So, dog bites man; what’s new? What’s new is a new study by Princeton researchers, reported this morning in the Chronicle of Higher Edudcation, that purports to find that legacies are more harmed by the preferences they receive than either athletes, on one hand, or blacks and Hispanics on the other.
The article, “Children of Alumni Are Uniquely Harmed by Admissions Preferences, Study Finds,” by the usually reliable Peter Schmidt, begins:
A new study by researchers at Princeton University has found that the children of alumni — commonly known as “legacies” — are far more likely than minority students or athletes to run into academic trouble in college if admissions preferences got them through the door.
The farther a selective college lowers the bar for a given legacy applicant — as measured by the gap between that applicant’s grade-point average and the mean for that institution — the lower the grade-point average that the student is likely to earn, according to a paper written by the two researchers who conducted the study, Douglas S. Massey, a professor of sociology and public affairs, and Margarita Mooney, a postdoctoral fellow in Princeton’s Office of Population Research.
What’s more, those selective colleges that are the most committed to admitting the children of alumni have the highest dropout rates among such students, says the paper, published in the current issue of the journal Social Problems.
The paper says the researchers found that students who had received extra consideration in admission because they are black, Hispanic, or athletes did not have the same academic problems as legacies, as measured by grades or retention rates, even if college policies of giving minority students and athletes extra consideration in admissions appeared to have some drawbacks.
The authors of the study clearly see their results as buttressing the case for racial and ethnic preferences, and one gets the feeling they are not disappointed by what they have found.
“We do not expect these findings to settle the debate on affirmative action,” Mr. Massey and Ms. Mooney wrote. “We do hope, however, that they enable readers to place the issue of minority affirmative action in a broader context, viewing it as just one of several programs to target a subgroup of students affirmatively.”
Perhaps the study itself is unclear, but for whatever reason I found the text of Schmidt’s article itself sometimes hard to follow. For example, I really don’t know what the following means:
In conducting their research, Mr. Massey and Ms. Mooney theorized that college policies of granting preference may affect students at both the individual and the institutional level. In other words, a student may be harmed or helped directly by the admissions preference he or she received, or indirectly through the broader consequences of a college’s policy of giving an admissions edge to certain groups of students. Legacy students appeared vulnerable on both fronts, earning lower grades than other students if they personally benefited from preferences and being more likely to drop out at those colleges that gave legacy applicants the biggest edge.
Still, it is clear that the study paints preferences to legacies in the worst light, with the effect (and perhaps the intent) of making racial preferences seem not so bad.
And yet most of the actual findings presented in the Chronicle article do not seem to me to support the conclusion that preferences to legacies are worse than preferences based on race or ethnicity. For example, consider these results:
The researchers found that the selective colleges gave less of an edge to legacies than either to athletes or to black or Hispanic students. About 70 percent of athletes, 77 percent of black and Hispanic students, and just 48 percent of legacies had SAT scores below their institution’s average. Among those students who appeared to have benefited from extra consideration in admissions, the average bump upward given to a legacy was about 47 SAT points, while athletes and black or Hispanic students appeared to get an edge equal to about 108 SAT points.
On the whole, legacies fared better than the other two populations studied in terms of grades; their mean grade-point average at the end of two years of college was 3.26, compared with 3.12 for athletes and 3.05 for students categorized as black or Hispanic. In terms of their retention rates, legacies were in the middle: 5 percent of athletes, 7 percent of legacies, and 11 percent of black or Hispanic students had dropped out by the end of their junior year.
In other words, there was less of a qualifications gap between legacies and other students than was the case with black or Hispanic students (I wonder if the comparison was to all black and Hispanic students, or only those who were admitted because of the preference they received); legacies received higher grades than black or Hispanic students; and the dropout rate for legacies was significantly lower than for blacks of Hispanics.
So, how are legacies “uniquely harmed” by the preferences they received? There seems, in short, to be something of a disjunction between what this study purports to find, expressed in the title to the Chronicle article, and at least the data that was presented in the article.
The authors of the study are not unaware of what strikes me as this tension between at least some of their data and some of their conclusions. Discussing minority students who attend institutions “with an especially aggressive affirmative-action policy,” they write:
The dropout rates of black and Hispanic students at such institutions were lower than at other colleges, but, on average, those students’ grades were not as high as those of their black and Hispanic peers who went elsewhere.
