12 — Taxes & the Tax Code

STAGE 1 — EVIDENCE BRIEF

Referee panel: Prof. Olivia Brandt (#57, statistics, C) · Prof. June Takahashi (#17, econometrics, C) · domain specialist Prof. Beatrice Lang (#13, accounting, C)

Scope. Thread 02 already handled deficits and entitlements, so this thread is about how the tax system is designed: who pays, what the base leaves out, how well the IRS collects, how capital and corporations are taxed, tariffs treated as taxes, and the alternatives (VAT, wealth tax). Today is 2026-09-24. Every figure below is given as the cited source states it, with the year of the data. Where we could not verify a number we say so.

---

A. Progressivity: who pays now

1. In 2022, the average federal tax rate (all federal taxes) was 20.6% for all households. It was −10% for the lowest quintile, which is negative because refundable credits exceed taxes paid. Rates for the bottom four quintiles rose in 2022 because the pandemic Recovery Rebates and the expanded CTC expired. — CBO, The Distribution of Household Income, 2022 (Jan 2026). https://www.cbo.gov/publication/62300 · PDF https://www.cbo.gov/system/files/2026-01/61911-Household-Income-2022.pdf
2. The top 1% paid 27% of all federal taxes in 2022, up from 14% in 1979. Their share of income before transfers and taxes rose from 9% to 18% over the same period. — CBO (same report, 2026).
3. **Federal income tax alone, TY2022:** the top 1% paid 40.4% of income taxes on 22.4% of AGI, at an average rate of 26.1%. The bottom 50% paid 3.0% of income taxes on 11.5% of AGI, at an average rate of 3.7%. — Tax Foundation summary of IRS SOI data (2025). https://taxfoundation.org/data/all/federal/latest-federal-income-tax-data-2025/
4. Measuring the top is contested. Piketty–Saez–Zucman have the top 1% after-tax income share rising from 9% (1960) to 15% (2019). Auten–Splinter find only about a 1-point rise, to 9%. Most of the gap comes from how each allocates unreported income (~2 pts), government consumption (~1.3 pts) and deficits. — Brookings primer. https://www.brookings.edu/articles/measuring-income-inequality-a-primer-on-the-debate

B. The 2025 reconciliation law (P.L. 119-21, "OBBBA", signed July 4, 2025)

1. Main parameters: the TCJA individual rates are made permanent (top rate 37%). CTC is $2,200, indexed. The SALT cap is $40,000 for 2025–2029, rising 1%/yr, with an income-based phase-down, and it reverts to $10,000 in 2030. The tips deduction is up to $25,000 and the overtime deduction up to $12,500 (premium portion only), both for 2025–2028 and both phasing out above $150k/$300k. The senior deduction is $6,000 per person aged 65+, phasing out at 6% above $75k/$150k, for 2025–2028. 100% bonus depreciation and domestic R&D expensing are permanent. The §199A 20% QBI deduction is permanent. The estate exemption is $15M, indexed. — Tax Foundation FAQ (2025). https://taxfoundation.org/research/all/federal/one-big-beautiful-bill-act-tax-changes/
2. Budget effect: +$3.4 trillion in deficits over 2025–2034 (revenues −$4.5T, direct spending −$1.1T), measured against CBO's January 2025 baseline. — CBO (2025). https://www.cbo.gov/publication/61570
3. Distribution including spending cuts (CBO): averaged over 2026–2034, resources fall about $1,200/yr (−3.1% of income) for the lowest decile. They rise about $800 (+0.8%) for the 5th decile and about $13,600 (+2.7%) for the top decile. The losses at the bottom come mostly from the Medicaid and SNAP cuts. — CBO letter (Aug 2025). https://www.cbo.gov/system/files/2025-08/61367-Distributional-Effects.pdf
4. Tax provisions only (TPC, 2026 vs. current law): after-tax income rises by +0.8% ($150) for the bottom quintile, +2.3% ($1,780) for the middle, +3.4% ($12,540) for the top quintile and +3.5% ($75,410) for the top 1%. Including debt service, the total deficit increase is $4.1T. Once the spending cuts and the financing are counted, about 60–70% of households are worse off, depending on how financing is assumed to be allocated. — Gale & Page, TPC preliminary assessment (Mar 2026). https://taxpolicycenter.org/sites/default/files/2026-03/OBBBA_Preliminary_Assessment_2026-03-23.pdf
5. **Measured against current policy (i.e., treating the TCJA extension as already given):** about one-third of households get no additional cut, almost half get less than $100, and about 20% get more than $1,000. Half of the top quintile get at least $1,000. The top 1% gain relatively little beyond the TCJA extension. — Yale Budget Lab (July 30, 2025). https://budgetlab.yale.edu/research/distribution-tax-cuts-new-tax-law
6. Growth estimates diverge. Tax Foundation projects a +1.2% larger long-run economy (source in #5). Yale Budget Lab counts debt crowd-out and finds real GDP 3.3% lower in 2054, debt/GDP at 194% vs 142% without the law, and the 10-year yield +1.4 pp. — Yale (2025). https://budgetlab.yale.edu/research/long-term-impacts-one-big-beautiful-bill-act-enacted-july-4-2025
7. Complexity: Tax Foundation estimates 7.1 billion hours of compliance time per year and $536B in monetized compliance cost (2025). It puts the cost of the new carve-outs at: overtime −$145.9B, senior −$134.8B, tips −$30.8B, car-loan interest −$28.8B (2025–2034), and adds that OBBBA creates about 6.9 million more itemizers than a clean TCJA extension would. — Tax Foundation (2025). https://taxfoundation.org/research/all/federal/obbba-income-tax-complexity-tax-breaks/

C. The tax base: tax expenditures

1. JCT puts tax expenditures at $2.3T in FY2026 ($11.7T over 2025–2029). The largest items in FY2026 are:

• retirement exclusions $355B
• preferential rates on capital gains and dividends $252B
• the employer health-insurance exclusion $240B
• CTC $128B
• ACA credits $105B
• charitable deduction $78B
• §199A $76B
• step-up in basis at death $73B
• EITC $67B
• SALT $60B

The mortgage interest deduction no longer makes the top ten, because the TCJA raised the standard deduction and capped the loan size. — CRFB summary of JCX-45-25 (2025/26). https://www.crfb.org/blogs/jct-projects-tax-expenditures-will-be-23t-2026

1. The JCT and Treasury figures for the health exclusion differ. For example, Treasury projected $252.4B for 2024, and part of the difference comes from how each counts the payroll-tax effect. — TPC Briefing Book. https://taxpolicycenter.org/briefing-book/what-are-largest-tax-expenditures

