you’re with us or against us,” Arsiniega says. She points out that Derek Chauvin, the officer who knelt on George Floyd’s neck until Floyd died, had at least 17 misconduct complaints against him. “The problem is that you get a combination of one person who is prone to violence and a culture that insulates them from any accountability.”

There is one area, however, in which police are legitimately at great risk: mental health disorders, including depression and PTSD, and suicide, Arsiniega says. While civilian assault does not empirically pose great risk to officers, officers are routinely exposed to other traumatic events, like gruesome traffic accidents, child abuse, domestic violence and, Arsiniega says, “stuff that haunts your dreams.”

However, police departments rarely track the data that could lead to healthier reforms for officers. For her research, the departments didn’t even maintain or track data on officer injuries; they had to request it from their workers’ compensation insurance carriers.

“Police data are designed and very effective at creating knowledge about the physical dangers of policing, especially from civilian assault, but are currently incapable of creating reliable knowledge about suicide and mental illness,” Arsiniega says.

“There’s a lot of space to have compassion for the police, but not from the things we so often hear them lamenting about.”