Mr. Massey and Ms. Mooney said the lower academic performance of students at such colleges provides support for “the social-subversion hypothesis,” which posits that “a large gap between minority students and others at an institution challenges the legitimacy of their presence on campus, thereby creating a social climate within which it is difficult for them to function effectively.”
Gee, is everything a function of “social climate”? Couldn’t the difficulty of students who are admitted to selective institutions with significantly lower academic qualifications than their non-preferred peers be caused by the simpler fact that they were not as prepared to do the work that is expected there?
I have a particular question, which I mentioned only parenthetically in my post, for any of you who may read, or have read, the study (or to the authors if they happen to see this): do M&M compare all legacy admits (52% of whom had qualifications equal to or better than the school average) to all blacks and Hispanics? Or in one or both cases, do they only look at those who would not have been admitted without the preferences? If that would have been too hard, wouldn’t it make more sense to compare the legacies with lower than school-average qualifications (48%) to the blacks and Hispanics with lower than school-average qualifications (77%), since those are the ones who would seem to have benefited most from the preferences. In addition, it would be good to know if there was any difference in the grades/success/failure between cohorts of minorities and legacies who received the same degree of preference, as measured by their distance below the school average for entering test scores/grades.
UPDATE [7 April 2007]
Thanks to a friendly reader who has lent me a copy of the published study, I have now had a chance to read it. I encourage as many of you who have access to it, and the stamina to read it, to do likewise: Douglas Massey and Margarita Mooney, “The Effects of America’s Three Affirmative Action Programs on Academic Performance,” Social Problems, Vol. 54, Issue 1, pp. 99–117.
Let me say first that I am not a social scientist, nor a quantitative anything, and thus I am in no position to judge the extensive and complex calculations that make up the body of the study. I would like to hear from any of you are. I do have some thoughts, however, on how these calculations are, if you’ll pardon the expression, framed, and what they may or may not mean.
In my original post above I cited data the authors reported that showed less of a qualifications gap between legacies and other students than there was between either athletes or minorities and others; that showed legacies had higher grades and lower drop-out rates than athletes or minorities; and thus asked how legacies were “uniquely harmed” by the preferences they received. In short, I mentioned what seems to me “to be something of a disjunction between what this study purports to find … and at least the data that was presented in the article,” and some “tension between at least some of their data and some of their conclusions.” After a close reading, I still think that is true.
It appears even clearer to me now that this study was designed to buttress the case for preferential treatment of minorities by arguing that it is less detrimental to its beneficiaries than is affirmative action for legacies. Let me hasten to add, however, that there’s nothing wrong with this, and I do not mean to imply that Massey’s and Mooney’s data are fudged or that there is anything unprofessional about their study. I do mean to suggest that the answers to the questions they pose, assuming them to be correct, do not support the conclusions they clearly want readers to reach.
Before addressing the distance between the questions asked and the conclusions reached, however, let me also point out that some of the ways the authors describe their project reveals some highly debatable assumptions. On the first page, for example, they define “affirmative action” as coming from
the legal requirement that institutional officials take concrete, identifiable, and positive (in other words, affirmative) steps to include historically excluded groups in their selection pools and to adopt mechanisms that enure their representation among those ultimately chosen.
Really? What legislation contains that “legal requirement”? I’m not aware of any legislation that requires institutions, public or private, to ensure “representation” (how much?) of historically excluded groups. They cite a 1996 book by John Skrentny (The Ironies of Affirmative Action) for this proposition. Although it’s been several years since I read Skrentny (a very interesting book, by the way), and I don’t have a copy at hand, I’d be surprised if he cited legal requirements to take this sort of affirmative action. Certainly the two presidential executive orders, which I’ve cited many times, do no such thing. In fact, they do the opposite, requiring government agencies and contractors to take affirmative steps to ensure that employees and applicants are treated “without regard” to race.
