it hurts, but it might not be the truth

I would like to get in an argument with Bryan Caplan.

However, Caplan is a smart guy, so to really come after him, you better bring the data. Bringing the data takes time, and time for blogging is currently limited to the time that it takes my SPSS syntax to compile. And even that is time I should be using to get my grad school applications in order. So I'll voice my worry, rather than undertake the data project it would take to dispute Caplan's argument.

Caplan is criticizing Tim Hartford's latest book, The Logic of Life, on the issue of statistical discrimination. I haven't read Mr. Hartford's book (for reasons mentioned above), so I have no intention of rescuing any specific point of his.

The part that I do want to take issue with is Caplan's claim that there is a sizable return on education for blacks. While I'm sure that this broad claim is accurate, I think there may be something to the distribution of this return that undermines Caplan's point. (I must admit I'm not entirely sure what his larger point is, he seems to be tiptoeing around something politically incorrect).

What gives me pause? I haven't been delving into compensation practices of various companies for an incredibly long time, but those companies that I have looked into have had some interesting features. If you divide job titles into quintiles based on their average salary and then look at what races make up what share of the resulting quintiles you get something like the chart below.

While blacks are well represented at the top most quintile, they are less well represented in the middle three. Now, a chart that looks like this is certainly not evidence of discrimination in a company, but it does tell us if there is discrimination, where to look for it. The place to look would be in those middle salary brackets, not the top.

What's that got to do with education? Well, jobs at the top part of the income distribution take a lot of education (MBAs, Masters, PHDs) and pay a lot too. But those middle three quintile jobs are of the kind that usually just take a four year-degree or experience. If my hunch is right, there is little payoff for blacks at medium levels of education, the kind attainable by most middle and lower middle class Americans. If you were told when you started school as an undergrad that it probably wouldn't pay off unless you stayed in school for at least six years, don't you think that would effect your decision to continue your education?

Now for all I know, Caplan has already thought of this and used a nonlinear model to do his regression. But he doesn't say so. If he wants to convince me, (and as an EEO professional I think I'm exactly who he should want to convince) then Caplan should look at this return to education at different levels of income.