Why is the count rising? Why do agencies not report? Seven essays refused to guess. Here is the careful version.
This series has been strict about one thing: measure first, explain never. That discipline earned the numbers their trust, and now it can be spent, carefully. This essay takes the two questions every reader eventually asks and answers them the same way the series counts: explanations ranked by the strength of their evidence, every figure sourced, and the difference between "the data shows" and "researchers believe" kept visible at all times. No verdicts on any person or department. Those are the rules.
The best-supported explanation is geographic. Killings by police have been declining in the biggest cities, most of which rewrote their use-of-force rules, added de-escalation requirements, and built crisis-response alternatives over the past decade, while rising in suburban and rural areas policed by thousands of smaller agencies where little changed.[src] The national total is just the sum of those two curves, and the non-reforming zone is much larger than the reforming one. This explanation also survives the measurement test: the rise shows up identically in two instruments with different rules that agree on 98 percent of the actual people, so it is happening in the world, not in the counting.
Most of the people police kill are armed, and public-health researchers have found that state-level rates of fatal police shootings track civilian firearm prevalence, even after controlling for crime, poverty, and urbanization, with the association strongest for shootings of people armed with guns.[src] A growing civilian gun stock mechanically raises the number of armed encounters. This is the least comfortable explanation on the list, because it predicts the trend continues even if no officer's conduct changes at all.
A stable fraction of the people killed, on the order of one in five, are in a mental-health crisis when police arrive,[src] and police remain the default responders to crises nobody else is dispatched to. The alternatives, co-responder units and crisis teams, exist mostly in the big cities of the first explanation, which is why this one compounds the geography rather than competing with it.
The same year-end analysis notes that killings rose while arrests and police stops fell,[src] which means the rate of killings per police contact is climbing faster than the raw count. That cuts against the simple story that more policing produces more deaths, and toward the encounters themselves becoming more dangerous, which is what the guns explanation predicts.
It is not crime: the record years of 2023 and 2024 arrived while national homicide figures were falling, a decoupling this series' own data shows. And it is not better counting: the instruments kept their rules fixed for a decade and agree person by person. The deep background, argued at book length by the criminologist Franklin Zimring, is that roughly a thousand deaths a year is what tens of millions of annual police contacts in a heavily armed country structurally produce.[src] The level moves slowly. The geography moves fast.
Start with the tell this series' own dataset carries. The federal collection has covered anywhere from 46 to 80 percent of the country's officers in a given year; in 2022, the 53 percent of officers it covered came from only about 37 percent of its departments. Read those two numbers together and the picture inverts: the big departments, the ones with the most killings, the most lawsuits, and the most to hide, overwhelmingly do report. The missing mass is thousands of small agencies, many under ten officers. If concealment were the driver, the pattern would run the other way. The evidence points somewhere more boring.
Congress already passed a mandate. The Death in Custody Reporting Act of 2013 requires states to report these deaths and lets the Attorney General dock 10 percent of a noncompliant state's federal justice grants.[src] That penalty has never once been applied, the department started collecting four years after its own statutory deadline, and its Inspector General wrote the implementation up in 2018.[src] [src] So the operating regime, as an agency experiences it, is simple: voluntary programs are voluntary, and mandatory programs are voluntary too.
The FBI's collection wants monthly submissions, including affirmative zero-reports in months when nothing happened, from agencies where the records department is one sergeant's afternoon. Government auditors found participation stuck, by their own measure, at 44 and 55 percent in the first two years, low enough that federal data-quality rules nearly forced the FBI to shut the entire collection down at the end of 2022.[src] [src] And because the FBI could not publish anything until coverage cleared its threshold, the agencies that did file got nothing back for five years. Reporting into a void is a hard habit to sell.
The same audit trail describes small agencies that simply did not know the program existed, records systems that could not produce the required fields, and states with no conduit between local departments and the federal form. The years in question also consumed agency reporting capacity with a separate mandatory overhaul of federal crime reporting. None of this is sinister. All of it is sufficient.
The FBI told auditors that some agencies are reluctant to share sensitive use-of-force data, and lawyers do advise caution about creating discoverable records.[src] It belongs on the list, and last: the big urban departments that draw the most lawsuits report at the highest rates, which is not what a concealment-first story predicts.
Put the two answers side by side and they turn out to be one answer. The rise is concentrated in small suburban and rural agencies. The federal blind spot is made of small suburban and rural agencies. Whatever each department's reason for silence, the practical effect is that the fastest-growing part of the phenomenon is the least officially visible part. That is this series' whole thesis, stated for the last time as a finding: the parts of the world the state cannot see are the parts that change without it noticing.