Pay equity: how to audit, explain and close pay gaps

Pay equity vs the gender pay gap

Pay equity means people doing equal work, or work of equal value, are paid the same unless a legitimate, job-related factor explains the difference. The gender pay gap is a blunter measure: the difference between the average or median pay of all women and all men in a workforce, regardless of role.

The two answer different questions. The pay gap tells you where women and men sit in your organisation. If senior and highly paid roles are mostly held by men, the gap will be large even if every individual is paid fairly for their job.

Pay equity tells you whether comparable people are paid comparably. It is the measure that equal pay law cares about, and the one a claimant's lawyer will test. It is also one of the tests any total rewards programme has to pass: a package that is generous on average but unequal between comparable people will not hold up.

National figures show how large the headline gap still is. Eurostat's provisional figure for 2024 puts women's gross hourly earnings on average 11.1% below men's across the EU (Eurostat). In the UK, the Office for National Statistics puts the median gap at 6.9% for full-time employees and 12.8% for all employees in April 2025 (ONS).

Why it matters now: the rules by region

Pay equity has moved from good practice to legal exposure in most large markets. The rules differ, but the direction is the same: more disclosure, and a heavier burden on employers to explain any difference they cannot justify.

European Union

The EU Pay Transparency Directive (2023/970) had a transposition deadline of 7 June 2026. Employers with 250 or more workers report their gender pay gap annually from 7 June 2027; those with 150 to 249 report every three years from the same date; those with 100 to 149 every three years from 7 June 2031 (European Parliament summary).

Where reporting shows a gap in average pay of at least 5% in any category of workers that the employer has not justified on objective, gender-neutral criteria, and has not corrected within six months of submitting the report, the employer must carry out a joint pay assessment with worker representatives (Article 10 of the Directive). The Directive also shifts the burden of proof: once a worker shows facts suggesting discrimination, the employer has to prove there was none.

Most member states missed the June 2026 deadline, so national laws are arriving at different speeds. Check the law in each country where you employ people rather than relying on the Directive text alone.

United Kingdom

Employers with 250 or more employees must publish six gender pay gap measures each year: mean and median hourly pay gaps, mean and median bonus gaps, the share of men and women receiving a bonus, and the gender split in each pay quartile (GOV.UK). The snapshot date is 5 April for private and voluntary sector employers and 31 March for most public authorities, with publication due within a year (GOV.UK).

United States

The federal Equal Pay Act requires equal pay for men and women doing substantially equal work in the same establishment. Job content, not job title, decides whether roles are comparable, and all forms of pay count, including bonuses, stock options and benefits (EEOC).

States add more. California requires private employers with 100 or more employees, at least one of them in California, to file annual pay data reports with its Civil Rights Department (California CRD). Massachusetts gives employers an affirmative defence if they completed a self-evaluation that is reasonable in detail and scope and can show reasonable progress towards closing gaps (Mass.gov).

India

The Code on Wages, 2019 prohibits discrimination on the ground of gender in wages for the same or similar work, and replaces the Equal Remuneration Act, 1976. The Code on Wages and the other three labour codes came into force on 21 November 2025, with some provisions waiting on rules (Press Information Bureau). Listed companies covered by SEBI's BRSR Core also disclose gross wages paid to women as a percentage of total wages (SEBI).

Australia, the wider APAC region and the Middle East

Australia's Workplace Gender Equality Agency publishes the gender pay gaps of individual employers with 100 or more employees; its first release came in February 2024 (WGEA). Elsewhere in APAC and across the Gulf, reporting duties vary widely by country. Multinationals usually apply a single global audit standard and layer local reporting on top. For what you must disclose to candidates and employees, see our guide to pay transparency.

How a pay equity audit works

A pay equity audit compares like with like, explains every difference it can on legitimate grounds, and isolates what is left. It has five stages, and the first two decide whether the result will stand up.

1. Define scope and the pay elements

Decide which countries, entities and groups are in scope, and which characteristics you will test. Gender is the starting point everywhere. Race and ethnicity are common in the US and UK, but collecting that data is restricted or unlawful in some countries, so check before you ask.

Test base pay, then total cash (base plus bonus), then long-term incentives. Gaps often hide in bonus and equity rather than in salary.

2. Assemble and clean the data

You need one row per employee: job, level, function, location, hire date, time in role, full-time equivalent, base pay, variable pay, equity, performance rating and the characteristics being tested. Convert everything to annualised, full-time equivalent figures in one currency.

Expect this to take longer than the analysis. Wrong job codes, missing levels and part-time pay recorded as full-time are the usual faults.

3. Check the job architecture

Pay equity can only be tested against a credible view of which jobs are comparable. If levels were assigned inconsistently, a regression will treat a levelling mistake as a legitimate explanation and hide the problem.

The EU Directive expects pay structures based on gender-neutral job evaluation. If your levels were set by negotiation rather than evaluation, fix that first; consistent salary bands by level are part of the same work.

