Racial and Ethnic Preference

Thomas J. Kane and
Thomas J. Kane Walter H. Gale Professor of Education and Economics - Harvard University
William T. Dickens
William T. Dickens University Distinguished Professor of Economics & Social Policy - Northeastern University

November 1, 1996

In the aftermath of the California Civil Rights Initiative vote, many more states are likely to reconsider the use of racial and ethnic preference in college admissions. This Brief summarizes the best available evidence on two issues vital to that burgeoning debate: the true extent of racial preference in college admissions and its impact on the careers of the intended beneficiaries. While the evidence of racial preference in admissions is strong at elite universities (those with average SAT scores in the top 20 percent), racial preference is less evident outside the elite sector. Despite the hopes of supporters of the CCRI and the fears of its opponents, the end of racial preference will have little impact on the college-going prospects of most high school students. But, contrary to the assurances of many of its opponents, racial preference does not do more harm than good for minority youth. Rather, selective institutions seem to enhance the earnings prospects and raise the college completion rates for both minority and nonminority youth who are admitted. Although this need not mean that the benefits of affirmative action exceed the costs, ending affirmative action is not likely to be a painless step for minority youth. Rather it is likely to lead to some redistribution of social benefits away from them. Finally, we discuss reasons why the perceived costs of racial preferences may be exaggerated and describe the conditions under which racial preferences could be an appropriate remedy for labor market discrimination.


As the labor market bestows larger rewards for educational attainment and the competition for admission at elite universities becomes more keen, racial and ethnic preferences in college admissions have become increasingly controversial – particularly at public institutions. In the summer of 1995, the Board of Regents of the University of California voted to end racial and ethnic preferences in college admissions, and in the spring of 1996, the Fifth Circuit Court of Appeals rejected the consideration of racial or ethnic identity for the purpose of promoting “diversity” on college campuses in its jurisdiction. The Supreme Court’s refusal to revisit that case has left college administrators in limbo, waiting for a legal standard yet to be defined.

As the labor market bestows larger rewards for educational attainment and the competition for admission at elite universities becomes more keen, racial and ethnic preferences in college admissions have become increasingly controversial – particularly at public institutions. In the summer of 1995, the Board of Regents of the University of California voted to end racial and ethnic preferences in college admissions, and in the spring of 1996, the Fifth Circuit Court of Appeals rejected the consideration of racial or ethnic identity for the purpose of promoting “diversity” on college campuses in its jurisdiction. The Supreme Court’s refusal to revisit that case has left college administrators in limbo, waiting for a legal standard yet to be defined.

While the country awaits clarification of the legal issues, policymakers ought to be asking whether racial and ethnic preferences are, nevertheless, worthwhile. Unfortunately, the strong emotions swirling around the issue have obscured the facts and undermined careful inquiry. The purpose of this policy brief is to clear away some of the misconceptions that are common in arguments about this issue and to present the best available evidence on the true impact of preferences. Although one’s response may differ depending upon subjective beliefs about fairness and the intangible value of greater racial equality, our intent is to simply pose the question: Do the benefits of affirmative action in college admissions justify the costs?

What Is the True Extent of Racial Preference in College Admissions?

Because college admissions committees pursue their work in secret, it is difficult to know much about the extent of racial preference in college admissions. Indeed, as we show in our forthcoming book, many of the statistics cited in the current debate can be quite misleading. For example, Richard Herrnstein and Charles Murray (authors of The Bell Curve) present differences in the average test scores of admitted students as evidence that whites and Asians face massive reverse discrimination. However, given differences in test scores and other characteristics among racial groups in the population, one might expect to see some differences in the mean characteristics of accepted students in different colleges even if admissions were race-blind. Test scores are only one of several indicators of academic potential considered by admissions officials. Therefore a completely race-blind admissions process could produce large differences in the mean test scores of admitted students as long as disadvantaged groups do worse on tests than they do on other measures of academic potential.1

The best way to uncover the extent of racial preference in higher education is to follow students from different racial and ethnic groups – with otherwise similar characteristics – and see how they fare when they apply to the same colleges. Using a sample of roughly 5,000 students from the high school class of 1982, Thomas J. Kane has studied differences in the likelihood of admission among blacks, Hispanics, and whites with similar test scores, high school grades, and family backgrounds.2 The data reveal not only the extent of racial preference in college admissions but also its effects on the early labor market careers of the students, who were interviewed again in 1992.

