The Center for Equal Opportunity-which opposes the use of racial and ethnic preferences and has published a series of studies exposing them at public universities, including those in California, Michigan, Colorado, North Carolina, and Washington, as well as the service academies at West Point and Annapolis-asked three authors to review The Shape of the River, each with a different perspective on the book, its methodology, and its arguments.

Robert Lerner is a sociologist and the coauthor of CEO's series on preferences in higher education. He criticizes the book's "data aggregation," which is "at an extremely high level . . . thus ignoring all individual differences among these colleges and universities," and its exclusion of non-black minority groups such as Asians and Hispanics. He notes that the authors' own data establish that "[t]he more selective the school is, the worse do black enrollees perform relative to their white counterparts." Dr. Lerner's studies have concluded-and Bowen and Bok do not dispute-that the more selective schools are the ones that use preferences the most. Lerner criticizes the failure of The Shape of the River to discuss the impact on those non-blacks who are denied admission to top-tier schools because they are wrong skin color and the effects of these decisions on society as well as on individuals. Lerner concludes that Bowen and Bok-who are, after all, former college presidents and long-time users of racial preferences-have failed to offer an objective study. He notes their "lack of rigor in research design, relative inattention to alternative hypotheses, non-presentation of important descriptive data, and ... unwillingness to confront in detail the arguments made by the critics of preferences." Dr. Lerner calls on Bowen and Bok-and on the colleges and universities studied-to make their admissions and other data available to all qualified researchers. "Honest discussion requires no less."

Dave O'Neill, an economist, shows why, "despite hundreds of pages of tables and charts, Bowen and Bok have no good statistical evidence on one of the key causal questions," namely how the beneficiaries of preferences "would have fared in the labor market and their careers" if they had not received that preference. He also notes problems of statistical bias in the data that Bowen and Bok do adduce, and shows how, by ignoring the use of race-conscious policies at non-elite colleges, "Bowen and Bok are just defining away the really tough social problems created by the use of racial preferences in higher education." Like the other two authors, O'Neill points out the book's Orwellian penchant for "terms like 'race sensitive policies,'" as opposed to "candid terms like 'race preferences'" which more clearly describe the policies at issue.

Roger Clegg is a lawyer who served as a deputy in the Justice Department's civil rights division and is now CEO's vice president and general counsel. He stresses the legal significance of one of the authors' quiet concessions: that preferences should no longer be defended as somehow "remedial," but only because of the value of "diversity." Except for the infamous decisions in the Japanese-American internment cases, no Supreme Court majority has ever recognized the legal sufficiency of a racial classification except in the remedial context; and no justice besides the late Lewis Powell has ever said that the diversity rationale passes the test of "strict scrutiny" that the Court now uses for all such classifications. Clegg also notes other major concessions by Bowen and Bok-that preferences are widely used at the top schools, and that they are not mere "tiebreakers" but a very heavy thumb on the scale. These latter points, incidentally, have been repeatedly documented by the CEO studies.

Part of the appeal of The Shape of the River may be its novelty: it is the first book purporting to marshal actual data in defense of affirmative action. But those who read the reviews collected in this Policy Brief will see that these data are often incomplete or misleading, in addition to being shielded in large part from other researchers. Not only that, but the data are often simply irrelevant-used to attack straw men, prove propositions not really in dispute, or offered as a distraction from the real issues.

What I find most infuriating about The Shape of the River, however, is its pessimism and condescension. Its white male authors are absolutely certain that there is no way for some racial and ethnic minorities to make a major contribution to American life unless standards are lowered for them. But these are the same minorities who have survived-and overcome-blatant discrimination against them for many years. What is most essential for their continued progress is simple nondiscrimination and respect-which comes from playing by the same rules and being treated the same as all other Americans. But this is what Messrs. Bowen and Bok would deny them.

THE EMPIRE STRIKES BACK

by Robert Lerner

Racial preference policies in higher education admissions are a hotly contested issue, yet have been subject to little serious study. Ever since sociologist Nathan Glazer wrote his path-breaking Affirmative Discrimination in 1975, many have suspected that most of America's colleges and universities employed racial preferences in admissions as a means of achieving some degree of racial proportional representation. Regardless of the name used to describe the policy, racial preferences describe an admissions process that weighs applicants unequally according to their race, judging applicants according to "the color of their skin," not merely "the content of their character."

But for the past 30 years the academic establishment has stoutly denied that racial preferences are actually used in admissions, or, if they are used, it is said that they have only a marginal impact on admissions. This claim has been difficult to test. In 1991, when Georgetown University law student Timothy J. McGuire exposed the use of racial preferences in law school admissions by pointing to the huge gap between black and white admittee LSAT scores, it created an enormous uproar, a uniform denial by academic authorities that such preferences operate, and, of course, the threat of punishment for the miscreant.1

Even Justice Lewis Powell, in his famous Bakke opinion, cited only the catalogue of Harvard College to justify his invocation of diversity as a criterion in admissions decisions. At no time did he have access to the data on the actual admissions process at Harvard or anywhere else. Colleges and universities are extremely resistant to any study of the existence and impact of racial preferences, and thus data are largely impossible to obtain. To put it bluntly, despite the fact that they are ostensibly institutions devoted to the growth and dissemination of knowledge, including knowledge about higher educational institutions, they have furiously resisted scrutiny or any kind of study that would lead to increased knowledge of whether or how racial preferences in institutions of higher education operate and to what effect.

The research appearing in The Shape of the River purportedly provides some of the missing hard quantitative empirical evidence on affirmative action issues, thereby helping shape the subsequent debate. Its authors, William Bowen and Derek Bok, are former presidents of Princeton and Harvard, respectively.

Supported generously by the Andrew W. Mellon Foundation,2 they undertook a massive empirical examination of the existence and consequences of what they call "race-sensitive" policies of admission to elite undergraduate colleges and universities. In addition to establishing the existence or nonexistence of racial preference policies, the authors propose to investigate their short-run and longer-run consequences in some detail. The short-run consequences include examining the effects of such preferences on college graduation rates and college grades (chapter 3). The longer-run consequences include their influence on graduate study and professional degrees (chapter 4), effects on employment, earnings, and job satisfaction (chapter 5), civic participation and life satisfaction (chapter 6), memories of the alma mater (chapter 7), and graduates' views of diversity and racial interaction (chapter 8).

To carry out these ambitious tasks, the authors have created a massive "restricted access data base"3 (p. xxviii) with the cooperation not only of the institutions studied, but also the College Board and the Higher Education Research Institute. The institutional data file contains individual enrollment and transcript records of the more than 80,000 full-time students who were enrolled at 28 college and universities in the fall of 1951, 1976, and 1989.4

The second part of the book's study is a survey of the life experiences of these students after graduation, especially the 1951 and 1976 cohorts. The details of the methodology are unclear; the presentation convoluted. It appears that the authors surveyed all the enrollees of the earlier two cohorts and a sample of the 1989 entering cohort that was selected from 17 of the schools. (P. 300.) At the liberal arts colleges studied, all enrollees were surveyed, while at the public colleges and universities all blacks, Hispanics, athletes, and a random sample of other students were included in the sample. (P. 300.)5 There was also a matching control group study of all college graduates carried out by the National Opinion Research Center in 1996. (P. 291.)

Finally, the study included two extensive matching projects to supplement the survey and institutional data collection efforts. The first was a matching project carried out in conjunction with the College Board. The information matched included extensive data from the Student Descriptive Questionnaire filled out by test-takers as well as test score information. The second was a matching project carried out in conjunction with the Higher Education Research Institute at UCLA, headed by psychologist Alexander Astin, which carries out annual surveys of college enrollees.6 (P. 313.)

The Shape of the River is packed with 326 pages of text in rather convoluted prose and many tables and graphics, as well as several extensive appendixes, and copious footnotes with much information important for understanding what the authors intended. While the work is so massive that a short review cannot possibly deal with every issue raised in those pages, not to mention discussing every result that is presented, I will cover as much as possible in a chapter-by-chapter discussion of the findings, prefaced with some general comments about the research design employed. I conclude with some overall comments on the moral tone of the book as a research study.

Problems with the Database

Assembling this extremely large body of information by the authors' research team was itself an impressive feat of organization. This does not mean, however, that the assembled database is without problems. For example, individual student records for 1951 apparently do not include the race of the enrollee. (P. 30.) Thus the data for that year are not used in the discussion of applications. (P. 291 n.2.) A still greater limitation of the data is that relatively complete applications files containing individual record information were available for only five schools in 1989 and for only two in 1976.7 These schools provided detailed and apparently complete information on all applicants for admissions, including those who were denied admission, those who were admitted but who chose to attend other colleges and universities, and those who enrolled. (P. 17.)

Even with these limitations, however, there is no convincing explanation why these particular three years and especially the starting year were chosen for the study. Students enrolling in 1951 would have been admitted subsequent to the WWII-era GI Bill of Rights (p. 293),8 but before the Brown decision and well before the mobilization of the civil rights movement of the 1950s and 1960s. To study the impact of preferential admissions, a more reasonable starting point might have been 1966, a point in time subsequent to the passage of the Civil Rights Acts of 1964 and 1965 but before the affirmative action/racial preference bandwagon began its seemingly inexorable movement. Lacking this kind of data makes it difficult to assess the book's claim that racial preference policies in the 1960s were necessary to raise the proportion of black enrollees to its current level. Equal protection or affirmative action without preferences might have sufficed. This bias is a major design flaw.

