Perhaps each group has some level of both and we shouldn't generalize to entire strata and population groups too much. Instead maybe we shouldn't be so quick to measure outcomes and just fix the problems. Instead we're using outcomes as a (I would argue) poor measure of how well we're fixing the problems we should be agreeing are problems on their own worth fixing. That's where a lot of the argument/debate gets lost, quibbling over correlations instead of just solving bad things.