The table reports racial and ethnic differences in the probability of admission for the typical four-year college applicant at different types of colleges. At the most selective four-year colleges – those with reported mean test scores in the top 20 percent of all four-year schools – blacks and Hispanics from the class of 1982 enjoyed a large advantage. For example, students with the average characteristics of those applying to a four-year college had a 60 percent chance of being admitted at an elite four-year school if they were white non-Hispanics. However, black or Hispanic applicants with the same characteristics had an 87 or 75 percent chance, respectively. At the most selective schools, race or ethnicity “mattered” about as much as a grade point average that was one letter grade higher or twice as much as being a member of the student government, and was equivalent to several hundred points on the SAT.


However, preference in admissions is much less evident at nonelite schools, where 80 percent of four-year college students enroll. At such schools – where the average four-year college applicant would have more than a 90 percent chance of acceptance – acceptance rates for black, Hispanic, and white students with the same characteristics differed by only a couple of percentage points, if at all. Indeed, these differences in the probability of admission were not statistically distinguishable from zero. Racial and ethnic preference in college admissions is sometimes large – but it is primarily restricted to elite institutions.

Yet students and parents using Herrnstein and Murray’s metric may still have concluded that racial preference was common at nonelite schools. Within these schools, the average black or Hispanic student had a combined SAT score 125 points below other youths admitted to the same institution. (The difference at elite schools was only slightly larger, at 159 points.) Despite apparently race-blind admissions, the student bodies at nonelite schools reflect the differences in test scores that exist in the population of all high school students. This should not be surprising, since nearly everyone who applies is accepted at most four-year colleges.

Presumably because of the racial preference practiced by elite schools, blacks do attend better colleges than whites with similar characteristics. However, even among those attending predominantly white institutions, the difference by race in the “quality” of college attended by those with similar test scores and high school grades is small. On average, blacks attended schools where mean SAT scores for all students were less than 55 points higher than those attended by whites with similar qualifications. That is a relatively small difference. For comparison, the mean SAT score of Harvard undergraduates in the early 1980s exceeded the mean for Yale undergraduates by over 70 points. The difference between Harvard and the median four-year college was just over 400 points.

If the Effects of Preferences Are Limited, Why Are They Perceived as Large?

Whatever the true extent of racial and ethnic preferences, their perceived importance is likely to be much larger. Whites and other “nonpreferred” groups probably underestimate how widely any costs of racial preference are likely to be shared. Parking for disabled drivers provides a useful analogy.3 Suppose that there was one parking space reserved for disabled drivers in front of a popular restaurant. Even though eliminating the reserved parking space would have only a minuscule effect on the average search for a parking space by nondisabled motorists, the sight of an open space might frustrate many passing motorists. If too many mistakenly believe that they would be parked if the space had not been reserved, the sum of the costs perceived by each individual passing motorist will be greater than the true costs.

For example, at Harvard, only about one in ten applicants is accepted to the undergraduate college. Many of the rejected applicants (and many more of those who did not bother applying) may falsely believe that they would have been accepted at Harvard if there were no racial preferences. Yet only 15 percent of the undergraduate student body is made up of blacks and Hispanics. Even if racial preferences ended and all of these students were forced to surrender their seats (this is clearly an overstatement, since some of these students would have been admitted without racial or ethnic preferences), the college would make room for only one or two more students for each one hundred that apply. If more than one or two of the ninety that were originally denied admission are convinced that they would have been admitted without racial preferences, then the perceived costs will overstate the true costs of reserving the space.

