Discrimination, in its sociological meaning, involves highly complex social processes. The term derives from the Latin discrimination, which means to perceive distinctions among phenomena or to be selective in one’s judgment. Cognitive psychology retains the first of these meanings, popular usage the second. Individual behavior that limits the opportunities of a particular group is encompassed in many sociological considerations of discrimination. But exclusively individualistic approaches are too narrow for robust sociological treatment. Instead, sociologists understand discrimination not as isolated individual acts, but as a complex system of social relations that produces intergroup inequities in social outcomes.
This definitional expansion transforms ”discrimination” into a truly sociological concept. But in its breadth, the sociological definition leaves room for ambiguity and controversy. Obstacles to consensus on what constitutes discrimination stem from two sources—one empirical, the other ideological and political. First, deficiencies in analysis and evidence limit our ability to trace thoroughly the dynamic web of effects produced by discrimination. Second, because social discrimination is contrary to professed national values and law, a judgment that unequal outcomes reflect discrimination is a call for costly remedies. Variable willingness to bear those social costs contributes to dissension about the extent of discrimination.
The broadest sociological definitions of discrimination assume that racial minorities, women, and other historical target groups have no inherent characteristics that warrant inferior social outcomes. Thus, all inequality is seen as a legacy of discrimination and a social injustice to be remedied. By contrast, political conservatives favor a far narrower definition that limits the concept’s scope by including only actions intended to restrict a group’s chances. For solid conceptual reasons, sociologists seldom follow suit (but see Burkey 1978, p. 79). First, an intentionality criterion returns the concept to the realm of psychology and deflects attention from restraining social structure. Second, the invisibility of intentions creates insuperable obstacles to documenting discrimination.


Sociologists’ understanding of intricate societal patterns sensitizes them to the fact that disadvantage accruing from intentional discrimination typically cumulates, extends far beyond the original injury, and long outlives the deliberate perpetration. Many sociologists distinguish between direct and indirect discrimination (Pettigrew 1985). Direct discrimination occurs at points where inequality is generated, often intentionally. When decisions are based explicitly on race, discrimination is direct. Indirect discrimination is the perpetuation or magnification of the original injury. It occurs when the inequitable results of direct discrimination are used as a basis for later decisions (”past-in-present discrimination”), or decisions in linked institutions (”side-effect discrimination”) (Feagin and Feagin 1986). Hence, discrimination is indirect when an ostensibly nonracial criterion serves as a proxy for race in determining social outcomes.
To illustrate with wages, direct discrimination exists when equally qualified blacks and whites or men and women are paid at different rates for the same work. Indirect discrimination exists when the two groups are paid unequally because prior discrimination in employment, education, or housing created apparent differences in qualifications or channeled the groups into better- and worse-paying jobs. This direct/indirect distinction resembles the legal distinction between disparate treatment and disparate impact. While intentional direct discrimination may have triggered the causal chain, the original injury is often perpetuated and magnified by unwitting accomplices. Intentionality criteria deny that the continuing disadvantage is a legacy of discrimination.
Years ago, Williams outlined the concept differently: ”Discrimination may be said to exist to the degree that individuals of a given group who are otherwise formally qualified are not treated in conformity with these nominally universal institutionalized codes.” (Williams 1947, p. 39, italics added). For Antonovsky (1960, p. 81), discrimination involves ”. . . injurious treatment of persons on grounds rationally irrelevant to the situation” (italics added). Economists use starker terms. Becker (1968, p. 81) held that economic discrimination occurs ”. . . against members of a group whenever their earnings fall short of the amount ‘warranted’ by their abilities” (italics added).
Two problems arise with these definitions. First, the assessment of ”abilities” and of what treatment is ”rationally” relevant or ”warranted” is no easy task. Critical examination of common practice uncovers many instances where formal qualifications and ”nominally universal institutionalized codes” prove not to provide a logical basis for distinctions. Employment testing litigation demonstrates that when hiring criteria once legitimized by tradition or ”logic” are put to scientific test, they often fail to predict job performance in the assumed fashion. Analogous fallacies have been identified in the conventional wisdom guiding admission to advanced education. Hence, nominally universalistic standards may provide an altogether illogical basis for decision making. If such misguided selection procedures also work to the disadvantage of historical victims of discrimination, these practices are not exempted from the charge of discrimination by their universalistic facade.
The second problem with these definitions is that they ignore another, prevalent form of indirect discrimination. Even where nominally universalistic standards do serve some legitimate social function, such as selecting competent workers, adverse impact of these standards on those who bear the cumulated disadvantage of historical discrimination cannot be disregarded.
The complexity of discrimination and unresolved issues about its definition impede easy application of social science methods to inform institutional policy. Apparently rigorous quantitative analyses often only camouflage the crucial issues, as critical examination of wage differential decompositions reveals.


