Cops do an intentionally-terrible job of collecting statistics, and their unions vigorously resist attempts to make them do a better job.
There's a Some More News episode about police car chases (TL;DW they're definitely net-harmful and should be avoided in all but the most extreme circumstances, but the public likes watching them and doesn't realize how bad they are, and police fucking love anything that gives them an excuse to Go Nuts and act like they're in an action movie, so they remain distressingly common in most states) that, as a necessary aside, also discusses how intentionally awful police stats collection is—and you'd think maybe it was just for that one issue, or SMN's researchers suck at their jobs, except you run into the exact same problem any time you look into deep reporting or research on basically any issue connected to US policing. What we have looks really bad, but also we're constantly working from (probably less-bad-looking than the median...) datasets from a few places that manage to actually collect meaningfully-complete statistics.
There's a Some More News episode about police car chases (TL;DW they're definitely net-harmful and should be avoided in all but the most extreme circumstances, but the public likes watching them and doesn't realize how bad they are, and police fucking love anything that gives them an excuse to Go Nuts and act like they're in an action movie, so they remain distressingly common in most states) that, as a necessary aside, also discusses how intentionally awful police stats collection is—and you'd think maybe it was just for that one issue, or SMN's researchers suck at their jobs, except you run into the exact same problem any time you look into deep reporting or research on basically any issue connected to US policing. What we have looks really bad, but also we're constantly working from (probably less-bad-looking than the median...) datasets from a few places that manage to actually collect meaningfully-complete statistics.