The four-fifths rule: how to calculate adverse impact at each hiring step

The four-fifths rule says that if one group's selection rate is less than 80% of the rate for the group selected most often, that is generally treated as evidence of adverse impact. It comes from the federal Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4(D), and it is a rule of thumb, not a legal definition. To use it properly you run it at every step of a hiring process that removes people, not only on final hires, and you treat a result built on small numbers with caution in both directions.

This page works through a full hiring funnel, explains why a balanced bottom line does not protect a biased step, shows how to tell a real signal from noise, and sets out where the rule stands federally in October 2026. How the rule applies to AI screening tools specifically is covered in Title VII disparate impact and AI hiring tools.

What the rule says, and what it does not

The key sentences of 29 CFR 1607.4(D) say:

• A selection rate for any race, sex or ethnic group that is less than four-fifths of the rate for the group with the highest rate "will generally be regarded by the Federal enforcement agencies as evidence of adverse impact."
• Smaller differences may still be adverse impact where they are significant in both statistical and practical terms, or where the employer's actions have discouraged applicants disproportionately.
• Larger differences may not be adverse impact where they are based on small numbers and are not statistically significant.
• Where numbers are too small to be reliable, evidence over a longer period or from similar circumstances elsewhere may be considered.

The agencies' Questions and Answers (1979) add the framing that matters most: the rule "is not intended as a legal definition, but is a practical means of keeping the attention of the enforcement agencies on serious discrepancies." Passing it does not prove a step is fair, and failing it does not prove discrimination. It tells you where to look.

Two definitions from 29 CFR 1607.16 set up the arithmetic. A selection rate is the proportion of applicants or candidates who are hired, promoted or otherwise selected. Adverse impact is a substantially different rate of selection that works to the disadvantage of members of a race, sex or ethnic group. The Q&A also says a person who voluntarily withdraws, formally or informally, at any stage is no longer an applicant, which is why a clean applicant flow log records withdrawals separately.

The calculation in four steps

Question 12 of the Q&A sets out the method:

1. Calculate the selection rate for each group: the number selected divided by the number of applicants from that group.
2. Find the group with the highest selection rate.
3. Calculate each group's impact ratio: its selection rate divided by the highest rate.
4. Check whether any impact ratio is below 0.80.

The groups are compared by sex (men and women) and by race and ethnicity, each against the highest group in its own comparison. The Guidelines' recordkeeping section, 1607.15(A)(2), says determinations should be made at least annually for each group that is at least 2% of the relevant labor force or workforce.

A worked example across a hiring funnel

An invented example: a company hires for a customer success role. 250 people apply who meet the basic qualifications. The table compares men and women at each step; the same method applies to race and ethnicity groups.

What the example shows:

• The overall result hides the problem. Men and women both end up with a 6% offer rate, but the phone screen passes women at 70% of the rate it passes men. That step is the one to examine.
• The highest group can change from step to step. At the panel interview women had the higher rate, so men are compared against women there. A ratio of exactly 0.80 is not below four-fifths, so it does not trigger the rule, though it is worth watching.
• The size of the gap can be expressed as a shortfall. For women to reach four-fifths of the men's 40% pass rate at the phone screen, they would need a 32% rate, or 32 of 100, which is 4 more people. Framing it this way helps a hiring team see how much a step would need to change.

Why a balanced bottom line is not a defense

The Guidelines contain a "bottom line" concept in 1607.4(C): where the total selection process shows no adverse impact, the agencies, in usual circumstances, will not expect the employer to evaluate each component. That is a statement about agency enforcement, not about liability. In Connecticut v. Teal, 457 U.S. 440 (1982), the Supreme Court held that a nondiscriminatory bottom line does not prevent a plaintiff from establishing a case about a particular step and does not provide a defense to one, because Title VII protects individuals rather than groups as a whole. Title VII's 1991 amendment, at 42 U.S.C. § 2000e-2(k), likewise focuses on a "particular employment practice."

The practical lesson: measure every step that removes people, including knockout questions on the application, resume review, phone screens, assessments and interviews. The funnel table above is the minimum useful report.

Small numbers: when the ratio is noise

The Q&A's Question 21 says that where selecting one different person would shift the result from adverse impact against one group to that group having the higher rate, the agencies will not assume adverse impact without more evidence. An invented example of the trap:

A ratio of 0.56 looks alarming, and one hire reverses it. Three habits help with small numbers:

• Run the flip test. Move one selection from the higher group to the lower group, and see whether the conclusion survives.
• Pool data across several requisitions for the same job and the same step, or across a longer period, as 1607.4(D) contemplates.
• Use a significance test as well as the ratio. The Guidelines treat statistical and practical significance together. Have someone qualified run the test; do not pick whichever measure gives the answer you want.

The opposite trap matters too. With very large numbers, a ratio of 0.85 can be statistically significant and, under 1607.4(D), can still be adverse impact. Large employers should not treat 0.80 as a safe harbor.

Where the rule stands in October 2026

Federal agency enforcement based on the four-fifths rule is unlikely in the near term, and the Guidelines may be gone from the EEOC's regulations in 2027. The arithmetic stays useful anyway. Courts in private cases decide for themselves what evidence of impact means, some state and city laws build impact ratios into their own requirements (New York City's Local Law 144 bias audits report them for automated tools), and an impact ratio is still the quickest way for a hiring team to spot a step that is filtering on something other than the job.

Running it in practice

1. Collect demographic data separately. Use voluntary self-identification kept away from the people making decisions. Never ask recruiters or interviewers to guess.
2. Log every step and every exit reason, so withdrawals are not counted as rejections.
3. Run the funnel table quarterly or per hiring wave for each job family, plus a bottom-line row.
4. Investigate any ratio below 0.80 that survives the flip test: what does the step measure, is it tied to the job, and is there an equally effective alternative with less impact?
5. Fix the step, not the numbers. Structured questions and anchored scoring, described in how to reduce bias in interviews, change outcomes at the step itself. Adjusting scores or setting different cutoffs by group is itself prohibited by Title VII, § 2000e-2(l).
6. Keep the analysis with your hiring records and decide with counsel whether it is prepared under privilege.