M&M’s goal is to refute what they describe as the “three basic charges against race-sensitive admissions”:
(1) affirmative action constitutes reverse discrimination that lowers the odds of admission for other, better qualified non-minority students; (2) affirmative action creates a mismatch between the skills of minority students and the abilities required for suc- cess at selective institutions, setting up minorities for academic problems; and (3) affirmative action stigmatizes members of the target group as less than fully qualified, which results in demoralization and substandard performance by students in the favored group who may, in fact, be very well qualified.
Note that this list omits the most basic criticism of racial preferences offered by all critics: that awarding burdens and benefits based on race, i.e., violating what Gunnar Myrdahl called the “American Creed,” the formerly fundamental principle that all Americans deserve to be treated “without regard” to race, creed, or color, is simply wrong and unfair. Compared to to what we critics believe to be its fundamental wrongness and unfairness, concerns about mismatch and stigma are almost trivial. Thus racial preference is simply discrimination — no need to stick a “reverse” in front of it — and is wrong whether it lowers the odds of admission for better qualified people by a little or a lot. The point, after all, is not the “odds” of any particular applicant being accepted; it is that no one should be denied anything on account of race.
With regard to “reverse” discrimination, the authors are wholly unpersuasive when they argue that
The first criticism—that affirmative action constitutes reverse discrimination — has not stood up to empirical scrutiny. Studies show that minority affirmative action generally has small and insignificant effects on the admission prospects of white students. In legal terms, the Supreme Court recently held that using race, as one of several factors, in college admissions is indeed constitutional and allowable under federal law (see Gratz v. Bollinger 2003; Grutter v. Bollinger 2003). In light of this decision, and owing to a lack of appropriate data at our disposal, we will not consider the reverse discrimination hypothesis further in this paper. [Citations omitted]
I find this claim bordering on the bizarre. First, even though Grutter (but not Gratz) upheld a preferential admission program, that is a defense only against a claim that preferences are illegal, not that they are wrong or unfair. In addition, The authors do not cite much of the critical literature, and neither they nor their cited sources refute what they do cite. Even if it were true that “affirmative action generally has small and insignificant effects on the admission prospects of white students,” the harm it inflicts is not on “white students” but individual students who would have been accepted had their race not been used against them (and, of course, to the society at large, which suffers a great loss by the undermining of the formerly fundamental principle of colorblind equality).
But, as I argued at some length here, I don’t believe the number of individuals whom racial preferences make victims of racial discrimination is in fact small at all. For example, using data provided by the University of Michigan, I demonstrated that
according to Michigan, 124 white, Asian, or unpreferred minority applicants were prevented from attending the UM law school in one year because of their race or ethnicity. The 2000 entering class of 400 students contained 42 students, or a bit over 10% of the class, who in Michigan’s estimation would not have been there if their race or ethnicity had not been taken into account. 27% of the “underrepresented minorities” who applied would have been accepted under a non-discriminatory, colorblind admissions system; 73% of those who were offered admission would not have been admitted without the racial preference they were given. Thus, 124 whites, Asians, etc., who would have been admitted under a race-blind admissions system were denied admission in order to produce a yield of 42 more “underrepresented minority” admits than a race-blind system would have produced, or about three race-based denials for every one of the preferentially admitted entering students.
Read the whole post to see how these numbers were derived.
According to Bowen & Bok, “a white student” has a 25% probability of being admitted to a selective college under the current regime of race preferences, but under a “race-blind” system that probability would increase “only” to 26.2%. But what if one also considers Asians and other non-preferred minorities? B&B don’t say. In any event, based on their numbers, for every thousand applicants to a selective college, 12 whites (Asians, etc., still invisible) are rejected only because of their race or ethnicity. Applying those numbers to Michigan’s 25,000 applicants every year to its freshman class, Michigan rejects 300 white applicants a year based exclusively on their race.
Again, those 300 are whites only. If the Asians rejected because they were not black or Hispanic were added, that total would be significantly higher. For one university in one year.
In short, even using data provided by advocates of racial preference, there is very good reason to conclude that the number of student applicants who have been deprived of admission to selective schools because of their race is not at all “small and insignificant,” as M & M breezily assert here.
Moving on, let us now consider M&M’s treatment of what they’ve referred to as the “second hypothesis.”
