Official data dramatically undercount the number of people killed by the police

See When Police Kill by Alex Tabarrok.

"When Police Kill is the 2017 book by criminologist Franklin Zimring. Some insights from the book.

Official data dramatically undercount the number of people killed by the police. Both the Bureau of Justice Statistics’ Arrest-Related Deaths and the FBI’s Supplemental Homicide Reports estimated around 400-500 police kills a year, circa 2010. But the two series have shockingly low overlap–homicides counted in one series are not counted in the other and vice-versa. A statistical estimate based on the lack of overlap suggests a true rate of around 1000 police killings per year.



The best data come from newspaper reports which also show around 1000-1300 police killings a year (Zimring focuses his analysis on The Guardian’s database.) Fixing the data problem should be a high priority. But the FBI cannot be trusted to do the job:

Unfortunately, the FBI’s legacy of passive acceptance of incomplete statistical data on police killings, its promotion of the self-interested factual accounts from departments, and its failure to collect significant details about the nature of the provocation and the nature of the force used by police suggest that nothing short of massive change in its orientation, in its legal authority to collect data and its attitude toward auditing and research would make the FBI an agency worthy of public trust and statistical reliability in regard to the subject of this book.

The FBI’s bias is even seen in its nomenclature for police killings–“justifiable homicides”–which some of them certainly are not.

The state kills people in two ways, executions and police killings. Executions require trials, appeals, long waiting periods and great deliberation and expense. Police killings are not extensively monitored, analyzed or deliberated upon and, until very recently, even much discussed. Yet every year, police kill 25 to 50 times as many people as are executed. Why have police killings been ignored?

When an execution takes place in Texas, everybody knows that Texas is conducting the killing and is accountable for its consequences. When Officer Smith kills Citizen Jones on a city street in Dallas, it is Officer Smith rather than any larger governmental organization…[who] becomes the primary repository of credit or blame.

We used to do the same thing with airplane crashes and medical mistakes–that is, look for pilot or physician error. Safety didn’t improve much until we started to apply systems thinking. We need a systems-thinking approach to police shootings.

Police kill males (95%) far more than females, a much larger ratio than for felonies. Police kill more whites than blacks which is often forgotten, although not surprising because whites are a larger share of the population. Based on the Guardian data shown in Zimring’s Figure 3.1, whites and Hispanics are killed approximately in proportion to population. Blacks are killed at about twice their proportion to population. Asians are killed less than in proportion to their population.

A surprising finding:

Crime is a young man’s game in the United States but being killed by a police officer is not.

The main reason for this appears to be that a disproportionate share of police killings come from disturbance calls, domestic and non-domestic about equally represented. A majority of the killings arising from disturbance calls are of people aged forty or more.

The tendency  of both police and observers to assume that attacks against police and police use of force is closely associated with violent crime and criminal justice should be modified in significant ways to accord for the disturbance, domestic conflicts, and emotional disruptions that frequently become the caseload of police officers.

A slight majority (56%) of the people who are killed by the police are armed with a gun and another 3.7% seemed to have a gun. Police have reason to fear guns, 92% of killings of police are by guns. But 40% of the people killed by police don’t have guns and other weapons are much less dangerous to police. In many years, hundreds of people brandishing knives are killed by the police while no police are killed by people brandishing knives. The police seem to be too quick to use deadly force against people significantly less well-armed than the police. (Yes, Lucas critique. See below on policing in a democratic society).

Police kill more people than people kill police–a ratio of about 15 to 1–and the ratio has been increasing over time. Policing has become safer over the past 40 years with a 75% drop in police killed on the job since 1976–the fall is greater than for crime more generally and is probably due to Kevlar vests. Kevlar vests are an interesting technology because they make police safer without imposing more risk on citizens. We need more win-win technologies. Although policing has become safer over time, the number of police killings has not decreased in proportion which is why the “kill ratio” has increased.

A major factor in the number of deaths caused by police shootings is the number of wounds received by the victim. In Chicago, 20% of victims with one wound died, 34% with two wounds and 74% with five or more wounds. Obvious. But it suggests a reevaluation of the police training to empty their magazine. Zimring suggests that if the first shot fired was due to reasonable fear the tenth might not be. A single, aggregational analysis:

…simplifies the task of police investigator or district attorney, but it creates no disincentive to police use of additional deadly force that may not be necessary by the time it happens–whether with the third shot or the seventh or the tenth.

It would be hard to implement this ex-post but I agree that emptying the magazine isn’t always reasonable, especially when the police are not under fire. Is it more dangerous to fire one or two shots and reevaluate than to fire ten? Of course, but given the number of errors police make this is not an unreasonable risk to ask police to take in a democratic society.

The successful prosecution of even a small number of extremely excessive force police killings would reduce the predominant perception among both citizens and rank-and-file police officers that police have what amounts to immunity from criminal liability for killing citizens in the line of duty.

Prosecutors, however, rely on the police to do their job and in the long-run won’t bite the hand that feeds them. Clear and cautious rules of engagement that establish bright lines would be more helpful. One problem is that police are protected because police brutality is common (somewhat similar to my analysis of riots).

The more killings a city experiences, the less likely it will be that a particular cop and a specific killings can lead to a charge and a conviction. In the worst of such settings, wrongful killings are not deviant officer behavior.
…clear and cautious rules of engagement will …make officers who ignore or misapply departmental standards look more blameworthy to police, to prosecutors, and to juries in the criminal process.

Police kill many more people in the United States than in other developed countries, even adjusting for crime rates (where the U.S. is less of an outlier than most people imagine). The obvious reason is that there are a lot of guns in the United States. As a result, the United States is not going to get its police killing rate down to Germany’s which is at least 40 times lower. Nevertheless:

[Police killings]…are a serious problem we can fix. Clear administrative restrictions on when police can shoot can eliminate 50 to 80 percent of killings by police without causing substantial risk to the lives of police officers or major changes in how police do their jobs. A thousand killings a year are not the unavoidable result of community conditions or of the nature of policing in the United States."