August 1996 € Volume 23 € Number 4

Building Equal Opportunity On Firmer Footing

A Case Against Affirmative Action

By Arnold Barnett

We argue here against extreme affirmative action, which we define as preferential treatment based on ethnicity and/or gender that routinely results in the selection of less-qualified indiaviduals over more qualified ones.

There is nothing biased about this definition: supporters of present policies need merely assert that, thus characterized, extreme affirmative action scarcely exists in the U.S. Indeed, the definition is meant to acknowledge that some measures identified with affirmative action are clearly desirable (e.g., advertising jobs widely rather than filling them through old-boy networks), and that we will limit discussion to the policy's more questionable aspects.

Our case against extreme affirmative action rests on four propositions. Because issues of race have, for better or worse, been central to the ongoing national controversy, we will emphasize ethnicity over gender in what follows.

Proposition 1:
It appears that extreme affirmative action is widespread in the United States.

Last year, The New York Times published mean SAT scores for four groups of undergraduates admitted to the University of California-Berkeley in the fall of 1993:

Mean Scores (on scale 400-1600)
Asians	1293
Whites	1256
Hispanics	1032
Blacks	994

Nor is such ethnic diversity in SAT scores an idiosyncrasy of Berkeley. Using statistics from the Consortium on Financing Higher Education (COFHE), Herrnstein and Murray reported in "The Bell Curve" [1994] that, among students entering 26 high-prestige colleges in 1990-91, the average white-black disparity in SAT scores was 182 points. At none of the schools was the mean advantage for whites less than 95 points. (While many people (including this author) have criticized aspects of "The Bell Curve," no one has questioned the accuracy of data like these.)

Because of the exhaustively documented correlation between SAT scores and subsequent academic success, the pattern just described raises the specter of extreme affirmative action. Still, the tests are not perfect measures of relevant qualifications; the case against taking them too seriously in this context was well-put by the columnist Molly Ivins: "Here's the deal: A black kid makes straight A's at an inner-city school but comes in 100 points lower on the SAT than a rich white kid who went to the best prep schools. How many points on a standard test are grit and determination worth? If you had your choice, which kid would you want as your lawyer?" -- April 19, 1996

A hypothesis is lurking behind Ivins' comparison ("grit defeats disadvantage"), and follow-up data allow us to explore it. The simplest measure of college success is whether the student completes the program. Unfortunately, statistics on that topic are not reassuring: non-Asian minority students fail to finish college at much higher rates than do whites. At Berkeley, the difference is a factor of 2.5 (40 percent vs. 16 percent); at MIT, the ratio has been 1.7 (24 percent vs. 14 percent). Among the randomly-selected individuals in the National Longitudinal Youth Survey, 66 percent of blacks who started college had not received their degrees by age 29; the comparable figure for whites was 37 percent. While these data can be interpreted in various ways, they hardly demonstrate that the differences in academic qualifications suggested by SAT scores simply wither away once students reach the campus.

What is sorely needed to inform the affirmative action debate is a comprehensive cost/benefit analysis of the ethnic admissions strategies employed by elite colleges. How many of their intended beneficiaries actually gain from such policies? (Some clearly flourish, but what about those who dropped out of college when they might readily have succeeded at other schools?) What are the effects on "bystanders" displaced by the policies? It is easy to make light of their plight ("if he doesn't get into Yale, he can always settle for Harvard"), but such nonchalance may be unfair. To get into Berkeley in recent years, a white or Asian applicant has needed a GPA of 3.9 out of 4. Faced with such requirements, might not lots of talented majority-group students say "what's the use?" and work less hard? What are the long-term ramifications if large numbers of youths thus "underinvest" in their human capital?

In the workplace, it is often hard to quantify either performance on the job or qualifications prior to hiring. But there is reason to fear that extreme affirmative action is abundant in that setting. For many years, the U.S. Labor Department promoted "race-norming" of scores on standard aptitude tests used in job referrals. Under this policy, each applicant's performance was reported as a percentile of performances within her own ethnic group. This extreme manifestation of the quota mentality -- in which all groups have the same distribution of scores by definition -- meant that an absolute score of 300 was treated as a rating of 45 in one ethnic group and an 89 in another. And millions of people took these tests.

Race-norming was banned by Congress in 1991, but the spirit behind it may live on. In a 1994 handbook about the hiring of air traffic controllers, FAA managers were advised that: "The merit promotion process is but one means of filling vacancies, which need not be utilized if it will not promote your diversity goals."

