Skip to Main Content

The FBI has more accurate data about bicycle thefts than it does about the number of Americans shot by law enforcement officers. Criminologist David Klinger once called the FBI’s estimates “wildly inaccurate to astronomically inaccurate.” The Department of Justice doesn’t have the numbers. Nor do the National Association of Police Organizations or other law enforcement organizations.

Earlier this year, The Guardian, a British newspaper, launched a website to determine the number of Americans killed by police. Called The Counted, it uses reporting by The Guardian staff and crowdsourcing to identify officer-involved deaths. Today, the number for 2015 stands at 1,055. Similar efforts have popped up, such as Mapping Police Violence and an ongoing investigation by The Washington Post.


Researchers at the Harvard T.H. Chan School of Public Health offer a different approach. Writing Tuesday in the online journal PLOS Medicine, they propose that law-enforcement-related deaths be treated as a “notifiable condition.” That would require hospitals, medical examiners, and other public health entities to quickly report such deaths, just as they report cases of meningitis, measles, or the plague.

We asked several experts if this approach makes sense.

(Note: The FBI declined to contribute to this First Opinion. A few hours after it was published, the Bureau said it would track incidents in which law enforcement officers injure or kill civilians.)


Laurie Robinson: States should take the lead
Brian Root: Accountability is an issue
David A. Klinger: “Notifiable condition” will miss many shootings
Philip Stinson: Multiple databases needed to triangulate
Robert J. Kaminski: Focus on deaths a shortcoming

By Laurie Robinson: Tracking law-enforcement-related deaths as a public health issue is an intriguing idea. I and likely many others would welcome the involvement of public health officials in collecting this information. I would hope, though, that it would be done in an objective manner. The Harvard paper refers to “police violence.” While there are all too many instances in which police are not justified in shooting civilians, there are also certainly many instances in which police use of firearms protects the public, as was the recent case in San Bernardino when law enforcement officers shot and killed two alleged terrorists after the couple were said to have killed 14 individuals and were potentially on their way to attack others.

This year, I had the honor of co-chairing the President’s Task Force on 21st Century Policing. We quickly realized that information about police use of force and about deaths of individuals in custody was lacking, and recommended better systems for collecting it. When we presented the report to President Obama, he zeroed in on this issue and wanted to know why we couldn’t make grants to police departments conditional on their providing accurate statistics. I explained that wouldn’t work because small departments — half of the 18,000 law enforcement agencies in the United States have fewer than ten sworn officers — don’t get direct federal grants.

The states, however, could make a huge difference in collecting more reliable criminal justice statistics. State legislatures can impose these kinds of regulatory controls over local governments. State legislatures could direct local police departments to collect these data and report them.

We don’t know if officer-involved shootings are becoming more common, or if we are just hearing about them more today. More, and more reliable, statistics from local law enforcement agencies and from public health officials would give us a clearer picture of law-enforcement-related shootings and deaths, as well as many other facets of law enforcement, and help us formulate good public policy.

Laurie O. Robinson is professor of criminology, law, and society at George Mason University and was co-chair of the President’s Task Force on 21st Century Policing.

By Brian Root: Without accurate data on the number of individuals killed by law enforcement officers, who was killed, what officers were involved, and the circumstances surrounding each incident, law enforcement leaders are hard-pressed to reduce the number of officer-involved deaths. Equally important, the lack of reliable data makes it difficult to secure accountability for these deaths. Officers involved in lethal force incidents, even those implicated in unjustified use of force, are rarely indicted, let alone convicted. In cases with clear officer wrongdoing, the government has an obligation to investigate and appropriately prosecute those responsible.

People who don’t work with criminal justice data are usually shocked that we don’t have basic numbers on officer-involved deaths. It isn’t a subjective or esoteric thing to measure: you count the number of people who have been killed. Having worked with criminal justice data for years, I know firsthand that they are a nightmare.

The main reason? Although we use the term “criminal justice system,” it isn’t a system, especially when compared with the public health system. It is thousands of disconnected agencies and jurisdictions, each responsible for its own data with no standardization between them. It’s not just law enforcement. Courts, prosecutors, defenders, and the like all have their own systems that don’t speak to each other. There is little impetus for the more than 18,000 law enforcement agencies to self-report police killings in a standard way. Worse, there is no body with the oversight, resources or political willpower to impose some kind of standardized data collection on officer-involved deaths. I’m skeptical about any solution for counting these events coming from the Department of Justice or Congress.

“Having worked with criminal justice data for years, I know firsthand that they are a nightmare.”

Brian Root, Human Rights Watch

Turning this into a public health issue makes good sense. For the most part, every death touches the public health system at the hospital, the coroner’s office, or elsewhere. Making a law-enforcement-related death a notifiable event would be one way we could get a grip on this problem.

Of course, whenever you are dealing with data, the definitions are essential. The CDC or whoever might implement the Harvard suggestion must define what a law-enforcement-related death actually is. Does someone have to be in the process of being arrested, or does a killing following a simple stop also qualify? Will we at least be able to gain data on which law enforcement agency was involved? In developing a reporting system, the devil will be in the details.

