What Is Adverse Impact? The Four-Fifths Rule and How It Applies to Job Advertising and AI Screening

Adverse impact is a substantially different rate of selection in hiring or other employment decisions that works to the disadvantage of a group defined by race, sex, or ethnic group. 

Under the federal Uniform Guidelines on Employee Selection Procedures (UGESP), the four-fifths (80%) rule is a rule of thumb: if a group’s selection rate is less than four-fifths (80%) of the rate for the group with the highest rate, enforcement agencies will generally regard that as evidence of adverse impact (29 CFR 1607.4(D)). It is a screening signal, not a legal verdict, and it can be applied to any step where people are selected or advanced, including AI screening and the applicant flow that job advertising produces.

What Is Adverse Impact?

Adverse impact (also called disparate impact) describes a neutral-looking practice that produces unequal results for a protected group. UGESP defines it as a substantially different rate of selection in hiring, promotion, or other employment decisions that works to the disadvantage of members of a race, sex, or ethnic group. The practice does not need to be intentional. What matters is the outcome and whether the practice is job-related and consistent with business necessity.

For talent acquisition teams, “selection” covers every gate: who applies, who passes a screen, who gets an interview, who gets an offer. Joveo’s guide to responsible AI in hiring covers how EEOC expectations and bias audits fit together.

What Is the Four-Fifths Rule?

The four-fifths rule compares selection rates between groups. Divide each group’s selection rate by the highest group’s rate. If the result (the impact ratio) is below 0.80, the agencies will generally treat the difference as evidence of adverse impact. The EEOC’s UGESP Questions and Answers describe it as a rule of thumb “not intended as a legal definition,” and note that it “merely establishes a numerical basis for drawing an initial inference.”

How to calculate it: a worked example

Suppose 450 people pass through an initial AI resume screen, and the screen advances some of them to a recruiter:

Group B’s impact ratio is 24% / 40% = 0.60, below 0.80, so the screen would be flagged for review. Group C’s ratio of 0.90 is above the threshold. A flag does not prove discrimination. It tells you to look at whether the step is job-related, whether a less discriminatory alternative exists, and whether the numbers are reliable.

What Are the Limits of the Four-Fifths Rule?

The rule is useful because it is simple, but the regulation and EEOC guidance are explicit that it is not the whole test.

• Small samples. UGESP says a gap larger than 20% may not indicate adverse impact when it rests on small numbers or lacks statistical significance. The EEOC Q&A gives an example where the rule is failed but the number of selections is too small to warrant a determination, because changing one hire would reverse which group has the higher rate.
• Smaller gaps can still matter. The agencies state that differences under 20% can still be adverse impact if they are significant in both statistical and practical terms. Passing the 80% test is not a safe harbor.
• Statistical significance. With large numbers, even small differences can be statistically significant, so significance testing is used alongside the rule. The Q&A notes the Guidelines do not rely primarily on a significance test, partly because small samples rarely reach the conventional .05 level.
• Unreliable data. When data is too thin, agencies may consider impact over a longer period or evidence from similar uses elsewhere.

In short, treat the 80% ratio as a trigger for deeper analysis, with significance testing, longer time windows, and legal review where needed.

How Does Adverse Impact Apply to AI Screening?

Any AI step that scores, ranks, or filters candidates produces selection rates, so the same arithmetic applies. Employers generally need to compute impact ratios for each gate where an algorithm influences who moves forward, keep the records, and revisit them as models and applicant pools change. Knockout rules deserve the same scrutiny, since a single required question can drive large differences in who advances (see Joveo’s explainer on knockout questions), and so do automated interviews (see how AI interviewers work, and whether they are fair).

How Does Adverse Impact Apply to Job Advertising?

Skew can start before anyone applies. Published research shows that ad platforms can deliver job ads unevenly even when the advertiser targets broadly:

• Ali et al. (2019), “Discrimination through optimization,” found that Facebook’s ad delivery can produce skewed outcomes by gender and race for real employment and housing ads, driven by the platform’s optimization and its predictions of ad relevance rather than the advertiser’s targeting choices.
• Imana, Korolova, and Heidemann (2021), “Auditing for Discrimination in Algorithms Delivering Job Ads,” built a method to separate skew caused by qualifications from skew caused by the platform. They found gender skew on Facebook that qualifications did not explain, and found no skew on LinkedIn.

On the policy side, Meta announced in March 2019 that advertisers running housing, employment, and credit ads could no longer target by age, gender, or ZIP code, and would have a much smaller set of targeting options. Meta’s current Special Ad Category requirements for employment ads should be checked directly in its advertiser documentation. In June 2022, the U.S. Department of Justice announced a settlement with Meta over housing ad targeting and delivery, its first case challenging algorithmic bias under the Fair Housing Act, which required Meta to redesign its delivery system for those ads.

The practical lesson for advertisers: you control targeting, but the platform controls delivery. That is why monitoring the applicant flow your campaigns produce matters. UGESP is aimed at selection procedures, so ad delivery is not itself a “selection rate” under the rule. But comparing who applies, by channel and by role, is a reasonable early-warning check, and recruiting practices that discourage certain groups can be relevant under UGESP. Pair it with a review of your job posting compliance checklist so that EEO language, pay transparency, and platform rules are covered together.

What Should Advertisers and Recruiters Monitor?

1. Applicant flow by source. Where voluntary demographic data is lawfully collected, compare who applies from each channel and role family.
2. Impact ratios at each gate. Calculate screening, interview, and offer rates by group, and keep dated records.
3. Significance and sample size. Do not act on, or dismiss, a ratio without checking numbers and time windows.
4. Platform changes. Ad delivery systems and AI models change, so rerun checks after changes.
5. Human review. Keep people accountable for decisions that affect who advances. Platforms built for agentic recruiting, such as Joveo, pair automation with reporting that helps teams see performance across channels, which supports this kind of oversight.

Frequently Asked Questions

What is adverse impact in hiring?

A substantially different rate of selection in hiring, promotion, or other employment decisions that works to the disadvantage of a group defined by race, sex, or ethnic group, whether or not it was intended.

What is the four-fifths rule?

A rule of thumb in UGESP: a selection rate for any group that is less than four-fifths (80%) of the rate for the group with the highest rate is generally regarded as evidence of adverse impact.

Is passing the 80% rule a guarantee of compliance?

No. EEOC guidance says smaller differences can still be adverse impact if they are statistically and practically significant, and the rule is not controlling in all circumstances.

Does the four-fifths rule work with small samples?

Not reliably. UGESP says larger gaps may not indicate adverse impact when based on small numbers or lacking statistical significance, and the EEOC Q&A says agencies will not assume adverse impact when chance could explain the gap, absent other evidence.

Can job ads create adverse impact?

Research by Ali et al. (2019) and Imana et al. (2021) found skewed delivery of job ads on Facebook. Monitoring applicant flow by channel helps advertisers spot patterns worth investigating.