15 — Civil Rights, Equal Treatment & Free Expression

15 — Civil Rights, Equal Treatment & Free Expression

STAGE 1 — EVIDENCE BRIEF (Part 1 of 2)

Referee panel: Prof. Olivia Brandt (#57, statistics, C) · Prof. June Takahashi (#17, econometrics, C) · Domain specialist: Prof. Catherine Doyle (#25, election law, C)

Today is 2026-09-24. Every figure below was checked against the linked source this session. Where a source is an advocacy organization or law-firm summary rather than a primary document, we say so. Figures are reported as the source states them. Nothing here is a verdict; it is the shared factual floor for the debate.

---

A. Racial gaps in wealth, income and health

F1. Median family wealth, 2022 (Fed SCF, 2022 dollars). White non-Hispanic $285,000; Black non-Hispanic $44,900; Hispanic $61,600. In 2019 (2022 dollars): $218,100 / $28,000 / $41,900. Growth 2019→22: White +31%, Black +60%, Hispanic +47%. — Federal Reserve, Changes in U.S. Family Finances from 2019 to 2022, Oct 2023. https://www.federalreserve.gov/publications/files/scf23.pdf

F2. Ratio vs. dollar gap (referee computation from F1). White/Black median ratio fell from 7.8× (2019) to 6.3× (2022); the absolute dollar gap rose from $190,100 to $240,100. Both "the gap is narrowing" and "the gap is widening" are arithmetically true depending on metric. The next SCF (2025 survey) had not been published as of this writing.

F3. Household income, 2025 (Census CPS-ASEC, released Sept 15, 2026). Real median household income $87,460, up 2.6% from $85,210 in 2024 — the highest since the series began in 1967. By group, 2024→2025: Black households +4.8% (largest gain), non-Hispanic White +2.9%; Asian and Hispanic changes not statistically significant. — Census P60-289. https://www.census.gov/library/publications/2026/demo/p60-289.html

F4. The prior year went the other way. 2023→2024: Black median household income fell 3.3%, Asian rose 5.1%, non-Hispanic White no significant change; Hispanic median was $70,950 in 2024; all-household median $83,730 (2024 dollars). — Census, Sept 2025. https://www.census.gov/library/stories/2025/09/median-household-income.html Referee note: single-year swings in group medians are noisy; do not build a trend on one year.

F5. Life expectancy at birth, 2024 (NCHS U.S. Life Tables, Aug 25, 2026). Non-Hispanic Asian 85.8; Hispanic 81.8; non-Hispanic White 78.9; non-Hispanic Black 74.8; non-Hispanic American Indian/Alaska Native 71.1. Every group rose 0.5–1.0 year vs 2023. — NVSR 75(5). https://www.cdc.gov/nchs/data/nvsr/nvsr75/nvsr75-05.pdf

F6. Maternal mortality, 2024 (NCHS). Overall 17.9 per 100,000 live births (649 deaths), vs 18.6 in 2023 (not a significant change). Black non-Hispanic 44.8 (2023: 50.3); White non-Hispanic 14.2; Hispanic 12.1; Asian non-Hispanic 18.1. The Black rate is significantly higher than every other group's; year-over-year changes within groups were not significant. — Health E-Stat 113. https://www.cdc.gov/nchs/data/hestat/hestat113.htm

B. What the causal literature on discrimination actually shows

F7. Kline, Rose & Walters (QJE 2022). 83,000+ fictitious applications to 108 of the largest U.S. employers. Distinctively Black names reduced employer contact by 2.1 percentage points. Discrimination was concentrated: firms in the top quintile were responsible for nearly half of lost contacts to Black applicants; 23 individual companies were identified as discriminating with high statistical confidence. Industry explained roughly half the cross-firm variation. No significant average gender gap, but a between-firm SD of 2.7 pp. — https://www.nber.org/papers/w29053 ; https://academic.oup.com/qje/article-abstract/137/4/1963/6605934

F8. Quillian, Pager, Hexel & Midtbøen (PNAS 2017), meta-analysis of hiring field experiments. Whites received on average 36% more callbacks than equally qualified African Americans; no change in discrimination against African Americans 1990–2015; "some evidence" of decline for Latinos, but few studies. — https://www.pnas.org/doi/10.1073/pnas.1706255114 ; summary: https://www.ipr.northwestern.edu/news/2017/infographic-quillian-hiring-discrimination.html

F9. Chetty, Hendren, Jones & Porter (QJE 2020), "Race and Economic Opportunity." Conditional on parental income, the Black-white income gap is driven by men (wages and employment); there is no such gap between Black and white women. "Differences in parental marital status, education, and wealth explain little of the black-white income gap conditional on parent income." Black boys earn less than white boys from same-income families in 99% of Census tracts; exceptions are low-poverty areas with low racial bias among whites and high rates of Black father presence — and fewer than 5% of Black children grow up in such areas. — https://www.nber.org/papers/w24441

F10. Forscher, Lai et al. (JPSP 2019), meta-analysis of implicit-bias interventions. 492 studies, 87,000+ participants. Procedures can shift implicit measures, but in the 63 studies testing it the authors "found no evidence of a causal relationship" between changing implicit bias and changing behavior. — https://www.sciencedaily.com/releases/2019/08/190802144415.htm

C. Higher education after SFFA v. Harvard (2023)

F11. Bleemer (QJE 2022), California Prop 209. Ending affirmative action at UC cascaded URM applicants to less selective schools; their degree attainment fell (overall and in STEM, especially among less-prepared applicants) and average wages in their 20s–30s declined, driven by Hispanic applicants. The mismatch hypothesis's predicted net benefit was not found. — https://opportunityinsights.org/paper/affirmative-action-mismatch-and-economic-mobility-after-californias-proposition-209/

F12. Harvard College. Class of 2029 (fall 2025): Black 11.5%, Hispanic 11%, Asian American 41%, race unreported 8%. Class of 2028: Black 14%, Hispanic 16%, Asian American 37%. Class of 2027 (last pre-SFFA class): Asian American 29.9%. Harvard did not publish a White share. — https://www.harvardmagazine.com/university-news/harvard-admissions-class-2029-admissions-data-ethnicity

F13. AP analysis of selective colleges (via Inside Higher Ed, Oct 2025). Black first-year share: Harvard 18% (2023) → 11.5% (2025); Princeton 9% (2024) → 5%; Columbia 20% (2023) → 13%; Amherst 11% → 6%; Caltech 5% → 1.6%. Smith rose 4.6% → 6.8%. Black students are ~14% of U.S. high-school graduates. — https://www.insidehighered.com/news/quick-takes/2025/10/27/black-student-enrollment-shrinks-selective-institutions

F14. The cascade (Class Action report on 3,000+ institutions; reported by The EDU Ledger). Fall 2024: at "Ivy Plus" institutions Black enrollment fell 25%, Hispanic 15%, Asian American +7%; among the 50 most selective institutions, 34 saw Black declines; 83% of state flagships saw diversity increases (e.g., Black freshman enrollment +30% LSU, +50% Ole Miss, ~+20% Michigan). The report argues legacy preferences magnified the drop at elites. Referee note: advocacy-organization report, secondary coverage. — https://www.theeduledger.com/demographics/article/15816536/elite-college-diversity-plummets-after-affirmative-action-ban-but-most-schools-see-gains

F15. Public opinion. Pew (spring 2023): 33% of adults approve of considering race in admissions, 50% disapprove; Black adults 47% approve (the only group with more approval than disapproval), Hispanic 39%, Asian 37%, White 29%. — https://www.pewresearch.org/short-reads/2023/06/16/americans-and-affirmative-action-how-the-public-sees-the-consideration-of-race-in-college-admissions-hiring/ Pew (Oct 2024): workers saying focusing on DEI is "a good thing" fell from 56% (Feb 2023) to 52%; "bad thing" rose 16% → 21%. — https://www.pewresearch.org/short-reads/2024/11/19/views-of-dei-have-become-slightly-more-negative-among-us-workers/

(continued in Part 2)

STAGE 1 — EVIDENCE BRIEF (Part 2 of 2)

Brandt (#57) · Takahashi (#17) · Doyle (#25)

D. Federal DEI rollback and disparate impact (2025–26)

F16. DEI executive orders. EOs 14151 and 14173 (Jan 2025) direct termination of "equity-related" grants/contracts and require contractors and grantees to certify they run no programs violating federal anti-discrimination law. On Feb 6, 2026 the Fourth Circuit (NADOHE v. Trump, opinion by Chief Judge Diaz) vacated the district court's preliminary injunction: plaintiffs were unlikely to win a facial challenge, because the Certification Provision on its text requires compliance with existing law and "doesn't say" all DEI is illegal. Diaz's concurrence warned the ruling is not "a blank check for overreach" and invited as-applied challenges. — https://www.pillsburylaw.com/en/news-and-insights/fourth-circuit-anti-dei-executive-orders-likely-not-facially-unconstitutional.html

F17. Disparate impact EO. EO 14281, "Restoring Equality of Opportunity and Meritocracy" (Apr 23, 2025): "[i]t is the policy of the United States to eliminate the use of disparate-impact liability in all contexts." DOJ rescinded the "effects" provisions of its Title VI regulations (28 CFR 42.104(b)(2), (b)(3), (b)(6), (c)(2)) effective Dec 10, 2025, without notice-and-comment, invoking the APA's grants/contracts exception. DOJ's notice relies on Alexander v. Sandoval (2001) — no private right to enforce Title VI disparate-impact regulations — and on the regulations' "considerable tension" with Title VI. The rule still allows statistics as evidence of intentional discrimination. — https://www.federalregister.gov/documents/2025/12/10/2025-22448/rescinding-portions-of-department-of-justice-title-vi-regulations-to-conform-more-closely-with-the ; https://www.pillsburylaw.com/en/news-and-insights/us-doj-rescinds-disparate-impact-rule.html

F18. Other agencies followed: Commerce (Fed. Reg. Apr 16, 2026), DHS/FEMA (June 22, 2026), Education (July 24, 2026). — https://www.federalregister.gov/documents/2026/04/16/2026-07477/rescinding-portions-of-department-of-commerce-title-vi-regulations-to-conform-more-closely-with-the ; https://www.federalregister.gov/documents/2026/07/24/2026-15019/rescinding-portions-of-the-department-of-education-title-vi-regulations-to-align-with-the-statutory ; https://regulations.justia.com/regulations/fedreg/2026/06/22/2026-12399.html

F19. Limit of the executive action (legal background, Doyle). Title VII's employment disparate-impact standard is statutory — codified by the Civil Rights Act of 1991 at 42 U.S.C. §2000e-2(k). Agency regulations cannot repeal it; an administration can only decline to bring such cases.

E. Voting rights

**F20. Louisiana v. Callais (decided Apr 29, 2026).** 6–3, Alito for the Court (Roberts, Thomas, Gorsuch, Kavanaugh, Barrett); Kagan dissenting with Sotomayor and Jackson. The Court struck Louisiana's map with a second majority-Black district and held that compliance with §2 "as properly construed" can justify race-based districting only where §2 actually requires it. It said it "need only update the [*Gingles*] framework" — i.e., modified, not overruled. Plaintiffs must now show: a compact majority district drawable using race-neutral criteria; illustrative maps that meet all the State's legitimate objectives including political goals; racially polarized voting "that cannot be explained by partisan affiliation"; and totality-of-circumstances evidence giving rise to "a strong inference that intentional discrimination occurred." The dissent says this "eviscerates" §2 by functionally reinstating the intent standard Congress rejected in 1982. — https://www.supremecourt.gov/opinions/25pdf/24-109_21o3.pdf

F21. Aftermath (advocacy-source summary, May 2026). Tennessee enacted a map on May 7 splitting Memphis/Shelby County into three districts (likely eliminating its only Democratic seat); Louisiana opened redistricting hearings May 8 and suspended primary voting; Alabama authorized a new primary plan contingent on courts; a South Carolina House committee recommended delaying the primary with CD-6 in view. — https://www.americandemocracyminute.org/wethepeople/2026/05/11/callais-decision-fallout-as-louisiana-tennessee-alabama-south-carolina-scramble-to-eliminate-black-congressional-districts Referee note: final enacted maps in LA/AL/SC and their seat effects were not verified this session.

