The Four-Fifths Rule, Explained With a Worked Example
The four-fifths rule is the standard first check for adverse impact in hiring, and the metric behind New York City’s bias-audit law. It is simple to calculate and easy to misread.
The rule
Under the U.S. Uniform Guidelines on Employee Selection Procedures, 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. It is a rule of thumb, not a legal verdict, but it is where regulators and plaintiffs start.
Step 1: selection rates
For each group, divide the number selected by the number who applied (or who were assessed by the tool).
Step 2: impact ratios
Divide each group’s selection rate by the highest rate, here Group A’s 30%.
Group C’s impact ratio of 0.67 is below 0.80, so the tool shows potential adverse impact against Group C.
Scores instead of pass/fail
If the tool outputs a score rather than a yes or no, New York City’s rules use a scoring rate: the share of each group scoring above the median score for the whole sample. You then compute impact ratios the same way.
Pitfalls
- Small samples. With few candidates, ratios swing wildly. Pair the ratio with a statistical-significance test before drawing conclusions.
- Aggregation hides problems. A tool can pass overall and fail for a specific job posting. Analyse by role and requisition as well as in total.
- Intersections. Sex and race together can show impact that neither shows alone. New York City requires intersectional categories.
- Proxies. Zip code, graduation year, employment gaps and accent can stand in for protected traits even when the tool never sees them.
- Population choice. A normalized test (equal groups, matched profiles) shows what the model does; a representative test (regional demographics) shows what real candidates will experience. You want both.
After a failed ratio
- Check statistical significance.
- Identify which features drive the gap.
- Check whether the criteria are job-related and whether a less discriminatory alternative exists.
- Fix, retrain or retire the tool, and document the decision.
Map your hiring AI against the laws that apply to you
Whether you build HR AI or deploy it, Fairly AI tests it, monitors the rules where you operate and keeps the evidence.
Schedule a discovery call