August 1996 € Volume 23 € Number 4

A View from the Ivory Tower

Inefficient, unfair afirmative action policies must be reformed, but the costs pale in comaprison to the waste of human potential among the disadvantaged

By Harold Pollack

Is there anything left to say about affirmative action? Merit, diversity, quotas -- the codewords drop into the word processor without conscious thought. Every argument, on every side, was presented in someone's brief in the Bakke case 20 years ago. Reading "The Bell Curve" and the deeply flawed White House affirmative action report, one might think that public debate has declined over the past 15 years.

Rather than weigh in with my own opinions, I will take the analytic high road by presenting four propositions informed by the affirmative action debate. Some readers may regard this rhetorical stance as mere subterfuge, a sly way to present partisan argument in the guise of dispassionate analysis. How cynical our profession has become.

Proposition 1:
Elections and public sentiment matter. This makes scholarship more, rather than less, important.

Until a few years ago, the conventional wisdom among supposed experts was that affirmative action was good fodder for political campaigns, but that the actual policies would proceed, basically unaffected by the larger public debate. Ending affirmative action as we know it -- like candidate Bill Clinton's pledge to end welfare as we know it -- was viewed as a pipedream by those in the know.

It is easy to see why smart people thought that affirmative action would never really change. Affirmative action was never popular among the electorate. Yet a solid elite consensus held that racial and gender preferences were needed. Such policies seemed the cheapest way to keep civic peace without enacting more drastic policies to serve disadvantaged groups.

Like welfare, affirmative action was entrenched within specialized bureaucracies. Bob Dole and other moderate Republicans supported gradual expansion of many affirmative action programs. Legislation such as the Americans with Disabilities Act brought such policies into new domains with bipartisan support. Companies such as MacDonald's viewed affirmative action as an expression of corporate citizenship -- one that nicely matched their desire to maintain good relations with important consumer groups.

What a difference elections and new judges make. Political opposition to affirmative action has taken on new potency. In California, a popular ballot initiative proposes to essentially outlaw affirmative action in education, employment and contracting.

Affirmative action is also under fire in the courts. Several recent decisions question the basic foundations of race and gender-conscious policies. Most recently, a federal appeals court sent a shock wave through higher education with Hopwood v. Texas.

In remarkably blunt language, the court ruled that race-conscious admissions policies at the University of Texas Law School were unconstitutional. Especially ominous to timid administrators, the court added: "(If) the law school continues to operate a disguised or overt racial classification system in the future, its actors could be subject to actual and punitive damages."

So the political and legal foundations of affirmative action are open to debate. Most readers of this magazine claim no special expertise about the moral and political questions affirmative action engages. However, we do know about important issues of measurement and implementation that concern voters and policymakers alike: What is the impact of affirmative action on black wages and employment? What happens to people who might be harmed by affirmative action? Where does affirmative action create excessive disparities in entry qualification, and how would we know?

Careful, empirically grounded analysis often enriches public debate by critically testing widespread assumptions that often turn out to be wrong. Ask yourself: Is the typical black or Latino college student an affirmative action recipient? The clear answer is no. Research by Thomas Kane indicates that affirmative action is aggressively pursued at the top tier of undergraduate colleges, but that preferences are basically irrelevant at most colleges. The fact is, most colleges are non-selective. They admit most students who choose to apply.

Proposition 2.
Beware over-generalization from extreme cases.

Often in life we draw inferences from extreme cases, and we do so for two good reasons. Outliers don't cost much to observe. They also carry real information. As Michael Walzer once wrote, the Nazis were an extreme case, but not an imaginary one.

The utility of this approach is clear from a contrived example. Suppose I am lecturing to 200 students, half of whom are Pakistani and half are Korean. During my talk, my mind unaccountably wanders to the question of which is the taller group. I can't stop to calculate the relevant group means, but I can find the tallest person in the room and assume that he came from the tallest group. In fact, this crude method works pretty well. If heights are normally distributed and the true population difference in mean height is one inch, I will be right about 75 percent of the time without expending any real effort (assuming that both groups have a standard deviation of four inches). One does much better by picking the group with two of the top three.

Similar algorithms are often useful in evaluating public policies. Yet outlier-based reasoning is dangerous when it neglects two important points.

  • First, programs that work for some populations might fail for others. You shouldn't abolish prison furloughs for pickpockets after Willy Horton goes on a rampage.

    This is especially important in evaluating affirmative action. Race and gender preferences often look foolish at the extremes of the income distribution, but they provide significant, if less visible, benefits for many people near the middle. Affirmative action is often irrelevant to desperately poor ghetto residents who are not qualified for the opportunities preferences open up. Affirmative action also looks bad at the high end of the distribution, because it benefits many people who don't need the help.

