Do differences in ability explain the minority wage gap?

I read a paper the other day about differences in ability and minority wage gaps [pdf]. The author, James Heckman, is a pretty brilliant economist, and he's written a lot on discrimination. Many social leaders don't like his findings - they're controversial - but it is hard to argue with the logic. This paper is pretty light, not even 20 pages of double spaced text, and it has a lot of footnotes. He relies on age−corrected AFQT measures to estimate skill.

In this paper, Heckman finds that adjusting wages to skill explains the wage gap between minorities and whites. The one small caveat is with black males - skill differences can't explain the entire wage differential with black male individuals, so something else is going on here, but skill differences do explain a "substantial fraction of the gap." The conclusion is that affirmative action won't close this gap, we have to close the educational gap earlier in life.

This paper only looks at central tendency statistics. So none of it is relevant for a single individual, but the paper is relevant to populations as a whole. That's a hard thing to hear for a several reasons.

• First, it says minorities have less skill, on average, than whites. That isn't a polite thing for anyone to say, but with the differences in educational attainment, it is difficult to dispute.
  
  
• Second, the resolution essentially gives up on the current generation of minorities, and focuses on helping their children - since "intervention" targeted at adults is far less effective than when targeted at children.
  
  
• Third, this finding suggests that affirmative action and strengthened anti-discrimination laws won't help (with the possible exception of a minor help for black males).
  
  
• Fourth, test scores between 3-4 year old children shows a significant disparity between white children and minorities. Which means intervention is needed earlier.
  
  
• Last, differences persists when controlling for individual, neighborhood (urban or suburban), and family characteristics (e.g. single parent, number of siblings).
  
  

This paper has lots of room for criticism. Some have argued that age−corrected AFQT tests don't measure skill, but instead measure race, wages, and schooling [pdf]. The explanation here is that the AFQT weighs math skills heavier than verbal skills, and minorities do relatively better with verbal skills and whites do relatively better with math skills. Additionally, this argument says the measure overstates the role job skills play in generating wages - e.g. the AFQT underpredicts wages for minorities.

PS: Sorry about the flower bullet points, that choice is out of my hands.