> What about using the proportion of black assailants as reported by the victims? Would that be relatively free of bias?

No. There is considerable history of victims not only misidentifying the perpetrator, but even inventing the offense to smear someone because of racial bias, personal animus, or other reasons. To the extent that there is a systematic racial bias throughout society, you'd expect it to be reflected in such reports.

A "considerable history"? You've got nothing here. And you're gonna have to deal with it. Victim of crime survey data is an actual thing we have. You're trying to make it untrue because of your own bias.