Tuesday, December 12, 2017

potential racial bias in Louisville police searches

The CJ writer misses the most important number: the rate of successful searches by race. (Here's the link to the study.)

If search rates differ but success rates are the same, bias is much less likely. (An even stronger example: loan default rates by borrowers vs. loan rejection rates by lenders. If default rates are higher for those in group X vs. Y, then group X is getting too many loans.)

In this case, searches on whites yielded greater success than on blacks, which leads a (weak) inference that blacks are searched too often relative to whites. (Why a weak inference? There are other variables in play.)


Post a Comment

Subscribe to Post Comments [Atom]

<< Home