"But there could be a racial bias that causes police to be more likely to stop black Americans." Even if this is true (and it's probably true to some extent),

"But there could be a racial bias that causes police to be more likely to stop black Americans." Even if this is true (and it's probably true to some extent), by focusing on 'interactions only' we can isolate the degree to which cops may be racially biased in at least those incidents wherein there is interaction. This would yield much more valuable results than the paper in the link. The linked paper is almost useless because there is a very large divide in who cops interact with. So many factors there including police density, income, poverty, education etc..