The Volokh Conspiracy

On Friday I discussed a body of research – all of it uncontroverted – that documents a serious flaw in affirmative action programs pursued by elite colleges. Students who receive large preferences and arrive on campus hoping to major in STEM fields (e.g., Science, Technology, Engineering and Math) tend to migrate out of those fields at very high rates, or, if they remain in those fields, often either fail to graduate or graduate with very low GPAs. There is thus a strong tension between receiving a large admissions preference to a more elite school, and one’s ability to pursue a STEM career.

Is it possible for contemporary American universities to engage constructively with this type of research? Recent events at Duke University suggest not.

The Duke study I described in Friday’s post (by economists Peter Arcidiacono and Esteban Aucejo, and by sociologist Ken Spenner, all of Duke) was motivated by an important question: do students who receive large admissions preferences “catch up” with their peers over their college years? This ties into an important premise of many preference programs – i.e., that the rich resources of an elite university will help to phase out prior preparation gaps between students of different races. Aggregate data at Duke suggested that the GPA gap across racial groups was, indeed, narrowing as college progressed, from over half-a-point black/white GPA gap in the first semester, to less than three-tenths of a point by the eighth semester.

Using data gathered by the university, Arcidiacono et al found that this narrowing was illusory. Courses taken by juniors and seniors were graded very leniently, and, more importantly, students who had bad grades in their freshmen year migrated in large numbers from STEM fields and economics to other majors, which generally had easier grading. When one adjusted for these effects, the relative achievement level of different groups was unchanged over the course of college. Thus, there was no silver lining to offset the science mismatch effect.

Importantly, the authors found that these patterns had nothing to do with race, but rather with a student’s level of academic preparation upon entry into Duke. White legacies admitted with large preferences showed the same patterns as blacks admitted with large preferences.

The paper offered no policy recommendations; like a large body of Arcidiacono’s earlier research on other social and educational issues, it simply presented intriguing results researched and analyzed in a conceptually clear and empirically careful way.

In mid-January 2012, the Chronicle of Higher Education ran a story on the article. Although the reporter, Peter Schmidt, was characteristically fair in summarizing the article’s findings, once the news reached Duke, the reaction was extreme. The Black Student Alliance denounced the research and staged a protest, suggesting that the research was actually an attack on black students and that data they had provided to the university had been misused. Seventeen black alumni wrote an open letter attacking the research as “misguided scholarship” whose results and methodology were “both flawed and incorrect”, though they provided no specifics. “We cannot sit idly by and allow this slander to be (mis)labeled as truth.” Duke faculty got into the act as well, sending angry, indignant emails to the authors and to the economics department.

The President of Duke, Richard Brodhead, finally weighed in on the controversy on March 22nd, at the Annual Meeting of University Faculty. He said he had decided to devote his talk to the issue of race in part because of the controversy generated by the study. He extolled the university’s progress in moving from exclusionary policies in the 1950s and before, to today having among the highest proportion of enrolled blacks of any elite university. He then went on:

“With respect to this January’s controversy I would say the following. I hope all members of this community recognize that it is not the proper function of the university to block expression from its faculty or enforce a correct view. Universities live through free and open debate; when someone thinks someone else has come to an erroneous conclusion, the remedy is to criticize it and offer a better account. On the other hand, I can see why students took offense at what was reported of a professor’s work. Generalizations about academic choices by racial category can renew the primal insult of the world we are trying to leave behind – the implication that persons can be known through a group identity that associates them with inferior powers. A further insult was that the paper had been included in an amicus brief submitted by opponents of affirmative action urging the Supreme Court to hear [Fisher v. University of Texas]….”

Brodhead’s remarks neatly stood reality on its head. The university’s policy of giving large preferences based on race had created a large academic preparation gap across racial lines (e.g., an average 150-point SAT gap, on the old 1600-point scale, between blacks and whites) and thus large differences in academic outcomes across racial lines; but careful research on the effect of academic preparation on these outcomes was offensive? Academic freedom was vital to the university’s life, but factually baseless slander against accurate research was understandable? And it was especially “insulting” to use such research in an amicus brief – i.e., a debate about public policy?

(As it happens, I know about the amicus brief mentioned by President Brodhead, because I coauthored the brief with Stuart Taylor. Both of us are, to be sure, critics of affirmative action, but neither of us are “opponents”, as I will discuss in a coming post. We cited Arcidiacono et al’s research in the brief pretty much in the same spirit that I discussed it in Friday’s post.)

Brodhead’s message was pretty clear: we won’t try to fire people who engage in honest research that identifies problems in affirmative action; but we will ostracize them, and thus strongly discourage such research. Other parts of the record suggest that Duke’s substantive response to the controversy will consist of providing additional funding to race-based student groups, and showing greater “sensitivity” to student complaints.

One might be tempted to put this behavior down to a particularly high level of intolerance at Duke or on Brodhead’s part (many Duke officials and faculty, including Brodhead, took political correctness to disgraceful lengths during the “lacrosse” scandal several years ago, when a number of white students were falsely accused of raping a black woman and Duke officials led the invidious attacks against them, even long after the prosecution had been discredited). But all of the facts of this latest episode at Duke, including Brodhead’s behavior, actually capture perfectly the dynamics of affirmative action discussions at all major universities.