This flaw could have been partially remedied by a time-series presentation showing black, Asian, Hispanic, white, and foreign enrollments for every year from 1951 to the present, which presumably is available for individual institutions as well as for the sample as a whole.

To see the kind of analysis that might have been done, consider this brief summary of the authors' incomplete qualitative discussion of the history of black enrollment at these schools. The history can be reconstructed as follows. In 1951, long before the advent of the civil rights movement, black students averaged 0.8 percent of the entering classes at the elite schools studied. (P. 4.)9 By 1965, enrollment at "selective New England colleges" had risen to 1 percent. (P. 5.) This probably understates the enrollment at the 28 relevant schools, since not all of them are selective New England colleges. By 1967, the percentage of black enrollees in Ivy League schools increased to 2.3 percent. (P. 7.) Again, this probably understates the true percentage enrolled in the complete sample of schools before the widespread adoption of racial preference policies.

A properly reconstructed history might suggest that the establishment of racial preference policies at America's leading colleges and universities appears to have taken place in 1968 or 1969, following the riots sparked by the assassination of Dr. King. By 1976 the proportion of black enrollees rose to 6.3 percent. (P. 7.) Apparently the year 1976 was chosen not only because the turmoil of the Vietnam War era was over (p. 293), but also because "the exuberance and strong ideological commitment of the late 1960s and early 1970s" which had led many colleges to select "truly disadvantaged students recruited from the ghettos" (p. 7) had been scaled down by then.10 This would tend to minimize the degree of preference found. As the authors candidly admit, however, the 1976 enrollees were still selected using preferential admissions policies. (P. 293.)

Another problem with the study is its aggregation of data. Bowen and Bok claim that they were given the data only with the understanding that individual institutions would not be identified as it is a "restricted access data base." (P.xxviii) This limits the authors to presenting their analysis at an extremely high level of aggregation, thus ignoring all individual differences among these colleges and universities. Clearly, there are massive differences between Yale University and Miami University (Ohio), between Bryn Mawr College and the University of North Carolina at Chapel Hill, and between Pennsylvania State University and Williams College that cannot be easily captured in overall summary analyses. Yet the authors include neither college or university averages nor measures of institutional variability. Except for measures of school quality, there is little discussion of regional or other differences.

Consideration of multiple racial and ethnic groups points to still another major flaw of the study. Unlike the studies I have coauthored for the Center for Equal Opportunity (CEO), Bowen and Bok largely exclude from consideration Hispanic and especially Asian applicants, admittees, and enrollees. Given the emphasis on the racial diversity that provides the underlying rationale for the book's preferred racial preference regime, the lack of attention paid to Hispanic and Asian applicants or enrollees is quite striking. One suspects that a reason for this lack of attention is that whites, like Asians, are likely to be admitted at far lower rates than blacks despite exhibiting outstanding academic performance, and it might be embarrassing to have to explain this away.11 This discrepancy is also quite noticeable in chapter 8-which is entitled "Diversity: Perceptions and Realities"-where questions are included about diversity or "race/culture" as well as getting along with "people holding different beliefs" (pp. 220-21), about which I will say more later.

The Contribution Is the Concession

These criticisms do not mean the authors' research has made no contribution to the study of racial preferences. The major contribution of Bowen and Bok to the discussion of preference policies is precisely the concession by leading members of the academic establishment that racial preferences in admissions favoring blacks and occasionally members of other minority groups at the expense of white applicants (and Asians) for admission are routinely used by all selective colleges and universities,12 private as well as public, for at least the past 25 years and possibly longer.13

As Bowen and Bok put it, all these colleges and universities take "race into account in the admissions process by accepting qualified black students even if they had lower grades and test scores than most white students." (P. 7.)14 Yet for the past 30 years, the very same academic establishment has denied the existence of massive preferences in favor of black applicants and against white applicants for admission to selective colleges and universities.

Despite the denials, this of course has been long suspected by many. More than 10 years ago, Harvard political scientist Robert Klitgaard presented data showing that, in 1971, black applicants to Harvard College with academic ratings of "2" or "3" had a 73.0 percent chance of admission, as compared to the 28.5 percent chance possessed by white applicants with the same academic qualifications.15

The authors' own data provides additional strong support for this view. Despite their lesser academic qualifications, 42 percent of black applicants but only 25 percent of white applicants in 1989 were admitted. (P. 26.) Bowen and Bok show this in more detail in a chart (p. 27) revealing that, for the five colleges and universities with detailed admissions data, the odds of admission of blacks relative to whites is substantially better than 2.5 to 1.

Since many advocates of preference policies prefer to downplay the extent of preferences, some examples drawn from the authors' own data might be useful in dispelling any remaining confusion. Bowen and Bok find that all black applicants with combined SAT scores of 1500 or better were admitted to these top colleges but only about 60 percent of similarly qualified white students were admitted. Also, 75 percent of black students with SATs of 1200-1249 were admitted as compared with only 25 percent of their white counterparts. (P. 27.) Finally, about 40 percent of whites with SAT scores of 1500 or better were rejected while 75 percent of blacks with (much lower) SAT scores of 1250-1299 were accepted. This is not a minimal degree of preference of blacks over whites, but a substantial showing of racial favoritism.

In fact, given the richness of the admissions data available, the authors' analysis of admissions data was disappointingly superficial. In addition to information on admissions and enrollment decisions, they have at least some information on the test scores of applicants, their high school grades, whether or not the applicant was a legacy, his or her athletic status, sex, home state or county, citizenship, place of birth, size and type of high school, family background (mother's and father's occupation and education), and even financial aid information. (P. 294.)

This raises two important questions with a methodological flavor: how to measure the extent of racial preferences in admissions and how to analyze any available data so as to estimate the degree of preference given, if any. The major problem for any study of preferential admissions is that this policy normally cannot be directly observed; therefore it must be inferred (or inferred not to exist) from external evidence.16 A second consideration is that-unless there is a fixed quota selection process, which is illegal under Bakke-preference, if any, is a matter of degree.

CEO Studies

A common method for ascertaining the existence of preferences is to compare average test scores and grades of enrollees by race and ethnicity. This procedure is admittedly imperfect, but it does establish some useful limits. If the averages differ substantially, then this showing is a first step toward demonstrating the existence of preferential policies.

This procedure, commonly used in many studies of racial preferences (including studies carried out by the Center for Equal Opportunity), induces a disingenuous howl of protest from Bowen and Bok quite out of keeping with the carefully controlled tone of the rest of the book. They claim unequivocally that using mean differences in test scores and grades as an indicator of discrimination is "seriously flawed" and "should not be used for this purpose." (P. 16.)17 The reason is because the large differences in test scores and grades that exist between blacks and whites in the general population and presumably in applicant pools will yield black-white differences in admittee qualifications even under race-neutral conditions.

But the size of the black-white difference in mean scores is itself a useful indicator of the extent of racial preferences, even if it is imperfect. The larger the difference is, the more likely that there is racial preference in admissions and the larger the amount of such preference is likely to be for the favored group. This conclusion follows as a matter of simple common sense. The term "racial preference" means admitting individuals of the proper skin color with lower grades and test scores over those with higher test scores and grades but with the wrong skin color.18 Using racial preferences in admissions will lower the average black score and grade point average and increase the difference between the black and white averages over what would have been the case if race were not a criterion in admissions. The greater the degree of preference afforded to blacks, the greater the black-white difference in average scores will be, because it will require admitting those in the applicant pool with even weaker qualifications. When racial preferences are removed, the differences in test scores and grades may not vanish totally but they will decline very substantially.

The situation is analogous to considering the size of a correlation coefficient as evidence for a causal relationship in ordinary social science research. The larger the size of the coefficient, the less likely it is that the correlation is explainable by other factors and the more likely it is not a chance finding. Finding a large correlation coefficient is not proof but is strong evidence of a causal relationship.19 The distributional problem does raise the question of how large a difference in means is required in order to provide strong evidence of preferential treatment or of discrimination. This problem exists for all statistical studies of discrimination. The answer is to some extent arbitrary, but it is useful to establish some kind of threshold value for inferring the existence of preferences.

CEO's studies of racial preferences in undergraduate admissions do not assume that every racial or ethnic difference in median test scores and grades (a more accurate measure than differences of means) is the result of racial preference. Nor in fact have they concluded that every school studied shows such preferences. In fact, they generally assume that if the difference is less than 30 points on either of the SATs, and less than 0.1 of a grade point on high school grades, the school's admissions policy likely does not use racial preferences. In order to infer the operation of racial preferences, they generally assume that all three indicators of academic merit must exhibit differences of at least the above size.

