Compare selection rates, group against group
The four-fifths rule does not look at how many people from each group were hired. It looks at the rate at which each group was selected from the pool that was actually considered. A group can be a small share of your applicants and still clear the rule, or a large share and still trip it, because the test is about the odds of being selected, not the headcount. The math is four short steps.
Running the numbers on one hiring round
Say you posted a role, two groups applied, and you want to check the screen before you make offers. You count only the people who actually applied for this role, not your whole workforce, and you run the rates.
Top rate 60%
The group with the highest selection rate had 80 applicants and 48 offers. 48 divided by 80 is 0.60, so the rate is 60%. This becomes the benchmark every other group is measured against.
Compared group 42%
A second group had 50 applicants and 21 offers. 21 divided by 50 is 0.42, so the rate is 42%. On raw counts this looks unremarkable, which is exactly why the ratio matters.
Impact ratio 0.70, a flag
42% divided by 60% is 0.70, or 70%. That sits below the four-fifths line, so this round is flagged for adverse impact. The flag does not say you discriminated; it says look at why the rates diverged and whether the screen is doing real work.
If the second group’s rate had landed at 48% instead, the ratio would have been 0.80 exactly, and the round would clear the rule. That small gap, six offers either way in this example, shows how close these calls can be and why the next two sections matter.
A screening flag, not a finding of discrimination
A flag means the process deserves a closer look, nothing more. The guidelines themselves are careful about this in both directions. A gap smaller than four-fifths can still count as adverse impact when it is significant in statistical and practical terms, and a gap larger than four-fifths may not count when it rests on small numbers that are not statistically significant. So the bare ratio is a first read, not the last word.
When a process is flagged, the legal question that follows under Title VII is whether the practice that caused the gap is job related and consistent with business necessity, and whether a less discriminatory alternative was available and skipped. A flag opens that inquiry. It does not settle it.
A pass is not a clean bill of health, and a flag is not proof. Clearing the rule does not show that a selection step is fair or job related; it can pass and still screen out a group for no good reason on a small sample. And failing it does not prove discrimination; it points you to the practice that produced the gap so you can test whether that practice is defensible. Treat the number as a flashlight, not a verdict.
Any decision that selects some people over others
Hiring is the textbook case, but the four-fifths rule travels. The same logic applies to who gets promoted, who is chosen for a development program, and who is kept or let go when a team is reduced. Anywhere you select a subset from a larger group, the selection rate exists and the check can run.
It also reaches the tools that make those decisions for you. Resume screeners, assessment cut scores, and AI hiring software all produce selection rates, and adverse impact is one of the first things they get tested against. An automated step that quietly screens one group at a lower rate carries the same exposure as a human one, and often more, because it runs at scale and leaves a record.
The rule gets unstable on small groups
The four-fifths rule was built for large applicant pools, and it wobbles badly on small ones. With a handful of candidates, one or two decisions swing the ratio across the line in either direction, which is the precise reason the guidelines say a gap based on small numbers that is not statistically significant may not be adverse impact at all. On a small team, run a significance test before you act on a bare ratio, and bring in help if a real decision rides on it. Three things trip up the calculation most often.
- Small pools swing the ratio.One extra hire can move a group above or below 80%. A flag from a tiny sample is noise until a significance test says otherwise, and the guidelines suggest running the analysis for any group that is at least 2% of the relevant workforce.
- The reference group can shift.The benchmark is whichever group has the highest selection rate in this round, not a fixed group you decide ahead of time. It can change from one hiring cycle to the next, so recompute it each time rather than assuming.
- The wrong population gets counted.Use only the people actually considered for the decision: applicants for a hire, candidates for a promotion, the pool in scope for a reduction. Folding in people who never entered the process distorts every rate and the ratio with it.
Federal enforcement has pulled back, the law has not
The rule lives inside a federal regulation, the Uniform Guidelines on Employee Selection Procedures, that the EEOC and three other agencies adopted in 1978 and that still appears in the Code of Federal Regulations. The thing that changed recently is enforcement posture, not the rule.
In April 2025, an executive order directed federal agencies to deprioritize disparate-impact enforcement and to review or unwind matters built on the theory. That is a real shift in how the federal government pursues these cases. It does not erase the underlying exposure, because an executive order cannot rewrite a statute or overrule the Supreme Court. Disparate impact is written into Title VII by the Civil Rights Act of 1991 and was recognized by the Supreme Court in 1971 in Griggs v. Duke Power. Private plaintiffs can still bring disparate-impact claims, and several state laws, including those in California, New York, and Illinois, keep the theory firmly in force.
