The Merits of Merit
Judith Lichtenberg and David Luban
How are people to be chosen for jobs and educational opportunities? Opponents of affirmative action draw a sharp contrast between two possible approaches: on the one side, employment and admissions procedures based on merit; on the other, affirmative action programs that dilute the merit principle. Either we can select the best candidate regardless of race or gender, or else we can allow group membership to influence our decision.
The ideal of "careers open to talents" dates from Napoleon's time, when it formed a centerpiece of the liberal assault on nepotism and hereditary privilege. To nineteenth-century reformers, the opposite of merit-based selection was not affirmative action for previously excluded groups; it was the awarding of posts on the basis of pedigree and connections. In one sense, today's merit principle is the reverse of the historical merit principle. During the age of aristocracy, the merit principle was a kind of affirmative action for previously excluded groups -- the talented sons of middle-class families -- whereas contemporary affirmative action is most often defended as a way for women and minorities to overcome the lack of pedigree or connections.
In another sense, however, the principle remains the same: then and now, it insists that what matters is what you know, not whom you know or what group you belong to. California Governor Pete Wilson has argued that it is a bedrock principle of American society that "individuals should be rewarded on the basis of merit." If Wilson is right, affirmative action programs are un-American and unjust, and the merit principle must reign supreme.
Critics of affirmative action like Wilson believe they know what merit is and how to find it. The defense of affirmative action, on the other side, is often understood to mean that "there is no such thing" as merit, or that merit is "socially constructed." Some critics of merit do make such claims. In so doing they tend to alienate those of us who hold the commonsense beliefs that some people really are more talented at various endeavors than others, that ability is not a purely subjective thing, and that neither talent nor differences in talent are entirely the product of culture or politics.
But these commonsense beliefs don't prove that opponents of affirmative action are right. Ability is a complex and problematic notion, and the merit principle, which identifies ability with entitlement, is doubly complex. Because it figures so prominently both in defenses of and attacks on affirmative action, merit demands sustained and careful analysis. In what follows we pose three sorts of challenges to the meritocratic ideal:
1. Detecting merit. Even if we assume that there is such a thing as merit, and that we agree about what it is, the methods and procedures employed to detect merit may be flawed or biased, even when they function properly.
2. Defining merit. Disagreement exists not only about what procedures best uncover merit, but also about what constitutes merit for a given job or for a coveted place in the freshman class. In general, we argue, meritocrats are wedded to an overly simple and overly narrow view of what constitutes merit.
3. Establishing merit's proper sphere of influence. We defend two claims. First, small differences in merit often get reflected in large differences in reward, violating a principle of proportionality that ought to be part of a reasonable merit principle. Second, we argue that merit, however defined, should not be the only factor employed in allocating jobs or admissions slots.
Detecting Merit
Assume for the moment that jobs and admissions slots ought to be distributed purely on the basis of merit. Assume also that agreement exists on the criteria of merit for these jobs and places. The criteria of merit might be simple or complex; of course they will vary from job to job or from field to field. Leaving aside these details, let's call whatever it is we believe constitutes merit "M." M, we'll suppose, gets measured on the M-scale. And let's call that which merit is merit for -- the particular job or admissions slot -- a "post." The merit principle says that posts rightfully go to the highest-M applicants. The question facing us, then, is how best to locate M.
The critique of meritocracy often begins with the observation that -- leaving affirmative action programs aside -- we don't in fact live in one. Coveted posts are often won not by those possessing the most M, but by those whose good connections have brought them to the attention of those with posts to distribute. A high school graduate whose mother works at the utility company hears about a job and has an in. The alumnus of Ivy College learns through friends or the alumni association of a recent graduate looking for a job and is pleased to strengthen the collegiate connection.
Often these processes are innocent, in the sense that it's natural, and even in many ways praiseworthy, to let someone you like know of a good job, and to help him get it if you can. Sometimes they are not. Whether innocent or not, we may think them largely inevitable; social institutions couldn't operate smoothly without personal contacts greasing the wheels. But connections and old-boy networks surely violate the principle of meritocracy -- that was the original point of the career open to talents -- and affirmative action programs can serve to counteract their power. In so doing, affirmative action can sometimes enhance the relationship between merit and posts, not diminish it, as critics imply.
But the role of connections in the acquisition of posts goes deeper than these remarks suggest. To see why, consider this example. The top-ranked law school in the country has no class rankings and grades its students pass-fail. Yet these students have no difficulty getting great jobs at top law firms. Why?
