Affirmative Action: The Good, The Bad (And the Ugly?)

The Chronicle of Education today reports on two studies (still unpublished) that were presented this week at a meeting of American Association for the Advancement of Science.

Douglas S. Massey, a professor of sociology and public affairs at Princeton, and Mary J. Fischer, an assistant professor of sociology at the University of Connecticut,

examined longitudinal student data from 28 selective colleges in an attempt to determine whether any evidence supported two of the most common criticisms of race-conscious admissions policies. Those are the “mismatch hypothesis,” which holds that such policies result in the admission of students who find themselves in over their heads academically, and the “stereotype-threat hypothesis,” which holds that such policies stigmatize all minority students as academically subpar, thereby placing them under a form of psychological pressure that undermines their academic performance.

The two researchers sought to gauge how much weight each college gave to applicants’ race or ethnicity by examining the difference between the average SAT scores of all students that they enrolled as freshmen in the fall of 1999 and the average SAT scores of the black and Hispanic members of that entering class. To try to measure how much of a role a particular student’s race or ethnicity played in his or her admission, the researchers looked at the difference between that person’s SAT score and the average for the entering class. (On average, black students’ SAT scores were 131 points below the average for all students at the 28 colleges, while Hispanic students’ SAT scores were 76 points below.)

The results, at least as summarized here, seem inconsistent.

The study found that those black and Hispanic students who had seemed to get the biggest break in admission actually tended to have slightly higher grade-point averages than other students, and were much less likely than other students to leave college. Their level of satisfaction with college was about the same as that of other students….

When all black and Hispanic students at an institution were examined collectively, however, evidence of “stereotype threat” emerged. The more a college used affirmative action, the lower were the grade-point averages of its minority students, and the more likely such students were to leave college and express dissatisfaction with their college experience. The negative correlation between a college’s commitment to affirmative action and the grade-point averages of its black and Hispanic students grew stronger the longer the students were in college, suggesting that the effects of “stereotype threat” mounted as the students became more accustomed to the campus culture.

I’m not sure what this means, but Prof. Massey seems to have no doubts.

“Affirmative-action programs don’t set minority students up to fail,” Mr. Massey said on Monday in an interview….

Mr. Massey said that, on balance, the positive effects of affirmative action on minority students outweighed the negative. Moreover, he said, colleges have found ways to counter the effects of “stereotype threat,” by, for example, hiring more minority faculty members.

You’ll not be surprised when I say that I have a few questions, although let me say again that my questions and concerns may well be addressed in the complete study, which I have not seen.

First, as I’ve had occasion to observe here before, comparing “the average SAT scores of all students that … enrolled as freshmen in the fall of 1999 and the average SAT scores of the black and Hispanic members of that entering class” minimizes the degree of preference awarded because the blacks and Hispanics were themselves members of the entering class. The more revealing comparison would have been between the scores of the blacks and Hispanics and the scores of those students (whites, Asians, others) not awarded any preference.

Second, if the grades of individual preferees increased with the amount of preference they received, why would the composite grades of all preferees go down as the amount of preference the school extended went up?

Third, how does Prof. Massey know that the explanation for the lower grades is “stereotype threat”?

Finally, on what evidence does Prof. Massey rely when he concludes “that, on balance, the positive effects of affirmative action on minority students outweighed the negative”? Did he find these “positive effects” in his study, or is it a conclusion not based on his survey results? Does he count the cost of compromising, or undermining, the principle that people should be treated without regard to race as a negative effect?

The second study, by Marta Tienda, a professor of sociology and public policy at Princeton, and Sunny Niu, a research associate in Princeton’s Office of Population Research, found that black and Hispanic students who attended integrated high schools in Texas were less likely to graduate in the top 10% of their class than those who attended minority-dominated schools, and that they were less likely to attend a selective college.

The researchers found, however, that the differences between high schools disappeared when the economic conditions of the students were taken into account, suggesting that it is the effects of high concentrations of poverty in minority high schools, rather than racism, that keeps many of their top graduates from enrolling in selective colleges.

I’m beginning to think that one of the strongest arguments for getting rid of racial preference is to be done with studies of it.