The second hypothesis, which argues that affirmative action sets up minority students for failure by placing them in academic settings where they are under-prepared, has been called the mismatch hypothesis because it posits a disconnect between the skills minority students possess and those they need for success at competitive institutions of higher education (Sowell 2004; Thernstrom and Thernstrom 1999). Although this hypothesis makes intuitive sense, it has not been supported empirically (see Holzer and Neumark 2000; Kane 1998). Bowen and Bok (1998), for example, found that blacks who attended selective institutions were more likely to graduate than their counterparts at less selective institutions. Sigal Alon and Marta Tienda (2005) found that minority students generally thrive at selective institutions, whatever their origins.
This is a woefully inadequate summary of the literature of “mismatch,” making no reference, for example, to evidence from many selective campuses that the graduation rate of minority students is far lower than that of students admitted without preference or to the various studies from the Center for Equal Opportunity documenting just how much of a “mismatch” in qualifications there is. The author of the most impressive recent evidence of “mismatch” is mentioned in the following paragraph, but even more unsatisfactorily:
In his assessment of minority affirmative action at U.S. law schools, Richard H. Sander (2004) argued that black students were substantially less qualified than whites and that, as a result, they clustered at the bottom of the class, dropped out at higher rates, and failed the bar more often. Based on these data, he concluded that affirmative action actually undermined the goal of producing black lawyers. The fact that blacks earn lower test scores, achieve lower grades, and graduate at lower rates is well-documented across all institutions, selective or not, but critics have argued that these facts alone provide no basis for concluding that affirmative action is causing academic problems (see Chambers et al. 2005).
M&M cite only one of Sander’s studies, none of the subsequent articles that support his findings, findings which they summarily dismiss because of what “critics have argued.” Well, yes, they have argued that, but Sander and others have replied at length and effectively to those arguments, replies which M&M simply ignore. This is not impressive scholarship.
M&M do cite a forthcoming study by Massey and Mary Fischer that purports to find that minority students with SAT scores below the “institutional average” earned better grades than minority students generally, but we’ll have to wait until that study appears to evaluate its claims. It will be interesting to see whether, or to what degree, the relative success of minority students with low SAT scores (compared to minority students with average or above SAT scores) conflicts with M&M findings in the current study, quoted in my post above, documenting the “lower academic performance” of minority students at institutions with an “aggressive affirmative action policy.”
With regard to the degree of preference legacies enjoy, recall, as quoted in my original post, that 52% of legacies had SAT scored equal to or higher than their institution’s average, compared to 23% of black and Hispanic students. Legacies, on average, were given a 47 point “bump upward,” compared to 108 points for blacks and Hispanics. Nevertheless, M&M write that “the children of alumni are found to benefit from exceptionally high admission rates.” What do they mean by “exceptionally high”? Here’s a clue:
William G. Bowen and Derek Bok (1998), legacies had a two to one admissions advantage over non-legacies. Cameron Howell and Sarah E. Turner (2004) document a similar advantage at the University of Virginia, where only 32 percent of regular applicants were admitted compared with 57 percent of alumni children. As a result, the freshman class of 2002 was 7 percent legacy, compared with 3 percent African American, even though the state is 20 percent black.
I especially like that “even though the state is 20 percent black.” M&M appear to imply here that the student body of the very selective University of Virginia should mirror the demographics of the state. But let’s leave that aside and concentrate on the “exceptionally high” 57% admission rate for legacies at UVa.
“Exceptionally high” compared to what? According to the latest figures I have seen (for students entering UVa in Sept. 2005), and cited here, 57.1% of black applicants were admitted (12.3% of all admittees), compared to 29.4% of non-black applicants. That amounts to an almost exactly 2 – 1 admissions advantage for blacks, presumably the same as the “exceptionally high” legacy admissions. And as I’ve cited a number of times here over the past few years, although UVa graduates a higher percentage of its black students than any other selective public university (and more than many selective private universities), blacks fail to graduate at twice the rate of whites and Asians. I would be surprised if legacies failed to graduate at that same rate, but I’ve never found any data on that for UVa.