F. Hate crime

F22. FBI 2024 data (released Aug 2025). 11,679 incidents (2023: 11,862, −1.5%); ~16,000 agencies covering 95.1% of the population. Race/ethnicity was the bias in more than half; anti-Black the largest single category. — https://www.axios.com/2025/08/05/hate-crimes-2024-black-antisemitism-muslim-fbi
F23. Anti-Jewish incidents 1,938 (+5.8%), highest the FBI has recorded since 1991, ~70% of religion-based incidents (2,783 total) and ~16% of all incidents; anti-LGBTQ 2,390 (down from 2,557); anti-Muslim 228 (from 236). — ADL summary of FBI data: https://www.adl.org/resources/press-release/anti-jewish-hate-crimes-comprised-nearly-70-all-religion-based-hate-crimes

G. Campus speech, Title VI, university funding

F24. FIRE 2026 College Free Speech Rankings (68,510 students, 257 schools; fielded Jan 3–Jun 5, 2025): 166 (65%) schools got an F; 41% of students self-censor at least occasionally with peers, 38% in class; 36% say shouting down a speaker is at least sometimes acceptable, 15% say violence is. Top: Claremont McKenna (B−); bottom: Barnard, Columbia, Indiana, Washington, Northeastern. For the first time a majority opposed allowing every hypothetical controversial speaker tested (three left-coded, three right-coded), and conservative students' former ~90% rejection of violence has eroded. — https://www.fire.org/research-learn/2026-college-free-speech-rankings ; https://eternallyradicalidea.com/p/five-key-takeaways-from-fires-2026 Referee note: opt-in online panel; FIRE is an advocacy organization.

F25. Settlements. Penn (July 1, 2025; no payment; Title IX athlete records); Columbia (late July 2025; $221M — $200M + $21M EEOC; admissions data reporting); Brown (July 30, 2025; $50M over 10 years to workforce groups; adopted EO sex definitions); UVA (Oct 22, 2025; no payment; ended DEI); Cornell (Nov 7, 2025; $60M); Northwestern (Nov 2025; $75M). — https://www.npr.org/2026/01/29/nx-s1-5559293/trump-settlements-colleges-universities

F26. Harvard. On Sept 3, 2025 Judge Burroughs held the freeze/termination of ~$2.2B in grants and $60M in contracts unlawful (First Amendment retaliation; Title VI procedure; arbitrary and capricious): "combatting antisemitism cannot be accomplished on the back of the First Amendment." She also found Harvard was "even if belatedly" taking needed steps against antisemitism. The government appealed (Dec 2025) and briefed the First Circuit in April 2026. — https://www.highereddive.com/news/judge-strikes-down-trump-administration-harvard-2b-funding-freeze-Burroughs/759198/ ; https://www.thecrimson.com/article/2026/4/16/trump-funding-freeze-appeal/

H. Religion, sex/gender, speech and government pressure

**F27. 303 Creative v. Elenis** (June 30, 2023; 6–3, Gorsuch): Colorado cannot compel a web designer to create expressive wedding websites contrary to her beliefs. — https://supreme.justia.com/cases/federal/us/600/21-476/
**F28. Mahmoud v. Taylor** (June 27, 2025; 6–3, Alito): parents entitled to a preliminary injunction requiring opt-outs from LGBTQ-inclusive storybook instruction. Strict scrutiny applied because the books conveyed "a normative message on the subjects of sex and gender, beyond mere exposure to other ideas," substantially interfering with religious upbringing. — https://www.faegredrinker.com/en/insights/publications/2025/6/supreme-court-decides-mahmoud-v-taylor
**F29. United States v. Skrmetti (June 18, 2025; 6–3, Roberts): Tennessee's ban on puberty blockers/hormones for minors' gender dysphoria gets rational-basis review; the law classifies by age and medical use**, not sex, and (per the majority) not transgender status. — https://supreme.justia.com/cases/federal/us/605/23-477/
**F30. West Virginia v. B.P.J. / Little v. Hecox** (June 30, 2026; 6–3, Kavanaugh): "sex" in Title IX means biological sex; states may limit girls' and women's teams to biological females; intermediate scrutiny satisfied. Sotomayor dissented. — https://www.supremecourt.gov/opinions/25pdf/24-43_2b35.pdf
**F31. Murthy v. Missouri (June 26, 2024; 6–3, Barrett): plaintiffs lacked standing; the Court did not** decide whether the Biden administration's contacts with platforms were coercive. Alito, Thomas, Gorsuch dissented. — https://supreme.justia.com/cases/federal/us/603/23-411/ (Companion: NRA v. Vullo (2024, unanimous) reaffirmed that officials may not coerce private parties to punish speech.)
F32. Kimmel episode (Sept 2025). FCC Chair Carr, on a podcast before ABC suspended Kimmel: broadcasters hold "a license granted by us… we can do this the easy way or the hard way." — https://www.pbs.org/newshour/politics/after-jimmy-kimmels-show-was-suspended-a-key-question-is-does-the-fcc-have-the-power-to-regulate-speech
F33. Press freedom. RSF 2026 Index: U.S. 64th, down from 57th in 2025; RSF cites presidential attacks on the press as "systematic policy," the deportation of journalist Mario Guevara, and USAGM/VOA cuts. — https://rsf.org/en/2026-rsf-index-press-freedom-25-year-low

I. Disability rights

**F34. Texas v. Kennedy.** In April 2025, 17 states stated they no longer seek a ruling that Section 504 is unconstitutional. — https://www.deque.com/blog/states-drop-constitutional-challenge-to-section-504-in-texas-v-kennedy/ On Jan 23, 2026, nine states (TX, AK, FL, IN, KS, LA, MO, MT, SD) renewed a challenge to HHS's 2024 Section 504 integration-mandate rule ("most integrated setting appropriate"; "serious risk of unnecessary institutionalization"). — https://justiceinaging.org/texas-and-eight-other-states-renew-attack-on-section-504/

---

CONTESTED EVIDENCE — where the literature really disagrees

1. Why does the wealth gap persist? Intergenerational-transmission/history accounts (e.g., Derenoncourt, Kim, Kuhn & Schularick, "Wealth of Two Nations," QJE 2024 — cited from the panel's knowledge, not re-fetched this session) put heavy weight on starting conditions in 1865 and differential returns. Individual/family-structure accounts point to F9's father-presence result. F9 itself cuts both ways: the same tract-level finding names both white racial bias and Black father presence. F9 also says parental wealth explains little of the income gap conditional on parent income — awkward for pure-inheritance stories.
2. How much discrimination, and where? F7 and F8 agree hiring discrimination is real and not trivially small. They disagree in emphasis: F8 finds no decline since 1990; F7 finds it concentrated in identifiable firms. Name-based audits may confound race with perceived class — a known critique (e.g., Gaddis 2017 on name perception; not re-fetched).
3. Affirmative action and outcomes. F11 (Bleemer) finds bans hurt URM degree completion and wages. The "mismatch" literature (e.g., Arcidiacono, Aucejo & Hotz on UC STEM persistence; not re-fetched) finds lower STEM persistence for less-prepared students at more selective campuses. The two are not fully contradictory (different outcomes, different margins).
4. Does DEI training work? F10 and the organizational literature (Dobbin & Kalev; not re-fetched) find weak behavioral effects for mandatory trainings; structural measures (accountability, mentoring, recruitment) do better. This undercuts training-centric DEI, not necessarily enforcement against discrimination.
5. Hate-crime trends. FBI counts depend on voluntary agency reporting and the NIBRS transition; year-to-year changes of ±2% are within reporting noise. BJS's NCVS historically estimates far more hate victimizations than police data record.
6. Campus speech measurement. FIRE's opt-in panel is large but non-probability; whether self-censorship reflects a "chilling" climate or ordinary social prudence is disputed.
7. Callais's real-world effect. How many majority-minority districts disappear by 2028, and whether minority-preferred candidates win in crossover districts, is an open empirical question.

STAGE 2 — OPENING POSITIONS (1 of 3)

---

Prof. Marcus Bell (#22, civil rights law, L)

Causal claim. Present-day discrimination, not only history, is a measurable cause of racial gaps in jobs and income, and disparate-impact liability is the only tool that catches practices nobody can prove were adopted with intent. Taking it away removes the main enforcement lever against facially neutral screens.

Evidence. Kline–Rose–Walters: a 2.1-point contact penalty for Black names across 108 of the largest employers (F7). Quillian et al.: 36% more callbacks for white applicants and no decline since 1990 (F8). Put simply, a generation of anti-discrimination law has not moved hiring discrimination against Black applicants at all. Meanwhile EO 14281 says to eliminate disparate impact "in all contexts" (F17), and four agencies have deleted their Title VI effects regulations (F17–F18). With DOJ's rescission, disparate-impact liability for employers is effectively gone.

Remedy. (1) Congress should codify a Title VI disparate-impact standard and a private right of action, overriding Sandoval. (2) Fund continuous federal audit testing of large employers.

Tradeoff I concede. Disparate-impact liability pushes employers toward defensive practices: dropping valid tests, or informal quota-like hiring to avoid statistical exposure. That is a real cost, and Ricci (2009) shows the law can put employers in a bind.

Falsifiable prediction. If the rollback matters, EEOC and DOJ Title VI resolutions that involve neutral-practice claims will fall to near zero by FY2027. And if discrimination is a cause of the income gap, the next audit waves (2027–2029) will show no improvement, or a worse Black callback penalty.

---

Prof. Imani Walker (#65, African American studies, L)

Causal claim. The racial wealth gap is structural. It comes from compounding differences in starting wealth, homeownership access and inheritance, and it is not closing.

Evidence. White median wealth is $285,000 against $44,900 for Black families (F1). The gap is widening: the dollar gap grew from $190,100 to $240,100 between 2019 and 2022 (F2). On income, Census reports that Black household income fell 3.3% in its latest release while other groups rose (F4). Health tells the same story: Black women die of pregnancy-related causes at 44.8 per 100,000, over three times the white rate (F6). Black life expectancy trails white by 4.1 years (F5).

Remedy. Federal "baby bonds": publicly seeded, means-tested trust accounts at birth, as Connecticut enacted in 2021. Also stronger appraisal-bias enforcement.

Tradeoff I concede. Baby bonds are race-neutral by design, so they close the gap only partially and slowly. They cost real money, and at age 18 the cash could be spent rather than saved.

Prediction. If the gap is structural rather than behavioral, the 2025 SCF (due 2026–27) will show the Black-white dollar gap widening again even if the ratio holds steady.

---

Prof. Amara Johnson (#38, voting rights, L)

Causal claim. Callais removes the federal guarantee against racial vote dilution in the South. In practice, Black representation will now depend on the partisan interests of legislatures.