Do Group Preferences Do More Harm than Good for Minority Groups?

The most devastating charge by opponents of group preferences is that they actually hurt the intended beneficiaries by enticing minority students to attend schools that are too demanding. According to this argument, racial preferences are at the root of high dropout rates and poor academic performance among minority students. If this were true, the decision to stop racial and ethnic preferences would be an easy one for college administrators, since ending it would benefit minority and nonminority youth alike. However, as usual, the choice is not likely to be so simple.

A widely cited article highlighted the high dropout rates of those admitted under the affirmative action program at Berkeley as evidence for the negative effects of affirmative action.4 However, such comparisons say virtually nothing about the effects of racial preference policies. The relevant question is not how well these students fared relative to the average student at Berkeley, but whether these students fared worse at Berkeley than they would have at a less selective institution.

Almost by definition, those admitted because of group preference will have weaker educational backgrounds than the average student. Studies show that students with weaker educational backgrounds can be expected to have lower grade averages and to be more likely to drop out no matter what school they attend. Moreover, African American students have lower college grades than white students with similar test scores and high school grades – wherever they attend school.5 Thus higher dropout rates for students granted preferences at elite schools need not mean that those students would be better off attending less demanding schools. We need to compare how similar students attending different schools fare.

Most research suggests that for both blacks and whites just attending (not necessarily completing) a higher-quality college is associated with higher earnings later in life – even after holding constant all available measurable characteristics such as high school grades, standardized test scores, and parental income and education.6 Such research also typically finds that both blacks and white receive similar payoffs to college quality.7 In fact, one recent study reports that the payoff to college quality is higher for blacks than for whites.8

Moreover, attending a higher-quality school need not lead to a higher dropout rate. Selective schools may offer students access to brighter classmates and more valuable learning experiences, which provide students with more motivation to stay in school. One study found that for both blacks and whites, attending a school with a higher mean SAT score is associated with higher – not lower – completion rates, even among those with similar high school performance and test scores.9 Another paper found that attending a school with a higher median test score is associated with higher completion rates for whites but lower completion rates for blacks.10 However, this anomalous result for blacks may be due to the fact that the authors do not separately identify those attending historically black institutions, which tend to have low mean test scores and higher completion rates. This makes it seem that attending schools with low mean test scores is good for blacks, when it is attending historically black institutions that provides the advantage.

At the same time, although the historically black institutions often have low mean SAT scores, they do seem to offer special opportunities for minority students. Compared with institutions with similar mean SAT scores, graduation rates are higher, and some research shows that those who attended historically black institutions have postgraduation earnings approximately 8 to 11 percent higher than similar youths attending the average predominantly white institution.11 But, as large as these benefits may be, the increment in earnings associated with attending a historically black institution is no larger than that associated with attending an institution witha mean SAT score 100 points higher.12 Attending a historically black institution seems to offer unique advantages to minority youth, but being admitted to a more selective institution also offers advantages to both blacks and whites.

Are Group-Based Preferences an Inappropriate Remedy for Discrimination?

A recent book gave voice to a common complaint about affirmative action: that an acceptable remedy for discrimination should either impose costs on known perpetrators or offer compensation to identifiable victims.13 Since affirmative action in college admissions rarely satisfies either criterion, the author argues that it is an inappropriate remedy for past or current discrimination. While superficially compelling, this argument is not so clear in the case of racial discrimination. Although we know of no one who would argue that blacks currently face substantial discrimination against them in admission to higher education, they may face discrimination in the labor market. We might wish to pursue affirmative action in higher education to compensate for labor market discrimination.

