The Four-Fifths Rule: How Regulators Flag a Hiring Test for Bias

Two candidate pools can pass a hiring test at rates eight percentage points apart and be entirely fine under federal guidelines. Two other pools can pass at rates eleven points apart and fail the same screen. The difference is not the size of the gap in points. It is a ratio, and it is worth working out by hand at least once, because the arithmetic is the whole rule.

Where the rule comes from

The four-fifths rule — also called the 80 percent rule — was set out in the 1978 Uniform Guidelines on Employee Selection Procedures, adopted jointly by the Equal Employment Opportunity Commission, the Civil Service Commission, the Department of Labor and the Department of Justice. It built on an earlier 1972 California state guideline and a 1971 technical advisory panel's recommendation. The purpose was practical: give employers and regulators a simple, consistent first screen for whether a selection procedure — a test, an interview cutoff, a resume screen, or anything else used to decide who advances — produces an adverse impact on a group defined by race, sex or national origin, without requiring a full statistical workup for every single hiring decision.

The arithmetic, worked through

Take 200 people sitting a cognitive ability test as part of a hiring process, in two equally sized groups of 100. In Group A, 50 pass. In Group B, 42 pass. Group B's pass rate is 42%; Group A's is 50%. Divide the lower rate by the higher one: 42 ÷ 50 = 0.84, or 84%. Since 84% is above the 80% line, this result does not trigger the four-fifths rule — despite an eight-percentage-point gap that looks, at a glance, fairly wide.

Now change one number. Group B passes 39 instead of 42. The point gap widens only slightly, from eight points to eleven. But 39 ÷ 50 = 0.78, or 78% — below 80%, which does trigger the rule. An eleven-point gap fails where an eight-point gap passed, because the rule was never comparing points to begin with. It compares what fraction one group's success rate is of the other's, and that ratio is most sensitive to change right around a 50% baseline pass rate.

A flag, not a verdict

Clearing or tripping the four-fifths ratio is not itself a legal finding. Under the disparate-impact framework built up in U.S. employment law since the Supreme Court's 1971 decision in Griggs v. Duke Power Co., an employer whose test trips the four-fifths rule can still defend it by showing the test is a valid, job-related predictor of performance and that no less-discriminatory alternative was practical for the role. A test that clears the ratio is not automatically safe either — the guidelines allow other statistical evidence to establish adverse impact even when the simple four-fifths screen does not flag it. Our explainer on whether IQ-style tests are legal to use in hiring works through that fuller legal picture; this piece is about the one arithmetic step inside it that is easiest to check yourself.

Why cognitive ability tests draw particular scrutiny

General cognitive ability tests are, across a wide body of research, among the strongest available predictors of job performance across many roles — a point our piece on what a cognitive test score at work actually predicts covers in more depth. The same body of research also finds these tests tend to produce some of the largest average score differences between demographic groups among common selection methods, considerably larger than, for instance, a structured interview. That combination — strong predictive validity alongside a larger average group gap — is exactly why cognitive ability tests sit squarely in the path of four-fifths analysis more often than many other selection tools, and why organizations that use one seriously invest in validating it for the specific role rather than treating it as a generic screen.

What validity generalization changes about the argument

It is worth being precise about what "strong predictor" means here, because it is doing real work in the argument. Meta-analyses pooling results across many jobs and many organizations consistently find general cognitive ability tests among the most valid single predictors of job performance available, more so for complex jobs than simple ones but positive across the range. That evidence is what makes the four-fifths rule genuinely load-bearing law rather than a paperwork exercise: it forces a real tradeoff, between a measure that predicts performance about as well as anything in the selection toolkit and a measure that also tends to produce a larger average score gap between some groups than most alternatives do. Neither half of that tradeoff is in serious dispute in the research; what organizations do about it is where the four-fifths rule, and the legal framework around it, actually bites.

Reducing adverse impact without abandoning validity

The four-fifths rule is often framed as a choice between a valid test and an equitable one, and personnel-selection research generally finds that framing too stark. Combining a cognitive ability test with other predictors — a structured interview, a job-knowledge test, a work-sample exercise — typically produces a smaller average group difference than the cognitive test scored alone, while preserving most of the overall predictive validity, because the added measures do not carry the same pattern of group differences and still track job performance in their own right. That is the practical lever most organizations actually pull: not discarding a validated test, and not ignoring adverse impact, but building a selection process out of more than one ingredient.

A newer version of the same math: auditing hiring algorithms

The same selection-rate-ratio logic has recently been written directly into law for automated hiring tools. New York City's Local Law 144, passed in 2021 with enforcement beginning July 2023, requires that an automated employment decision tool be independently audited for bias before an employer can use it, with the audit results published and candidates notified in advance. The audits performed under the law calculate an impact ratio for each group — the same selection-rate-ratio concept the four-fifths rule uses, applied to whatever the algorithm is doing rather than to a single test score. Our piece on the EU AI Act's deadline for hiring assessments covers a parallel regulatory approach on the other side of the Atlantic — different law, same underlying question of whether a selection tool treats groups proportionately.

The same logic behind a very different kind of cutoff

Every pass or fail line on a hiring test is, underneath the label, a percentile cutoff drawn on somebody's score distribution — the identical logic behind the percentile cutoffs this site covers for gifted-program eligibility or Mensa admission, just applied to an employment decision instead of a school placement or a society membership. Our percentile calculator runs that same computation, without a hiring decision riding on the answer. The arithmetic of where a cutoff falls does not change depending on what it is being used to decide. What changes is the weight riding on it.

The next time a hiring test's pass rates get compared across two groups, the honest first question is not whether the point gap looks big. It's the ratio — and now you know how to check it in under a minute.

Common questions