Skill Guide

Statistical testing for adverse impact - the four-fifths rule, chi-squared tests, and regression-based impact analysis

Statistical testing for adverse impact is a set of quantitative methods used to determine whether a personnel selection procedure (e.g., hiring, promotion) has a significantly different impact on members of a protected class, using the four-fifths rule as a practical guideline and chi-squared and regression analyses as formal inferential tests.

This skill is highly valued as it ensures compliance with equal employment opportunity (EEO) laws, mitigates legal risk, and validates the fairness of talent processes, directly protecting the organization from costly litigation and reputational damage while building a defensible, meritocratic culture.

1 Careers

1 Categories

9.0 Avg Demand

15% Avg AI Risk

How to Learn Statistical testing for adverse impact - the four-fifths rule, chi-squared tests, and regression-based impact analysis

Focus on: 1) Defining and calculating the selection rate and impact ratio for the four-fifths rule. 2) Understanding the basic principles of a chi-squared (χ²) test of independence and when to apply it. 3) Grasping the concept of a binary logistic regression model with a protected class indicator as a predictor.

Move to practice by: 1) Applying these methods to real HR data from a single hiring cycle, correctly identifying the comparison groups and interpreting p-values. 2) Learning common pitfalls like small sample sizes and base rate issues. 3) Practicing the interpretation of results for a non-technical legal or HR audience.

Master the skill by: 1) Designing and implementing ongoing adverse impact monitoring systems integrated into HRIS. 2) Conducting multi-variate regression analyses controlling for legitimate, job-related predictors (e.g., experience, test scores) to isolate the impact of group membership. 3) Advising leadership on the trade-offs between statistical significance and practical significance, and developing remediation strategies when adverse impact is found.

Practice Projects

Beginner

Project

Adverse Impact Analysis for a Single Job Requisition

Scenario

You are provided with a spreadsheet containing applicant data for an Analyst role: applicant ID, gender (M/F), and outcome (Hired/Not Hired). There were 200 male applicants (20 hired) and 100 female applicants (10 hired).

How to Execute

1. Calculate the selection rate for each group: Males: 20/200 = 10%; Females: 10/100 = 10%. 2. Calculate the impact ratio: lower rate / higher rate = 10%/10% = 1.0. 3. Apply the four-fifths rule: since 1.0 ≥ 0.8, there is no indication of adverse impact. 4. Document your calculation and conclusion in a 1-page memo.

Intermediate

Project

Chi-Squared Analysis of Promotion Decisions

Scenario

You have promotion outcome data for three racial/ethnic groups (White, Hispanic, Asian) across the company. The numbers are: White: Promoted=150, Not Promoted=850; Hispanic: Promoted=40, Not Promoted=160; Asian: Promoted=30, Not Promoted=170. Is there a statistically significant association between race/ethnicity and promotion outcome?

How to Execute

1. Formulate the null hypothesis (H0): There is no association between race/ethnicity and promotion outcome. 2. Use a software tool (e.g., Excel, Python's SciPy) to perform a chi-squared test of independence on the contingency table. 3. Calculate the chi-squared statistic and corresponding p-value. 4. Interpret the results: if p < 0.05, reject H0 and conclude there is a statistically significant association, warranting further investigation into the promotion process.

Advanced

Project

Regression-Based Impact Analysis with Covariates

Scenario

A high-volume hiring process for customer service representatives shows potential adverse impact against a protected group based on the four-fifths rule. You need to determine if this impact persists after controlling for legitimate, job-relevant predictors: a pre-employment assessment score and years of relevant experience.

How to Execute

1. Build a binary logistic regression model with the hiring decision (Hired=1, Not Hired=0) as the dependent variable. 2. Include the protected class indicator (e.g., Group=1, Other=0) as a primary independent variable of interest. 3. Include the assessment score and years of experience as control covariates. 4. Examine the coefficient and statistical significance of the protected class variable. If it remains significant after controlling for the covariates, it suggests the selection procedure may have an impact beyond what is explained by job-relevant factors.

Tools & Frameworks

Statistical Software & Languages

Python (pandas, SciPy, statsmodels)RAdvanced Excel / Google Sheets (with Analysis ToolPak)SPSS/SAS

Used for data manipulation, calculating selection rates, running chi-squared tests, and executing logistic regression models. Python and R are preferred for scalable, automated analyses.

HRIS & Analytics Platforms

Workday People AnalyticsOracle HCM AnalyticsVisierTableau

Enterprise platforms for extracting applicant flow data, tracking hiring metrics, and often built-in or easily integrable adverse impact monitoring dashboards.

Regulatory & Methodological Frameworks

Uniform Guidelines on Employee Selection Procedures (1978)EEOC Compliance ManualSIOP Principles for the Validation and Use of Personnel Selection Procedures

Provide the legal and professional standards defining adverse impact, acceptable methodologies, and best practices for demonstrating job-relatedness and business necessity.

Interview Questions

Answer Strategy

Test understanding of practical vs. statistical significance and sample size effects. Strategy: Explain that the four-fifths rule is a practical guideline, while the chi-squared test assesses statistical significance, which is influenced by sample size. A p-value of 0.12 is not statistically significant at the 0.05 level, suggesting the observed difference could be due to chance, especially with a small sample. Next steps include: 1) Examining the sample size and power analysis. 2) Documenting the analysis for the compliance file. 3) Considering a review of the selection procedures for potential improvements regardless of statistical significance, as the practical ratio is concerning.

Answer Strategy

Tests advanced analytical and communication skills. The core competency is translating technical analysis into business context. Sample Response: 'I would first frame the analysis as a risk management and quality control exercise. I'd build a model predicting promotion outcomes, including demographic indicators for protected groups and controlling for legitimate predictors like performance ratings, tenure, and skills assessment scores. I would present the findings not as a verdict of discrimination, but as an analysis of variance: showing that, for example, 90% of the promotion decisions are explained by the legitimate factors, and examining whether any remaining variance is systematically related to protected group status. I'd visualize this to make it intuitive, and then recommend a review of the specific decision points where the controlled-for factors might not fully account for the observed patterns.'