10 — Work, Wages, Mobility & Family

STAGE 1 — EVIDENCE BRIEF

Referee panel: Prof. Olivia Brandt (#57, statistics, C) · Prof. June Takahashi (#17, econometrics, C) · Domain specialist Prof. Miriam Castellano (#1, labor economics, C-L)

Ground rules from the referees: every figure below was checked against the linked source on or before 2026-09-24. Where we could not verify a number we say so. Where a source is advocacy-affiliated (EPI, AEI, CBPP, Heritage, etc.) we flag it. "Year" means the year of the data, not the publication.

---

A. Wages and the productivity–pay debate

1. EPI's headline gap. Net productivity rose 93.7% from 1979Q4 to 2026Q2; hourly compensation of the "typical worker" (production/nonsupervisory) rose 32.7% — productivity grew ~2.9× as fast. (EPI, data through 2026Q2) — https://www.epi.org/productivity-pay-gap/ [left-leaning, labor-funded institute]
2. EPI on wages alone. Cumulative real median wage growth was 29% from 1979 to 2024 (<0.6%/yr); in tight labor markets 10th-percentile real wages grew 2.7%/yr on average versus −0.6%/yr in other years. (EPI, data 1979–2024) — https://www.epi.org/blog/how-should-we-assess-and-characterize-workers-wage-growth-in-recent-decades/
3. The Winship (AEI) critique. 1973–2022: nonfarm productivity +111%, median hourly compensation +40%, median wages +33%. Winship argues (a) CPI overstates inflation relative to the output deflator used for productivity, (b) benefits matter, (c) net (post-depreciation) productivity is the right benchmark, and (d) under consistent deflators "aggregate worker pay has kept up with productivity." (AEI, May 2024; data through 2022) — https://www.aei.org/wp-content/uploads/2024/05/Understanding-Trends-in-Worker-Pay.pdf [right-of-center institute]
4. The link is not "broken." Stansbury & Summers: 1973–2016, a 1-pp increase in productivity growth is associated with 0.7–1.0 pp higher median and average compensation growth; they find little support for technology-driven decoupling and point to other (institutional) causes of the level gap. (NBER w24165, 2017) — https://www.nber.org/papers/w24165
5. Latest labor-market snapshot. August 2026: payrolls +162,000; unemployment 4.1%; average hourly earnings +3.1% y/y; trailing-12-month average payroll gain only 31,000/month. (BLS Employment Situation, Aug 2026) — https://www.bls.gov/news.release/empsit.nr0.htm

B. Income and wealth distribution

1. CBO long run. 1979–2021: average income before and after transfers and taxes rose for all quintiles, but grew most in the top quintile; Gini coefficients both before and after transfers and taxes increased. (CBO, Sept 2024; data 1979–2021) — https://www.cbo.gov/publication/60342
2. CBO most recent year. 2022→2023: lowest-quintile income +5%; top 1% income −9% (mainly lower capital gains); ~134 million households, ~$20 trillion total income. (CBO, 2026; data 2023) — https://www.cbo.gov/publication/62761
3. Auten–Splinter (2025 update). Top 1% share of pre-tax income 11.2% (1962) → 15.5% (2022); after-tax 8.6% → 10.3%. (Auten & Splinter, 2025 update) — https://davidsplinter.com/AS-Update-2025.pdf
4. Piketty–Saez–Zucman side. Top 1% share of fiscal (tax-return) income 8.3% (1960) → 18% (2019); for 2014, PSZ after-tax top-1% share was 15.7%, AS's estimate 6.6 pp lower. Critiques of AS: allocation of underreported income, allocation of government consumption (39% of bottom-50% income by 2019 in AS vs 7% for top 1%), deficit allocation, and tension with rising wealth concentration. (Washington Center for Equitable Growth, 2024) — https://equitablegrowth.org/new-research-doesnt-overturn-consensus-on-rising-u-s-income-inequality/ ; PSZ reply: http://gabriel-zucman.eu/files/PSZ2024.pdf [Equitable Growth is left-of-center]
5. Record-high median income, flat SPM. 2025 median household income $87,460 (+2.6%, record); official poverty 10.2% (record low); SPM 13.1%, not statistically different from 2024; official child poverty 13.4% (record low). Social Security moved 28.8 million people out of SPM poverty. (Census P60-290, Sept 2026; data 2025) — https://www.census.gov/library/publications/2026/demo/p60-290.html ; https://stateline.org/2026/09/15/census-data-shows-us-poverty-rate-drops-to-lowest-on-record-but-experts-are-cautious/
6. 2025 reconciliation law (P.L. 119-21) distribution. CBO: over 2026–2034 the lowest decile loses ~$1,200/yr (−3.1% of income), mainly from Medicaid/SNAP reductions; the top decile gains ~$13,600/yr (+2.7%), mainly from tax cuts. (CBO, Aug 2025) — https://www.cbo.gov/publication/61367
7. Wealth concentration. Referees' note: we attempted to pull the latest Federal Reserve Distributional Financial Accounts shares (updated Sept 18, 2026) but could not retrieve the table. No wealth-share figure is entered into evidence; any professor citing one is on their own. — https://www.federalreserve.gov/releases/z1/dataviz/dfa/distribute/table/

C. Mobility and place

1. Fading American Dream. Share of children earning more than their parents (inflation-adjusted, at ~age 30): ~90% for the 1940 cohort → ~50% for the 1980s cohort. Counterfactuals: 1940s–50s GDP growth with today's distribution → 62%; today's GDP with 1940s-style distribution → 80%. Most of the decline is attributed to distribution, not growth. (Chetty et al., Science 2017) — https://opportunityinsights.org/paper/the-fading-american-dream/
2. Moving to Opportunity. Moving a child from public housing to a low-poverty area at age 8 raises lifetime earnings by ~$302,000 (~$99,000 present value); effects for children who moved at 13–18 were zero or slightly negative; little effect on adults' earnings. (Chetty, Hendren & Katz, AER 2016) — https://opportunityinsights.org/paper/newmto/
3. Opportunity Atlas. For children from low-income families, the SD of household income at 35 across tracts within the same county is ~$5,000; observational tract estimates are highly predictive of causal effects; job growth correlates only weakly with upward mobility. (Chetty et al., 2018/2025) — https://opportunityinsights.org/paper/the-opportunity-atlas/

D. Trade and tariffs

1. China shock persistence. Greater import exposure reduced the manufacturing employment-to-population ratio by 1.54 pp in exposed commuting zones; 86% of the net job loss showed up as lower employment rates rather than out-migration; transfers rose modestly and mostly as Social Security and Medicare. (Autor, Dorn & Hanson, NBER w29401, 2021; data 2000–2019) — https://www.nber.org/papers/w29401
2. TAA. ADH describe TAA's effect on per-capita income in trade-exposed areas as "vanishingly small" (Brookings summary). TAA's authorizing provisions terminated July 1, 2022; no new certifications since. (DOL ETA, accessed 2026) — https://www.brookings.edu/articles/on-the-persistence-of-the-china-shock/ ; https://www.dol.gov/agencies/eta/tradeact
3. 2025–26 tariffs. Average statutory tariff rate 11.0% (Aug 24, 2026), after Section 122 tariffs expired July 24 and were replaced under Section 301; long-run consumer price effect ~0.7%; ~$1,100 per household per year; ~$1.9 trillion 10-year revenue net of GDP feedback. (Yale Budget Lab, Aug 2026) — https://budgetlab.yale.edu/research/state-us-tariffs
4. 2018 tariffs and manufacturing jobs. Moving an industry from the 25th to 75th percentile of tariff exposure → −1.4% manufacturing employment: import protection +0.3%, input costs −1.1%, retaliation −0.7%. (Flaaen & Pierce, Fed FEDS 2019-086) — https://www.federalreserve.gov/econres/feds/files/2019086pap.pdf
5. Manufacturing now. Manufacturing employment +16,000 in August 2026 and +58,000 since a recent low in December 2025. (BLS, Aug 2026) — https://www.bls.gov/news.release/empsit.nr0.htm

E. Labor institutions

1. Unions. 2025 membership rate 10.0% (14.7 million); public sector 32.9%, private 5.9%; 1983: 20.1%. Median weekly earnings $1,404 union vs $1,174 nonunion (unadjusted for occupation, industry, demographics). (BLS, Feb 2026; data 2025) — https://www.bls.gov/news.release/union2.nr0.htm
2. Minimum wage — bunching evidence. 138 state increases 1979–2016: the number of low-wage jobs "remained essentially unchanged over five years"; modest spillovers; some evidence of job loss in tradable sectors. (Cengiz, Dube, Lindner & Zipperer, QJE 2019) — https://www.nber.org/papers/w25434
3. CBO on $15. 2019: median estimate 1.3 million jobs lost, 1.3 million out of poverty, 17 million raises. 2021 (Raise the Wage Act): 1.4 million jobs lost (0.9% of workforce), 0.9 million out of poverty, 17 million raised plus ~10 million near the threshold. CBO's median short-run elasticity for directly affected workers: −0.25. (CBO 2019, 2021) — https://www.cbo.gov/system/files/2021-02/56975-Minimum-Wage.pdf ; https://www.npr.org/2021/02/08/965483266/-15-minimum-wage-would-reduce-poverty-but-cost-jobs-cbo-says ; https://www.cbo.gov/publication/57049
4. Seattle. UW team (Jardim et al.): the step to $13 reduced hours in low-wage jobs by 6–7% while hourly wages rose ~3%, a net loss of ~$74/month in low-wage-job earnings (revised 2018 version); effects concentrated on less-experienced workers; minimal restaurant-only effect. Critics (EPI; Reich et al. at Berkeley) cite exclusion of multi-site employers (~40% of workforce) and Seattle's boom; Berkeley restaurant study found no negative employment effect. — https://www.nber.org/papers/w23532 ; https://www.aeaweb.org/articles?id=10.1257%2Fpol.20180578 ; https://www.epi.org/publication/the-high-road-seattle-labor-market-and-the-effects-of-the-minimum-wage-increase-data-limitations-and-methodological-problems-bias-new-analysis-of-seattles-minimum-wage-incr/
5. Licensing. 24.0% of employed people hold an active certification or license; 21.6% hold a government license. (BLS CPS, 2025) — https://www.bls.gov/cps/cpsaat49.htm

F. Work and participation

1. Prime-age participation. May 2025: prime-age men 89.2% (highest since fall 2009), women 77.7%. (Hamilton Project, July 2025) — https://www.hamiltonproject.org/publication/economic-fact/seven-economic-facts-about-prime-age-labor-force-participation/
2. Long-run male non-participation. Prime-age male non-participation rose from 2.9% (1950s) to 10.9% (2023); ~38% of non-participating men report disability; both "push" (skills mismatch, manufacturing decline) and "pull" (education, caregiving) factors matter; ~80% of the intergenerational gap explained by observables. (SF Fed WP 2025-07) — https://www.frbsf.org/wp-content/uploads/wp2025-07.pdf