In order to ascertain what might happen in the real world, I carried out a simulation of this general proposition on the University of Michigan at Ann Arbor's 1995 applicant data. The original white-black differences in admittee medians20 are: 100 points for verbal SATs, 130 for math SATs, and 0.40 of a grade-point for high school grade point average. I simulated an admissions procedure for admitting students based solely on their grades and SAT scores, a race-neutral admissions process, and computed the median differences in SATs and grades on my "admittees."21 I found that the white-black differences in test scores and grades had shrunk drastically. The mean difference in math SATs was now only 50 points, the average SAT verbal difference was now only 30 points, and the GPA difference was 0.0. These results indicate that about two-thirds of the difference in median test scores and all the difference in high school grades could be attributed to the operation of racial preferences, while the remainder of the difference is the result of the black-white test score gap. As the black-white mathematics gap is the largest in the general population, it seems reasonable to put the least weight on that difference and the greatest weight on differences in grade point averages in attempting to ascertain the existence of racial preferences.22

Therefore Bowen and Bok are wrong that differences of median yields no useful information, but they are right that using mean differences in scores is not the optimal procedure for ascertaining the extent of racial preferences in admission if other good data are available.

Of course, the best, most immediately useful information is documentary evidence from administrative memos and records. As a practical matter, however, this is never made available to the outside researcher, even though there is no good social-science reason for keeping college and university admissions policies as closely guarded as national security secrets. The next best kind of information consists of applicant data in the form of individual records that allow the researcher to model directly the admissions process. This kind of information is also never routinely released by colleges and universities. CEO has obtained its information only through freedom of information act requests. Even Bowen and Bok say that they were prohibited from naming the five colleges and universities that provided them with relatively complete admissions data-and carried this policy over to not discussing individual results for the remainder of their study (which used the full sample of 28 schools).

Once individual data have been obtained, the best procedure for ascertaining the existence of racial preferences is to estimate the odds of admission by using both academic and non-academic criteria as independent predictor variables in a regression equation. As a practical matter, this is best done by computing logistic regression equations.23 These odds ratios provide the most precise estimate of the extent of racial preferences in operation (if any). They estimate the effects of race, in varying combinations, on the odds of admission. The procedure allows the researcher to hold constant such factors as test scores, grades, and any other available criterion variables. For example, one might compare the odds of admission for blacks as compared to whites, or Asians as compared Hispanics, holding grades and test scores constant.

CEO's reports have used precisely this procedure in studies of racial preferences at public colleges and universities in Michigan, the service academies at West Point and Annapolis, and North Carolina.24

Unfortunately, the authors' analysis of admissions (p. 17) fails to follow this procedure. Despite its very rich data lode, the analyses are lacking because: (a) the authors do not identify the individual schools studied, (b) they do not present any logistic regression equations which would estimate the odds of admission controlling for all the information they have available, and (c) they present information only for 1989, the beginning of the period of heightened interest in racial preferences (p. 9). Step (b) is of singular importance because it is often claimed that race is only "one factor" in admissions and that athletic and alumni preferences are at least as important. This canard is defensible only because of the secrecy of admissions procedures and data.

If Bowen and Bok had carried out the analysis in the fashion suggested, I surmise that they would have obtained results comparable to those we found at the University of Michigan at Ann Arbor, where we concluded that the black-to-white odds of admission are on the order of 170 to 1. Similarly, it would have been possible to examine directly the relative impact of legacies (i.e., alumni/ae children), athletes, region of residence, race and ethnicity (including Asian and Hispanic applicants), as well as academic qualifications as predictors of admission.

The authors do present selected data on legacies and athletes which is of interest. They state that, overall, legacies are admitted at twice the same rate as other candidates. When SAT scores are taken into account, however, the legacy benefit is far less than the racial preference given.25 Thus, among applicants with combined SATs of 1100-1199, 22 percent of all legacies were accepted, versus 18 percent of all white applicants but 40 percent of all black candidates. At combined SAT levels of 1300 or greater, 60 percent of legacies are admitted, 70 percent of blacks are admitted, and 24 percent of all non-legacies are admitted. (P. 28.) One would like to know what the black legacy admissions rate is, but the authors provide no information.

Athletes receive even more preference than do legacies. Bowen and Bok report that 78 percent of those identified by coaches as promising candidates for college sports teams were admitted, a figure which includes an 84 percent admit rate for those athletes with SATs of 1150 or higher. (P. 29.) Again, one would like to know what the black versus white athlete rate of admission is, but the authors provide no information here either.26

Consequences of Preferences

Having disposed of the false claim that weighty racial preferences do not exist in elite college and university admissions, the next step is to ascertain their consequences, which is the subject of the remainder of the book.

Two common measures of college performance are graduation rates and college grade point averages. Despite the authors' extensive efforts to show how wonderful the performance of black enrollees is, it is clear from their data that blacks graduate from these institutions at lower rates than do whites and that Asians have the highest graduation rates of all. (P. 56.) This race and/or ethnic effect (Asians graduating at higher rates than whites, blacks graduating at lower rates than whites) is independent of the quality of the school, parental socioeconomic status, sex, SAT score, high school class rank, and whether or not the school is a woman's college. (P. 381.)

One might expect that the best schools have the lowest rates of graduation and that at these schools blacks have the lowest rates of graduation, but this is not the case. In order to show this, Bowen and Bok divide their sample of 28 schools into three categories of selectivity based upon their SATs: SEL-1, SEL-2, and SEL-3.27 It turns out that students at the very best schools graduate at the highest rates. (P. 376.) While this is true for students of all races, it is especially true for black students. The highest black graduation rates in the entire sample are for students enrolled at the most selective schools with the worst composite SAT scores, namely those of less than 1,000. (Pp. 61-62.) At the same time, the lowest graduation rates in the sample are among black students with the same SAT scores at SEL-3 schools. Why this occurs is not explained, but it is asserted to contradict the view, which Bowen and Bok term the "fit" hypothesis, that blacks will graduate at the lowest rates at the best schools.

One might suspect that these students are kept in schools by an elaborate series of remedial and tutorial programs, including extensive "personal attention, counseling, support." (P. 58.) Surely the authors, as former presidents of SEL-1 institutions, could describe in detail the programs that keep these students in school.28 These might have even been made part of the data set, so that individual students who have received remediation might be studied separately.29 Or do remedial students, including those who receive racial preferences, receive the equivalent of a social promotion, at least as far as graduation is concerned?

It may be that schools learned from the experiences of the late 1960s and early 1970s when what Bowen and Bok call the "strong ideological commitment" to preferences "led many colleges to place an emphasis on recruiting truly disadvantaged students recruited from the ghettos." (P. 7.) Since "absorption of black students into higher education did not prove to be a simple matter" (p. 7)-most of them could not survive academically in the elite college and university environment and dropped out-special programs might have been created to keep them in school. The improvement in graduation rates from 1976 to 1989 might reflect the creation and expansion of remedial programs for all disadvantaged students. (P. 69.) Again, the authors, despite their long experience as academic administrators, are silent.

This line of argument is further supported in a different section of the chapter entitled "Institutional Initiatives," which discusses the many special programs available to help black and sometimes Hispanic students on various campuses. These include the 21th Century Program at the University of Michigan, the Mellon Minority Undergraduate Fellowship Program, the University of Colorado program for minority engineering students, and the Meyerhoff Program at the University of Maryland Baltimore County. (Pp. 86-87.) This is only a scattering of the myriad of such programs in existence, both at individual colleges and sponsored by external funding sources. (Of course, as these particular programs are "race-sensitive," no white students need apply, but there is no reason to assume that all remedial programs are "race-sensitive.")

Next, Bowen and Bok analyze differences in undergraduate grades. This is an important contribution of the book. Because these data are nearly impossible for outsiders to obtain and are a much more sensitive measure of academic performance than are graduation rates in an era when most students at elite colleges and universities graduate, the results are worth scrutinizing closely. The average cumulative GPA of black enrollees at all 28 schools was 2.61 on a 4-point scale, while the average for their white counterparts was 3.15. The resulting difference of 0.52 of a grade point is quite large, as Bowen and Bok readily admit. (P. 72.)30 It does suggest that those admitted because of preferences are likely to do much worse than their non-preferred fellows who would have been admitted anyway. Unlike the situation for graduation rates, the selectivity of the school strongly and negatively predicts class rank. The more selective the school is, the worse do black enrollees perform relative to their white counterparts. (P. 383.)

Interestingly, while Bowen and Bok make much of the failure of the "fit" hypothesis for the case of graduation rates, they fail to apply it to class ranking (grades), where it apparently works.31 The programs that foster retention, including the blacks-only programs discussed above, are apparently unable to improve the grades of black students to match their white counterparts.

Black Underperformance

Nor is this all. Bowen and Bok find the widely-recognized black underperformance phenomenon in their data.32 Although until quite recently it was commonly alleged that these tests were biased against blacks, in reality black students with the same SAT scores as whites earn substantially lower college grades than do these whites. (P. 77.)33 This gap appears to be the greatest for the highest scoring blacks.34 The authors consider two different possible explanations: (1) inadequate precollegiate preparation not reflected in high school grades and SATs and (2) experiences in college.

The former possibility is not discussed at great length. But it is worth remembering that all black students with high SAT scores were admitted to these colleges and universities, versus only about 60 percent of their white counterparts. It may be that these white students have something lacking in their black counterparts, such as superior study skills. This is only a suggestion, but one that merits further investigation not undertaken by Bowen and Bok.

The second possibility leads to consideration of some interesting research efforts undertaken by Claude Steele and Joshua Aronson on "stereotype vulnerability."35 The suggestion here is that, where a negative stereotype exists, individuals to whom the stereotype may apply must deal with the possibility that their actions will confirm it. This, in turn, impairs individual performance. In the case of blacks, when the negative stereotype of black achievement is invoked, it will purportedly lower black performance regardless of other factors. Steele and Aronson carry out some interesting experiments which appear to confirm this view.

