See also The Wall Street Journal, The Evidence of Things Not Seen, May 16, 2003.
Michigan Bar Journal An Economic Alternative to Affirmative Action, May, 2003.

First Amendment Issues Raised by U-M Admininstration Statements on "New" Admissions Essays: Must be "risky, but not too risky."

by Chetly Zarko
ANN ARBOR, MI
October 29, 2003
On September 17, 2003 a public panel discussion was convened by the University of Michigan administration. The panel participants were all University administrators or researchers who have taken public positions in favor of race preferences. They made various statements that call into question the real nature of change since June, raise first amendment questions related to the weighting of the new essay admissions system, and lead to questions about veracity of previous statements by the same individuals.
The panel, titled Affirmative Action Rulings:  How Will They Impact Our Campus was recently replayed on local cable. The videotapes provide a veritable goldmine of priceless quotes.  The author attended this panel in September after requesting in writing that the University balance the discussion by including some of the known opponents of race preferences, but was told the purpose of the panel was strictly "informational." (Perhaps even more reason to include an alternative viewpoint) The information was designed to address issues related to the legal decisions and the August 28th announcement of a new "holistic" admissions policy relying heavily on essay questions related to a prospective student's "diversity contribution."
Theodore Spencer Director, Undergraduate Admission, had a remarkable thing to say when asked by an audience member later in the session to more precisely define what the evaluative criteria in the new "holistic" process was.  "We will look for them [essay writers] to take a bit of risk, but we don't want them to take too much risk to where they can offend the reader or someone else.  We will look for all those kinds of things that the U still feels are important ... The only thing we've done ... we've rearranged them so that ... they are no longer given points ..."  So there you have it, a prima facie admission that the content of the essay will be judged on unconstitutional grounds ("offensiveness"), and that the point system's values have not been changed ("the only thing we've done ... rearranged them") except to take off the quantifications.  This statement was perhaps the most remarkable of the evening.
Earlier in the session, Dean of Students and co-moderator Edward Willis kicked off the question and answer session with a prepared question to each panelist.  Willis began with a question to Professor Emerita Patricia Gurin, in clear reference to an opinion analysis this author had written for The Wall Street Journal on May 16th. That piece had criticized Gurin for making rosy claims about a dataset she used, when her husband had co-authored an executive summary 5 years earlier reporting a very 'mixed bag' of conclusions, including that racial tension actually increased on campus over a student's four years, that blacks self-segregated, and that numerical diversity alone had no effect (without 'diversity classes'). The Journal piece also criticized U-M for withholding under FOIA the raw data that would have proved or disproved these concerns conclusively, and that scientists had been consistently refused access to this raw data.
Willis asked, "Some opponents" say that the Michigan Student Study data [that she used in her 1999 expert testimony to the Court in Grutter] show "increased racial tension.  Are these arguments valid?"  Gurin, obviously expecting the question, immediately replied, almost with a laugh, "I want to make five points ... quickly." I remember thinking at the time that this was a bit of overkill, but then again, the charges against her research from many angles are quite serious. Gurin continued,  "the claim of increased tension ... has no way to be tested.  We don't have a baseline, before there was any race-conscious admissions policy. We have no real evidence about there was increased racial tension because race conscious policies or not."  Certainly, Gurin is quite correct about the lack of a baseline to compare against non racial admissions policies, but this criticism also affects the question of whether race conscious admissions have educational benefits, meaning that her expert testimony assertions had no baseline.  When you attack your critics, you must be careful not to do so with a double-edged sword. Of course, the dataset provides itself with a weak internal baseline, by comparing the freshman, sophomore, and senior year's questions (all of whom experienced the ongoing U-M system of diversity); and the hidden result never reported in Gurin's expert testimony is that perceived racial tension on campus actually increased from freshman to senior year. Her better defense could have been her second point, something U-M has stuck to since May.  She argued that the data show a difference between student perceptions of campus climate and students perceptions of their own relationships.  She admits that 25% (and increasing over time) of white students perceive campus tension, whereas only 1-7% (how precise!) of the same students perceive racial tension in their own lives.  Here's the kicker, and its a new insight into the study and Gurin's beliefs about it.  Gurin says the campus tension does not come racial diversity, comes from "it probably""graffiti" and the "media", but not from students interacting.  How does she know?  "We've tried to understand where the tension comes from ... none of the regressions work" in finding the answer.  This is a significant revelation.  Gurin just admitted that she and her group looked, and looked hard, for an answer to the troublesome data.  They ran regressions, none of which were reported to the Court or peer reviewers. They just couldn't find it.  So she's left with a "probably ... graffiti and the media", (this of course all ignores the failure to account for set response and attribution bias which could explain why perceptions of external reality and personal reality differ, but that is a more technical concern) but most significantly, she never reported the regressions that don't work to the Court.
An unexplained variable is certainly something the judge and world was entitled to know about. Given the new concession that she looked and experimented but couldn't find an answer, and never reported these regressions in her expert testimony; this author, now, for the first time, will make the assertion that Gurin committed academic fraud as defined by her profession's standards. I believe that Ms. Gurin's motivations may been sub-conscious, or that she firmly believed she was in the right, but very few people think of themselves as ill-motivated (the essence of attribution bias, BTW). Scientific integrity requires that one report the known weaknesses in one's own work.  
Gurin's third, fourth, and fifth points all seem to blend into the same argument, that we should expect tension when we put different people together, and that the tension is perhaps good in creating an opportunity to overcome obstacles and pre-existing systemic differences (this "societal benefit" is far afield from the notion that there are "educational benefits" in diversity). This integrationist argument is correct of course - there is a benefit from the conflict diversity creates - but Gurin's husband concluded based upon the same dataset that African Americans were actually more likely to "self-segregate," and other University policies ("multicultural lounges" designed for specific groups, funding race-specific group activities, etc.) increase this segregation.