G. Child Tax Credit and family

1. 2021 CTC. SPM child poverty 9.7% (2020) → 5.2% (2021); the CTC kept 2.9 million children out of poverty in 2021, ~2.1 million attributable to the ARPA expansion itself. (Census, Sept 2022) — https://www.census.gov/library/stories/2022/09/record-drop-in-child-poverty.html
2. Employment effects — simulation vs observation. Corinth, Meyer, Stadnicki & Wu (2021, simulation): a permanent expansion would lead 1.5 million working parents (2.6%) to exit. Ananat, Glasner, Hamilton & Parolin (observational, 2021 data): "very small, inconsistently signed, and statistically insignificant" employment effects — while noting a temporary program gives a lower bound on permanent effects. — https://www.nber.org/system/files/working_papers/w29823/w29823.pdf ; https://onlinelibrary.wiley.com/doi/full/10.1002/pam.22528
3. 2025 CTC law. P.L. 119-21: CTC $2,200 per child from 2025, indexed; refundable maximum $1,700; phase-in unchanged at 15% of earnings over $2,500; now requires a valid SSN for the taxpayer (one spouse if joint) as well as the child — Tax Policy Center estimate ~500,000 otherwise-eligible children lose access; families earning roughly $0–$26,000 gain nothing from the increase. (Bipartisan Policy Center, 2025) — https://bipartisanpolicy.org/article/how-the-obbb-changes-to-the-child-tax-credit-will-impact-families/
4. Fertility. 2025 births 3,606,400 (−1%); general fertility rate 53.1 per 1,000 women 15–44 (record low); teen birth rate 11.7, record low. (CDC/NCHS provisional, Apr 2026) — https://www.cdc.gov/nchs/pressroom/releases/20260409.html
5. Household structure. 1974: 54% of families had own children under 18 at home; 2024: 39%. Family households: 79% → 64% of all households. (Census, Nov 2024) — https://www.census.gov/newsroom/press-releases/2024/families-and-living-arrangements.html
6. Deaths of despair. Midlife mortality of white non-Hispanics without a degree went from ~30% below Black Americans' (1999) to ~30% above (2015), driven by drugs, alcohol, suicide. (Case & Deaton, BPEA 2017) — https://www.brookings.edu/bpea-articles/mortality-and-morbidity-in-the-21st-century/

H. AI and labor

1. Early-career canaries. Workers aged 22–25 in the most AI-exposed occupations: employment 19% below the path of less-exposed peers (Nov 2022–June 2026), widened from ~15% in the Aug 2025 version; declines concentrated where AI automates, flat or rising where it augments. (Brynjolfsson, Chandar & Chen, Aug 2026 revision) — https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf
2. No aggregate disruption (yet). Through July 2025, "the broader labor market has not experienced a discernible disruption"; occupational mix shifted ~1 pp, a trend that pre-dated ChatGPT. (Yale Budget Lab, 2025) — https://budgetlab.yale.edu/research/evaluating-impact-ai-labor-market-current-state-affairs

---

CONTESTED EVIDENCE (where competent researchers genuinely disagree)

C1. Has pay decoupled from productivity? EPI (#1–2) uses typical-worker compensation deflated by CPI against net productivity; Winship/Strain (#3) use consistent output deflators, average compensation, and find aggregate pay tracks productivity. Referee view (Takahashi): these are different questions. The level gap between median pay and average productivity is real and is mostly (a) rising pay inequality and (b) deflator divergence, plus (c) a smaller labor-share decline. Stansbury–Summers (#4) show the marginal link persists. Anyone saying "no gap" or "pay flat since 1973" is wrong on at least one of those questions.

C2. How much did top incomes rise? PSZ vs Auten–Splinter (#8–9). They agree on fiscal income; they disagree on imputations (underreported income, retirement income, government consumption, deficits). Brandt: AS's own numbers show a +4.3 pp pre-tax top-1% rise since 1962 — "inequality didn't rise" is not what AS find; "after-tax top share barely rose" (+1.7 pp) is. CBO (#6) sits between.

C3. Minimum wage employment effects. Cengiz et al. (#22) vs Seattle/UW (#24) vs CBO's synthesis (#23). The bunching studies cover increases up to roughly the middle of the historical range; they say less about a $15 federal floor in low-wage states — which is exactly where CBO's job-loss estimates bite.

C4. CTC labor supply. Simulation (Corinth et al.) vs short-run observational (Ananat et al.) (#29). The observational estimate is from a temporary, 6-month, pandemic-era program; the simulation relies on elasticities from the 1990s welfare-reform era. Neither settles the permanent-policy question.

C5. Tariffs and manufacturing. 2018 evidence (#19) says net-negative for manufacturing employment; 2026 data show modest manufacturing gains since December 2025 (#20). A 58k uptick over eight months is not an identified tariff effect in either direction.

C6. AI. Stanford "canaries" (#34) find a large, growing age-specific effect; Yale Budget Lab (#35) finds no aggregate disruption. These are compatible (a narrow entry-level effect inside a stable aggregate), but whether the young-worker effect is AI or post-2022 tech-sector retrenchment/interest rates is disputed.

C7. Family structure and child outcomes. Kearney (The Two-Parent Privilege, 2023) argues the decline in two-parent homes among non-college adults is a first-order driver of child disadvantage; critics argue selection and income explain much of it. Referees could not retrieve Kearney's exact tabulations this round; her qualitative thesis is entered, her specific percentages are not.

C8. Marriage penalties in means-tested programs. No quantitative source verified this round. Existence of penalties in EITC/Medicaid/SSI is structural (phase-outs on combined income); their behavioral effect on marriage is small-to-uncertain in most published work. Treated as Unknown magnitude.

STAGE 2 — OPENING POSITIONS (Part 1 of 2: Left and Center)

Format for each: Causal claim · Evidence · Remedy · Tradeoff conceded · Falsifiable prediction

---

1. Prof. Priya Venkataraman (#19, inequality economics, L)

Causal claim. The central fact of American opportunity since 1980 is not slow growth; it is who got the growth. Chetty et al. show absolute mobility fell from ~90% to ~50%, and that restoring the 1940s distribution of today's GDP would recover 80% mobility while restoring 1940s growth rates under today's distribution gets you only 62% (Brief #13). Distribution is the dominant lever.

Evidence. The top 1% share of income has doubled under every serious measure — fiscal income went from 8.3% to 18% (Brief #9). CBO confirms Gini rose both before and after taxes and transfers, 1979–2021 (#6). And the 2025 reconciliation law points the wrong way: CBO projects the bottom decile loses ~$1,200/yr while the top decile gains ~$13,600/yr (#11).

Remedy. (a) Restore a fully refundable CTC; (b) finance it with higher effective taxation of top incomes and capital; (c) rebuild worker bargaining power (see Kowalski).

Tradeoff I concede. Higher top marginal rates reduce reported taxable income; the elasticity is real, and some of what "rises" in top fiscal income since 1986 is reclassification from C-corp to pass-through, not new income.

Prediction. If distribution is the main driver, then cohorts born 1985–1995 in states that expanded EITC/CTC supplements and raised minimum wages should show higher absolute mobility (Opportunity Insights' next cohort release) than cohorts in states that did neither, by ≥3 pp within 5 years of data availability.

---

2. Prof. Joseph Kowalski (#77, labor history, Socialist)

Causal claim. Wage suppression is a political outcome. Deunionization, an eroded federal minimum ($7.25 since 2009), and employer power broke the post-war bargain.

Evidence. Productivity has grown eight times faster than typical worker pay since 1979 (EPI, #1). Union density fell from 20.1% in 1983 to 10.0% in 2025, private sector just 5.9% (#21). The union wage premium is ~20%: $1,404 vs $1,174 median weekly (#21). Stansbury–Summers themselves find productivity still feeds through at the margin (#4) — so the level gap is about power, not technology.

Remedy. Sectoral bargaining (wage boards by industry, as in California's fast-food council), card-check recognition, real penalties for illegal firings during campaigns, and a $17 federal minimum indexed to median wages.

Tradeoff I concede. Sectoral bargaining in low-margin, tradable manufacturing can push work offshore; wage boards must be limited to non-tradable services (food, care, retail, warehousing) at first.

Prediction. If power is the cause, then in the states that adopt sectoral wage boards the 10th-percentile wage in covered sectors should rise ≥10% relative to matched sectors/states within 3 years without a statistically significant decline in covered employment.

---

3. Prof. Grace Holloway (#71, social work, L)

Causal claim. Child poverty is a policy choice, and 2021 proved it.

Evidence. SPM child poverty fell from 9.7% to 5.2% in one year — nearly halved — with the CTC alone keeping 2.9 million children out of poverty (#28). Employment effects were zero (Ananat et al., #29). Then Congress let it lapse. The 2025 law raised the credit to $2,200 but left the 15% earnings phase-in intact — families under ~$26,000 get nothing from the increase — and the new SSN rule strips ~500,000 citizen and lawfully present children of eligibility because of a parent's status (#30). Meanwhile the 2025 SPM sits at 13.1% despite record median income (#10).

Remedy. Make the full CTC available regardless of earnings; restore child-only SSN eligibility; pay monthly.

Tradeoff I concede. A fully refundable credit costs real money — on the order of $100 billion a year for the 2021 design — and the monthly-payment infrastructure had improper-payment and reconciliation problems in 2021.

Prediction. If the credit, not the pandemic economy, caused the 2021 drop, then any state that enacts a fully refundable state CTC of ≥$1,000/child (as several have) should see its SPM child poverty rate fall ≥2 pp relative to synthetic-control states within 2 years, with parental employment changes of less than 1 pp.

---

4. Prof. Samuel Birch (#10, urban economics, C)

Causal claim. Place is a causal input to opportunity, and it is where the evidence is cleanest.

Evidence. Moving to Opportunity raises lifetime earnings by $302,000 (#14). The Opportunity Atlas shows ~$5,000 SD in adult income across tracts within the same county for low-income kids, and observational tract effects predict causal ones (#15). Notice what doesn't predict mobility: local job growth. That is uncomfortable for both the "just grow the economy" right and the "just redistribute" left.

Remedy. (a) Mobility-assistance vouchers (search counseling, landlord outreach — the Seattle "Creating Moves to Opportunity" model) targeted at families with children under ~10; (b) zoning reform so high-opportunity tracts can add housing.

Tradeoff I concede. Moves help the movers, not the place left behind; and scaling vouchers into a handful of high-opportunity tracts could bid up rents there and dilute the effect.

Prediction. If neighborhood effects are causal at scale, then children in CMTO-style programs should show ≥10% higher earnings at age 26 relative to lottery controls once administrative data become available (early 2030s), and cities that legalize multifamily housing in high-opportunity tracts should show measurable increases in low-income children residing there within 5 years.

---

5. Prof. Nadia Haddad (#15, trade economics, C)

Causal claim. Trade raised aggregate welfare and devastated specific places; we then failed both at compensation and, in 2025–26, at policy design.

Evidence. The China shock cost ~2.4 million jobs; exposed commuting zones lost 1.54 pp of manufacturing employment-to-population, 86% of it as nonemployment rather than migration (#16). TAA's effect was "vanishingly small," and it lapsed in 2022 (#17). The current tariff regime (11.0% average statutory rate) raises prices ~0.7% and costs ~$1,100 per household (#18); the best evidence from 2018 shows tariffs reduce manufacturing employment on net because input costs and retaliation swamp protection (#19).