There's no mystery from the law firms' point of view. They know that the law school accepts only 5 percent of its applicants. The firms are willing to take the law school's word for it that these students are loaded with M. The students' applications were read by a team of talented law professors; for a firm to conduct a similar evaluation would cost tens of thousands of dollars. So the firms are willing to free ride on the law school's prior search.
We might describe the method used by the law firms in identifying attractive associates as a rational search procedure. It's not the search procedure that provides the most detailed and accurate information about the M of all possible candidates. But it provides very good information about M at low cost, and that makes it rational.
The important point is this: a rational search procedure will regularly select some candidates with less M than others it passes over. For the procedure will give only perfunctory glances at candidates from lesser law schools, some of whom have more M than those from top-ranked schools. That doesn't mean the procedure is irrational: from the firm's point of view, the slight gain in accuracy from a more inclusive search isn't worth the added cost.
Still, we are justified in rejecting or modifying certain entirely rational procedures if we find that they have unacceptable social costs. If minority candidates from disadvantaged backgrounds are likely to have lesser credentials that do not always reflect lesser M, that is a reason to adopt affirmative action programs to counteract this flaw in rational search procedures.
The instruments employed to detect M may be deficient in other ways as well. The controversy surrounding standardized tests has become familiar. Some critics argue that a strong cultural bias against women and minority groups infects tests like the SAT; others think this view is nothing more than PC-think. Perhaps more significant than possible bias in the tests themselves is that higher-income students, among whom blacks and Hispanics are underrepresented, attend high schools that self-consciously teach to the SAT, and they further boost their performance with expensive test-preparation courses that raise scores while presumably leaving M untouched.
Such flaws in procedures designed to reveal M need not reflect ill will or personal prejudice towards women or minorities. Indeed, they show that the controversial term "institutional racism" -- invoked to support the view that discrimination does not always disappear when personal bias does -- can have a well-defined meaning.
On the other hand, some obstacles to evaluating a person's M objectively can be located in people's (not necessarily conscious) attitudes, rather than in procedures and institutions. For example, surveys of male managers, business students, and college professors show that, in Deborah Rhode's words, "identical resum‚s are rated significantly lower if the applicant is a woman rather than a man." Other studies indicate that "both male and female subjects have given lower ratings to the same artwork or scholarly articles when the artist or author is thought to be a woman."
Even if there is such a thing as M, and even if we agree about what it is, finding out who has it isn't always easy.
Defining Merit
As we have noted, the most familiar objections to the merit principle challenge the very idea of merit, arguing that there is no such thing, or that it is "socially constructed." (Rightly or wrongly, these two assertions are often taken to be equivalent.) We shall defend a more modest view. Modest or not, however, it is a view that critics of affirmative action implicitly deny.
First, M is not a single property; for practically any post, there are several ingredients that contribute to doing well in it. Second, there can be legitimate disagreement both about what the appropriate ingredients in M are and about their relative importance. It follows that comparing the M of two candidates for a post is always a matter of weighing the relative significance of different qualifications; the comparison sometimes involves disputes about whether particular qualifications are relevant at all.
Take the example of an academic job in a large state university. Such a position involves both teaching and research responsibilities; thus evidence of merit in each of these areas is relevant. There is, of course, disagreement about the relative weight each should bear, with many people arguing that teaching ability should count for more, and research for less, than it usually does in appointment decisions. But beyond this fundamental division are other points of dispute: what counts as good research (quality, quantity, subject area, degree of specialized interest or relevance); the variety of settings in which a person might or might not be a good teacher (small, medium, and large classes; introductory or advanced students; graduate students and undergraduates); collegiality, administrative ability, and general reliability. To think it is generally clear who has the most M, and that the question is only whether to hire that person rather than someone else who satisfies other criteria (such as "diversity") quite distinct from merit, is to be confused by a very narrow view of merit. (Our own experience suggests that many academics are confused in this way, not only favoring a highly restricted view of merit but insisting that it alone captures M objectively. These are the people who talk about how "smart" candidates are and how nothing else really matters and who act as if they can peer into candidates' brains to see them percolating.)
An important conclusion to be drawn from this example is that merit is always a functionalnotion. Many people hold a picture of merit that is wholly individual and personal: the person with the most merit is the smartest, or the fastest, or possesses some other quality that can be defined without reference to the role these qualities play. But this picture is false. You can't decide what merit is for a university teaching position, for example, without knowing what the purposes to be served by such a position are. And that in turn requires an account of the goals and roles of a university and the services it provides. Needless to say, such questions are controversial.