Bowen and Bok acknowledge the potential utility of this explanation but largely dismiss it for reasons that are not clear. Yet in a different part of the book, they claim that the unwillingness of selective institutions to discuss racial preferences is in part due to the fact that "the standing of black students would be lowered in the eyes of their white classmates if the differences were publicized." (P. 265.) Bowen and Bok add: "More than a few black students unquestionably suffer some degree of discomfort from being beneficiaries of the admission process . . . [and] many high-achieving black graduates continue to see reassurance that 'they have made it on their own.'" (P. 265.)36 This would lead to their stigmatization and, one assumes, the reinforcement of the stereotype of black non-achievement and the accentuation of whatever "stereotype threat" problems exist for black students. This supports the Steele view and adds weight to the notion that preferential policies are not cost-free even for their supposed beneficiaries.

There is a third possible explanation that Bowen and Bok do not consider seriously. (P. 85.) This is that black students, knowing that their admission to an elite college or university means they have it made for the rest of their lives, have little incentive to work as hard as their white counterparts.37 Despite the authors' refusal to deal with the hypothesis, anecdotal evidence supporting it is provided by one of their own interviews. Bowen and Bok report the following comment by a black enrollee at the University of Michigan (Ann Arbor) from an inner-city high school: " 'I thought I would be able to skate through, like I did in high school.'" (P. 79.) All in all, it is rather amusing to see an economist (Bowen) refusing to consider the possible hypothesis that incentives affect personal behavior, when practically all of economics is based on this assumption.

There is also independent evidence for the idea that effort devoted to studying matters in student achievement. Psychologist Alexander Astin, in a large-scale national sample survey of college students, not surprisingly finds that number of hours per week spent studying is positively correlated with students' college GPA, controlling for a wide variety of other variables.38 Despite their collaboration with Astin, Bowen and Bok failed to measure or even consider this variable. Could it be that they were afraid as to what they might find?

There is also a more general problem. Bowen and Bok analyze only their total enrollee data set rather than also analyzing separately the collection of five schools with complete admissions data. Had they used these data also, it would have allowed for separate analyses of the performance of those blacks who were preference admits and those who were not preference admits. These analyses, quite important in determining the effects of preferences on subsequent performance, were never carried out. Such analyses would have also allowed comparative study of the academic achievements or non-achievements of legacies and athletes,39 who, while receiving admissions preference, receive less than do blacks. These comparisons might have served as a kind of non-racial control study.40

Several other notable indicators of college performance are omitted. Bowen and Bok provide no analysis of college honors-either Phi Beta Kappa or Latin Honors (cum laude, magna cum laude, summa cum laude)-by race and ethnicity. There is also no discussion of GRE, LSAT, MCAT, GMAT, or NTE41 scores by race or any other way, despite the fact that the authors had the assistance of the College Board, which administers these tests. Do blacks perform as poorly on these indicators as they do on SATs ? Do college grades and attendance at elite colleges and universities help them close the score gap? Analysis of such questions would contribute to examining further the "fit" hypothesis and to demonstrating any negative effects from preferences.

The authors' omission of graduate test data is especially interesting in light of the fact that black graduates were more likely to earn legal and medical degrees than their white counterparts. (P. 99.) Even more strikingly, blacks are more likely to earn law, medical, and business degrees at top-tier universities than are their white counterparts. (P. 102.) As Bowen and Bok admit, graduate, law, medical, and business schools' own racial preference policies may account for these differences. (P. 109.)

If we accept the newly-minted liberal conventional wisdom that the black-white SAT gap is real and predictive of both academic and job performance outcomes, then it follows that blacks privileged by racial preference policies both in school and on the job will occupy the bottom of the distribution in their respective professions. The widespread operation of racial preferences guarantees that this overrepresentation remains in effect at least for the time being. The stereotype that purportedly causes high-achieving blacks to experience "stereotype threat" will remain in being as well for as long as do racial preference policies.

A major contribution of the authors' research is their showing that it is worth a good deal to attend one of their elite institutions, because there is a kind of halo affect that accrues to their graduates regardless of their actual abilities, skills, and performances either before or after attending college. For example, Bowen and Bok note that going to an SEL-1 or SEL-2 school increases the odds that a student would earn a professional or doctoral degree. (P. 114, 387-91.) Similarly, attending one of the elite institutions provides "a real wage premium for students" who graduate. (P. 128.) Of course, the findings make gaining admission to these institutions even more desirable than would be the case if this halo effect were absent.

This highlights another shortcoming in the book, however, namely its failure to discuss the fate of the white and Asian students who are not admitted due to preferential treatment except in quite disparaging terms. (Pp. 36-37.) In a rare criticism of any academic practice or policy, Bowen and Bok even attack "early admissions" policies because whites are reportedly more likely to make use of them than are blacks. (Pp. 38-39 n.24.) They do concede, however, that these rejected white and Asian students would have done at least as well as their black counterparts, including having higher incomes. (P. 282.) These students would likely have also received better college grades. Presumably they would have also learned more, thus fulfilling the normal admissions criterion of being better able to make use of the opportunities provided. But this thought does not occur to Bowen and Bok.

Diversity

The authors' chapter on diversity is highly problematic. While Bowen and Bok concede that they have by no means resolved the issue of exactly how diversity is supposed to have enriched education and although they define and even ask about diversity of ideas, they focus as expected on racial diversity, and there on black-white differences. (P. 219.) This focus is even more peculiar with the authors' acknowledgement that, quoting writer Anthony Appiah, "racial identity is regarded as important by many middle-class African Americans . . . in part because there are so few cultural differences between them and comparable white Americans." (P. 220.)

Bowen and Bok proceed to argue that if race-neutral admissions were adopted, the social gains from cross-racial interaction would decrease. This proposition is supported by a statistical analysis purportedly showing that if the percentage of blacks dropped below 5 percent the proportion of whites who know blacks "well" would decline from 60 percent to 49 percent. (P. 234.) Even if one accepts the view that this kind of interaction is important and accepts the inability of students on such campuses to find these individuals on their own, both of which are questionable, this modest drop appears to be of little consequence.

While this chapter raises some interesting questions, it is the questions that aren't asked by the authors that are significant. In the first place, Bowen and Bok never ask their respondents directly about such issues as affirmative action, quotas, racial preferences and the like. Why not? It cannot be because no one has ever asked such questions before. There is a wealth of survey questions and data on the subject. For example, sociologist Seymour Martin Lipset notes that Gallup has asked questions about racial preferences six times between 1977 and 1991.42 (Incidentally, Lipset notes that Americans are overwhelmingly against such preferences.) As political scientists Paul M. Sniderman and Thomas Piazza, authors of a major study explaining attitudes towards preferences, put it: "to give an account of people's reaction without taking into account what they are reacting to is misleading in a fundamental way."43 One can only conclude that Bowen and Bok really did not want to know how their respondents felt about these issues or about how they might feel if they came to discover they were either advantaged or disadvantaged by racial preferences. The reality, referenced in the books cited above, is that large majorities of the American public oppose racial preferences.

This discussion of the social benefits of interaction across racial lines is misleading, finally, because Bowen and Bok ignore a rather basic proposition in the social psychology of personal interaction. It is a well-established principle that it is equal-status contact among persons of different races that reduces intergroup prejudice. Psychologist Gordon Allport writes that "prejudice . . . may be reduced by equal-status contact between majority and minority groups in the pursuit of common goals." (P. 267.) Sociologist Robin Williams lists as one of "the durable conditions of mutually acceptable cultural pluralism . . . a substantial amount of equal-status collaborative interaction across ethnic lines."44 As long as racial preferences exist or are thought to exist, the equal-status requirement is violated and the true benefits from such contacts are denied. Deep down, Bowen and Bok must know this, for it explains their observation, quoted above, as to why black-white interaction is strained after it is discovered that one group received preferential treatment while the other group did not.

Still another problem is that Bowen and Bok never seek to show the academic benefits of diversity for students. One way to do this might have been to relate differences in diversity or interaction to student achievement, either in college grade point average or in postgraduate test scores or grades. My supposition is that one of the reasons that Bowen and Bok did not do this is because this kind of diversity has absolutely no effect on test scores or other measures of academic achievement. Alexander Astin has found that "academic outcomes are not affected by peer measures," which means that the racial composition of the college or university has no discernible effect on the grades or test scores of white students.45 As noted above, this also may be why Bowen and Bok do not make use of graduate-level test scores.

Here are a couple of other findings that are of interest. Despite all of their preoccupation with leadership, Bowen and Bok find that civic participation rates do not correlate with SAT scores or college class rank. (P. 165.) Leadership may be a desirable attribute, but it is clearly a non-academic accomplishment. Also, the analysis of political attitudes that suggests that black male elites are in the middle of the road is simply laughable. (P. 174.) If the authors had bothered to ask about voting behavior, rather than asking only about whether or not their respondents voted, I suspect that they would have found a uniformly liberal cast here. But it is worth noting that, even using these restricted measures, graduates of the elite colleges and universities are substantially more liberal on both social and economic issues than are college graduates as a whole. (P. 413.) Could this be the result of political correctness? Unfortunately, Bowen and Bok do not attempt to correlate these views with respondents' views of diversity, as discussed in their chapter 8. This would likely have had the effect of showing that graduates of the institutions studied are not uniformly in favor of diversity, especially if it is defined as racial proportionalism (as Bowen and Bok de facto define it).