    Arnold Barnett suggests (his Proposition 2) that affirmative action is a "public relations disaster." Nowhere is this more true than in the area of set-asides, preferential broadcast licenses and other subsidies to wealthy recipients. These visible and occasionally corrupt programs help to convince the public that affirmative action is more inefficient and arbitrary than it actually is.

    These programs are also costly. Making heroic assumptions from the recent White House review, I estimate that the federal government spent about $2 billion to assist women and minority entrepreneurs over the past year. Many of these costs are essentially invisible because they come in the form of slightly inflated prices in government contracts, or as slightly reduced prices for properties the government sells. The federal government spends almost as much on set-asides as it does for Head Start or WIC nutritional supplements for poor pregnant women and children.

    Some people believe that providing subsidies to prosperous entrepreneurs is a good strategy to alleviate problems of low wages and lack of upper mobility that afflict many minority groups. Maybe so, but this argument enjoys the empirical backing one associates with other forms of trickle-down economics.

    More generally, I am struck that many affirmative action "horror stories" involve a rather small group of relatively educated and affluent recipients. It seems that every time I see a newspaper story about affirmative action, it concerns admission to Harvard, partnerships at a New York law firm, or conflict between reporters at the Washington Post. Ironically, affirmative action is more fervently, and sometimes more foolishly pursued in these settings than it is within the society at large.

    We hear about such cases because they occur in the lives of journalists, professors and other elites that play a disproportionate role in shaping public opinion. Yet such cases are often misleading. The typical affirmative action recipient is a mail carrier or a cashier. Affirmative action on the assembly line at Southern New England Telephone makes less dramatic newspaper copy than whether Allen Bakke could go to medical school, but the ultimate consequences of such policies are probably more profound.

    A second problem with drawing inferences from extreme cases is that the best policy will still produce occasional disasters. Bureaucratic justice is dispensed by fallible, busy people who face significant uncertainty. Under any reasonable policy, there will be some welfare mother in Brooklyn who kills her child. Afterwards, there will always be some expose about how the authorities should have intervened.

    Right now, conservative foundations are combing the land in search of appealing, disadvantaged whites who were denied important opportunities because of affirmative action. Many resulting cases reflect the consequences of bureaucratic snafu or unusual circumstance rather than inherent problems of affirmative action itself.

    Cheryl Hopwood is the current poster child of the anti-affirmative action cause. She was denied admission to the University of Texas Law School even though she had worked 30 hours per week while attending school and caring for her disabled child. Virtually every admitted black or Latino student had lower grades and LSAT scores than she did.

    Regardless of what you think about affirmative action, this seems outrageous; so outrageous that there is more to the story. Although the story was widely reported in the national press, few pundits seem to have noticed that Hopwood submitted a very sloppy application. As the original court judgment notes, "Hopwood's application file contains no letters of recommendation. Additionally, her responses to the questions are brief and do not elaborate on her background and skill. She provided no personal statement with the application."

    Hopwood applied with outstanding grades, but these were from a junior college while most other applicants had attended four-year institutions. Many affirmative action "horror stories" are like this; not so horrible once the full facts are known.

    Ironically, gross favoritism and inefficiency associated with affirmative action were more common 20 years ago than they are today. In the early 1970s, universities and firms faced strong external pressure to improve racial diversity. The available pool of female and minority applicants was much weaker than it is today, while administrators were less savvy about the true costs and benefits associated with affirmative action.

    Proposition 3:
    (suggested by Richard Zeckhauser) If you care about both diversity and merit, there are strong arguments for quotas. That's one reason implicit quotas are so common.

    Even staunch defenders of affirmative action swear on a stack of Bibles that they oppose quotas. Few organizations are dumb enough to enact illegal explicit quotas. Yet in practice, many organizations establish policies that are almost the same thing. They do this because they want to achieve diversity, and at the least cost to their other organizational goals. Ironically, if judges and policymakers drive preferences underground by requiring ostensibly neutral criteria, they may encourage a less meritocratic outcome than we see today.

    This apparent paradoxical follows straight from the logic of optimization theory. Suppose that a state college cares about diversity; call this D to provide the appearance of mathematical rigor. The college also cares about strictly academic criteria such as SAT scores, denoted by s. Suppose further that the college's past decisions can be represented as some well-behaved objective function U(D, s).

    Free from legal and political constraint, the optimal policy is to divide the applicant pool into separate race and gender groups. Within each group, one then ranks the candidates based upon meritocratic criteria, rejecting every applicant who falls below some group-specific threshold. The actual thresholds are set based upon the relative quality of the different applicant pools, and the school's self-perceived tradeoff between diversity and other goals. Notice that given any level of ethnic and gender diversity, this algorithm yields the most academically qualified admitted group.

    What happens if some judge tells the university that it can't do this anymore? Maybe the university will genuinely ignore race and gender. This strikes me as unlikely. In selective institutions, such a policy would dramatically reduce minority representation, even if such factors as prior economic disadvantage were taken into account. It is not clear that most Americans, even affirmative action opponents, could really accept such an outcome.

