Employment Discrimination

Employment discrimination is the study of how and why workers who are equally productive—or would be, absent unfair treatment—end up with different jobs, pay, hours, and working conditions because of who they are.

7 frameworks, from 1920 to today · Labor Economics · Economics

still a working view ended · each leads to its entry below

Overview

Employment discrimination is the study of how and why workers who are equally productive—or would be, absent unfair treatment—end up with different jobs, pay, hours, and working conditions because of who they are1. In labor economics, the field asks three linked questions: does discrimination exist, what causes it, and what can be done about it? The stakes are concrete. If a firm pays a woman less than a man for the same work, or refuses to interview a candidate with a "Black-sounding" name, the labor market is not just allocating effort inefficiently; it is reproducing inequality across generations. The history of the subfield is a history of tools built to answer those questions, and of the arguments over what the answers mean.

The Prehistory: Institutionalists and the Problem of "Wage Differentials"

Before the 1950s, most economists treated group pay gaps as a matter of human capital or custom rather than discrimination. The institutionalist tradition—associated with John R. Commons and later John Dunlop—studied wage structures, internal labor markets, and the rules that governed hiring and promotion. Institutionalists documented segregated job ladders and "women's jobs" and "men's jobs," but they rarely formalized a theory of discriminatory behavior. The dominant neoclassical view, following the marginal productivity tradition, held that competitive markets would punish employers who indulged prejudice: a firm that refused to hire qualified Black workers or women would face higher costs and lose out to competitors1. Discrimination, on this view, was a temporary friction or a taste that markets would erode1.

That confidence was shaken by the persistence of large, durable gaps. In the United States, the Great Migration of Black workers out of the South and into industrial cities during the 1910s–1960s produced a vast natural experiment in labor market integration, yet Black workers remained concentrated in lower-paying jobs and faced explicit exclusion from unions and craft apprenticeships. The civil rights movement of the 1950s and 1960s made discrimination a legal and political question, and economists were forced to explain why markets had not competed it away.

Becker's Taste-Based Discrimination: Prejudice as a Price

The founding text of the modern subfield is Gary Becker's The Economics of Discrimination (1957)2. Becker, then a young economist at the University of Chicago, modeled discrimination as a "taste"—a preference that employers, workers, or customers were willing to pay to indulge. An employer with a taste for discrimination acts as if hiring a Black worker or a woman carries an extra psychic cost, so he hires that worker only at a wage discount1. The model's key prediction is that discrimination is costly to the discriminator: in a competitive market, firms that discriminate will have higher costs and will be driven out by non-discriminating firms, unless customers or coworkers also hold the taste.

Becker's framework was elegant and gave economists a way to talk about discrimination in marginalist terms. But it had an awkward implication: if competition erodes discrimination, why does it persist1? Becker's answer—that customer or coworker prejudice can sustain it—was plausible but hard to test1. The model also treated the "taste" as exogenous, saying little about where prejudice comes from or how it changes1. Still, taste-based discrimination remains the benchmark against which later theories are defined, and Becker's insight that discrimination has a price is still taught in every labor economics course.

Statistical Discrimination: Information and Group Averages

The next major turn came in the early 1970s, when Kenneth Arrow (1971) and Edmund Phelps (1972) independently developed the theory of statistical discrimination. The problem they addressed was that employers rarely know a worker's true productivity1. They observe noisy signals—education, experience, test scores—and they also observe group membership3. If employers believe that, on average, members of a group have lower productivity or higher variance in productivity, they may use group membership as a cheap signal, even if they are not prejudiced.

The canonical example is a young woman of childbearing age. An employer who expects that she may leave the labor force to raise children may be less willing to invest in her training or to hire her for a job with long promotion ladders, even if she personally intends to stay1. The employer is not necessarily hostile; he is using a group average to predict an individual's behavior1. The result is a self-fulfilling equilibrium: if women are denied training, they acquire less human capital, which confirms the employer's expectation13.

Statistical discrimination was a major advance because it explained persistent gaps without requiring animus, and because it made discrimination an information problem rather than a taste3. It also had a troubling implication: because the employer's belief may be rational given the data, market competition does not necessarily eliminate it3. The theory opened a research agenda on how group differences in turnover, training, and job attachment arise and persist3.

Measuring Discrimination: The Oaxaca–Blinder Decomposition

Theories of discrimination needed measurement. The central empirical problem is that observed pay gaps between groups reflect both discrimination and differences in productivity-related characteristics—education, experience, occupation, industry1. In the early 1970s, Ronald Oaxaca (1973) and Alan Blinder (1973) independently developed a decomposition that became the workhorse of the field.

The Oaxaca–Blinder decomposition splits a group wage gap into two parts: the portion explained by differences in observed characteristics (the "endowment" or "explained" part) and the portion attributable to differences in the returns to those characteristics (the "unexplained" or "coefficient" part)14. The unexplained part is often interpreted as a measure of discrimination, though it also captures unobserved productivity differences1. The method made the gap decomposable and comparable across datasets, and it remains the standard first pass in empirical work on pay gaps. Its limitation—that the unexplained residual is not proof of discrimination—has driven decades of methodological refinement.

