I think this "the dog that didn't bark" phenomenon may be a much more important phenomenon than meets the eye. People's perception of the world is a function of their observations of reality - for starters, the vast majority of people don't even have any exposure to things like police data (other than the typically misleading narratives that they might encounter in the news), but then even the small majority of people who do dig deeper into the details, these people too will leave with a distorted perception of reality if only a subset of the data is revealed, and they do not know that it is only a subset.

The end result of all of this complexity is that people think the world is one way, but it is actually another, and the vastness of the gap between the two could be anywhere from small and insignificant, to gigantic and crucially significant, if one applies this idea to not just police data, but everything (the entirety of reality is subject to this flaw, and many others that we do not realize).

Add the fact that even the businessmen and regulators, influential people, politicians and judges, they all build their image of the world mostly on the same sources as everyone else - and it's becoming worryingly apparent that all the control systems in societies are operating with very low visibility into the things they're controlling.

Imagine a PID controller keeping a boiler from overheating. Now imagine the temperature input to that controller is a thermometer mounted to the door of a metal cabinet next to the boiler, and thus it reports anything related to the boiler temperature only when somebody is looking for something in that cabinet (when the opened metal doors touch the boiler and conduct heat). Doesn't sound like a good control system.

They could and should document and account for this somewhere in the data set.

It wouldn’t solve the problem but it would allow the bias to be visible.