Colleges and universities are committed to the mythology that diversity happens merely because they want it and put resources into it, and that all admitted students arrive with all the prerequisites necessary to flourish in any way they choose. Administrators work hard to conceal the actual differences in academic preparation that almost invariably accompany the aggressive use of preferences. Any research that documents the operation and effects of affirmative action therefore violates this “color-blind” mythology and accompanying norms; minority students are upset, correctly realizing that either the research is wrong or that administrators have misled them. In this scenario, administrators invariably resort to the same strategy: dismiss the research without actually lying about it; reassure the students that the researchers are misguided, but that the university can’t actually punish the researchers because of “academic freedom”. Note that in this dynamic, “academic freedom” becomes a device to protect the administration, not the faculty doing the research!

Imagine if Brodhead had said something like this. “Arcidiacono, Aucejo, and Spenner have done exemplary research here. Their data is accurate (after all, they got it from us) and other experts tell me that their analysis is correct on every point. Importantly, they have demonstrated that race itself does not affect outcomes at Duke, validating our efforts to create an environment where discrimination is either absent altogether, or so minimal as to not affect academic performance. But they have also demonstrated that relative levels of academic preparation do matter, not for every individual but in a way that is somewhat predictable when we look across large numbers of students. They matter not only for grades, but perhaps for learning itself, and apparently they have an influence on one’s choice of major, and chances of successfully sticking with that major. This merits further research, but assuming these findings hold up, they suggest three things that we in the administration should be doing.”

“First, we should provide academic support counseling for students who wish to pursue STEM fields but enter Duke with lower-than-average levels of academic preparation; through careful and rigorously evaluated support efforts, we may be able to improve successful persistence in these fields. Second, we should work with our sister institutions to evaluate how students who struggle at Duke would have fared at other schools, and we should all pool data to evaluate post-graduate outcomes, so that we better understand how the tradeoffs that students encounter at Duke affect their long-term success and career goals. And third, as we learn more, we should make those findings transparent to our applicants. If a high school senior we choose to admit hopes to become a physicist, we should help that student have the best possible information about his actual prospects at Duke, so that she can weigh the advantages and disadvantages of a Duke degree. This transparency is not only fairer to our students, but I can think of no better way of maintaining accountability among Duke’s administrators and faculty.”

Words in this vein would be both moral and intellectually honest. Why, exactly, do university leaders find it impossible to say them?

Some varied sources on the events at Duke:

Peter Arcidiacono’s blog

A representative critique of the research

Commentary by KC Johnson, keen observer of political correctness at Duke

Some of the most significant recent work on affirmative action concerns a phenomenon called “science mismatch”. The idea behind science mismatch is very intuitive: if you are a high school senior interested in becoming, for example, a chemist, you may seriously harm your chances of success by attending a school where most of the other would-be chemists have stronger academic preparation than you do. Professors will tend to pitch their class at the median student, not you; and if you struggle or fall behind in the first semester of inorganic chemistry, you will be in even worse shape in the second semester, and in very serious trouble when you hit organic chemistry. You are likely to get bad grades and to either transfer out of chemistry or fail to graduate altogether.

This idea was  first advanced by Dartmouth psychologist Rogers Elliott (and coauthors) in 1996, and using data from several Ivy League schools, he demonstrated that, indeed, attrition rates from the sciences were highly associated with comparatively lower academic preparation, which in turn was highly associated with receiving an admissions preference. His data suggested that a given student was far more likely to achieve a science degree if she attended a school where her pre-college credentials were close to the median science student.

Virginia psychologists Frederick Smyth and John McArdle provided an even stronger demonstration of these points in a 2004 article. Making use of the same data Bowen & Bok used in Shape of the River, they were able to compare similar students who were interested in “STEM” fields (an acronym for science, technology, engineering and math), and who attended schools with either similar peers, somewhat more prepared peers, or much more prepared peers. Smyth and McArdle found strong evidence of science mismatch. Among their key conclusions: had all the black and Hispanic students in their sample enrolled at schools where their credentials were close to the class-wide averages, then 45% more of the women minorities, and 35% more of the men minorities, would have completed STEM degrees.

Duke economists Peter Arcidiacono, Esteban Aucejo, and Ken Spenner last year completed a study that looked at a number of ways that differences in admissions standards at Duke affected academic outcomes. In one of many useful analyses they did, they found that 54% of black men at Duke who, as freshmen, had been interested in STEM fields or economics, had switched out of those fields before graduation; the comparative rate for white men was 8%. Importantly, they found that “these cross-race differences in switching patterns can be fully explained by differences in academic background.” In other words, preferences – not race – was the culprit.

In research conducted by FTC economist Marc Luppino and me, using data from the University of California, we have found important peer effects and mismatch effects that affect students of all races; our results show that one’s chances of completing a science degree fall sharply, at a given level of academic preparation, as one attends more and more elite schools within the UC system. At Berkeley, there is a seven-fold difference in STEM degree completion between students with high and low pre-college credentials.