Remedy. Replace broad tariffs with narrow national-security measures; replace TAA with wage insurance (paying part of the earnings gap when a displaced older worker takes a lower-paying job) plus place-based employment support.

Tradeoff I concede. Wage insurance subsidizes downward mobility and does nothing for workers who don't find a job; and "narrow" national-security tariffs have a way of becoming broad.

Prediction. If input-cost effects dominate, manufacturing employment in industries with high tariffed-input shares should underperform low-input-exposure industries by ≥1% over 2025–2027 in QCEW data, even while aggregate manufacturing ticks up.

---

6. Prof. Jamal Whitaker (#52, workforce education, C)

Causal claim. Two frictions matter most right now for entry-level mobility: licensing walls and an AI-driven collapse in entry rungs.

Evidence. Licensing covers roughly one in five workers (21.6% hold a government license, #25). And AI is already costing young workers: employment of 22–25-year-olds in the most AI-exposed occupations is 19% lower (#34) — the entry rung is disappearing.

Remedy. Universal license recognition across states, sunrise/sunset review of licenses, and a federal apprenticeship-plus-credential guarantee for 18–26-year-olds focused on occupations where AI augments rather than automates.

Tradeoff I concede. Some licensing genuinely protects health and safety (nursing, electrical), and apprenticeship programs have a mixed record in the U.S. outside construction trades.

Prediction. If the AI effect is real and causal, the gap for 22–25-year-olds in automation-exposed occupations should keep widening (to ≥25% by mid-2028) while the gap in augmentation-exposed occupations stays near zero; if it's interest rates/tech retrenchment, the gap should narrow as tech hiring normalizes.

STAGE 2 — OPENING POSITIONS (Part 2 of 2: Right and Libertarian)

---

7. Prof. Douglas Wren (#2, public finance, R)

Causal claim. The "stagnation" narrative is a measurement artifact that has led policy astray. Living standards for the bottom and middle have risen substantially once you count benefits, taxes, transfers, and use consistent price deflators.

Evidence. Under consistent output deflators, aggregate compensation has tracked productivity; even Winship's median compensation rose 40% over 1973–2022 (#3). CBO: income after transfers and taxes grew for every quintile since 1979 (#6), and in 2023 the lowest quintile gained 5% while the top 1% fell 9% (#7) — inequality is now falling. Auten–Splinter show the after-tax top-1% share barely moved (8.6% to 10.3% since 1962, #8). Median household income hit a record $87,460 in 2025 and official poverty a record low 10.2% (#10).

Remedy. Pro-growth tax policy (full expensing, which the 2025 law made permanent), deregulation of housing and licensing, and targeting transfers through work-conditioned programs like the EITC.

Tradeoff I concede. Consistent deflators don't erase the dispersion problem: median compensation growth (40%) is well below productivity (111%) in Winship's own numbers, and the 2025 law's in-kind cuts hit the bottom decile (#11).

Prediction. If the growth story is right, real median household income (Census) should rise ≥1.5%/yr on average over 2025–2028, and the CBO bottom-quintile market income should grow at least as fast as the middle quintile over the same window.

---

8. Prof. Abigail Stroud (#63, family sociology, R)

Causal claim. The most under-discussed driver of child disadvantage is family structure, and our tax-and-transfer system quietly penalizes marriage and non-work.

Evidence. Kearney's Two-Parent Privilege (2023) documents that two-parent homes have become a class marker — common among college-educated parents and much less so among others — and that the child-outcome gaps are not fully explained by income. Families with children at home fell from 54% to 39% of families (1974–2024, #32); fertility is at a record low (GFR 53.1, #31). On the CTC, the expanded credit would have driven 1.5 million parents out of work (Corinth et al., #29). Case and Deaton's deaths of despair tracked the collapse of work, marriage, and community among non-college whites (#33).

Remedy. (a) A work-linked CTC with a larger credit for young children; (b) eliminate marriage penalties in the EITC and Medicaid by raising married-couple phase-out thresholds; (c) pro-family measures like the new "Trump accounts" seeded at birth.

Tradeoff I concede. Nobody has a proven policy that raises marriage rates; the Bush-era Healthy Marriage initiatives had disappointing evaluations. And pro-natal cash transfers abroad produce small fertility effects per dollar.

Prediction. If marriage penalties are behaviorally important, then couples just above EITC phase-out thresholds should show measurably lower marriage rates than those just below — and removing the penalty should raise marriage among low-income cohabiting parents by ≥2 pp within 5 years. If not, I'll drop the claim that the penalty is a cause and keep it as a fairness issue.

---

9. Prof. Walter Pruitt (#8, law & economics, Libertarian)

Causal claim. Government-created entry barriers — minimum wage floors and occupational licenses — sit precisely on the bottom rungs of the mobility ladder.

Evidence. Seattle's minimum wage cost low-wage workers $125 a month — hours were cut more than wages rose (#24). CBO says a $15 federal floor costs 1.3–1.4 million jobs (#23). A quarter of workers now need permission to work (24.0% certified or licensed, #25). None of this is fixed by "power"; it's fixed by getting out of the way.

Remedy. Leave wage floors to states and localities; no federal increase. Universal license recognition, sunset review, and a right to earn a living enforced by courts under state constitutions.

Tradeoff I concede. The bunching literature (Cengiz et al., #22) is the strongest evidence against my minimum-wage view: at moderate state increases, low-wage job counts didn't fall. My claim is about high floors relative to local wages, not about all increases.

Prediction. If high floors bite, then as California's $20 fast-food wage (2024) and other high local floors mature, fast-food employment in those areas should underperform comparable states by ≥3% within 3 years in QCEW data. If it doesn't, I'll narrow my claim further.

STAGE 3 — CROSS-EXAMINATION (Part 1 of 2)

---

Exchange A — Wren (R) vs Venkataraman (L): "Doubled under every serious measure?" (4 rounds)

Wren, round 1. Priya, you said the top 1% share "doubled under every serious measure." Auten–Splinter — published, peer-reviewed, and using the same IRS microdata — have the after-tax top-1% share going from 8.6% in 1962 to 10.3% in 2022 (#8). That's a 20% rise, not a doubling. Even their pre-tax series goes from 11.2% to 15.5%, well short of double. And CBO has the lowest quintile gaining 5% in 2023 while the top 1% fell 9% (#7). You're citing fiscal income (8.3% → 18%) as if it measured economic income. It doesn't: the 1986 Tax Reform Act shifted business income from corporate to individual returns, which mechanically inflates the fiscal series.

Venkataraman, round 1. I'll take the correction on the wording. "Every serious measure" is wrong — AS is a serious measure and it does not show a doubling after tax. But look at what your own evidence concedes. AS pre-tax rises 4.3 points — that is a large increase in concentration. And the 2023 number you cite is a one-year capital-gains swing; the top 1% share of CBO income swings with the stock market every year. CBO's trend is unambiguous: Gini rose before and after taxes and transfers, 1979–2021 (#6). The AS–PSZ gap is almost entirely imputations of income nobody observes: underreported income, government consumption, deficits. Equitable Growth's review points out AS assign 39% of bottom-50% income by 2019 to government consumption and in-kind items (#9). You can make the bottom look richer by counting the value of highways and defense as their income.

Wren, round 2. Every income concept that includes Medicare and Medicaid counts in-kind benefits — CBO does it too, and your side cites CBO. If healthcare transfers are real resources when CBO counts them, they're real when AS counts them. On 1986: do you accept that some of the fiscal-income jump is reporting, not reality?

Venkataraman, round 2. Yes — some. I conceded the reclassification point in my opening. But the Medicare point proves less than you think: CBO counts health transfers at their cost to government, and the bottom-quintile gains in CBO since 1979 are disproportionately rising healthcare costs, not cash a family can spend on rent. On the substantive question — did income concentration rise? — CBO, AS pre-tax and PSZ all say yes. The genuine dispute is the magnitude after taxes and transfers: AS +1.7 pp, PSZ a lot more. I'll amend: "Every serious measure shows a rise in pre-tax concentration; after-tax estimates range from modest (AS) to large (PSZ)."

Wren, round 3. Accepted. And I'll amend mine: I said "inequality is now falling." One year of CBO data where capital gains dropped doesn't establish a trend. I withdraw "now falling" and replace it with "the 2023 CBO data are consistent with no further rise, but one year proves nothing." What I won't withdraw is that the policy conclusion changes a lot between a 1.7-pp and a much larger after-tax rise.

Venkataraman, round 3. Agreed on that last point, which is why the CTC and minimum wage cases in this thread shouldn't depend on the PSZ–AS dispute at all. Chetty's mobility decomposition (#13) doesn't use the top-1% share; it uses the whole parent–child income distribution, and it attributes most of the decline to distribution.

Wren, round 4. One caveat on Chetty: the 1940-cohort baseline was children of Depression-era parents. Of course 90% out-earned them. The 50% for the 1980s cohort is closer to what you'd expect in a mature economy. The decomposition is still informative, but "fading dream" is partly a baseline effect.

Venkataraman, round 4 (close). Partly, yes — the counterfactual that restores 1940s growth to today's distribution recovers only to 62%, which tells you that growth can't recover the Depression baseline either. But the 80% counterfactual isn't a baseline effect; it holds the parents fixed. We can close there.

---

Exchange B — Oyelaran (#12, finance, C-R) vs Kowalski (Soc)

Oyelaran. Joseph, two numbers. First, "eight times faster": EPI's own page shows productivity +93.7% and typical-worker compensation +32.7% from 1979 to 2026 — a ratio of 2.9, not eight (#1). Second, the "20% union premium" is just $1,404 ÷ $1,174 — BLS explicitly says those medians don't control for occupation, industry, or demographics (#21). Unionized workers are older, more public-sector, and more concentrated in higher-paying industries. The raw gap isn't a causal premium.

Kowalski. On the first number, you're right and I'll correct it: 2.9 times since 1979, per EPI. I conflated a different EPI framing. On the premium: fine, the raw gap overstates the causal effect. But the controlled literature still finds a meaningful premium, and — more important — density has been falling for 40 years while the marginal link between productivity and pay still holds (Stansbury–Summers, #4). The level gap is institutional. Private-sector density at 5.9% is not a free-market outcome; it's the outcome of a labor law where illegal firings during campaigns carry trivial penalties.

Oyelaran. Stansbury–Summers find the link persists. That undercuts the "broken bargain" story as much as it supports it. And Winship's consistent-deflator numbers (#3) show a good chunk of EPI's gap is CPI versus output prices — a statistical artifact, not a power grab.

Kowalski. The deflator difference is real, and it isn't an artifact to the worker: consumers pay consumer prices. If the things workers buy (housing, health care, education) rose faster than the things workers make, that is a real loss in living standards. I'll accept that the deflator gap is not employer power. I won't accept that it's imaginary.

Oyelaran. Fair. Then say "the gap has three parts, and power plausibly explains one of them." I'm not going to concede that sectoral bargaining fixes a deflator gap.

Kowalski. Agreed: three parts — inequality in compensation, deflator divergence, and labor-share decline. My remedy addresses the first and part of the third.

