An excellent op-ed here.
The connections he draws between (1.) discrimination, (2.) inequality, and (3.) mobility are very important.
This has cropped up in a lot of the racial disparities work I've done at Urban. People are always shocked when they find fairly high racial parity in a controlled setting. For example, one project I was on looked at racial differences in the transition to adulthood and some important variables we looked at had to do with "risky behavior". Drugs, teen sex, fighting, carrying weapons, etc. There wasn't any significant difference between black and white youth on these points when we controlled for youth characteristics. More recently (in a brief that will be coming out soon) when we looked at black males in the Great Reccession we found that given a certain educational level employment outcomes were relatively close (wages may not have been - we did not look at those).
I'm not saying this happens everywhere or even that the conditional differences were exactly the same in all of our cases. But it often surprises people that these things can be close.
But that conditional difference really just tells you about is discrimination, bias, or unequal treatment for one reason or another on a particular indicator. Given a set of characteristics, what is the disparity, in other words? What drives a lot of the disparities in society is that often blacks have much less of this "set of characteristics" that we're controlling for than whites do. So we found that controlling for parental education, family structure, and income, blacks and whites are about as likely to be arrested. The problem is black youth are much more likely to have parents with less education, come from an unstable family structure, and grow up in poverty. The real factor driving a lot of later racial disparities in the U.S. is that too damned many black kids grow up poor.
That's why talking about inequality and mobility (the way Krugman does) is so important for getting a full sense of racial disparity in the country. Simply talking about discrimination only gives you a partial answer. It doesn't matter if a young black man with a college degree and a young white man with a college degree get treated comparably in the labor market (they don't, but let's say they do). That doesn't matter if considerably fewer young black men get college degrees in the first place.
Conditional differences only get you so far. This is a major problem with how a lot of people trained in economics think about disparities. We're too quick to just toss things into a regression. The problem is that we don't live in a counter-factual world. We can't hold everything else constant. And in the case of racial disparities, what we can't hold constant matters a lot.