Who counts the officers? The same broken way, it turns out. That is the point.
This essay exists because a reader asked the fairest question the series had not answered: nobody accurately counts the officers who are killed or seriously hurt by the people they serve, so where is that essay? Here it is, held to the same rules as the other ten. And the finding is the strongest possible vindication of the question: the count of what happens to police fails in almost exactly the ways essays one through eight documented for the count of what happens by police. Voluntary reporting. A definitional split that produces totals six times apart. Volunteer counters outrunning the official ledger. If this series is right that America has a counting problem, the problem should show up on both sides of the ledger. It does.
The FBI has counted officer deaths and assaults since 1937 under a program now called LEOKA, Law Enforcement Officers Killed and Assaulted. Its strong half is felonious killings, officers intentionally killed in the line of duty: 46 in 2020, then a spike to 73 in 2021, the most since the September 11 attacks, then 61, 60, and 64 in the three years after, a stretch the FBI itself describes as consistent.[src] [src] Deaths are rare, public, and investigated, so this half of the ledger is about as good as American police data gets.
The hollow half is everything short of death. The assault and injury count is voluntary, agency by agency, exactly like the use-of-force collection in essay two. In 2023 the assault figures came from 10,884 agencies employing about 600,000 officers,[src] out of roughly 18,500 American police agencies, the same denominator this series has used all along.[live] That is about six in ten agencies. So when the 2024 report announces the highest officer assault rate in ten years,[src] the honest reading is the one this series taught: that is the rate among whoever showed up, and nobody can say what the country's real number is, because four in ten agencies said nothing at all.
It gets worse in a familiar way. When the FBI retired its old summary reporting system in 2021 and required agencies to file through the newer NIBRS pipeline, LEOKA fell into the same hole as the national crime count: the number of agencies reporting officer assault data collapsed from around 16,000 to roughly 5,000 in a single year before beginning to recover.[src] For that stretch, the official record of violence against police in America ran on about a quarter of its former base. Whole states went quiet. Any trend line drawn through 2021 crosses a data void, on the officers' side of the ledger, not just the public's.
Every dynamic from essay one has a mirror image. Because the official count is partial, private counters filled the gap, and the oldest of them predates every civilian tracker in this series: the Officer Down Memorial Page, run as a nonprofit since 1996, keeping a page for every officer, and every K9, who dies in the line of duty.[src] Alongside it stands the National Law Enforcement Officers Memorial Fund, whose annual fatality reports are the numbers most often quoted in the press.[src]
And, exactly as in essay four, the counters disagree, not because anyone is wrong but because they define differently. The FBI's felonious count excludes accidents and illness. The memorials include them, with differing rules. Watch what that does to a single year:
| Counter · rule | Officer deaths, 2021 |
|---|---|
| FBI LEOKA · feloniously killed only[src] | 73 |
| NLEOMF · all line-of-duty deaths, including illness and accident[src] | 458 |
Seventy-three versus 458 is a wider definitional gap than anything in the civilian record this series measured. And the memorials disagree with each other too: for 2020, ODMP counted 374 line-of-duty deaths while NLEOMF counted 295, because their inclusion rules differ.[src] [src] Five counters and five totals was the civilian story. The officer story is the same story with the same moral: the definition is the disagreement.
What pulled those totals apart was the hardest definitional question the officer ledger has ever faced: COVID-19 was the leading cause of line-of-duty deaths among American officers in 2020 and again in 2021, with hundreds of deaths attributed to duty-related exposure.[src] [src] Whether a virus contracted somewhere in a week of shifts counts as dying in the line of duty is not a fact you can look up. It is a definition somebody has to write, with presumption rules and evidence standards, and the counters wrote different ones, so the country got different totals. Readers of this series will recognize that instantly: it is the arrest-related-death question, the off-duty-shooting question, the was-it-the-restraint question, transposed to the other side of the ledger.
The point of this essay is not that one count matters and the other does not. It is that the same design failures produce both broken counts, which means the same design, essay seven's four parts, fixes both: a mandate, because voluntary reporting never finishes on either side; a written definition, because "line of duty" is doing as much silent work as "killed by police" ever did; the open-source review method, which the memorial pages have effectively been running on the officer side since 1996; and an audit against the independent lists, which on this side of the ledger already exist and already disagree usefully.
A state that counted well would know what happens to everyone in the encounter, the person on the ground and the officer standing over them, in the same ledger, under written definitions, audited by outsiders. Nothing in that sentence takes a side. That is what makes it a measurement standard instead of a position.
What this essay has said: the official count of violence against officers is strong for deaths and hollow for everything else; the assault figures are voluntary, cover about six in ten agencies, and cratered to about 5,000 reporters in 2021; the volunteer memorials predate and outrun the official ledger and disagree with each other and with the FBI on definition, 73 versus 458 for the same year; and the same four-part spec this series wrote for the civilian count applies symmetrically.
What this essay has not done: compared dangers, weighed one side's deaths against the other's, or drawn any conclusion about who is threatened by whom. Those are exactly the arguments this country runs on numbers it does not have, which is the series' whole complaint. No scoreboard is offered here, because no honest one can currently be built, and that, precisely, is the finding.