I've been reading Ira Katznelson's new book, When Affirmative Action Was White, which revisits the affirmative-action debate in a rather novel fashion, by arguing that many New Deal and post-WWII federal policies benefited whites almost exclusively during the 1930s, 40s, and 50s. Hence the title. Programs like the GI Bill, however well-meaning in theory, created in practice "a massive transfer of privilege to white Americans." It's a surprisingly engaging book, all told, and I'd recommend it. Nevertheless, in a Commentary review, Dan Seligman argues that Katznelson's thesis shouldn't imply that "reverse" affirmative action, directed at African-Americans, is a necessary and appropriate response:

A related problem is the considerable amount of time that has elapsed between the grievances Katznelson cites and the future remedy he proposes. How, in particular, are specific injuries to identifiable victims today to be traced back to the 1930's and 40's? In describing the workings of this and other features of his program—an exercise relegated to the book’s final three pages—Katznelson becomes noticeably blurry. One possibility, he suggests, would be grants to individuals whose parents or grandparents were harmed.

A "less administratively burdensome" possibility would be "assertive federal policies" designed to help those "poor Americans who face conditions produced by the constellation of patterns of eligibility and administration the South placed inside the most important New Deal and Fair Deal programs." "Constellation of patterns"? This seems a long way from specificity.

This sort of argument gets tossed around a lot. A related complaint, equally silly, is the idea that, say, a Korean immigrant "fresh off the boat" shouldn't have to pay for injustices committed by mostly-dead white Americans. Well, true. But that's not the point! No policy on earth, as far as I know, could possible remedy past injuries to specific and identifiable people; it's just too difficult. Katznelson's framing this all wrong. The choice isn't between atoning for past wrongs and failing to do so (or doing so in an equally unjust manner). The choice here is between a world in which whites have a disproportionate advantage in life, and a world in which that advantage is somewhat lessened. As Katznelson shows, past policies have created the first type of world; race-based affirmative action, insofar as it works, aims for the second. It's not perfect by any means, but unless someone can come up with a better alternative, it's the best option available.

(And as far as I can tell, income-based affirmative action just isn't always a good substitute. Even if racism didn't exist anymore, just note that among families making under $15,000 a year, roughly 40 percent are black while 17 percent are white. But the wealth disparity between races is considerable: whites with incomes under $15,000 have an average net worth of $63,836; blacks have an average net worth of $16,152, putting the former in a much better position to take advantage of income-based affirmative action. [Wealth-based AA, in turn, seems harder to pull off.] See this post, and this paper, for the statistics. On the other hand, Ross Douthat makes the point that elite colleges, for instance, could stand to attract more low-income students of all colors. Very true. Again, I would never say race-based AA is perfect, but it seems to do better than the alternatives.)