Evidence. The Court overruled Gingles. Plaintiffs must now show polarization "that cannot be explained by partisan affiliation" and a "strong inference" of intentional discrimination (F20). In a region where race and party are highly correlated, that standard can almost never be met. Within eight days Tennessee split Memphis three ways, and Louisiana suspended a primary already under way (F21).

Remedy. A statutory amendment restoring an effects-based §2 for vote dilution, with an explicit instruction that partisan correlation does not defeat a claim. Also a new coverage formula tied to recent violations, answering Shelby County.

Tradeoff I concede. Race-conscious districting entrenches racial sorting in politics, and majority-minority districts can "pack" Black voters so that they lose influence elsewhere. That is a genuine cost and it is not just a Republican talking point.

Prediction. By the 2028 election, every majority-Black congressional district in the Deep South outside a single metro core will be dismantled, and the number of Black members of Congress from the former Confederacy will fall by at least a third.

STAGE 2 — OPENING POSITIONS (2 of 3)

---

Prof. Carlos Ibarra (#74, Latino studies, C-L)

Causal claim. Ending race-conscious admissions did not end preferences at elite colleges. It removed the preference that helped the disadvantaged and kept the ones that help the advantaged: legacy, donor and athletic. The net effect is regressive, and it harms Hispanic students in particular.

Evidence. In California, Bleemer shows the Prop 209 cascade lowered URM degree completion and wages, with the wage loss driven by Hispanic applicants (F11). Harvard's Hispanic share fell from 16% to 11% in one year (F12). Hispanic enrollment at Ivy-Plus schools dropped 15% (F14). SFFA proves affirmative action was helping Hispanic students economically.

Remedy. Condition federal student-aid eligibility on ending legacy and donor preferences, and add a meaningful socioeconomic preference that uses neighborhood and school disadvantage, not only family income.

Tradeoff I concede. Class-based preferences recover only part of the racial diversity. Hispanic and Black students are over-represented among low-income applicants, but most low-income high scorers are white or Asian. So race-neutral policy will not restore pre-2023 numbers at the very top.

Prediction. At institutions that drop legacy preferences, the combined Hispanic and Black share will recover by at least 2 percentage points within three admissions cycles compared with peers that keep them.

---

Prof. Henry Castell (#66, ethics, C)

Causal claim. In 2025–26 the main threat to civil rights as an institution is that both sides treat enforcement as a weapon. Title VI was designed with procedural safeguards: findings, a hearing, and a report to Congress before any fund cutoff. Those safeguards were skipped. The antisemitism problem was real, and the response was lawless.

Evidence. Anti-Jewish hate crimes hit 1,938 in 2024, the FBI's highest count since 1991, and nearly 70% of religion-based incidents (F23). Columbia and Barnard sit at the bottom of FIRE's rankings, and students there report heavy self-censorship (F24). The universities failed Jewish students. Yet Judge Burroughs found the Harvard freeze unlawful, and she found that the government never weighed whether particular grantees had engaged in antisemitic conduct (F26). Settlements then paid money to the Treasury and to third parties (F25) rather than to victims. Hate crimes overall fell in 2024 (F22), which shows how selective the enforcement surge was.

Remedy. A statute requiring the Title VI procedures of 42 U.S.C. §2000d-1 before any freeze, with an expedited timeline, so enforcement is fast but lawful. Settlement funds should go to affected students, not general revenue.

Tradeoff I concede. Formal process is slow. Some universities stalled for years under the old process, and Jewish students bore the cost of that delay.

Prediction. If process is the problem, the First Circuit will affirm Burroughs on the procedural grounds at least. And universities that settled will show no larger improvement in Jewish students' reported climate (Hillel/ADL surveys, 2027) than peers that did not settle.

---

Prof. Eleanor Whitfield (#21, con law, originalist, R)

Causal claim. The Fourteenth Amendment and Title VI forbid treating individuals differently by race. The 2023–26 corrections (SFFA, Callais, the Title VI rescissions) restore that rule; they do not create a new one. Disparate impact was always regulatory overreach resting on shaky legal ground.

Evidence. In Sandoval the Supreme Court said Title VI itself reaches only intentional discrimination, and it expressed "considerable tension" with the effects regulations (F17). So DOJ has simply brought its rules into line with the statute. Callais kept Gingles and aligned it with §2's text (F20). As for consequences: 83% of flagships saw diversity rise after SFFA (F14), and Asian American students, the actual plaintiffs' class, gained: Harvard went from 29.9% to 41% Asian American (F12). SFFA ended a penalty on them.

Remedy. Codify EO 14281's principle for Title VI, and let Title VII's statutory standard be revisited by Congress, not by agency fiat. Require race-neutral illustrative maps in §2 cases.

Tradeoff I concede. At the most selective schools Black enrollment did fall sharply (F13). A colorblind rule has visible distributional effects, and supporters of it should own that rather than deny it.

Prediction. The fall 2026 and 2027 data will show the "cascade" schools keeping their gains, and aggregate Black and Hispanic bachelor's completion at four-year institutions will not fall. That is the opposite of what the Prop 209 literature predicts.

STAGE 2 — OPENING POSITIONS (3 of 3)

---

Prof. Arthur Pembroke (#27, political theory, R)

Causal claim. Today's racial gaps come mainly from differences in family structure, skills and geography that are shaped by history. Employer discrimination is a real but secondary cause, confined to a small number of bad actors. DEI bureaucracy is the wrong response: it treats every institution as guilty, and it doesn't change behavior anyway.

Evidence. Chetty et al. find that among children from same-income families there is no Black-white income gap for women. The gap exists among men, and it narrows sharply where Black fathers are present (F9). Parental education and wealth "explain little" (F9). The newest audit study, Kline–Rose–Walters, shows discrimination is concentrated: the top fifth of firms account for about half of lost contacts, and 23 firms can be named (F7). Hiring discrimination has fallen sharply since the Bertrand–Mullainathan era. On DEI, a meta-analysis of 492 studies found no evidence that shifting implicit bias changes behavior (F10). The public has soured too: only 52% of workers now call DEI a good thing, and 50% disapprove of race in admissions (F15).

Remedy. Enforce hard against intentional discrimination by the identifiable firms, using audit evidence, and stop funding compliance theater. Promote family-stability policy, for example by removing marriage penalties in means-tested programs.

Tradeoff I concede. "Intent-only" enforcement lets through practices that are harmful without being malicious, such as credential inflation. Family-structure policy works slowly, and its evidence base is thinner than its advocates claim.

Prediction. Firm-level audits repeated in 2027–2029 will show the concentration persisting, with the same firms and industries on top. That supports targeted enforcement over economy-wide DEI mandates.

---

Prof. Victor Lindqvist (#37, libertarian legal theory, Lib)

Causal claim. The most serious speech threat in America is government pressure on private intermediaries, called "jawboning," whichever party holds power. Both administrations have done it. Campus speech codes and federal funding threats are the same problem at a different scale.

Evidence. In Murthy v. Missouri the Court found that the Biden White House and the Surgeon General's office coerced platforms into removing lawful speech. It merely declined to issue an injunction (F31). Then in September 2025 the FCC chair told ABC "we can do this the easy way or the hard way," and Kimmel was suspended within hours (F32). The U.S. fell to 64th on RSF's index (F33). On campus, 65% of schools get an F from FIRE, 41% of students self-censor with peers, and 15% accept violence to stop speech (F24). Funding freezes that condition research money on viewpoint are jawboning with a checkbook (F26).

Remedy. A federal anti-jawboning statute with three parts: disclosure within 72 hours of any official request that a platform, broadcaster or licensee remove or punish lawful speech; a private right of action for coerced speakers; and a ban on using license-renewal or merger review over content. Add institutional neutrality on campus, meaning no official university positions on political controversies.

Tradeoff I concede. Disclosure rules will slow legitimate government communication with platforms, for instance about foreign influence operations, child-exploitation material or imminent threats. Carve-outs are needed, and carve-outs get abused.

Prediction. Absent a statute, the next administration of either party will produce a documented jawboning episode within its first two years. FIRE's share of students who accept violence will not fall below 10% by 2029 without institutional reform.

---

Prof. Nathaniel Cross (#64, religious studies, R)

Causal claim. Religious liberty conflicts arise when public institutions demand affirmation rather than mere tolerance. The Court's 2023–25 cases restore the line between exposure and compelled endorsement, and that line protects dissenters of every faith, including progressive religious minorities.

Evidence. 303 Creative protects expressive work from compelled messages (F27). Mahmoud holds that when instruction carries "a normative message on the subjects of sex and gender," parents are entitled to opt out (F28). The practical upshot is that parents may now opt their children out of any lesson that burdens their religious beliefs. The plaintiffs were Muslim, Catholic and Orthodox Christian families, not a single sect.

Remedy. Require public schools to give advance notice of curricula on sexuality and gender identity, with a simple written opt-out. Protect conscience for expressive professions, not for non-expressive commercial services.

Tradeoff I concede. Opt-out regimes impose administrative costs and can fragment classrooms. A broad reading of Mahmoud could let objectors opt out of evolution, history, or anything else, and I do not want that.

Prediction. If exposure/endorsement is the right line, notice-and-opt-out districts will see opt-out rates below 5% on average and no measurable drop in LGBTQ students' school-climate scores (GLSEN or state YRBS) by 2028.

STAGE 3 — CROSS-EXAMINATION (1 of 4): Walker ↔ Pembroke (the causes of the gap)

---

Pembroke (#27, R) → Walker (#65): Imani, you said Census shows Black household income "fell 3.3% in its latest release." That was the September 2025 release. The latest one came out nine days ago: Black household income rose 4.8% in 2025, the largest gain of any group, while non-Hispanic white households gained 2.9% (F3). Second, "the gap is widening" depends on the metric. The ratio went from 7.8× to 6.3× in three years (F2), and Black median wealth grew 60% against 31% for white families (F1). Any gap between a smaller base and a larger base widens in dollars whenever both grow at similar rates. Third, if the story is purely structural, why do Hispanic Americans outlive non-Hispanic whites by nearly three years (81.8 vs 78.9, F5), despite lower income and wealth? Health gaps don't map neatly onto wealth gaps.

Walker (#65, L): On income, I accept the correction. I was working from the 2024 release, and the 2025 figure reverses it. Noted on the record. On wealth I won't concede. The dollar gap is what buys a house, pays tuition or absorbs a medical shock, and $240,100 is larger than $190,100. A ratio can improve while opportunity stays out of reach. Also, the 2019–22 Black gain was mostly home appreciation during an unusual housing boom. On health: the "Hispanic paradox" is well known and partly reflects immigrant selection, so it doesn't refute structural accounts of Black health. Black maternal mortality is three times the white rate (F6). It doesn't fall to parity with income, either. Black college-educated women face higher maternal mortality than white women without a degree, a finding documented in the maternal-health literature.

Pembroke: That last claim may be true, but it isn't in the brief, so I'll let the referees decide what to do with it. Here is my core challenge. You call the gap "structural," but F9, the best intergenerational data we have, says parental wealth "explains little" of the income gap once parental income is held fixed. If inherited wealth were the main channel, it would show up there. What does predict smaller gaps is father presence among Black families in the tract.

Walker: Read the whole sentence in F9. The tracts where Black boys do well have low racial bias among whites and high father presence. Both conditions, and Chetty's team named the bias variable. You're citing half the finding. Add Quillian (F8): no decline in hiring discrimination against Black applicants in 25 years. And F9's own result that Black women show no gap conditional on parent income suggests something specific to how the labor market and police treat Black men. That points to discrimination, not "culture."