D. Tax gap and enforcement

1. The tax gap for TY2022 is $696B gross and $606B net. The voluntary compliance rate is about 85%. Of the $381B individual underreporting gap, $179B (47%) comes from income with little or no information reporting, including $117B from nonfarm sole proprietors. — IRS Pub. 5869 (Oct 2024). https://www.irs.gov/pub/irs-pdf/p5869.pdf; CRFB summary https://www.crfb.org/blogs/irs-estimates-606-billion-tax-gap-2022
2. Audit returns by income: Boning, Hendren, Sprung-Keyser & Stuart (QJE 2025) find that audit spending on above-90th-percentile taxpayers returns over $12 per $1, against about $5 below the median. The average initial revenue is $2.17 per $1. Specific deterrence (higher reported tax in later years) produces at least 3× the initial audit revenue. — NBER w31376. https://www.nber.org/papers/w31376
3. The IRS's own ROI re-estimate: $390B (old method) versus $851B (broader method) over FY2024–2034 from IRA-era investment. — IRS Pub. 5901 (Feb 2024). https://www.irs.gov/pub/irs-pdf/p5901.pdf
4. The 2025 staffing collapse: examination and collection staff fell from 27,217 (FY2024) to 19,612 (FY2025), a 27% drop, and to 17,517 by Jan 10, 2026. Individual exam starts fell 30%. Audits of returns over $400k fell 27%. Enforcement revenue went from $98.7B to $93.8B. — TIGTA via The Tax Adviser (Aug 2026). https://www.thetaxadviser.com/news/2026/aug/irs-enforcement-activity-fell-despite-record-tax-collections-tigta-says/
5. Yale estimates the IRS cuts reduce revenue by ~$861B over 2026–2035, based on 27,636 fewer employees and more than 3,600 revenue agents lost (~31% of audit staff). — Yale Budget Lab. https://budgetlab.yale.edu/research/weakened-irs-has-substantial-consequences
6. Direct File had about 140,803 users in the 2024 pilot and expanded to 25 states in 2025. The IRS told states in Nov 2025 that it would not be available for 2026, citing cost and low uptake. The 2025 usage count is [unverified] by this panel. — Tax Notes (Nov 5, 2025) https://www.taxnotes.com/featured-news/irs-shutters-direct-file-citing-cost-and-low-uptake/2025/11/05/7t7q0; Wikipedia summary https://en.wikipedia.org/wiki/IRS_Direct_File

E. Capital gains, step-up, "buy-borrow-die"

1. Fox & Liscow (J. Public Econ., 2025):

• For the top 1%, the income tax base captures only ~60% of economic income (71% after adjusting for inflation), and about half for the top 0.1%.
• New borrowing each year is only 1–2% of economic income, while unrealized gains are 20–40× larger.
• The dominant pattern is therefore "buy, save, die", not "buy-borrow-die".

https://www.sciencedirect.com/science/article/abs/pii/S0047272725002178; TaxVox (June 15, 2026) https://taxpolicycenter.org/taxvox/richs-real-tax-trick-isnt-buy-borrow-die

1. CBO options (estimated before OBBBA, 2026–2035):

• realize gains at death: $570B
• carryover basis: $230B
• cap the health exclusion at the 50th percentile of premiums: $1.2T; at the 75th percentile: $630B
• carried interest as ordinary income: $15B
• +1 pt on the corporate rate: ~$140B
• broaden the NIIT: $440B

— CRFB summary of CBO. https://www.crfb.org/blogs/cbos-revenue-savings-options

1. Carried interest: CBO/JCT scored ordinary-income treatment at $14.0B over 2019–2028 (Dec 2018; https://www.cbo.gov/budget-options/54795). Yale Budget Lab (May 4, 2026) used new partnership capital-account data and re-estimated the Wyden-Whitehouse-King bill at $87.7B (previously $47.5B). It puts a broad version at about $100B. https://budgetlab.yale.edu/research/refining-revenue-estimates-taxing-carried-interest

F. Corporate tax

1. Chodorow-Reich, Zidar & Zwick (JEP 2024 / NBER w32672):

• Firms with the mean tax change increased domestic investment 20%.
• Aggregate tangible corporate investment rose 11%, and the long-run corporate capital stock ~7%.
• The predicted long-run wage gain is ~$750 per worker, "an order of magnitude below" the 2017 CEA's $4,000–$9,000.
• Corporate revenue fell ~40%, and the investment response was "far too small" to offset the cost.

https://www.nber.org/papers/w32672

1. Kennedy, Dobridge, Landefeld & Mortenson (AER, Sept 2026): the TCJA raised firm investment, sales, employment and payroll, but 87% of short-run private income gains went to the top 10%. https://aeaweb.org/articles?id=10.1257/aer.20240404
2. Global minimum tax:

• June 28, 2025: a G7 statement endorsed a "side-by-side" system that fully excludes US-parented groups from the IIR and UTPR, conditional on removing §899 from H.R. 1. §899 was then dropped. — Treasury. https://home.treasury.gov/news/press-releases/sb0181
• Jan 5, 2026: the OECD Inclusive Framework adopted the side-by-side package. Foreign domestic minimum top-up taxes (QDMTTs) still apply to US firms' foreign subsidiaries, and the safe harbor depends on the US keeping GILTI/CAMT. — FACT Coalition. https://thefactcoalition.org/policy-brief-oecd-side-by-side-system/

G. Tariffs as taxes

1. Incidence: the NY Fed estimates that nearly 90% of the 2025 tariff burden fell on US firms and consumers through Nov 2025. Foreign exporters' share rose from 6% (Jan–Aug) to 14% (Nov). — Liberty Street Economics (Feb 2026). https://libertystreeteconomics.newyorkfed.org/2026/02/who-is-paying-for-the-2025-u-s-tariffs/
2. Yale's one-year retrospective (Apr 2, 2026) cut its estimate of the short-run price effect from 2.3% to 0.5–1.0%. Its estimated pass-through to consumers is 40–76%. The effective tariff rate averaged about 14.8% (daily, Apr–Dec 2025). https://budgetlab.yale.edu/research/one-year-tariff-analysis-what-we-got-right-what-changed-and-what-we-learned
3. ***Learning Resources v. Trump* (Feb 20, 2026), decided 6–3: IEEPA does not authorize tariffs. About $166B in IEEPA duties is being refunded through the CAPE system, launched Apr 20, 2026. Replacement §122 tariffs (15%, 150 days) were subsequently challenged. The reported May 2026 CIT ruling against them is from a secondary source (Wikipedia) and is not independently verified** by this panel. https://www.supremecourt.gov/opinions/25pdf/24-1287_4gcj.pdf; https://en.wikipedia.org/wiki/Learning_Resources,_Inc._v._Trump

