AI Resume Screening and the Four-Fifths Rule Explained

A man typing on a laptop at a desk

Photo: Olu Eletu, CC0 (public domain), via Wikimedia Commons.

Your AI resume screener advanced 40 of 100 male applicants and 24 of 100 female applicants. Nobody told it to prefer anyone. Is that a problem? Regulators have used a one-line arithmetic check since the late 1970s to decide whether a gap like that deserves a closer look. It is called the four-fifths rule, and it is simpler, and more limited, than most people assume.

The four-fifths (80%) rule compares selection rates. Divide each group's selection rate by the rate of the most-selected group. If the result is below 4/5 (80%), the EEOC's Uniform Guidelines treat it as generally evidence of adverse impact. It is a rule of thumb, not a legal definition: small samples can mislead, and smaller gaps can still be unlawful when they are statistically and practically significant.

By the Decision Makers Hub editorial team. Last updated November 2026. This article is general information, not legal advice. Have employment counsel review how adverse impact analysis applies to your tools and workforce.

Key takeaways

  • Selection rate is the number selected divided by the number of applicants in each group.
  • The impact ratio is a group's selection rate divided by the highest group's selection rate. Below 0.80 is the red flag.
  • The EEOC says the 80% figure is not a legal definition but a practical means of focusing attention on serious discrepancies.
  • Small numbers can trigger the rule without proving much, and larger samples can show real impact even when the ratio passes.
  • Run the check at every stage the AI touches: screening, assessment, interview and offer.

What we keep hearing. "Our tool is bias-free," usually followed by a nod and a signed order form.

Our take. "Bias-free" is a label, not a test. Ask for selection rates by group, the date they were measured and who measured them. Then run the four-fifths check on your own applicants.

What does the four-fifths rule actually say?

The federal Uniform Guidelines on Employee Selection Procedures (29 CFR 1607.4D) state that a selection rate for any group that is less than four-fifths (or eighty percent) of the rate for the group with the highest rate will generally be regarded as evidence of adverse impact, while a higher rate generally will not. The same section qualifies the threshold in both directions: smaller differences may still constitute adverse impact if significant in both statistical and practical terms, and greater differences may not if they rest on small numbers and are not statistically significant. You can read the text on Cornell's Legal Information Institute.

How do you calculate it?

The EEOC's Uniform Guidelines Q&A (question 12) gives four steps: divide the number selected from each group by that group's number of applicants; identify the group with the highest rate; divide each group's rate by that highest rate to get an impact ratio; then check whether any ratio falls well below 4/5.

ExampleApplicantsSelectedSelection rateImpact ratio
EEOC example: White applicants804860%1.00 (highest)
EEOC example: Black applicants401230%0.50 (below 0.80)
Our illustration: men1004040%1.00 (highest)
Our illustration: women1002424%0.60 (below 0.80)

The EEOC's own Q&A also shows a case where the rule flags only one group out of several: with rates of 60%, 45%, 48% and 51%, the ratios are 75%, 80% and 85% against the highest, so only the 45% group falls under four-fifths. A spreadsheet with two columns (applicants and selected) and one formula (group rate divided by the top rate) is enough to run the check.

Why is it only a rule of thumb?

The EEOC is direct about the limits. Question 11 says the 80% figure "is not intended as a legal definition, but is a practical means" of drawing attention to serious discrepancies. Question 19 says the rule merely establishes a numerical basis for an initial inference, and does not mean differences under 20% are tolerated. Three cautions follow.

  • Small numbers. The EEOC gives an example where 3 of 20 applicants versus 1 of 10 fails the four-fifths test but is too small to warrant a determination of adverse impact. If changing one selection would reverse which group has the higher rate, the agencies generally will not assume adverse impact (question 21), although a sustained pattern over time would change that.
  • Statistical significance. The EEOC says significance testing is usually unnecessary except with large numbers of selections (question 22). With very large applicant pools, a gap that passes four-fifths can still be statistically and practically significant.
  • Group size. Question 16 says determinations should be made for each group that is 2% or more of the labor force in the relevant labor area.

How does it apply to AI resume screening?

An AI screener can touch thousands of applicants at several stages, so a gap that would be a rounding error in one requisition becomes a large pattern at scale. Our editorial advice: calculate impact ratios separately for each stage the tool influences (resume screen, skills assessment, interview invitation, offer), because a pass at one stage can hide a fail at another. Keep the vendor's numbers separate from your own, since your applicant mix is what matters.

New York City goes a step further for hiring tools used in the city. Its Local Law 144 requires an independent bias audit that calculates and publishes impact ratios, and we walk through it in how to run a bias audit under NYC Local Law 144. The employer, not only the vendor, is responsible for how a tool is used. A May 2025 ruling in Mobley v. Workday shows how seriously courts are treating these claims: on May 16, 2025, a federal judge in California granted preliminary certification of a nationwide age-discrimination collective action over Workday's AI screening tools, per a Holland & Knight summary. Those are allegations, not findings, and Workday says the suit lacks merit.

What should you do when the ratio falls below 0.80?

  1. Check the numbers. Confirm the counts, the stage, the time window and whether the sample is large enough to matter.
  2. Look at the job-relatedness of the criteria. Ask what the tool actually scores and whether it relates to the job. Moving to skills-based hiring often means swapping proxy criteria for evidence of the work itself.
  3. Keep a human reviewer who can disagree. A reviewer who accepts every ranking adds no protection. See human in the loop AI for how to design real oversight.
  4. Talk to the vendor and counsel. Ask for the validation evidence, then involve counsel before changing thresholds, because adjustments can create their own legal issues.
  5. Document and repeat. Write down what you found and changed, then run the check on a schedule. It belongs inside your AI governance framework, and the same discipline powers a pay equity audit.

Frequently asked questions

What is the four-fifths rule in hiring?

It is a rule of thumb under the federal Uniform Guidelines: a selection rate below four-fifths (80%) of the highest group's rate is generally regarded as evidence of adverse impact.

Is failing the four-fifths rule illegal?

Not automatically. The EEOC says the 80% figure is not a legal definition but a practical means of focusing attention. Small samples can fail the rule without proving much, and smaller gaps can still be unlawful when they are statistically and practically significant.

How do I calculate an impact ratio?

Divide the number selected by the number of applicants for each group, find the highest group rate, then divide each group's rate by it. A result below 0.80 is the signal.

Does the rule apply to AI screening tools?

Adverse impact analysis applies to selection procedures, including software and algorithms. Calculate impact ratios at each stage the tool influences, and check local rules such as NYC Local Law 144.

Is the vendor responsible for the results?

Do not assume so. The employer decides how and where to use a tool and carries its own legal duties. Ask vendors for validation evidence, and have counsel review your obligations.

Sources

Comments

Popular Posts

Why Workday Onboarding Breaks Down for Frontline Employees

Interview Tips for Beginners: A Step-by-Step Prep Plan

Are Biometrics Better Than Passwords? What NIST Says

How to Improve the Customer Experience (CX)

How to Market Your Handyman Services to Small Business Owners

Are Mobile Apps Tracking Your Every Move? How to Check

Predictive Scheduling Laws: Where They Apply and How to Comply

End-of-Day Routines That Boost Productivity at Work

Can I Learn Epic EHR Online? Training Paths Explained

How AI Assistants Enhance Decision-Making in HCM Integration