The Use-of-Force Register

The problem

The United States does not know how many people its police kill. It knows roughly, because other people count. The Washington Post kept a database of every fatal police shooting from 2015 and recorded more than 10,000 over the decade, and 2024 was the highest year in it; then the Post stopped counting on January 1, 2025. Mapping Police Violence, a nonprofit, counted at least 1,365 people killed by police in 2024, up from 1,329 in 2023, one every six and a half hours, with only ten days in the year on which nobody was. These are the best numbers the country has, and none of them is the government's.

The government has tried twice and failed twice. In 2019 the FBI launched the National Use-of-Force Data Collection, to which any of the country's roughly 18,000 law enforcement agencies may report, if they like. The Bureau's rules allow it to publish full national figures only when agencies covering 80 percent of officers participate. In June 2024 it stood at 72 percent. In 2021 it came close to being shut down for missing the lower threshold that federal data collections must meet to survive. Five years on, the count exists but the country cannot read it.

The second attempt was Congress's. The Death in Custody Reporting Act of 2013 requires every state to report every death during arrest or in custody to the Department of Justice, on pain of losing a tenth of its Byrne grant. When the Government Accountability Office checked in 2022, nine states had reported no deaths at all for fiscal 2021; in four of those states GAO found 124 on its own. Comparing the states' filings against public records, GAO identified nearly 1,000 deaths that should probably have been reported and were not, and found that 70 percent of the records that were filed were missing at least one required element. The Department had not determined whether a single state was in compliance, and the penalty has never been applied.

№ 055 argued that a stop needs a reason on the record, and № 056 that a case needs to be closed. This is the third rule of the badge: force needs a count. Not because a number ends an argument, but because without one the country has the same argument after every incident and never learns whether things are getting better or worse.

The proposal

A national Use-of-Force Register. Every use of force by a law enforcement officer that results in death or serious injury, and every discharge of a firearm at a person, reported within 30 days, under one definition, as a condition of every federal dollar an agency or its state receives. Published quarterly, by agency. Reconciled every year against death records and press reports by an independent statistical office, which publishes the gap.

How it would work

1. One definition. The FBI's existing collection already defines a reportable incident: death, serious bodily injury, or a firearm discharged at or in the direction of a person. Keep that definition and write it into statute, so that a chokehold in Ohio and a chokehold in Oregon are the same event on the same form.

1. Every dollar, not a tenth of one grant. The 2013 Act's penalty was a tenth of one grant stream, and it was never imposed. Under this proposal, an agency that does not report is ineligible for any federal law enforcement funding, and a state in which more than a set share of agencies do not report loses its share too. States have every tool they need to compel their own departments. They will use them when the money depends on it.

1. Thirty days, by agency, in public. Reports are filed within 30 days of the incident. The register publishes quarterly, naming the agency, the type of force, the outcome, and whether the person was armed, with the officer identified by a persistent anonymous number so that a pattern is visible without a name. Anyone can see how many times their city's police shot at someone this year, and how many of those people died.

1. The reconciliation. Numbers reported by the people being counted are never enough on their own. Every year the Bureau of Justice Statistics, which already runs the arrest-related deaths programme, matches the register against death certificates, medical examiner records, and press reports, exactly as GAO did by hand in 2022, and publishes, by state, the number of deaths it found that the register did not. A state with a gap explains it in public. The gap is the score.

1. The decertification link. An officer who is dismissed or decertified for excessive force is entered in the existing National Decertification Index, which is today voluntary and incomplete, and a department may not hire an officer without checking it. The register and the index are cross-referenced by the same anonymous number, so that an officer with a record of force in one state cannot start fresh in another.

1. Serious injury, not only death. The nonprofit counts are of the dead, because death is what a newspaper can verify. The register includes serious injury, which is far more common and almost never counted, and which is where most of the pattern is.

The numbers

What exists. 1,365 deaths in 2024 by the nonprofit count; more than 10,000 fatal shootings in a decade by the newspaper's; a federal collection at 72 percent participation that cannot publish; a statutory death count with nine states filing zero. There is no national figure at all for serious injuries.

What it costs. The FBI already runs the collection and the Bureau of Justice Statistics already runs the death programme. The additional cost is the reconciliation, the publishing, and enforcement of the condition, on the order of $50 million a year, against about $130 billion the country spends on local policing and several billion in federal grants to it. It is the cheapest proposal in this block, because the machinery is built and is simply not being used.

What the penalty is worth. Federal law enforcement grants to states and localities run in the low billions of dollars a year. No state will forgo them over a reporting form. The condition is not designed to be imposed. It is designed to make imposition unnecessary.

The honest objections

"A count of police killings will be used to inflame, not to inform." It will be used both ways, as every public number is. But the alternative is not silence; it is the present, in which the counting is done by advocacy groups and newspapers and the police have no official number with which to answer them. A department that shoots fewer people than its size would predict has, today, no way to prove it. The register gives it one.

"Small departments can't carry another reporting burden." A department that has a reportable incident has, by definition, already written an incident report; the register asks that one copy be sent to one place within a month. Most small agencies will file nothing in most years, because they have nothing to file. The burden falls on the agencies with the incidents, which is where it belongs.

"The federal government cannot commandeer state police." It cannot, and this proposal does not. It attaches a condition to money, which the Supreme Court has permitted for as long as there has been federal money, provided the condition is clear and related to the grant. A count of the use of force is related to a policing grant in the most direct way there is. A state that would rather not report may decline the money, as now.

"Counting deaths does nothing to prevent them." On its own, no. But every reform this country has tried in policing has been argued from anecdote, because there was nothing else, and the results have been argued from anecdote too. Aviation did not become safe by resolving to be careful; it became safe by reporting every incident and studying the count. A register is not the reform. It is the instrument that lets anyone tell whether a reform worked.

"Officers will be identified from the anonymous number." The number is persistent so that patterns can be seen, and it is anonymous so that individuals are not. That is the same design used for physicians in malpractice data and for pilots in incident reporting, and it has held. Where an incident is already public through a court case, the name is already public. The register does not add to that; it adds the incidents that never reach a court.

Sources

• Federal Bureau of Investigation, National Use-of-Force Data Collection, participation updates and publication thresholds, 2024, fbi.gov
• U.S. Government Accountability Office, Deaths in Custody: Additional Action Needed to Help Ensure Data Collected by DOJ Are Utilized (GAO-22-106033), December 2022
• Death in Custody Reporting Act of 2013, Pub. L. 113-242; Congressional Research Service, Death in Custody Reporting Act: Background and Legislative Considerations (R47559)
• The Washington Post, Fatal Force database, 2015–2024, and the announcement ending the collection, washingtonpost.com
• Campaign Zero, Mapping Police Violence, 2024 annual report, mappingpoliceviolence.org
• Bureau of Justice Statistics, Census of State and Local Law Enforcement Agencies, agency counts, bjs.ojp.gov
• International Association of Directors of Law Enforcement Standards and Training, National Decertification Index, iadlest.org
• Proposals № 051 (The Citizen Card); № 055 (The Warrant Standard); № 056 (The Clearance Rate)