H. Alternatives: VAT and wealth tax

1. CBO, 5% VAT, 2025–2034: a broad base raises $3.38T. A narrow base (excluding food, health, new housing and postsecondary education) raises $2.18T. Both figures are net of the income/payroll-tax offset. — CBO (Dec 2024). https://www.cbo.gov/budget-options/60961
2. Wealth taxes: the number of OECD countries levying a net wealth tax fell from 12 (1990) to 4 (2017). In 2023, Switzerland raised 1.16% of GDP (4.3% of tax revenue), Spain 0.21% of GDP and Norway 1.5% of tax revenue. — OECD data via Euronews (Sept 2025). https://www.euronews.com/business/2025/09/29/wealth-taxes-in-europe-who-collects-them-and-how-much-do-they-raise

---

Contested evidence: where the literature genuinely disagrees

• Top income shares (PSZ vs. Auten–Splinter, #4). Both sides agree inequality rose. They disagree on how much, and how much taxes and transfers offset it.
• OBBBA growth (Tax Foundation +1.2% vs. Yale −3.3% by 2054, #10). The difference is mostly about debt crowd-out and about whether temporary provisions get extended. Neither is a forecast of observed data.
• Corporate tax incidence. CZZ (#23) and Kennedy et al. (#24) agree investment rose and that the worker gains were far smaller than CEA 2017 projected. Some still argue the long-run capital-deepening effect is under-measured in a 5–7 year window.
• Enforcement ROI (#15–18). Marginal returns fall as audit coverage scales. The IRS $851B figure includes IT-modernization and service gains that are harder to verify than direct exam revenue. The Yale $861B loss estimate is model-based.
• Tariff pass-through (#26–27). The NY Fed measures import-price incidence; Yale measures consumer-price incidence. Both can be true at once: importers absorbed part of the cost in their margins, at least temporarily.
• Wealth-tax efficacy. European repeals (#30) are consistent with both "capital flight and administrative failure" and "design failure (low thresholds, exemptions)". Saez–Zucman argue a US tax with a high threshold and no exemptions would differ. Critics (e.g., the OECD 2018 report) favor taxing capital income and inheritances instead.
• Buy-borrow-die (#20). This is the newest evidence (Fox–Liscow 2025/26). It is one study using SCF/IRS-linked data, and replication is pending.

— Brandt, Takahashi, Lang

STAGE 2 — OPENING POSITIONS (Part 1 of 2)

---

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

Causal claim. The federal system is progressive at the bottom and in the middle. It goes soft at the very top because the income tax does not reach the largest source of top-end income: unrealized appreciation. OBBBA then shifted resources upward.

Evidence.

• CBO: over 2026–2034 the lowest decile loses about $1,200 a year (−3.1%) while the top decile gains about $13,600 (+2.7%) (Brief #7).
• TPC: the top 1% get $75,410 in 2026 from the tax provisions alone (#8).
• JCT: the step-up in basis costs $73B a year (#12).
• On the mechanism, I'll say it bluntly: buy-borrow-die is the central tax-avoidance strategy of American billionaires. They buy appreciating assets, borrow against them to fund consumption, and die with the step-up wiping out the gain.

Remedy.

1. Tax gains at death above a $5M per-person exemption, with deferral for family farms and businesses (CBO: ~$570B over 10 years, #21).
2. A minimum tax on very high incomes that counts unrealized gains.
3. Restore IRS enforcement funding.

Concession. Realization at death creates liquidity and valuation problems for illiquid closely-held businesses. A deferral regime adds complexity, and some revenue leaks through trusts.

Falsifiable prediction. If the income base really misses a large part of top-end income, then when SOI data for TY2026–27 arrive we should see top-0.1% reported AGI grow more slowly than Fed Distributional Financial Accounts wealth for the same group. If the two track each other, my mechanism is weaker than I think.

---

Prof. Rosa Delgado-Finch (#7, political economy, Soc)

Causal claim. Concentrated wealth is concentrated political power, and an income tax cannot correct a stock problem. We need an annual tax on the stock of wealth.

Evidence.

• Switzerland raises 1.16% of GDP from its wealth tax, 4.3% of all tax revenue (#30). That shows a well-run wealth tax is administrable at scale.
• The European repeals (12 countries down to 4) were driven by lobbying by the wealthy, not by capital flight. The taxes were riddled with exemptions because the rich wrote them.

Remedy. A 2% annual tax on net worth above $50M and 3% above $1B. Use a comprehensive base with no asset exemptions, plus a 40% exit tax on renunciation and third-party valuation reporting.

Concession. Valuing private businesses and art every year is hard. Revenue estimates depend heavily on avoidance elasticities, which Saez–Zucman put low and critics put high.

Falsifiable prediction. If a US-style high-threshold wealth tax is administrable, Norway's and Spain's taxable-wealth bases should not show migration-driven erosion of more than 10% within 5 years of their recent rate increases. Larger erosion would count against me.

---

Prof. Miriam Castellano (#1, labor economics, C-L)

Causal claim. The cheapest revenue in America is tax that is already legally owed. Gutting the IRS in 2025 was a revenue cut disguised as a spending cut.

Evidence.

• The net tax gap is $606B a year (TY2022). Nearly half of individual underreporting is in income with little information reporting (#14).
• Audits of the top 10% return more than $12 per $1 (#15).
• Exam and collection staff fell 27% in one year, and audits of returns over $400k fell 27% (#17).
• CBO has confirmed the 2025 IRS cuts will cost about $861B over ten years.
• Separately, killing Direct File took away a free public option that 86% of pilot users said raised their trust in the IRS.

Remedy.

1. Multi-year mandatory funding for enforcement, with a statutory floor on revenue agents and targeting guidance above $400k.
2. Information reporting on business receipts (for example, a lower 1099-K threshold).
3. Restore Direct File.

Concession. Marginal returns fall as coverage rises, and heavy-handed enforcement erodes trust. The IRA's 2022 rhetoric overpromised on how fast the IRS could hire.

Falsifiable prediction. If staffing drives collections, enforcement revenue should fall further in FY2026–27 (below FY2025's $93.8B in nominal terms) even as total receipts grow.

---

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

Causal claim. America's tax problem is the base, not the rates. We have $2.3T a year of tax expenditures, and they mostly subsidize the already-comfortable: owner-occupiers, people with generous employer health plans, and high-bracket savers.

Evidence.

• JCT puts tax expenditures at $2.3T in FY2026 (#12).
• The employer health exclusion alone costs $240B a year.
• The mortgage interest deduction is still one of the top five tax expenditures. It inflates house prices in exactly the high-cost metros where we need supply.

Remedy. A classic broaden-base-lower-rates swap:

• Cap the health exclusion at the 75th percentile of premiums (CBO: ~$630B over 10 years, #21).
• Turn the mortgage interest deduction into a capped, refundable first-time-buyer credit.
• Let the tips, overtime and senior deductions sunset in 2028 as scheduled, and put the savings into a higher standard deduction.

Concession. Capping the health exclusion hits older workforces and high-cost regions, and unions will fight it. Base-broadening coalitions historically fall apart because every carve-out has a lobby.