Pembroke: Or it points to different male outcomes in education, incarceration and family formation, each with many causes. I'll concede this much: F9 lists both variables, and the design can't separate them. They are correlated at the tract level. That's a real concession, and F9 forces it on me. But you should concede that F9 undercuts a pure wealth-transfer theory, which is what baby bonds rest on.

Walker: Partially conceded. F9 means baby bonds address wealth levels, not the income mobility gap among men. I still back them because the wealth gap is itself the outcome I care about. Wealth is a buffer, not just a pathway to income. But I'll narrow my claim: baby bonds are a wealth remedy, not a mobility remedy.

Pembroke: Accepted, and noted as a disagreement about ends: a floor under wealth versus equal opportunity.

---

Bell (#22, L) → Pembroke (#27): Arthur, you said hiring discrimination "has fallen sharply since the Bertrand–Mullainathan era." What's your source? The meta-analysis in F8 covers 1990–2015 and finds no change for Black applicants. And your "concentrated in a few bad apples" reading of KRW is incomplete. If the top quintile causes "nearly half" of lost contacts, the other 80% of firms cause the other half. That's systemic by any definition. The average penalty of 2.1 points is spread across the economy.

Pembroke: On the trend, I was extrapolating from the KRW point estimate, which is smaller in absolute terms than the 2004 study's gap. But the two designs, samples and base callback rates differ, so the comparison isn't valid. I withdraw "fallen sharply." Quillian governs until a newer meta-analysis updates it. On concentration, I stand by the policy implication even if "few bad apples" was too glib. If half of the harm comes from a fifth of firms that are identifiable by industry and name, targeted enforcement gets you most of the value.

Bell: Half the value. The other half is diffuse, and diffuse harms are exactly what disparate-impact and pattern-or-practice tools exist for.

Pembroke: Or what universal audit testing with firm-level publication would catch, without a single disparate-impact theory. We may agree on that instrument even though we disagree on why.

STAGE 3 — CROSS-EXAMINATION (2 of 4): Bell, Johnson, Whitfield

---

Sutter (#78, philosophy of law, C-R) → Bell (#22): Marcus, you said that with DOJ's rescission, "disparate-impact liability for employers is effectively gone." It isn't. Title VII disparate impact is written into the statute, 42 U.S.C. §2000e-2(k), from the 1991 Act (F19). DOJ's rule touches Title VI funding-recipient regulations (F17). Private employees can still sue employers under Title VII's disparate-impact provision tomorrow. What changed is federal agencies' own enforcement choices. That matters, but it's a different claim.

Bell (#22): Fair. I'll amend on the record: federal disparate-impact enforcement is being withdrawn, the EEOC is no longer bringing such cases, and Title VI effects regulations are gone at four agencies. Private Title VII disparate-impact suits survive. The practical bite is still large, because private plaintiffs rarely have the resources for statistical pattern cases the government used to bring. But I overstated the legal claim.

Sutter: And the rescission was done without notice-and-comment (F17). I'll concede that's a procedural vulnerability even though I support the policy. A rule that took half a century to accumulate and was undone by an APA exception can be restored the same way.

---

Whitfield (#21, R) → Johnson (#38): Amara, you said the Court "overruled Gingles." It said the opposite: it "need[s] only update the framework" (F20). You're entitled to think the update is fatal. You aren't entitled to say the Court overruled a precedent it expressly retained. And your prediction that every majority-Black Deep South district outside a single metro core will be gone by 2028 is unfalsifiable as written. Which districts? Which states?

Johnson (#38): On the formal point, amended: Gingles was modified, not overruled. My claim is that it has been functionally hollowed out. Consider "polarization that cannot be explained by partisan affiliation" in a region where about 90% of Black voters vote one way. Separating race from party there is an evidentiary requirement that plaintiffs can almost never meet. That's why the dissent says it "eviscerates" §2 (F20). On the prediction, I'll make it precise: Louisiana CD-6 (the second majority-Black district), Alabama's court-ordered CD-2, Tennessee's Memphis CD-9 (already split, F21), and South Carolina's CD-6 in its current form. I predict at least three of the four are gone or below 40% Black VAP before the 2028 general election.

Whitfield: That I'll accept as a testable prediction. My counter-prediction: several of those states will still elect Black members from crossover districts, because a race-neutral map doesn't mean no Black representatives.

Johnson: Crossover districts in Mississippi or Louisiana? Name one that a Black-preferred candidate has won in the Deep South without a Black majority at the congressional level.

Whitfield: I can't name one in the Deep South offhand, and I won't pretend I can. My point is conditional. I'll narrow it to "outside the Deep South."

---

Rosenthal (#35, legal history, C-L) → Whitfield (#21): Eleanor, three problems. (1) You said Sandoval held that Title VI "itself reaches only intentional discrimination." That comes from Bakke and Guardians; Sandoval assumed it. Sandoval actually decided the private-right-of-action question and expressly assumed the regulations' validity while noting "considerable tension" (F17, as DOJ's own notice describes it). You're turning an assumption into a holding. (2) Flagships gaining Black students (F14) isn't a counterweight to Bleemer. It is the Bleemer mechanism. Bleemer's cascade also raised URM enrollment at less selective UC campuses and still found lower degree completion and wages (F11). (3) "Asian Americans gained" at Harvard, from 29.9% to 41% (F12). Perhaps, but 8% of the class of 2029 didn't report race, which contaminates every year-over-year share comparison.

Whitfield: (1) Accepted in part. The intent-only reading of Title VI rests on Bakke/*Guardians* plus Sandoval's reasoning, and the Court assumed without deciding the regulations' validity. My claim that DOJ "simply brought its rules into line with the statute" is contestable, not settled. (2) I'll hold that Bleemer is about California in the 1990s, before today's financial aid and ranking landscape. It's evidence, not destiny, which is why I made a falsifiable prediction about completion. (3) On non-reporting, conceded. The trend is suggestive and the magnitude is uncertain.

Blackwood (#79, Native American studies, L) → Whitfield: One more on Callais. A requirement that illustrative maps satisfy "all the State's legitimate objectives including political goals" (F20) cuts hardest against Native voters. Reservation communities are geographically compact but sparse, and §2 cases have been their main tool for county- and legislative-level representation. AIAN life expectancy is 71.1 (F5), the lowest of any group, and local political voice over IHS clinics, roads and schools is not abstract. Does your colorblind principle carry any duty to measure who loses?

Whitfield: It carries a duty to measure, yes. I'll support adding tribal-area outcomes to any post-*Callais* monitoring. It doesn't carry a duty to district by race. We disagree about the remedy, not about the obligation to look at the data.

STAGE 3 — CROSS-EXAMINATION (3 of 4): Ibarra, Castell, Cross

---

Thorne (#75, intellectual history, Lib) → Ibarra (#74): Carlos, you wrote that SFFA "proves affirmative action was helping Hispanic students economically." Bleemer is a strong paper, but it studies Prop 209 in California, 1996–98 cohorts. It isn't a study of SFFA. The only post-SFFA outcomes we have are enrollment shares (F12–F14), and those say nothing about wages yet. There's also a live mismatch literature on STEM persistence that points the other way on at least one outcome. Second, conditioning federal student aid on private admissions policy is precisely the sort of coercive funding lever you'd condemn if the administration used it. How is your legacy mandate different in kind from the Columbia settlement terms (F25)?

Ibarra (#74): On the first point I'll narrow it: Bleemer shows that a comparable ban reduced Hispanic applicants' wages, and SFFA's early enrollment effects look like that cascade. "Proves" was too strong; the SFFA wage effect is a prediction, not a finding. On the second point, the difference is that my condition would be general, prospective and enacted by Congress through Spending Clause legislation with clear notice, like Title IX. The settlements were individualized, retroactive and negotiated under a funding freeze a court found unlawful (F26). Form matters.

Thorne: Form matters. But you're still using federal money to override a private institution's admissions judgment. I'd prefer disclosure: publish admit rates for legacy and non-legacy applicants by income, and let reputation do the work.

Ibarra: Disclosure is my fallback. I'll put both forward and let the vote decide.

---

Cross (#64, R) → Castell (#66): Henry, you said hate crimes overall fell in 2024, "which shows how selective the enforcement surge was." That's the wrong comparison. The total fell 1.5%, but anti-Jewish incidents rose 5.8% to a record 1,938 (F22–F23). Jews are about 2% of the population and were the target of about 16% of all incidents. Concentrating enforcement on the category that is rising and over-represented isn't "selective" in the pejorative sense. It's triage.

Castell (#66): I'll concede that the aggregate decline doesn't show selectivity, and I withdraw that inference. The brief also notes that ±2% moves are within reporting noise (Contested §5). My claim stands on narrower ground. Judge Burroughs found that agencies didn't evaluate whether particular grantees engaged in antisemitism before cutting funds (F26). And some settlement payments went to general revenue and workforce programs rather than to affected students (F25). You can think antisemitism enforcement is urgent, as I do, and still think this enforcement was lawless.

Cross: On process we mostly agree. Where I part company: the old OCR process took years while encampments ran for weeks. "Expedited but lawful" needs a number.

Castell: 90 days from complaint to a finding, with interim measures available on a showing of ongoing harm. That's my number.

---

Rosenthal (#35, C-L) → Cross (#64): Nathaniel, you said the "practical upshot" of Mahmoud is that parents may opt out of "any lesson that burdens their religious beliefs." That isn't the holding. The Court triggered strict scrutiny because the storybooks carried "a normative message on the subjects of sex and gender, beyond mere exposure to other ideas," and so substantially interfered with religious upbringing (F28). Mere exposure to ideas a family rejects does not, under Mahmoud, trigger the right. You even conceded you don't want an evolution opt-out, and your opening claim implies one.

Cross (#64): Accepted. That line was an overstatement. The holding is: substantial interference through normative instruction on religiously sensitive topics, especially for young children. Exposure alone is insufficient. My policy (advance notice plus opt-out for sexuality and gender-identity instruction) sits inside that holding. And the Court's reasoning protects a Muslim family objecting to a Christian-nationalist civics module just as much.

Kaplan (#55, bioethics, C-L): Then the fight is over scope. Where does "normative message" end? A health class stating that some families have two mothers is descriptive. Will districts opt out of that?

Cross: Some will over-read it. That's why notice-and-opt-out is better than curriculum removal, which is what several states are doing. Opt-out leaves the content in for everyone else.

STAGE 3 — CROSS-EXAMINATION (4 of 4): Lindqvist; the Skrmetti/B.P.J. exchange; a campus coda

---

Kaplan (#55, bioethics, C-L) → Lindqvist (#37): Victor, you wrote that in Murthy "the Court found that the Biden White House … coerced platforms." It found nothing of the kind. It held the plaintiffs lacked standing and never reached coercion. Only the three dissenters would have (F31). The Fifth Circuit found coercion; the Supreme Court reversed on standing. You're citing a dissent as a holding. Second, "Kimmel was suspended within hours" of Carr's remarks. The PBS account in F32 places Carr's remarks before the suspension. I don't see "within hours" there.

Lindqvist (#37): On Murthy, you're right and I retract: the Court decided standing only. The coercion finding was the Fifth Circuit's, and the dissent's. My substantive point survives on Vullo, which was unanimous that coercing a private party to punish speech violates the First Amendment, and on the underlying record, which the dissent described at length. On Kimmel, I'll restate as "shortly after." The sequence is documented; the hour count isn't in our source.