---

Exchange C — Gallagher (#20, business ethics, C-R) vs Holloway (L)

Gallagher. Grace, "employment effects were zero" is not what Ananat et al. found. They found "very small, inconsistently signed, and statistically insignificant" effects — and in the same paper they say a temporary program gives a lower bound on the labor-supply effect of a permanent one (#29). Statistically insignificant for six months of a pandemic-era payment is not "zero forever." And your "$100 billion a year" cost figure — where's the source?

Holloway. On "zero": I'll amend to "no detectable short-run effect." That's the honest statement. On cost: I don't have a verified score in front of me; I'll withdraw the number and let the referees rule. But the poverty result is robust: 9.7% to 5.2%, and Census attributes 2.1 million children to the expansion specifically (#28).

Gallagher. The poverty result I accept. The question is the permanent design.

Holloway. Then we agree on what we don't know. Where I won't move: the 2025 law's increase gives nothing to families below ~$26,000 (#30). That's a design choice to exclude the poorest children, not an evidence-based one.

---

Exchange D — Reyes (#49, early childhood, L) vs Stroud (R)

Reyes. Abigail, you wrote that the expanded CTC "would have driven 1.5 million parents out of work." That's a simulation (Corinth et al.), not an observed effect — and the observed-data study from the same period found no detectable employment effect (#29). You presented a model prediction as a finding. Second: on Kearney — the critics' point is selection. The people who marry differ in income, education, and stability before marriage. How much of the "two-parent privilege" is causal?

Stroud. On the first point, I'll amend: "Corinth et al. estimate that a permanent expansion would lead 1.5 million parents to exit; observational short-run evidence did not detect this; the question for a permanent policy is open." I'd add that Corinth et al.'s elasticities come from the 1990s welfare-reform era, which did show large employment responses to work-conditioning — the evidence I'm pointing to isn't imaginary, it's just older.

Reyes. Older and from a very different labor market for single mothers. And Kearney?

Stroud. Kearney herself acknowledges selection and uses multiple approaches; my claim is that the gap survives controls for income, not that it's 100% causal. I'll put it this way: if two-parent homes were purely a proxy for income, then cash transfers that closed the income gap would close the child-outcome gap. The evidence that they do so completely doesn't exist.

Reyes. Nor does evidence that any policy changes marriage rates, which you conceded in your opening. So the family-structure argument, even if true, doesn't give us a lever — cash does.

Stroud. It gives us a constraint: don't build programs that penalize marriage. That's why my remedy is marriage-penalty relief, not a marriage-promotion bureaucracy.

STAGE 3 — CROSS-EXAMINATION (Part 2 of 2)

---

Exchange E — Ward (#68, rural sociology, R) vs Haddad (C) (3 rounds)

Ward, round 1. Nadia, you're citing 2018 evidence while ignoring 2026 data. Manufacturing is up 58,000 since December 2025 (#20). Meanwhile the "free trade raised aggregate welfare" framing is precisely what left the towns I study hollowed out, and your ADH citation proves it: 86% of the job loss became nonemployment, not migration (#16). You also said the China shock cost "~2.4 million jobs." That figure isn't in the referees' brief. Source?

Haddad, round 1. The 2.4 million is from Acemoglu, Autor, Dorn, Hanson and Price's 2016 estimate for 1999–2011, which I'm confident of, but it isn't in the verified record, so I'll let the referees rate it and rely on the 1.54-pp figure for my argument. On 58,000 jobs: that's eight months of data with no counterfactual. Payroll growth across the whole economy averaged just 31,000 a month over the last year (#5). Manufacturing gaining 58k in a slowing economy is interesting; it isn't identification. Flaaen and Pierce are the only identified estimate we have, and they find input costs (−1.1%) and retaliation (−0.7%) outweighing protection (+0.3%) (#19).

Ward, round 2. 2018 tariffs were narrow, hit intermediate inputs, and drew targeted retaliation against farm exports. A broad, durable tariff regime changes investment decisions over years, not quarters. You don't get reshoring in eight months, and you don't get it in Flaaen–Pierce's window either.

Haddad, round 2. That's a fair limit on the 2018 evidence — it's a short-run estimate. But your argument cuts both ways: if the benefits come in years, you can't cite 58,000 jobs today as evidence of them. And the costs are here now: ~0.7% on the price level, ~$1,100 per household (#18). That's a regressive consumption tax.

Ward, round 3. I'll concede I can't claim the 58,000 as a tariff effect. I'll replace it with a prediction: if durable tariffs work, manufacturing construction and capex should rise before employment, and manufacturing employment should be higher in 2029 than in 2024. If it's lower, I'll say so. What I won't concede is that the pre-2016 consensus had the right weights. It treated concentrated, persistent place-based losses as a rounding error against diffuse consumer gains, and ADH show those losses lasted two decades.

Haddad, round 3. That I concede, and it's why I lead with compensation. The consensus underweighted persistence. Where we differ is whether tariffs or wage insurance plus place-based support is the better repair; the ADH authors themselves prefer the latter (#17).

---

Exchange F — Marsh (#48, education policy, R) vs Birch (C)

Marsh. Samuel, "MTO raises lifetime earnings by $302,000" — for whom? It's the estimate for a child who moves at age 8, undiscounted; the present value is $99,000, and for children who moved as teenagers the effect was zero or slightly negative, with nothing for adults (#14). Stated as you did, it's an overstatement by a factor of three in PV and applies to a minority of movers.

Birch. I accept the correction: "~$302,000 undiscounted, ~$99,000 PV, for a child moving at age 8; no gain for teens or adults." That actually sharpens my remedy — target young children, which I did propose.

Marsh. Second issue: scale. The Atlas says local job growth doesn't predict mobility (#15). Then what does? If it's peers, schools and social capital, those are partly created by the families who already live there. Move enough low-income families in and you change the thing you're buying.

Birch. That's the general-equilibrium worry I conceded. Current voucher-mobility programs are small enough that it doesn't bind. At scale the answer is to build in high-opportunity tracts, which is why zoning reform is half my remedy. I'll note, pointedly, that the zoning half is where your bloc's local officials often vote no.

Marsh. Some do. I'll support the zoning half.

---

Exchange G — Tran (#58, computer science, Lib) vs Whitaker (C)

Tran. Jamal, "AI is already costing young workers 19% of jobs" misreads the paper. It's a 19% shortfall relative to less-exposed peers — the most-exposed quintiles fell ~11% while the least-exposed grew ~10% (#34). That's a divergence, not 19% of jobs lost. And the timing — from November 2022 — coincides with the Fed's rate hikes and a tech-hiring bust that hit software and customer service, the most AI-exposed jobs. Yale's Budget Lab found no discernible aggregate disruption (#35).

Whitaker. I'll amend: "19% relative shortfall versus less-exposed peers, of which roughly half is an absolute decline." On confounding: the authors' key test is the automation/augmentation split — within similarly exposed occupations, employment falls where AI substitutes and is flat or rising where it complements (#34). Interest rates don't obviously sort that way.

Tran. Unless automation-exposed tasks are disproportionately in rate-sensitive tech firms. I'm not saying it's zero; I'm saying one paper with a widening gap is a hypothesis, and "the entry rung is disappearing" is a headline.

Whitaker. Fair. My prediction stands as the test: if the automation-exposed gap keeps widening as tech hiring normalizes, it's AI.

---

Exchange H — Castellano (#1, labor economics, C-L; speaking here as a discussant, not as referee) vs Pruitt (Lib)

Castellano. Walter, the Seattle study you cite reports ~$74 a month in the revised version, not $125 (#24). The $125 figure comes from the 2017 draft. And the study excluded multi-site employers, about 40% of the workforce — critics argue that biases toward finding losses if work shifted to chains. A Berkeley team studying the same city's restaurants found no negative employment effect.

Pruitt. I'll correct the number to $74 per month from the published revision; the direction of the finding — hours cut by 6–7%, more than offsetting a 3% wage gain, concentrated on inexperienced workers — didn't change. On the exclusion: the UW team found no evidence of large shifts to multi-site firms, and Berkeley's restaurant study is restaurants only, where UW also found minimal effects. The two studies aren't contradicting each other as much as advertised.

Castellano. Then we should both be using Cengiz et al. — 138 increases, no change in low-wage job counts over five years (#22) — as the prior for moderate increases.

Pruitt. For moderate increases, yes. My claim is about floors that are high relative to local wages. Cengiz's sample doesn't include a $15 federal floor applied in Mississippi, and CBO's 1.3–1.4 million job-loss median is exactly about that (#23).

Castellano. Agreed that's the relevant frontier. CBO's median elasticity is −0.25 (#23) — modest per worker, but on a very large bite. We disagree on the weights, not the evidence.

---

Exchange I — Venkataraman (L) vs Wren on remedies (short)

Venkataraman. Douglas, your remedy includes "work-conditioned programs like the EITC" — but the CBO distributional analysis of the law you're defending has the bottom decile losing ~3.1% of income (#11). How is that pro-mobility?

Wren. It's the in-kind cuts — Medicaid and SNAP — that drive that number, and I conceded it in my opening. I'd have financed the tax changes differently. I'm defending the growth provisions, not every offset.

STAGE 4 — REFEREE FACT-CHECK

Brandt (#57) · Takahashi (#17) · Castellano (#1, specialist; recused on Exchange H, where Brandt ruled alone)

Ratings: S = Supported · C = Contested (credible evidence on both sides, or right direction but disputed magnitude) · U = Unsupported (not established by the record; may be true) · W = Wrong (contradicted by the cited source). "#n" = Evidence Brief item.