4. Build comparison cohorts

A cohort, sometimes called a pay analysis group, is a set of employees whose work is similar enough to be compared directly, such as all mid-level software engineers in one country. Cohorts that are too broad blur real gaps. Cohorts that are too narrow contain too few people to test.

5. Measure, test and review outliers

Calculate the unadjusted gap for each cohort, run the adjusted analysis, then look at individuals. A cohort with no significant average gap can still contain one person paid well below peers with the same profile, and that person is your next claim.

Regression, explained plainly

Multiple regression estimates how much of each person's pay is explained by legitimate factors, and then shows how much difference remains between groups once those factors are held constant. That remaining difference is the adjusted pay gap.

Think of it as a fair-pay predictor. The model learns, from your own data, how pay typically rises with level, location, time in role and similar factors. It then predicts each employee's pay and compares the prediction with what they are actually paid.

Reading the output

• Gender coefficient. When pay is modelled on a logarithmic scale, this is roughly the percentage difference between women and men with the same profile.
• Statistical significance. Whether the difference is larger than chance would produce in a group of that size. Small cohorts rarely show significance, which is not proof of fairness.
• R-squared. The share of pay variation the model explains. A low value suggests important factors are missing or pay is being set inconsistently.
• Residuals. Each person's distance from predicted pay. Large negative residuals mark individuals to review.

An illustrative example

Illustrative only. Northwind Fabrication, a fictional manufacturer, has 120 engineers in one country. Women's median base pay is 9% below men's: that is the unadjusted gap. Most of it reflects more men in senior levels.

After the model controls for level, location and time in role, women are paid about 3% less than men with the same profile. That 3% is the adjusted gap. It is smaller, but it is the part Northwind cannot explain, and the part it must fix.

Remediation and budgeting

Remediation means raising the pay of people paid less than the model says they should be, then fixing the processes that created the gap. Pay is never cut to close a gap; the EEOC is explicit that employers may not lower one sex's wages to equalise pay.

Setting the adjustment

There are two common approaches. The first raises everyone in the affected group whose pay sits below predicted pay, up to the predicted level. The second sets a tolerance and adjusts only those below it. The first costs more and closes more of the gap; the second is cheaper but leaves smaller differences in place.

Before approving either, re-run the model with the proposed increases applied. Adjustments based on a single averaged gap can over-correct some people and miss others.

Building the budget

The cost is the sum of the individual adjustments, plus their knock-on effect on bonus, pension and other pay-linked elements. Calculate it in each local currency.

Illustrative only. If Northwind's model identifies 14 engineers below predicted pay by an average of 4%, on salaries averaging 80,000 in local currency, the base cost is about 44,800 a year before pay-linked costs.

Keep the remediation budget separate from the merit budget. If equity adjustments come out of the merit pool, everyone else's increase shrinks, and the fix becomes a grievance.

Fixing root causes

• Set starting pay from the band and the role, not from candidates' previous salary.
• Require approval and a written reason for any offer, promotion or increase outside guidelines.
• Review promotion rates and time-to-promotion by group, not only pay.
• Train managers to explain pay decisions against your compensation philosophy.

Legal privilege and governance

In the US and some common-law countries, organisations often commission a pay equity analysis through legal counsel, so that the analysis may be protected by legal privilege while the organisation works out what it means and what to do.

Privilege is not automatic. It usually depends on the analysis being requested by counsel for the purpose of legal advice, and it can be lost if results are widely shared or relied on publicly. The rules differ between jurisdictions, so take advice locally before you start, not after you see the numbers.

Privilege has limits elsewhere too. Statutory reports under the UK regulations and the EU Directive must be produced and published regardless. Massachusetts law makes the self-evaluation itself a defence, which means it may need to be disclosed to be used.

Who should see what

• Keep the working model and individual-level findings to a small group: counsel, the reward lead and the analyst.
• Give leadership aggregated findings and the costed remediation plan.
• Decide in advance what you will tell employees, and whether you will publish results voluntarily.

How often to run it

Run a full pay equity audit at least once a year, finishing early enough that the remediation can be funded in the next salary review. Annual is also the cadence of the main reporting regimes in the UK and, for the largest employers, the EU.

Between audits, monitor. Test every offer, promotion and off-cycle increase against the model before approval. Gaps reopen quickly through new hires, because starting pay is where negotiation and prior salary have most influence.

Run an extra audit after events that reshape pay: an acquisition, a restructure, a new job architecture, a large hiring wave or a market adjustment for one job family.

Software vs consultant

Pay equity software is best for repeatable monitoring once the data and job architecture are sound. A consultant is best when the foundations need building, the population spans several countries, or the work needs independent judgement and legal coordination.

The two are not exclusive. A common pattern is a consultant-led first audit that fixes the job architecture and sets the model, followed by software that keeps it current.

Whichever you choose, ask how the model treats small cohorts, which factors it controls for by default, and whether it can show individual-level results. A dashboard that reports a single company-wide adjusted gap is not an audit.