Say What? (14)

  1. Kimberly February 22, 2005 at 3:57 pm | | Reply

    One thing I find odd is the assumption (presumably made by the researchers) that all minority students with below-mean SAT scores were admitted due to racial preferences. Certainly, they could have had other factors to balance out their application – which would explain the higher college GPA for some of them.

  2. anon February 22, 2005 at 4:51 pm | | Reply

    Given Massey is not known as an idiot and has surely taken sociological methods, I suspect he knows damn good and well how invalid it is to compare the mean scores of blacks and Hispanics to a group (the overall mean) that includes…blacks and Hispanics. Surely it would be good to have those categories be mutually exclusive. I suspect (shock) he is not interested in seeking truth.

    Of course, it would be helpful to see the study (and it seems wrong of the Chronicle to give the results when the things hasn’t been peer-reviewed and published) but I wonder why he is switching units of analysis–from looking at individuals at the bottom of the AA pile to looking at schools and their overall minority results? There could be a good reason, but I’m wondering if there isn’t a statistical trick he is using, for example perhaps he is “controlling” for race and/or socio-economic status in looking at the success of the lowest-scorers on the individual level part. Then at the school-level he can drum up the “stereotype threat” to counter the arguement that minorities do worse the more AA there is. I would love to get my hands on that data and the study.

    Just love the statement that the “stereotype threat” is countered by having more minority faculty members. Is this hiring by AA, or just by actively seeking qualified minority professors? Guess it’s just Massey’s way of promoting his political agenda at a higher level as well.

  3. Gabriel Rossman February 22, 2005 at 7:07 pm | | Reply

    I’m a sociologist at Princeton. I don’t work with Doug and I haven’t read the study, but I can give my take based on John’s description. The short version is that it sounds like legitimate research, but I’ll go through point by point.

    — Why not use the white/Asian SAT as a baseline?

    The study isn’t about the magnitude of racial preference, but about the impact of racial preference. Statistics is about angles, not positions, so across the board cuts to a variable don’t matter. Since white/Asian SAT and population SAT are not the same but highly correlated, recalibrating the magnitude of preference against the former probably wouldn’t change anything. I agree that the way they did it has its downsides, but I think it is theoretically defensible.

    — Statistical tricks.

    I’d have to read the paper to be sure, but it sounds like everything they did is totally mainstream. It sounds like they used some kind of multilevel analysis, a common and universally accepted technique. It can be sensitive to colinearity between the levels of analysis, but I’d doubt that’s a big problem here. As for the idea that they were controlling for race and class, I don’t know if they did, but I think it would be appropriate to do so since the mismatch hypothesis is not that poor or black students do poorly, it’s that underqualified students do poorly.

    — Where did he get the idea that it’s stereotype threat?

    From the contextual level analysis. I think it’s actually a pretty clever way to estimate it. The argument would basically go that the less qualified are the minorities at school X, the lower opinion of them faculty and peers will have and the greater the stress minorities at school X will endure as a result.

    You could concievably interpret the contextual effects other ways, for instance, racial homophily in study partners (especially if there are segregated dorms or fraternities). Still, stereotype threat is a plausible interpretation of the results.

    — How does he know that minority faculty members counter stereotype threat?

    This sounds like he was speculating, though in principle it’s a testable hypothesis.

    — How does he know that AA is good overall?

    By good overall, he means good for the proximate outcomes of underrepresented minorities. He is not docking points for the moral dubiousness of race consciousness, or a lot of other negative externalities. As for how does he know it’s good in the limited way he’s defining “good,” yes, this is a numbers thing. You can simply combine the micro effect and the macro effect to get the net effect.

    — Why is the Chronicle publicizing a non-peer-reviewed paper?

    Most conferences are peer-reviewed, albeit not as strictly as journals. It is very common in the social sciences to consider conference papers as newsworthy.

    — Where can I get my hands on the paper and data?

    Usually all you have to do to get a paper is send a polite email asking for it. Sometimes people aren’t ready to release a paper, but they often are.

    On the other hand, the data is harder to get. If they used the data I’m thinking of, it’s very closely held — partly for the privacy of both the schools and the students.

    — You suspect that Massey is not interested in the truth

    I know a fair number of partisan hacks (no I won’t identify them), but Doug isn’t one of them. In my experience, he’s open-minded and professional. I simply don’t believe he would fudge his results to get a palatable finding.