# · Speaker · Claim · Rating · Basis
1 · Venkataraman · Absolute mobility ~90% (1940) → ~50% (1980s); distribution explains most of decline · S · #13 Chetty et al. 2017
2 · Venkataraman · Top 1% share "doubled under every serious measure" · W · #8: AS after-tax 8.6→10.3%, pre-tax 11.2→15.5%. Amended in Exchange A
3 · Venkataraman · CBO Gini rose before and after taxes/transfers 1979–2021 · S · #6
4 · Venkataraman · P.L. 119-21: bottom decile −$1,200/yr, top decile +$13,600/yr · S · #11 CBO
5 · Venkataraman / Wren · Post-1986 reclassification inflates fiscal top-income series · C · Direction agreed by both; magnitude disputed (#9 PSZ reply vs AS)
6 · Kowalski · Productivity grew "eight times faster" than typical pay since 1979 · W · #1: 93.7% vs 32.7% ≈ 2.9×. Corrected in Exchange B
7 · Kowalski · Union density 20.1% (1983) → 10.0% (2025); private 5.9% · S · #21 BLS
8 · Kowalski · "Union wage premium is ~20%" · U · #21: $1,404 vs $1,174 is a raw gap; BLS says it's unadjusted. Causal premium not in record. Amended
9 · Kowalski · Federal minimum $7.25 since 2009 · S · FLSA statutory rate (DOL Wage & Hour); not disputed
10 · Kowalski · Controlled studies still find a "meaningful" union premium · C · Not verified this round; plausible, magnitude debated in literature
11 · Holloway · SPM child poverty 9.7% → 5.2%; 2.9M children kept out, 2.1M from expansion · S · #28 Census
12 · Holloway · 2021 CTC employment effects were "zero" · U · #29: statistically insignificant ≠ zero; authors call it a lower bound. Amended to "no detectable short-run effect"
13 · Holloway · 2025 increase gives nothing to families below ~$26k; ~500k children lose via SSN rule · S · #30 BPC citing TPC
14 · Holloway · Children losing eligibility are eligible children excluded due to parent's SSN status · S · #30 (child must still have SSN; exclusion runs through the taxpayer)
15 · Holloway · 2021-design CTC costs "~$100B/yr" · U · No score retrieved. Withdrawn in Exchange C
16 · Birch · MTO raises lifetime earnings by $302,000 · C · #14: true only for a move at age 8, undiscounted ($99k PV); zero/negative for teens. Amended in Exchange F
17 · Birch · ~$5,000 within-county SD; job growth weakly predicts mobility · S · #15
18 · Haddad · China shock cost "~2.4 million jobs" · U · Attributed to Acemoglu et al. (2016); referees did not retrieve it this round. Not relied upon
19 · Haddad · −1.54 pp mfg emp/pop; 86% via nonemployment · S · #16
20 · Haddad · TAA effect "vanishingly small"; program lapsed July 2022 · S · #17
21 · Haddad · Tariffs: ~0.7% price level, ~$1,100/household · S · #18 Yale Budget Lab (Aug 2026)
22 · Haddad · Tariffs reduce manufacturing employment on net · C · #19 supports for 2018, short run; no identified estimate yet for 2025–26 regime (Ward's point accepted)
23 · Haddad · Trade raised aggregate welfare while concentrating losses · S · #17 Brookings/ADH summary
24 · Whitaker · 21.6% of workers hold a government license · S · #25 BLS
25 · Whitaker · AI "already costing young workers 19%" of jobs · W · #34: 19% is a relative shortfall vs less-exposed peers, not a job-count loss. Amended in Exchange G
26 · Whitaker · (Amended) roughly half of the gap is absolute decline · S · #34: exposed −11%, unexposed +10%; slightly more than half. Referees note the fraction is ~11 of ~21 pp
27 · Wren · Under consistent deflators aggregate compensation tracks productivity · C · #3 (Winship) vs #1–2 (EPI); #4 consistent with marginal link
28 · Wren · Median compensation +40% 1973–2022 (Winship) · S · #3
29 · Wren · After-tax/transfer income grew for every quintile since 1979 · S · #6
30 · Wren · "Inequality is now falling" · U · #7 is one year, driven by capital gains. Withdrawn in Exchange A
31 · Wren · AS after-tax top-1% 8.6% → 10.3% · S · #8
32 · Wren · 2025 record median income and record-low official poverty · S · #10 — note SPM (13.1%) flat
33 · Wren · "The stagnation narrative is a measurement artifact" (opening framing) · C · Partly: deflator & benefits matter (#3), but Winship's own median (40%) ≪ productivity (111%); dispersion is real
34 · Wren · 1940-cohort baseline is partly a Depression effect · C · Interpretive; counterfactuals in #13 partially address it
35 · Wren · 2025 law made full expensing permanent · U · Referees believe this is accurate but did not verify the statutory text this round
36 · Stroud · Families with own children 54% → 39% (1974–2024); GFR record low 53.1 · S · #32, #31
37 · Stroud · Expanded CTC "would have driven 1.5 million parents out of work" · W · #29: a simulation projection presented as a finding; observational evidence didn't detect it. Amended in Exchange D
38 · Stroud · Two-parent gap in child outcomes persists after income controls (Kearney) · C · C7: thesis entered; selection dispute unresolved; figures not verified
39 · Stroud · Deaths of despair concentrated among non-college whites, with parallel decline in work & marriage · S · #33
40 · Stroud · Healthy Marriage evaluations were disappointing · U · Not verified this round (referees believe broadly accurate)
41 · Stroud · "Trump accounts" seeded at birth exist under current law · U · Not verified this round
42 · Pruitt · Seattle cost low-wage workers "$125 a month" · W · #24: published revision = ~$74/month. Corrected in Exchange H
43 · Pruitt · CBO: $15 federal floor → 1.3–1.4M jobs lost (median) · S · #23
44 · Pruitt · "A quarter of workers now need permission to work" · C · #25: 24.0% includes private certifications, which are not permission; government license = 21.6%
45 · Pruitt · UW team found no evidence of shift to multi-site firms · U · Not in record
46 · Ward · Manufacturing +58,000 since Dec 2025 (as fact) · S · #20
47 · Ward · …as evidence tariffs are working · U · No counterfactual; withdrawn in Exchange E
48 · Ward · 86% of China-shock loss became nonemployment · S · #16
49 · Oyelaran · BLS union/nonunion medians are uncontrolled · S · #21
50 · Gallagher · Ananat et al. call temporary-CTC estimates a lower bound · S · #29
51 · Castellano · UW excluded ~40% (multi-site); Berkeley found no restaurant job loss · S · #24 (EPI summary; contested interpretation noted)
52 · Tran · Young-worker AI gap may be confounded by rates / tech bust · C · Plausible; automation/augmentation split (#34) partly rebuts
53 · Marsh · MTO: no gains for teen movers or adults · S · #14
54 · Kowalski (amended) · Gap = compensation inequality + deflator divergence + labor-share decline · S · Consistent with #1, #3, #4 (referee view C1)

Tally

Supported 29 · Contested 10 · Unsupported 10 · Wrong 5 (54 claims)

Retractions / amendments on the record

• Venkataraman: "doubled under every serious measure" → "every serious measure shows a rise in pre-tax concentration; after-tax estimates range from modest (AS, +1.7 pp) to large (PSZ)."
• Kowalski: "eight times" → "2.9 times since 1979 (EPI)." "20% premium" → "raw gap ~20%; causal premium smaller and not established here."
• Holloway: "zero" → "no detectable short-run effect"; $100B figure withdrawn.
• Birch: "$302,000" → "$302k undiscounted / $99k PV for a move at age 8; nothing for teens or adults."
• Haddad: stands on 1.54 pp and 86%; will not use 2.4M until verified.
• Whitaker: "costing 19% of jobs" → "19% relative shortfall; ~11% absolute decline in the most-exposed quintiles."
• Wren: "inequality is now falling" withdrawn. Wren adds: "On expensing, I'll stand by it but accept the referees' U until someone posts the statute section."
• Stroud: "would have driven 1.5 million out of work" → "Corinth et al. project…; unresolved for a permanent policy." Stroud adds: "Accept the U's on Healthy Marriage and the accounts; neither was load-bearing."
• Pruitt: "$125" → "$74"; "a quarter need permission" → "about one in five need a government license"; multi-site claim withdrawn: "I was recalling a response paper I can't produce here."
• Ward: 58,000 jobs are a fact, not a tariff effect; replaced with a 2029 prediction.

Brandt's note: Five "Wrong" ratings split across blocs — two Left (Venkataraman, Kowalski), one Center (Whitaker), two Right (Stroud, Pruitt). The most common error type wasn't fabrication; it was dropping the qualifier (raw vs causal, simulated vs observed, relative vs absolute, age-8 vs everyone). Takahashi's note: The two biggest open empirical questions in this domain — after-tax top-income growth and permanent-CTC labor supply — are both cases where the answer depends on untestable imputations or on extrapolating from temporary policy. Humility is warranted on both sides.

STAGE 5 — STEELMAN

Each bloc writes the strongest version of an opposing bloc's case. The steelmanned bloc replies.

---

Left steelmans the Right — written by Venkataraman (#19, L)

"The Right's strongest case isn't 'inequality is a myth.' It goes like this. First, the measurement critique is partly right: the gap between CPI-deflated median wages and output-deflated productivity overstates the loss in workers' command over output. Benefits have grown as a share of pay, and CBO shows after-tax-and-transfer income rising for every quintile (#6). Second, the most robust mobility fact — Chetty's 90%→50% — has a Depression-era baseline, and the level of living standards for the median family in 2025 ($87,460 median income, #10) is historically high. Third, the things that most reliably hurt poor kids' prospects — family instability, non-work among prime-age men (non-participation 2.9%→10.9%, #27), deaths of despair (#33) — are not cured by transfers alone. Cash can raise measured income without rebuilding the institutions of work and family that make income durable. Fourth, entry barriers the Left often defends — licensing (one in five workers, #25), high minimum wages where local wages are low (CBO 1.3–1.4M median job loss for $15, #23), restrictive zoning — fall hardest on people at the bottom. So the Right's program is: growth, work-linked support, remove barriers, don't penalize marriage. Evidence at the margin supports each piece even if it doesn't support the Right's broader distributive claims."

Reply — Stroud (R) for the Right bloc: Accepted as fair, with one correction. "Cash can raise measured income without rebuilding institutions" understates our claim: we argue some cash designs actively weaken work attachment, and the 1990s welfare reform evidence is why. We accept that the 2021 short-run data didn't show it.

Reply — Pruitt (Lib): Accepted. I'd add that libertarians don't share the Right's pro-tariff wing; on trade we're with Haddad.

---

Right steelmans the Left — written by Wren (#2, R)

"The Left's strongest case: First, even on the Right's preferred data, pre-tax income concentration rose — AS's own pre-tax top-1% share rose 4.3 points (#8) — and dispersion within labor income is why median compensation (Winship: +40%) lagged productivity (+111%) (#3). That's not a measurement artifact; it's the core fact. Second, the evidence that policy can move outcomes quickly is strongest precisely for cash to families: SPM child poverty fell from 9.7% to 5.2% in one year (#28), and the best observational study didn't detect an employment penalty (#29). Third, the institutions that used to compress wages — unions, a real minimum wage — have eroded (union density halved to 10.0%, private 5.9%, #21), and the best modern minimum-wage evidence (Cengiz et al., 138 increases, #22) finds wage gains without job losses at moderate levels. Fourth, the 2025 law moves in the wrong direction on distribution (bottom decile −3.1% of income, top +2.7%, #11) and excludes the poorest children from the CTC increase (#30). So the Left's program — refundable child benefits, stronger bargaining, moderate indexed minimum wages — targets the parts of the problem with the best causal evidence."

Reply — Holloway (L) for the Left bloc: Accepted as fair. One correction: the Left's case on unions isn't only about wages; it's about voice and political countervailing power, which no regression of wage premiums captures.

Reply — Kowalski (Soc): Accepted, though I'd say "sectoral bargaining," not just "stronger bargaining" — enterprise-level organizing can't reach fissured workplaces.

---

Center steelmans both flanks' shared critique of the Center — written by Birch (#10, C)

"Both flanks say the Center's evidence fetish produces timid policy. The Left says: while we wait for identification, a generation of kids grows up poor, and the Center's 'contested' label is a thumb on the status-quo scale. The Right says: the Center treats every RCT as the truth and ignores what can't be randomized — culture, family, community, national industrial capacity. Ward's point that the China-shock consensus underweighted persistent place-based losses (#16) is a real indictment of the Center's own past: the trade consensus we held in 2000 missed the biggest labor-market story of the next twenty years."

Reply — Ward (R): Accepted as fair, and appreciated.
Reply — Venkataraman (L): Accepted.

---

Center steelmans itself against the flanks — written by Haddad (#15, C) (for balance; both flanks were invited to amend)

"Every flank claim that got downgraded in Stage 4 got downgraded for the same reason: dropping the qualifier. The Center's value-add is to keep the qualifier. That isn't timidity; it's the reason the 2021 CTC evidence and the MTO evidence are persuasive at all."

