I enjoyed reading this book. It taught me a lot about the role that race and class play in the admissions process of elite colleges in the USA. And like Mr. Blume, i was impressed with the methodological and conceptual thoroughness of the analysis.
However, on a number of issues, the authors seem to overstate the results of their analysis in order to arrive at a conclusion that is friendlier to a pro-Affirmative Action position that seems warranted. For example:
1) In Chapter 6, the authors attempt to evaluate the "mismatch hypothesis", a proposal by critics of AA such as Sander (2004) and Thomas Sowell, which argues that when, thanks to AA programs at elite colleges, African-Americans and Hispanics with low academic preparedness (e.g., low SAT scores, low GPA) are placed in competition against highly-prepared whites and Asians, they will be worse off than if they had attended a less-selective, hence less competitive college, where their class rank would be higher and where they would be more likely to graduate; against the "selectivity" hypothesis of Bowen & Bok (1998), which argues that when academically-underprepared minorities are admitted to elite colleges, whatever costs they may incur because of their AA-based admission are outweighed by the benefits, particularly the notion that they are more likely to graduate from a highly selective college.
In making this comparison, the authors conduct some very thorough regression analyses, and some interesting facts are surfaced. First, the analysis shows that both blacks and hispanics are each about 45% less likely to graduate within 6 years than are whites and asians (p.234, 238). Second, contra Bowen & Bok (1998), once numerous factors are controlled for, there is NO statistically-significant (p < .05) "selectivity" effect. That is, school selectivity is not significantly related to graduation rates. Also, their analysis of class rank effects shows that, as we might expect, black students tend to graduate with a class rank 17 percentile points lower than "otherwise equivalent" white students at the same school (p. 249).
These results are consistent with the "mismatch" hypothesis and contradict the "selectivity" hypothesis. They suggest that under-prepared students, like those admitted by AA, are far less likely to graduate from elite schools, and have significantly lower class rank, and that this is not counter-balanced by higher overall chances of graduating. And yet, on page 239 the authors conclude that "nevertheless, the totality of the evidence...suggests the advantages of school selectivity outweigh the costs ... Under no reasonable interpretation can our results be viewed as supporting the academic mismatch hypothesis" (pp. 237-238).
They also conclude that "Our judgment - based on the available evidence and commonsense considerations - is that, in most instances, the positive effects of school selectivity override the negative consequences of lower class rank" (p.259). They support this conclusion by citing research by other authors about the career-advancement and income and prestige advantages of graduating from an elite university. True, there's little doubt that it's better, from a career perspective, to graduate with a 3.3 GPA and a top-10% class rank from Harvard than to do the same from the University of Maryland. But, is it better to graduate with a 2.4 GPA and bottom-25% class rank from Harvard than the better GPA/rank from Maryland? That's the relevant real-world "mismatch hypothesis" comparison, and the authors have no evidence to counter it. It also stands to reason that any advantages to graduating from Harvard (and other super-elite colleges) only accrue if you actually graduate, and will tend to be greater if your class rank/GPA are higher, and yet their own data shows that blacks/hispanics graduate at a significantly lower rate and with lower class rank. This does support the "mismatch" view, which argues that it's better to actually graduate from, say, the U. of Alabama than to drop out of Harvard.
In summary, most of the evidence generated by the authors in Chapter 6 support the mismatch thesis, and yet the authors reject this in favor of a pro-AA conclusion.
2) The same problem is present in their discussion of the "educational benefits of diversity", the pro-AA claim that swayed the Supreme Court in 2003 to uphold the constitutionality of the use of racial preferences in college admissions. Going in to their analysis, the authors accept that there are such benefits, despite analyses in the literature that suggest that the evidence in favor of such benefits is weak and hardly compelling (e.g., Pidot, Stanford Law Review, 2006). The authors attempt to address this issue by focusing on one question in their data-set, which asks students "how much they have learned from students of different racial and ethnic backgrounds who attended the same college or university as yourself" (p. 308-309). The authors call this a measure of "learning from difference".
To me, there are obvious problems with using this question as an outcome variable. The first is that students may not be qualified to determine what they've learned, much less where they learned it. Usually, in say a classroom setting, that is determined by independent appraisal (e.g., a professor gives an exam to students and then grades it). Also, there is also an obvious possible political-bias to this question: students who read it will likely be able to tell that it is an assessment of diversity initiatives on campus, and their political values, either pro- or anti- AA, could very well influence their response.
Tellingly to me, while the authors recognize this political-bias possibility, they discuss it only as a possible influence on white-student attitudes (i.e., because whites may perceive AA to harm them, this may cause them to understate the amount they learned from difference - see footnote 14, page 310). However, this also would seem to work the other way as well (i.e., because blacks/hispanics may perceive AA to benefit them, they may tend to overstate the amount they actually learned from difference).
Their data show that whites tend to say they learned less from difference (26% say they learned a lot from it) than do blacks (36%) or hispanics (45%). Asians also tended to say they learned a lot from difference more than whites from difference (38% - all on page 310).
In their data analysis of this issue, the authors use 4 diversity-related factors as causes of learning from difference: socializing with members of other races, having an other-race roommate freshman year, having an other-race close friend, and dating someone from another race. The presentation of their results are again skewed in a pro-AA direction. For example, the authors present a prominent bar graph (p. 312) showing that those who engaged in all 4 diversity experiences were much more likely to say they learned a lot from difference, and draw the bold conclusion that "The take-away lesson is evident. When students seize opportunities to interact frequently with .. other-race peers.. and when this interaction occurs on more than a casual or superficial level, positive learning experiences can result. Our findings lend support therefore to claims that there are educational benefits to diversity" (p. 313).
However, their subsequent regression analysis, which is not reported prominently in table form, shows that only 2 of the 4 diversity experiences, dating and frequent socializing, had a significant impact on learning from difference. And of these, let's face it: Dating is likely to produce knowledge about difference that is substantially intimate in nature, and thus perhaps not as relevant to ordinary educational outcomes.
Possibly the worst aspect of this "learning from difference" discussion is that it doesn't confront the fact that *where* something is learned (i.e., from "difference" or "similarity") isn't nearly as important as *what* is learned. The author's own analysis has already shown that, as measured by GPA, chances of graduating, and class rank, blacks and hispanics learn considerably less while at elite universities than whites, despite saying they learned considerably more from difference than whites. And Asians are in between - like whites, they graduate with higher GPAs, etc. but like blacks/hispanics say they learned a lot from difference. These contradictory findings suggest little relationship between "where" and "what", and thus it's not clear how important "learning from difference" is in the scheme of college learning, even if we can take the student's self-appraisal at face value.
And finally, if the benefits of "learning from difference" are alleged to be primarily social, not strictly academic, then it is curious that in the regression analysis findings(p.322), blacks report being 38% LESS satisfied with their campus social experiences than whites- the opposite of what we'd expect given that blacks say they learned more from difference than whites.
Finally, the authors demonstrate some pretty amazing naivety about the relationship between academic preparedness (as measured by things like SAT scores and high school GPA) and the presence of under-represented minorities on elite campuses. In a very interesting analysis, the authors conduct simulations that determine the impact on black/hispanic enrollment under different hypothetical conditions, such as eliminating race-based AA, enhancing class-based AA, etc. One such simulation involves comparing the current situation (blacks/hispanics suffer an achievement gap and colleges have race-based AA) with a simulated scenario in which race-based AA is eliminated, but there is no achievement gap (e.g., on average, blacks/hispanics have the same SAT scores, etc.
Read more ›
