I have a slightly kindlier view of affirmative action than Becker does, though I agree with most of his points: specifically, that affirmative action harms the ablest beneficiaries of it by casting doubt on their ability; that it places many of them in situations in which they are bound to fail or cluster at the bottom because they have been admitted to a school or given a job that is above their level; and that remedial education is unlikely to be effective after early childhood.
But there are three areas in which preferences that are often (though not necessarily correctly) described as affirmative action seems to me defensible:
1. Situations in which race, sex, ethnicity, etc. is a legitimate job qualification. An obvious case is casting a black man as Othello and a white woman as Desdemona in a performance of Othello. A subtler case is making sure that in a prison or jail the vast majority of whose inmates are black there are some black correctional officers in supervisory positions; this is important for alleviating racial tensions. (See my opinion in Wittmer v. Peters, 87 F.3d 916 (7th Cir. 1996).) I would go further and say that if an all-male prison wanted to hire just male guards, or an all-female prison just female guards, I would permit this, although the courts disagree.
2. Situations in which a private firm or other private entity practices affirmative action in response to customer preference or to ward off adverse legal action. A firm that sells primarily to blacks might want to give a preference to black applicants for sales positions or insist that its advertising agencies include black models in their advertisements for the firm's products. Similarly, if a firm fears that it will be sued for discriminating against blacks, it should be allowed to favor blacks in hiring in order to reduce its legal risks.
These are situations in which affirmative action is prima facie efficient because it is being adopted voluntarily by a private, competitive institution, presumably as a profit-maximizing, cost-minimizing response to competitive pressures. (Of course not all private institutions are commercial, but the noncommercial ones, universities for example, do face competitive pressures and do need to economize.) I don't think government should interfere with such choices.
This by the way is not to say that firms controlled by blacks, say, should be permitted to discriminate against whites. That would not be affirmative action, which refers instead to discrimination in favor of the group that controls the discriminator.
3. Situations in which the only beneficiaries of affirmative action are black. Most of my examples of affirmative action have involved blacks rather than women, Hispanics, etc. My reason for that choice is that realism requires recognition that blacks are, for whatever reason or combination of reasons, in far the worst position, so far as health, prosperity, educational achievement, intermarriage, and other measures of success and integration, of any other major group in American society. Women, Jews, Asians, and other traditional victims of discrimination or newcomers or outsiders have all advanced to positions of essential parity with male WASPs, but blacks have lagged badly in relative terms. A situation in which 12 percent of the population is lagging badly behind the rest of the population is not healthy. I don't think affirmative action for blacks does much to promote their integration and sense of belonging in this society, but it probably does a little (notwithstanding Becker's correct point about the negative effect of affirmative action on self-esteem). Without affirmative action, elite educational institutions and other elite institutions (probably including the officer corps of the military) would have virtually no blacks, and this would underscore the gulf in achievement in a dramatic way that would be potentially harmful to social peace. Colin Powell was a beneficiary of affirmative action and his well-deserved public success and prominence is probably good for black morale.
Category 3 is perhaps the only "real" affirmative-action category. Such a classification would be consistent with Becker's treatment.
What is unfortunate is that although the only real case for affirmative action (outside my first two categories) concerns blacks, naturally other groups, seeing the potential benefits of discrimination in their favor, have climbed on the affirmative action bandwagon, often with ludicrous results, notably in the case of white, well-to-do, accentless, fully integrated Americans of Hispanic ancestry. I well remember a conference I attended at which a law professor named Delgado advocated affirmative action for persons of color, among whom he counted himself. However, as if often true of persons of Spanish ancestry, his black hair crowned a very pale face. He was in fact the whitest person in the room, had no foreign accent, and by his presence had in effect converted "people of color" into a purely political category.
So I would like to see affirmative action confined and diminished, but I would not press for its abolition. I especially would not favor judicial abolition in the name of equal protection of the laws. Not that a powerful legal case can't be made; but it seems odd that the courts should strain to intervene on the side of majority rights, since the majority should be able to protect its interests in the democratic political process, without having to run to the courts. In addition, it is doubtful that the courts could effectuate a ban on affirmative action. As Becker points out, states have tried and failed, since such a ban can be circumvented in a variety of ways, such as by reducing or eliminating the weight given to meritocratic criteria in selecting for colleges or jobs.
One final point. Given my category 2 above, the problematic area of affirmative action is largely confined to government employment, public universities, public contracting, and other government activities. In many of those activities, personnel practices are not meritocratic, but political, nepotistic, or simply slack and inefficient because of lack of economic incentives. If criteria are not meritocratic to begin with, injecting affirmative action may not reduce efficiency. I infer that the aggregate social costs of affirmative action probably are not great--though neither are the benefits.