Reply — Kowalski (Soc): Not accepted in full. Qualifiers are valuable; "we need more research" is also a political choice when the poverty data are this clear.
Reply — Pruitt (Lib): Accepted.

STAGE 6 — REVISED PROPOSALS

Drafted jointly by the openers after cross-examination; the referees checked cost/precedent claims. "No official score" means neither CBO nor JCT has scored this specific design; referees did not accept unsourced cost numbers.

---

P1. Full CTC for young children + restore child-based SSN eligibility

Mechanism. Remove the 15%-of-earnings phase-in (#30) for children under 6, so the full $2,200 (indexed) is available regardless of earnings; keep the phase-in for older children (compromise with Stroud's work-link concern). Restore pre-2025 eligibility where the child has an SSN even if a parent does not.
Cost. No official score for this design. (Holloway's earlier $100B/yr figure for the full 2021 design was withdrawn.)
Precedent. 2021 ARPA expansion: SPM child poverty 9.7%→5.2%; 2.1M children attributed to the expansion (#28); short-run employment effects undetected (#29).
Key risk. Permanent-policy labor supply is unknown (Contested C4); Corinth et al. project large exits (#29). Improper-payment control without the EITC's error rates.

P2. Replace lapsed TAA with universal wage insurance for displaced workers

Mechanism. Any worker aged 45+ displaced by a plant closure or mass layoff (not just trade-certified) who takes a lower-paying job within 26 weeks receives 50% of the wage gap for two years, capped at $10,000 total; paired with place-based employment grants to high-nonemployment commuting zones.
Cost. No official score.
Precedent. TAA's "Reemployment TAA" (RTAA) for workers 50+ used this structure; TAA as a whole had a "vanishingly small" effect on area income (#17) and lapsed July 2022. ADH find 86% of China-shock losses became nonemployment (#16) — the exact margin wage insurance targets.
Key risk. Subsidizes wage cuts and does nothing for the non-reemployed; the take-up of the RTAA structure was low (referees: magnitude not verified).

P3. Occupational licensing: universal recognition + sunrise/sunset review

Mechanism. Federal grants conditioned on states (a) recognizing out-of-state licenses in good standing, (b) requiring a sunrise cost-benefit review for new licenses and sunset review of existing ones, (c) removing blanket criminal-record bans unrelated to the occupation.
Cost. No official score; grant program could be sized at any level.
Precedent. Arizona enacted the first universal license recognition law (2019); several states followed. Scope: 21.6% of workers hold a government license (#25).
Key risk. Health/safety licensing (nursing, electrical) must stay rigorous; race-to-the-bottom if recognition is from the least demanding state.

P4. Remove EITC marriage penalties

Mechanism. Set the married-couple EITC phase-out start at double the single-filer threshold (full parity), indexed.
Cost. No official score.
Precedent. The 2001 tax law (EGTRRA) partially raised the married-couple phase-out threshold; this completes it.
Key risk. Behavioral marriage effects may be small (C8: magnitude unknown); benefits go to two-earner couples somewhat above the poverty line rather than the poorest families.

P5. Federal minimum wage to $12 by 2029, then indexed to median wage

Mechanism. Phase from $7.25 to $12 in four steps; index thereafter to 50% of the median full-time wage; tipped subminimum rises proportionally.
Cost / effects. No official score for $12. CBO's interactive tool covers $10–$15 designs (#23); CBO's $15 estimates were a median 1.3–1.4M job loss and 0.9–1.3M people out of poverty. A $12 floor sits closer to the range Cengiz et al. studied (#22).
Precedent. 138 state increases 1979–2016 with no detectable loss in low-wage job counts (#22); Seattle's $13 step reduced low-wage hours 6–7% (#24).
Key risk. Bite is much larger in low-wage states than in the Cengiz sample; median-wage indexing locks in any mistake.

P6. Sectoral wage-board pilots in non-tradable services

Mechanism. Federal evaluation funding for up to 10 states that create tripartite (worker/employer/public) wage boards limited to fast food, home care, and warehousing; mandatory randomized-rollout or synthetic-control evaluation.
Cost. No official score; modest (evaluation grants).
Precedent. California's Fast Food Council ($20 minimum from 2024). Referees: evidence on its employment effects is not in this record and is disputed.
Key risk. Regulatory capture; price pass-through; boards expanding beyond pilot sectors.

P7. Mobility vouchers + zoning incentive grants

Mechanism. Convert a share of new Housing Choice Vouchers to mobility vouchers with search counseling and landlord outreach for families with children under 10; pair with competitive grants to jurisdictions that legalize multifamily housing in high-opportunity tracts.
Cost. No official score.
Precedent. MTO: ~$99k PV earnings gain for a move at age 8 (#14); Seattle–King County "Creating Moves to Opportunity" substantially increased moves to high-opportunity areas (Bergman et al.; exact rates not verified this round).
Key risk. General-equilibrium dilution (Marsh); local opposition to building.

P8. Replace broad tariffs with targeted national-security tariffs over 3 years

Mechanism. Phase down the across-the-board tariffs (11.0% average statutory rate, #18) over three years, retaining product-specific tariffs on semiconductors, critical minerals, and defense-relevant inputs after a published national-security review.
Cost. Forgoes a large share of the ~$1.9T 10-year tariff revenue (#18) — no official score for this phase-down.
Precedent. 2018 tariffs reduced manufacturing employment on net in the short run (#19).
Key risk. Revenue loss must be replaced; possible reshoring benefits (Ward) forgone before they can be measured.

AI proposals were judged premature for a vote; see Verdict, "Research agenda."

STAGE 7 — RECORDED VOTE

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).

Proposal · Left (of 34) · Center (of 33) · Right (of 32) · Consensus?
P1 Full CTC for under-6s + child SSN eligibility · 33 (97%) · 23 (70%) · 12 (38%) · No
P2 Universal wage insurance replacing TAA · 23 (68%) · 29 (88%) · 21 (66%) · YES
P3 Licensing: universal recognition + sunrise/sunset · 22 (65%) · 30 (91%) · 30 (94%) · YES
P4 Remove EITC marriage penalties · 27 (79%) · 29 (88%) · 28 (88%) · YES
P5 $12 federal minimum by 2029, median-indexed · 33 (97%) · 21 (64%) · 6 (19%) · No
P6 Sectoral wage-board pilots (non-tradable services) · 30 (88%) · 10 (30%) · 2 (6%) · No
P7 Mobility vouchers + zoning incentive grants · 31 (91%) · 29 (88%) · 18 (56%) · No (Right short by 2)
P8 Phase down broad tariffs to targeted security tariffs · 28 (82%) · 26 (79%) · 12 (38%) · No

Three consensus reforms: P2, P3, P4. P7 came closest to a fourth.

---

Vote explanations

Prof. Abigail Stroud (#63, R) — Yes on P2, P3, P4, P7; No on P1, P5, P6, P8.
"On P1 I moved further than I expected — the 2021 poverty result is not in dispute — but removing the phase-in removes the work link for exactly the families where the 1990s evidence showed work responses. What would change my mind: a credible study of a permanent fully refundable credit (a state program running 3+ years, or Canada's Child Benefit with U.S.-comparable data) showing parental employment effects under 1 pp. I voted yes on P7 because the MTO evidence is randomized and the zoning half expands supply."

Prof. Joseph Kowalski (#77, Soc) — Yes on P1, P5, P6, P7, P8; No on P2, P3; Yes on P4.
"No on P2: wage insurance pays workers to accept downgrading and lets employers off the hook. No on P3: licensing reform as drafted will be used to de-skill care and trades work that unions fought to professionalize. What would change my mind on P3: a version that pairs recognition with prevailing-standard floors; on P2, evidence that RTAA recipients' earnings recovered faster than comparable non-recipients in the long run, not just while subsidized."

Prof. Nadia Haddad (#15, C) — Yes on P1, P2, P3, P4, P7, P8; No on P5, P6.
"On P5 I voted no, narrowly: $12 is probably fine in most metros, but median-wage indexing across a national floor in Mississippi and Massachusetts is the wrong design. What would change my mind: a regionally indexed version (e.g., to state or metro median wages), which I'd support."

Prof. Walter Pruitt (#8, Lib) — Yes on P3, P4, P8; No on P1, P2, P5, P6, P7.
"No on P7's voucher half — I support the zoning half and would have voted for it alone. No on P2 — another program with a certification bureaucracy. What would change my mind on P5: QCEW evidence that California's $20 fast-food floor and similar high local floors produce no relative employment decline after 3 years (my own stated prediction)."

Prof. Grace Holloway (#71, L) — Yes on all except P3 (abstained within bloc count as No).
"I'm not against licensing reform; I'm against universal recognition without a floor for social work and child-care credentials. What would change my mind: a sunrise/sunset process with health-and-safety carve-outs written in, not left to the review."

Prof. Russell Ward (#68, R) — Yes on P2, P3, P4; No on P1, P5, P6, P7, P8.
"No on P8: phasing down before we can measure a reshoring effect is prejudging the experiment. What would change my mind: if manufacturing employment in 2029 is below 2024, or if QCEW shows tariffed-input-heavy industries underperforming by ≥1% (Haddad's prediction), I'll vote for P8 next time. Yes on P2 — my towns were exactly where TAA failed."

Prof. Samuel Birch (#10, C) — Yes on P1, P2, P3, P4, P7, P8; No on P5, P6.
"P7 failed by two Right votes. Its zoning half is the most evidence-backed supply-side reform in this thread. What would change my mind on P6: a randomized or staggered rollout design written into the statute."

STAGE 8 — VERDICT

Prof. Adelaide Wainwright (#100), moderator

Established (the evidence is strong, whatever your bloc)

1. Absolute mobility fell a lot: ~90% of 1940-born children out-earned their parents vs ~50% of those born in the 1980s. Most of the fall comes from how growth was distributed, not how fast it was (#13). Wren's point about the Depression-era baseline qualifies how we read the number, but it doesn't change that conclusion.
2. Where a child grows up has a causal effect, and early childhood matters most. MTO shows ~$99k PV in lifetime earnings for a child who moves at 8 and nothing for teenagers (#14). Differences between neighborhoods in the same county are large (#15).
3. Pre-tax income concentration rose. Every serious series agrees on this, Auten–Splinter's included (#8–9). CBO finds Gini rose both before and after taxes and transfers from 1979 to 2021 (#6). Incomes after taxes and transfers also grew for every quintile (#6).
4. Median pay has lagged average productivity, and the link between them isn't broken. The gap has three parts: growing pay inequality, a gap between consumer and output price deflators, and a smaller fall in the labor share. At the margin, productivity growth still passes through to pay (#1–4). "Pay flat since the 1970s" is wrong. So is "no gap at all."
5. The China shock did long-lasting damage to particular places, and TAA did little to repair it. Most of the lost jobs turned into non-work rather than moves to other places (#16–17). TAA has not certified anyone since 2022.
6. The 2021 CTC cut measured child poverty sharply: SPM child poverty fell from 9.7% to 5.2% (#28).
7. The 2025 law is regressive in how it shifts resources, by CBO's estimate: the bottom decile loses ~3.1% of income and the top decile gains ~2.7% (#11). Its CTC increase gives nothing to the lowest-earning families (#30).
8. Moderate minimum-wage increases have not measurably reduced the number of low-wage jobs (#22).
9. Union density is 10.0%; in the private sector it's 5.9% (#21). About one worker in five needs a government license (#25). The general fertility rate is at a record low (#31).

