The Four-Fifths Rule (80% Rule): Origin, Application, and Limits

The four-fifths rule — also known colloquially as the “80% rule” — is the most cited and most misunderstood number in U.S. employment selection. It was codified in 1978 by the Uniform Guidelines on Employee Selection Procedures as a practical screen for adverse impact: if the selection rate for any protected group is less than four-fifths (80%) of the selection rate for the highest-selecting group, the procedure produces presumptive adverse impact and triggers further scrutiny. Almost every adverse-impact analysis in HR, every plaintiff complaint, and every defense brief reference this single threshold.

The rule has lasted nearly five decades because it is fast, transparent, and reasonably aligned with practical disparity at moderate sample sizes. But it was never intended as a safe harbor and never functioned well as one. Modern selection analytics, court precedent, and EEOC enforcement guidance treat the four-fifths rule as one indicator among several — useful but neither sufficient nor exclusive. Employers who rely on it alone systematically miss disparities that are statistically significant at large N, and conversely raise alarms about ratio failures that are not practically meaningful at small N. The rule is best understood as the front door to adverse-impact analysis, not the entire house.

Origin: The Uniform Guidelines and the 1978 Compromise

The Uniform Guidelines on Employee Selection Procedures (29 C.F.R. Part 1607) were jointly issued in 1978 by the Equal Employment Opportunity Commission, the Civil Service Commission, the Department of Labor, and the Department of Justice. They consolidated previously fragmented federal guidance on testing, screening, and selection into a single framework that applied across executive-branch enforcement. Section 4(D) introduced the four-fifths rule in plain language: a selection rate “less than four-fifths (4/5) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact, while a greater than four-fifths rate will generally not be regarded by Federal enforcement agencies as evidence of adverse impact.”

The phrase “generally be regarded” did substantial work. It signaled that the rule was a presumption, not a definition, and that other evidence — including statistical significance tests — could rebut it in either direction. The rule emerged as a compromise between mathematical precision and operational simplicity: a single ratio that line HR staff and small employers could compute by hand was deemed more useful than a complex test no one would apply.

How the Rule Works in Practice

Computing the four-fifths ratio is mechanical. For each demographic group of interest, divide the number selected by the number who applied at that step. The highest-resulting rate becomes the reference. Each other group’s rate is then divided by the reference rate. If any quotient falls below 0.80, the rule flags a potential adverse impact. The mechanic is the same whether the step is an initial resume screen, a cognitive assessment, an interview, or a final offer.

A worked example: ~120 men and ~80 women apply for an engineering role; ~60 men and ~32 women pass the technical screen. Selection rates are 50% for men and 40% for women. The ratio is 40/50 = 0.80, exactly at the threshold. A small change in either direction — one fewer woman passing — would tip the ratio below four-fifths and trigger adverse-impact concern.

Research Evidence on the Rule’s Limits

The technical literature has long documented the four-fifths rule’s sample-size sensitivity. Cohen, Aamodt, and Dunleavy (2010) demonstrate that at sample sizes below approximately ~30 per group, the rule misclassifies cases at high rates: small absolute differences produce ratios below 0.80 by chance, while real differences are masked by sampling noise. At very large sample sizes (>10,000), the rule frequently passes when statistical significance tests detect real, if small, disparities. The classical statistical fix — Fisher’s exact test or a z-test of two proportions — addresses the chance-versus-real-difference question that the four-fifths rule cannot answer alone.

Schmidt and Hunter (1998) provide the broader meta-analytic context: the most valid selection methods (work samples, structured interviews, cognitive ability) have well-documented mean group differences that often produce four-fifths failures even when the methods are job-related and consistent with business necessity. Sackett and Lievens (2008) reinforce that adverse-impact diagnosis and validity defense are separate questions, and that conflating them — using the four-fifths rule as either a green light or a red light — short-circuits proper analysis.