One of the arguments made by Bowen and Bok in defending preferences is that they have dramatically expanded the black middle class beyond what it would have otherwise been. But the notion that the graduates of these elite colleges and universities constitute in quantitative terms the middle class is ludicrous; there are too few of them. In any event, a majority of black enrollees (63 percent in 1976 and 71 percent in 1989) were already middle class according to Bowen and Bok. (P. 341.) Thus, in 1976, 46 percent of black enrollees had at least one parent with a B.A. while, in 1989, 64 percent of black enrollees had at least one parent with a B.A. (P. 341.) While white enrollees were both better educated and richer, only 26 percent of black enrollees were of low socioeconomic status in 1976 and only 14 percent of 1989 black enrollees were of low socioeconomic status. (P. 49.)

Policies of racial preferences at elite colleges and universities do not create the middle class. Instead, policies of preferential admissions at elite colleges and universities tend to enrich the children of middle-class or upper-class parents. This result is hinted at when Bowen and Bok dismiss the suggestions that class-based admissions preferences can be substituted for race-based ones. (P. 49 and passim.) Needless to say, they do not choose to emphasize the implications of this finding.

Some Concluding Thoughts

Although Bowen and Bok have assembled some important data and have contributed some important analyses to the study of racial preferences in colleges and universities, the book overall is seriously flawed. The problem is that the authors pursue two very different and conflicting aims in their single-volume study. First, The Shape of the River presents the results of a large-scale empirical study of the existence and effects of racial preferences in higher education. Second, the book is an apologia for policies of racial preferences in higher education admissions, including those carried out by the authors when they were university presidents. This amalgam of administrator and social scientist promised unique opportunities for increasing understanding of how preference policies work, since most social scientists are not policymakers, but it also presented a danger that the former policymakers would feel constrained to defend their policies regardless of the data or any other considerations. This is what happened.

Its tangled apologia for academic institutions renders the book more of a polemical legal brief-defending any and all policies that the academic establishment has carried out for the past 30 or so years-and less the objective social science research study that would have been desirable.

This counsel-for-the-defense approach is copiously illustrated in several ways. The large number of superficial research efforts reported could and should have made for a much larger book or a far more detailed discussion of each major point. Many of these studies exhibit a lack of rigor in research design, relative inattention to alternative hypotheses, non-presentation of important descriptive data, and an unwillingness to confront in detail the arguments made by critics of preferences. These problems indicate an intent to establish their view of preferences rather than simply to present analyses of their data.

The polemical thrust of the book is also illustrated in the authors' characteristic use of language to describe preferences and admissions policies. For example: (1) the authors deny they favor "quotas" (p. xxiv), (2) they have a great deal of trouble in using the word "racial preference," preferring the term "race-sensitive admissions," and (3) what they really mean by "diversity" is racial proportional representation.

The first point is tendentious at best. Quotas, goals, and preferences are convertible quantities. Any set of preferences can be manipulated to produce a desired quota or goal for a given applicant pool, simply by adjusting the size of the preference given to members of favored groups. This is because the larger the size of the preference given, the greater is the proportion of the preferred group which is admitted. Bowen and Bok know this full well, but perhaps since strict quotas are illegal and preferences may or may not be, the distinction likely turns upon legal technicalities rather than on meaningful empirical differences.

The second point also presents a problem. Bowen and Bok have difficulty with the notion of race-neutral admissions, placing the term "race-neutral" in quotes. They continuously blur together policies with differential consequences for racial and ethnic groups with policies that explicitly single out individuals of different races differently. Again, Bowen and Bok are perfectly aware of the differences between these two kinds of policies, but choose except when they absolutely must to ignore them. This point illustrates the authors' obvious relish for euphemism. Bowen and Bok have great difficulty with the notion of racial preference, which they practically never use.46 They prefer a term of their own invention, "race-sensitive admissions," as the term of art. This may be a useful argument to make in a legal brief but it is laughable as serious social science. What does this term mean? What "the right people" are sensitive to is that many whites and Asians will be discriminated against by these elite colleges and universities. Admissions officers are "sensitive" to their race, all right.47 Any reader even semi-alert to euphemism will detect the linguistic legerdemain of Newspeak.

The third point involves euphemism, too. As was noted in the discussion of "diversity," Bowen and Bok invoke this term of art talismanically, but what they are really interested in is black proportional representation. Diversity of opinion, which they acknowledge, is never discussed at length. As noted previously, they largely ignore Asian and to a lesser extent Hispanic enrollees, even though together with foreign students these make up about a quarter of the enrollees at the colleges and universities studied.

Worse than all these semantic slights of hand, however, is the real problem with racial privileging policies and with the authors' attitudes towards studying them. They cannot be carried out openly and honestly and cannot be studied publicly with full disclosure of all data. Bowen and Bok note that the unwillingness of selective institutions to discuss racial preferences, to discuss their admissions policies openly and honestly, and to make routine presentations of the kind of data discussed here is due to the fact that "the standing of black students would be lowered in the eyes of their white classmates if the differences were publicized." (P. 265.) But what about the truth?

This attitude of hostility towards openness, unfortunately, is one that Bowen and Bok appear to endorse. An indication of their attitudes towards outside research by others is given in their discussions of others' efforts to study preferences. If one were to follow the authors' rule that differences in mean or median scores between racial and ethnic groups are never to be used to infer the existence or absence of preferences and combine it with the intense secrecy that has covered admissions policies, it would provide a fine means of insulating admissions committees from any accountability. Indeed, it would for all practical purposes eliminate the study of this vital subject for all except the favored few.

At no point do Bowen and Bok state that these matters should be studied by others and that colleges and universities need to open up their files to outside researchers. While they make much of their "restricted access database," they provide no indication that they favor study of preferential policies by other researchers who do not or need not share their point of view. Nor do they indicate in any way an intention to make their data publicly available for other researchers to analyze.

This is contrary to the scientific spirit of free inquiry, public criticism, and disclosure48 and to the philosopher Charles Sanders Peirce's famous injunction, his ninth lesson on the history of science: "Do Not Block the Road of Inquiry."49

Supporting this kind of secrecy is wrong because it inhibits serious study of the problem of racial preferences in colleges and universities. My own view is that admissions committees at public universities should be required, and private colleges and universities be strongly encouraged, to publish their "admissions matrices" (point scoring systems). Both types of institutions should allow qualified researchers the same access to (suitably redacted) individual records granted to Bowen and Bok. Honest discussion requires no less.

MUDDY RIVER

by Dave O'Neill

William Bowen and Derek Bok are both former presidents of what can only be called elite universities-Princeton and Harvard. And their obviously elitist attitudes makes this a maddening book to read. It drips with noblesse oblige and uses terms like "race sensitive policies" and "selective colleges and universities" to avoid using candid terms like "race preferences" and "elite schools."

This book has been enthusiastically received by adherents of affirmative action who have been disheartened by the growing opposition to racial preferences in school admissions, job hiring, and the like. The New York Times sighed with relief, stating that the book "provides striking confirmation of the success of affirmative action in opening opportunities and creating a whole generation of black professionals who are now leaders in their fields and communities."1 But unfortunately for the fans of racial preferences this claim is simply not true. In fact the book deals with a sector of higher education (the most elite private colleges) and a group of black college entrants (from the very top of the black SAT distribution) for which the imposition of racial preferences in admissions raises the least problems for public policy. And even for their limited group, despite hundreds of pages of tables and charts, Bowen and Bok have no good statistical evidence on one of the key causal questions: if the (subgroup of) black students who were admitted only because of the preference policy had not gotten into the elite colleges and universities, then how would they have fared in the labor market and their careers?

Why the Study Is Irrelevant for the Tough Issues of Racial Preferences

Bowen and Bok focus exclusively on a subgroup of blacks from the top of their ability distribution who score relatively high on the SATs and on a subgroup of elite, mostly private colleges and universities which voluntarily adopt admissions policies that favor blacks. This means that any findings of their study cannot be applied to the many public institutions that have to deal with blacks from a much lower part of their distribution and are mandated either to ensure certain quotas of black students (as was the state university system in California before Proposition 209) or have had to lower entry standards to ensure minimal representation of minorities. These latter situations are what might be called the "trenches" in the battles surrounding racial preferences in higher education. Bowen and Bok are operating in an area of racial

preferences that has caused some controversy, but it is mostly controversy among intellectuals concerned about private elite colleges and universities. But Bowen and Bok think they are at the center of controversy over race preferences. In their own words from the preface:

Within the realm of higher education, we are concerned only with academically selective colleges and universities. The main reason is that the debate surrounding race-sensitive admissions is primarily within these institutions. In colleges and professional schools that admit nearly every qualified applicant, there is little to debate (although there may be arguments over how "qualified" should be defined, and whether the same definition is applied to white and black candidates). It is when there are strict limits on the number of places in an entering class and far more qualified applicants than places, that the choices become difficult and the issue of whether to give weight to race comes to the forefront. (P. xxvi.)

