Four-Fifths Rule: How to Calculate Adverse Impact (2026)

The four-fifths rule is the quickest test US employers have for spotting a selection step that may be screening out one group more than others. It takes a calculator and the pass counts from your ATS. This guide shows the calculation with a worked example, explains where the rule breaks down, and sets out how to run it across a hiring funnel.

This is a plain-English explainer written for talent teams. It is not legal advice; if a ratio comes out low, talk to employment counsel before you change anything.

What is the four-fifths rule?

The four-fifths rule is a threshold in the federal Uniform Guidelines on Employee Selection Procedures (1978): a selection rate for any race, sex or ethnic group that is less than four-fifths (80%) of the rate for the group with the highest rate will generally be regarded as evidence of adverse impact.

The Guidelines were adopted jointly by the EEOC, the Department of Labor, the Department of Justice and the then Civil Service Commission, and they sit in the Code of Federal Regulations at 29 CFR part 1607. They define three terms you need:

• Selection rate: the proportion of applicants or candidates who are hired, promoted or otherwise selected.
• Adverse impact: a substantially different rate of selection that works to the disadvantage of members of a race, sex or ethnic group.
• Selection procedure: any measure used as a basis for an employment decision. That covers tests, work samples, interviews, application forms and experience or education requirements.

The EEOC's official Questions and Answers on the Guidelines say the rule "is not intended as a legal definition" and describe it as a rule of thumb. It gives you a first inference, nothing more. It also does not mean 20% discrimination is acceptable; the same Q&A says so directly.

How do you calculate the four-fifths rule?

You calculate it in four steps: work out each group's selection rate, find the highest rate, divide every group's rate by the highest, and check whether any result falls below 0.80.

1. Selection rate per group. Number selected ÷ number who applied or reached that stage.
2. Highest rate. Pick the group with the highest selection rate. This is your benchmark, and it is not always the largest group.
3. Impact ratio. Each group's selection rate ÷ the highest rate.
4. Compare with 0.80. A ratio below 0.80 is a flag for adverse impact.

The EEOC's own example: 80 White applicants with 48 hired is a 60% rate, and 40 Black applicants with 12 hired is 30%. 30 ÷ 60 is 0.50, well under 0.80, so adverse impact is indicated.

A worked four-fifths rule calculation

Here is a hypothetical example for an online assessment stage, by race and ethnicity. The numbers are illustrative.

Three things to notice.

• The benchmark is the highest rate, whichever group has it. Here the smallest group sets it, so a few more or fewer passes in that group move every other ratio.
• The Black group is flagged. 25.0% ÷ 35.0% = 0.71.
• The shortfall is small in headcount. To reach 0.80, the Black group would need a 28.0% pass rate, which is about 34 of 120 instead of 30. Four people. That is useful context when you investigate, and it is exactly why the next section matters.

Run the same calculation separately for sex (women and men). The Guidelines ask for records by sex and by race or ethnic group, and the EEOC Q&A says records should also let you identify people by sex combined with race or ethnicity.

Is the 80% rule enough on its own to prove adverse impact?

No. The 80% rule is a starting inference, and the Guidelines themselves say statistical significance and sample size can override it in either direction.

Section 1607.4(D) of the Guidelines sets out both directions:

• Smaller differences can still count where they are significant in both statistical and practical terms, or where your actions discouraged applicants from a group.
• Larger differences may not count where they rest on small numbers and are not statistically significant.

Why small samples mislead

With small numbers, one hire can swing the ratio across the line. Take a hypothetical senior role:

A ratio of 0.40 looks alarming, yet one decision moves it to the line. The EEOC Q&A gives a similar example and says that where one different selection would reverse the result, the agencies would not assume adverse impact without other evidence. The Q&A also says a lower rate that persists over time could still form a pattern. When numbers are small, the Guidelines allow you to look at a longer period or at how the same procedure performed elsewhere.

Why large samples can hide a real gap

At high volume, the opposite problem appears. The EEOC Q&A describes a screen that passed 90% of Hispanic applicants and 96% of non-Hispanic applicants. The ratio is about 0.94, comfortably above 0.80, yet the agencies would treat it as adverse impact because the difference is statistically and practically significant across a large number of people.

What to run alongside the ratio

Run a significance test whenever you compute a ratio. A two-proportion test or Fisher's exact test (better for small counts) tells you whether the gap is likely to be chance. In the worked example above, the Black and Asian pass rates differ by about 1.5 standard errors, which is short of conventional significance with these numbers. So the ratio flags the stage, the test says the evidence is not yet strong, and the sensible response is to keep collecting data and review the assessment content. The Q&A says to keep collecting impact information until you can determine whether adverse impact exists.

How do you monitor the four-fifths rule across a hiring funnel?

Calculate the impact ratio at every decision point (screen, assessment, interview, offer) as well as for the whole process, because an overall pass can hide one stage that screens a group out.

The Guidelines start with the "bottom line": if the overall process for a job shows no adverse impact, the agencies generally will not expect you to evaluate each component. The courts take a stricter view. In Connecticut v. Teal (1982) the Supreme Court held that a favourable bottom line was no defence for a test that screened out Black candidates at a higher rate and was not shown to be job-related. Title VII's disparate impact section also refers to each particular employment practice. Monitoring each stage is the safer habit.