Take a simple example: suppose that 1,000 pairs of white and black men with similar qualifications each apply for jobs at 1,000 different establishments chosen at random. Suppose that the white applicants were offered jobs at 80 percent of them and the black applicants were offered jobs at 20 percent of them. It would be safe to conclude that discrimination had occurred, since the probability of such a large difference occurring by chance in the aggregate is quite remote. However, since any given employer made only two hiring decisions and each applicant made only one application, it would be nearly impossible to discern which applicants had been the subject of discrimination and which employers had discriminated. Even without discrimination, each applicant would have been quite likely to be turned down purely by chance. In fact, studies very similar to this have been performed. One of the best found that whites received 50 percent more job offers than blacks.14 Unless we are willing to conduct many audits of each employer and observe many applications by every candidate, we are likely to be left with a quandary: we can have strong evidence that a crime had occurred, but have no clear identification of specific perpetrators or victims.

Is it just to leave the victims uncompensated even if we can not identify them or the perpetrators individually? This is a fundamental problem implicitly faced every day by our justice system. Applying the standard of “innocent until proven guilty” is appropriate in a criminal trial, primarily because the costs of an unjust conviction are so much higher than the costs of an improper acquittal. But when combating the effects of lingering discrimination, setting such a high burden of proof may leave too many victims uncompensated. Preferences impose a cost on the members of nonpreferred groups, but the magnitude of these costs must be weighed against the injustice that must be endured by the victims of known discrimination who, as a practical matter, are nearly impossible to identify individually.

Moreover, the victims of discrimination need not be limited to the specific individuals unfairly denied a job. With enough discrimination, the market wage for all black males could be affected by discrimination against a few individuals, since nondiscriminating employers would have the luxury of choosing from a flood of black applicants. The larger the number of discriminating employers, the larger will be the fall in the wage that nondiscriminating employers will have to pay. In that event, the group may be forced to pay a price even when only a subset of individuals is directly subjected to discrimination. Similarly, whites as a whole can benefit from discrimination even if specific individuals do not discriminate or do get one of the jobs from which blacks have been excluded.

Therefore, for at least two reasons, it may be appropriate to offer group-based remedies: when individual acts of discrimination are too difficult to identify with any certainty and when the effects of individual acts of discrimination are reflected in the market wage for the group.

One clear difficulty with using racial preferences to compensate for discrimination is that it is hard to define a “stopping rule.” When do acts of discrimination recede far enough into the past or become so rare as to become irrelevant? The answer may be that, as long as audit studies continue to turn up evidence of substantial racial discrimination in labor and housing markets, one could make an argument for the use of group-based remedies, operating through the educational system, to allow minority youth to better equip themselves for a labor market in which they will be at a disadvantage.15


The legal standing of the “diversity” rationale for group preferences is in limbo. Justice Powell first advanced the “diversity” rationale in college admissions in the Bakke decision as a justification for racial preference policies even when an institution can not demonstrate evidence of discrimination in its recent past. Justice Powell’s reasoning did not earn the endorsement of a majority of the court: a majority of justices could only agree that simple quotas were not acceptable, but they could not agree on a definition of when seemingly more subtle racial preferences were acceptable. However, his rationale has served as the primary legal justification for racial preference in college admissions for more than a decade and a half. Recently, in its decision in the Hopwood case, the Fifth Circuit Court of Appeals seemed to dismiss Powell’s reasoning and adopted a much tougher standard for acceptable racial and ethnic preferences. The final resolution of this issue will have to wait until other cases involving the diversity rationale reach the Supreme Court. However, it is certainly worth noting that if the Fifth Circuit’s standard is upheld, a vast majority of public institutions will have a difficult time making the legal case in defense of their racial preference policies.

Independent of the legal questions involved, we as a nation ought to be asking ourselves whether group preferences in college admissions serve a worthwhile social purpose at an acceptable social cost. Although there has been surprisingly little systematic effort to assess their impact, there is some evidence that racial and ethnic preferences in college admissions benefit the groups they are intended to help. Whatever the merits of ending affirmative action in college admissions, such a stroke is likely to lead to some redistribution of the social pie.

