Herrnstein and Murray's (HM) The Bell Curve famously reported that, controlling for cognitive ability, most or all of the black-white earnings gap disappears. An interesting chapter in Intelligence, Genes, and Success: Scientists Respond to The Bell Curve challenges this result - but not in the way you'd expect. The chapter is "The Hidden Gender Restriction: The Need for Proper Controls When Testing for Racial Discrimination" by Cavallo, El-Abbadi, and Heeb (CEH).
The paper begins by replicating HM's original result. So far, so good. But, CEH explain, HM's analysis smuggles in the statistical assumption that the effect of being black is the same for men and women. And it's not. Controlling for all other variables, black men still earn substantially less than you'd expect.
Oh, and black women earn substantially MORE than you'd expect.
This is a very interesting finding - and one that I noticed myself a few years before I read this chapter. But it's arguably even more contrarian than HM's finding. If subpar earnings are evidence of discrimination against black males, are superpar earnings evidence of discrimination in favor of black females?
When you put it that way, you start to wonder if it makes sense to automatically label statistically unexplained earnings gaps between groups as "discrimination." And the answer, of course, is No. Jews (Jewish men, to be precise) earn vastly more than you would expect given their education, IQ, and so on, and it hard to believe the reason is philo-Semitism.
So why do black women earn more than you'd expect, while black men earn less? One story I've heard is that black females are a "two-fer" for affirmative action purposes. But the effect is way too large for this to be plausible.
A better explanation, I think, comes from novels (and movies) like Terry McMillan's Waiting to Exhale. McMillian depicts a world of successful, independent black career women, and the black men who disappoint them. In her novels, women and men partake in radically different subcultures: On average, the women are achievement- and future-oriented; on average, the men are not. These cultural differences in turn lead to differences in earnings.
The McMillan story is extremely consistent with my first-hand observation. But I freely admit that I've only got a few data points. Anyone care to supplement first-hand observation, one way or the other?