I would see another more likely explanation than your "likely prejudice and racism." I think it's more a function of police interactions. So, instead of using total number of each race in your calculation denominator, you need to use total numbers of each race involved in police interactions (let's say arrested). That will give a more true indication and change your conclusion.