The fair-lending layer

Protected-class A/E & disparate-impact testing

A model can be blind to a protected attribute and still produce disparate impact through correlated features. The actuarial answer is not to inspect the model but to cut its realized error stream by the protected class and test the outcome. This page runs three escalating tests on the same synthetic book the tool uses — the four-fifths / adverse-impact ratio, a significance test, and the actuarial contribution: a decomposition of the disparity into the part explained by a legitimate rating factor and the unexplained residual.

The protected-class labels here are a synthetic stand-in for any ECOA / Title VII basis (race, ethnicity, sex, age ≥ 40, …). The model never uses them; they only reveal how the already-drawn errors fall across groups. The math is identical whatever the real basis.

Flip to the control basis to see a clean pass.

Adverse-outcome rate by group

Where does the disparity live?

Adverse rate, disadvantaged vs advantaged group, by cohort.

Explained vs. residual — the legitimate-actuarial-basis test

Per-cohort A/E cut

Within each cohort, A/E ≈ 1.0 means the group is treated as validated. A residual A/E gap inside a cohort is disparity the cohort factor cannot explain.

Governance reading. A raw disparity that is fully explained by a legitimate rating factor (residual ratio ≈ 1) may have a defensible actuarial basis — but the factor itself must still be justified and the outcome monitored. A residual disparity that survives after controlling for the rated factor has no legitimate actuarial basis on this evidence and is the governance red flag under ECOA and Colorado SB21-169. This analysis quantifies impact; it does not by itself establish legal liability, which is a fact-specific legal determination.