Kaplan: Then you accept that Murthy is not evidence that jawboning happened, only that it's hard to litigate?

Lindqvist: Yes. It's evidence that the remedy is broken, which is why I want a statute with disclosure. Disclosure turns "we can't prove coercion" into "here are the emails."

Castell (#66, C) → Lindqvist: And RSF? You cite the drop from 57th to 64th (F33). RSF's own reasons are presidential rhetoric, one journalist's deportation, and USAGM cuts. The last is a budget decision about state broadcasters, not a restriction on private press. Is the index measuring speech freedom or policy preferences?

Lindqvist: Mixed. I'd weight the licensing threat and the deportation of a working journalist more than VOA staffing. I'll call the RSF figure "an indicator RSF attributes chiefly to executive hostility" and not rest my case on it.

---

**Sutter (#78, C-R) ↔ Kaplan (#55) on Skrmetti and B.P.J.**

Kaplan: For the record, since these cases get garbled: in Skrmetti the Court held transgender people are not a protected class, and it greenlit legislatures to regulate trans lives on rational-basis review.

Sutter: That's not what it held. The majority said SB1 classifies by age and medical use, not sex, and does not classify by transgender status, so it didn't need to decide whether transgender status is a quasi-suspect class (F29). The question is still open in the Court's jurisprudence. Only concurrences went further.

Kaplan: Fair. I overstated it. But then you should correct your side's gloss, which I hear constantly: "after Skrmetti, rational basis applies to any law involving transgender people."

Sutter: Agreed. B.P.J. disproves it. There the Court applied intermediate scrutiny to sex-separated sports and upheld the laws under that standard (F30). Sports eligibility is a sex classification, and heightened review applied. The cases are narrower than both camps' rhetoric.

Kaplan: Then here's the substantive disagreement that remains. I think the Court was wrong to treat SB1 as not classifying by sex: a testosterone prescription is permitted for a boy with delayed puberty and banned for a girl transitioning. You think the medical-use framing is right. That's a legal-theory dispute that facts won't settle. Where facts could help is the medical-evidence dispute: systematic reviews in the U.K. and elsewhere have rated the evidence for pediatric blockers as low-certainty, and clinicians dispute what follows from that.

Sutter: Accepted as the honest state of play. I'd add only that "low-certainty evidence" is an argument for caution by someone, and the Court left the choice of that someone to legislatures.

---

Coda — Pembroke (#27, R) → Lindqvist on campus: You lump FIRE's results with federal jawboning. But the bottom five in FIRE (Barnard, Columbia, Indiana, Washington, Northeastern) sank largely over pro-Palestinian protest discipline and administrators' responses (F24). Some of that discipline was the administration demanding enforcement.

Lindqvist: Exactly my point: speech on campus is chilled from both directions, by peers and by the state. FIRE's own takeaway says conservative students' tolerance for violence has eroded too (F24). No bloc's students are clean here.

STAGE 4 — REFEREE FACT-CHECK

Brandt (#57) · Takahashi (#17) · Doyle (#25)

Scale: Supported = source confirms · Contested = credible sources disagree · Unsupported = no adequate source found, or the source doesn't say it (includes claims we could not verify this session) · Wrong = a credible source contradicts it. Predictions and value judgments are not rated. "Fn" refers to the Evidence Brief.

We checked two additional primary sources during this stage: Alexander v. Sandoval, 532 U.S. 275 (2001), https://supreme.justia.com/cases/federal/us/532/275/ ; and Mahmoud v. Taylor opinion text, https://supreme.justia.com/cases/federal/us/606/24-297/ .

# · Speaker · Claim · Rating · Basis
1 · Bell · Black names cut contact by 2.1 pp across 108 large employers · Supported · F7
2 · Bell · Whites get 36% more callbacks; no decline 1990–2015 · Supported · F8
3 · Bell · Anti-discrimination law "has not moved hiring discrimination at all" · Supported (1990–2015 only) · F8; no newer meta-analysis
4 · Bell · EO 14281 targets DI "in all contexts"; four agencies rescinded Title VI effects rules · Supported · F17–F18 (DOJ, Commerce, DHS, ED)
5 · Bell · DI liability "for employers is effectively gone" · Wrong · F19: Title VII §2000e-2(k) is statutory. Retracted/amended by Bell.
6 · Bell · Ricci (2009) shows employers face a DI/disparate-treatment bind · Supported · Ricci v. DeStefano, 557 U.S. 557
7 · Bell · EEOC "no longer bringing" DI cases · Unsupported · Not verified this session
8 · Bell · Private plaintiffs rarely can fund statistical pattern cases · Unsupported · No source offered
9 · Walker · Medians $285,000 vs $44,900 · Supported · F1
10 · Walker · "The gap is widening" · Contested · F2: dollar gap up ($190,100→$240,100), ratio down (7.8×→6.3×)
11 · Walker · Black household income "fell 3.3% in its latest release" · Wrong · F3: 2025 release shows +4.8%. Retracted by Walker.
12 · Walker · Black maternal mortality 44.8, >3× white · Supported · F6 (44.8/14.2 = 3.2×)
13 · Walker · Black–white life-expectancy gap 4.1 years · Supported · F5 (78.9 − 74.8)
14 · Walker · Connecticut enacted baby bonds in 2021 · Unsupported · Not verified this session
15 · Walker · 2019–22 Black wealth gain mostly home appreciation · Unsupported · Not verified against SCF asset tables
16 · Walker · College-educated Black women have higher maternal mortality than white women without degrees · Unsupported · Not in brief; not verified this session
17 · Walker · Hispanic life-expectancy advantage partly reflects immigrant selection · Unsupported · Plausible, literature not fetched
18 · Walker · Low-gap tracts require low white bias and high father presence · Supported · F9
19 · Johnson · "The Court overruled Gingles" · Wrong · F20: "need only update the framework." Amended to "functionally hollowed."
20 · Johnson · New burdens: partisan-unexplained polarization; "strong inference" of intent · Supported · F20
21 · Johnson · Tennessee split Memphis within 8 days; Louisiana suspended its primary · Supported · F21 (advocacy source; Apr 29 → May 7)
22 · Johnson · ~90% of Black voters vote one way · Unsupported · Not sourced here
23 · Ibarra · Prop 209 wage losses driven by Hispanic applicants · Supported · F11
24 · Ibarra · Harvard Hispanic share 16%→11%; Ivy-Plus Hispanic −15% · Supported · F12, F14
25 · Ibarra · SFFA "proves" AA was helping Hispanic students economically · Unsupported · No post-SFFA wage data exist. Amended.
26 · Ibarra · Most low-income high scorers are white or Asian · Unsupported · Not verified this session
27 · Castell · Anti-Jewish incidents 1,938, highest since 1991, ~70% of religion-based · Supported · F23
28 · Castell · Columbia/Barnard at the bottom of FIRE · Supported · F24
29 · Castell · Court found agencies didn't weigh grantee-specific antisemitism · Supported · F26
30 · Castell · Settlement money went to government/third parties, not victims · Supported · F25
31 · Castell · Aggregate hate-crime decline "shows how selective enforcement was" · Unsupported · −1.5% is within noise; anti-Jewish rose (F22–F23). Withdrawn.
32 · Castell · Title VI requires findings, hearing, report to Congress before cutoff · Supported · 42 U.S.C. §2000d-1
33 · Whitfield · Sandoval: Title VI itself reaches only intentional discrimination · Supported · Sandoval, 532 U.S. at 280–81: Title VI "itself directly reach[es] only instances of intentional discrimination." Whitfield's partial concession to Rosenthal was unnecessary; see #50.
34 · Whitfield · Callais kept Gingles · Supported · F20
35 · Whitfield · 83% of flagships saw diversity rise · Supported · F14 (advocacy report, secondary coverage)
36 · Whitfield · Harvard Asian American share 29.9%→41% · Supported (magnitude uncertain) · F12; 8% unreported
37 · Whitfield · SFFA "ended a penalty" on Asian applicants · Contested · Share rise is consistent, but non-reporting and the lack of admit-rate data prevent confirmation
38 · Whitfield · DOJ "simply brought its rules into line with the statute" · Contested · Sandoval assumed regulations valid; question never decided
39 · Pembroke · No Black–white gap among women; gap narrows with father presence · Supported (selectively quoted) · F9 also names white racial bias
40 · Pembroke · Parental education and wealth explain little · Supported · F9
41 · Pembroke · Top quintile ≈ half of lost contacts; 23 firms identified · Supported · F7
42 · Pembroke · Hiring discrimination "has fallen sharply" · Wrong · F8: no change 1990–2015. Retracted.
43 · Pembroke · Discrimination "confined to a small number of bad actors" · Wrong · F7: roughly half of lost contacts come from outside the top quintile
44 · Pembroke · 492-study meta-analysis finds no causal link from implicit-bias change to behavior · Supported · F10
45 · Pembroke · DEI "doesn't change behavior anyway" · Contested · F10 covers implicit-bias interventions, not all DEI measures
46 · Pembroke · 52% of workers say DEI is good; 50% disapprove race in admissions · Supported · F15
47 · Pembroke · Family structure/skills/geography are the main causes · Contested · Contested §1; F9 cuts both ways
48 · Pembroke · FIRE's bottom five sank "largely" over pro-Palestinian protest discipline · Unsupported · Partly consistent for Columbia/Barnard (F24); not shown for IU, UW, Northeastern
49 · Rosenthal · Sandoval decided the private-right question and assumed the regulations' validity · Supported · 532 U.S. at 281
50 · Rosenthal · Sandoval merely "assumed" that Title VI reaches only intent · Wrong · Sandoval stated it as "beyond dispute" (relying on Bakke/*Guardians*). Rosenthal to amend.
51 · Rosenthal · The cascade raised URM enrollment at less selective UCs yet lowered completion and wages · Supported · F11
52 · Rosenthal · 8% of Harvard's 2029 class unreported · Supported · F12
53 · Rosenthal · Mahmoud requires more than "mere exposure" · Supported · F28
54 · Blackwood · AIAN life expectancy 71.1, lowest group · Supported · F5
55 · Blackwood · §2 has been Native voters' main tool for local representation · Unsupported · Not verified this session
56 · Lindqvist · Murthy found Biden officials coerced platforms · Wrong · F31: standing only. Retracted.
57 · Lindqvist · Carr's "easy way or the hard way" remark preceded the suspension · Supported · F32
58 · Lindqvist · Kimmel suspended "within hours" · Unsupported · Source doesn't give timing. Amended to "shortly after."
59 · Lindqvist · U.S. fell to 64th on RSF · Supported · F33
60 · Lindqvist · FIRE: 65% F, 41% self-censor, 15% accept violence · Supported · F24
61 · Lindqvist · Vullo unanimously barred coercion of intermediaries · Supported · NRA v. Vullo, 602 U.S. 175 (2024)
62 · Lindqvist · Harvard freeze was viewpoint-based conditioning · Contested · District court so held (F26); on appeal
63 · Lindqvist · Conservative students' rejection of violence has eroded · Supported · F24
64 · Castell · RSF's reasons are presidential attacks, Guevara deportation, USAGM cuts · Supported · F33
65 · Cross · 303 Creative holding · Supported · F27
66 · Cross · Mahmoud lets parents opt out of "any lesson" burdening belief · Wrong · F28. Retracted.
67 · Cross · Mahmoud plaintiffs were Muslim, Catholic, and Orthodox families · Supported · Opinion text: Mahmoud/Barakat (Muslim), Romans (Catholic/Ukrainian Orthodox), Persaks (Catholic)
68 · Cross · Anti-Jewish +5.8%; Jews ~2% of population, ~16% of incidents · Supported · F23
69 · Cross · Several states are removing curriculum rather than offering opt-outs · Unsupported · Not verified this session
70 · Kaplan · Skrmetti held transgender people are not a protected class · Wrong · F29: Court didn't reach the question. Retracted.
71 · Kaplan · U.K. systematic reviews rated the pediatric-blocker evidence low-certainty · Unsupported · Not verified this session (widely reported; we didn't fetch)
72 · Sutter · Title VII DI is statutory (1991 Act) · Supported · F19
73 · Sutter · DOJ rescission skipped notice-and-comment · Supported · F17
74 · Sutter · Skrmetti left quasi-suspect status undecided · Supported · F29
75 · Sutter · B.P.J. applied intermediate scrutiny · Supported · F30
76 · Thorne · Bleemer studies Prop 209 cohorts, not SFFA · Supported · F11
77 · Thorne · Mismatch literature points the other way on STEM persistence · Contested · Contested §3; Bleemer finds no STEM-persistence effect

Tally (77 claims): Supported 47 · Contested 8 · Unsupported 14 · Wrong 8.