  4. retrofuturistic February 22, 2005 at 8:22 pm | | Reply

    Am I correctly interpreting your comment correctly about Professor Tienda’s “research”? (Yes, I am using the ol’ sarcasm quotes there.)

    Is her point realy that minority students who attend a minority-majority are more likely to graduate in the Top 10% of their class than those who attend fully-integrated high schools? If a school is 100% African-American, then the entire Top 10% will be African-American, by definition. If your school is 30% African American and say, by way of example, that academic performance is randomly distributed throughout the student population, then 30% of the Top 10% should be African American. And that would be substantially less than the guaranteed dominance of minorities in a majority-minority school. If I’m understanding Professor Tienda’s research correctly, she has merely stated a truism. Without delving into the incomprehensive academic-ese in which she has likely gussied-up her sociological findings, I don’t understand what possible break-through point she believes she is making.

    And her research is based on Texas high schools, right? Texas has a state law guaranteeing college admission for the Top 10% of each and every high school graduating class. That would seem to be an important detail and I wondering if Professor Tienda willfully opted to ignore it (or if it influenced her choice of Texas as the state to be studied). So, of course, you could go to a more “selective” school in Texas merely by one’s status as a graduate in the Top 10% of one’s class in a Texas high school, which, in a minority-dominated Texas high school would most assuredly be a minority.

    Years ago, I began suspecting that there may be a problem with affirmative action when I realized that none of its defenders could speak openly and honestly about the costs and benefits. (And if one did, you would only need to see the example of Larry Summers to see what happens when you speak truth to power about affirmative action.) Like someone justifying remaining in a bad relationship, if you have to lie to justify your action or your policy choice, chances are your policy choices (like relationship choices) are a terrible mistake.

    But, then again, I’m just an old-fashioned liberal who is still naive enough to believe the old saw about equal justice under the law for everybody.

  5. John Rosenberg February 22, 2005 at 9:10 pm | | Reply

    Gabriel – I will happily defer to your expertise in this area, and I recognize that some of your comments are directed to other commenters and not to me. Still, I’ll respond to a couple of your points:

    — Where did he get the idea that it’s stereotype threat?

    From the contextual level analysis. I think it’s actually a pretty clever way to estimate it. The argument would basically go that the less qualified are the minorities at school X, the lower opinion of them faculty and peers will have and the greater the stress minorities at school X will endure as a result. You could concievably interpret the contextual effects other ways, for instance, racial homophily in study partners (especially if there are segregated dorms or fraternities). Still, stereotype threat is a plausible interpretation of the results.

    This still seems like a stretch. If less qualified students (as determined by their distance below the average entering SAT scores) do less well, why assume that “stereotype threat” is the explanation? If something in the study itself points to this as the explanation, that’s another matter, but in the absence of that something it seems safer to assume that less qualified students do less well because they are less qualified.

    — You suspect that Massey is not interested in the truth

    I sujspect that this comment is not aimed at me, but let me nevertheless affirm that I certainly didn’t mean to imply that. What I suspect is that some of Prof. Massey’s conclusions — that on balance affirmative action is good; that stereotype threat explains poor performance — were not derived from “the numbers” in his study. This suspicion (and, not having seen the study, it is only that) doesn’t mean I think he doesn’t care about the truth. It means that some of his opinions quoted in the article sound like they are just that, opinions, and not statements of research results. That doesn’t mean they are wrong, untrue, or in any way illegitimate. But it would mean they don’t necessarily have the authority of science behind them.

  6. Gabriel Rossman February 22, 2005 at 10:08 pm | | Reply


    If they simply found that black students do poorly, I’d agree that it would be premature to call this a stereotype threat effect. But the paper found a macro penalty, not a micro penalty. Basically, they’re claiming that blacks do relatively poorly at schools with a large skill gap between groups, regardless of whether the individual student is underqualified. So it’s not that “less qualified students” do less well, it’s that students from less qualified groups do less well.