Falsifiable prediction. If the carve-outs are extended past 2028, tax expenditures as a share of GDP will be higher in JCT's 2029 report than in the 2025 report.

---

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

Causal claim. The 2025 tariffs were the largest tax increase in decades, and they were imposed without a vote of Congress. As taxes go they are regressive, inefficient and unstable.

Evidence.

• The NY Fed finds about 90% of the burden fell on US firms and consumers (#26).
• The effective tariff rate at the April 2025 peak announcement was 22.4%, the highest since 1909 (#27).
• Consumers have borne the full cost, and consumer prices are about 2.3% higher as a result — roughly $3,800 per household.
• The Supreme Court held 6–3 that IEEPA did not authorize them (#28).

Remedy.

• Any tariff raising revenue above a de minimis threshold should need express congressional enactment within 150 days.
• JCT and CBO should score tariffs in the same distributional tables as income taxes.

Concession. Short-run pass-through to consumers has so far been partial, and some tariffs (on Chinese strategic goods) have a national-security rationale that is not about revenue.

Falsifiable prediction. If tariffs are a consumption tax on Americans, then once importers' margins are exhausted, core goods PCE inflation should stay above its 2017–19 trend through 2027 even with the IEEPA duties refunded.

STAGE 2 — OPENING POSITIONS (Part 2 of 2)

---

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

Causal claim. The 2017 corporate reform (21% rate plus full expensing) worked. It fixed an internationally uncompetitive rate, raised investment, and OBBBA was right to make expensing permanent. Taxing capital is the most growth-destructive way to raise revenue.

Evidence.

• Chodorow-Reich, Zidar & Zwick find investment rose 20% at the average treated firm and aggregate corporate investment rose 11% (#23).
• Kennedy et al. find firm sales, employment and payroll all rose (#24).
• The TCJA raised household wages by roughly $4,000, as the CEA predicted.
• Corporate inversions stopped after 2017.

Remedy.

• Keep 21%.
• Keep 100% expensing and R&D expensing permanent (done).
• Defend the Pillar Two side-by-side deal.
• Resist any rise in the capital-gains rate.

Concession. The reform was not self-financing at a static score, and corporate receipts fell sharply in 2018–19.

Falsifiable prediction. With expensing now permanent, nonresidential fixed investment as a share of GDP should run at least 0.5 pp above its 2015–2019 average over 2026–2029. If it doesn't, the "permanence premium" argument is wrong.

---

Prof. Raymond Kessler (#18, macroeconomics, R)

Causal claim. The US system is already the most progressive income-tax system in the OECD. The problem is not "fair share"; it is an income tax that double-taxes saving. We should move toward a consumption base.

Evidence.

• The top 1% paid 40.4% of federal income taxes on 22.4% of AGI, and the bottom 50% paid 3% (#3).
• Across all federal taxes the top 1% pay 27% (#2).
• OBBBA will raise long-run GDP by about 1.2% (Tax Foundation, #10). The Yale negative number is an artifact of assuming the debt is never addressed.

Remedy. A phased move to a progressive consumption tax: a cash-flow business tax plus an X-tax-style progressive wage tax. It would replace the corporate income tax and the taxes on capital gains and dividends. I oppose a VAT on top of the income tax; that is a money machine.

Concession. A consumption base is less progressive at the very top unless bequests are taxed. The transition is brutal for existing wealth held in already-taxed forms.

Falsifiable prediction. If saving is as tax-elastic as I claim, household saving rates should respond measurably to the permanence of lower rates. Personal saving should hold at or above 5% of disposable income through 2028.

---

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

Causal claim. The tax code's biggest cost is not its rates but its complexity. Every carve-out (tips, overtime, seniors, SALT) is a political favor paid for in compliance hours and lost trust.

Evidence.

• Americans spend 7.1 billion hours a year on compliance, costing $536B (#11).
• OBBBA adds 6.9M itemizers.
• Direct File cost taxpayers more per return than the private Free File program would have, so ending it was reasonable.
• A 38,000-page code (a figure I'll concede is folklore) rewards those who can afford lawyers, which is itself regressive.

Remedy.

• A single-rate income tax with a large standard deduction, no itemized deductions and no carve-outs.
• Repeal the tips, overtime, senior and car-loan provisions now rather than in 2028.
• Oppose wealth taxes: they require an intrusive annual valuation state.

Concession. A flat rate with a large exemption is progressive in average rates but less so at the top, and the transition hurts itemizers in high-tax states.

Falsifiable prediction. If complexity is the binding cost, the new deductions should show low take-up among eligible low-income filers (below 60% of eligible tipped workers claiming in TY2025) and heavy use of paid preparers.

---

Prof. Martin Oyelaran (#12, finance, C-R)

Causal claim. The left's capital-tax story rests on myths. "Buy-borrow-die" is mostly folklore, and carried interest is a symbolic loophole worth almost nothing. The real distortion is lock-in: the step-up makes investors hold assets until death and misallocates capital.

Evidence.

• Fox & Liscow show the top 1% borrow only 1–2% of economic income a year (#20).
• Taxing carried interest as ordinary income raises a trivial ~$14B over a decade (CBO/JCT, #22).
• The step-up costs $73B a year (#12). That is real money, and it arises precisely because it rewards not selling.

Remedy. A deal: replace the step-up with carryover basis (CBO: $230B, #21) and index capital-gains basis to inflation prospectively. Leave carried interest alone as a rounding error.

Concession. Carryover basis was enacted in 1976 and repealed before it took effect because the records problems were real. Indexation creates arbitrage with nominal interest deductions.

Falsifiable prediction. If lock-in is real, realizations by holders aged 75+ should rise measurably in the years after any step-up repeal, relative to younger holders.

---

Nine openings recorded. Cross-examination follows. — Moderator Wainwright (#100)

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

---

Exchange 1 — Oyelaran (C-R) challenges Venkataraman (L) on "buy-borrow-die" (four rounds)

Oyelaran (R1): Priya, you called buy-borrow-die "the central tax-avoidance strategy of American billionaires." Fox & Liscow (J. Pub. Econ. 2025) built the best dataset we have on this. The top 1% borrow 1–2% of economic income per year, and their unrealized gains over the same period are 20–40 times larger (Brief #20). Borrowing is a sideshow. They don't need to borrow, because they already have plenty of taxable income to live on.

Venkataraman (R1): Fox–Liscow cover the top 1%, a group of 1.3 million households. "Billionaires" are about 800 people. Borrowing may well be more concentrated at the extreme top, where wage and dividend income is smallest relative to wealth. The ProPublica-leaked cases were billionaires, not dentists in the 99th percentile.

Oyelaran (R2): Then you're conceding that your claim rests on anecdotes about a few hundred people, not on data. Fox–Liscow also look at the top 0.1% and find the base captures about half of economic income there. The gap grows as you go up, but the mechanism they identify is still saving, not debt. You'd need a dataset of billionaire margin loans, and neither of us has one.