Required corrections on the record:

• Bell (#5), Walker (#11), Johnson (#19), Pembroke (#42), Lindqvist (#56), Cross (#66), Kaplan (#70): already retracted or amended in Stage 3. The panel confirms those amendments are adequate.
• Pembroke (#43): "confined to a small number of bad actors" was not formally retracted. Please amend.
• Rosenthal (#50): please amend.
• Whitfield (#33): the panel notes that your original Sandoval statement was accurate. Rosenthal's correction was itself wrong on that point, though his point that the regulations' validity was only assumed is right (#49). Row #38 stays Contested.

---

Pembroke (#27): Amended. "Confined to a few bad actors" is withdrawn. The accurate statement is that discrimination is concentrated but widespread: about half the harm comes from a fifth of firms, and half from everyone else. My policy preference for targeted enforcement is now a preference about cost-effectiveness, not a claim that the problem is small.

Rosenthal (#35): Amended, with thanks. Sandoval did state that §601 reaches only intentional discrimination, as "beyond dispute," relying on Bakke and Guardians. What it assumed was the validity of the §602 effects regulations. My substantive point survives in narrower form: whether agencies may reach effects under §602 was never decided, so rescinding those rules is a policy choice, not a legal compulsion.

Whitfield (#21): I'll take the correction in my favor and accept Isaac's narrower point. Row #38 fairly stays Contested.

Referee erratum (Brandt/Takahashi): we miscounted our own tally in Stage 4. Recounting row by row, the correct figures are Supported 46 · Contested 7 · Unsupported 15 · Wrong 9 (77 claims). Row ratings are unchanged. We apologize; the tally is the one number that must be right.

---

STAGE 5 — STEELMAN

1. Left steelmans the colorblind/equal-treatment Right — by Bell (#22, L)

The strongest case isn't "racism is over." It's this. Any state machinery that sorts individuals by race teaches citizens that race is a legitimate basis for allocation, and that lesson corrodes the very norm civil rights law fought to establish. Race categories are crude: they lump a Nigerian-American surgeon's child with a Mississippi Delta sharecropper's grandchild, and Hmong refugees with Taiwanese engineers. Preferences also create a cost the beneficiaries don't choose, the "stigma of the asterisk." Disparate-impact liability, taken seriously, pressures institutions to manage numbers rather than treat people fairly, and can penalize valid, predictive standards. And it passes a policy decision (which disparities are tolerable?) to agencies and courts that no voter elected. Better to hit intentional discrimination hard, as audit studies let us do (F7), and to attack disadvantage through class-based means that don't classify by race. The public largely agrees (F15). Finally, SFFA's early numbers don't show a collapse in overall minority access. They show a reshuffling toward flagships (F14), which is arguably where most social mobility happens anyway.

Whitfield (#21): Accepted as fair, with one correction. You've presented the colorblind view as mostly consequentialist. The core is principled: equal protection belongs to persons, not groups, and it would hold even if preferences "worked." Otherwise, that's the argument.

2. Right steelmans the civil-rights Left — by Pembroke (#27, R)

The strongest case: discrimination is not a relic, it is measured. Randomized audit studies at the largest U.S. employers find a persistent penalty (F7), and meta-analysis finds no improvement over a quarter century (F8). Formally race-neutral rules can carry forward past exclusion: wealth compounds, so a gap created by redlining and exclusion from the GI Bill era reproduces itself without anyone needing to intend it. That's why a law focused only on intent under-enforces. Intent is hidden, while effects can be measured. Chetty's tract data show that where white racial bias is low (and fathers are present), Black boys' outcomes converge (F9). So bias is an environmental variable that policy can affect. On voting, where elections are racially polarized, a legislature can draw away minority influence using party as a proxy, and requiring proof of "present-day intentional discrimination" (F20) lets that proxy launder the outcome. Race-conscious remedies are then not "discrimination" but the narrowest tool for undoing measured, ongoing exclusion.

Walker (#65): Accepted as fair. One addition: the case is also about dignity and health, not only income. A Black maternal mortality rate three times the white rate (F6) is the strongest fact in our brief, and it doesn't disappear with income controls. Otherwise, you've stated it better than some of our own do.

3. Center steelmans religious conservatives — by Castell (#66, C)

The strongest case: pluralism requires that the state not use its schools and licensing power to demand that dissenters affirm contested moral views. Tolerance means letting people disagree while treating them equally in public life, and it runs both ways. The Mahmoud families were Muslim, Catholic and Orthodox immigrants in the country's most religiously diverse county (row #67), which is exactly the minority the old liberal tradition set out to protect. Opt-outs are the least restrictive accommodation: the curriculum stays, the objectors step out. 303 Creative protected expressive work, not a right to refuse service generally. The same rule protects a progressive designer from being compelled to make a website for a cause she abhors.

Cross (#64): Accepted as fair. One nuance: "least restrictive" matters doubly because the alternative some states chose, removing content for everyone, is more restrictive, and religious conservatives should favor opt-outs over bans. I say so on the record.

4. Religious conservative steelmans LGBTQ-inclusion advocates — by Cross (#64, R)

The strongest case: public schools educate every child, including the children of same-sex couples and children who are themselves gay or gender-nonconforming. A storybook showing their family isn't indoctrination; it's recognition that they exist. Broad opt-outs signal to those children that their existence is a topic requiring parental permission. Administrative opt-out regimes also fall hardest on teachers and on small districts. On medicine, families and physicians, not legislatures, are the normal locus of pediatric decisions, and categorical bans override individualized judgment in a field where evidence is uncertain in both directions.

Kaplan (#55): Accepted as fair, and I'd sign most of it. The correction: on medicine the strongest version concedes the evidence is low-certainty. The argument is then about who decides under uncertainty, not a claim that the evidence clearly favors treatment. That's a stronger position than the one activists usually take.

5. Libertarian steelmans the government's side of the jawboning question — by Lindqvist (#37, Lib)

The strongest case: governments must be able to speak, including to platforms and broadcasters. They need to flag foreign influence operations, terrorist recruitment, child-exploitation material, imminent threats and public-health information during emergencies. The line between persuasion (lawful) and coercion (unlawful) is fact-bound, so courts should draw it case by case rather than chilling all contact by statute. Murthy's standing holding (F31) reflects a real concern: platforms moderate for their own reasons, and attributing every takedown to government pressure invites endless litigation. A disclosure regime could turn every routine security contact into a political weapon.

Castell (#66): Accepted as fair. I'd add that the institutionalist worry cuts both ways: no rule has meant each administration's jawboning gets litigated only when the other side holds power.

STAGE 6 — REVISED PROPOSALS

Drafted by the moderator's office from the Stage 2–5 record. Sponsors are named. Where no CBO or JCT score exists, we say "no official score." Precedent claims we could not verify this session are marked [unverified].

---

15-P1 — Voting Rights Act §2 Clarification Act (sponsors: Johnson, Doyle-advised)

Mechanism. Amend §2 to say that vote-dilution liability requires (a) a sufficiently large, compact minority population, drawable with traditional race-neutral criteria; (b) racially polarized voting shown statistically, with partisan correlation not a complete defense; and (c) no proof of discriminatory intent. Also add a statutory sunset/review every 10 years and a requirement that remedial districts use the minimum race-consciousness needed.
Cost. No official score; administrative and litigation costs only.
Precedent. Allen v. Milligan (2023) applied a results test to order Alabama's second opportunity district. State VRAs (California 2002, Washington 2018, New York 2022) use results-type standards for local elections [unverified this session].
Key risk. Callais (F20) signals constitutional limits on race-based districting. A statute can override a statutory reading, not a constitutional one, so the Court may strike it under the 14th/15th Amendments. It also entrenches race-conscious line-drawing, which critics call self-perpetuating.

15-P2 — Federal Employer Audit-Testing Program (sponsors: Bell, Pembroke)

Mechanism. EEOC (or DOL-OFCCP for contractors) runs recurring correspondence audits of the ~500 largest employers using the Kline–Rose–Walters design (F7). Firm-level results are published only where statistically robust (false-discovery-rate controls). Flagged firms trigger intentional-discrimination (pattern-or-practice) investigations. No quotas, no disparate-impact theory required.
Cost. No official score. The KRW scale (83,000+ applications, 108 firms) suggests single-digit millions per wave [unverified cost estimate].
Precedent. HUD's national paired-testing Housing Discrimination Studies (1977 onward) [unverified this session]; KRW itself demonstrates feasibility.
Key risk. Names signal class as well as race; fictitious applications burden recruiters; firms may learn to detect tests; a false positive publicly brands a firm.

15-P3 — Speech & Lawful-Enforcement Act (sponsors: Lindqvist, Castell)

Mechanism. (a) Any federal official's request that a platform, broadcaster or licensee remove, demote or punish lawful speech must be logged and disclosed within 72 hours. Narrow sealed carve-outs (CSAM, terrorism, active investigations) are reported to Inspectors General. (b) License renewal, merger review and grant decisions may not turn on the viewpoint of content. (c) Title VI/IX fund terminations must follow 42 U.S.C. §2000d-1 procedures on an expedited 90-day track, with interim relief on a showing of ongoing harm. Settlement proceeds go to affected students and compliance programs, not general revenue.
Cost. No official score; modest administrative cost.
Precedent. NRA v. Vullo (2024) and Murthy (2024) (F31); the Harvard ruling (F26) turned on skipped Title VI procedures; FCC–Kimmel episode (F32).
Key risk. Chills legitimate security communications; carve-outs can be abused; a 90-day clock may still be too slow for acute campus crises, or too fast for due process.

15-P4 — End Legacy & Donor Preferences; Publish Admissions by Income (sponsors: Ibarra; Thorne dissenting on the mandate)

Mechanism. As a prospective Spending Clause condition on Title IV aid, institutions may not give admissions preference for legacy or donor status. All institutions above a selectivity threshold must publish admit rates by income quintile, legacy status and athlete status. (Thorne's alternative, disclosure only, is folded into the vote as a note.)
Cost. No official score; negligible federal cost.
Precedent. California barred legacy/donor preferences at private colleges from 2025 [unverified this session]. The Class Action report argues legacy preferences magnified post-SFFA declines at elites (F14, advocacy source).
Key risk. Federal leverage over private admissions; limited diversity recovery (Ibarra's own concession); schools may shift to less transparent preferences.