    It helps to think of an example. Let’s say Joe College is black and has an SAT of 1500. He’s admitted to school A and school B, each of which has an average SAT scores of 1500 so he’s as well qualified as anybody else there. At school A the average black student has an SAT of 1450 and at school B the average black student has an SAT of 1300. At school A, black students are pretty close to average so everyone respects them and if Joe went there he wouldn’t encounter any stereotypes so he’d do pretty well there. At school B though there’s a big skill gap so faculty and classmates don’t really respect blacks so Joe would encounter a lot of stereotypes that would hinder his performance even though as an individual he’s as competent as anyone else.

    So basically what the paper is calling sterotype threat is the “stigmatizes qualified minorities” argument by another name. The reason that they still conclude that AA is good for minorities is because these macro effects are swamped by the individual level effects (which go exactly opposite of the mismatch hypothesis). I’m not sure exactly why this is, but my guess would be that “mismatched” students are in more selective colleges and these schools provide much better outcomes. (And no, I do not think they should be controlling for school selectivity since the policy question they’re addressing is whether one is better off at the best school or the best fitting school).

    I didn’t mean to imply that you were calling Massey a liar and I apologize for the lack of clarity. I was responding to the anonymous poster on that point. In fact, I agree with your position that his numbers are honest but that in parts he probably speculated beyond the specific findings of his study. (Although I should add that there is nothing improper about this so long as one brackets one’s suspicions and opinions from one’s findings and as far as I can tell, Doug more or less stuck to this).

  7. Gabriel Rossman February 22, 2005 at 10:25 pm | | Reply


    It’s pretty silly to say Tienda ignored the 10% law since it’s not just in the paper, it’s in the TITLE, “Capitalizing on Segregation, Pretending Neutrality: College Admissions and the Texas Top 10% Law.”

    As I understand it, Hopwood and the ensuing 10% law is exactly what got her interested in studying the state’s education, plus it probably doesn’t hurt that she’s a Texan herself.

    You can get the paper from the project’s webpage.

  8. John Rosenberg February 22, 2005 at 10:48 pm | | Reply

    Gabriel – As usual I find your exposition lucid and compelling, and it encourages me to look forward to reading the study itself.

    If I understand what you (explicating Massey) are saying, it is that 1500 SAT Joe College will work on a par with his peers if he attends a college where the SAT average of the entering class, and the entering blacks and Hispanics, is in the same neighborhood as his 1500, but if attends a college where the black and Hispanic average is significantly below the class average he won’t do as well.

    By contrast, as i understand what you cum Massey are saying, if Bill College, who scored 1150 on his SAT, attends a college where the entering class average is 1500, he’ll do fine. (Could that be because there is affirmative actin in grading as well as admissions at selective schools?)

    In any event, if this is all true it is indeed interesting.

  9. retrofuturistic February 23, 2005 at 3:25 am | | Reply

    I stand corrected by Mr. Rossman on one of the basic premises on which I have attacked Professor Tienda’s research. However, the conclusion referenced above still seems over-simplistic. In fact, correct me if I’m wrong (and I’m sure you will), but wasn’t the very INTENT of the Texas law to serve as an affirmative action substitute and to give an extra boost to students who attend majority-minority schools which, as a class, at least aming public schools, under-perform in terms of academic performance. Given the structure of the Texas “affirmative action” plan, it was a given that minority students were more likely to finish in the Top 10% of their class in a majority-minority school, since, again, majority-minority schools are, by definition, have minority students as a numerical majority.

    Too much affirmative action researtch is conclusion-driven; i.e., rather than going where the research and open mind leads, the academician will tortute the data until it screams out in favor of affirmative action. Professor Tienda did a backflip or two to reach her conclusion. First, she stated that segregration is responsible for increasing the pool of minority (i.e., African American and Latino) students to go to the flagship schools of the State of Texas. If I understand the Texas plan, its very purpose is to try to squeeze at least one advantage (increasing the pool of African American and Latino students attending flagship universities) out of an otherwise negative situation (the extreme segregation in the Texas piublic school system). She correctly recognizes that there are tremendous costs to attending secondary school in a segregated environment which more than offset this limited advantage. So far so good. But then she goes off the rails. She criticizes the Texas approach as somehow shortchanging minority students from less-segregated environments, saying that these students need traditional “affirmative action” and “aggressive recruitment” (which, really, can’t we admit is just a code word for “quotas”) in order to succeed. The leap that is made is that she appears to conclude that minority students from integrated or majority-majority schools deserve just as much “affirmative action” as their counterparts from segregated schools. This is simply defining “diversity” as skin color without looking at the whole person as to who needs more of a helping hand. The Texas approach is crude, but it least gives the greater helping hand to minority students who have come out of segrtegated schools. And, to me, a minority student from a segregated environment definitely needs more of a helping hand. And, despite her reaching a contrary conclusion, I see nothing in her research to indicate that I am prescribing the wrong medicine for a real problem.