Data Notice: The 80% threshold and projected ratio outcomes discussed here reflect federal guidance and meta-analytic patterns. Specific case outcomes — including ~30-per-group sample-size guidance and projected misclassification rates — are summary indicators based on long-running research, not jurisdictional rulings. Application of the four-fifths rule in any specific matter requires legal counsel and case-specific analysis.

Practical Workflow: Combining the Rule with Significance Testing

Court-defensible programs apply the four-fifths rule as the first lens in a layered approach. Step one: compute selection rates and the four-fifths ratio for every funnel step and every protected group with sufficient sample. Step two: regardless of the four-fifths result, run a two-tailed Fisher’s exact test or z-test of proportions and report the p-value. Step three: compute and report effect sizes — Cohen’s d for continuous outcomes, odds ratios for binary outcomes — so that practical magnitude is visible alongside statistical significance. Step four: when any test flags a concern, investigate the underlying selection step; when all three converge on a flag, treat the disparity as material and either redesign the tool or document validity.

This layered approach pairs naturally with a structured-measurement infrastructure. Tools like /score/ and /assess/ produce auditable per-candidate evidence that supports both the diagnostic side (clean group comparisons) and the validity-defense side (clear job-relatedness). When upstream selection relies on validated tests and structured rubrics rather than free-form judgment, downstream adverse-impact analysis becomes more interpretable, not less necessary.

Common Pitfalls

Three errors recur. The first is over-reliance: programs that report only the four-fifths ratio and stop there miss significant disparities at large N and chase noise at small N. The second is the “passed-the-rule” defense: employers cite a four-fifths pass and decline to investigate further, even when significance tests, effect sizes, or candidate complaints suggest a real issue. The third is the bottom-line aggregation problem — running the rule only at final hire while ignoring the funnel — which Connecticut v. Teal (1982) explicitly rejected.

A fourth, more technical pitfall: applying the rule across roles that should be analyzed separately. Aggregating engineering, sales, and operations hires into one ratio masks role-specific disparities that, examined alone, would fail. Mature programs disaggregate by role, location, and time period.

AIEH Portable Credentials and Funnel Mechanics

The four-fifths rule applies at every selection step, including those that occur before formal application — early sourcing filters, recruiter pre-screens, and resume parsing. AIEH’s Skills Passport shifts some of that early-funnel filtering from credential proxies (years of experience, school prestige) to validated capability evidence. The mechanical effect is that selection-rate differentials at the earliest funnel step — historically a major contributor to bottom-line adverse impact — become tied to measured skill rather than demographic-correlated proxies. Programs adopting portable credentials should still apply the four-fifths rule (and significance tests) to every step, because the rule is a diagnostic that runs continuously regardless of upstream tool choice. For broader context on shifting from proxies to evidence, see skills vs credentials and hiring bias mitigation.

Takeaway

The four-fifths rule has earned its longevity by being computable, transparent, and reasonably useful at moderate sample sizes. Its limits — sensitivity at small N, insensitivity at large N, silence on practical magnitude — are well documented, and modern selection analytics layer significance testing and effect sizes on top. The right way to use the rule is as the first lens in a multi-test framework, applied at every funnel step, disaggregated by role and location, and paired with validity evidence when any flag is raised. Employers should resist both extremes: treating the rule as a hard threshold that ends inquiry on either side, and dismissing it as obsolete. It is what it has always been — a rule of thumb, useful in its place, dangerous when stretched. Application to specific employment decisions should be reviewed with qualified counsel.

Sources

• Uniform Guidelines on Employee Selection Procedures (1978). 29 C.F.R. Part 1607, Section 4(D). EEOC, DOL, DOJ, CSC.
• Cohen, D. B., Aamodt, M. G., & Dunleavy, E. M. (2010). Technical Advisory Committee Report on Best Practices in Adverse Impact Analyses. Center for Corporate Equality.
• Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2), 262–274.
• Sackett, P. R., & Lievens, F. (2008). Personnel selection. Annual Review of Psychology, 59, 419–450.
• Connecticut v. Teal, 457 U.S. 440 (1982).
• Watson v. Fort Worth Bank & Trust, 487 U.S. 977 (1988).