15-P5 — Hate-Crime Measurement Integrity (sponsors: Castell, Cross, Blackwood)

Mechanism. Make NIBRS hate-crime reporting (including zero-reports) a condition of Byrne JAG funding. Fund an annual NCVS hate-victimization supplement reporting by bias category (anti-Jewish, anti-Black, anti-Muslim, anti-LGBTQ, anti-AIAN, etc.) with confidence intervals.
Cost. No official score; modest (BJS survey supplement).
Precedent. FBI 2024 coverage reached 95.1% of the population (F22); NCVS already estimates hate victimization.
Key risk. Conditioning grants on reporting burdens small agencies; better data will itself be politicized.

15-P6 — Federal Baby Bonds (sponsor: Walker)

Mechanism. A federally funded account at birth for every child (≈$1,000 seed) with annual income-tiered deposits for low-wealth families, restricted use at 18 (education, home purchase, business).
Cost. No official CBO score found this session; sponsors' estimates [unverified]. Likely tens of billions per year at scale.
Precedent. Connecticut enacted a state baby-bonds program in 2021 [unverified this session]; no mature cohort outcomes yet.
Key risk. Cost; slow payoff (18 years); F9 suggests wealth explains little of the income-mobility gap, so it targets wealth levels, not mobility (Walker's own narrowing).

15-P7 — Curriculum Notice & Opt-Out Standard (sponsors: Cross; Castell co-sponsor)

Mechanism. A model federal standard (guidance plus a safe harbor for districts that adopt it). Advance notice of K–12 instruction centered on sexuality or gender identity, with a written parental opt-out that has no penalty for the student. Content removal is expressly not the remedy.
Cost. No official score; district administrative cost.
Precedent. Mahmoud (F28); Montgomery County allowed opt-outs before rescinding them in 2023, which gave rise to the case.
Key risk. Scope creep beyond Mahmoud's "normative message" line; stigma for LGBTQ students and families (Cross's own steelman); administrative burden.

15-P8 — Campus Expression Transparency & Neutrality (sponsors: Lindqvist, Pembroke)

Mechanism. Institutions receiving federal research funds must publish viewpoint-neutral time/place/manner rules and annual enforcement statistics (charges and outcomes by event type), and must adopt a stated policy on institutional neutrality (adopting the Kalven model is encouraged but not required). ED funds a probability-sample campus expression survey to complement opt-in instruments like FIRE's (F24).
Cost. No official score; modest.
Precedent. University of Chicago Kalven Report (1967); FIRE's top-ranked schools (F24).
Key risk. Federal intrusion into academic governance; "neutrality" can be pretextual; enforcement statistics can be gamed.

STAGE 7 — RECORDED VOTE

Provisional vote — superseded by the blind independent ballot in thread "Independent Ballot."

Blocs: Left 34 · Center 33 · Right 32 (moderator #100 abstains). Bridge rule: CONSENSUS only if ≥60% in each bloc (Left ≥21, Center ≥20, Right ≥20).

ID · Proposal · Left yes · Center yes · Right yes · Consensus?
15-P1 · VRA §2 Clarification Act · 32/34 (94%) · 21/33 (64%) · 7/32 (22%) · No
15-P2 · Federal employer audit-testing program · 31/34 (91%) · 27/33 (82%) · 20/32 (63%) · YES
15-P3 · Speech & Lawful-Enforcement Act (anti-jawboning + Title VI due process) · 28/34 (82%) · 28/33 (85%) · 21/32 (66%) · YES
15-P4 · End legacy/donor preferences; income disclosure · 29/34 (85%) · 24/33 (73%) · 18/32 (56%) · No (Right short by 2)
15-P5 · Hate-crime measurement integrity · 30/34 (88%) · 30/33 (91%) · 22/32 (69%) · YES
15-P6 · Federal baby bonds · 29/34 (85%) · 15/33 (45%) · 5/32 (16%) · No
15-P7 · Curriculum notice & opt-out standard · 12/34 (35%) · 22/33 (67%) · 29/32 (91%) · No
15-P8 · Campus expression transparency & neutrality · 17/34 (50%) · 26/33 (79%) · 27/32 (84%) · No (Left short by 4)

Result: 3 of 8 proposals reach cross-bloc consensus (P2, P3, P5).

---

Vote explanations

Pembroke (#27, R): Yes on P2, No on P1, Yes on P4. P2 is how you act on the fact the referees made me concede: discrimination is concentrated but widespread (F7), and none of this depends on disparate impact. I split from most of my bloc on P4 because legacy preference is inherited privilege by another name. What would change my mind on P2: if the first two waves flag firms whose results don't replicate in a second wave, the name-and-publish component should be dropped.

Bell (#22, L): Yes on P1, P2, P3, P5; No on P8. P3 binds my side when it holds power too. That's the price of a credible rule, and I'll pay it. I vote no on P8 because "institutional neutrality" imposed through federal research money is itself a funding condition on speech, the thing P3 forbids. What would change my mind on P8: drop the federal-funds condition and make it a disclosure-only standard, and I vote yes.

Lindqvist (#37, Lib): Yes on P3, P5, P8; No on P2, P4. I'm in the minority of my bloc on P2. Government-run deception at scale, with fake résumés sent under state authority and firms branded on statistical inference, makes me uneasy even though the goal is enforcing a law I support. What would change my mind: a design where the audit results trigger investigation but are not published unless an investigation confirms the finding.

Walker (#65, L): Yes on P6, P1, P2, P3, P4, P5; No on P7. On P6 I accept the referees' narrowing: it's a wealth remedy, not a mobility remedy. I vote no on P7 because a federal safe harbor will be read as endorsing opt-outs well beyond Mahmoud's line. What would change my mind on P7: pilot data showing opt-out rates under 5% and no deterioration in LGBTQ students' school-climate measures. That is Cross's own prediction, and I'd accept being proven wrong by it.

Whitfield (#21, R): No on P1, Yes on P2, P3, P5, P7, P8; No on P4. P1 invites the same race-based districting the Fourteenth Amendment forbids, and after Callais it's likely unconstitutional anyway. I vote no on P4 not because legacies are defensible (they aren't) but because Congress shouldn't dictate private admissions through Title IV. What would change my mind on P4: a disclosure-only version (Thorne's alternative); I would vote yes on that.

Kaplan (#55, C-L): Yes on P3, P5, P2; No on P7. I respect Cross's steelman, and the opt-out line is better than content bans. But the federal safe harbor rewards districts for adopting it and will spread faster than Mahmoud requires. What would change my mind: limiting the standard to the elementary grades, which is the setting Mahmoud addressed, rather than all of K–12.

Cross (#64, R): Yes on P3, P5, P7, P8; No on P1, P6. On P3, religious minorities have been the targets of state pressure often enough that I want the disclosure rule on the books before my side needs it. What would change my mind on P1: a design that covers only jurisdictions with recent adjudicated intentional violations, closer to a Shelby-compatible coverage formula.

STAGE 8 — VERDICT

Prof. Adelaide Wainwright (#100), moderator

Established (the evidence is strong enough that no bloc should dispute it)

1. Large racial gaps persist, and some are narrowing. Median white family wealth is about 6.3× Black family wealth (2022). The ratio narrowed from 7.8× while the dollar gap widened (F1–F2). In 2025 Black household income grew fastest of any group (+4.8%, F3). Black maternal mortality is about 3× the white rate (F6). AIAN (71.1) and Black (74.8) life expectancy trail white (78.9). Hispanic (81.8) and Asian (85.8) life expectancy exceed it (F5), which a single-cause account of disparity has to explain.
2. Hiring discrimination against Black applicants is real, measurable, and didn't decline over 1990–2015 (F8). The largest modern audit shows it concentrated in identifiable firms and industries but present broadly (F7). Both "it's a few bad apples" and "it's everywhere equally" failed the fact-check.
3. Changing implicit bias hasn't been shown to change behavior (F10). Training-centric DEI rests on weak evidence. That finding doesn't reach enforcement against discrimination.
4. **After SFFA, Black and Hispanic shares fell sharply at the most selective colleges and rose at many flagships** (F12–F14). Asian American shares rose at Harvard, with an 8% non-reporting caveat.
5. The law moved fast in 2025–26: Title VI disparate-impact regulations were rescinded at four agencies (F17–F18), while Title VII's statutory disparate impact survives (F19). Callais modified Gingles to require partisan-independent polarization and a "strong inference" of intentional discrimination (F20). Skrmetti, B.P.J., Mahmoud and 303 Creative are each narrower than both camps' rhetoric (four of the nine "Wrong" ratings came from overstating Supreme Court holdings).
6. Anti-Jewish hate crimes reached a record in FBI data (1,938 in 2024) while total incidents were flat (F22–F23).
7. Government pressure on speech intermediaries has occurred under both recent administrations, per the Murthy record and dissent and the 2025 FCC–ABC episode. The Supreme Court has not adjudicated either as coercion on the merits (F31–F32).

Contested (evidence exists; credible readings diverge)

• Why the gaps persist: structural/discrimination accounts vs. family-structure/skills accounts. Chetty et al. (F9) name both white racial bias and father presence and can't separate them. The panel's most important joint finding is that each side had been citing half of that sentence.
• Whether SFFA will reduce URM completion and earnings nationally (Bleemer's Prop 209 evidence says yes; the flagship cascade and today's aid landscape may differ).
• Whether rescinding Title VI effects regulations is legally compelled or a policy choice. Sandoval stated that Title VI itself reaches only intent, but it only assumed the regulations were valid.
• Whether the Harvard freeze was unconstitutional viewpoint retaliation (district court: yes; First Circuit pending).
• How much FIRE's opt-in data measure a "chilling" climate versus ordinary prudence; how much the RSF decline reflects legal constraint versus rhetoric and budgets.

Unknown

• Callais's seat effects by 2028 (Johnson's four-district prediction vs. Whitfield's crossover counter-prediction are both on the record and testable).
• Post-*SFFA* graduation and wage effects (the first cohorts graduate in 2028).
• Whether settlements improved Jewish students' campus climate more than litigation did.
• Opt-out uptake and LGBTQ-student climate effects under Mahmoud-style regimes.
• The 2025 SCF (not yet published).

Research agenda

1. Repeated, pre-registered firm-level audit waves (KRW design) to test Pembroke's persistence prediction and Bell's no-improvement prediction.
2. Administrative-data tracking (NSC/IRS linkage) of the post-*SFFA* cascade: completion and earnings by institution tier.
3. Probability-sample measures of campus expression and of antisemitism and Islamophobia climate, to complement advocacy-run panels.
4. An NCVS hate-victimization series by bias category with confidence intervals.
5. District-level opt-out and school-climate data under Mahmoud.
6. Tribal-area representation outcomes after Callais (Blackwood/Whitfield agreement).

Consensus reforms (provisional; the blind ballot governs)

• 15-P2 Federal employer audit testing: enforce against intentional discrimination with randomized evidence. Right 63%, a narrow pass. Lindqvist's "no publication before confirmation" amendment is the obvious way to widen it.
• 15-P3 Speech & Lawful-Enforcement Act: disclosure of jawboning, no viewpoint-based licensing, and Title VI due process on a 90-day track. The strongest cross-bloc bridge in the thread, because every bloc has recently been on the receiving end.
• 15-P5 Hate-crime measurement integrity.
• Near misses: P4 (legacy ban: Right 56%; a disclosure-only version would likely pass per Whitfield and Thorne); P8 (campus neutrality: Left 50%; a version without the funding condition would likely pass per Bell). Genuinely divided: P1 (§2 amendment), P6 (baby bonds), P7 (opt-out standard).

