REGULATORY GUIDE
EEOC Uniform Guidelines on Employee Selection Procedures
The federal framework for evaluating employment selection procedures—and the origin of the four-fifths rule.
Published
1978
Jurisdiction
Federal (Nationwide)
Agencies
EEOC, DOL, DOJ, CSC
Citation
29 CFR Part 1607
Overview
The Uniform Guidelines on Employee Selection Procedures were jointly adopted in 1978 by four federal agencies: the Equal Employment Opportunity Commission (EEOC), Department of Labor, Department of Justice, and Civil Service Commission. They provide a framework for determining whether employment selection procedures—including tests, interviews, and other assessments—have an adverse impact on protected groups.
Though written decades before AI hiring tools existed, the Guidelines remain the authoritative federal framework for evaluating whether any selection procedure, including AEDTs, discriminates against protected classes. The Guidelines introduced the four-fifths (80%) rule, which is now the cornerstone of bias auditing under NYC LL144 and other state laws.
The Four-Fifths Rule
Section 4D of the Uniform Guidelines establishes the four-fifths rule as a "rule of thumb" for determining adverse impact:
"A selection rate for any race, sex, or ethnic group which is less than four-fifths (4/5) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact."
— 29 CFR § 1607.4(D)
How the Four-Fifths Rule Works
The calculation is straightforward:
- Calculate the selection rate for each demographic group (number selected ÷ number of applicants)
- Identify the group with the highest selection rate
- Calculate the impact ratio for each other group (that group's selection rate ÷ highest group's selection rate)
- If any impact ratio is less than 0.80 (80%), there is evidence of adverse impact
Example Calculation
| Group | Applicants | Selected | Selection Rate | Impact Ratio |
|---|---|---|---|---|
| White | 100 | 30 | 30% | 1.00 |
| Asian | 50 | 14 | 28% | 0.93 |
| Hispanic | 40 | 10 | 25% | 0.83 |
| Black | 60 | 12 | 20% | 0.67 |
In this example, Black applicants have an impact ratio of 0.67 (20% ÷ 30%), which is below the 0.80 threshold, indicating potential adverse impact.
Important Nuances
The Four-Fifths Rule Is Not a Safe Harbor
An impact ratio above 0.80 does not guarantee a procedure is non-discriminatory. The Guidelines state that the four-fifths rule is a "rule of thumb" and that smaller differences may still constitute adverse impact if statistically significant. Conversely, a ratio below 0.80 may not indicate discrimination if the difference is not statistically significant due to small sample sizes.
When Adverse Impact Is Found
Under the Guidelines, if a selection procedure shows adverse impact, the employer has three options:
- Validate the procedure: Demonstrate through a validity study that the procedure is job-related and consistent with business necessity
- Modify the procedure: Adjust the selection procedure to eliminate or reduce the adverse impact
- Adopt an alternative: Use a different procedure with less adverse impact that still serves the employer's legitimate needs
Relationship to Modern AI Laws
NYC Local Law 144 explicitly incorporates the four-fifths rule from the Uniform Guidelines as its benchmark for adverse impact. The Colorado AI Act and other emerging state laws also reference the Guidelines as the standard for evaluating algorithmic discrimination.
How Modern Laws Build on the Guidelines
- NYC LL144: Requires annual bias audits using selection rate analysis and the four-fifths rule
- Colorado AI Act: Requires impact assessments that analyze algorithmic discrimination risk
- EEOC AI Guidance (2023): Confirms that the Uniform Guidelines apply to AI and algorithmic hiring tools
How Paritas Implements the Guidelines
Every Paritas audit is built on the statistical framework established by the Uniform Guidelines:
- Selection Rate Analysis: We calculate selection rates for all demographic categories (sex, race/ethnicity, intersectional)
- Impact Ratio Calculation: We compute impact ratios using the highest-selected group as the reference
- Three-Tier Classification: We classify results as PASS (≥0.90), MONITOR (0.80-0.89), or FLAG (<0.80)
- Statistical Significance Testing: We supplement the four-fifths rule with Fisher's exact test, z-tests, and confidence intervals to account for sample size
- Practical Significance: We assess whether statistically significant differences are also practically meaningful
Ensure your AEDT is compliant.
Get a bias audit grounded in the federal framework that's been the standard for employment discrimination analysis since 1978.