Tuesday, January 10, 2012

Skipping Evidence in Favor of Conclusions

Tonight's Chronicle of Higher Education features a story of great policy relevance. Under the headline "Study Disputes Claims That Preferentially Admitted Students Catch Up," author Peter Schmidt describes the results of an unpublished paper by Duke researchers as calling "into question other studies that play down the academic difficulties initially experienced by the beneficiaries of race-conscious admissions." The paper, Schmidt says, has been marshaled by critics of affirmative action as they seek a Supreme Court ruling knocking the policy down.

My own reading of the paper is that drawing such conclusions from this work, and highlighting them with such an inflammatory headline ("preferentially admitted students"?) is grossly inappropriate. While the authors document (1) racial/ethnic variation in the relationship between initial academic preparation and later major-switching and (2) that major-switching accounts for the diminishing racial/ethnic gap in GPA during college, there is absolutely no evidence presented that the use of race in college admissions is the driving factor behind such switches.

But I can see where the reporter got such ideas. The authors frame the paper from page 1 in terms of the debate over racial preferences. There were other options, including framing it in terms of unpacking the significant fluctuation in students' choices of college majors over time (e.g. as documented by Jerry Jacobs in Revolving Doors), and the implications of that fluctuation for labor market outcomes. Or they could have thought in terms of the debates over student learning, and how this may help us understand racial differences in rates of learning, such as those laid out in Academically Adrift. Instead, the authors open with a discussion of how affirmative action is "promoting access" to the "less prepared,"suggesting that such admits would need to "catch up" over time. This is the language of affirmative action critics, not researchers who recognize that the literature on testing and admissions hardly indicates that students who are admitted through the use of some form of preference may not be less prepared at all, and thus have no reason to catch up.

The first finding presented in the paper is that even though they start "behind" white students at Duke (behind as proxied by their first semester GPA, which is hardly solely a function of intelligence or preparation), black and Hispanic students make steady gains in GPA over time. One could look at this and think many things-- for example, perhaps over time minority students learn the "system" and figure out how to reap its rewards. (For the record, even white students do this, and have for decades--read about the "grade point average perspective" in Howard Becker's classic Making the Grade.) Or maybe black students benefit disproportionately from mentoring and other attention on campus. There are many possible explanations, but the authors turn to two in particular--variance and course selection. And in the end, they hang their hats on major migration--black students catch up on their GPA by switching to easier majors, and this is because they are disproportionately weaker students. Controlling for switching explains almost all of the black/white convergence in grades, say the authors. And so, they write, "Attempts to increase representation [of minorities] at elite universities through the use of affirmative action may come at a cost of perpetuating under-representation of blacks in the natural sciences and engineering." There you have it: we oppose affirmative action because it hurts black people.

Ok, let's say it together: COME ON. There are no alternative explanations proffered by these very fine academics (one of whom trained at UW-Madison)? And seriously, issuing such a policy proclamation based on a study of one single, highly unusual private university?

It is easy to come up with many, and next to impossible for the authors to rule them out. So black students have lower first year GPAs than white students, and this explains why they'd be more likely to switch majors over time. This can be true even WITHOUT affirmative action. Only if we hold a very high bar that says "we will only admit students above this bar so that they will not change majors" would this be stopped. Moreover, have the authors considered the potential that Duke's affirmative action program is the problem-- not affirmative action itself? As Doug Massey and his colleagues have pointed out, programs of all kinds that admit students but fail to adequately support their success are bad programs-- period. That does not mean the admissions practice itself is bad-- it means the university is hypocritical. It wants the glory of claiming diversity without devoting sufficient resources and effort required to ensure that all who are admitted fully succeed.

These are but a few conjectures- and deciding on which is right is not my job. My point is straightforward: this paper begins and ends with a political agenda, and it's being used as such. There is no more reason to think the patterns observed are attributed to the use of race in Duke's admissions decisions than to discriminatory practices in the allocation of university resources to its students, or to inadequate teacher quality on the part of its professors in some majors. Who knows the reason? Certainly not these researchers.