As is always the case with affirmative action, ironies abound. Although young blacks are about one-seventh as likely as young whites to eventually earn a Ph.D. in STEM fields, academically strong blacks in high school are more likely than similar whites to aspire to science careers. And although a U.S. Civil Rights Commission report in 2010 documented the “science mismatch” phenomenon in some detail, President Obama’s new initiative to improve the nation’s production of scientists neither recognizes nor addresses mismatch.

Unlike my research on law school mismatch, science mismatch has been the subject of little media attention and hardly any academic dispute. In fact, I am unaware of any research that directly disputes any of the studies described above. (If readers know otherwise, please advise.) In other words, there appears to be a strong consensus on the existence of a mismatch problem among all those who have actually studied the interaction of affirmative action and degrees in STEM fields. Across the country, it is plausible that science mismatch undermines the careers of tens of thousands of promising scientists. Yet I am aware of no university that has either acknowledged the problem or taken direct steps to address it.

Science mismatch is, of course, relevant to the general affirmative action debate in showing that preferences can boomerang on their intended beneficiaries. But it also has a special relevance to Fisher v. University of Texas. The university’s main announced purpose in reintroducing racial preferences in 2004 was to increase “classroom” diversity. The university contended that, even though over a fifth of its undergraduates were black or Hispanic, many classrooms had no underrepresented minorities. It sought to use direct (and very large) racial preferences to increase campus URM numbers and thus increase the number of URMs in classes that lacked them. But science mismatch shows that this strategy, too, can be self-defeating. The larger a university’s preferences, the more likely it is that preferenced students will have trouble competing in STEM fields and other majors that are demanding and grade sternly. These students will tend to drop out of the tough fields and congregate in comparatively less demanding ones. Large preferences, in other words, can increase racial segregation across majors and courses within a university, and thus hurt classroom diversity.

More tomorrow.

As some readers will recall, a little more than seven years ago I published an analysis of law school affirmative action in the Stanford Law Review. The article was the first to present detailed data on the operation and effects of racial preferences in law schools (focusing on blacks).

I also laid out evidence suggesting that large preferences seemed to be worsening black outcomes. I argued that this was plausibly due to a “mismatch effect”; students receiving large preferences (for whatever reason) were likely to find themselves in academic environments where they had to struggle just to keep up; professor instruction would typically be aimed at the “median” student, so students with weaker academic preparation would tend to fall behind, and, even if they did not become discouraged and give up, would tend to learn less than they would have learned in an environment where their level of academic preparation was closer to the class median.

I suggested that the “mismatch effect” could explain as much as half of the black-white gap in first-time bar passage rates (the full gap is thirty to forty percentage points). I also suggested that “mismatch” might so worsen black outcomes that, on net, contemporary affirmative action was not adding to the total number of black lawyers, and might even be lowering the total number of new, licensed black attorneys.

The article generated intense interest, debate, and criticism, though even most critics conceded that I had gotten the facts right. Several well-known empirical scholars in law schools published essays that purported to disprove the mismatch hypothesis. For awhile, many defenders of affirmative action seemed to assume that the article would inevitably provoke a crisis in legal academia, and while attempting to seize the moral high ground in the debate, they attracted even more publicity to the article.

After several months, however, it became clear there would be no widespread calls, among either law students or law faculty, for further inquiry and reform, and things died down. Those unhappy with the “mismatch” article – and that included the vast majority of law school and university administrators – decided the best strategy was to (a) ignore the issue and (b) use their best efforts to prevent the further release of data such as I had used in the original article. There was another, smaller burst of attention when I published a follow-up article  about affirmative action in law firms, and its similar tendency to boomerang on the intended beneficiaries; but otherwise, public debate about mismatch faded away.

It is about to come back.

Over the past few years, there has been a steadily growing stream of empirical research on affirmative action, much of it taking up the mismatch question.  Some social scientists, like Peter Arcidiacono at Duke University and Frederick Smyth at the University of Virginia, were interested in this subject and producing valuable research well before my Stanford article appeared.  Others, like Doug Williams at Sewanee University or Robert Zelnick at Boston University, were intrigued by some of the issues that arose out of the public mismatch debate and the questions raised in the debate.  Still others have been attracted by the “natural experiments” in affirmative action created by the bans on racial preferences adopted in half-a-dozen states.  I have worked closely with Jane Yakowitz (soon to join the law faculty at the University of Arizona) and public-spirited lawyers to pry loose data relevant for studying affirmative action.

Cumulatively, these scholars have produced a remarkable body of research (some of which can be found here) on the workings and effects of affirmative action. And the Supreme Court’s decision (by granting cert to Fisher v. University of Texas) to revisit the subject of racial admissions preferences in higher education will undoubtedly fuel interest in this work. Over the next few days on The Volokh Conpsiracy, I’m going to highlight some of this research and the challenges facing a healthy public discussion of the mismatch issue.