Adverse Impact Calculator

The Adverse Impact Calculator assesses selection rates, computes adverse impact ratios, and tests differences for statistical significance between demographic groups.

About the Adverse Impact Calculator

This tool evaluates differences in hiring, promotion, or other selection outcomes between defined groups. It focuses on two common standards: the four-fifths rule and statistical significance. You supply basic inputs, like applicants and selections per group, and the calculator returns an adverse impact ratio, p-values, and interval estimates.

The calculator supports multiple testing approaches. It uses exact tests for small samples and standard two-proportion methods for larger data. It also highlights assumptions, such as independence and binomial sampling, so you can judge fit. When assumptions are strained, the tool flags caution.

Use it during audit prep, quarterly monitoring, or after process changes. The goal is to connect results to action. If you identify a disparity, the next step is to review job-relatedness, validate assessments, and adjust processes where appropriate.

How the Adverse Impact Method Works

Adverse impact analysis compares selection rates between a reference group and one or more comparison groups. The basic idea is simple: compute each group’s selection rate, form a ratio, and judge the result against practical and statistical thresholds. The method respects sampling variation and the distribution of counts.

• Compute selection rate per group: selections divided by applicants at the same stage.
• Choose a reference group, typically with the highest selection rate or the historically favored group.
• Form the adverse impact ratio (AIR): comparison group rate divided by reference group rate.
• Check the four-fifths rule: an AIR below 0.80 signals potential adverse impact.
• Test statistical significance: compare two proportions using an exact test or z-test.
• Contextualize with confidence intervals to understand precision and uncertainty.

These steps balance practical and statistical views. A ratio below 0.80 is a red flag, but significance depends on sample size and variability. The calculator shows both, so you can interpret effects and risk in tandem.

Adverse Impact Formulas & Derivations

The core computations rely on simple proportions, with extensions for inference. Selection outcomes are modeled using a binomial distribution. For large samples, normal approximations are efficient; for small samples or sparse data, exact methods are more reliable.

• Selection rate: r = x / n, where x = selections and n = applicants for a group.
• Adverse impact ratio (AIR): AIR = r_comp / r_ref. Four-fifths rule flags AIR < 0.80.
• Difference in rates: Δ = r_ref − r_comp. Standard error (approx): SE(Δ) = sqrt[r_ref(1 − r_ref)/n_ref + r_comp(1 − r_comp)/n_comp].
• Two-proportion z-test (pooled): z = (r_ref − r_comp) / sqrt[p(1 − p)(1/n_ref + 1/n_comp)], where p = (x_ref + x_comp)/(n_ref + n_comp).
• Exact test option: Fisher’s exact or binomial exact test uses the hypergeometric/binomial distribution to compute a p-value without normal assumptions.
• Confidence interval for Δ (approx): Δ ± z* × SE(Δ). For the ratio, a log method can be used when needed to stabilize variance.

The calculator selects formulas based on your inputs and settings. It defaults to exact tests when small counts or extreme proportions appear. For larger samples, it uses the faster normal approximation with optional continuity correction.

Inputs, Assumptions & Parameters

Start with clear data and a well-defined comparison. You will define groups, provide applicant and selection counts, and select a reference group. Then choose a test method and a significance level that aligns with your policy or review standard.

• Applicants per group (n): the number considered at the specific stage.
• Selections per group (x): the number advanced, hired, or promoted at that stage.
• Reference group: the group against which other groups are compared.
• Significance level (alpha): the threshold for statistical tests (commonly 0.05).
• Method: exact test for small samples or two-proportion z-test for larger samples.
• Continuity correction: toggle on/off for the z-based confidence intervals and tests.

Watch for edge cases. Very small n, zero selections, or 100% selection can strain assumptions. The tool will switch to exact methods and suggest adding data or aggregating periods when possible. Treat results near thresholds with caution, and review the underlying process.

Using the Adverse Impact Calculator: A Walkthrough

Here’s a concise overview before we dive into the key points:

1. Define your analysis stage, such as screening, assessment, or final offer.
2. Select your groups and specify the reference group.
3. Enter applicants (n) and selections (x) for each group.
4. Choose a test method and set the significance level (alpha).
5. Review the selection rates, adverse impact ratios, p-values, and confidence intervals.
6. Adjust method options or groupings if samples are very small or imbalanced.

These points provide quick orientation—use them alongside the full explanations in this page.

Worked Examples

Example 1: A company screens 20 men and 10 women for a role. It advances 10 men and 3 women. Men’s selection rate is 10/20 = 50%. Women’s rate is 3/10 = 30%. The AIR is 0.30/0.50 = 0.60, which is below 0.80. A two-proportion test yields a non-significant result with this small n (the difference is not statistically conclusive). What this means: The four-fifths rule flags a potential issue, but the evidence is statistically weak; review process steps and gather more data.