Once we look at merit in this way, however, it becomes clear that some considerations opponents of affirmative action contrast with merit may be a legitimate part of it. For example, being a member of a certain group should count as a qualification or "plus factor" for a given post when members can serve as role models. The role-model argument is especially relevant to educational institutions, whose mission may include shaping students' values and aspirations. It isn't simply that a black teacher, say, can provide a motivating example to black children. That idea was rejected by Justice Powell, in Wygant v. Jackson Board of Education, on the grounds that it would provide a rationale for segregation. Rather, the argument is that a black teacher plays an exemplary role not only for black children but for all children, as well as for parents and the larger community. (Likewise for the male elementary school teacher.) Moreover, contrary to the view of some critics, this argument does not rest on stereotyped judgments about the beliefs or attitudes of group members. The example of a female professor affirms possibilities of achievement and recognition, whether or not she holds what are thought to be the right political views.
There is a further argument for sometimes counting group membership as an element of merit. Affirmative action opponents frequently argue that the market will eventually eliminate discrimination. A business that discriminates against women and minorities will be at a competitive disadvantage, it is said, because it will lose capable employees to firms that don't discriminate. Over time, businesses that wish to remain competitive will be compelled to cease discriminating.
We think this argument is naive. But even if market pressures were sufficient to eliminate discrimination, the argument leads to conclusions that would dismay opponents of affirmative action. Common sense suggests that very few women want to be the only female employee in a business, just as very few minorities want to be the only minority employee. Even someone who doesn't mind anomaly and isolation will justifiably suspect that an employer with very few women or minorities in responsible positions has a glass ceiling. Having minority or female employees in responsible positions thus gives firms a competitive edge in the recruitment and retention of women and minorities. But that implies that being a woman or minority is, in and of itself, a valuable asset a component of merit for those positions.
All of which is to say that the distinction between merit and group membership, or talent and connections, is far more dubious than proponents of the merit principle suppose.
Merit, Proportionality, and Reward
Even if we agree about what merit is, and even if we agree about how to find it, merit alone is an inadequate distributive principle, for two reasons. In this section we discuss the first, in the next the second.
Consider that associated with various posts are a range of attractions, including salary, pleasant working conditions, security, fringe benefits, enhanced future opportunities, and social status. Let's bundle all these together and call them rewards.
Every post requires a certain amount of M to do the job successfully; let's rank posts on the P-scale according to how much M they require. Then one version of the merit principle says that the higher you are on the M-scale, the higher you deserve to be on the P-scale. Let's assume, further, that rewards can also be ranked, on the R-scale: higher Rs mean higher salary, status, and the like. Now it certainly isn't true that posts ranking higher on the P-scale necessarily rank higher on the R-scale as well. Classical philology, which requires a working knowledge of at least five languages, is a lot harder than legal scholarship -- it's higher on the P-scale -- but law teachers do much better than philologists on the R-scale. Indeed, the starting salary for beginning law teachers approximately equals the maximum salary for a full professor of classics.
It probably is true, however, that within professions, and to some degree across professions, there is some rough correlation between the P-scale and the R-scale. The extent of this correlation will depend, in part, on how widely and unevenly dispersed the range of rewards is in a given economy. The U.S., for example, has a more polarized distribution of incomes and wealth than other industrialized countries; the rich are richer and the poor are poorer. (And according to economists Robert Frank and Philip Cook, rewards are becoming increasingly dispersed in our "winner-take-all" society.) Ben & Jerry's used to insist that its highest-paid executive would never make more than seven times what its lowest-paid worker makes. In large firms, however, the CEO often makes 300 times what the lowest-paid worker makes.
On one construal, the merit principle says that people who rank higher on the M-scale should rank higher on the P-scale. This reading of the merit principle best corresponds with the notion of careers open to talents. Should people higher on the M-scale also rank higher on the R-scale? Perhaps. But suppose one person has slightly more M than another. By the merit principle, she should end up with a job slightly higher on the P-scale. The best anthropology graduate student should get the job in the first-tier department, and the less capable should go to lower-tier departments. In a winner-take-all economy, however, the difference in rewards between two people differing slightly in M can be unjustifiably great. The slightly more talented student lands the increasingly scarce tenure-track job, where, with a moderate teaching load, she can do her research and get tenure; the only slightly less talented student moves from one temporary job to another and never has the time or security to establish her career.
