Testing for 'Mismatch'

If members of some minority groups are admitted to elite colleges because of affirmative action -- and don't perform as well as they expected -- does this show a serious flaw in efforts to diversify student bodies?

Critics of affirmative action answer in the affirmative, and this is the basis of the controversial "mismatch" theory -- namely that affirmative action doesn't actually help its intended beneficiaries because they may struggle academically where admitted instead of enrolling at less competitive institutions where they might excel. Mismatch is heatedly debated -- in part because of the political potency of the argument. After all, it allows critics of affirmative action to say that they aren't just worried about white applicants, but about black and Latino students, too.

In a paper released Friday, four scholars at Duke University (three in economics and one in sociology) propose a new way to test for mismatch. They say that much more information is needed than has typically been available in the past. But because they were able to obtain this information for Duke, they argue that a mismatch test is possible. They propose a test in which applicants admitted to an elite university are asked to predict their first-year grades and are then told the average grades earned by members of similar ethnic and racial groups admitted under similar circumstances. In this situation, they argue, students admitted under affirmative action could make an informed judgment on whether they were being mismatched.

The data released by the scholars in explaining their idea could be quite controversial. Private colleges and universities historically release very little information, broken down by race and ethnicity, about the admissions qualifications and subsequent performance of students. Getting even SAT averages by race can be difficult. Duke provided the researchers not only with SAT averages, but with admissions officers' average rankings of admitted students on a five-point scale, by race, as well as the students' own projected first-year grades and actual grades.

Generally, the data show that Asian admitted students had better rankings and scores than all other groups, although their advantage over white students was modest. But Asian and white applicants are generally far above other applicants. And while all groups, on average, overestimated their academic performance in their first year at Duke, black and Latino students had the largest gaps between the performance they expected and what they achieved.

The study, "Does Affirmative Action Lead to Mismatch," is by Peter Arcidiacono, Esteban M. Aucejo, Hanming Fang and Kenneth I. Spenner, and was released by the National Bureau of Economic Research. (An abstract is available here, as is information on how to download the study for $5.)

Via e-mail, the researchers collectively answered questions about their study, and said that they are trying to change the way mismatch theory is measured. Some of the early, controversial studies have used statistics such as graduation rates or passage rates on bar exams.

The Duke researchers argue for a broader examination of mismatch, including some sense of students' actual perspective. And they say that existing data -- even the detailed information from Duke -- is not enough to demonstrate that mismatch exists or doesn't.

"As is amply clear from the introduction and the literature review in our paper, the literature on mismatch is very important, but still in very early stage," the Duke scholars said. "Researchers are not even settled on what 'mismatch' should mean, let alone making firm statements about whether there is 'mismatch.' As you read through our paper, you will notice that we take a very nuanced view about what should 'mismatch' mean: We think it should be measured by whether the minority students are made worse off, in utility terms, by attending elite universities. Since utility can not and is not captured by any single outcomes, in our formulation mismatch should not be judged from any single outcomes."

Asked if they are generally supporters or critics of affirmative action, the scholars replied: "We are neither supporters nor critics (or in a sense, both critics and supporters) of affirmative action. We, as social scientists, have only one goal: contribute to the understanding of the implications of affirmative action in admissions practices by proposing new concepts, new tools and new data sets. We firmly believe that higher education and its stakeholders are best served if we understand the process."

The data on which the paper is based come both from Duke admissions and from the university's Campus Life and Learning Project, which surveys cohorts of students on a range of issues. This gave the Duke researchers an unusual mix of data -- including both student views of their academic performance, actual performance and admissions analysis.

Here are the data used by the scholars from a recent Duke class cohort to show the gaps in admissions rankings, test scores, students' own projected grades, actual grades and other factors. The admissions rankings are based on a five-point scale used by the staff to evaluate applicants, and cover assessments of applicants academic achievement, quality of their curriculum and various other factors.

Duke Admissions and Academic Performance Statistics by Race and Ethnicity

The authors of the study offer two ways to determine if mismatch is real and a problem. One would be for the admissions office to conduct an experiment in which admitted minority students are divided into two random groups. One group would receive the standard letter. The other would receive information related to the admissions officers' rankings and possible impact on post-enrollment academic performance. "If we observe that the enrollment rate for the second group is smaller than the first group, this will prove that the university's private information may have generated mismatch," the paper says.

A second test the authors offer would involve asking admitted students what they expect their grade-point average to be their first year, and then telling them what it is likely to be.

The authors stress that this information Duke has doesn't mean that there is mismatch, and that it can actually be used so that entering students know what they getting into and make informed decisions. "To the extent that a university with active affirmative action programs is concerned about potential mismatch, it suggests that releasing more information to their applicants about how the admission officers feel about their fit with the university will minimize possibilities for actual mismatch," the paper says. "More transparency and more effective communication with the students, and possibly pre-enrollment sit-ins in college classrooms, etc. can help minority students enrolling an an elite university potentially find out that they would have been better off elsewhere."

A key point stressed by the authors is that their views are theirs as scholars, and do not represent Duke's opinions.

Christoph Guttentag, dean of undergraduate admissions, said he has yet to read the paper, but he noted that Duke's graduation rates are extremely high across ethnic and racial groups.

"I'd say that our ratings system is a tool -- no more and no less -- for gathering, organizing, and interpreting some of the information we use in the admissions decision. There are many additional factors that we consider in making our decisions, which is why we read and discuss each individual application rather than simply let a rating system determine the decision," he said.

As for potential mismatch, he said: "We believe that every student we admit has the preparation and attitude to be fully successful at Duke. We certainly don't admit anyone about whom we have doubts."