Assessments of discrimination produced by decomposing gross race or gender differences in wages or other social outcomes are common in sociology (e.g., Corcoran and Duncan 1978; Farley 1984, Rosenfeld and Kalleberg 1990) as well as in economics (e.g., Gill 1989). One segment of the gross intergroup differential is defined by its empirical linkage to ”qualifications” and other factors deemed legitimate determinants of social rewards. The second, residual segment not de-monstrably linked to ”legitimate” determinants of the outcomes often is presented as the estimate of discrimination. However, in the absence of better information than usually available and greater agreement on what constitutes discrimination, no unique estimate is possible. Through their choice of control variables to index ”legitimate” determinants of social outcomes and their interpretation of findings, researchers wittingly or unwittingly shape their answers. Any appearance of scientific certitude is an illusion. For example, estimated proportions of the gender earnings gap caused by discrimination in the United States range from Sanborn’s (1969) 10 percent or less to Blinder’s (1973) 100 percent. Predictably, each has been challenged (Bergman and Adelman 1973; Rosensweig and Morgan 1976).
An Illustrative Decomposition Study. An analysis of gender differentials in faculty salaries at a large university illustrates the difficulty of separating ”legitimate” wage differentials from inequity (Taylor 1988). About 90 percent of a $10,000 gender difference in faculty salaries was empirically linked to three factors widely considered legitimate determinants of faculty pay: academic rank, age, and discipline. Women tend to hold lower academic rank, to be younger, and more often to be affiliated with poorly-paid disciplines than men. Insofar as women’s lower salaries are linked to rank, age, and discipline, is the salary differential untainted by gender discrimination? Conventional wage differentials imply an unequivocal yes. If a simple answer is given, it should be no. But in truth, when policy makers ask for dollar estimates of inequity that the institution is obliged to remedy, the answers are neither unequivocal nor simple.
If the university’s promotion system has operated fairly, a gender gap reflecting differences in rank may be warranted. If gender bias has existed in the university’s promotion system, depressing the average academic rank of women faculty, the resulting deficit in women’s salaries reflects indirect discrimination. Attention should then be directed to the offending promotion processes. However, direct salary adjustments may also be in order, because gender bias in promotions weakens the link between rank and merit. Inequitable depression of women’s ranks would not necessarily lessen their actual contributions to the faculty, just their status. Using rank as a universalistic determinate of salary would then undermine the goal this practice is claimed to promote—the matching of rewards to contributions.
Salary differences tied to the age differential of female and male faculty also raise troublesome questions. If the relative youth of women faculty reflects lower retention and higher turnover as a result of discriminatory review processes or generally inhospitable conditions, salary differentials tied to age differences are again examples of indirect discrimination. The evidence would signal a need for institutional efforts to improve the retention of women faculty. But here it is not clear that salary adjustments are warranted. Because faculty contributions may be a function of experience, application of the universalistic age criterion is arguably reasonable. Any gender gap in salary tied to age differentials could, then, be both a legacy of discrimination and a reasonable conditioning of rewards on contributions. Complicating the issue further is the fact that affirmative action efforts often meet with greatest success in recruiting junior candidates. Thus, without supplementary data, it is not even clear whether an age-linked gender gap in salary reflects continuing institutional discrimination or affirmative hiring.
Gender differences in salary associated with discipline present even more complicated interpretational problems. Women and men are distributed across academic disciplines in a fashion that mirrors the gender distribution across occupations. Disciplinary differences in average salary likewise mirror wage differentials across occupations. But are the differing occupational distributions simply a matter of gender differences in preferences or abilities with no implication of discrimination? Or are women steered away from lucrative fields, so that gender differentials in salary linked to disciplinary affiliation represent indirect discrimination in training, recruitment, and hiring? Or does the pattern of occupational wage differentials mirrored in disciplinary differences represent direct discrimination, an influence of gender composition per se on occupational wage structures (England et al. 1988)? In the latter case, assignment of responsibility for remedy presents particular problems. The university, like any other single employer, is simultaneously vulnerable to competitive forces of the wider labor market and a constituent element of that market. Defiance of the market by a single organization is costly to that organization; adherence by all organizations to the broader occupational wage structure perpetuates gender inequity and carries broader costs.
This faculty salary study did not examine the role of scientific ”productivity.” But the inclusion of productivity measures among the control variables raises further difficulties. On standard ”productivity” measures, women faculty average lower scores than men (Fox 1990). Thus, an institutional study of salary differentials might find some segment of the male/female salary gap linked to productivity differences. Fox’s research demonstrates, however, that gender differences in scientific productivity reflect contrasting levels of resources that institutions provide to male and female faculty. Like age and rank, the gender difference in productivity may itself be a product of institutional discrimination. Thus, salary differentials based on male/female productivity differences also may represent indirect discrimination. The new ingredient here is that institutional shaping of productivity is subtle. Scientific productivity is ordinarily seen as an outgrowth of talent and effort, not potentially gender-biased institutional resource allocation. Solid documentation of this indirect discrimination process offers another challenge for researchers.