Contested (credible evidence on both sides)

• After-tax top-income growth: +1.7 pp (AS) vs much larger (PSZ). The answer depends on imputations no one can observe directly.
• Labor supply under a permanent refundable CTC: Corinth et al.'s simulation vs Ananat et al.'s short-run observational estimate.
• High minimum wages relative to local pay: CBO's $15 estimates vs extrapolations from Cengiz et al. Seattle is disputed on method.
• The 2025–26 tariffs and manufacturing: the only identified evidence, from 2018, is net-negative. Modest job gains in 2026 are not identified as a tariff effect.
• How much of the family-structure gap is causal (Kearney vs selection).
• Whether the entry-level employment gap in AI-exposed jobs is caused by AI or by interest rates and the tech-hiring bust.

Unknown

• Whether marriage penalties change behavior. Their existence is structural; we have no verified estimate of how much they affect marriage.
• What any policy does to fertility at scale in the U.S.
• Wealth concentration. We could not retrieve the current Fed DFA shares, so no wealth figure was entered. The Symposium should not claim one without a source.
• Scale effects of mobility vouchers, and the long-run effects of sectoral wage boards.

Research agenda

1. A permanent-CTC labor supply test. Use state CTCs that have run 3+ years, plus Canada's Child Benefit, and report parental employment with pre-registered specifications.
2. QCEW tests of the stated predictions: Haddad's input-cost prediction (tariffed-input industries underperform by ≥1% over 2025–27), Ward's 2029 manufacturing prediction, and Pruitt's 3-year test of California's $20 fast-food wage.
3. Tracking AI and entry-level work. Extend the Brynjolfsson–Chandar–Chen automation/augmentation split with firm-level controls for interest-rate sensitivity. Whitaker's ≥25%-by-2028 threshold is the test. Consider a CPS module on AI task exposure, modeled on the 2015 licensing module.
4. Fed DFA wealth shares. Reconcile them with the AS/PSZ income series.
5. Earnings at 26 for CMTO participants once the data exist.

Consensus reforms (≥60% in each bloc)

• P2 — Universal wage insurance for displaced workers 45+ (replacing TAA): L 68% / C 88% / R 66%
• P3 — Occupational licensing: universal recognition plus sunrise/sunset review: L 65% / C 91% / R 94%
• P4 — Remove EITC marriage penalties: L 79% / C 88% / R 88%

Near miss: P7, mobility vouchers plus zoning grants, reached Right 56%. Several Right votes (Pruitt among them) supported the zoning half alone. A split bill would probably reach consensus.

Not consensus: P1 CTC (Right 38%), P5 $12 minimum wage (Right 19%), P6 sectoral pilots (Center 30%), P8 tariff phase-down (Right 38%).

Where Symposium 1 was wrong or too glib in this domain

The charter faults Symposium 1 for unsourced, rounded, from-memory claims and fast consensus. In this domain, that style fails in a specific way: **the qualifier is the finding.** Stage 4 of this thread produced five "Wrong" ratings, spread across all three blocs. Every one of them came from dropping a qualifier:

• raw vs causal (union premium)
• 2.9× vs "eight times" (productivity gap)
• simulated vs observed (CTC exits)
• relative vs absolute (AI and young workers)
• 2017 draft vs published revision (Seattle)

Any Symposium 1 claim in these forms is withdrawn: "the productivity–pay gap is a myth" or "pay has decoupled from productivity," "inequality has doubled" or "inequality hasn't risen," "the minimum wage kills jobs" or "the minimum wage has no job cost," "the CTC makes people quit" or "the CTC has no work effect." None of them survives contact with the sources. Symposium 1 also appears to have treated trade policy as a settled efficiency question. This thread's record shows the pre-2016 consensus underweighted how long place-based losses persist. Ward was right about that, even though the 2026 data don't yet show that tariffs are the fix.

Bottom line. The evidence is strongest where outcomes can be measured directly: place effects in early childhood, what cash does to child poverty, how persistent trade shocks are, and how modest minimum-wage increases play out. Those findings support a Center-Left diagnosis of distribution and place. The strongest findings about barriers point to Center-Right remedies: licensing, zoning, and marriage penalties. The three consensus reforms fall exactly where those two sets of evidence overlap.

Independent Audit

Audit Panel: Prof. Laura Pennington (#97, operations research, C), lead, and Prof. Olivia Brandt (#57, statistics, C). Brandt was a referee on this thread and ruled alone on Exchange H (#42–45, #51). She recuses from every Stage 1 and Stage 4 ruling, including those. Pennington made the findings below independently. Corrections are appended. Nothing above has been deleted or edited.

1. Source check (15 claims from the Evidence Brief and the Verdict)

# · Claim · Figure in thread · Verified figure · Status · URL
1 · EPI productivity–pay gap · 93.7% vs 32.7%, 2.9× (1979Q4–2026Q2) · Same · Confirmed · epi.org/productivity-pay-gap/
2 · Winship (AEI), 1973–2022 · productivity +111%, median compensation +40%, median wages +33% · All three numbers are in the paper. But the paper's like-for-like comparison with the 111% (same net-value-added deflator) is median compensation +50%. The 40% and 33% use a cost-of-living deflator. · Minor discrepancy (see §2c) · aei.org/wp-content/uploads/2024/05/Understanding-Trends-in-Worker-Pay.pdf
3 · Chetty et al., absolute mobility · ~90% → ~50%; 80% counterfactual · Same. The 62% counterfactual is not on the landing page. We did not open the full paper. · Confirmed · opportunityinsights.org/paper/the-fading-american-dream/
4 · Moving to Opportunity (MTO) · $302k / $99k PV for a move at age 8; teens zero or negative; adults little effect · Same (3% discount rate) · Confirmed · opportunityinsights.org/paper/newmto/
5 · CBO on P.L. 119-21 · bottom decile −$1,200 (−3.1%); top decile +$13,600 (+2.7%) · Same · Confirmed · cbo.gov/publication/61367
6 · Census 2025 · median $87,460 (+2.6%, record); official 10.2%; SPM 13.1% flat; child 13.4%; Social Security lifts 28.8M · Same · Confirmed · census.gov/…/p60-290.html; stateline.org/2026/09/15/…
7 · Auten–Splinter · pre-tax 11.2 → 15.5; after-tax 8.6 → 10.3 · Same (+4.3 pp / +1.7 pp) · Confirmed · davidsplinter.com/AS-Update-2025.pdf
8 · BLS union data, 2025 · 10.0% (14.7M); public 32.9%; private 5.9%; 1983 20.1%; $1,404 vs $1,174 · Same · Confirmed · bls.gov/news.release/union2.nr0.htm
9 · Seattle (UW) · $74/month in revised version; hours −6–7%, wages +3% · Same. The abstract was revised in May 2018. The 2017 draft said $125. · Confirmed · nber.org/papers/w23532
10 · Stanford "canaries" · 19% shortfall vs less-exposed; widened from ~15% in the Aug 2025 version; −11% vs +10% · 19% as of June 2026. The paper gives the earlier kept-pace figure as 15% at the July 2025 data vintage, and says earlier versions reported 13% (Jul 2025) and 16% (Sep 2025). · Minor discrepancy (label only) · digitaleconomy.stanford.edu/…/Canaries_August2026.pdf
11 · 2025 CTC law · $2,200 indexed; $1,700 refundable; 15% over $2,500; SSN rule; ~500k children; $0–26k gain nothing · Same · Confirmed · bipartisanpolicy.org/article/how-the-obbb-changes-to-the-child-tax-credit…
12 · 2021 CTC · SPM child poverty 9.7 → 5.2; 2.9M children; 2.1M from ARPA · Same · Confirmed · census.gov/library/stories/2022/09/record-drop-in-child-poverty.html
13 · Yale Budget Lab tariffs · 11.0% statutory rate; ~0.7% on prices; ~$1,100 per household; ~$1.9T · Same · Confirmed · budgetlab.yale.edu/research/state-us-tariffs
14 · BLS, August 2026 · +162k; 4.1%; AHE +3.1%; 12-month average 31k; manufacturing +16k / +58k since Dec 2025 · Same · Confirmed · bls.gov/news.release/empsit.nr0.htm
15 · TAA · terminated Jul 1 2022; no new certifications · Same ("unable to certify new workers") · Confirmed · dol.gov/agencies/eta/tradeact

Counts: 13 Confirmed · 2 Minor discrepancy · 0 Not supported · 0 Could not access.

2. Internal consistency

a. Tally. We re-counted every row: S = 29, C = 10 (#5, 10, 16, 22, 27, 33, 34, 38, 44, 52), U = 10 (#8, 12, 15, 18, 30, 35, 40, 41, 45, 47), W = 5 (#2, 6, 25, 37, 42), total 54. This matches the posted tally. Brandt's note says the five Ws split 2 L / 1 C / 2 R, which is correct.

b. Vote math. We checked all 24 cells, and every percentage is right after rounding: 37.5% shows as 38%, 18.75% as 19%, 6.25% as 6%. Every consensus label is correct. P2 passes (Right 21/32 = 65.6%). P3 passes (Left 22/34 = 64.7%). P4 passes. P7 fails with Right 18/32, two short of 20. No mislabels. Holloway's vote on P3 is recorded as "abstained within bloc count as No". That is consistent with using the whole bloc as the denominator, but it should be stated as a general rule, not left in one footnote.

c. The Verdict misstates the five Wrong ratings (the most important correction). "Where Symposium 1 was wrong" says all five Wrong ratings came from dropped qualifiers, and its list includes "raw vs causal (union premium)". The union premium (#8) was rated U, not W. The list also leaves out an actual W: Venkataraman's "doubled under every serious measure" (#2). Corrected list of the five Ws: "doubled under every serious measure" (pre-tax vs after-tax / series choice), "eight times" vs 2.9×, simulated vs observed (CTC exits), relative vs absolute (AI and young workers), and 2017 draft vs published revision (Seattle). The "eight times" error is also a wrong number, not a dropped qualifier, so "every one came from dropping a qualifier" should read "four of five."

d. Winship's comparison mixes deflators. Brief #3, Wren's conceded tradeoff, fact-check #33 and the Right-steelmans-Left section all set "median compensation +40%" against "productivity +111%". In Winship's own like-for-like comparison, using the same deflator, median compensation rose 50% against 111%. Correction: use 50% vs 111% when comparing with productivity. 40% is the cost-of-living-deflated figure. The conclusion that dispersion is real still holds. The gap is somewhat smaller than the thread says.

e. Rating consistency. Kowalski's "controlled studies still find a meaningful union premium" (#10) was "not verified this round" but rated C. Three claims in the same position were rated U: Wren on expensing (#35), and Stroud on Healthy Marriage (#40) and Trump accounts (#41). By the thread's own definitions (U = "not established by the record; may be true"), #10 should be U. Correction: #10 is U. The revised tally is 29 S / 9 C / 11 U / 5 W. None of these four claims carried weight in the Verdict.

f. Roster. Marsh is #48, education policy, R, and is correct. Pruitt (#8, Lib) and Wainwright (#100) are consistent. Castellano served as both referee and discussant, and the thread disclosed her recusal on Exchange H.

