Race-Neutral by Construction: A Transparent, Gaming-Resistant Mechanism for Disadvantage-Based College Admissions in the United States

In Students for Fair Admissions v. Harvard (2023) the U.S. Supreme Court ended race-conscious college admissions, intensifying interest in race-neutral alternatives that select for disadvantage. The leading alternative — class-based admissions — is well argued in the policy literature but has repeatedly foundered on operationalisation : the College Board’s opaque “adversity score” collapsed a person into a single number and was withdrawn, while naive additive indices are trivially gameable by advantaged applicants. We propose a transparent, closed-form admissions mechanism that (i) organises disadvantage factors by verifiability into three tiers, (ii) replaces raw achievement with a cross-fitted context-adjusted residual (“distance travelled”) that is ability-informative but environment-neutral, (iii) combines these through a bounded, published index, and (iv) resists manipulation by anchoring the score on exogenous and verified data with a per-tier cap on self-attested factors. We prove that the manipulation gain any applicant can extract is bounded by a single published constant, and confirm the bound empirically. On a synthetic population the mechanism admits a class of higher latent ability than test-score ranking while roughly quadrupling the share of admits from disadvantaged (bottom-half) environments, and cuts the displacement of genuinely disadvantaged applicants by an adversary by ~ 85% relative to a naive index. On real data from the Stanford Education Data Archive (about 12,400 districts) we show that socioeconomic context explains 74% of achievement variance (grounding the approach), that the residual is environment-neutral (correlation with socioeconomic status falls from 0.77 to 0.02), and — decomposing the race–achievement gap — that context captures 80–89% of it while a real remainder persists, quantifying both the promise and the honest limit of class-based admissions. We discuss governance, contestability, and the ethical boundaries of adversity measurement.