Critical reflection on this decomposition study highlights a set of interrelated points about the complex nature of discrimination and unresolved issues of remedy.
1. In American society today, the injuries of indirect discrimination are often far more extensive than those of direct discrimination. This conclusion does not imply that direct discrimination no longer exists (Reskin 1998). The continued operation of direct forms of discrimination is indicated by employment complaint records. American women and minorities have filed almost 1.5 million job discrimination complaints since 1965 (Blumrosen 1996, p. 4). In 1994 alone, over 150,000 such complaints were filed; 91,000 to local and state agencies and 64,000 to the U.S. Equal Employment Opportunity Commission (Leonard 1994, p. 24). Of course, not all complaints reflect genuine discrimination; on the other hand, not all discrimination prompts formal complaints. Many major corporations were found guilty of direct race or gender discrimination in the 1990s. And employment audits using paired, equally qualified applicants reveal widespread direct discrimination (Reskin 1998, pp. 25-29).
2. Apparently reasonable universalistic principles may on closer examination be unnecessary or even disfunctional. Scrutiny of employment criteria prompted by the Supreme Court’s 1971 Griggs v. Duke Power Co. decision has provided useful models for challenging nominally universalistic standards. Where it is possible to substitute standards that do as well or better at screening or evaluation without adversely affecting historical targets of discrimination, there are gains for all involved.
3. When gaps in actual qualifications are a legacy of discrimination, more extensive remedies are needed. Where training deficits impair employability, or inadequate preparation impedes admission to higher education, attention should be given to the earlier schooling processes that generated these deficiencies. This form of remedy aids future generations. In the meantime, compensatory training can reduce the liabilities of those who have already fallen victim to inferior schools.
4. Microcosms cannot escape the discriminatory impact of the societal macrocosm. Just as salary differences across academic disciplines reflect general occupational wage structures, institutions are often both prey to and participant in broader social forces. Narrow, legalistic approaches to remedy are inadequate for addressing this dynamic of discrimination.
5. Empirical research on group discrimination must mirror the phenomenon in its variety and complexity. The regression decomposition approach has proven useful but has its limitations (see also Dempster 1988, and the ensuing commentary). Regression analyses could provide more pertinent information if based on more homogeneous job groups (Conway and Roberts 1994) and on structural equation models that test reciprocal causation. Most important, if the aim is to guide policy, a framework far more complex than the dichotomous discrimination-or-not approach is required. The sociological arsenal of methods offers other promising approaches. Research that traces the actual processes of institutional discrimination is essential (e.g., Braddock and McPartland 1987, 1989). Also needed is attention to victims’ perceptions of discrimination (e.g., Feagin and Sikes 1994) and investigation of the changes generated by anti-discrimination efforts. Another approach involves cross-national comparative research, which we consider below.


Direct racial and gender discrimination in the United States has declined in recent decades— more slowly in the 1980s and 1990s than in the 1960s and 1970s. But what caused this decline? Many factors were involved, but governmental intervention was an important impetus. For example, blacks in South Carolina made dramatic economic gains in manufacturing during the late 1960s. Heckman and Payner (1989) demonstrated that human capital, supply shifts, and tight labor markets could not explain the sudden improvements. It was federal anti-discrimination programs, they concluded, that made a decisive contribution to the gains.
More general assessments also show that antidiscrimination legislation did reduce direct job discrimination nationally (Burstein 1985). It did not, however, eliminate the problem. Nor did such laws effectively attack indirect discrimination. For this more difficult problem, affirmative action programs were needed and have had some success (Reskin 1998). The resistance to such programs, however, underscores the difficulty of establishing effective remedies for the more subtle forms of discrimination.


