Adverse impact is a substantially lower selection rate for one race, sex or ethnic group than for another in a hiring or promotion process. Under the US four-fifths rule, a group's rate below 80 percent of the highest group's rate is generally treated as evidence of it.
How to calculate adverse impact
The method comes from the US Uniform Guidelines on Employee Selection Procedures, at 29 CFR 1607.4(D). The EEOC's Questions and Answers on the Uniform Guidelines set out the same four steps.
- Selection rate for each group:
selection rate = number selected / number of applicants. - Find the highest rate among the groups.
- Impact ratio for each group:
impact ratio = group's selection rate / highest selection rate. - Compare to 0.80. A ratio below four-fifths is generally regarded by federal enforcement agencies as evidence of adverse impact. A ratio above it generally is not.
The Guidelines add two qualifications. Smaller differences can still be adverse impact when they are significant in both statistical and practical terms. Larger differences may not be, when they rest on small numbers and are not statistically significant.
Worked example: a knockout question
Brightwater Clinics, a fictional healthcare group, adds a knockout question to its application for practice managers: "Do you have five or more years of management experience?" Applicants who answer no are screened out. After one quarter:
| Group | Applicants | Passed the knockout | Selection rate | Impact ratio |
|---|---|---|---|---|
| Group 1 | 300 | 150 | 50.0% | 1.00 |
| Group 2 | 200 | 70 | 35.0% | 0.70 |
| Group 3 | 40 | 18 | 45.0% | 0.90 |
Group 2's ratio, 35 / 50 = 0.70, is below 0.80, so the question shows adverse impact for that group. Group 3 passes the rule, but with only 40 applicants, a few people answering differently would move its ratio a lot, so the team watches it rather than drawing a conclusion.
The next step is not to drop Group 2's applicants into a separate cutoff (see the mistakes below). It is to ask whether five years is actually needed for the job. If the team's best current practice managers include people hired with three years, the knockout sits above the real hiring bar, and lowering it to three years is a less discriminatory alternative that serves the same purpose.
Why adverse impact matters
Under Title VII, a neutral-looking practice that causes adverse impact can be unlawful unless the employer shows it is job-related and consistent with business necessity, and even then a plaintiff can point to a less discriminatory alternative. This is the theory of disparate impact, written into the statute at section 703(k). The Uniform Guidelines tie the two together: where a selection procedure has adverse impact, the employer is expected to have evidence that it is valid for the job.
Federal enforcement priorities changed in 2025. Executive Order 14281 directs federal agencies to deprioritize enforcement based on disparate-impact liability. An executive order does not amend the statute, and state and city rules still apply. As of October 2026 the EEOC also lists a final-stage rule, RIN 3046-AB43, to rescind the Uniform Guidelines, so check the current text before relying on the four-fifths rule. New York City's bias audit rule, for one, requires impact ratios for automated hiring tools. This is a summary, not legal advice.
For teams using AI candidate scoring, impact ratios are the first number to check, because a tool can screen thousands of people with one hidden criterion.
Adverse impact vs disparate treatment
| Adverse impact (disparate impact) | Disparate treatment | |
|---|---|---|
| What happened | A neutral rule produces unequal outcomes | People are treated differently because of a protected trait |
| Intent required | No | Yes, intent or motive is central |
| Typical evidence | Selection rates and impact ratios | Comments, documents, comparisons of similar people |
| Typical defence | The practice is job-related and necessary | A legitimate non-discriminatory reason |
Common mistakes
- Treating 0.80 as a safe harbour. The EEOC's Q&A calls the four-fifths rule a rule of thumb, not a legal definition of discrimination. Passing it does not prove a process is fair.
- Checking only the final hire. Measure each stage. A knockout question can cause impact that later stages hide.
- Ignoring sample size. With small groups, one person can swing the ratio. Use significance tests and look across more periods.
- Fixing it with separate cutoffs. Title VII bars adjusting scores or using different cutoff scores by race, color, religion, sex or national origin. Change the criterion, not the threshold for one group.
- Not keeping the data. You cannot compute rates you did not record. See auditability.
Finding impact is the start of the work; bias mitigation covers what to change. With an automated tool, you can only test whether a criterion is job-related if the tool shows which inputs drove each score, which is the point of explainable AI. Running these checks on a schedule, with a named owner and a record, is part of AI governance.
How ScoringFactory approaches it
ScoringFactory ranks candidates against a team's bar and cites each score to the record, so a team can see which criterion moved a candidate and test whether that criterion is job-related. It does not reject anyone; the hiring team decides. Current data and access practices are on the trust page.
Frequently asked questions
What is the four-fifths rule?
The four-fifths rule is a US guideline from the Uniform Guidelines on Employee Selection Procedures. If a group's selection rate is less than 80 percent of the rate for the group with the highest rate, federal enforcement agencies generally treat that as evidence of adverse impact. It is a rule of thumb, not a legal definition of discrimination.
How do you calculate adverse impact?
Divide the number selected by the number of applicants for each group to get selection rates. Divide each group's rate by the highest group's rate to get its impact ratio. A ratio below 0.80 indicates adverse impact under the four-fifths rule. For example, rates of 30 and 50 percent give 30 / 50 = 0.60.
Is the four-fifths rule enough to prove a hiring tool is fair?
No. The EEOC describes it as a practical rule of thumb for spotting serious gaps, not proof either way. Small samples can make ratios unreliable, large samples can make small gaps significant, and a tool can pass overall while one stage or one intersectional group fails. Use it with significance tests and stage-by-stage checks.
What should you do if a selection step shows adverse impact?
Find which criterion causes the gap, then check whether that criterion is truly needed to do the job. If a less discriminatory alternative serves the same purpose, switch to it. Do not adjust scores or cutoffs for one group, which Title VII prohibits. Document the analysis. This is not legal advice; involve employment counsel.
Sources
- 29 CFR 1607.4: Information on impact (four-fifths rule), eCFR, U.S. Government, current text
- Questions and Answers to Clarify and Provide a Common Interpretation of the Uniform Guidelines on Employee Selection Procedures, U.S. EEOC, current text
- Executive Order 14281, Restoring Equality of Opportunity and Meritocracy, Federal Register via GovInfo, 2025
- Title VII of the Civil Rights Act of 1964, U.S. EEOC, current text
- 42 U.S.C. 2000e-2(k), burden of proof in disparate impact cases (Title VII, US Code 2023 edition), US Government Publishing Office
- EEOC RIN 3046-AB43: Rescission of Uniform Guidelines on Employee Selection Procedures, Office of Information and Regulatory Affairs (reginfo.gov), 2025