Does affirmative action help the disadvantaged students it is supposed to benefit?

The Supremes are going to take up an affirmative action (AA) suit. One argument against AA may seem to be the result of benevolent concern for under-represented groups. It employs mismatch theory.

“Mismatch theory” says that AA clearly gets some students into better schools than they are otherwise qualified for. But such students are then at a disadvantage, since those who get in on merit will out perform them and leave them at a loss in class after class.

As Dan Slater in the NY Times says:

[Mismatch] is the idea that affirmative action can harm those it’s supposed to help by placing them at schools in which they fall below the median level of ability and therefore have a tough time. As a consequence, the argument goes, these students suffer learningwise and, later, careerwise.

And Slater seems to think a 1991 study of the law bar is significant, though very imperfect:

In 1991, more than 27,000 incoming law students — about 2,000 of them black — completed questionnaires for the B.P.S. (Bar Passing Study)and gave permission to track their performance in law school and later on the bar.

Among other things, the questionnaire asked students (a) whether they got into their first-choice law school, (b) if so, whether they enrolled at their first choice, and (c) if not, why not.

Data showed that 689 of the approximately 2,000 black applicants got into their first-choice law school. About three-quarters of those 689 matriculated at their first choice. The remaining quarter opted instead for their second-choice school, often for financial or geographic reasons. So, of the 689 black applicants who got into their first choice, 512 went, and the rest, 177, attended their second choice, presumably a less prestigious institution.

Those who went to their second choice schools did significantly better.

Duke researchers have weighed in and they argue that it is STEM fields that mismatch causes problems. Stem courses build on expertise acquired in earlier classes, and problems can multiply in very serious ways.

As far as I know, this is pretty much the whole mismatch argument, made famous in fact by Clarence Thomas. I’m really interested in hearing what you think of it. I’m going to restrict myself to pointing out that the mismatch argument proponents don’t seem very concerned about all the others who get into universities on something other than academic merit. For starters, legacy students and athletes are among them. Perhaps also famous actors, members of royal families, the children of presidents, and so on. And since the worry is that students who benefit from AA will be below the median, shouldn’t we worry about all the other students there too?

8 thoughts on “Does affirmative action help the disadvantaged students it is supposed to benefit?

  1. AA action is intended to improve outcomes and if it doesn’t that is cause for concern. The “beneficiaries” of AA are being deceived, led down a path that isn’t the best for them.

    The consequences of mismatch to the rich and powerful are their own business.

  2. Interesting that the ‘median’ is the cut off point. I was below the median in my year, and it was a culture shock. But I am not a particualrly disadvantaged group and I still found it quite tough (after being top in my tiny school).

  3. If the point is to impugn the motives of AA opponents, then sure, point out that athletes and legacies are in the same boat.

    I personally think universities *should* curtail their preferential treatment of athletes and legacies, so I’m not all that interested in the analogy.

  4. The questions of (1) whether a mismatch effect exists, and (2) if so, how strong it is (and in particular, whether it is strong enough to outweigh competing benefits of affirmative action), don’t strike me as questions on which it’s reasonable to have strong a priori leanings one way or the other, nor do they strike me as questions that must be answered in the same way in every area. That is, we shouldn’t rule out ex ante that mismatch theorists are right about law school, but that there’s no similar, comparably large effect for undergraduates. (At the very least, undergrads who graduate with a poor gpa still have a credential that is useful in a way that law students who graduate but don’t pass the bar do not.)

    Also, I think it’s a bit unfair to say that *Slater* seems to think a 1991 study is very important; from what I gather from reading the piece, the work he’s writing about (starting with a 2005 stanford law review article, and then continuing with responses to that article, and responses to responses) has discussed the study a lot. If Slater’s piece hadn’t given it a prominent position, he’d have been misrepresenting the debate.

    As for the point that “mismatch proponents don’t seem very concerned about all the others who get into universities on something other than academic merit”, I’ll ask you what you asked Hamish: could you give any evidence supporting your claim? I don’t know what Richard Sander’s views on legacy admissions and athletes are (he’s the one who wrote the 2005 paper that seems to have sparked the academic debate about the mismatch effect) but my impression is that lots of opponents of racial affirmative action are also proponents of narrowly merit-based criteria for admissions. That is, they oppose preferences for legacy students as much as for racial minorities. Whether this tends to be true specifically of believers in a mismatch effect or not, I don’t know.

    It seems to me that there are good empirical questions here, and that impugning the motives of mismatch theorists, rather than rolling up one’s sleeves and doing the social science necessary to determine the answers to questions (1) and (2)–or if one is not a social scientist, acknowledging that the answers to those questions depends on an ongoing empirical debate–is a distraction.

  5. Based on the Slater piece — and especially the quotation — these seem to be the premises of the mismatch theory argument:

    (1) Students admitted with the benefit of AA are expected to fall below the median level of ability.
    (2) Students who fall below the median level of ability are expected to “have a tough time.”
    (3) Students who have a tough time are expected to learn less.
    (4) Students who have a tough time are expected to “suffer careerwise.”
    (5) Students who are expected to learn less and (or?) suffer careerwise are harmed.