Here is a hypothetical funnel for one role, by sex:

The overall ratio of 0.83 passes. The assessment stage, at 0.75, does not, and later stages happen to make up for it. That is the pattern a stage-by-stage check is designed to catch.

A monitoring routine that works

• Collect demographic data properly. Use voluntary self-identification, stored separately from the application and not used in hiring decisions. The EEOC Q&A describes this approach.
• Define each stage's denominator. Use the people who reached that stage, and decide in advance who counts as an applicant.
• Calculate per job, at least annually. The Guidelines ask for impact determinations at least once a year, job by job, for each group that is at least 2% of the relevant labour force or workforce. Monthly or quarterly checks catch problems sooner in high-volume hiring.
• Run the test on each stage and overall. Record the ratio, the counts and a significance result.
• Write down what you did about a flag. Note the stage, the cause you found and the change you made.

Small employers have lighter duties. Under the Guidelines, users with 100 or fewer employees keep simpler annual records and, per the EEOC Q&A, are not expected to make adverse impact determinations. The ratio is still a cheap way to check your own process.

What should you do if a stage fails the four-fifths rule?

Investigate the stage before you change it: confirm the numbers, test for significance, check whether the step measures what the job needs, and look for an equally valid step with less impact.

1. Check the data. Wrong denominators, duplicate applicants and missing demographic records can all produce a false flag.
2. Look at the content. Does each item or question relate to the work? Check requirements such as experience thresholds or degree filters, and every test section, against the tasks in the job description. Our guide to assessing cognitive ability covers how to use reasoning tests responsibly.
3. Confirm job-relatedness. Under the Guidelines, a procedure with adverse impact is treated as discriminatory unless it has been validated. Title VII lets an employer defend a practice that is job-related for the position and consistent with business necessity.
4. Consider alternatives. Where two procedures are substantially equally valid, the Guidelines say to use the one with lesser adverse impact. A structured interview scored with an interview scorecard, or a work sample of real tasks, is often the alternative teams test first.
5. Do not adjust scores by group. Title VII makes it unlawful to adjust scores, use different cut-off scores, or otherwise alter test results on the basis of race, colour, religion, sex or national origin. Fix the step itself.

Job simulations are one way to keep a stage tied to the work. Cohesyve takes a company's job description, sources candidates, does the outreach, puts them through AI job simulations of the real work and returns a ranked shortlist. Every candidate for a role gets the same simulation, which makes a stage easier to explain if a ratio is questioned. You still run the four-fifths check on your own applicant data. Plans are a subscription from $199/month billed annually, with 10 free candidates to start.

Does the four-fifths rule still apply in 2026?

For now, yes, though its federal status is changing. As of late October 2026 the EEOC's version of the Uniform Guidelines, including the four-fifths rule, is still in the Code of Federal Regulations, and Title VII still sets out disparate impact liability in statute.

Federal policy has moved quickly:

The EEOC rule's text was not public at the time of writing, and the Guidelines' recordkeeping rules are the subject of a separate, earlier-stage proposal. Until a final rule is published in the Federal Register and takes effect, the current Guidelines stand. The EEOC's 2023 technical assistance on applying the four-fifths rule to AI tools is also no longer on its website.

Rescinding the Guidelines would not remove disparate impact liability, which comes from Title VII itself and can still be pursued by private plaintiffs. State and city rules continue to apply, and New York City's bias audit rule uses the same impact-ratio arithmetic. The calculation in this guide stays a useful check either way.

New York City is the clearest example. Local Law 144 requires a bias audit within one year before using an automated employment decision tool, with impact ratios across sex, race/ethnicity and intersectional categories. The city's rule defines the impact ratio as a category's selection rate divided by the most selected category's rate, and says this is consistent with the four-fifths section of the Uniform Guidelines. Categories under 2% of the audit data may be excluded. The law does not require any specific action based on the results, though anti-discrimination law still applies. Our roundup of AI hiring laws in 2026 covers the other jurisdictions.

If you buy assessments, ask vendors for pass rates by group on your own population. Our guides for enterprise assessment platforms and high-volume hiring list the questions to put in an RFP.

Common questions

What is the 80% rule for adverse impact? It is another name for the four-fifths rule: if a group's selection rate is below 80% of the highest group's rate, federal agencies generally treat that as evidence of adverse impact.

Is the four-fifths rule a law? It is part of federal regulations (29 CFR 1607.4(D)) that guide enforcement agencies, but the EEOC's own Q&A says it is not a legal definition of discrimination, and the EEOC is moving to rescind that part of the regulations. Courts and agencies also weigh statistical significance.

Which groups does it cover? Race, sex and ethnic groups. The Guidelines do not apply to age discrimination under the ADEA or to disability under the Rehabilitation Act, which are handled under other rules.

Do I compare each group with the largest group? No. You compare each group with the group that has the highest selection rate, whatever its size.

What if a group has very few applicants? The ratio becomes unstable. Pool data over a longer period, run Fisher's exact test, and keep monitoring. Under the Guidelines, groups below 2% of the relevant labour force or workforce do not need a separate determination.

Does passing overall mean every stage is fine? No. Check each stage. The Supreme Court held in Connecticut v. Teal that a good overall result does not excuse a step that is not job-related.

Can I fix a low ratio by changing the cut-off for one group? No. Title VII prohibits different cut-off scores or adjusted scores by race, colour, religion, sex or national origin. Change or replace the step for everyone.