|Discrimination in the Post-Civil Rights Era: Beyond Market Interactions|
Journal of Economic Perspectives, Spring 1998
Comment of Glenn C. Loury
The literature reviewed in the papers in this symposium suggests that discrimination against women and blacks still exists. In credit, housing, and labor markets there are measurable differences in the terms at which similarly situated men and women, and comparably placed persons of different racial groups, transact. We cannot confidently assume that markets free from monopoly conditions will automatically eliminate such discrimination. Each of the authors, in their respective ways, makes this point, and it is a point worth making. But, this is not really a very surprising observation, nor does it present an insuperable challenge to economic theory.
From my perspective, there is something a bit odd about the way the research problem is defined in this literature. The fact that price discrimination based on gender or racial identity offends the arbitrage principle, but is readily observed in society, is taken to constitute an anomaly that critics of the neoclassical orthodoxy in economics seize upon, and that defenders of this orthodoxy need to explain away. Thus, the inability of Beckers original theory of market discrimination (Becker, 1957) to account for the persistence of wage differences between blacks and whites, or men and women, of comparable skills is said to be a problem for economic science. At the same time, evidence of declining productivity-controlled wage difference between the races or sexes is supposed to afford a vindication.
It seems to me that this way of thinking about the problem is too narrow. With respect to wages, for example, the focus is on the demand side of the labor marketon the employers so-called "taste for discrimination." The underlying normative idea is that discrimination is bad, a legitimate object of regulatory intervention, and a significant contributor to the scourge of race and sex inequality in society. Implicit here is the notion that if inequality were due to supply side differencesin the skills presented to employers by blacks and whites, for examplethen the resulting disparity would not raise the same moral issues, or give a comparable warrant for market intervention. With respect to housing markets, the view is that residential segregation induced by the discriminatory behavior of realtors is a more severe problem than the segregation that comes about because of the freely chosen decisions of fairly treated market participants.
Focusing attention on the issue of racial inequality, I want to use this comment to propose an alternative approachor, at least, to suggest a shift of emphasis. It has become very easy to produce theoretical explanations of price discrimination against blacks in credit, labor, and consumer goods markets, using the tools of modern economics. If anything, there are more such accounts than available data can distinguish among. It is not so much the intellectual resources of economics, but rather the political and moral resources of society, that are called into question by the phenomenon of persistent racial inequality. Moreover, while discrimination against blacks still exists, it is neither as pervasive, nor as significant an explanation for racial inequality in the United States, as was the case in decades past. The skills gap between black and white workers is substantial, may be growing, and is certainly a key factor in accounting for racial inequality in the labor market (ONeill, 1990; Neal and Johnson, 1996; Thernstrom and Thernstrom, 1997). There is evidence that this skills gap is linked to social and cultural influencesgeographic segregation, deleterious social norms and peer influences, poor educational qualitythat have a racial dimension (Cutler and Glaeser, 1997; Akerlof, 1997). In this view, racial inequality is substantially a supply side problem. A focus on price discrimination does not do it justice.
Some scholars (e.g., Darity and Williams, 1985) have been justifiably concerned that an approach to the problem of racial inequality that focuses less on employer discrimination and more on skills differences could foster dangerous stereotypes, and undermine normative arguments for policies that narrow the racial wage gap. In the decade after the enactment of federal anti-discrimination laws, the researchers who began to find evidence of a decline in labor market discrimination (Freeman, 1973; Wilson, 1978) were criticized for giving aid and comfort to political conservatives. Unfortunately, this reaction accepts the implicit normative assumption that racial inequality based on skill disparities is not as important a moral problem, warranting as vigorous a corrective intervention, as inequality based on price discrimination in the labor market. But, as an ethical matter, that assumption is not compelling, and it should be challenged.