The National Notifiable Diseases Surveillance System gathers data from 57 state, local, territorial, and tribal health departments. To get a complete picture of law-enforcement-related deaths, we would still need voluntary participation from many public health agencies. But it could be a system that we can push toward something more standardized and universal.

Brian Root, PhD, is a quantitative analyst for Human Rights Watch in New York City.

By David A. Klinger: Law enforcement in the United States is a fractured system. There are 18,000 or so municipal, county, state, and specialized departments, such as transportation or housing authority police. For better or for worse, there is no direct federal oversight of these departments, nor is there a requirement to report law-enforcement-related deaths. In order to understand when and why police use deadly force, we need to know how often it occurs and the circumstances around it.

“We need to think more catholically about the issues surrounding officer-involved shootings.”

David Klinger, University of Missouri-St. Louis

I applaud the Harvard team’s suggestion. It is a good start. But it will miss many important events. Among individuals who are shot by police officers, more survive than die. There are also many instances in which police officers discharge their weapons and all of the rounds miss their target. By counting dead bodies, we will not capture these shootings.

We need to think more catholically about the issues surrounding officer-involved shootings. The best organization for collecting statistics on police shootings is likely the Bureau of Justice Statistics. It has a long history of collecting data about various crime and justice matters. The bureau needs to roll out a reporting system that law enforcement agencies would agree to use for officer-involved shootings. It may need some teeth to get agencies to comply. Knowing details such as the age and race of the person shot, how many other suspects were present, how many officers were involved, how many officers fired, how many rounds were fired, and the like would give us a clearer picture of these shootings.

David A. Klinger, PhD, is a professor of criminology and criminal justice at the University of Missouri-St. Louis, a senior fellow at the Police Foundation in Washington, D.C., and author of Kill Zone: A Cop’s Eye View of Deadly Force.

By Philip Stinson: The lack of reliable statistics on officer-involved shootings has been a problem for years. Until recently, no one has really cared to notice. Now there’s a real hunger for these numbers, but no one has them.

Getting accurate numbers on officer-involved shootings will be difficult. Law enforcement agencies are supposed to file information about officer-involved shootings, but few do, and there is no way to force them to do it.

The suggestion by the Harvard School of Public Health team to make law-enforcement-related deaths so-called notifiable conditions is one way to come at the problem, in part because these deaths are truly a public health issue.

“The lack of reliable statistics on officer-involved shootings has been a problem for years. Until recently, no one has really cared to notice.”

Philip Stinson, Bowling Green State University

But I believe we will need several different databases using different methods to collect the data. My research group and others, for example, use Google alerts to capture reports of officer-involved shootings. This is a good complement to public health reporting. Having several databases lets us triangulate to the real numbers.

While I understand the need to collect flu and meningitis data in real time, I’m not sure that officer-involved shooting data needs to be collected in real time, at least not from a sociological perspective. Taking a bit longer may yield more accurate data.

Philip Stinson, PhD, is associate professor of criminal justice at Bowling Green State University in Ohio.

By Robert J. Kaminski:

The proposal from the Harvard School of Public Health research team to collect data on law enforcement-related deaths is an excellent idea. The Federal Bureau of Investigation’s Supplementary Homicide Data program and the Bureau of Justice Statistics’ Deaths in Custody Reporting program capture about half of police justifiable homicides and deaths from all causes of persons in the practical custody of law enforcement, respectively. Having another data source that examines law enforcement-related fatalities from a public health perspective would help paint a more complete picture of such incidents. While the collection of data on deaths is of the upmost importance, I have some reservations that the proposed effort would, at least initially, be limited to fatalities (the vast majority of which are firearm related).

A major focus of my research concerns violence against law enforcement officers. In 2012, I presented research on firearm violence directed against police to the National Officer Safety and Wellness Group in Washington, DC. One major point of my presentation was that including nonfatal shootings (woundings and near misses), which are far more common than fatal shootings, may have important implications for identifying high-risk locations and times, tracking changes in trends and patterns, and identifying causes and prevention efforts.

In my analysis of New York City Police Department officers shot and killed and shot and wounded over a 30-year period, for example, shows that examining only fatal shootings seriously underestimates the incidence of firearm-related violence directed at police and fails to reveal a rising trend in shootings from the early 1980s through the early 1990s. An analysis of officer-involved fatal shootings, woundings, and near misses in the Las Vegas and North Las Vegas Police Departments yielded different results depending on whether it included only fatalities, fatalities plus woundings, and fatalities plus woundings plus near misses.

I understand the Harvard team’s concerns about the burden of attempting to collect information on nonfatal injuries at this juncture. Perhaps that burden could be reduced by restricting the collection of these data to nonfatal firearm injuries, especially since the vast majority of law enforcement-related deaths are firearm related.

Robert J. Kaminski, PhD, is associate professor of criminology and criminal justice at the University of South Carolina.