Becker’s model: employer discrimination as a downward shift in the demand for a group’s labour, lowering their wage and employment while costing the employer forgone value; the discrimination coefficient; statistical discrimination; why a pay gap is not a measure of discrimination; the Equality Act and pay gap reporting. Through the ONS gender pay gap for 2026.
On 22 October 2026 the ONS published the gender pay gap for April 2026, from the Annual Survey of Hours and Earnings. Median hourly pay for full-time employees was £21.51 for men and £20.05 for women, a gap of 6.8%, continuing a slow decline. Across all employees the gap was 12.3%, because more women work part-time and part-time jobs pay less an hour; among part-time employees alone, women earned 2.1% more than men. The gap was 4.1% at ages 30 to 39 and 8.2% at 40 to 49, and largest among the highest-paid occupations.
Since 2017 employers with 250 or more staff have had to publish their own gaps; in 2025/26, 78% reported higher median pay for men. None of these numbers compares a man and a woman doing the same job for the same employer, which is what the Equality Act 2010 regulates. They compare the typical man with the typical woman across the whole labour market. Economists therefore separate two questions: how much of the gap is discrimination, employers paying equally productive workers less because of who they are, and how much is differences in occupation, hours and experience, which may themselves have discrimination somewhere behind them.
Source: ONS, Gender pay gap in the UK: 2026, 22 October 2026; House of Commons Library, The gender pay gap, 2026.
Labour market discrimination is treating workers differently in hiring, pay or promotion because of a characteristic, such as sex, ethnicity, age or disability, that does not affect their productivity. In Gary Becker’s model, taste-based discrimination means an employer, fellow workers or customers have a discrimination coefficient d: they act as if employing the group costs the wage plus d. The employer’s demand for that group’s labour is then MRP − d rather than MRP, so the group is paid less and employed less. Statistical discrimination is different: employers who cannot observe an individual’s productivity judge them by the average of their group, for instance assuming women will take career breaks, which penalises individuals who would not.
A pay gap is the difference in average or median pay between two groups across the labour market. It reflects discrimination together with occupational segregation, hours, experience and qualifications. Equal pay, the same pay for equal work with the same employer, has been required by law since 1970 and is now in the Equality Act 2010; a pay gap can exist with no equal pay breach, and discrimination can exist in hiring and promotion with no pay gap inside a grade.
In April 2026 the median full-time woman employee earned £20.05 an hour and the median man £21.51: a gap of 6.8%. Across all employees the gap was 12.3%.
The full-time gap was 4.1% for employees aged 30 to 39 and 8.2% for those aged 40 to 49. Among part-time employees women earned 2.1% more than men. The Equality Act 2010 makes it unlawful to pay a woman less than a man for equal work.
Is the 6.8% gap a measure of discrimination?
A pay gap is an outcome. Discrimination is one cause among several. This lesson draws what discrimination does to a group’s wage and employment, why competition should erode it and often does not, and how to separate it from the rest of the gap.
The labour market for one group of workers. Supply slopes up; demand is their marginal revenue product, what each extra worker adds to revenue. Without discrimination the market clears at E1: £20 an hour, 100,000 employed, and the wage equals MRP. Profit-maximising employers hire anyone whose output is worth more than their pay, whoever they are.
Check: Supply of this group’s labour is W = 10 + 0.1L and demand, their marginal revenue product, is W = 40 − 0.2L. Without discrimination the wage and employment are:
10 + 0.1L = 40 − 0.2L gives L = 100 and W = 20. Each worker is paid what the last one adds.
Becker’s model. Employers with a discrimination coefficient d = £9 value this group’s work at MRP − 9, so demand falls to D2. The market clears at E2: £17 an hour and 70,000 jobs. The group is paid 15% less than equally productive workers and 30,000 of them are not hired at all. Discrimination shows up as both a pay gap and an employment gap.
Check: Employers now carry a discrimination coefficient of £9 against this group, acting as if each worker costs the wage plus £9. Their demand becomes W = 31 − 0.2L. The new wage and employment are:
10 + 0.1L = 31 − 0.2L gives L = 70 and W = 17. The group loses £3 an hour and 30,000 jobs.
The model’s sharpest result. At E2 the 70,000th worker is worth £26 and paid £17: the employer forgoes £9 an hour to indulge a prejudice. A firm without it can hire the group cheaply and earn more. In a competitive market discrimination is costly and should shrink. That it persists tells you markets are not that competitive, or that the discrimination comes from customers or co-workers, whose tastes the employer passes on.
Teacher tip: Ask who pays for discrimination here. The group loses £3 an hour and 30,000 jobs; the employer loses £9 an hour on the last worker. Students expect the discriminator to gain. Becker’s insight is that they do not.