3. Balance

The five Ws split evenly (2 L, 1 C, 2 R) and every one matches its source. The imbalance is in the U ratings. The Right collected 6 of 10 (#30, 35, 40, 41, 45, 47), and three of those were given only because the referees did not check them. A Left claim in the same position got C (§2e). This is a small procedural tilt against the Right. We found no tilt in substance: the Verdict pairs the Center-Left diagnosis (distribution, place, cash for children) with Center-Right barrier remedies (licensing, zoning, marriage penalties), and all three consensus reforms are cross-bloc. Established #8 (moderate minimum-wage increases) rests on Cengiz et al. alone, but Pruitt conceded the point for moderate increases, so it is fairly placed.

4. Grade: B+

The sources are excellent: 13 of 15 confirmed, and the vote and tally arithmetic has no errors. We deducted because the Verdict's list of the five Wrong ratings is itself wrong (it includes a U and drops a W), the Winship comparison mixes deflators, and "not verified" claims were rated inconsistently across blocs.

— L. Pennington (#97), lead; O. Brandt (#57), recused on Stage 1 and Stage 4 rulings, including Exchange H

Verdict v2 (audit-corrected)

Prof. Adelaide Wainwright (#100), moderator. This post supplements the Stage 8 verdict (post 245) and the Independent Audit (post 250). It does not edit either. I re-checked every open item on 2026-09-24. Ratings use the standard scale: Supported / Contested / Unsupported (includes "could not verify") / Wrong.

1. Resolution of unverified and audit-flagged items

Item · Origin · Result · Figure and source
Fed DFA wealth shares · Brief #12, Verdict "Unknown" · Now verified · Q2 2026 shares of household net worth: top 1% 32.5% (top 0.1% 15.0%), 90–99th 36.4%, 50–90th 28.8%, bottom 50% 2.3%. The top-1% share was 31.1% a year earlier and 22.8% in 1989 Q3. FRED DFA release table, Q2 2026; FRED series WFRBST01134
Kearney's tabulations · C7, #38 · Now verified (figures) · The share of children living with married parents fell from 77% to 63% (1980–2019). By mother's education: college 90% → 84%, high school 83% → 60%, less than high school 80% → 57%. Aspen ESG, Kearney. How much of the gap is causal remains Contested.
Marriage-penalty sizes · C8, P4 · Now verified (size); behavioral effect still unknown · In Fragile Families data, 40–50% of parents of 1-year-olds faced an EITC penalty of $1,200–$2,100 (2001 $). Bonuses of $400–$500 reached about half as many parents (Rachidi, IFS, Dec 2015). In a worked example, a $16k single mother who marries a $25k earner loses ~90% of her EITC, about $2,800 (Urban, 2013). Neither source finds a measured behavioral effect.
Controlled union wage premium (Kowalski #10; also bears on #8) · Rated C, then U by the audit · Now verified · Regression-adjusted union–nonunion gap: ~11.8% (2020) and ~13.5% (2019), the lowest since 1973. It has declined with density. The private-sector gap is roughly double the public-sector gap (Macpherson & Hirsch, IZA DP 14398). The raw BLS gap is 19.6%.
China shock "~2.4M jobs" (Haddad #18) · U · Now verified · "Net job losses of 2.0 to 2.4 million" from rising Chinese import competition, 1999–2011 (Acemoglu, Autor, Dorn, Hanson & Price, NBER w20395).
Full expensing permanent (Wren #35) · U · Now verified · OBBBA "permanently restores 100% bonus depreciation for qualified property acquired and placed in service after Jan. 19, 2025" (Grant Thornton).
Healthy Marriage evaluations disappointing (Stroud #40) · U · Now verified · Building Strong Families at 36 months: no effect on marriage (21% vs 21%). Treatment couples were less likely to remain romantically involved (57% vs 60%). Oklahoma City was the one positive site (OPRE 2012-28A).
"Trump accounts" exist (Stroud #41) · U · Now verified · Created by OBBBA. There is a $1,000 pilot contribution for children born 2025–2028. More than 4M children were enrolled and 1M+ pilot elections made as of Mar 31, 2026 (IRS IR-2026-42; Fed. Reg. 2026-04534).
CMTO move rates (P7) · "not verified" · Now verified · Moves to high-upward-mobility areas rose from 15% (control) to 53% (treatment) (Bergman et al., NBER w26164).
Wage-insurance take-up (P2) · "magnitude not verified" · Now verified (historical) · In two-thirds of states, ≤5% of TAA participants received wage insurance (ATAA) in FY2006. At the case-study sites, no more than 1 in 5 of potentially eligible workers received it (GAO-07-994T).
Chetty "62%" counterfactual · Audit: not on landing page · Partly verified · The abstract says growth alone "cannot restore" 1940s mobility and a 1940-style distribution "would reverse more than 70% of the decline" (NBER w22910). That is consistent with the 80% counterfactual. The exact 62% remains unseen.
Winship comparison · Audit §2d · Corrected · Like-for-like, with the same deflator: median compensation +50% vs productivity +111%. The +40% figure is cost-of-living-deflated.
Canaries "~15% in Aug 2025 version" · Audit · Corrected (label) · 15% is the figure at the July 2025 data vintage. Earlier versions reported 13% (Jul 2025) and 16% (Sep 2025). 19% as of June 2026 stands.
Holloway "~$100B/yr" (#15); UW multi-site claim (#45) · U, withdrawn · Still unverifiable · Not retrieved. Both were withdrawn on the record.
California $20 fast-food evidence (P6) · "not in this record" · Still unverifiable · Not attempted within the search budget. It is disputed in the literature.

2. Rating normalization and revised tally

Changes, starting from the audit-corrected baseline (29 S / 9 C / 11 U / 5 W, with #10 moved to U):

• #10 (Kowalski): "controlled studies still find a meaningful premium" goes U → S (~12%).
• #8 (Kowalski): "the union wage premium is ~20%" goes U → W. The economy-wide regression-adjusted premium is ~12%. The ~20% is the raw gap, which BLS flags as unadjusted.
• #18 (Haddad): U → S.
• #35 (Wren): U → S.
• #40 (Stroud): U → S.
• #41 (Stroud): U → S.
• #38 (Stroud, Kearney) stays C. The figures are now verified, but the causal claim is still disputed.

Tally: 29 S / 10 C / 10 U / 5 W (as posted) → 29 / 9 / 11 / 5 (audit) → 34 S / 9 C / 5 U / 6 W (54 claims).

The audit found a procedural tilt: three Right claims were U only because they had not been checked. That tilt is resolved, since all three are now S. Remaining Right U ratings are #30, #45 and #47, all withdrawn. Left Wrongs are now #2, #6 and #8.

3. Corrected verdict

Established

1. Absolute mobility fell from ~90% (1940 cohort) to ~50% (1980s cohort). Most of the fall is due to distribution. A 1940-style distribution would reverse more than 70% of the decline (#13).
2. Place has causal effects, and they are largest in early childhood.

• MTO: ~$99k present value for a child who moves at age 8, and nothing for teens (#14).
• CMTO counseling raised moves to high-opportunity areas from 15% to 53%.

1. Pre-tax income concentration rose on every serious series (AS pre-tax +4.3 pp), and CBO's Gini rose both before and after taxes and transfers. After-tax income grew for every quintile.
2. Wealth concentration is high and rising. (New; this was Unknown.)

• Top 1% share of net worth: 32.5% in Q2 2026, up from 31.1% a year earlier and 22.8% in 1989.
• Bottom 50% share: 2.3%.
• Source: Fed DFA.

1. Median pay has lagged average productivity, but the link is not broken. (Corrected.)

• Like-for-like, median compensation is +50% vs productivity +111% (Winship 1973–2022). The +40% figure uses a cost-of-living deflator.
• EPI's figure is 2.9× (1979–2026).
• The gap has three components: pay inequality, divergence between the price deflators, and a smaller decline in labor share. At the margin, productivity growth still passes through to pay.

1. The China shock caused lasting place-based damage: 2.0–2.4M net jobs lost 1999–2011, and 86% of it showed up as nonemployment. TAA did little. New: its wage-insurance component historically reached ≤5% of participants in most states.
2. The 2021 CTC cut SPM child poverty from 9.7% to 5.2%.
3. The 2025 law is regressive in how it shifts resources, by CBO's estimate: bottom decile −3.1% of income, top decile +2.7%. It also made 100% expensing permanent and created Trump accounts.
4. Moderate minimum-wage increases have not measurably reduced the number of low-wage jobs (#22).
5. Institutions and demographics:

• Union density is 10.0% overall and 5.9% in the private sector.
• The controlled union wage premium is ~12% economy-wide (2020), not ~20%.
• About one worker in five needs a government license.
• The GFR is at a record low.
• Children living with married parents fell from 77% to 63% (1980–2019). The fall was steepest among children of non-college mothers: 83% → 60% for high school.

1. Government marriage-promotion programs have not raised marriage. Building Strong Families had no effect at 36 months.

Contested

• After-tax top-income growth (AS vs PSZ).
• Labor supply under a permanent refundable CTC.
• High minimum wages relative to local pay.
• The 2025–26 tariffs and manufacturing.
• How much of the family-structure gap is causal. Kearney's figures are verified. Selection is still disputed.
• Whether AI is the cause of the entry-level gap.

Unknown

• Whether marriage penalties change behavior. Their size is now documented: $1,200–$2,100 (2001 $) for 40–50% of new parents in Fragile Families. Their behavioral effect is not.
• Fertility effects of policy at U.S. scale.
• Scale effects of mobility vouchers. Long-run effects of sectoral wage boards, including California's $20 fast-food floor.

Correction to "Where Symposium 1 was wrong." The list of Wrong ratings is now six:

1. "Doubled under every serious measure" (pre-tax vs after-tax; choice of series).
2. "Eight times" vs 2.9× (a wrong number).
3. "~20% union premium" vs ~12% regression-adjusted (raw vs causal).
4. Simulated vs observed CTC exits.
5. Relative vs absolute (AI and young workers).
6. 2017 draft vs published revision (Seattle).

Five of the six came from dropping a qualifier. "Every one" was an overstatement.

Conclusions that changed

• (a) Wealth concentration moves from Unknown to Established. This does not resolve the AS–PSZ income dispute. It does undercut any reading of Auten–Splinter as "concentration didn't rise."
• (b) The union evidence narrows. A real but moderate premium (~12%) weakens the magnitude of Kowalski's power story. It leaves the direction intact, and it bears on P6.
• (c) P2 (wage insurance) gets a verified risk. Historical take-up was very low, so automatic enrollment matters.
• (d) P4 (marriage penalties) gets a documented size but still no demonstrated behavioral payoff. Its case now rests on fairness more than on incentives, as Stroud herself anticipated.
• (e) The consensus results recorded in Stage 7 are unchanged.

— A. Wainwright (#100)