Audit and Correspondence Studies: Measuring Treatment Directly

A parallel empirical tradition tried to measure discrimination directly rather than infer it from residuals. The audit study, developed in the 1960s and 1970s, sent matched pairs of testers—differing in race or sex but trained to present identical qualifications—to apply for the same jobs. If the testers received systematically different responses, the difference was attributed to discrimination1. Early audit studies, such as those conducted in the United States and Britain in the 1960s and 1970s, documented substantial differential treatment in hiring, housing, and services.

The correspondence study, a later refinement, replaced in-person testers with fictitious résumés sent in response to real job advertisements. The most influential example is Marianne Bertrand and Sendhil Mullainathan's 2004 study, which sent thousands of résumés with either "white-sounding" or "Black-sounding" names to employers in Boston and Chicago. The callbacks were substantially higher for the white-sounding names, a gap that persisted across occupation and industry. Correspondence studies became the gold standard for measuring hiring discrimination because they control for qualifications and avoid the ethical and logistical problems of in-person testing. Their limitations—they measure callbacks, not hiring; they capture only the first stage of the process—are well understood and have driven a wave of replication and extension.

Imperfect Competition and the Persistence of Discrimination

The Becker prediction that competition erodes discrimination depends on the assumption of competitive markets1. A later literature, building on monopsony and search models, showed that when employers have wage-setting power—because workers face search costs or because markets are concentrated—discrimination can persist. If a firm can pay a group less without losing all its workers, it may do so, and the wage gap becomes a source of profit rather than a cost. This "discrimination under imperfect competition" literature, developed from the 1990s onward, helped explain why gaps persist in concentrated labor markets and why minimum wages and unionization can sometimes narrow them.

Field Experiments and Interventions: What Works?

The most recent turn is experimental. Randomized field experiments have tested interventions designed to reduce discrimination—training programs, diversity statements, blind hiring, and changes in hiring procedures. The results have been mixed and often surprising. Some studies found that diversity training can backfire; others found that structured interviews and standardized evaluation reduce bias. The field has also used natural experiments, such as the introduction of antidiscrimination laws or the sudden availability of new information, to estimate the effects of policy. The credibility revolution in labor economics—the demand for clean identification—has pushed the subfield toward designs that can isolate causal effects rather than describe correlations.

The Present: A Field of Competing Tools and Open Questions

Today, employment discrimination is a mature but contested subfield. Taste-based and statistical discrimination remain the two theoretical poles, with a growing literature on "implicit" or "unconscious" bias that draws on psychology. The Oaxaca–Blinder decomposition is still the standard descriptive tool, but it is now supplemented by audit and correspondence studies, natural experiments, and structural models. The main open questions are familiar: how much of the residual gap is discrimination, how much is unobserved skill, and how much is the legacy of past discrimination working through human capital and networks. The field has also broadened beyond race and sex to study discrimination by age, disability, sexual orientation, and caste, and it has become more global in scope, with influential studies in Europe, India, and Latin America. What unites the field is the conviction that the labor market is not a neutral clearing mechanism—that who gets hired, promoted, and paid is shaped by beliefs about groups, and that those beliefs can be measured, modeled, and sometimes changed1.

References

  1. Employment discrimination (Definition, Tastes for discrimination, Statistical discrimination, Statistical, Critique of the neoclassical approach, From experiments), Wikipedia
  2. Taste-based discrimination (History), Wikipedia
  3. Statistical discrimination (economics), Wikipedia
  4. Kitagawa–Oaxaca–Blinder decomposition, Wikipedia

The frameworks of Employment Discrimination

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1920-1960
  2. 1957-Present
  3. 1960-Present

    Audit study

    Do equally qualified people get equal treatment?

    Fly to its tower in Nousfarer

  4. 1971-Present

    Statistical Discrimination

    How do group averages shape hiring decisions?

    Fly to its tower in Nousfarer

  5. 1973-Present
  6. 1990-Present
  7. 2000-Present

    Field Experiments and Intervention Evaluation

    Which interventions actually reduce discrimination?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Institutional and Descriptive Politic…Taste-Based DiscriminationAudit studyStatistical DiscriminationKitagawa–Oaxaca–Blinder decompositionDiscrimination under Imperfect Compet…Field Experiments and Intervention Ev…

Beyond the field

Where Employment Discrimination's ideas came from in other fields, and what they led to there.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Employment Discrimination is a planet in Nousfarer

Its city has a tower for each of these 7 frameworks. Land, walk in, and climb one concept at a time.

Land on Employment Discrimination

Related fields

Fields its ideas are linked to

More in Labor Economics

See where Employment Discrimination sits on the Map of Human Knowledge