In other words, according to Bowen and Bok, there is no problem of a race-based admission policy at, say, the City University of New York in New York City because every qualified applicant is admitted. The fact that the minimal qualification standard at CUNY has plummeted over the years, for the sole purpose of increasing black and Hispanic enrollment, does not qualify as a policy of giving weight to race. But Bowen and Bok are just defining away the really tough social problems created by the use of racial preferences in higher education.

No Evidence on Key Questions

The diligent reader, willing to plow through appendix tables, can assemble a useful description of what happened to blacks who went to the elite colleges. We are inundated with glowing material on how many of the blacks went on to professional schools, how they excelled in community services, and how they earned very high incomes-much higher than the average of all black college graduates. This goes on and on page after page and chapter after chapter. Much less trumpeted are the "disturbing" facts that show the graduation rates and especially the grade levels of the black students being much lower than that of white students, even holding constant SAT scores and other variables. It is typical of the Bowen-Bok style that they brush aside the possibility that this poor academic performance of blacks might be an unintended effect of the race preference policies-for example, that these policies may lower their motivation to perform because they feel patronized by the policies. Instead Bowen and Bok dwell on factors like the insecurities of black students, their difficulties in adjusting to new environments, etc., etc. For them racial preferences can do no wrong.

But what is the significance for public policy toward racial preferences of all this, mostly glowing, material? What would be useful to have is statistical evidence on two causality questions: (1) if the blacks who entered the elite schools because of the racial preference policy (not all the blacks because that would include blacks who would have been accepted without a preference policy) had not been able to, then how would their careers have turned out?; and (2) do race preference policies tend in Shelby Steele's term "to demoralize" black students (and perhaps non-students as well)? On both these questions Bowen and Bok claim to have important evidence.

With respect to question (2), it doesn't make much sense to debate Bowen and Bok. They have convinced themselves that black graduates' answers to survey questions about how they felt about the "diversity" policies of the schools they attended is going to elicit their true feelings about racial preference policies in admissions. As noted they refuse to consider the possibility that this demoralization effect might be producing the poor black academic performance in their data. Their total commitment to racial preferences makes them incapable of doing a meaningful analysis of the demoralization question.

On question (1) we must take up their claim in detail because on the surface the statistical comparison they make might appear to shed some light on the real question. Bowen and Bok never attempt to identify the subgroup of individual blacks who were accepted by the elite schools but would not have been if they had not had racial preference policies. The problem is that in the elite schools the preference policy is implemented without different cutoff scores; hardly anyone is accepted in these schools that has a combined SAT score less than 800. These schools favor blacks by taking a much higher fraction of black applicants at each very high SAT level. Thus in the combined SAT interval 1250-1300 these schools accept about 75 percent of the blacks who apply but only about 25 percent of the whites who apply. The amount of preference is quite large: for a black applicant to have a 50 percent chance of acceptance he has to have an SAT of only 1125 while for a white to have the same probability he must score at least 1475. Bowen and Bok calculate that if the preference program were dropped the black acceptance rate would fall from 42 percent to 13 percent. Thus they estimate that only 30 percent of blacks accepted under race preferences would have been able to make it with no preferences. Bowen and Bok do not mention the possibility that blacks would respond to the elimination of the preference policy by working harder to raise their SAT scores, making the percent accepted greater than 13. (This happened among black college athletes when the NCAA required higher grades for athletes to be eligible for sports.)

Although Bowen and Bok estimate the number of blacks who would not have been accepted, they do not try to identify who the individual blacks are in their data. Without being able to make this separation, Bowen and Bok cannot directly answer the key question (1) above. This basic shortcoming of their data and analysis is not stated in a straightforward way anywhere in their long book. And perhaps more egregious is that they never attempted to obtain a sample of blacks with high SAT scores who went to non-elite colleges. They commissioned the National Opinion Research Center to do a survey that presumably could have produced this data, but it did not.

Bias Problems

Instead Bowen and Bok compare all the blacks who went to the elite colleges with all black college graduates of the same age from the U.S. Census. Of course the blacks who went to the elite schools have much higher average earnings-81 percent higher. As an estimate of the impact of the race preference policies at the elite schools on black earnings, we know this comparison suffers from two important biases: (1) the blacks from the general college population have much lower SAT scores (nationally 70 percent of black college entrants score below 400 on the verbal SAT, while only 12 percent of the blacks who went to the elite schools did), and (2) the comparison lumps together blacks who would and would not have been accepted with race-neutral admission policies. (The group that would have been accepted presumably would have higher earnings.)

The authors' response to these bias problems is to cite the "literature" they reviewed which they say does control for differential ability. Thus in their entire book the best evidence they have on the most crucial issue relating to the race preference policies of their elite colleges comes from a review of existing studies! But given the authors' obvious biases in favor of preferences, how can one rely on their unsubstantiated assertions about the findings of these studies? The "findings" of this literature are not presented in any detail; even the select schools they covered are not given. The entire discussion takes up one paragraph. (P. 128.)

The other evidence on this crucial question that Bowen and Bok present are comparisons of their high SAT blacks who went to the most selective of the colleges within their highly select group of colleges with blacks who went to the less selective of them. They use multiple regression techniques to hold constant SAT scores and other variables and conclude that, for black males, there is about a 10 percent increase in their earnings from attending the most select schools rather than the least select. Bowen and Bok argue that blacks with a given SAT score who attend the more selective schools are being given more racial preference than those who attend the least select and, therefore, this is evidence that racial preferences lead to higher earnings.

There are two problems with this evidence. One is statistical. It is very likely that they were not able to adjust fully for ability differences between a 1000 SAT black accepted by the most select schools and a 1000 SAT black accepted by the less selective schools. After all, the most selective schools accept a small number of blacks with this SAT level while the less selective accept many more. So it is likely that any residual ability differential would be large enough to account for the 10 percent net effect claimed by the authors (this is also suggested by the fact that the partial regression coefficients for the degree of selectivity variables are not statistically significant).

The other problem is substantive and pragmatic. Do we want the country to keep a policy that is in direct conflict with one of the most treasured of our political principles-that the government should not favor any racial or religious group-and which could be having significant demoralization effects on blacks, simply because it might raise the earnings capacity of high SAT blacks by a few percentage points?

Conclusion

In sum, the Bowen and Bok study, for all its data and unending quantitative analyses, tells us nothing about the most crucial question connected with the use of racial preferences in college admissions. If all the preference policies that have been mandated on public colleges and universities were eliminated, what would happen to blacks? The trend of public opinion seems to be that blacks as a group would be better off. It is perhaps most ironic that there is some Bowen and Bok data that indicate that even their elite private colleges and universities are backing off their racial preference policies. This is suggested by the fact that the SAT distribution of black applicants in their 1976 entry cohort is much lower than the black applicants to these same elite schools in the 1989 admissions cohort. The overall black SAT distribution probably did not improve by enough over this period to explain this increase among elite school applicants.

Thus it is possible that the schools themselves have tightened up on how much preference they are giving blacks. Bowen and Bok would do a service if they would release more data so this hypothesis could be checked out.

OLD MAN QUOTA

by Roger Clegg

Confusion and Dishonesty

The confusion begins with the book's title, The Shape of the River. Mark Twain wrote that a riverboat pilot must "know the shape of the river perfectly." Are authors William Bowen and Derek Bok making an analogy to how perfectly a college admission officer must know the lives of those whose applications he is considering? Or to the twists and turns that a person's life takes as he goes from high school to college to graduate or professional school to the workplace? Either way, it is not a very good metaphor, although it nicely illuminates the elite's fatal conceit that it can know and manage everything. Merely selecting students who will make good sociologists or engineers is not challenging enough. Through the college admissions process, Bowen and Bok see themselves as engineering lives and even society itself. "It is helpful, in our view, to think of admissions decisions as having many of the attributes of long-term investment decisions involving the creation of human and social capital." (P. xxvi n.4.)

The book's dishonesty also begins early and never ends. Throughout, the discrimination defended by the authors is euphemistically called "considering race" or being "race sensitive." But the authors themselves are race obsessed. They declare, "Neither of the authors of this study has any sympathy with quotas or any belief in mandating the proportional representation of groups of people, defined by race or any other criterion, in positions of authority." (P. 283.) But, in fact, that is exactly what they believe in. By the end of the paragraph, they are calling the country to action because "present racial disparities in outcomes are dismayingly disproportionate." (P. 284.)

The authors style their book as an honest contribution of much needed data to the debate over affirmative action. But the pose is a dubious. Advance copies of the book were circulated to the authors' allies but apparently not to those likely to be skeptical or critical, and a nationwide campaign to place favorable news stories and op-eds about the book followed-not the usual approach for honest scholarship. The data relied on are, to a large extent, not available to potential critics, and the secondary sources cited are frequently "forthcoming." The authors' assertion to the contrary notwithstanding, critics of preferences have marshaled considerable data, but Bowen and Bok generally ignore them. There is no mention, for instance, of the well publicized studies by the Center for Equal Opportunity. While Linda Wightman's study defending preferences by law schools is repeatedly cited, no mention is made of Stephan Thernstrom's and Gail Heriot's devastating rebuttals of that study. And so on.