The conventional wisdom regarding equal opportunity policy is that the elimination of racial discrimination will result, eventually, in a significant reduction of racial economic inequality. This view derives from an analysis of labor markets in which the core unit of analysis is the individual, not the social group. The market is seen as a place where agents act more or less independently, each seeking to make the best of the opportunities at hand. This way of thinking has been very fruitful for economics, but it cannot adequately capture the ways that racial inequality persists over time. In actuality, individuals are embedded in complex networks of affiliationsthey are members of nuclear and extended families; they belong to religious and linguistic groupings; they have ethnic and racial identities; they are attached to particular localities. The isolated individuals posited in economic theory are really socially situated. Being situated in this way matters for the problem at hand because ones location within the network of social affiliations substantially affects ones access to resources that are essential for personal development. Opportunity travels along the synapses of these social networks.
Thus, a newborn is severely handicapped if its parents are relatively uninterested in (or incapable of) fostering the youngsters intellectual development in the first years of life. A adolescent with prodigious talent, but whose social peer group disdains those activities that must be undertaken for that talent to flourish, is at some risk of never achieving his or her full potential. An unemployed person without friends or relatives already at work in a certain industry may never hear about the job opportunities available there. In every case, the individuals inherited social situation plays a major role in determining his or her ultimate economic success.
In earlier work (Loury, 1977, 1981, 1987) I have suggested an extension of human capital theory designed to provide a richer context within which to analyze group inequality. This theory builds upon the empirical observations of students of social mobility concerning the importance of family and community background in determining individual achievement. In this theory, ones investment in productive skills depends on ones position in the social structure, due to imperfect capital markets for educational loans, social externalities mediated by residential location and peer associations, and psychological processes that shape a persons outlook on life. As a result, familial and communal resourcesthat is, social and cultural capitalexplicitly influence a persons acquisition of human capital. Some important part of racial inequality, in this view, is seen to arise from the way that social segregation along racial lines makes an individuals opportunities to acquire skills depend on previous and contemporaneous skill attainments by others in the same social group.
There is fairly strong support in the literature for this view, as it applies to the lagging economic position of blacks. Akerlof (1997) provides a theoretical argument, supported by a wealth of evidence from sociology and anthropology, for the notion that concerns for status and conformity are primary determinants of individuals educational attainment, childbearing, and law breaking behavior. Anderson (1990) provides an ethnographic account of life in inner city Philadelphia in which peer influences are seen to significantly constrain the acquisition of skills by adolescents in those neighborhoods. Waldinger (1996), in a study of immigrant labor in New York City, concludes that poor blacks suffer less from the racism of employers than from the fact that they do not have access to the ethnic networks through which workers are recruited for jobs in construction and service industries. Glaeser and Cutler (1997), comparing U.S. cities with varying degrees of racial population concentration, find blacks to be significantly disadvantaged by residential segregation. They estimate that a 13% reduction in segregation would eliminate about one-third of the black-white gap in schooling, employment, earnings, and unwed pregnancy rates. Mills and Lubuele (1997) argue that a central problem for students of urban poverty is to explain why "low income black residents actually or potentially eligible for jobs that have moved to suburbs (have) not followed such jobs to the suburbs."
All of this suggests the inadequacy of the discrimination/anti-discrimination framework when discussing racial inequality. Conventional price discrimination against blacks in labor markets is not the primary source of race disparities, and available methods for fighting such discrimination have little power to reduce the economic gap between the races. In any event, for over a decade now the bulk of the federal anti-discrimination apparatus has been occupied with the complaints of non-black plaintiffspersons bringing claims under age, disability, and sex discrimination statutes (Smith, 1993). It has been shown that in one important areatransactions between business concerns penalties for practicing racial favoritism are negligible, because the very illegality of discrimination has yet to be clearly established (Suggs, 1991). Moreover, given the informational asymmetry between employers and enforcement agents, there are theoretical limits to how aggressive anti-discrimination policy can be before one encounters significant efficiency costs (Coate and Loury, 1993 (a)). Unlike Epstein (1992), I do not think we should repeal the Civil Rights Acts or dismantle the Equal Employment Opportunity Commission. But, I do think that, if the concern is economic inequality between the races, then looking mainly through the lens of labor market discrimination is unlikely to shed much light on the problem.