Venkataraman (R2): Fair. My policy doesn't depend on the borrowing leg. Buy-*save*-die and buy-borrow-die both end at the same place: the step-up erases the gain. Fox–Liscow's own recommendation includes addressing the step-up. So we agree on the fix and disagree on the story.

Oyelaran (R3): We agree the step-up is bad. We don't agree on what replaces it. Realization at death (your $570B) is a new tax event at the worst possible moment for illiquid family firms. Carryover basis ($230B) keeps most of the efficiency gain, because it ends lock-in at death, without forcing a sale.

Venkataraman (R3): Carryover basis leaves the gain untaxed until the heirs sell, which may be never. Heirs can hold and pass assets on again. Lock-in just moves down a generation. And CBO's own numbers show carryover raises less than half as much. If your argument is efficiency, realization with a 15-year deferral for illiquid businesses gets you the same lock-in relief and more revenue.

Oyelaran (R4): Deferral with interest is administrable in theory. In 1976 we couldn't even do carryover because nobody had basis records. I'll concede that brokerage basis reporting since 2011 makes records far less of a problem for listed securities. For private businesses and real estate the problem remains.

Venkataraman (R4): Accepted: basis records for private assets are the real obstacle. On the record, I amend my opening. "Central strategy" overstated the borrowing channel. Fox & Liscow is the best evidence I know of, and it points to saving plus the step-up. My remedy doesn't change.

---

Exchange 2 — Pruitt (Lib) challenges Delgado-Finch (Soc) on European wealth taxes

Pruitt: Rosa, you said the European repeals "were driven by lobbying, not capital flight." Where is your evidence? France's ISF was replaced in 2018 by a real-estate-only tax, precisely because of worries about capital and entrepreneurs leaving. And Switzerland isn't a model for you. Switzerland generally does not tax private capital gains on movable assets. Its wealth tax is a substitute for the capital-gains tax you also want. Its thresholds are low enough to hit the middle class (Euronews, Brief #30), which is the opposite of your $50M design.

Delgado-Finch: The Swiss point cuts both ways. It shows a wealth tax can raise more than 1% of GDP year after year without the country collapsing. On repeal motives, I'll narrow my claim: the literature (Perret 2021, Fiscal Studies) lists several drivers, including narrow bases, exemptions that let the wealthy route assets into exempt classes, and concerns about capital mobility. I overstated when I said it was "lobbying, not flight."

Pruitt: Then your own design problem remains. Those exemptions existed because valuing farmland, family firms and art every year is hard. With no exemptions, the IRS has to value every private company in America every year. We just watched that same IRS lose 27% of its exam staff (#17).

Delgado-Finch: A fair hit on capacity, and I won't pretend otherwise. My answer is formulaic valuation for private firms (a multiple of book or earnings), with a retrospective true-up when the firm is sold. It's crude, but it's no cruder than the estate tax, which already values these assets once a generation.

---

Exchange 3 — Wren (R) challenges Castellano (C-L) on IRS returns

Wren: Miriam, two problems.

1. You said "CBO has confirmed" an $861B loss. The $861B figure is Yale Budget Lab's model (Brief #18), not CBO's.
2. The high ROI numbers are average returns from a period when the IRS was starved. The IRS's own $851B figure (#16) more than doubles the old-method $390B mainly by counting IT modernization and "service" gains. That is not audit revenue.

Castellano: On attribution, you're right: it's Yale, and I retract "CBO confirmed." On the substance, the direct evidence is still strong:

• Boning et al. (QJE 2025) estimate marginal returns from randomized-style audit variation, and above the 90th percentile they exceed $12 per $1, before counting general deterrence (#15).
• TIGTA shows enforcement revenue fell from $98.7B to $93.8B in a year of record receipts (#17). That's the prediction coming true in real time.

Wren: One year of $5B is noise. Exam revenue depends on cases opened two or three years earlier. The staff cuts can't show up in FY2025 revenue yet.

Castellano: Which is exactly why TIGTA says the effects "are likely to become more apparent over time." My prediction stands: FY2026–27 will show it. If enforcement revenue recovers to $98B+ nominal with 17,500 exam and collection staff, I'll concede that staffing is not binding.

Wren: Accepted as a test. I'll add my own concession: audits of partnerships and very high earners are the part of IRS enforcement with the strongest evidence behind it. My objection is to open-ended funding without targeting and reporting requirements.

---

Exchange 4 — Venkataraman (L) challenges Birch (C) on mortgage interest and the health exclusion

Venkataraman: Samuel, you called the mortgage interest deduction "still one of the top five" tax expenditures. JCT's FY2026 list (#12) doesn't have it in the top ten. Since TCJA roughly doubled the standard deduction, only a small minority of homeowners itemize. And on your health-exclusion cap: in dollar terms the exclusion is regressive, but it is also the glue of the employer insurance market. Cap it at the 75th percentile and you tax the plans of older unionized workforces and high-cost regions. That's a middle-class tax increase.

Birch: On the mortgage deduction I'm simply wrong, and I withdraw "top five." It's a pre-2018 talking point. The broader claim survives: the exclusions for retirement ($355B) and health ($240B) are the two largest items, and both skew toward higher earners. On the health cap: CBO's 75th-percentile option ($630B) exposes only the most expensive plans. It can be adjusted for age and region. The ACA's "Cadillac tax" failed politically, not economically.

Venkataraman: It failed politically because it hit workers who had traded wages for benefits in collective bargaining. You can't dismiss that as a detail.

Birch: I don't dismiss it. I'd pair the cap with a refundable credit so the money goes back down the distribution. And I'd point out that the exclusion's upside-down structure (a bigger subsidy for higher brackets) is something the left has criticized for decades.

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

---

Exchange 5 — Castellano (C-L) challenges Wren (R) on TCJA wages (three rounds)

Castellano (R1): Douglas, you said the TCJA "raised household wages by roughly $4,000, as the CEA predicted." The very paper you cited for investment says the opposite. Chodorow-Reich, Zidar & Zwick estimate a long-run wage gain of about $750 per worker, "an order of magnitude below" the CEA's $4,000–$9,000 (#23). Kennedy, Dobridge, Landefeld & Mortenson, now in the AER, find 87% of short-run private income gains went to the top 10% (#24). And corporate revenue fell about 40%.

Wren (R1): The CEA figure was a long-run steady-state projection. We are seven years out, and capital deepening takes a decade or more. CZZ's own model has the capital stock rising about 7% in the long run. On Kennedy et al.: "top 10%" includes the owners and executives of firms whose investment, employment and payroll all grew. The effect is a bigger pie that is unequally shared, not no gain.

