The Supreme Court banned race-conscious admissions in June 2023. Two admissions cycles later, the data from the country's most selective colleges show shifts that are larger, more persistent, and more uneven than early models predicted.
Five schools sit at the top of every admissions list: Harvard, Yale, Princeton, Stanford, MIT. Together they enroll roughly 7,800 freshmen a year. Their pre-SFFA Black share averaged near 12%. Two cycles after the ruling, the average has fallen by a third.
The largest single drop belongs to MIT — from 15% Black to 5% in Year 1, and 6% in Year 2. Princeton's collapse arrived a year late: Black enrollment slid from 9% to 5% only in the Class of 2029, the lowest at the school since 1968.
Most early commentary assumed the second cycle would soften the shock. It didn't.
The Class of 2029 was supposed to be a regression to the mean — the first full cycle where every admissions officer worked under the new rules. Instead, three patterns emerged.
At Harvard, Stanford, and Yale, declines continued. At Princeton, the cliff was delayed: Year 1 looked nearly normal, Year 2 cratered. Only at Cornell, Amherst, and (per polling) Brown did Year 2 hint at partial recovery.
Yale's case is instructive. Year 1 looked like a non-event — 14% Black, 19% Hispanic, both within prior bounds. Year 2 rewrote the picture: Hispanic enrollment fell to 13%, a 28% drop from baseline. Year 1 had reflected applications and yield; Year 2 was the first year admissions policy fully cleared.
If Black and Hispanic shares fell, somebody's share rose.
SFFA was brought by Asian-American plaintiffs alleging that Harvard's admissions process held them to higher standards. After the ruling, Asian enrollment rose at every HYPSM school except Yale — and Yale's Year 1 dip turned into a Year 2 rebound to 30%.
Harvard recorded a 38% increase in Asian enrollment over baseline. Stanford's Class of 2029 has the largest Asian and white shares in three years. MIT's Asian share spiked to 47% in Year 1 before IPEDS methodology pulled the figure back to 38% — not a real reversal, just multiracial students getting reclassified.
Methodology matters here. Schools that switched from self-reported (multiple identities counted) to IPEDS single-category reporting between 2027 and 2028 produced numbers that aren't directly comparable. Year-over-year deltas for Harvard, Duke, and MIT need that asterisk.
Tier 2 schools saw the same pattern, sometimes sharper.
EdWorkingPapers' system-wide study found URM enrollment at the Ivy+ cohort fell 18.9% — about three times the drop measured across all highly selective schools.
Cornell saw the steepest individual decline: Black enrollment fell from 11.7% to 4.3% in Year 1 and only edged up to 4.8% in Year 2. Columbia's Year 1 Black share dropped to 12% from 20%. Brown's CDS data showed Black enrollment falling from 9% to 5% before Year 2 polling suggested a rebound to 12% — if confirmed, the strongest recovery signal in the tier.
Schools that refused to publish are part of the story too. Dartmouth declined to release a full Class of 2029 demographic profile. UChicago and Caltech have published nothing. Duke held its data for two months before federal reporting pressure forced disclosure — under a methodology that erased its baseline.
Liberal arts colleges fragmented in unexpected ways.
Tier 4 selective privates produced the most varied response in the data set. Williams was the lone outlier in the wrong direction: Black and Hispanic enrollment rose slightly in Year 1, while Asian and white enrollment fell — the opposite of the HYPSM pattern. Strong financial-aid guarantees may explain the divergence.
Middlebury saw the sharpest LAC drop: students of color from 35% to 26%, a 26% decline. NYU lost 43% of its Black share in Year 1 and stopped publishing Year 2 numbers. Amherst is the clearest recovery story — Black enrollment rebounded from 3% to 6% under IPEDS, and to 12% under self-reporting, suggesting genuine policy adaptation rather than methodological noise.
The takeaway: selectivity tier alone doesn't predict response. Institutional posture — financial aid, recruiting reach, supplemental essays, public messaging — produced visible differences within tiers.
Public flagships moved in the opposite direction.
Public flagships logged an 8% URM increase system-wide — against 3.2% overall enrollment growth. At less selective publics, the gains were dramatic: LSU's Black first-year enrollment rose 30%, Ole Miss saw a 50% increase, UT Knoxville and South Carolina each posted 33%+ Hispanic gains.
These are real but mostly sit outside the simulation's selective-public set. At UVA, the closest comparable Tier 5 case, Hispanic enrollment rose 25% while Black enrollment fell 12% — a moderate, mixed signal. The UC system saw small Black and Hispanic gains at Berkeley and UCLA, though test-blind admissions limits comparability.
The simulation's current T5/T6 URM multipliers (1.30 / 1.25) overstate the effect. The real flagship signal closer to 1.08. Outsized gains live at non-selective publics that don't appear in the model.
The system absorbed the displaced students — just not at the same level.
The EdWorkingPapers study tracked where high-achieving URM students actually went. Diversity wasn't eliminated; it was redistributed downward. Students with SAT 1550–1600 moved from Ivy+ to other highly selective schools and selective publics. The 1400–1550 cohort moved from Ivy+ into selective publics directly.
The result: at universities with graduation rates above 80%, Black enrollment fell 1.6 percentage points and Hispanic 1.0pp. At less-selective four-year institutions, both groups rose roughly 5.9%. The pipeline didn't lose students; it relocated them.
This is the part the simulation gets approximately right at the system level — a downward
cascade across selectivity tiers — but underestimates at the top. The Tier 1–3
SFFA multipliers in sim.js capture about half the observed magnitude.
There's one more variable that complicates everything: who's even reporting.
Of the 50+ selective colleges that released demographic profiles for the Class of 2028, fewer than half did so for the Class of 2029. By January 2026 the Murphy tracker held only 29 schools with comparable Year 2 data.
The "unknown race" category compounds the problem. Princeton's share rose from 7.7% to 8.2%; across the data set, 5–10% of students now decline to report. That's enough to mask a 2–3 percentage-point shift in any minority category.
IPEDS' new ACTS supplement, with first reporting due March 2026, will require all selective institutions to disclose race-by-sex applicant data going back to 2019. That dataset will land in Fall 2026 and is likely to be the first comprehensive picture of the post-SFFA admissions environment.
Two cycles after the Supreme Court banned race-conscious admissions, the empirical picture at the most selective colleges is sharper than first models suggested. Black enrollment at HYPSM has fallen by roughly a third on average. Hispanic enrollment has fallen by half that. Asian enrollment has risen at most schools by a comparable margin.
The most important pattern in Year 2 is the absence of a clean rebound. Most schools with falling minority shares saw the same or further declines. A small number — Cornell, Brown, Amherst — have begun to recover. Princeton illustrates the opposite: a cliff that arrived a year late, suggesting that pipeline effects from race-conscious recruiting take longer than one cycle to drain out.
The simulation's current SFFA multipliers underestimate magnitude at HYPSM and Ivy+ by roughly a factor of two on Black enrollment, while overstating the public-flagship countertrend by roughly threefold. Splitting URM into separate Black and Hispanic multipliers — effects that differ by a factor of two in the data — is the cleanest improvement available before the IPEDS ACTS dataset arrives in fall 2026.