There is another, more fundamental, reason to broaden the discussion of racial inequality beyond the context of market discrimination. In cities across the country, and in rural areas of the Old South, the situation of the black underclass and, increasingly, of the black lower-working classes, is bad and getting worse. This is certainly a race-related problem. The plight of the underclass is not rightly seen as another (albeit severe) instance of economic inequality, American style. But, conventional market discrimination is only one, small, part of it. These black ghetto dwellers are a people apart, susceptible to stereotyping, ridiculed for their cultural styles, isolated socially, experiencing an internalized sense of helplessness and despair, with limited access to communal networks of mutual assistance (Anderson, 19990; Wilson, 1996). Their purported criminality, sexual profligacy, and intellectual inadequacy are the frequent objects of public derision. In a word, they suffer a stigmatized, pariah status (Goffman, 1963). It should not require enormous powers of perception to see how this degradation relates to the history of black-white race relations in this country.
Here is where the implicit normative model that accompanies the emphasis on market discrimination is most seriously flawed. The civil rights strugglethat succeeded brilliantly to win for blacks the right to be free of discriminationfailed utterly to secure a national commitment toward eradicating the effects of historical discrimination. When those effects manifest themselves among poor blacks in patterns of behavior bearing on the acquisition of skills that are self-limiting, the tendency of many who think only in terms of market discrimination is to argue that society is not at fault. There is a grain of truth in the conservatives insistence that, while overt racism was implicated in the past, it is behavioral differences that lie at the root of racial inequality in contemporary America (Thernstrom and Thernstrom, 1997). But the deeper truth is that, for quite some time now, the communal experience of the African slaves and their descendants has been shaped by political, social, and economic institutions that, by any measure, must be seen as oppressive. When we look at "underclass culture" in the American cities of today we are seeing a product of that oppressive history. In the face of the despair, violence, and self-destructive behavior of these people, it is morally obtuse and scientifically naïve to argue, as some conservatives now do, that if they would just get their acts together then we would not have such a horrific problem in our cities. Yet, for the same reason, it is a mistake to argue against racial inequality, as some liberals do, based mainly on the supposed severity of ongoing market discrimination.
There are significant market failures, having nothing to do with price discrimination per se, that play a powerful role in perpetuating racial inequality. Consider the problem of residential segregation (Massey and Denton, 1993). Compelling theoretical arguments (Schelling, 1978: Chp. 4), and recent computer simulations (Wayner, 1998), suffice to show that, absent invidious behavior by market participants, a mild desire for people to live near members of their own race can lead to a strikingly severe degree of segregation in the aggregate. Adding class concerns to these models only strengthens their predictions of geographic clustering. Moreover, residential location is not the only venue in which segregation occurs. Linguists studying speech patterns in urban centers have uncovered quite strong evidence of race and class separation (Labov, 1982). There is reason to suspect that race differences in communicative styles could play an important role in accounting for the adverse labor market outcomes of low income blacks (Lang, 1986; Cornell and Welch, 1996; Charles, 1997; Wilson, 1996).
Even though residential and social segregation based on race may be a natural outcome, in some sense, it could still have morally disturbing consequences. And, even if those consequences manifest themselves mainly on the supply side of the labor market, a strong case could still be made for doing something about them. That case need not be based solely on equity grounds. Indeed, once social and psychological externalities influencing the preferences and the skills of market participants are admitted into the analysis, the conventional results in welfare economics concerning the efficiency of market outcomes are no longer generally valid. Consider the following "stories" illustrating how such social and psychological externalities might operate:
(1) Dissonance associated with holding values at some distance from ones peers. The individual experiences a loss of utility from such "deviance," which is gradually ameliorated by adjustment of behavior toward greater conformity with the "norm." This "norm" is inferred from the apparent preferences of those with whom the individual has frequent contact. This being the case for all individuals in the model, the "norm" is itself the resultant of these individual choices. Moreover, when social structure (i.e. the pattern of individuals associations) is such that racial group members are more likely to associate within as opposed to between groups, it is to be expected that distinct "norms" can prevail simultaneously among the distinct groups. Generalizations about differences between groups in attitudes toward work, family life, criminal participation, and the like may thus be empirically correct, but morally irrelevant. Moreover, the existence of multiple self-sustaining norms raises the possibility of Pareto-improvements in welfare associated with norm-shifting interventions. (Akerlof, 1997; Sunstein, 1996)
(2) Causal attribution based upon socially influenced cognition. Individuals infer from the random results of sequential encounters with the market the extent to which "effort," as opposed to "ability," or "luck" accounts for market rewards. Each individual adjusts effort in line with current perceptions of the return to it, balanced against costs. Neither "ability" nor "luck" are observable. However, market outcomes for oneself and others to whom one is socially connected can be observed, providing information about the distributions "ability" and "luck" for group members. In this way, the degree to which an individual believes that bad personal outcomes can be corrected with greater personal "effort" can vary positively with the aggregate performance of other members of his group. Self-fulfilling pessimism about the returns to effort for certain activities (e.g., academic pursuits (Steele, 1992)) are possible in a model like this.
