The government does run one official count. It has the biggest reach on paper and the least to show for it.
In 2019 the FBI started a program called the National Use-of-Force Data Collection. Its reach is wider than any other effort in this series. It is meant to capture every time police use force that kills someone, badly injures them, or fires a gun at them. On paper, it should be the count that settles the argument.
One choice at the center of it breaks the whole thing. Reporting is optional. A police department sends in its numbers only if it feels like it. And the FBI made itself a rule: it will not release a national total until the departments that do report employ at least 80 percent of the country's police officers. Below that line, the number stays locked in a drawer.[src]
It started at 46 percent and took until 2024, its sixth year, to reach the line, then dropped below it again the next year.[src]
Watch out for a trick in these numbers, because two different things get called "coverage." The 80 percent is the share of officers. Counted by departments, participation is far lower: in 2022 only about 6,750 of roughly 18,500 departments reported, near 37 percent.[src] A handful of big-city departments cover a lot of officers, so the officer number always looks healthier than the department number. Most small departments simply never send anything.
For its first five years the federal count basically sat in the dark. It had a form, a database, and a target it kept missing, because the data it needed sat inside thousands of departments that were never required to hand it over. The government wrote itself the widest rule in the country and then had no way to make anyone follow it.
So who filled the gap while the official number stayed locked away? The newspaper, the nonprofit, and the volunteer from No. 01. That is the whole problem in one sentence.