There are roughly 800,000 cops in the United States...

[HereSince1628]

59. Cops as the base of the percentage seems wrong, indeed percentages seem minimally useful

It considers the risk of bad policing events compared to all cops. But cops, good and bad, are involved in many policing events everyday, and it's quite likely that even bad cops don't do 100% bad policing...

If you want to know the unconditioned risk of a bad policing event, such events should be measured against all policing events...a comparative base that would be many multiples of the actual number of police...and which would make the problem of bad events seem even smaller.

Much of society has a very low tolerance for bad policing events. But, what is needed/wanted is to really reduce the occurrence of these events, not to make the proportion of these events look smaller by playing with a denominator in a ratio.

Reducing occurrences requires identifying and managing hazards that contribute to bad events, so that the hazards can be anticipated and managed to reduce the occurrence of the actual events.

Because not all policing events are as likely to result in a bad policing event, initial detection of hazards that contribute to bad events must be studied among good and bad policing events where a suspect factor is present as well as where the suspect factor is absent.

Simple percents don't work very well for this.

Simple statistical tests like chi-square contingency tables can be used to analyze such data to detect factors that co-occur in patterns with the bad event, although in their simplest form chi-square type analysis doesn't distinguish features among bad events that co-occur in patterns but which are spurious with respect to causation.

Addressing spurious factors doesn't reduce risk, it misdirects resources. The analysis of the problem requires a bit more sophistication. And it should be obvious that referencing poorly considered percentages reduces the opportunity to obtain meaningful understanding.

Moreover, even after all the statistical analyses are completed, not every contributing hazard may be amenable to management. Management of a particular risk can only be directed and tractable hazards.

What is and isn't tractable isn't a question of statistics.