I am not suggesting that "stories" of this kind are the explanations of racial inequality. But, I hold that they are not implausible accounts of how social segregation might support behavior patterns that lead to a skills gap between racial groups. Furthermore, these stories and others similar "tales" that could be told suggest that policies directed at reducing the skills gap might be just as morally required, and even more efficacious, than policies directed against such market discrimination as may still exist.
There is one further general point that I would like to make about models of economic discrimination. It is useful to distinguish between the taste-based theories of discrimination that originate with Becker (1957), and information-based theories of statistical discrimination derived from the work of Arrow (1972), Phelps (1972), and Aigner and Cain (1977). My own view, elaborated at greater length in Coate and Loury (1993), is that information-based models afford a more realistic account of the market discrimination encountered by blacks today. Race is an easily observable trait, difficult to alter or conceal, that, as an empirical matter, is correlated with some hard-to-observe traits about which employers, lenders, police officers and others are concerned. Wilson (1996) presents evidence based on employer interviews to show that both black and white employers are reluctant to hire black, urban young males who exhibit lower class behavioral styles. Applebaum (1996) discusses the widespread use by police officers of racial identity as information guiding their law enforcement decisions. Indeed, the dramatic disparity between the races in the rates of arrest and incarceration for criminal offenses (Tonry, 1995) must be taken into account when discussing racial differences in the labor market experiences of males (though the direction of causality is not obvious).
In no way do I mean to suggest that statistical discrimination should be tolerated because it is so widespread. To the contrary, such discrimination can be quite damaging to both the efficiency and the equity of market allocations. This is due to the very real possibility that the statistical generalizations lying at the heart of such discrimination can be self-fulfilling prophecies (Lundberg and Startz, 1983; Coate and Loury, 1993). For this reason, I would support the view that statistical racial discrimination, even when based on empirically valid beliefs, should be treated with legal suspicion. However, the enforcement problem in this area appears to be formidable; so, a fair amount of discrimination based on racial stereotypes is taking place. It is not difficult to give straightforward economic accounts of how this process might work.
An Adverse Selection Argument: Suppose only a very few taxi drivers will pick up young black men after a certain hour. Given that behavior by taxi drivers as a class, what "types" of young black men will be hailing taxis? If the set of those seeking taxis when the odds are low that a given taxi will stop for them has an especially large fraction of potential robbers in it, the drivers might rationaly be reluctant to stop. If, on the other hand, most drivers willingly picked-up young black men, then this behavior on their part might induce a less threatening set of black males to rely on taxi transportation after dark, confirming the rationalilty of the drivers more tolerant behavior.
A Moral Hazard Argument: Suppose employers fire workers who make a certain number of mistakes on the job, since they infer from this that the worker has not tried hard enough to learn the job. Suppose workers can affect the probability of making a mistake by exerting high rather than low effort during the training period, but employers do not observe this level of effort. If employers a priori beliefs are that blacks are more likely to be low effort trainees than are whites, then they will set a lower threshold for blacks on the number of mistakes needed to trigger dismissal. (The optimal decision rule is to terminate when the posterior likelihood that an employee was a low-effort trainee rises above a fixed level, which happens with fewer mistakes for those believed a priori more likely to have exerted low effort.) But, knowing that they can be fired if even after exerting high effort they have the bad luck to make a few mistakes, more black employees may elect not to exert high effort in the first place, confirming the employers initial beliefs.