    A more likely scenario is that colleges would tinker with the official criteria to give greater weight to those factors in which minorities do relatively well. Some of these changes would provide welcome inducement for universities to adopt broader and more socially accepted criteria than race or gender alone. Yet one should expect unexpected consequences, since the essential purpose would be to find a legitimate approximation for the same race and gender patterns admitted today.

    Most notably, such changes would apply less meritocratic criteria within every racial and ethnic group, including the group of white males who were not the presumed subjects of affirmative action in the first place. Schools would also be tempted to employ more subjective or poorly measured criteria such as "leadership skills" that would provide cover to take race into account.

    Proposition 4:
    We don't know the economic impact of affirmative action. There is no evidence that affirmative action creates serious adverse incentives in school or in the workplace.

    We don't know much about the real impact of affirmative action on the U.S. economy or on the specific individuals affected by such policies. The recent White House report, for example, is remarkably devoid of serious empirical analysis.

    I believe that there is a boring reason and an interesting reason for our ignorance about these visible and controversial programs. The boring reason is captured in the phrase, "Don't ask, don't tell."

    The more interesting explanation is that we know surprisingly little about the core processes that affirmative action seeks to influence: hiring, firing and promotion, why jobs are organized as they are. The internal labor markets of large organizations falls in the cracks between labor economics and operations research. Traditional tools of operations research are poorly suited to understanding equilibria and incentives. Economists have trouble unpacking how problems of lumpiness and group production create the need for large organizations in the first place.

    Our ignorance is most obvious in the debate over how affirmative action alters incentives to work hard or to acquire useful skills. An especially damaging criticism is that affirmative action dulls incentives for excellence because beneficiaries are effectively evaluated by less demanding standards.

    This argument does not follow from its factual premise. An interesting analytic feature of large organizations is that they reward people based upon relative, not absolute, performance. Candidates for promotion at AT&T are rewarded based upon how they stack up against other candidates. Relative compensation is especially common in settings where absolute productivity is difficult to measure, but where one can rank the performance of workers doing comparable jobs.

    Thus, the young lawyer works 70-hour weeks in the hope of making partner. Why doesn't she work 60 hours or 80 hours? Economic theory says that she works until her marginal disutility of effort balances the expected gains associated with her eventual promotion. A similar process operates in the allocation of prizes such as admission into elite colleges. The marginal expected value of studying for the SAT is pretty low if I am a high school star, or if I am unlikely to get into college even if I improve my score.

    Relative compensation runs into problems within a heterogeneous workforce. Our lawyer won't work hard if her current efforts will not appreciably improve the likelihood that she will be promoted. If she is a star, she will advance even if she slacks off a bit. If she is unskilled, if her boss doesn't like her, she may give up. In both cases, pure "meritocracy" is inefficient because the marginal expected payoff to the individual for her effort is too low.

    In the presence of such heterogeneity, firms want to sort workers so that each person competes against others of similar competitive standing. If that is impossible, the optimal policy is to impose a "handicap" whereby workers who start at a disadvantage get an added bonus if they perform well.
    I think this is the most compelling commonsense argument for affirmative action. If women or minorities face real competitive disadvantages -- due to poor prior training, draining family obligations or explicit discrimination -- they may not work hard out of a fairly rational calculation that their effort will go unrewarded. Most poor black and Latino teenagers know perfectly well that their academic skills lag behind their middle-class peers. Whatever might be wrong with affirmative action, such policies send a positive message to these young men and women: If you don't give up, you have a shot to get a decent job, to get into a selective college.

    This brings me to the concluding note of my essay. Conservatives note, correctly, that affirmative action brings real costs when less qualified workers are sometimes placed into important jobs. It is all too easy to find glaring examples of inefficiency, unfairness and hypocrisy in prevailing programs. Sometimes these costs are too large, and so programs must be reformed.

    It is, however, important to remember that the costs of affirmative action pale in comparison to the most glaring inefficiency in the U.S. economy: the incalculable waste of human potential among disadvantaged segments of our society. It seems to me that affirmative action has drifted from its most important goal: to provide opportunities for poor and working-class people. Affirmative action in hiring or college admission that advances this goal should be openly defended. Race and gender-conscious programs for relatively advantaged people should be eliminated. The resulting policies would remain imperfect. I suspect, at least I hope, they would command a stronger public mandate than they do today.

    Harold Pollack is a Robert Wood Johnson Health Policy Scholar at Yale University.

    E-mail to the Editorial Department of OR/MS Today:

    OR/MS Today copyright 1997, 1998 by the Institute for Operations Research and the Management Sciences. All rights reserved.

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

    Web Site Copyright 1997, 1998 by Lionheart Publishing, Inc. All rights reserved.
    Web Design by Premier Web Designs, e-mail