Check: At E2 the 70,000th worker produces £26 of revenue an hour and is paid £17. A new employer with no prejudice could:
Becker’s point: discrimination costs the discriminator profit. Non-discriminating firms have an advantage and should expand at their expense.
The pay gap is bigger than the discrimination in it. Women are concentrated in lower-paid occupations, work part-time more often and lose experience to career breaks: 4.1% in the thirties, 8.2% in the forties. Statistical discrimination links the two: if employers assume any young woman may leave, they under-promote all of them, and the assumption becomes self-fulfilling. The diagram draws only d; the rest sits in the supply and MRP curves.
Check: The full-time gap is 4.1% at ages 30 to 39 and 8.2% at 40 to 49, and the all-employee gap is 12.3% against 6.8% full-time. This pattern suggests that a large part of the gap reflects:
The gap widens with age and with part-time working, which points to hours and experience; equal pay breaches would not follow that pattern.
Policy. The Equality Act 2010 makes unequal pay for equal work unlawful, which raises the cost of d. Pay gap reporting exposes large gaps to staff, customers and applicants. Childcare support, shared parental leave and flexible working attack the career-break part of the gap. Each policy either raises the cost of d, shifting D2 back toward MRP, or changes the supply and MRP curves themselves.
Check: Mandatory gender pay gap reporting for employers with 250 or more staff, introduced in 2017, is best evaluated as:
Reporting works through reputation and labour supply, not law. The full-time gap has fallen slowly; how much is due to reporting is hard to isolate.
The student shifted supply left, which raises the group’s wage to £26. Employer discrimination acts on demand: employers value the group’s work at MRP − d, so demand shifts down and the wage falls. A leftward supply shift describes something else, such as the group withdrawing from the market, and gives the wrong sign on the wage.
Evaluate the view that the gender pay gap is mainly the result of discrimination by employers.
Labour market discrimination is paying or hiring equally productive workers differently because of a characteristic unrelated to their productivity. In Becker’s model employers with a discrimination coefficient d value a group’s work at marginal revenue product minus d, so demand for the group shifts down, their wage falls and fewer are employed: on the diagram from £20 and 100,000 to £17 and 70,000. If the gender pay gap were mainly this, women doing the same work as men would be paid less by the same employers, and the gap would be roughly constant across ages and hours.
The ONS figures do not fit that shape. In April 2026 the full-time median gap was 6.8%, the all-employee gap 12.3% because more women work part-time, and among part-time employees women earned 2.1% more than men. The full-time gap was 4.1% at ages 30 to 39 and 8.2% at 40 to 49, and largest in the highest-paid occupations. That pattern points to occupational segregation, part-time hours and career breaks after childbirth rather than to unequal pay for equal work, which has been unlawful since 1970. Decomposition studies that control for occupation, hours and experience typically leave an unexplained residual of a few percentage points, and even that residual is an upper bound on direct discrimination rather than a measure of it. Becker adds a theoretical reason to expect direct discrimination to be limited: an employer who pays a worker £17 for £26 of output forgoes £9 an hour, so competitors without the prejudice can hire the group cheaply and win.
**However,** the composition that explains the gap is not independent of discrimination. Statistical discrimination, judging each woman by the average risk of a career break, lowers promotion and training for all women and makes the break more likely, so segregation into lower-paid roles can be discrimination moved one step back. Customer and co-worker prejudice are passed on by employers who are not prejudiced themselves, and where employers have monopsony power, as in many public services, competition does not erode the gap as Becker predicts. Pay gap reporting since 2017 found 78% of large employers with higher median pay for men, which equal-pay law alone has not closed.
On balance the view is wrong if it means employers paying women less for the same work: that is a small part of the 6.8% gap, and the law and competition both press on it. It is partly right if discrimination is read more widely, to include the hiring, promotion and household decisions that sort women into part-time and lower-paid work. The honest answer is that the measured gap is mainly composition, and composition is partly discrimination; policy that targets only equal pay will leave most of the gap where it is.
Pay gap figures: ONS, Gender pay gap in the UK: 2026 (released 22 October 2026, Annual Survey of Hours and Earnings): full-time median hourly earnings excluding overtime £21.51 for men and £20.05 for women, a gap of 6.8%; all employees 12.3%; part-time −2.1%; full-time gap 8.2% at ages 40 to 49 and 4.1% at 30 to 39. House of Commons Library, The gender pay gap: 78% of reporting employers had higher median pay for men in 2025/26. The wages and employment on the diagram are invented. Diagrams drawn with Rough.js and perfect-freehand, both MIT licensed.