An Outside Option Argument: Let car dealers bargain with black and white buyers based on beliefs about the distributions of reservation prices in the two populations. If it is thought that black buyers have higher reservation prices than whites, then dealers will be tougher when bargaining with blacks than with whites. But, given that other dealers are harder bargainers with blacks, a black buyer anticipates less favorable alternative opportunities than a white buyer to the offer of a given dealer, and so, other things equal, may rationally agree to a higher price. (Whether or not this occurs in equilibrium of a bargaining-matching game depends on the details of the bargaining protocol.) This behavior confirms the dealers initial presumption that "color" predicts bargaining power.
These stories are not difficult to produce, and what is more, they have a certain ring of truth about them (at least, to my ear). Here is one final example: Suppose capital is provided by lenders to entrepreneurs who choose investment projects. The exact nature of the projects chosen is not fully observable by lenders. Specifically, projects take at most two periods to complete, and may require both start-up and interim finance. There area two types of projects: Type A projects have a higher net rate of return than type B, are likely to require interim finance, but the need for interim finance conveys no information about the ultimate project outcome. Type B projects have a lower likelihood of requiring interim finance, but the need for interim investment in these projects is a bad signal about the ultimate outcome. Lenders in the interim loan market cannot observe the type of project selected, but can observe the race of the prospective borrower. Entrepreneurs provide their own start-up financing, choose the project type, but must rely on interim lenders if additional financing is needed. Under these circumstances, it an be shown that, for appropriate parameter values, if entrepreneurs believe interim finance will not be forthcoming, they will purse type B projects, while if interim lenders think type B projects are being chosen, they will not provide interim loans. Alternatively, entrepreneurs confident of receiving interim financing will pursue A projects, and lenders who think type A projects are being chosen will provide interim financing. In this way, equilibria arise in which loan opportunities depend on an entrepreneurs' race, because lenders (correctly) associate race with the unobserved aspect of project choice.
The key to all of these examples is that the likelihood function used by Bayesian agents when engaging in racially based statistical discrimination is endogenous. The beliefs on which people base their differential reactions to blacks and whites are the result of some equilibrium-determined factora price, an outside opportunity, a participation decision, an effort choice. Yet, (the distribution of) that factor varies with race only because, in the equilibrium, it is believed by agents to be associated with race.
There is an especially disturbing variant of this process, which I might refer to as persistent prejudice. Consider two situations: (1) a homogeneous population, each member of which will, on a given occasion, engage in a criminal activity with probability 1/10; and, (2) a heterogeneous population harboring 10% bad guys who engage in criminal activity on every occasion, and 90% good guys who never do. Assume that in actual fact, situation (2) obtains, but that law enforcement personnel erroneously belief situation (1) is the correct model. Let an enforcement agent observe the behavior of a randomly chosen individual from the population, and also be informed about the aggregate rate of criminal offending. No matter what the observed outcome at the individual level, this enforcement agent never changes his beliefs, because in the aggregate 10% of the population is always reported to have engaged in criminal activity. Moreover, such an agent has no incentive to "tag" and keep track of the individual histories of the persons he encounters, since, given his model of what is generating his observations, he anticipates no gain from such tracking. But, unless data of this kind were collected and studied by law enforcement personnel, they would never learn that their model is the wrong one!
The example, while stylized, is not unrelated to the historical problems of race, as they have developed in our society. Race has salience in our culture, which makes it operate powerfully in many venues, precisely because it is common knowledge that people are taking it into account. Moreover, there are stigmas associated with race which alter the economic opportunities of blacks in subtle but profound ways. It is my conviction that the notion of market discrimination is simply not rich enough to capture the many troubling variants of this phenomenon.
Aigner, Dennis and Glen Cain, "Statistical Theories of Discrimination
in Labor Markets," Industrial and Labor Relations Review,
January 1977, 30, 175-87