Beyond racial and gender discrimination in the United States, the same basic concerns and principles arise for other nations and targets. Discrimination against Western Europe’s new immigrant minorities is pervasive (Castles 1984; MacEwen 1995; Pettigrew 1998). Both direct and indirect discrimination are involved, though the indirect forms are largely unrecognized in Europe.
Investigators have repeatedly uncovered direct discrimination in England (Amin et al. 1988; Daniel 1968; Gordon and Klug 1984; Smith 1976). Controlled tests reveal the full litany of discriminatory forms involving employment, public accommodations, housing, the courts, insurance, banks, even car rentals. Employment discrimination poses the most serious problem. In every European Union nation, minorities have far higher unemployment rates than the majority group. In 1990 in the Netherlands, Moroccans and Turks had unemployment rates above 40 percent compared with the native Dutch rate of 13 percent (Pettigrew and Meertens 1996). During the 1974-1977 recession, West German manufacturing reduced its labor force by 765,000—42 percent of whom were foreign workers (Castles 1984, p. 148).
As in the United States, there are many reasons for minority unemployment disparities. The ”last-in, first-out” principle selectively affects the younger minority workers. Typically less skilled, they are more affected by job upgrading. Minorities also are more likely to be in older, declining industries. But these patterns are not accidental. Planners put minorities into these industries for cheaper labor precisely because of their decline. In addition, these multiple factors offer insufficient explanations for the greater unemployment of minorities. Veenman and Roelandt (1990) found that education, age, sex, region, and employment level explained only a small portion of the differential unemployment rates in the Netherlands.
Indirect discrimination arises when the inability to obtain citizenship restricts the opportunities of non-European Union minorities. It limits their ability to get suitable housing, employment, and schooling. A visa is necessary in order to travel to other European Union countries. In short, the lives of Europe’s non-citizens are severely circumscribed (Wilpert 1993). Castles (1984) contends that the guest-worker system that brought many of the immigrants to Europe was itself a state-controlled system of institutional discrimination. It established the newcomers as a stigmatized ”out-group” suitable for low-status jobs but not citizenship. Widespread indirect discrimination was inevitable for these victims of direct discrimination.
Anti-discrimination remediation has been largely ineffective in Europe. Basic rights in Germany are guaranteed only to citizens. So, the disadvantages of non-citizenship include limited means to combat discrimination. There is extensive German legislation to combat anti-Semitism and Nazi ideology, but these laws have proved difficult to apply to non-citizens. The German constitution explicitly forbids discrimination on the basis of origin, race, language, beliefs, or religion—but not citizenship. Indeed, the Federal Constitutional Court has ruled that differential treatment based on citizenship is constitutional if there is a reasonable basis for it and if it is not wholly arbitrary. Hence, a German court upheld higher taxes for foreign bar owners than German bar owners. And restaurants can refuse service to Turks and others on the grounds that their entry might lead to intergroup disturbances (Layton-Henry and Wilpert 1994).
Few means of combatting discrimination are available in France either. Commentators often view discrimination as ”natural,” as something universally triggered when a ”threshold of tolerance” (seuil de tolerance) is surpassed (MacMaster 1991). Without supporting evidence, this rationalization supports quotas and dispersal policies that limit minority access to suitable housing.
The Netherlands, United Kingdom, and Sweden have enacted anti-discrimination legislation that specifically applies to the new immigrant minorities. And the Dutch have instituted modest affirmative action programs for women and minorities (De Vries and Pettigrew 1994). Not coinci-dentally, these countries make citizenship easier to obtain than Germany. Yet this legislation has been largely ineffective for two interrelated reasons. First, European legal systems do not allow class action suits—a forceful North American weapon to combat discrimination. Second, European efforts rely heavily on individual complaints rather than systemic remediation. Britain’s 1976 Act gave the Commission for Racial Equality the power to cast a broad net, but individual complaints remain the chief tool (MacEwen 1995).
It is a sociological truism that individual efforts are unlikely to alter such systemic phenomena as discrimination. Mayhew (1968) showed how individual suits and complaints are largely non-strategic. Minorities bring few charges against the worst discriminators, because they avoid applying for jobs with them in the first place. Complaints about job promotion are common, but they are made against employers who hire minorities. Thus, effective anti-discrimination laws must provide broad powers to an enforcement agency to initiate strategic, institutionwide actions that uproot the structural foundations of discrimination. Mayhew’s American analysis proves to be just as accurate in Western Europe.


A comprehensive understanding of societal discrimination in both North America and Western Europe must encompass two propositions.
1. The long-lasting character of discrimination means that the effects typically outlive the initiators of discriminatory practices. Apart from its importance to the law, this feature of modern discrimination has critical implications for sociological theory. Discrimination is fundamentally normative; its structural web operates in large part independent of the dominant group’s present ”tastes,” attitudes, or awareness. Hence, models based primarily on individual prejudice or ”rationality,” whether psychological or economic, will uniformly understate and oversimplify the phenomenon.
2. Discrimination is typically cumulative and self-perpetuating. For example, an array of research on black Americans has demonstrated that neighborhood racial segregation leads to educational disadvantages, then to occupational disadvantage, and thus to income deficits (Pettigrew 1979, 1985). To be effective, structural remedies must reverse this ”vicious circle” of discrimination. Affirmative action programs are one such remedy.
Seen in sociological perspective, then, discrimination is considerably more intricate and entrenched than commonly thought. The complexity of discrimination presents major challenges to social-scientific attempts to trace its impact. This complexity also precludes any one-to-one correspondence between perpetration and responsibility for remedy. Broad social programs will be necessary if the full legacy of direct and indirect discrimination is finally to be erased.

Next post:

Previous post: