You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Labour-market discrimination occurs when workers who are equally productive receive different pay or face different employment opportunities because of a characteristic unrelated to their productivity — gender, ethnicity, age, disability, religion or sexual orientation. This is not the same as the legitimate wage differentials of Lessons 1–3, which arise from genuine differences in marginal revenue product. Discrimination is, at root, a market failure: prices (wages) fail to reflect the true value of what workers produce, so labour is misallocated, output is lost, and the distribution of income is distorted by something other than productivity. This lesson formalises discrimination using the demand-side (MRPL) framework you already know, examines the two great theories — Becker's taste-based model and the Phelps–Arrow statistical model — shows how each predicts a lower wage for the discriminated group, and evaluates whether competitive markets or government policy is the more powerful remedy. Throughout, the connective tissue is the idea that discrimination drives a wedge between a worker's productivity and their pay.
This lesson sits within Section 4.1.6 — The labour market of the AQA A-Level Economics (7136) specification, and feeds directly into the analysis of wage differentials, income distribution and inequality in 4.1.7 (Lessons 8–10). Discrimination is a leading explanation of why pay gaps persist even between equally productive workers.
Exam Tip: The cleanest way to earn analysis marks is to model discrimination, not just describe it: show that taste-based discrimination lowers the perceived MRPL of the discriminated group, shifting their labour-demand curve left and depressing their wage. Translating a social phenomenon into the demand-and-supply diagram is exactly the AO3 skill examiners reward.
Key Definition: Labour-market discrimination occurs when equally productive workers are treated differently — in pay, hiring, promotion or conditions — because of a characteristic unrelated to their productivity, such as gender, ethnicity, age, disability, religion or sexual orientation.
The crucial qualifier is equally productive. If two groups differ in pay because they differ in skills, experience or hours (genuine MRPL differences), that is a legitimate differential, not discrimination. Discrimination is specifically the unexplained residual — the pay or employment gap that remains after controlling for every legitimate productivity factor. This distinction is the heart of the topic and a frequent exam discriminator.
Although the gender and ethnic pay gaps are the most studied, discrimination operates across many dimensions, each with its own economic mechanism. Age discrimination affects both younger workers (assumed inexperienced or unreliable) and older workers (assumed less adaptable or more costly), and is increasingly significant as working lives lengthen. Disability discrimination combines lower demand for disabled workers' labour with the additional living and working costs disability can impose, producing a substantial disability employment gap. Discrimination by religion or belief and by sexual orientation operates through the same taste-based and statistical channels analysed below. The shared analytical thread is that, in every case, a characteristic unrelated to productivity is allowed to influence pay, hiring or promotion — driving the same wedge between productivity and reward that the rest of this lesson models.
The gender pay gap is the most widely reported measure, and it is routinely misunderstood.
Key Definition: The gender pay gap is the difference between the average (mean or median) earnings of all men and all women, expressed as a percentage of men's earnings. Equal pay is the separate principle that a man and a woman must be paid the same for the same or equivalent work.
The two are different concepts, and a large gender pay gap can coexist with full compliance on equal pay. The reason is composition: men and women are distributed differently across occupations and seniority levels. If women are concentrated in lower-paid occupations and under-represented in senior roles, the average woman earns less than the average man even if every employer pays men and women identically for identical jobs. The UK median full-time gender pay gap has fallen substantially over recent decades — from the high teens in the late 1990s to single figures more recently — but progress has slowed, and the gap is much larger in some sectors (finance, construction) than others (hospitality). Since 2017, larger UK employers have been required to publish gender-pay-gap data annually, a transparency measure rather than a binding requirement to close the gap.
Exam Tip: A high-scoring answer states explicitly that the gender pay gap is not the same as unequal pay for equal work, and that most of the gap reflects occupational segregation and vertical segregation (under-representation at senior levels) rather than illegal pay discrimination. This conceptual precision is worth real marks.
Gary Becker (1957), in The Economics of Discrimination, produced the first formal economic model. His insight was to treat prejudice as a preference — a "taste for discrimination" — that the prejudiced party is willing to pay to indulge. There are three sources:
Because employer discrimination lowers the perceived marginal revenue product of the discriminated group, it shifts their labour-demand curve leftward, from D to Ddiscriminated. With an upward-sloping market supply curve, the equilibrium wage for the group falls from W to Wd and their employment falls from L to Ld. The group is paid below the value of its true marginal product — the wage no longer reflects productivity.
Becker's most important — and most examinable — result is that in a competitive product market, employer discrimination is costly to the discriminator. A prejudiced firm voluntarily shrinks its labour pool and hires preferred but no-more-productive workers, raising its costs. A non-discriminating rival can hire equally productive workers from the disfavoured group at the lower wage Wd, undercut the discriminator, and win market share. Competition should therefore bid up the wage of the discriminated group and gradually drive the discriminating firms out of business — discrimination should erode over time.
This prediction is partly borne out: pay gaps have narrowed over the long run. But discrimination has not disappeared, which tells us something important about why Becker's tidy mechanism is incomplete. Customer discrimination cannot be competed away (the firm loses revenue, not just incurs a private taste). Monopsony power (Lesson 4) blunts the competitive pressure that is supposed to punish discriminators. And institutional inertia, networks and search frictions slow the entry of non-discriminating firms. The mature evaluation is that competition exerts some downward pressure on taste-based discrimination but is insufficient on its own — which is precisely the gap that government policy is designed to fill.
Exam Tip: Becker's "competition erodes discrimination" prediction is a magnet for evaluation questions. State the mechanism (prejudiced firms have higher costs and are undercut), then qualify it (customer discrimination, monopsony, frictions, the persistence of gaps). Concluding that markets help but cannot finish the job sets up the case for intervention.
Edmund Phelps (1972) and Kenneth Arrow (1973) developed a model that, strikingly, requires no prejudice at all. It arises purely from imperfect information.
Key Definition: Statistical discrimination occurs when an employer, unable to observe an individual worker's true productivity, uses an observable group characteristic (gender, ethnicity, age) as a proxy for unobservable individual traits (productivity, reliability, likely tenure), and treats individuals according to the group average.
Suppose an employer cannot cheaply observe how committed or productive a particular applicant will be, but believes — rightly or wrongly — that one group is on average more likely to leave for family reasons. Faced with two otherwise identical candidates, the employer favours the candidate from the group with the more favourable average. No animus is involved; the behaviour is privately rational given costly information. Yet it is unjust to the individual who does not match the group average, and it produces the same outcome — a lower wage and fewer opportunities for the disfavoured group — as taste-based discrimination.
Statistical discrimination has two features that make it especially pernicious and hard to remedy:
flowchart TD
A[Imperfect information on individual productivity] --> B[Employer uses group average as a proxy]
B --> C[Disfavoured group offered less training and promotion]
C --> D[Lower realised productivity for the group]
D --> E[Apparent confirmation of the original belief]
E --> B
Exam Tip: The key evaluative contrast is that Becker's discrimination is competed away over time, whereas statistical discrimination can be self-reinforcing and so persists even in competitive markets. Pairing the two theories — one optimistic about markets, one pessimistic — is a strong way to build a balanced 25-mark answer.
Discrimination also operates through occupational segregation — the concentration of groups into particular occupations and levels.
| Type | Description | Illustrative pattern |
|---|---|---|
| Horizontal segregation | Groups concentrated in different occupations | Women over-represented in nursing, primary teaching and care; men in engineering, construction and IT |
| Vertical segregation | A group concentrated at lower levels of the hierarchy | Women under-represented in board and chief-executive roles |
Barbara Bergmann's (1974) crowding hypothesis explains how segregation by itself depresses wages even without direct pay discrimination. If women (or any group) are excluded from a wide range of occupations and crowded into a narrow set, the resulting excess supply in the "crowded" occupations depresses wages there, while the artificially restricted supply in the excluding occupations raises wages there. The pay gap then emerges from the allocation of workers across occupations, not from unequal pay within any single occupation — which is exactly why the gender pay gap can be large even where equal-pay law is obeyed.
flowchart TD
A[Group excluded from many occupations] --> B[Crowded into a narrow set of occupations]
B --> C[Excess labour supply in crowded occupations]
C --> D[Depressed wages there]
A --> E[Restricted supply in excluding occupations]
E --> F[Higher wages there]
D --> G[Pay gap between groups]
F --> G
A crucial analytical and evaluative skill is separating the part of a pay gap that reflects genuine productivity differences (human capital) from the unexplained residual that signals discrimination. Human capital theory (Becker, 1964) holds that pay differences can arise legitimately from differences in education, training, experience and hours — all of which raise MRPL. Part of the observed gender pay gap, for instance, reflects the motherhood penalty: career interruptions and a greater likelihood of part-time work reduce accumulated experience and hence productivity-related pay. That portion is not, in the technical sense, discrimination — though it may itself be the product of unequal social structures and of statistical discrimination feeding back into women's incentives to invest in skills.
Key Definition: The unexplained residual is the portion of a measured pay gap that remains after controlling for every observable productivity factor (education, experience, hours, occupation). It is the standard empirical proxy for discrimination, since by construction it cannot be attributed to legitimate MRPL differences.
The figures are hypothetical, chosen to illustrate the decomposition. Suppose the raw (unadjusted) gender pay gap in a firm is 15%. A statistical study controlling for education, experience, hours and occupation finds that productivity-related factors account for 9 percentage points of the gap. The unexplained residual is then:
Unexplained gap=15%−9%=6%
This 6-percentage-point residual is the part most plausibly attributable to discrimination (whether taste-based or statistical). The decomposition matters enormously for policy: the explained 9 points might be tackled by human-capital and family policy (childcare, flexible working, encouraging continuous careers), whereas the unexplained 6 points is the proper target of anti-discrimination law and enforcement. A common exam error is to treat the whole raw gap as discrimination; the sophisticated point is that only the residual is, and that the explained part may itself be the downstream consequence of discrimination operating through expectations and opportunities (the self-fulfilling channel above).
Exam Tip: When a question gives you a raw pay gap, distinguish the explained component (legitimate MRPL differences) from the unexplained residual (the discrimination proxy), and match each to a different policy. This decomposition is one of the clearest ways to demonstrate AO3 analysis on a discrimination question.
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.