Example 2: A larger pool includes 300 applicants in Group A and 100 in Group B. Hires are 90 for A and 15 for B. Group A’s rate is 90/300 = 30%. Group B’s rate is 15/100 = 15%. The AIR is 0.15/0.30 = 0.50, well below 0.80. With these sizes, a two-proportion test is significant (p around 0.003), suggesting a real disparity beyond sampling noise. What this means: Both practical and statistical indicators point to adverse impact; review job-relatedness, validation evidence, and decision steps right away.

Assumptions, Caveats & Edge Cases

Adverse impact analysis depends on statistical and process assumptions. The conclusions are strongest when data are complete, stages are well-defined, and groups are comparable. Use the findings to guide inquiry, not as a final legal determination.

• Independence: Each decision is assumed independent, with a stable process during the period analyzed.
• Binomial model: Selection counts are treated as binomial outcomes; rate estimates assume consistent criteria.
• Comparability: Compare groups at the same stage with the same criteria and opportunity to be selected.
• Small samples: Exact tests help, but precision is limited; results can swing with a few cases.
• Multiple comparisons: Testing many groups or stages inflates false positives; plan your analysis and document choices.

If the context is complex, consider stratifying by job family, location, or stage, or conduct a pooled analysis with care. Document your assumptions and any data exclusions. Use additional validation work to explain observed differences.

Units Reference

Adverse impact reporting mixes counts and proportions. Consistent units reduce confusion and make trends easier to spot. The table below lists common quantities and how the calculator reports them.

Use counts for input and proportions for interpretation. When CIs are wide, collect more data or aggregate across consistent periods. If AIR is near 0.80, examine both the CI and the p-value to judge stability.

Tips If Results Look Off

Unexpected outputs often trace to a data or settings mismatch. Review the basics first, then check assumptions. Use the list below to troubleshoot before drawing conclusions.

• Verify that applicants equal selected plus not selected for each group.
• Confirm the chosen reference group; it drives the AIR.
• Look for zeros or 100% selections; switch to exact tests when present.
• Ensure you are analyzing a single stage and time window.
• Check rounding; tiny differences can flip results near thresholds.

If issues persist, consider aggregating periods, stratifying by role, or reviewing how groups were coded. Document any decisions so future analyses stay consistent.

FAQ about Adverse Impact Calculator

What is the four-fifths rule and when does it apply?

The four-fifths rule flags potential adverse impact when a group’s selection rate is less than 80% of the reference group’s rate. It is a practical screen, not a legal conclusion.

Which statistical test does the calculator use?

It uses a two-proportion z-test for larger samples and switches to an exact method for small or sparse data. You can also select your preferred method in settings.

Can I analyze more than two groups?

Yes. Choose one reference group and compare each additional group to that reference. The tool reports AIR, p-values, and intervals for each comparison.

Does passing the 80% rule mean no risk?

No. Passing the 80% rule reduces concern but does not end review. Consider statistical significance, confidence intervals, job-relatedness, and validation evidence.

Adverse Impact Terms & Definitions

Adverse Impact

A substantially different rate of selection in hiring, promotion, or other employment decisions that works to the disadvantage of a protected group.

Selection Rate

The proportion selected at a stage, calculated as selections divided by applicants for a specific group.

Adverse Impact Ratio (AIR)

The comparison group’s selection rate divided by the reference group’s rate; values below 0.80 indicate potential adverse impact.

Reference Group

The group used as the benchmark in comparisons, often the group with the highest selection rate or historically favored group.

Four-Fifths Rule

A practical threshold stating that an AIR below 80% signals possible adverse impact and warrants further review.

Statistical Significance

A measure of whether observed differences are unlikely to be due to chance, based on a test and a chosen significance level.

Fisher’s Exact Test

An exact method for 2×2 tables that computes a p-value without relying on normal approximations; useful with small samples.

Confidence Interval

A range of values that likely contains the true effect (difference or ratio), reflecting uncertainty due to sampling.

Sources & Further Reading

Here’s a concise overview before we dive into the key points:

• Uniform Guidelines on Employee Selection Procedures (29 CFR Part 1607)
• EEOC Guidance: Employment Tests and Selection Procedures
• NIST Handbook: Two Proportions and Tests of Significance
• OFCCP FAQs: Understanding Adverse Impact
• SHRM: How to Conduct an Adverse Impact Analysis

These points provide quick orientation—use them alongside the full explanations in this page.