Education premiums and disadvantaged groups

A lot of you have probably been hearing about this new study by Dwyer, Hodson, and McCloud suggesting that women earn higher returns to education than men, which is part of the reason that men are more likely to drop out (Greg Mankiw highlights a piece of it here).

This is similar to my findings with Marla McDaniel in our Review of Black Political Economy article on the impact of a high school diploma on employment for youth who did not go on to college (so a completely different segment of the labor market from this work on college). The two big takeaway from the paper were that black high school graduates had comparable employment profiles to white high school dropouts (I had a similar conclusion using a different dataset in this recent briefing paper). But the other notable finding was that although black graduates do worse, the added benefit of graduating is greater than for whites.

It wasn't one of those awesome deal-with-the-endogeneity papers but it had a lot of things that you don't normally have like mental health scores, cognitive ability scores, measures of the quality of home life, etc. Kitchen sink regressions aren't great, but when you start getting at that kind of stuff I feel like the results are more robust. If anything it's an underestimate I'd expect because obviously these students aren't going to the same quality schools.

So this phenomenon seems to be popping up a lot. Why?

One answer is a catch-up-growth kind of convergence story that you could borrow from macro, but I doubt that's a major role since abilities are fairly well controlled for.

I think the answer probably has to do with differential statistical discrimination in the labor market, as laid out in this excellent paper on racial disparities in unemployment by Ritter and Taylor, which really guided a lot of my thinking in writing up the RBPE paper. White workers - particularly young ones without much work history - have an easier time signalling their productivity to white employers than minorities (the same would go for women signalling their productivity to male employers). Statistical discrimination is never a complete strategy in the hiring process but it's likely to be more heavily relied upon when hiring minority workers. So you can think of getting a degree either as jumping to a higher distribution where employers are still using statistical discrimination, or as the higher tail of the distribution getting a better signal out about their productivity. Education doesn't matter as much for white or male workers because more of their signal was already picked out from the noise.