Toward the end of the book, the authors make a disingenuous plea for institutional autonomy. Admission decisions are tricky, they say, and ought to be left to the experts. But surely the authors will concede a distinction-as a matter of both law and policy-between state and private schools. And for private schools, does this mean that Bowen and Bok now favor repeal of the Civil Rights Act of 1964? I doubt that's really what the authors want. More likely, they favor a regime where discriminating against some groups (whites, Asians) is permissible, but discrimination against other groups (blacks, Hispanics) remains flatly prohibited. This would literally "deny... the equal protection of the laws," to quote the Equal Protection Clause of the Constitution's Fourteenth Amendment, and shows that what Bowen and Bok really want is not autonomy as a matter of principle, but to allow what they want to do and prohibit what they don't.

Bowen and Bok do, however, make two concessions of enormous importance in their book. The first is that race is frequently a heavy consideration in college admissions-more than just the "tiebreaker" we are sometimes led to believe by the proponents of preferences. Blacks have an "appreciably greater chance than whites of being admitted"-no, "considerably greater." "In the upper-middle ranges of SAT scores, in particular, the admission probability for black applicants was often three times higher than the corresponding probability for white applicants." (P. 26.) Thus, one table in the book shows the relative likelihood of whites, then blacks, being admitted to a top-tier school with different SAT scores. The difference in odds is dramatic. For a white applicant with a combined SAT score in the 1250-1299 range, for instance, the odds of admissions are 25 percent; for a black, 75 percent. (P. 27.)

Of course it would be futile for Bowen and Bok to dispute the point that there are significant differences between the relative qualifications of their black and white admittees. They concede that the overall graduation rate at their schools was 79% for blacks, versus 94% for whites. (P. 56.) Further, the average rank of black matriculants was at the 23rd percentile, versus the 53rd percentile for whites. The difference is GPAs, Bowen and Bok concede, is "very large." (P. 72.) Despite this, Bowen and Bok assert near the end of their book, "The data assembled in this volume should dispel any impression that the abilities and performance of the minority students admitted to selective colleges and universities have been disappointing." (P. 256.) Really? Even when their "test scores and high school grades [are] substantially below those of their classmates"? (Pp. 256-57.) Bowen and Bok lamely contend that this simply "reflects the extraordinary quality of the white and Asian applicants who have been attracted to leading institutions in ever greater numbers." (P. 257.) All students are above average.

Diversity Justifies Discrimination

The second major concession made by Bowen and Bok is that the justification for racial preferences can no longer be based on the existence, past or present, of discrimination. Rather, they pin their hopes on the diversity rationale. (Pp. 7, 283.) This is a key concession for two reasons. As a policy matter, the diversity rationale is extremely weak. As a legal matter, the Supreme Court in recent years has recognized no defense for racial classifications except the remedying of past discrimination. Several courts of appeals have explicitly rejected the diversity rationale. Thus, Bowen and Bok are hinging their legal and policy defense on the late Justice Lewis Powell's 1978 opinion-joined by no other justice-in Regents of the University of California v. Bakke.

Bowen and Bok believe that diversity is essential because people of a particular skin color are needed to represent politically, sell to, manage and motivate, and treat medically those of like skin color. (Pp. 12-13.) They want diversity because it makes students "able to understand how others think and function." (P. 279.) They really believe that people should make gross generalizations, based on race, about how people "think and function."

Early on, Bowen and Bok list two considerations deemed important for higher education that they view as justifying the use of racial preferences: (1) the desire for a student body "with a wide diversity of backgrounds, experiences, and talents" that will "contribute to the development of their fellow students"; and (2) the need "to attract students who seem especially likely to utilize their education to make valuable or distinctive contributions to their professions and to the welfare of society." (Pp. 24, 25.) Let us assume that these two considerations are valid and that admission officers might have some hope of predicting which students will meet them. It remains utterly baffling why race should be used as a proxy for either one. "Of course, it would be an oversimplification to assume that all African Americans, any more than all midwesterners or all Lutherans, represent anything resembling one point of view," Bowen and Bok concede, about two hundred pages later. "The scope of our study, however, is too broad to allow us to investigate such finer grained levels of diversity." (P. 219.) So, the authors realize they are stereotyping, but justify doing so because ... what? Imagine their horror if an employer acknowledged that not all blacks were shiftless and lazy, and justified his generalization to the contrary only with some arm-waving and an opaque assertion that his data were "too broad to investigate such finer grained levels."

If someone has overcome obstacles put in his path because of his race, that can be considered, just as it can be considered that he overcame obstacles put there for some other reason. But Bowen and Bok are dishonest in suggesting that this is all they want to do. (P. 279.) There is no particularized inquiry into whether a black applicant has or has not been discriminated against. Instead, preferences are awarded to these 18-year-olds-born not into slavery or the Jim Crow era, but in 1980-because they have a particular skin color, period.

"The C&B [College & Beyond: the elite schools participating in the book's study] minority graduates with advanced degrees are the backbone of the emergent black and Hispanic middle class," assert Bowen and Bok. (P. 116.) But this is just an assertion; there is no data or evidence to support the claim. That is, there is no information about how much-quantitatively or qualitatively-of the black and Hispanic middle class is made of C&B graduates, let alone C&B graduates who received preferences. (Indeed, there is very little data in this book about Hispanics at all.) Bowen and Bok then melodramatically conclude that their graduates "have also gained the training that will allow them to offer medical services to traditionally underserved communities and give political leadership to struggling urban constituencies." (P. 116.) There are no data to support the claim that the best way to have improved medical care in the ghetto is by the use of preferences, let alone that the quality of political leadership there-or anywhere else-has improved because of Ivy League training.

Bowen and Bok argue that black and Hispanic professionals are needed as role models and to provide leadership, medical care and legal advice, and other networks in their communities. (Pp. 12-13, 116, 156.) The assumption, of course, is that not only is American society segregated, but there is no reason to suppose it will not continue to be. We should not expect white children to idolize black adults or black children to follow white role models, and we should assume that white professionals will not be interested in serving blacks. Certainly we cannot expect blacks to be led politically by anyone except other blacks.

Additional "evidence" of the value of diversity is drawn from a survey of the students themselves. Many of them said that being able to get along with people of different races and cultures was important, and that college helped them do this. Respondents were asked if they knew "well" two or more students of a different race. "We did not attempt to define what knowing someone 'well' meant, but rather hoped that our wording would encourage those being surveyed to ponder their answer, rather than merely give what might have been an easy and expected response." (Pp. 230-31.) As to whether the authors succeeded or not, who knows? Bowen and Bok conclude from their data that "a drop in the share of black enrollment from, say, 8 percent to 4 percent could be expected to reduce the percentage of white matriculants who knew well two or more black students from 61 percent to 53 percent." (Pp. 234-35.) That doesn't sound like the end of the world, and perhaps those 8 percent will meet some Asians, Oklahomans, or Appalachians. In all events, you don't have to go to Harvard to learn to get along with people of other races.2

If diversity can justify preferences in favor of certain groups that are "underrepresented," then surely it can justify negative consideration of those that are "overrepresented." There is simply no way to justify the former without justifying the latter, as the two authors know, given the unhappy history of the Ivy League's ceiling on overrepresented Jews. It is also unclear why prejudice will be fought, rather than fostered, by deliberately mixing less qualified blacks in with superqualified whites.

Damn, We're Good

The overarching tone of The Shape of the River is, "Hey, we college presidents are doing a great job!" But buried is this self-congratulatory book is a depressing premise. The SAT scores (and, to a lesser extent, high school grade point averages) for black students are so bad and will continue to be so bad, believe Bowen and Bok, that our top schools must award preferences to ensure their admission; so bad that-even if preferences aren't used-the median score for blacks admitted will always run significantly below that for whites admitted; so bad that the average black admittee given a preference is not far below the average black not given one; so bad that, controlling for economic status, poor whites greatly outscore poor blacks.

And the solution proposed by Bowen and Bok to all this is to pretend that black applicants are better qualified than they are. They justify this in part on political and practical grounds, but politically and practically the causes of black underperformance-high rates of illegitimate births, poor study habits, and bad public schools-are less likely to be addressed for so long as preferences paper over these realities. At the beginning of their book, Bowen and Bok write that the "river" started flowing at "the frozen moment in time when seventeen-year-olds . . . sat down with Number 2 pencils to take the SAT." (P. xxii.) Of course, it started flowing 17 years and nine months earlier. In a rare moment of political incorrectness, they acknowledge the importance of "parental involvement in their children's education" and even suggest that children may benefit from "the presence of a 'stay-at-home Mom.'" (P. 80.) How much of that support will there be in single-parent homes?

In classic post hoc, ergo propter hoc reasoning, Bowen and Bok imply that preferences are the reason for the increased numbers from the 1960s to the 1990s of black college students, law students, and medical students, as well as black professionals, executives, managers, administrators, physicians, and engineers-and even members of Congress and other black elected officials. (Pp. 9-10.) Mightn't the ban on discrimination in the Civil Rights Act of 1964 and the 1965 Voting Rights Act have had something to do with all this?