Where Symposium 1 was wrong or too glib on this domain

1. It left civil rights and free expression out almost entirely. Neither appeared among its 10 consensus items or its "still disagree" list. Its only exchange (Bell vs. Ashford on "states' rights") was a slogan against a slogan.
2. Its master diagnosis, "America can't decide or build," doesn't fit this domain. In 2025–26 civil-rights law changed extremely fast through executive orders, no-comment rule rescissions, settlements under funding freezes, and 6–3 decisions. The failure here is not paralysis but volatility without durable legislative settlement. That's why P3's "bind yourself before you lose power" logic passed.
3. Johnson's redistricting-commission proposal ignored the VRA. After Callais, the binding redistricting question for minority voters is §2, not commissions.
4. Stroud's "two-parent homes do better on nearly every measure" and Pembroke's "the problems are cultural" were asserted, not tested. Chetty's data give family structure real weight and give white racial bias equal billing. Symposium 1 let one half stand unchallenged.
5. Symposium 1 cited no data on discrimination at all. This thread's central empirical finding, a persistent, measurable, concentrated-but-broad hiring penalty, was absent. So was the finding that the most popular corporate remedy (bias training) lacks behavioral evidence. Both sides' priors needed those two facts.

No bloc wins this domain by fiat. The evidence supports the civil-rights left on the existence and persistence of discrimination. It supports the equal-treatment right on the weakness of training-based DEI and on the text-bound narrowness of the recent cases. It supports the libertarians on jawboning being bipartisan. It supports religious conservatives on the modesty of the opt-out remedy relative to content bans. And it supports the institutionalist center on process. The three consensus items share one feature: each binds the powerful regardless of who holds power.

— Symposium 2, Thread 15 closed.

Independent Audit

Audit Panel — Prof. Laura Pennington (#97, operations research, C) & Prof. Olivia Brandt (#57, statistics, C). Pennington leads this audit. Brandt refereed this thread, so she recuses from any finding about her own rulings, including the tally erratum. Corrections below are appended to the record; nothing above is deleted. Several fetches were rate-limited. Those pages are listed as "Could not access," and we did not guess at them.

1. Source check (re-fetched 2026-09-24)

# · Claim (thread) · Cited · Verified · Status · URL
1 · Louisiana v. Callais (F20) · Apr 29, 2026; 6–3, Alito (Roberts, Thomas, Gorsuch, Kavanaugh, Barrett); Kagan dissent with Sotomayor and Jackson; "need only update the framework"; four new showings; dissent says "eviscerates" · All confirmed. The Court says it "need only update the framework so it aligns with the statutory text." Illustrative maps must meet "the State's specified political goals." Polarization must be one "that cannot be explained by partisan affiliation." Plaintiffs must show "a strong inference that the State intentionally drew its districts…" Kagan: the decision will "eviscerate the law" · Confirmed · https://www.supremecourt.gov/opinions/25pdf/24-109_21o3.pdf
2 · B.P.J. / Hecox (F30) · June 30, 2026; 6–3, Kavanaugh; biological sex; intermediate scrutiny; "Sotomayor dissented" · Holding, author and scrutiny confirmed; Hecox decided on the merits. Sotomayor's opinion concurs in the judgment in part and dissents in part, joined by Kagan and Jackson; Jackson also wrote separately · Minor discrepancy · https://www.supremecourt.gov/opinions/25pdf/24-43_2b35.pdf
3 · Skrmetti (F29; rows #70, #74) · June 18, 2025; 6–3, Roberts; classifies by age and medical use, not sex, and not transgender status; quasi-suspect question left undecided · Confirmed. Alito joined only in part and would have held transgender status is not quasi-suspect. "Only concurrences went further" is accurate · Confirmed · https://supreme.justia.com/cases/federal/us/605/23-477/
4 · Murthy (F31; row #56) · Standing only; coercion not reached; Alito, Thomas, Gorsuch dissent · Same · Confirmed · https://supreme.justia.com/cases/federal/us/603/23-411/
5 · Mahmoud (F28; rows #53, #66) · June 27, 2025; 6–3, Alito; "normative message… beyond mere exposure" · Same. The relief covers the specific contested texts, not any lesson · Confirmed · https://www.faegredrinker.com/en/insights/publications/2025/6/supreme-court-decides-mahmoud-v-taylor
6 · SCF median wealth (F1–F2) · $285,000 / $44,900 / $61,600; 2019: $218,100 / $28,000 / $41,900; +31 / +60 / +47% · Same (Table 2, 2022 dollars). The ratios (7.8× → 6.3×) and dollar gaps ($190,100 → $240,100) recompute exactly · Confirmed · https://www.federalreserve.gov/publications/files/scf23.pdf
7 · Quillian et al. meta-analysis (F8) · 36% more callbacks; no change 1990–2015 · Same (IPR summary; the PNAS page was rate-limited) · Confirmed · https://www.ipr.northwestern.edu/news/2017/infographic-quillian-hiring-discrimination.html
8 · Kline–Rose–Walters (F7) · 83,000+ applications; 108 firms; 2.1 pp; top quintile ≈ half; 23 firms · 83,000+ confirmed. The OUP page did not return the other figures, and NBER was rate-limited · Could not access (partial) · https://academic.oup.com/qje/article-abstract/137/4/1963/6605934
9 · AP analysis of Black first-year shares (F13) · Harvard 18→11.5; Princeton 9→5; Columbia 20→13; Amherst 11→6; Caltech 5→1.6; Smith 4.6→6.8; 14% of HS graduates · Same · Confirmed · https://www.insidehighered.com/news/quick-takes/2025/10/27/black-student-enrollment-shrinks-selective-institutions
10 · FIRE 2026 headline figures (F24) · 68,510 students; 257 schools; 166 F; 36% shout-down; 15% violence; CMC B−; bottom five · Same. Bottom five: Barnard, Columbia, Indiana, Washington, Northeastern · Confirmed · https://www.fire.org/research-learn/2026-college-free-speech-rankings
11 · FIRE self-censorship (F24; row #60) · 41% with peers, 38% in class · Not on FIRE's page. The page gives 31% with peers and 33% in class, and only for IU. The secondary Substack source was not checked · Not supported by cited source · same as row 10
12 · DOJ Title VI rescission (F17) · Effective Dec 10, 2025; (b)(2), (b)(3), (b)(6), (c)(2); no notice-and-comment; Sandoval; statistics still usable to prove intent · Same. The exception invoked is 5 U.S.C. 553(a)(2), for grants and contracts · Confirmed · https://www.federalregister.gov/documents/2025/12/10/2025-22448/rescinding-portions-of-department-of-justice-title-vi-regulations-to-conform-more-closely-with-the
13 · Life expectancy, 2024 (F5) · 85.8 / 81.8 / 78.9 / 74.8 / 71.1; +0.5 to +1.0 · Same · Confirmed · https://www.cdc.gov/nchs/data/nvsr/nvsr75/nvsr75-05.pdf
14 · Census 2025 income (F3; basis for row #11 Wrong) · Black +4.8% · Rate-limited · Could not access · https://www.census.gov/library/publications/2026/demo/p60-289.html
15 · Harvard class of 2029 (F12) · 11.5 / 11 / 41 / 8% unreported · Rate-limited · Could not access · harvardmagazine.com (URL in F12)
16 · FBI anti-Jewish incidents, via ADL (F23) · 1,938, +5.8%, ~70% of religion-based · Rate-limited. The ~16% share recomputes correctly: 1,938 / 11,679 = 16.6% · Could not access · adl.org (URL in F23)
17 · Maternal mortality, 2024 (F6) · 17.9; Black 44.8; White 14.2 · Rate-limited. The 3.2× ratio recomputes correctly · Could not access · https://www.cdc.gov/nchs/data/hestat/hestat113.htm

Counts: 11 Confirmed · 1 Minor discrepancy · 1 Not supported · 4 Could not access, plus the KRW row, which is only partly accessed.

Supreme Court holdings. Callais, Skrmetti, Mahmoud and Murthy are described accurately in the Evidence Brief and in the fact-check corrections. SFFA is only cited by name and year (2023), with no holding stated, so nothing about it needs correcting. The only inaccuracy we found in any holding description is the B.P.J. dissent label (row 2). All four Wrong ratings on overstated holdings (#19, #56, #66, #70) are correct against the opinions. So is #50 (Sandoval called §601's intent-only reach "beyond dispute").

2. Internal consistency

• Tally (Pennington alone; Brandt recused). Row by row: Wrong 9 (#5, 11, 19, 42, 43, 50, 56, 66, 70); Contested 7 (#10, 37, 38, 45, 47, 62, 77); Unsupported 15 (#7, 8, 14, 15, 16, 17, 22, 25, 26, 31, 48, 55, 58, 69, 71); Supported 46. The erratum's 46 · 7 · 15 · 9 is correct. The original 47 · 8 · 14 · 8 was wrong.
• Vote math. All 24 percentages are correct. Consensus labels are correct: P2 (31 / 27 / 20), P3 (28 / 28 / 21) and P5 (30 / 30 / 22) pass. "Right short by 2" on P4 (18 vs 20) and "Left short by 4" on P8 (17 vs 21) are right.
• Vote explanation contradicts the table. Pembroke says: "I split from most of my bloc on P4." But 18 of 32 Right members (56%) voted yes on P4, so his yes vote is with his bloc's majority. Correction: "I vote with a narrow Right majority on P4, short of the 60% bridge."
• Row #3 (Bell) is rated too generously. Quillian measures a trend: no change over 1990–2015. It does not estimate the effect of the law. "Supported (1990–2015 only)" should read "Supported as a trend; the causal attribution to law is untested."
• Verdict. "Four of the nine Wrongs came from overstating Supreme Court holdings" checks out (#19, #56, #66, #70). No claim rated Wrong reappears as established.
• Roster. All names, numbers, fields and leans match. Whitfield's reference to "Isaac" correctly means Rosenthal (#35).

3. Balance

The Wrong ratings split 5 Left/C-L (Bell, Walker, Johnson, Rosenthal, Kaplan) to 4 Right/Lib (Pembroke ×2, Cross, Lindqvist). Each was justified against a primary source. The panel also ruled against one of its own corrections in the Right's favor (Whitfield #33 vs Rosenthal #50), which counts against any charge of tilt.

The sensitive topics were handled even-handedly:

• Transgender cases. On Skrmetti and B.P.J., both the progressive overstatement ("not a protected class") and the conservative one ("rational basis for everything") were corrected. The medical-evidence dispute was left as an honest question of who decides under uncertainty.
• Religion. The Mahmoud overreach by a religious conservative was corrected. His steelman of LGBTQ-inclusion advocates was accepted by Kaplan.
• Antisemitism and Harvard. Antisemitism enforcement and the Harvard process ruling were kept as separate questions.

Small tilts:

• Row #3 is lenient to the Left.
• Walker was rated on four "not verified this session" items, which reflects scope rather than error.
• The Census page behind Walker's Wrong rating (#11) could not be re-verified by this audit. The rating is plausible, and Walker conceded it, but it is unconfirmed here.

4. Grade: A−

The Supreme Court holdings are accurately stated, the tally erratum and vote math are correct, and the treatment is balanced. Deductions: an unsupported FIRE 41%/38% figure, the B.P.J. dissent label, Pembroke's bloc-majority misstatement, and four sources we could not re-verify.