Computes fair lending disparity metrics from aggregate approval counts. Applies the 4/5ths adverse-impact rule (29 CFR §1607.4(D)), two-proportion z-statistic (one-tail and two-tail at α = 0.05), odds ratio, and standardized mean difference. Takes aggregate group counts only — zero PII by construction. No individual applicant records are processed. Table: EEOC-UGSP-29CFR1607-2024.
29 CFR §1607.4(D)4/5ths RuleZero PIIAdverse ImpactZ-StatisticOdds Ratio
🔒 ZERO PII BY CONSTRUCTION. Inputs are aggregate counts only — no individual applicant records, names, demographic identifiers, or scores. All computation runs locally in your browser. No data is transmitted.
ⓘ Not legal or compliance advice. Disparity metrics are screening tools; statistical significance does not establish causation or legal liability. Consult legal counsel and a qualified statistician before use in supervisory or litigation contexts.
Group Labels
Labels are used for output readability only. They do not affect computation. Use generic labels such as "protected_class" and "control_group" — do not enter actual demographic categories here.
Group B is the reference (higher-approval) group for the 4/5ths ratio denominator.
Aggregate Approval Counts
Enter aggregate counts from your HMDA LAR, fair-lending statistical sample, or examination dataset. These are cohort totals, not individual records.
Count of approvals (or originations) in the protected-class cohort.
Total applications received from the protected-class cohort.
Count of approvals in the reference cohort.
Total applications received from the reference cohort.