At many institutions, as I’m sure M&M recognize, the qualifications of legacies are indistinguishable from those of the non-legacies. At Harvard, for example, legacies tend to make up about 12 or 13 percent of entering classes, and their rate of admission is definitely higher than the admission rate for non-legacies. According to Harvard College Dean of Admissions William Fitzsimmons, “[t]heir admission rate is between 34 and 35 percent … in contrast to a rate of nine percent for the Class of 2011 as a whole.” Fitzsimmons also said, however, that “[a]t Harvard, SAT scores for Harvard legacy students are ‘virtually identical’ to those of the rest of the student body.” He did not say, at least in this article, that legacies earn lower grades or drop out in higher numbers than non-legacies.
Or consider the case of legacy and minority preferences at Middlebury College (which I discussed here), where the admission rate for legacies in the class of 2006 was 45% and the rate for the class as a whole was 27%. Somewhat like Harvard, however, the average SAT scores of the 30 legacies in that class was 33 points higher than the non-legacies average. No doubt M&M would regard Middlebury’s legacy admission rate as “exceptionally high” — which it is, except when it is compared to the 60% admission rate there that year for black and Hispanic applicants.
Obviously the applications of legacies received preferential treatment at Middlebury, but I doubt they fared as poorly as the preferentially admitted minorities. The Journal of Blacks in Higher Education reports, for example, that Middlebury has one of the worst records for graduating its preferentially admitted minorities: its 2006 data reveals that 92% of whites graduated compared with only 72% of blacks, a 20% differential that was exceeded by only two institutions (one of which is the University of Michigan, with a 21% differential). At Middlebury, in short, blacks failed to graduate at a rate three and a half times greater than whites. Since some of the blacks at Middlebury would have been admitted even without any racial preference, the failure rate for the preferentially admitted blacks is no doubt even higher than three and a half that of the white failure rate.
Finally, M&M are concerned with “stereotype threat.”
If white students believe that many of their black peers wouldn’t be there were it not for a “lowering” of academic standards under affirmative action, and more importantly, if black students perceive whites to believe this (see Torres and Charles 2004), then affirmative action may indeed undermine minorities’ academic performance by increasing the psychological burden they experience. Fischer and Massey (forthcoming) also addressed the issue of stereo- type threat and found that affirmative action policies did, in fact, heighten the level of psychological threat to black students and contributed to their under-performance.
In the present study, we expand on this earlier work by measuring simultaneously the academic effects of all three of America’s large affirmative action programs.
Regarding M&M’s assumptions, note that “lowering” is in quotes above, presumably to indicate that the standards aren’t really lowered. Everything here seems to be a matter of perception — if white students believe … ; if black students believe … ; if black students believe white students believe … , etc.
Specifically, the expectation that others will draw on negative stereotypes in making evaluations increases performance anxiety to put extra pressure on students whenever they are called upon to demonstrate what they have learned and how well they can think. Students who believe that professors and other students are prejudiced against them come to feel as if they are carrying the entire group on their backs every time they are asked to perform. Massey and Fischer (2005) showed that students who saw out-group members as biased and prejudiced experienced a much higher performance burden because they viewed their performance as reflecting well or badly on the entire group, and that this psychological burden lowered grades significantly.
In addition to whatever effects institutional affirma- tive action may have in undermining academic performance through stereotype externalization, it may also lead to other pressures that inhibit scholarly achievement. If a group is perceived to be on campus “illegitimately,” such perceptions could produce a tense social atmosphere characterized by inter-group disputes that poison relations on campus and make life difficult for members of a targeted group, yielding social pressures that undermine academic performance. We label this the social subversion hypothesis because affirmative action is hypothesized to undermine the social legitimacy of group members on campus.
Or, in the alternative, it may simply be that students who enter selective schools with significantly lower academic qualifications on average perform less well than more qualified students. M&M do find that “the greater the gap on campus between minority SAT scores and those at the institution generally, the lower the grades earned by individual minority students at that institution,” but then they immediately hasten to explain that this lower performance is not caused by lower qualifications:
It appears that minority affirmative action as applied in the institutions under study does serve to undermine the perceived legitimacy of minority students’ presence on elite campuses. Thus, institutions that grant a large SAT admissions bonus to underrepresented minorities may unwittingly undermine minority grade performance to the extent that non-minority students and professors view the presence of minority students on campus as undeserving and unearned, or seen as being there only because standards were “lowered” or “relaxed” to admit them.