The process by which small differences in merit lead to large differences in reward can be relatively simple, as in the case just described, or it can be more complex and iterated. To see the latter, recall our notion of a rational search procedure: a procedure used by an employer or a school that provides good information about candidates cheaply by attending to easy-to-detect surrogates for quality. On this model it's rational, for example, for a professional school to rate applicants from first-tier colleges higher than those from lesser schools, even though some applicants from the latter may be better than some from the former.
Unfortunately, the use of such procedures tends over time to magnify the discrepancies between merit and success. Consider a high school senior (let's call him Gold) who barely squeaks into an Ivy League college. Perhaps he is a shade more talented than his closest competitor (let's call him Bronze); or maybe his father is an alumnus. Whatever the reason, on graduating he has a much better chance of being admitted to a top-flight law school. And this credential, in turn, bolsters his chances of obtaining a prestigious internship, and then a federal clerkship, and ultimately a successful legal career. Meanwhile, the less fortunate Bronze finds himself at a disadvantage at each of these thresholds. We see in this tale the cumulative effect of one rational search procedure after another, which amplifies a minute difference in ability into enormous differences in posts and rewards. A reward at each stage becomes a credential for the next.
It is of course possible that the difference in M between Gold and Bronze actually increases over time. Gold may have benefited from a superior education, better fellow students and teachers, and a more intense intellectual environment, so that even if Gold and Bronze began college with virtually indistinguishable M, they are no longer so similar. In that case, however, their story speaks to the enormous advantages conferred by prestigious and high-powered environments that are capable of producing M; and it speaks against the idea that M is some pristine quality immune to alteration. Given the opportunity to inhabit such environments, the less advantaged can also enhance their M, and therefore can deserve their excellent posts even on the most exacting use of the merit principle.
Those who care about merit ought to care about proportionality between merit and reward. Indeed, the most plausible interpretation of the merit principle says that a person should be rewarded in accordance with or in proportion to her merit. But in many contexts in our society, small differences in merit translate into large differences in reward. This is merit run amok, and ought to disturb anyone who genuinely cares about the merit principle.
Putting Merit in Its Place
Disproportionate rewards provide one reason to think that merit has been given too large a role in conferring benefits. The other is that merit isn't the only value or principle that plays or ought to play a part in the distribution of posts.
Plays or ought to play: the phrase may catch us up. Which is it? The defender of meritocracy seems to argue that merit is the only principle that ought to play a part, even though we know that others do in fact. Yet part of the argument for the view that merit ought not to be the only factor in such decisions is that it never has been and that no one really believes that it should be. To believe that merit alone should rule is to say that we should abolish seniority and veterans' preference in employment decisions; it is to say that we should disregard geographic diversity and preference for the children of alumni in college admissions.
Now it may well be that some of these policies ought to be abandoned. Legacy preference -- the policy of elite colleges giving preference in admissions to the children of alumni -- is a form of affirmative action that favors the already favored; it must serve as an embarrassment to meritocrats, who suggest that but for affirmative action the merit principle would reign, supreme and alone. But few people would deny that some weight should be given in employment or admissions decisions to seniority, veteran status, or geographic diversity. In accepting these practices, we implicitly admit that other values count besides merit.
Almost no one, then, really believes that -- to put it as Barbara Bergmann has --only the merit principle has value and that it should never be traded off for any other consideration. The question is rather which other values or principles count, and how much. If geographic diversity is a legitimate value in college admissions, why should racial or ethnic diversity be less legitimate? Given our history, we may well think that accepting more black students will better help a college create a cosmopolitan campus environment than accepting more students from North Dakota. Merit (properly detected and properly defined) ought to count a great deal, but a pure meritocracy is a society that is difficult to imagine, and one that few of us would care to inhabit.
Sources: Barbara Bergmann, In Defense of Affirmative Action (Basic Books, 1996); Robert Frank and Philip Cook, The Winner-Take-All Society (Free Press, 1995); Susan Sturm and Lani Guinier, "The Future of Affirmative Action: Reclaiming the Innovative Ideal," California Law Review, vol. 84 (1996) (for sources on the debates about standardized tests); Deborah Rhode, Justice and Gender (Harvard University Press, 1989) (for sources on studies of unconscious bias).