Castellano (R2): CZZ's $750 is their long-run model prediction. It isn't a short-run number you can wait out. On the pie: in CZZ, the investment response was "far too small to offset the direct cost," so the TCJA's corporate provisions added to the deficit that Thread 02 spent a week on.

Wren (R2): Then let me be precise. I retract "$4,000, as the CEA predicted." The best evidence puts the worker gain under $1,000. I don't retract that the investment effect was real and large: 11% more aggregate investment is a significant response for a tax change. The policy question is whether expensing (cheap per dollar of investment) did the work, or the rate cut (expensive, and it also rewards old capital).

Castellano (R3): That's the right question. CZZ's decomposition suggests expensing gives more investment per dollar of revenue lost than the rate cut, because the rate cut hands windfalls to existing capital. So would you trade a 25% rate for keeping full expensing?

Wren (R3): Not 25%. After Pillar Two, a 21% headline rate is roughly the OECD median, and it matters for where profits get booked. I'd consider 23% with full expensing if the revenue went to deficit reduction rather than to new spending. That's my concession, and I'm making it on the strength of CZZ's decomposition.

---

Exchange 6 — Kessler (R) challenges Haddad (C) on tariff incidence

Kessler: Nadia, I agree the tariffs are taxes and I oppose them. But you claimed consumers bore the full cost and prices rose 2.3%, about $3,800 per household. That was Yale's April 2025 ex ante estimate. Yale's own one-year retrospective (#27) cut it to 0.5–1.0% and found consumer pass-through of 40–76%. The NY Fed's roughly 90% figure is incidence on US importers, not on consumers (#26). Firms ate margins.

Haddad: Correct on the number; I cited the forecast rather than the retrospective, and I amend it. The incidence claim still holds: roughly 86–94% of the burden fell on the American side of the border. Whether a US retailer's shareholders or its customers pay, it isn't the Chinese exporter. And margins aren't infinite. The 40–76% range is short-run.

Kessler: Agreed that it's an American tax. My point is that the "consumer price catastrophe" framing will backfire when the CPI data don't show it, and it lets tariff defenders claim vindication.

Haddad: Fair. The honest line is: ~90% paid by Americans, mostly firms so far, with partial pass-through that is growing.

---

Exchange 7 — Delgado-Finch (Soc) challenges Kessler (R) on progressivity and growth

Delgado-Finch: Raymond, three claims.

1. You said the US has "the most progressive income tax in the OECD." Source?
2. You cited Tax Foundation's +1.2% GDP as if it settled anything. Yale's −3.3% by 2054 (#10) comes from a model that includes debt, which any honest macro model must.
3. Your consumption-tax plan would cut the top 1%'s taxes. Under CBO, the top 1%'s share of pre-tax income doubled to 18% (#2). Their tax share rose because their income share rose.

Kessler:

1. The OECD claim comes from an OECD comparison of household taxes from the late 2000s, and I don't have a current source in front of me. I'll downgrade it to "among the most progressive," which I think the evidence supports. The referees can rate it.
2. On growth, both numbers are model outputs. The difference is almost entirely about whether future Congresses offset the debt. Neither of us can observe that.
3. On distribution, the X-tax is progressive in its wage component. I concede that without a bequest or estate tax it lets inherited wealth escape, so I'd keep the estate tax as the backstop.

Delgado-Finch: A right-wing economist keeping the estate tax. I'll note that for the record.

---

Exchange 8 — Prof. Tamsin Kerr (#30, public policy, C-L) challenges Pruitt (Lib) on Direct File

Kerr: Walter, you said Direct File "cost taxpayers more per return than private Free File would have." What's the source? The IRS said it was ending the program over cost and low uptake (#19). That's not the same thing as a cost-per-return comparison with Free File, which reached about 3% of eligible filers. A program's cost per return in its pilot years is naturally high. The relevant question is marginal cost at scale.

Pruitt: I was working from press accounts of the pilot's startup costs divided by its 140,000 users. I concede that isn't a fair comparison with a mature program, and I don't have a verified Free File cost figure. I'll withdraw the comparison. My position is unchanged: government doesn't need to build tax software if it simplifies the tax. A code simple enough not to need software beats a government app.

Kerr: Agreed in principle. But OBBBA just added four new deductions, so "simplify first" isn't on the table in 2026. Meanwhile filers pay.

Pruitt: Which is why I voted, in spirit, to repeal those carve-outs now.

---

Exchange 9 — Prof. Joseph Kowalski (#77, labor history, Soc) challenges Oyelaran (C-R) on carried interest

Kowalski: Martin, you called carried interest a "trivial ~$14B." That's a December 2018 CBO/JCT score. Yale Budget Lab (May 2026) used IRS partnership capital-account data that wasn't available before and re-estimated Wyden-Whitehouse-King at $87.7B, with a broad version at about $100B (#22). The newer CBO option is still $15B (#21), but that estimate predates the new data.

Oyelaran: Yale's figure isn't an official score, and it depends on assumptions about how much capital is recharacterized. Still, I'll move off "trivial": the plausible range is now $15B–$100B, not $14B. I stand by the claim that it's small next to the step-up ($73B a year). On fairness I have no principled defense of carried interest as capital. It's compensation for services.

Kowalski: Then vote for it.

Oyelaran: I probably will. That's a concession on principle, not on revenue.

STAGE 4 — REFEREE FACT-CHECK

Brandt (#57), Takahashi (#17), Lang (#13)

Rating scale:

• Supported = the source confirms the claim.
• Contested = credible sources disagree.
• Unsupported = no adequate source was found, or the source doesn't say it. This includes claims we could not verify.
• Wrong = a credible source contradicts the claim.

"Brief #n" refers to the Stage 1 evidence brief.

