The short answer
Adverse impact is "a substantially different rate of selection in hiring, promotion, or other employment decision which works to the disadvantage of members of a race, sex, or ethnic group," as defined in the federal Uniform Guidelines on Employee Selection Procedures (29 CFR 1607.16). It matters because, under federal equal employment law, a selection procedure with adverse impact is discriminatory unless it has been properly validated or is otherwise justified.
Adverse impact in hiring explained
Adverse impact is about outcomes, not intent. A resume filter, a test or an interview can be applied the same way to everyone and still screen out one group at a much higher rate. Lawyers often call the legal theory "disparate impact"; "adverse impact" is the Guidelines' term for the measured gap.
The EEOC's Questions and Answers on the Guidelines describe a four-step check:
- Calculate the selection rate for each group: people selected divided by applicants from that group.
- Find the group with the highest rate.
- Calculate each group's impact ratio: its rate divided by the highest rate.
- See whether any ratio is substantially lower, usually below four-fifths (80%). If so, adverse impact is indicated in most circumstances.
The 80% line is the four-fifths rule, which the EEOC calls a rule of thumb, not a legal definition.
Example
The EEOC's own example uses hiring rates of 60% for white (non-Hispanic) applicants, 45% for American Indian applicants, 48% for Hispanic applicants and 51% for Black applicants:
On these numbers alone, adverse impact is indicated for American Indian applicants but not for the other groups. The EEOC notes that the calculation is not purely arithmetic and other factors can matter.
Common mistakes
- Checking only the end result. The EEOC says adverse impact is checked first for the overall selection process for each job, and then for the individual steps if the overall process shows it. Some situations call for checking steps even when the total looks fine.
- Treating 80% as a safe harbor. The Guidelines say smaller differences can still be adverse impact when they are significant in both statistical and practical terms.
- Trusting tiny samples. Large gaps based on small numbers may not be statistically significant. The Guidelines allow looking at a longer period or similar use elsewhere.
- Not keeping the data. If an employer has not kept the required records, federal agencies may infer adverse impact when a group is underused in the job.
AI tools are measured the same way. Our guide to AI bias audits shows how this math appears in a vendor's audit report.
Related terms
- Four-fifths rule: the 80% rule of thumb for flagging a substantial difference.
- Impact ratio: a group's selection rate divided by the highest group's rate.
- Selection rate: the share of a group's applicants who are selected.
- Bias audit: an independent test of a tool for this kind of disparity.
- Structured interview: a consistent, job-related format that is easier to validate.
Questions
Is adverse impact the same as discrimination?
Not automatically. Adverse impact is evidence that a selection procedure disadvantages a group. Under federal law, a procedure with adverse impact is discriminatory unless it has been properly validated or is otherwise justified, and employers may instead switch to a lawful alternative that removes the impact.
Does adverse impact require intent?
No. Adverse impact looks at the results of a selection procedure, so a test or screen applied identically to everyone can still have it. Intentional unequal treatment is a separate issue.
How often should employers check for adverse impact?
The Uniform Guidelines call for annual adverse impact determinations for groups that make up at least 2% of the relevant labor force or workforce. Many teams also check after changing a test, a tool or a rubric. In New York City, AI tools covered by Local Law 144 need a bias audit at least once a year.
Curious how an AI interview reports scores so you can run this math on your own pipeline? Book a 20-minute demo, or try a sample interview yourself.
This is not legal advice.