    Nevertheless, I appreciate the fact that you’ve corrected a major error on my part in expressing my concerns with Professor Tienda’s research. Even if you destroyed one of my underlying premises.

  10. Tim Gannon February 23, 2005 at 8:19 am | | Reply

    If the lower ranked students that were admitted using affirmative action did well in college, then why doesn’t the college admit all students with that score?

    OR why doesn’t the college admit ALL students above that score?

    In either case, this study shows that these students should do better than average.

    The point is that the “elite” colleges are not admitting students that would otherwise do well at their school. If that is the case then either the school is too easy, or the standards for admission are too high.

  11. Gabriel Rossman February 23, 2005 at 11:31 am | | Reply


    I believe your interpretation of the results is correct. Massey sent me his paper, so I can now make more precise examples.

    Recall the assumptions. School A and school B both have student mean SATs of 1500. At school A the black mean is 1450 and at school B it’s 1300. Joe and Bill are both typical black students in every way (ie, assume their control variable traits are all at the black mean). The only difference is that Joe has a 1500 SAT and Bill has a 1300 SAT.

    Based on the regression equation we would predict the following cumulative GPAs and probabilities of graduating:

    Joe @ A — 2.95 — 82%

    Joe @ B — 2.85 — 66%

    Bill @ A — 3.05 — 96%

    Bill @ B — 2.95 — 92%

    It doesn’t make sense that Bill would do better than Joe, but I’m pretty sure that’s what the model is predicting. My hunch is that the reason is that individual AA (school SAT – ego SAT) is capturing school selectivity. In other words, Bill probably wouldn’t get into school A and Joe probably would have gone to Harvard instead of school A. Even though the individual examples are a little screwy, I still think the paper is pretty good evidence that the mismatch hypothesis is wrong (at least for undergraduate education) since any mismatch effect gets swamped by the dramatically better outcomes at selective schools.

    This matches my personal experience. I went to college at UCLA and I’ve TA’d at Princeton and the difference is night and day. At moderately selective state schools you’re more or less a number whereas at top tier private schools they go to extraordinary lengths to keep you on the right track. I had a Bill College type kid in one of my courses and when he ran into problems the dean called and worked out a plan to make sure he could get back on track. (He still earned his grade, just w an extension). I’m certain that at UCLA, Bill would have just dropped out even though his credentials would have been a better fit there.

    I don’t think there’s much affirmative action in grading. It could be general grade inflation though, so looking at class rank instead of GPA may produce different results. In my opinion, the more important outcome is retention, and there’s no alternate definition of retention that could change the results.

  12. Jason February 25, 2005 at 10:47 am | | Reply

    Great comments on this thread! Mine will pale in comparison but here goes…

    Perhaps an alternative explanation for the “stereotype threat” would be social pressure that african-americans place on each other to not “act white.” I assume that this pressure is less at the graduate school level than at the undergraduate level, but is unlikely to be completely eliminated.

    If the african-american students at a school have credentials that match the general student population, then there won’t be a problem. Doing well won’t be seen as “acting white” but will be seen as typical performance of the group.

    However, if the majority of african-american students enter the school with skills significantly below those of the general population, then perhaps they will pressure each other to not stand out from the group through superior performance.

    This could lead the higher-credentialed african-american students to reduce their effort level so as to not risk alienation from their peer group.

  13. “Diversity,” Princeton-Style November 6, 2012 at 2:20 pm |

    […] sensitive” admissions. A few of my criticisms of Massey’s work(s) can be found here, here, here, and here (see especially the third and fourth links). Posted in And another […]

  14. […] few of my criticisms of Massey’s work(s) can be found here, here, here, and here (see especially the third and fourth links). 00   Read the […]

Say What?