Michigan’s Mangled Empirical Claims in the Schuette Affirmative Action Case

October 23, 2013
Guest Post
by William Kidder, the Assistant Provost at UC Riverside. Mr. Kidder’s post represents his personal views and not necessarily those of the UC administration. Mr. Kidder has a book review of Mismatch forthcoming in the Texas Law Review and a policy brief on the impact of Michigan’s Proposal forthcoming through the UCLA Civil Rights Project. His article on California’s Proposition 209 was published last spring in the Journal of College and University Law.
In discussing scientific evidence, Justice Breyer articulated a bare minimum standard that judges need to meet in order to protect the public interest and parties to litigation: “Consider the remark made by physicist Wolfgang Pauli. After a colleague asked whether a certain scientific paper was wrong, Pauli replied, ‘That paper isn’t even good enough to be wrong.’ Our objective is to avoid legal decisions that reflect that paper’s so-called science.  The law must seek decisions that fall within the boundaries of scientifically sound knowledge.”  
Regrettably, in last week’s oral argument in Schuette v. Coalition to Defend Affirmative Action, there were times where Michigan’s solicitor general John Bursch advanced arguments about higher education that aren’t “even good enough to be wrong” and that, if accepted, would leave the Court in a wilderness outside of the boundaries of sound statistics and social science knowledge. 
In response to Justice Sotomayor’s question about the impact of California’s affirmative action ban, Bursch claimed, “The statistics in California across the 17 campuses in the University of California system show that today the underrepresented minority percentage is better on 16 out of those 17 campuses.  It’s not at Berkeley; they haven’t gotten there yet; but its better on the rest.” There are only ten UC campuses, not seventeen.  Of these campuses nine UC campuses (and eight with undergraduates) permit a comparison of pre- and post-affirmative action periods.
In contextualizing Bursch’s claims, it is also helpful to address African Americans, American Indians and Latinos separately. Most directly responsive to Bursch’s dubious claim are total enrollment figures (though that means combining undergraduate, graduate and professional school students). Compared to a baseline of 1996 (before California’s affirmative action took effect), the percentage of African Americans in 2012 is lower on a majority of UC campuses: Berkeley, Davis, Los Angeles, San Diego and San Francisco. The negative impact of prohibiting affirmative action is greatest at the most selective campuses that disproportionately train future leaders.  At UC Berkeley African Americans were 5.1 percent of students in 1996 and only 3.3 percent in 2012, fifteen years after the campus implemented myriad rigorous race-neutral efforts to improve diversity. At UCLA African Americans were 5.8 percent of total enrollment in 1996, and 3.9 percent in 2012.
With respect to American-Indian students, the percentage enrolled at the University of California was lower in 2012 than in 1996 on seven of nine campuses: Berkeley, Davis, Los Angeles, Riverside, San Diego, San Francisco, and Santa Barbara.  In summary, even with the remarkable expansion of 72,000 seats (and a new campus) at UC during this span, the percentage of African- American and American-Indian students enrolled in the UC system was still lower in 2012 than it was in 1996.
Moreover, in law and medicine Proposition 209 turned back the clock on several decades of progress: The percentage of African Americans at UC Law Schools was twice as large in 1976 compared to 2012 (7.7 percent v. 4.0 percent) and the percentage of African Americans in UC medicine and health science programs was nearly one-third larger in 1976 compared to 2012 (6.1 percent v. 4.2 percent).
The percentage of Latinos enrolled at UC Berkeley and UCSF was lower in 2012 than in 1996.  While it is true that at seven UC campuses the percentage of Latinos was higher in 2012 than in 1996, as was noted in the Schuette amici brief by the UC President and Chancellors, “[T]hese bare figures must be considered in light of the explosive growth in California’s Latino student population … from 1995 to 2012, the percentage of public high-school graduates who were Latino jumped from 30.0 to 46.2 percent.” 
In light of all the above, no wonder then that several Schuette amici, including a social science brief by the UCLA Civil Rights Project, a group of UC faculty admissions experts and the UC President/Chancellors all conclude that energetic efforts at race-neutral alternatives have not been enough to offset the negative impact of California’s Proposition 209.
Michigan’s arguments about graduation rates and “mismatch” are likewise not “even good enough to be wrong” when judged against the preponderance of peer-reviewed scholarship. Bursch claimed that underrepresented minority students at UC “have a graduation rate that is 20 to 25 percent higher than it was before California’s Proposition 209.”  This claim relies upon a paper by UCLA law professor Richard Sander as well as the Mismatch book by Sander and Stuart Taylor Jr., a book referenced in Chief Justice Roberts’ remarks.  
Here petitioner relies upon a six-page unpublished “paper” by Sander that makes claims repeated in the Mismatch book released through a non-academic publishing house. As Professor Onwuachi-Willig and I note in a forthcoming piece for the Texas Law Review, Sander’s book and unpublished paper provide a “apples-to-oranges” analysis of UC graduation rates because summer graduates are included in the post-209 data but excluded from many of the pre-209 years. 
More fundamentally, because the UC system was becoming more selective over time, African Americans’ graduation rates at UC were already climbing significantly in the years before Proposition 209 and would most certainly have continued to do so even if the affirmative action ban had never come about. The very data Sander and Petitioner cite show that UC’s African American freshmen six-year graduation rate rose nine points between 1992 to 1997 and then rose by a less impressive four additional points in an equivalent span after the affirmative action ban (1998 to 2003). Likewise, for Latinos’ UC graduation rate improved five points before Proposition 209 and only a couple more points after Proposition 209.  The amici brief by the UC president and chancellors points out petitioner’s argument represent a “textbook example of the dangers of mistaking correlation for causation.”
The deeper point is that the weight of peer-reviewed studies undercut the claims by the Petitioner and by Sander and Taylor. The evidence outside California cited in the Mismatch book for the proposition that affirmative action lowers graduation rates consists of two stale studies using 1970s datasets.  By contrast, a large body of contemporary studies (that employ a range of empirical methods) concludes that affirmative action at highly selective universities tends to enhance the probability that black and Latino students will earn bachelor’s degrees. The literature on this point has grown from Bowen and Bok’s seminal book The Shape of the River (1998) to include Alon and Tienda (2005), Small and Winship (2007), Fisher and Massey (2007), Melguizo (2008), Espenshade and Radford (2009), Bowen, Chingos and McPherson (2009), Cortes (2010), Backes (2012), Kurlaender and Grodsky (2013) and others. (See this single source, which includes many of the above-referenced articles.)
Though petitioner was conspicuously silent regarding Michigan graduation rates, recent federal data on the six-year graduation rates for the 2005 entering class show that at the University of Michigan Ann Arbor African Americans (78 percent) and Latinos (88 percent) had higher graduation rates with affirmative action than did the African Americans and Latinos at UC without affirmative action, including UC’s most selective campuses of Berkeley (71 percent, 81 percent) and UCLA (76 percent, 85 percent). 
Finally, the cumulative weight of the social science evidence similarly disfavors the mismatch hypothesis at the law school level.  In last year’s Fisher v. University of Texas case, a preeminent group of eleven social scientists, including a couple members of the National Academy of Sciences who are leading methodologists, concludes that Sander’s mismatch scholarship “fails to satisfy the basic standards of good empirical social-science research” and that the Sander and Taylor brief in Fisher “misrepresents the acceptance of [Sander’s] hypothesis in the social-science community and, ultimately, the validity of mismatch.”  (See also AERA article links above.)