Bowen and Bok also find evidence of black success in the high number of graduates who go on to graduate programs and professional schools. "The major implication of this part of the analysis for race-sensitive admissions is that many black matriculants who would have been rejected under a strict race-neutral admission policy went on to earn advanced degrees, including professional and doctoral degrees." (P. 110.) But as Bowen and Bok grudgingly acknowledge: "One can interpret these data as a further indication that graduate and professional schools gave preference to black candidates from the C&B schools, and they undoubtedly did." (P. 114; see also pp. 44-46, 103.)

As critically important as they believe preferences to be, Bowen and Bok nonetheless try to reassure us that they are quite limited in scope. They assert that racial preferences are used by "only about 20 to 30 percent" of four-year colleges and universities, and that the other schools "accept all qualified candidates" and thus don't use racial preferences. (P. 15.) As Dave O'Neill points out, however, the definition of "qualified candidate" at many other institutions has been lowered to ensure sufficient "diversity" in the student body. And it is not very reassuring to learn that one out of every four colleges has a policy of formal racial discrimination.

The book is aimed in large part at college administrators, and the message to them is, "Stay the course!" Indeed, the message is to stay the course, no matter what some pesky judge might say the law is. It is "hard to imagine" private colleges and universities "passively accepting" a judicial mandate that racial preferences end, write Bowen and Bok, standing in the schoolhouse door. (P. 288.) They go on say that we might as well allow college administrators to use outright preferences, because banning them "is likely to bring forth ingenious efforts" to continue them surreptitiously in ways that might "have a wide variety of other consequences, not all of them benign." (P. 289.)

Bowen and Bok are not only stubborn, but elitist as well:

Shifting from race-sensitive admissions to class-based admissions, therefore, would substantially reduce the minority enrollments at selective institutions while changing dramatically their overall student profiles. Our data show that students with low socioeconomic backgrounds are less likely than students of equivalent ability from high socioeconomic backgrounds to complete their studies, attain professional or doctoral degrees, and earn high incomes. As a result, although a class-based system might reward applicants handicapped by poor schools, troubled neighborhoods, and similar burdens, it would surely hinder selective institutions in attempting to prepare the most talented minority students for eventual positions of leadership in government, business, and the professions. (P. 271.)

Unanswered Questions

The Shape of the River's principal thesis is: Blacks who get into good schools, and who graduate, do well in life, even though they got in by means of a racial preference. Not a big surprise. But the authors conclude from this that preferences are conducive to the general welfare and ought to be continued.

Assuming that there is something important to be gained by having greater "diversity" (a dubious assumption, as we have seen), there are three big questions that we would want answered before arriving at the authors' conclusion: (1) What would have happened to these students had they not benefited from racial preferences?; (2) What would have been the social contributions of the non-black students who would have been admitted instead, in the absence of preferences?; and (3) What were the other costs to society of the use of preferences? We will see that the authors try at length-though not completely satisfactorily-to answer the first question. Meanwhile, they all but ignore the second question and deal with the third very poorly indeed. It is as if someone wrote a defense of segregated universities that consisted of data demonstrating what fine contributions were made to society by their alumni, all the while ignoring the social costs of segregation and the foregone opportunities of black students-for themselves and the rest of us.

When, less than ten pages from the end of their book, the authors finally devote exactly seven sentences to what might have been lost from turning down the white and Asian students, they quickly point out that a lot of the white women would just be housewives, and then lamely conclude: "Would society have been better off if additional numbers of whites and Asian Americans had been substituted for minority students in this fashion? That is the central question, and it cannot be answered by data alone." (Pp. 282-83.)

The chapter on "Civic Participation and Satisfaction with Life" is likewise uncompelling. Bowen and Bok stress the high levels of civic participation by black matriculants of their school: they are "even more active than their white classmates." (P. 158.) But there is plenty of room for respondents to define the different kinds of "civic participation" in any way that they like; and the black-white gaps are quite small in most categories (and the gaps that do exist might be explained by the desire for black respondents to show that they are indeed giving something back to the community). It has been suggested that civic participation is inversely proportional to professional success ("I'm too busy to volunteer"), and so greater black civic participation may not be such a happy statistic. In addition, when national data are analyzed, it turns out that people who didn't go to the authors' elite schools "were more likely to participate in community and social service activities, youth organizations such as scouting and Little League, religious activities, and groups such as the PTA that operate within elementary and secondary schools." (P. 157.) Graduates of the elite schools, on the other hand, spend their time with "professional and trade associations, college-related functions such as fund-raising and student recruitment, cultural and arts activities, and environmental and conservation programs." (P. 157.) This suggests that perhaps what should be abolished is not preferences, but the elite schools themselves.

Hurting Black Students

Much of The Shape of the River is devoted to refuting a claim sometimes made by the critics of preferences, namely that such affirmative action has actually harmed its purported beneficiaries, since they are less likely to graduate from schools where they are less qualified than the rest of the student body than from schools where they aren't. That is by no means the only, or even a top, claim of those criticizing preferences, but it obviously stings Bowen and Bok. Being unfair to whites they can live with-but don't tell them they're hurting African Americans.

There is evidence from their data that those admitted because of preferences are as likely to graduate from the school granting the preferences as from a lower-tier school they might have gone to instead. (Bowen and Bok compare the graduation rates for blacks with similar SAT scores, GPAs, and income at the different schools.) But this is only because everyone is more likely to graduate from their particular top-tier schools. Once you get in, those schools try hard to keep you from leaving-harder than lower-tiered schools.

On the other hand, the data definitely do not show that the beneficiaries of preferences are as likely to graduate as other students: to the contrary, as noted above, they are indeed less likely to graduate. "At leading schools of law, business, and medicine, approximately 90 percent of black students complete their studies successfully," continue Bowen and Bok. (P. 257.) But, according to Stephan Thernstrom, a study "of the nation's top ten law schools found that the grades of the average black student fell in the bottom 8 percent of the class," and blacks admitted without preferences were three times less likely to fail the bar (a 9.8 percent failure rate versus 27.1 percent).3 Gail Heriot writes that "white medical students passed the National Board of Medical Examiners Part I exam at a rate of 87.7 percent, African Americans at 48.9 percent." Is this because of preferences? It would appear so: "matching academic credentials (in this case, the MCAT score and undergraduate GPA) made the discrepancy all but disappear."4 In another context, Bowen and Bok acknowledge that "[n]o one benefits" when black candidates are "'overpromoted' by firms or individuals too eager to 'do the right thing' or even 'to look good.'" (P. 284 n.8.) PhD's, heal thyselves.

Bowen and Bok argue that their data support the proposition that blacks at selective schools (who are likely to have received preferences) are happier than those at less selective schools (who probably didn't). Therefore, they say, a "mismatching" of students and colleges doesn't make the students less happy. But in this discussion they consider only the students who actually graduate; by excluding those who failed to do so, they are removing from consideration those who were most likely "mismatched" in the first place. (Pp. 198-99.) Later in the chapter, Bowen and Bok concede that "class rank is clearly related to regrets about choice of college." (P. 207.)

Bowen and Bok assert that preferences won't be abused because no one along the river is interested in unqualified people. (P. 117.) Thus, they claim that no black would be hired in the private sector if the company thought he or she were incompetent. Perhaps not, but certainly companies are willing to hire people who are less competent if doing so will placate the bean-counters in the federal government's civil rights bureaucracy, or avoid criticism from civil rights leaders like Jesse Jackson and Kweisi Mfume.

There are two additional caveats. First, it may not always be the case that preferences don't disserve those given them. In a study of the state of Washington by the Center for Equal Opportunity, for instance, the graduation rate for blacks admitted to the more selective school (the University of Washington) was lower than that for those admitted to the less selective school (Washington State University) in the same system: 29 percent versus 44 percent.5 Second, it is quite plausible that the students selected by top-tier schools were better qualified than the students selected by lower-tier schools, even if the students had similar SAT scores and grades. Bowen and Bok say as much elsewhere. (Pp. 58, 63.) After all, the top-tier schools had some reason for selecting one and not the other, right? If this is true, then maybe the graduation rate for students would have even higher at the second-tier schools-higher than those who actually went there, of course, and maybe even higher than the graduate rate at the top-tier schools.

I will offer Bowen and Bok this consolation though: I don't think it really disserves blacks students even if sometimes they are admitted to schools where they are less likely to graduate than other schools, so long as the students know the odds. After all, any student takes a risk when he goes to a tough school. If my son got into MIT and the local community college, I would not hesitate to urge that he attend the former, even though his chances of graduation are better at the latter. I'm sure most parents would agree.

Beating Up a Straw Man

Bowen and Bok also devote a great deal of time to demolishing a straw man-that the opponents of preferences would base all admission decisions solely on SAT scores (or, perhaps, SAT scores and high school grade point averages).

There is a widespread misconception that scores and grades represent the only truly valid considerations in deciding whom to admit to a selective institution. Thus, one often hears that students with top scores and the highest grades should be admitted "on the merits," as if these measures were the sole legitimate basis for admission and that other considerations were somehow insubstantial or even morally suspect. This is patently false. (Pp. 24-25.)

What is "patently false" is the assertion that there is any "widespread misconception" that only scores and grades are valid considerations or that anyone believes them to be the "sole legitimate basis for admission." Those opposing racial preferences would not require admission officers to ignore an applicant's extracurricular activities or teacher recommendations, non-academic accomplishments and hardships in his life, special talents or experience. Anything can be considered, so long as it's not race.

This straw man is an important distraction bec