A 1992 job posting by the U.S. Forest Service contained an extraordinary sentence: "Only unqualified applicants will be considered." (No, this was not a typo.)

Given that the government requires its many contractors to file frequent "goals and timetables" about affirmative action, one can easily imagine that attitudes and practices in the public sector carry over to the private one.

Let us be clear: we do not have data that prove irrefutably that extreme affirmative action pervades the United States. But neither are there persuasive data that make the contrary point. And there is enough circumstantial evidence of strong ethnic preferences that, faced with uncertainties about the national picture, we need not give the benefit of the doubt to the supporters of present arrangements.

Proposition 2:
The number of individuals who perceive they have suffered under affirmative action may greatly exceed the number who have actually suffered.

Affirmative action, in other words, may be a public-relations calamity that creates a large class of pseudo-victims. Consider the two statements:

(1) "I would have gotten that job had I been a member of a minority group."

(2) "I would have gotten that job had it not been for affirmative action."

Are these statements equivalent?

One might initially think so, but a simple example makes clear that they are not. Suppose that an organization gets 100 applications for 12 identical positions, and is able to rank the applicants from 1 (highest) to 100 based on their qualifications. It winds up hiring the three highest-ranked minority candidates (whose rankings were 5, 14 and 33) and the nine highest-ranked majority candidates (whose ranks were 1-4 and 6-10).

Assume that, without affirmative action, the firm would simply have hired candidates 1-12.
In this situation, there are exactly two majority-group candidates (namely, 11 and 12) who can legitimately claim that affirmative action deprived them of job offers. But each of the 21 majority candidates with ranks from 11 through 32 (11-13 and 15-32) can correctly assert that, had he been a member of a minority group, he would have been hired in preference to number 33. Anyone making this last claim, of course, tacitly assumes that, while he has become a minority-group member, all higher-ranked members of the majority group stay there.

Statements (1) and (2) differ drastically in their implications: there are over 10 times as many "reverse discrimination" victims under the latter standard as under the former. Which number is more appropriate? Presumably the lower one, because the higher depends on a "partial derivative fallacy," under which the applicant and the applicant alone gets to shift ethnic groups. Absent such a metamorphosis, almost all rejected applicants would have fared no better under ethnic neutrality than under affirmative action.

But if there are exaggerated beliefs about the displacement caused by affirmative action, then why not go after the misconceptions instead of the policy? Alas, that is easier said than done. A surprising number of intelligent people are befuddled by even simple mathematical arguments. Moreover, those rejected for jobs rarely learn their standings within the applicant pool, and, on a matter so closely tied to self-esteem, many of them may believe that they just missed getting an offer. Thus, even if all applicants in the case above understood that only candidates 11 and 12 were disadvantaged by affirmative action, fully 30 of those rejected might feel certain that their (undisclosed) rank was either 11 or 12.

We should not underestimate the bitterness of those who see themselves as affirmative-action victims. The Columbia Journalism Review recounts that, when the Houston Post folded, reporters making inquiries about other jobs were told that none were available but then quickly asked, "Are any minority reporters looking?" Some of us have heard similar answers when we asked about assistant professorships for our Ph.D. students. In today's economic environment, job applicants who encounter such reactions may believe that affirmative action has deprived them of not just particular job offers, but of any chance of pursuing careers in fields like journalism or academia. Such treatment may strike them as incomprehensible when they have never discriminated against anyone, and when they are too young to have any recollections of segregated lunch counters, police dogs or church bombings, let alone lynchings.

And such personal frustrations can have broader consequences. They can harden attitudes on other policy issues more closely related to minority progress than is extreme affirmative action. (Perhaps it is no coincidence that the April 19, 1996, issue of Time has a cover story titled "Back to Segregation.") If so, the indirect effects of affirmative action may harm disadvantaged minorities more than its direct effects could ever help them.

Admittedly, this argument is rooted in pragmatic considerations rather than fundamental questions of justice. But, when the stakes are so high, we should hesitate to scoff at pragmatic arguments.

Proposition 3:
The official case for affirmative action is empirically weak and conceptually loose.

Among official defenses of affirmative action, perhaps the most important is the 1995 review of that policy prepared for President Clinton. (It is the first document listed at a major affirmative-action web site.) The grave problems with that document heighten misgivings about the policy it attempts to defend.

