Equality Act 2010 in hiring: adverse impact and how to check for it

The Equality Act 2010 has been UK law for over a decade, but the way it shapes hiring practice has changed sharply in recent years. The rise of structured assessment, AI-assisted scoring, and bias monitoring tools has shifted what "compliance" means in practice. Doing the right thing now means more than not actively discriminating; it means being able to demonstrate that your selection process treats different groups consistently, and being prepared to evidence that if challenged.

This article covers what adverse impact is, how to check for it using the four-fifths rule, and what to do if a check flags one. We are not lawyers; this is not legal advice.

What the Equality Act protects

The Equality Act 2010 protects nine characteristics: age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex, and sexual orientation.

In hiring, the Act prohibits four kinds of conduct: direct discrimination (treating someone less favourably because of a protected characteristic), indirect discrimination (applying a policy or practice that disadvantages people with a protected characteristic), harassment, and victimisation.

Indirect discrimination is the one that matters most for selection processes. A policy that looks neutral on its face can still be unlawful if it disadvantages a protected group, unless the policy is a "proportionate means of achieving a legitimate aim."

What adverse impact means

Adverse impact is the statistical pattern that flags possible indirect discrimination. It occurs when a selection process produces materially different outcomes for different protected groups, even if the process is the same for everyone.

A common example: an assessment that requires verbal fluency in English may produce systematically lower pass rates for candidates whose first language is not English. The assessment is the same for every candidate. The outcome is not.

Adverse impact does not automatically mean a process is unlawful. Indirect discrimination is only unlawful if it cannot be objectively justified. But adverse impact is the warning signal that says: a closer look is required.

The four-fifths rule

The four-fifths rule (sometimes called the 80% rule) is a simple statistical test used to flag adverse impact in selection. It originated in US federal guidance but is widely used in UK hiring practice as a triage threshold.

The rule: a protected group's selection rate should be at least 80% of the rate for the group with the highest selection rate. If it falls below 80%, the process likely shows adverse impact and warrants investigation.

The calculation in plain numbers. Say 100 men and 100 women apply for a role. 30 men are shortlisted (30%) and 18 women are shortlisted (18%). Divide the female rate by the male rate: 18% divided by 30% equals 60%. Sixty percent is below the 80% threshold, so this process flags adverse impact against women.

Now say 30 men and 27 women are shortlisted. Female rate divided by male rate: 27% divided by 30% equals 90%. Ninety percent is above the 80% threshold, so the process does not flag adverse impact.

The four-fifths rule is not itself a legal test. UK courts and tribunals do not apply it as a rule of law. But it is the most widely used practical screen, and any responsible hiring process should run it as part of standard monitoring.

When the four-fifths rule is unreliable

The rule has known limitations and should never be the only check.

It is unreliable on small samples. A sample of 5 or 6 candidates per group can produce wildly variable ratios just by chance. Most guidance suggests at least 30 candidates per group before the rule produces meaningful signal.

It does not account for chance variation. A selection process can show 79% by random fluctuation rather than systematic bias. Statistical significance tests are more rigorous but harder to apply.

It compares two groups at a time. Adverse impact analysis ideally checks every protected characteristic against every other, not just one comparison.

For these reasons, the four-fifths rule is best treated as a triage signal that something may need investigation, not a definitive verdict on whether a process is lawful.

What to do if a check flags adverse impact

Adverse impact does not require you to abandon the selection process. It requires you to investigate and either fix it or justify it.

Three responses, in order of preference.

Investigate the cause. Is there a specific element of the process producing the disparity? An interview question, a screening criterion, a test format, a particular assessor? Often adverse impact comes from a small, fixable component rather than the process as a whole.

Redesign or remove that element. If a step in the process is producing disparate outcomes without clear job-relatedness, change it. Adverse impact reduction is rarely about lowering standards; it is usually about removing a process feature that does not actually predict performance.

Justify the process if you cannot redesign it. Some selection criteria are job-related and cannot be substituted (a heavy-lifting role genuinely requires lifting capability). In these cases, you need to be able to show two things: the criterion is genuinely necessary for the role, and there is no less-discriminatory way to test for it. This is what UK law calls "objective justification."

Document everything. Whatever you decide, document it. The Fair Work Agency, employment tribunals, and equality bodies will assess your process based on what you can evidence, not what you intended.

What you need in place

Five practical things.

Track demographic data on candidates where you can. You cannot run an adverse impact check without the data. Most candidates will provide voluntary demographic information at application stage if you explain why and assure confidentiality.

Run the four-fifths check at the end of each major recruitment round. Once you have at least 30 candidates per protected group, run the calculation. Save the results.

Document selection criteria up front. If a criterion is job-related, write down why before the round starts, not after a challenge.

Train assessors and interviewers consistently. Inconsistent assessment is one of the largest sources of unintentional adverse impact.

Have a process for what happens if a check flags impact. Who reviews? What gets changed? Who signs off? Document this in your hiring policy.

Where to look for the detail

The Equality and Human Rights Commission's Employment Statutory Code of Practice is the authoritative UK guide and is available at equalityhumanrights.com. Acas guidance on discrimination is at acas.org.uk/discrimination-and-the-law. The full Equality Act 2010 is at legislation.gov.uk/ukpga/2010/15/contents.

If your organisation has not run an adverse impact check on its hiring process in the last 12 months, it should. The cost of running the check is minimal. The cost of defending an indirect discrimination claim with no monitoring data is significant.

How PRODICTA fits in

PRODICTA's scoring engine is designed with Equality Act compliance built in. No candidate is penalised for spelling, grammar, or writing style. Every score is anchored to behavioural anchors rather than free-form AI judgement, so the same evidence produces the same score regardless of who is reviewing it.

For employer accounts, PRODICTA also runs three adverse impact checks across hiring rounds: range breadth, floor score, and cliff edge. Where data is sufficient, the four-fifths rule is applied to flag any protected group whose selection rate falls below the threshold. The output is the Bias-Free Hiring Certificate, a downloadable PDF that documents the methodology, the checks performed, and the results, with a reference number for audit.

This is what evidence-based fair hiring looks like in practice. To see how it works, the demo at prodicta.co.uk/demo includes the EDI dashboard.