The United States does not reliably count the people its police kill. The authoritative national numbers are built by a newspaper, activists, and volunteers, not the state, and they disagree because they measure different things.
Aggregates only. This hub reconciles five cited public efforts and publishes derived counts. It does not re-host any person-level records. Updated 2026-07-23.
Five independent projects count people killed by police in the United States. No two report the same total, and the reason is in the last column: each one decided the word "killing" means something different.
| Counter | Who runs it | Kind | What it counts | Span | Status | License | |
|---|---|---|---|---|---|---|---|
| Fatal Force[src] | The Washington Post | Newspaper | Fatal shootings by an on-duty police officer only. Excludes deaths in custody, off-duty shootings, and every non-shooting death (Tasers, restraint, vehicles). | 2015-2024 | Ended (tracking stopped 2024-12-31) | CC BY-NC-SA 4.0 | |
| Mapping Police Violence[src] | Campaign Zero | Activist | All people killed by police by any means, including off-duty officers and deaths in custody. Broader than shootings. | 2013-present | Active | Free for non-commercial use with attribution | |
| Fatal Encounters[src] | D. Brian Burghart and volunteers | Volunteer | All deaths during interactions with police, by any cause. The widest net of the five, and the one that seeded the others. | 2000-2021 | Volunteer updates wound down after 2021 | Free to download and reuse | |
| National Use-of-Force Data Collection[src] | Federal Bureau of Investigation | Federal | All uses of force resulting in death or serious bodily injury, or the discharge of a firearm at a person. Voluntary agency participation, so its totals stay sealed until coverage is high enough. | 2019-present | Active, but thresholded | Public data (federal) | |
| The Counted[src] | The Guardian | Newspaper | All killings by law enforcement, including shootings, Tasers, vehicles, and custody deaths. A two-year effort that ran alongside the early Post database. | 2015-2016 | Ended after 2016 | Journalistic database |
The Post recorded 10,430 people over 2015-2024 (the full v2 dataset, aggregated), an average above 1,000 a year, with 2024 the highest annual count on record.[src] Fatal Encounters holds more than 28,000 records reaching back to january 2000, the deepest history of the five.[src]
The clearest way to see it is one year, two counters. In 2015 the Post logged fatal shootings only, 995 of them. Mapping Police Violence, which starts from the Post's cases and adds the rest, logged 1099 killings by all means. Same country, same year, 104 people apart, and that gap, the non-shooting killings, shows up every year of the decade.
The one counter the government runs, the FBI National Use-of-Force collection, has the broadest definition of all and the least to show for it. For most of its life it could not publish a national total, because too few agencies chose to report. It first cleared its own coverage threshold in 2024.[src]
See the coverage curve →Every figure on this page is frozen into a versioned, hashed release (v0.4.0, CC-BY-4.0). It carries the five counters, their annual totals, the federal coverage series, the record-level overlap, the federal-validation, stratified-undercount, international, and disparity tables, and the source ledger, as CSV and JSON. It publishes derived aggregates, never a person-level roster.
| File | SHA-256 | |
|---|---|---|
| datasets.json | 4c1cee2264500c67... | download |
| datasets.csv | bcaeecd36e731b0c... | download |
| annual.json | 913238269b8b08e4... | download |
| annual.csv | 19592c3c3f12d88e... | download |
| officer_coverage.json | bba4c9e917165b7a... | download |
| officer_coverage.csv | 7dd4c4d0a0d97690... | download |
| agency_coverage.json | 81f5f3f07c1e8696... | download |
| agency_coverage.csv | 8f4d48bbff7733b1... | download |
| overlap.json | d2a560f0e849e04f... | download |
| overlap.csv | 26bd97ebba40093e... | download |
| federal_validation.json | 2a4abd2c1cf18272... | download |
| federal_validation.csv | aff353bf94fcfed2... | download |
| stratified_undercount.json | 4966ebefa6d21fad... | download |
| stratified_undercount.csv | f29d53f30dde5d83... | download |
| international.json | 1369386dff3b19d4... | download |
| international.csv | 99dd6bd132d492ee... | download |
| disparity.json | 6640d6d23702133d... | download |
| disparity.csv | 4774242cf6a3f976... | download |
| sources.json | a1e9c7a8ec14905e... | download |
| sources.csv | 079304856949148e... | download |
Also in the bundle: manifest.json, SHA256SUMS.
Each row recomputes a headline fact from the committed release. If a number on this page is wrong, a row here goes red.
A check did not pass. The page and the release have drifted.
I build the AI tools and automations a team adopts, and the full-stack apps and data pipelines under them, and ship them to production. Tell me the problem in a sentence and I'll give you an honest read on fit within a day.
Work with me →