The report to the President justifies affirmative action by citing statistics about the corrosive effects of past and continuing discrimination. These statistics systematically fail to make their intended point. Three examples:

1. The report states that "the average income for Hispanic women with college degrees is less than the average for white men with high school degrees." At face value, that disparity might suggest discrimination, but it can hardly be accepted at face value. Age is clearly correlated with income, and the men in the survey might on average be several years older than the women. Many of the women may have entered professions like teaching that, rightly or wrongly, have for a long time yielded lower salaries than various blue-collar jobs. Does the statistic distinguish full-time from part-time work? Differences in cost of living in different parts of the country? One wonders whether any of the original discrepancy would survive the appropriate adjustments.

2. The report also notes that "in 1993, less than 3 percent of college graduates were unemployed, but whereas 22 percent of whites had college degrees, only 12 percent of African-Americans and 9 percent of Hispanics did." The numbers are presumably accurate, but what point are they trying to make? That employers are insisting on college degrees in situations in which they aren't really needed? That colleges are discriminating against minorities in admissions and other processes? To put it simply, why are these statistics more an argument for affirmative action than for (say) urgent efforts to improve inner-city high schools?

3. The report cites "audit studies" in which people of different ethnicities but otherwise identical credentials apply for the same job. Such studies are meant to provide direct estimates of the prevalence of illegal discrimination. An audit in the Washington, D.C., area, for example, estimated that whites advanced further than their black counterparts in 20 percent of the pairings. (The report did not note instances in which the black applicant got further than the white, a phenomenon that sometimes appreciably reduced the net white-black disparity.)

These audit studies fall short of classic double-blind scientific experiments. "Applicants" in an audit know its purpose, and may have some prior beliefs about the persistence of hiring discrimination. It is not to question the integrity of the participants to suggest that, perhaps unconsciously, their beliefs might cause subtle changes in their behavior (e.g., during interviews), and might thereby compromise the matched experimental design.

Beyond statistics, many theoretical arguments advanced for affirmative action are weakened by an unwillingness to make salient distinctions. One can plausibly assert, for example, that ethnic diversity is important in an undergraduate student body. But is the argument equally potent for a graduate program in chemical engineering? Given a long history of distrust between minorities and police, minority-group status might be a legitimate asset for an aspiring police recruit. But does the same case apply without modification for open-heart surgeons or air-traffic controllers?

The president's report assures us that affirmative action is not meant to benefit "unqualified" job applicants. (One hopes that the Forest Service takes note.) But job seekers can rarely be dichotomized into the qualified and the unqualified; much more commonly, the credentials of candidates fall on a continuum. The real issue is whether marginally qualified candidates should sometimes get preference over better qualified ones.

And, while it is widely agreed that affirmative action should not last forever, advocates are vague about when and how it should end. Attempts to replace gender/ethnic preferences with those based solely on economic disadvantage have met strong resistance. Even a black youth from Scarsdale deserves special consideration, it is contended, because the effects of three centuries of relentless prejudice are not erased by a few years in the suburbs. Granted, this position deserves respect, but is there any sense about how long it can be invoked?

As advanced, many conceptual arguments for affirmative action seem imprecise and unbounded. That ambiguity can only increase uneasiness about the way the policy is being implemented.

Proposition 4:
There is no inconsistency between opposing extreme affirmative action and yet believing that the policy has been "good for America," and that an instantaneous move to "color-blindness" would be unwise.

Those who invoke pragmatic arguments for moving away from affirmative action should acknowledge that similar arguments can be made in its behalf. After centuries of discrimination, perhaps the policy was a necessary "shock therapy" to persuade minorities that the nation was fully committed to racial equality. One can sympathize with recent victims of reverse discrimination, but recognize that they were proceeded by far more numerous victims of direct discrimination.

At the same time, extreme affirmative action need not be extinguished at once, with no thought about which policies might replace it. (Speaker Newt Gingrich has made this point more than once.) Highly abrupt changes in policy could be demoralizing if not alarming to minorities, and might undo much of the goodwill the "shock therapy" was meant to create.

But, much as continuation of extreme affirmative action may have stunted other civil-rights initiatives, its diminution could have the opposite effect. It is noteworthy that Berkeley -- which is slated to end ethnic preferences in admission -- plans to work with high schools in impoverished areas to prepare students there to meet its standards. Given that these schools are disproportionately black and Hispanic, the policy is not colorblind in any meaningful sense. And, if it succeeds, it could produce an ethnically diverse campus on a much firmer footing than one tied to 250-point "discounts" on SAT scores.

Everyone in INFORMS passionately supports equal opportunity. But we must not flinch from asking whether the best way to achieve equal opportunity in the 21st century is to phase out affirmative action at the end of the 20th.

Arnold Barnett is a professor of operations research at MIT's Sloan School of Management.

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com