Even controlling for all the variables in the model, however, minority group members still earn lower grades than other students, suggesting the effect of other, unmeasured factors in determining grade performance.
Again, note that despite the large gap in test scores, M&M refer to standards being “lowered” or “relaxed,” not lowered or relaxed. And it’s still not clear to me why M&M posit “other, unmeasured factors” rather than their measured lower academic qualifications than other groups of students to explain their lower grades.
I will leave it to others better versed in the nuances of social science to evaluate M&M’s models, but to this outsider it appears that what they concentrated on was the students’ own perceived burdens (caused, or not, by factors like “stereotype threat”), how many hours they reported studying, etc., and it was the results in those areas that they used to reject not only the “stereotype threat” hypothesis but also the “mismatch” theory, except, oddly, for legacies:
… in terms of school leaving, the effects of affirmative action hypothesized by critics are generally not sustained. Neither of the two intervening variables (hours studied and performance burden) influence the odds of departure….
Likewise, we find no evidence that an institution’s commitment to racial affirmative action raises the odds that a minority student will leave school…. For minority students, attending a school where the application of affirmative action criteria have produced a large SAT gap between minorities and other students actually appears to lower the odds of leaving school.
Among athletes and legacy students, however, the critique is sustained. Attending an institution where these groups have lower average scores than others at the institutions does raise the odds that individual members of those groups decide to leave school….
At the individual level, there is little evidence that affirmative action itself has any direct effect on minority grade performance…. By granting these students an admissions bonus, colleges and universities do not appear to be setting them up for academic failure by placing them in situations where they are ill prepared to compete. Only among legacies does this story appear to hold: the greater the gap between an individual legacy student’s SAT and the institutional average, the lower the grades earned.
Here is the nub of M&M’s conclusions:
Our systematic analysis of the academic performance of affirmatively targeted groups found
little support for the mismatch hypothesis. Other things equal, minorities and athletes who received an apparent SAT admissions bonus did not earn lower grades or leave school at higher rates than other students on campus. Affirmative action programs thus do not appear to set up either minorities or athletes for academic failure by dumping them unprepared into a very competitive academic environment. Ironically, the only evidence we find of a skills mismatch is for the children of alumni. The greater the gap between a legacy student’s SAT and the institutional average SAT, the lower the grades he or she earned, though the effect size was modest.
There is no attempt here to explain these findings. That is, M&M make no effort to explain why steep admissions preferences have no negative effects, in their analysis, on the performance of minorities and athletes but they do for legacies. If the findings are accurate, perhaps the explanation, or part of it, is that institutions that extend preferences to all three groups care more about the success of minorities and athletes. After all, no school wants its star quarterback to flunk out, nor, I suspect, do any have Deans of Legacy Affairs and enormous support staffs to tutor and monitor their legacy students.
But the larger question about this study is the more glaring conflict it presents, but does not explain, with the by now voluminous data revealing large gaps in grades and graduation gaps between preferentially admitted minorities and other students. How, to take a couple of small examples that I discussed here, can M&M’s models explain things like these that happened in California after racial preferences were (at least legally) eliminated:
One largely unsung result of barring race preferences is the significant improvement in the graduation rate of minorities. Even the Journal of Blacks in Higher Education, an ardent supporter of racial preferences, noted in 2002 that the University of California at Berkeley’s “black student graduation rate has improved by 12 percentage points over the past seven years.” During that same period the black graduation rate at the University of Michigan fell by 5%.
UCLA law professor Richard Sander notes that black students at UC San Diego had a four-year graduation rate of 26 percent in 1995-1996 and a 52 percent rate in 1999-2001.
It would be useful for M&M not simply to dismiss Richard Sander’s findings because they have been criticized by critics but to say where and why his mismatch findings are incorrect. Similarly, I would love to see Sander publish a review of the work of Massey and his associates.
Finally, I don’t doubt that M&M’s models are impressive (though I confess, again, that I don’t altogether understand them). What I doubt is that they explain very much about the accumulating data from various sources about the performance of preferentially admitted minorities.