A Distinctly American Problem Needs Systematic Investigation

Every police killing, like every plane crash, warrants careful review by a federal agency.

An illustration of a police officer with a spotlight on them.
Getty / The Atlantic

Aviation deaths once looked like an intractable problem. Then the federal government began probing every plane crash with an eye toward preventing future loss of life. Our skies got much safer as a result. A similar approach could reduce police killings. A federal agency should investigate every single killing and significant injury caused by American police officers, who have long killed people at higher rates than cops in many other wealthy democracies.

Police killings and protests against them have loomed large in United States politics for at least the past seven years. Right now the nation is focused most closely on the trial of Derek Chauvin, who infamously knelt on George Floyd’s neck, even as new protests erupt in Minneapolis over the killing of Daunte Wright, who was shot to death by a police officer who says she intended to discharge her taser. On Thursday, the city of Chicago released footage of the fatal police shooting of 13-year-old Adam Toledo.

The number of police killings of unarmed people appears to have dropped since The Washington Post began keeping track in 2015. Officers killed 95 that year, then 54 in 2019 and 55 in 2020, according to the newspaper’s database. The drop might represent progress; it might also be a fluke. Regardless, total fatal shootings by on-duty police show no such decline. From 2015 to 2019, the newspaper recorded just under 1,000 such incidents a year nationwide; last year’s total was 1,021.

Officialdom’s primary response to police shootings and other uses of deadly force is currently backward-looking and legalistic. Local authorities review a killing to determine whether laws and department policies were followed. The most egregious police killings renew protests that succeed in generating attention, statements of concern from corporations, and gestures of solidarity from progressives, but not in reducing police killings. That cycle fuels anger, fear, polarization, and civic dysfunction, including occasional riots, with cultural effects that most Republicans, Democrats, conservatives, progressives, and libertarians dislike. Most everyone would benefit from more constructive responses all around.

A faction within the Black Lives Matter movement is proposing potentially valuable policy changes at the local level. Well-researched initiatives such as Campaign Zero and 8 Can’t Wait would encourage community oversight of cops, more restrictive policies toward the use of force, and other specific reforms that appear likely to save lives. But among fellow activists, they are losing mindshare to another faction that advocates for defunding the police, a nonstarter among the broader public. “The only thing most people can seem to agree on—even at the height of the protests after Floyd’s death—is that they’re against the idea of defunding the police,” FiveThirtyEight reported recently in an assessment of survey data. “And this remains true today, even among Black Americans and Democrats.”

The most constructive way that the federal government responds to avoidable loss of life is arguably in its treatment of aviation. Whenever a plane crash occurs, big or small, headline-grabbing or obscure, a team of experts is dispatched to reconstruct exactly what happened. The aim isn’t to advance a legal process or punish wrongdoers, but to figure out which changes, if any, could prevent it from happening again.

“Aviation is safe in large part because it learns from its disasters,” my colleague James Fallows, himself a recreational pilot, has argued. The NTSB’s painstaking collection and evaluation of evidence after each accident can take months or even years, but the investigations yield insights that save lives. “From the dawn of commercial aviation through the 1990s,” Fallows writes, “1,000 to 2,000 people would typically die each year in airline crashes. Today, the worldwide total is usually about one-tenth that level.”

What if every police killing triggered that sort of response?

Focusing “on only the immediate causer” of a police killing “and the narrow time frame that defines [the officer’s] actions” is inadequate, the University of Virginia law professor Barbara E. Armacost argued in the Ohio State Law Journal, because “the killing of unarmed civilians by police results from multiple causes, both human and systemic, that set the stage for the tragic moment when the shot was fired.” A broader, NTSB-style approach would not ignore any factors.

Paul Heaton, an economist who analyzes the criminal-justice system, argued in a 2017 essay that an NTSB for police killings would offer many advantages over the status quo. Perhaps the most significant is the ability a federal agency would have to learn from incidents all over America. “Often system-level factors that contribute to unwanted outcomes are only apparent after aggregating across multiple incidents, each of which appear unique and idiosyncratic when viewed in isolation,” he wrote, noting that “each locality has a limited set of incidents from which to draw useful lessons,” while “aggregation of information can likely produce better insights.”

Another advantage, he argued, is that a federal agency’s investigative protocols could be consistent from case to case—unlike a local district attorney who, in deciding whether to seek the conviction of a police officer “is making procedural decisions in real time while investigating an incident that differs in important ways from more routine investigative business.”

Finally, an NTSB-like approach would be separate from the criminal process. As Heaton wrote:

The current investigative paradigm focuses almost exclusively on whether use of police force was legally justified, which is an adversarial approach that discourages open sharing of information. Moreover, it fails to recognize that many incidents that could be perceived as legally defensible—such as situations in which officers mistakenly believe themselves to be in imminent physical danger—are undesirable from a societal perspective, and in some cases can be prevented.

An NTSB-style investigation wouldn’t preclude criminal charges for police, any more than it prevents the criminal prosecution of a pilot who drank on the job before a plane crash. But adding a neutral, fact-finding agency separate from the legal process would produce transparent recommendations even when an officer was acquitted or a DA declined to file charges.

During her presidential campaign, Vice President Kamala Harris aired an idea much like the one Heaton outlined. She vowed to create “a National Police Systems Review Board, which would collect data and review police shootings and other cases of severe misconduct, and work to issue recommendations and implement safety standards based on evidence revealed in these reviews.”

Some skeptics doubt that a proposal of this sort is politically viable. Others question whether it would overextend federal authority. Laurie Robinson, a criminologist at George Mason University and co-chair of the Obama administration’s Task Force on 21st Century Policing, told the Marshall Project that “it would be unusual, as opposed to NTSB's power over a regulated interstate business, to provide this kind of power over a very local function.” To work, the proposed federal agency would require power to compel cooperation in its fact-finding. (Reflecting a focus on prevention only, parts of an NTSB investigation are inadmissible in judicial proceedings.) But even if states and localities retained all other control over policing policy, an NTSB-style agency could do good merely by gathering data; improving transparency; distilling insights and best practices; and keeping police brass, lawmakers, and the public informed. Plus, everyone who cares about decreasing the number of people cops kill every year would have something to focus on beyond a process of criminal accountability ill-suited to bringing about progress.

Early on, perhaps a mere fraction of jurisdictions would follow federal investigators’ advice. But if those jurisdictions saw sustained declines in police killings, their example would be powerful.

Current investigations of police killings are neither independent nor broad enough in scope to determine whether many shootings could have been avoided, nor are they oriented around using findings in individual cases to identify patterns that would save lives nationwide. We must either change our approach or continue to allow preventable police killings, costing lives and undermining faith in the criminal-justice system. An NTSB for police killings could solve those problems.

Let’s try it.