The pullback deepened in June 2026. On 9 June 2026 the Justice Department’s Office of Legal Counsel, at the EEOC chair’s request, issued an opinion concluding that the EEOC’s longstanding disparate-impact guidelines are unconstitutional as applied: in the opinion’s view, impact evidence should function mainly as evidence of intentional discrimination rather than a freestanding basis for liability, and an employer’s defense need only show a reasonable business purpose rather than strict business necessity. An OLC opinion binds the executive branch’s enforcement posture; it does not amend Title VII, overrule Griggs or the 1991 Act, or bind federal courts, private plaintiffs, or state agencies. The practical read is unchanged: the check still takes minutes, private and state exposure still exists, and a flagged practice you cannot connect to the job is still a risk worth fixing.
So the federal scrutiny has eased while the risk in court and under state law has not. For an employer the practical move is unchanged: run the check, and when a process is flagged, confirm the practice that caused it is job related before you rely on it. This is also an area in motion, so the guidelines themselves could be revised; the date on this note is when these facts were last confirmed.
Before you change a selection process, defend a flagged one, or act on an adverse-impact finding, get legal advice. The analysis turns on facts a calculator cannot see, and the stakes are highest in a layoff or a termination, where a flagged selection can become the center of a claim. This is general information.
Where this comes from
Primary sources
- Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4(D). The federal regulation that defines the four-fifths rule: a selection rate below four-fifths (80%) of the highest group’s rate is generally regarded as evidence of adverse impact, while smaller gaps can still count if statistically and practically significant and larger gaps may not if based on small numbers. Adopted jointly in 1978 by the EEOC, the Civil Service Commission, the Department of Labor, and the Department of Justice. ecfr.gov, 29 CFR 1607.4Checked 2 June 2026
- Title VII disparate impact, 42 U.S.C. 2000e-2(k). The statute, added by the Civil Rights Act of 1991, that makes disparate impact unlawful and sets the burden-shifting test: the claimant shows a practice causes a disparate impact, the employer must show it is job related and consistent with business necessity, and the claimant may still prevail by showing a less discriminatory alternative existed. Rooted in Griggs v. Duke Power Co., 401 U.S. 424 (1971). law.cornell.edu, 42 U.S.C. 2000e-2Checked 2 June 2026
- EEOC, Employment Tests and Selection Procedures. The agency’s guidance explaining that determining whether a selection procedure has a disparate impact ordinarily requires a statistical analysis, and that a flagged procedure must be shown to be job related and consistent with business necessity. eeoc.gov, employment tests and selection proceduresChecked 2 June 2026
- Executive Order 14281, Restoring Equality of Opportunity and Meritocracy. Signed 23 April 2025 (90 FR 17537), directing federal agencies to deprioritize disparate-impact enforcement and to review matters built on the theory. The order does not amend Title VII or overrule Supreme Court precedent, so private and state disparate-impact claims remain available. federalregister.gov, EO 14281Checked 21 July 2026
- Department of Justice, Office of Legal Counsel opinion on the EEOC’s disparate-impact guidelines, 9 June 2026. Concludes, for the executive branch, that the EEOC’s disparate-impact guidelines are unconstitutional as applied, treats impact evidence primarily as evidence of intent, and softens the employer’s defense from business necessity to a reasonable business purpose. Not a court ruling: it does not amend Title VII, overrule Griggs, or bind courts, private plaintiffs, or state enforcement. justice.gov, OLC opinionsChecked 21 July 2026
Tools that run on this rule
Run the check, then act on it
A selection that fails the four-fifths rule can turn a layoff into a disparate-impact claim.
The free HR Operating System Readiness Scorecard profiles your hiring, compliance, and workforce planning, then shows where selection and adverse-impact risk is sitting in your process. No sign-up to see your profile.
Common questions
A quick check for unequal selection. You compare how often each group is chosen, and if one group is selected at less than 80% of the rate of the group chosen most often, the process gets flagged for a closer look. It comes from a 1978 federal regulation and is sometimes called the 80% rule.
No. A flag is the start of the analysis, not the conclusion. It points to a practice that produced a gap. The next question is whether that practice is job related and consistent with business necessity, and whether a less discriminatory alternative was skipped. A flagged process can be perfectly lawful once that is shown, and an unflagged one is not automatically safe.
The April 2025 executive order told federal agencies to deprioritize disparate-impact enforcement, but it did not change the law. Disparate impact is written into Title VII by the Civil Rights Act of 1991 and recognized by the Supreme Court in Griggs. A June 2026 Justice Department opinion goes further for federal enforcers, treating impact mainly as evidence of intent, but it binds only the executive branch. Private lawsuits and several state laws keep the theory in force, so the safe practice, running the check and confirming a flagged practice is job related, has not changed.
Use it with care. The ratio is unstable when only a few people are involved, because one decision can swing it past the line. The guidelines themselves say a gap based on small numbers that is not statistically significant may not be adverse impact, so on a small pool run a significance test rather than acting on the bare ratio, and get help if a real decision depends on it.
Any selection decision. Hiring is the common example, but promotions, training selections, and who is kept in a reduction all produce selection rates you can compare the same way. It also applies to automated screeners and AI hiring tools, which is why those get tested for adverse impact.