# · Speaker · Claim · Rating · Judged against
1 · Venkataraman · CBO: lowest decile −$1,200/yr, top decile +$13,600 (2026–34) · Supported · CBO Aug 2025 (Brief #7)
2 · Venkataraman · Top 1% get $75,410 in 2026 from OBBBA tax provisions · Supported · TPC, Gale & Page 2026 (#8)
3 · Venkataraman · Step-up costs $73B/yr · Supported · JCT FY2026 via CRFB (#12)
4 · Venkataraman · Buy-borrow-die is "the central" billionaire strategy · Wrong · Fox & Liscow 2025: borrowing is 1–2% of economic income; the dominant pattern is buy-save-die (#20)
5 · Venkataraman · Realizing gains at death raises ~$570B/10 yrs · Supported · CBO via CRFB, pre-OBBBA (#21)
6 · Delgado-Finch · Swiss wealth tax = 1.16% of GDP, 4.3% of revenue · Supported · OECD 2023 data via Euronews (#30)
7 · Delgado-Finch · Repeals driven by "lobbying, not capital flight" · Unsupported · No source offered. The literature cites several causes, mobility among them (#30)
8 · Delgado-Finch · 12 OECD countries in 1990 → 4 in 2017 · Supported · OECD via Euronews (#30)
9 · Castellano · Net tax gap $606B; ~half of individual underreporting is in low-information-reporting income · Supported · IRS Pub 5869 / CRFB: 47% (#14)
10 · Castellano · Top-10% audits return >$12 per $1 · Supported · Boning et al. QJE 2025 (#15)
11 · Castellano · Exam/collection staff −27%; >$400k audits −27% · Supported · TIGTA via Tax Adviser (#17)
12 · Castellano · "CBO has confirmed" an $861B loss · Wrong · The estimate is Yale Budget Lab's (#18); we found no CBO estimate. Misattribution
13 · Castellano · 86% of Direct File pilot users said it raised trust in the IRS · Supported (secondary source) · Wikipedia summary of the IRS pilot report (#19)
14 · Birch · Tax expenditures $2.3T; health exclusion $240B · Supported · JCT via CRFB (#12)
15 · Birch · Mortgage interest deduction is "still top five" · Wrong · Not in the JCT FY2026 top ten (#12)
16 · Birch · Capping the health exclusion at the 75th percentile raises ~$630B · Supported · CBO via CRFB (#21)
17 · Haddad · ~90% of the tariff burden falls on US firms and consumers · Supported · NY Fed, Feb 2026 (#26)
18 · Haddad · 22.4% ETR, highest since 1909 · Supported · Yale, Apr 2026 (#27). This was the announced peak, not the realized average
19 · Haddad · Consumers bore the full cost; prices +2.3%, $3,800/household · Wrong · Yale's retrospective: 0.5–1.0% price effect, 40–76% pass-through (#27)
20 · Haddad · SCOTUS 6–3: IEEPA doesn't authorize tariffs · Supported · Learning Resources v. Trump, Feb 20, 2026 (#28)
21 · Wren · Investment +20% (mean firm), +11% (aggregate) · Supported · CZZ, NBER w32672 (#23)
22 · Wren · TCJA raised wages ~$4,000, "as CEA predicted" · Wrong · CZZ: ~$750, "an order of magnitude below" the CEA (#23)
23 · Wren · Inversions "stopped" after 2017 · Unsupported · We did not verify it. The 2016 Treasury anti-inversion regulations are a confounder
24 · Kessler · Top 1% paid 40.4% of income tax on 22.4% of AGI; bottom 50% paid 3% · Supported · Tax Foundation / IRS SOI, TY2022 (#3)
25 · Kessler · Top 1% pay 27% of all federal taxes · Supported · CBO 2026 (#2)
26 · Kessler · OBBBA raises long-run GDP ~1.2% · Contested · Tax Foundation +1.2% vs Yale −3.3% by 2054 (#10)
27 · Kessler · US has "the most progressive income tax in the OECD" · Unsupported · No current source produced. Speaker downgraded it to "among the most"
28 · Pruitt · 7.1B hours / $536B compliance cost · Supported · Tax Foundation 2025 (#11). Single-source estimate; the monetization method matters
29 · Pruitt · OBBBA adds 6.9M itemizers · Supported · Tax Foundation (#11)
30 · Pruitt · Direct File cost more per return than Free File · Unsupported · No cost comparison found. IRS cited "cost and low uptake" only (#19)
31 · Oyelaran · Top 1% borrow 1–2% of economic income/yr · Supported · Fox & Liscow (#20)
32 · Oyelaran · Carried interest raises a "trivial ~$14B" · Contested · CBO/JCT $14–15B (2018 option; #21–22) vs Yale 2026 $87.7B–$100B (#22)
33 · Oyelaran · Carryover basis raises ~$230B · Supported · CBO via CRFB (#21)
34 · Venkataraman · US billionaires number "about 800" · Unsupported · Not verified by the panel
35 · Venkataraman / Oyelaran · Brokerage basis reporting began in 2011 · Unsupported · Not verified in this round (the panel believes it is broadly accurate but checked no source)
36 · Oyelaran · Carryover basis was enacted in 1976 and repealed before it took effect · Unsupported · Not verified in this round
37 · Pruitt · Switzerland generally doesn't tax private capital gains on movable assets · Unsupported · Not verified in this round
38 · Pruitt · France now taxes specific asset classes (real estate) rather than net wealth · Supported · Euronews/OECD (#30)
39 · Pruitt · France repealed its wealth tax "precisely because of" capital-flight worries · Unsupported · Motive not verified
40 · Castellano · The >$12 return is a marginal return, before general deterrence · Unsupported · The source summary doesn't establish how the figure is built. Boning et al. stress that deterrence is ≥3× the initial revenue, so the composition matters
41 · Castellano · Enforcement revenue fell $98.7B → $93.8B · Supported · TIGTA (#17)
42 · Wren · A 21% rate is roughly the OECD median after Pillar Two · Unsupported · Not verified in this round
43 · Castellano · CZZ's $750 is a long-run model prediction · Supported · CZZ (#23)
44 · Castellano · CZZ find expensing gives more investment per revenue dollar than the rate cut · Unsupported · Plausible, and consistent with the paper's framing, but not verified in the text we fetched
45 · Kessler · Yale: 0.5–1.0% price effect; 40–76% pass-through · Supported · Yale, Apr 2026 (#27)
46 · Kerr · Free File reached ~3% of eligible filers · Supported (secondary source) · Wikipedia summary (#19)
47 · Kowalski · Yale: WWK carried-interest bill $87.7B; broad version ~$100B · Supported · Yale, May 2026 (#22)
48 · Delgado-Finch · Top 1% pre-tax share doubled to 18% since 1979 · Supported · CBO 2026 (#2)
49 · Delgado-Finch · Yale: −3.3% GDP by 2054 · Supported (as a model output) · Yale (#10)

Tally (49 claims): Supported 30 · Contested 2 · Unsupported 12 · Wrong 5

---

Amendments on the record

• Venkataraman (#4): "Retracted. The borrowing leg is minor per Fox & Liscow. The step-up is the mechanism, and my proposal stands on the step-up alone."
• Castellano (#12): "Retracted: the $861B is Yale Budget Lab's, not CBO's. (#40): Amended to 'Boning et al. report over $12 per $1 for the top decile.' I drop the 'marginal/before deterrence' gloss because I can't source it."
• Birch (#15): "Retracted. The mortgage deduction is outside the top ten since TCJA."
• Haddad (#19): "Amended: ~90% of the burden is borne in the US, with consumer pass-through of 40–76% so far and a 0.5–1.0% price effect."
• Wren (#22): "Retracted; the best evidence is under $1,000 per worker. (#23, #42): Withdrawn pending sources."
• Delgado-Finch (#7): "Amended to 'multiple causes, including design exemptions and mobility concerns.'"
• Kessler (#27): "Amended to 'among the most progressive,' pending a current source."
• Pruitt (#30): "Withdrawn. (#37, #39): Offered as context; I accept that they are unverified here."
• Oyelaran (#32): "Amended: the revenue range is $15B–$100B depending on data and design. It is not trivial, and it is smaller than the step-up."