Audit studies continue to suggest that discrimination lingers in the labor market. However, it still remains to be seen whether the benefits of preferences are worth the costs being imposed on the youth in nonpreferred groups. As we argued above, the public choice may be distorted by the fact that whatever costs are being born by nonminority youth, the perceived costs are probably larger, simply because too many students are likely to believe that they are the ones who would have been admitted if racial preferences were ended. Rather than respond to such misperceptions, policymakers and voters should take a sober look at the actual benefits and costs before they take the step of ending racial preference in college admissions.


1 William T. Dickens, Thomas J. Kane, and Charles L. Schultze, Does The Bell Curve Ring True? (Brookings, forthcoming). [Back]

2 Thomas J. Kane, “Racial Preferences in College Admissions,” Brookings, May 8, 1995. [Back]

3 We thank George Akerlof for this analogy. [Back]

4 John Bunzel, “Affirmative Action Admissions: How It ‘Works’ at the University of California-Berkeley,” Public Interest (Fall 1988), pp. 111-29. [Back]

5 Anne T. Cleary, “Test Bias: Prediction of Grades of Negro and White Students in Integrated Colleges,” Journal of Educational Measurement, vol. 5 (Summer 1968), pp. 115-24; James Crouse and Dale Trusheim, The Case against the SAT (University of Chicago Press, 1988); Michael Nettles, A. Robert Theony, and Erica J. Gosman, “Comparative and Predictive Analyses of Black and White Students’ College Achievement and Experiences,” Journal of Higher Education, vol. 57 (May-June 1968), pp. 289-318; Kane, “Racial Preferences in College Admissions”; Frederick E. Vars and William G. Bowen, “SAT Scores, Race and Academic Performance: New Evidence from Academically Selective Colleges and Universities,” Mellon Foundation, June 25, 1996; Robert Klitgaard, Choosing Elites (Basic Books, 1985); and Leonard Ramist, Charles Lewis, and Laura McCamley-Jenkins, Student Group Differences in Predicting College Grades: Sex, Language and Ethnic Groups (New York: The College Board, 1994). [Back]

6 Dominic J. Brewer, Eric Eide, and Ronald Ehrenberg, “Does It Pay to Attend an Elite Private College? Cross Cohort Evidence on the Effects of College Quality on Earnings,” Working Paper 5613 (Cambridge, Mass.: National Bureau of Economic Research, June 1996); and Estelle James and others, “College Quality and Future Earnings: Where Should You Send Your Child to College?” American Economic Review, vol. 79 (May 1989, Papers and Proceedings, 1988), pp. 247-52. [Back]

7 Kane, “Racial Preferences in College Admissions.” [Back]

8 Kermit Daniel, Dan Black, and Jeffrey Smith, “College Quality and the Wages of Young Men,” University of Pennsylvania, Wharton School, June 1995. [Back]

9 Kane, “Racial Preferences in College Admissions.” [Back]

10 Linda Datcher Loury and David Garman, “College Selectivity and Earnings,” Journal of Labor Economics, vol. 13 (April 1995), pp. 289-308. [Back]

11 Ronald Ehrenberg and Donna Rothstein, “Do Historically Black Institutions of Higher Education Confer Unique Advantages on Black Students?” in Ronald Ehrenberg, ed., Choices and Consequences: Contemporary Policy Issues in Education (Ithaca: ILR Press, 1994); and Jill Constantine, “The Effects of Attending Historically Black Colleges and Universities on Future Wages of Black Students,” Industrial and Labor Relations Review, vol. 48 (April 1995), pp. 531-46. Both Ehrenberg and Rothstein and Constantine report significant impacts of historically black colleges on completion rates. However, only Constantine reports positive wage effects as well. [Back]

12 Kane, “Racial Preferences in College Admissions.” [Back]

13 Richard D. Kahlenberg, The Remedy: Class, Race, and Affirmative Action (Basic Books, 1996). [Back]

14 Margery Austin Turner, Michael Fix, and Raymond Struyk, Opportunities Denied, Opportunities Diminished, Report 91-9 (Washington: Urban Institute, 1991). [Back]

15 Turner and others, Opportunities Denied. [Back]