    Right off the bat, one logical weakness of the argument is that the four empirical premises 1-4 are statistical generalizations.* That A is correlated with B and B is correlated with C does not imply that A is correlated with C. So then we can’t support this argument by empirically investigating each of premises 1-4 on its own. But it also means we can’t go empirically investigate one of the subconclusions on its own (say, students admitted with the benefit of AA are expected to learn less, as the 1991 law bar study does), since this ignores the mechanism that (purportedly) connects cause to effect. In other words, the 1991 study gives some support to one of the implications of mismatch theory but fails to investigate the explanation given by mismatch theory.

    * That’s assuming that “level of ability” can be measured and “harm,” in premise 5, cannot.

    Next, exactly who are we comparing in premise 5? A graduate of Prestigious U law who just barely passed the bar exam might have learned less and suffer careerwise compared to the median fellow Prestigious U law graduate. But she will probably still be better off, in terms of lifetime wealth, than a graduate of Regional U law; and much better off than a graduate of For-Profit Technical College. The Prestigious U alumni network by itself will give her opportunities that her Regional U counterpart won’t have. The 1991 study, since it compares “first choice” with “second choice,” seems to be comparing students who went to Prestigious U with those who went to Slightly Less Prestigious U.

    Finally, I would challenge the individualistic assumptions about well-being used by mismatch theory. The aim of AA, at least as I would like to defend it, is to redress the deeply historical oppression suffered by a class of people. (Using “class” in the broad sense that includes race, gender, sexual orientation, and so on.) It is not to benefit individual members of that class, at least as individuals. So instead of “Does this policy help or harm these individuals?,” we should be asking “Does this policy help or harm this historically oppressed class?” (Unfortunately, the Regents v. Bakke decision rejected this kind of rationale for AA.)

  6. In response to Dan,

    My impression is that the data the mismatch theorists use is a bit more concrete than you seem to be suggesting. I thought it was all about bar passage rates (granted, this is about law school–as far as I know the only area where there’s been any serious study of this), rather than more generic stuff about learning and “suffering careerwise”. Now suppose the mismatch theorists were arguing like this:

    (1) law students who fall well below the median at their law school in undergrad gpa/lsat tend to have lower grades in law school
    (2) law students with lower grades in law school tend to pass the bar at lower rates, so
    (3) law students who fall well below the median level at their school in undergrad gpa/lsat tend to pass the bar at lower rates.

    This would be a logically invalid argument, for the reason you point out; because probabilistic correlation isn’t transitive, you can’t logically prove that A is correlated with C by showing that A is correlated with B and that B is correlated with C. But (i) it wasn’t clear to me that this was the (whole) strategy mismatch theorists were using. There are other ways to look at the effects of racial preferences in law school admissions on bar passage rates. E.g., looking at how bar passage rates in california changed after proposition 209 passed. From doing some searching in the article that started the debate I can see that proposition 209 is discussed, though I don’t know what conclusions about it are drawn. But also, (ii) while probabilistic correlation isn’t logically guaranteed to be transitive, I think it’s often a reasonable methodological presupposition to expect it to be, in the absence of special evidence to the contrary. If A is positively correlated with B and B is positively correlated with C, but A is not positively correlated with C, then there must be some special explanation of this. E.g., if the transitivity of probabilistic correlation fails in this particular case, then whatever factor leads students to generally pass the bar at lower rates when they have lower grades, must not be operating for members of racial minorities, or it must be counterbalanced by some factor leading members of racial minorities who are below the median at their law school in undergrad gpa/lsat to be more likely to pass the bar than people with similar grades in law school. That’s certainly possible, but the fact that it is a logical possibility doesn’t mean that establishing (1) and (2) in the argument above would be irrelevant to (3)–it would at least be the sort of thing that would suggest that further research ought to be done.

    As for your 5, I think you’re totally right to ask this question, but I also think that mismatch theorists (or at least Sander) do consider it. After the original post made me curious, I did some googling, and found a series of blog posts where Sander expands on the article. One thing he said, that I found interesting, was that his conclusions were actually different for the top tier of law schools than for less prestigious law schools. E.g., his research suggested (at least to him–obviously it’s controversial) that a student who was admitted to Harvard law, but who without affirmative action, would only have gotten into, e.g., UCLA, would still do better career-wise (in terms of bar passage and likelihood of employment after law school) to go to Harvard. At schools like Harvard and UCLA, bar passage rates are so high to start with that almost nobody is seriously at risk of failing the bar (even if they’re well below the median in terms of grades), but Harvard students do have extra advantages in terms of networks/prestige. It was only when you went a bit further down the hierarchy of law schools that the mismatch effect was significant.

    Lastly, while I think I’m sympathetic to what you say in your final paragraph, I’m not sure how you’re imagining affirmative action in law school might help some oppressed minority as a class, even while not systematically benefiting individual members of that class. I could see how this might be if students being admitted to better law schools, but then not going on to pass the bar, was still beneficial for some minority as a class (even if not beneficial for those individual students), but I’m not sure why this would be.

  7. In a sense, I’m not sure what to say about the mismatch argument as presented in the OP. It seems to rest almost entirely on empirical issues. I mean, I can say at least this much: mismatch seems to be an issue with some athletes. I’ve had a number of moneymaking-sports athletes (e.g. football, men’s basketball, etc.) in my courses who simply lacked the skills to succeed at a four year public research university. Proportionally, more moneymaking-sports athletes seem to fall in this camp than students who either play other sports or who are not athletes.

    Aren’t we simply lacking the data to say for sure whether this is a general feature of AA programs?

Comments are closed.