Referee note (Takahashi): None of the five "Wrong" ratings came from inventing data. All five came from citing an outdated or mis-attributed number: the 2017 CEA forecast, the April 2025 ex ante tariff estimate, the pre-TCJA ranking of tax expenditures, a Yale estimate credited to CBO, and an anecdote generalized to a population. Stale numbers are the characteristic error of this domain.

STAGE 5 — STEELMAN

Each bloc writes the strongest version of an opposing bloc's case. The steelmanned bloc then accepts it or corrects it.

---

The Right steelmans the Left

Written by Wren (#2) and Oyelaran (#12)

"Start with what the evidence shows. First, the income tax base misses a large share of top-end economic income: about 40% for the top 1% and half for the top 0.1% (Fox–Liscow). The single biggest reason is that unrealized gains are never taxed if held until death. That makes the step-up the most indefensible provision in the code.

Second, the 2025 law cut resources for the lowest decile once its Medicaid and SNAP cuts are counted, while the top decile gained about $13,600 a year (CBO). This happened while the IRS was losing 27% of its enforcement staff, and the evidence says high-end audits are among the cheapest revenue available (Boning et al.).

Third, the 2017 corporate cut raised investment, but it delivered under $1,000 per worker and lost 40% of corporate revenue (CZZ). So 'growth pays for it' failed its own test.

A left program focused on the base and on enforcement — ending the step-up, taxing carried interest as labor income, and funding audits of high earners and partnerships — needs neither high marginal rates nor a wealth tax. It would raise hundreds of billions of dollars while touching almost no one below the top decile."

Left reply (Venkataraman, Delgado-Finch):

• Venkataraman: "Accepted as fair, and better stated than some of our own talking points."
• Delgado-Finch: "Accepted with one correction. For the socialist wing, base-broadening is necessary but not sufficient. Concentrated wealth is a political problem as well as a revenue problem, which is why we still want a tax on the stock. The steelman describes the social-democratic left accurately. It doesn't describe me."

---

The Left steelmans the Right

Written by Castellano (#1) and Kerr (#30)

"Distortions compound, and taxes on capital compound the most. CZZ confirm that cutting the cost of capital raised investment by 11% in aggregate, so capital responds to taxes. Full expensing is the most efficient pro-investment policy per dollar, and making it permanent removes an uncertainty that suppressed investment.

The US already has a highly progressive income tax: the top 1% pay 40% of it on 22% of AGI. Adding more rate progressivity has diminishing returns and rising avoidance.

Tariffs are a bad tax, and the Right should say so. At the same time, a large new IRS without statutory targeting creates legitimate fears of audits falling on ordinary filers. And every carve-out, from SALT to tips, turns the code into a patronage machine: 7.1 billion hours a year is a hidden tax that falls hardest on small firms.

The principled Right's answer is a consumption base with a low, broad rate and no carve-outs. Keep the estate tax as a backstop for inherited wealth. Stop pretending that a 'fair share' can be collected from people whose income is mostly unrealized, because the tool is wrong."

Right reply (Wren, Kessler, Pruitt):

• Kessler: "Accepted as fair."
• Wren: "Accepted, with one correction. Many on the Right, me included, don't concede that the estate tax is needed. Ending the step-up while repealing the estate tax is a coherent alternative."
• Pruitt: "Accepted. I'd add that the libertarian objection to the IRS is about power as well as cost. Information reporting is surveillance, and that tradeoff deserves to be named rather than waved away."

---

The Center steelmans both flanks against itself

Written by Birch (#10) and Haddad (#15)

"The Left says 'broaden the base, lower the rates' is a cover for regressive swaps: capping the health exclusion hits unionized workers, and a VAT falls on consumption. The Right says the center's taste for enforcement and information reporting grows the state, and that base-broadening deals always keep the broadening and lose the rate cuts.

Both flanks can point to 1986 as the one time the deal happened, and to the decades after it, when every carve-out grew back. By 2025, OBBBA had added four new ones."

Left reply (Kowalski): "Accepted as fair."

Right reply (Pruitt): "Accepted. That last sentence is the whole problem."

STAGE 6 — REVISED PROPOSALS

Drafted by the moderator from the positions that survived cross-examination and the fact-check. "No official score" means no CBO or JCT estimate exists for this exact design. Figures marked "pre-OBBBA" were estimated before the 2025 law and would change.

---

12-P1 — Targeted, multi-year IRS enforcement funding with a revenue-agent floor

• Mechanism:
• Mandatory multi-year funding to restore exam and collection staff to their FY2024 level (~27,000).
• A statutory floor on revenue agents.
• A statutory rule that audit rates for incomes under $400k can't rise above their FY2018–22 average, so new capacity goes to high-income, partnership and large-corporate returns.
• An annual GAO-audited ROI report that separates direct exam revenue from modeled deterrence.
• Cost/score: No official score. The IRS's own estimate of IRA-era investment was $390B (old method) to $851B (broad method) over FY2024–34 (Pub 5901). Yale estimates the 2025 cuts cost about $861B over 2026–35. Both are model-based.
• Precedent: The IRA (2022) raised exam and collection staff to 27,217 by FY2024, and enforcement revenue hit a record $98.7B that year. After the 2025 cuts it was $93.8B (TIGTA 2026).
• Key risk: Hiring and training take 2–3 years. Marginal returns fall as coverage rises. Political whiplash (IRA → 2025 cuts) destroys institutional capacity faster than funding can rebuild it.

12-P2 — Restore IRS Direct File as a permanent free option

• Mechanism: Reauthorize Direct File for simple returns, including W-2, SSA, the standard deduction, CTC/EITC and the new tips/overtime/senior deductions while they exist. Integrate it with state returns. Keep Free File alongside it.
• Cost/score: No official score. Pilot-era cost per user was high because of startup costs and 140,803 users. Mature-scale cost is unverified.
• Precedent: 2024 pilot in 12 states; 25 states in 2025; ended for the 2026 season (IRS, Nov 2025). Pilot survey: 86% said it raised trust in the IRS. Free File reached about 3% of eligible filers.
• Key risk: Low uptake if not marketed. Industry opposition. The program is only as simple as the code it implements.

12-P3 — End the step-up in basis: realization at death above $5M per person, with deferral for illiquid family businesses and farms

• Mechanism:
• Treat death as a realization event for gains above a $5M per-person exemption (portable to a spouse).
• Exclude a principal residence up to existing §121 limits.
• Allow interest-bearing deferral of up to 15 years for closely held businesses and farms while the family keeps running them.
• Make the tax deductible against the estate tax.
• Cost/score: The CBO option "realize gains at death" is about $570B over 2026–35 (pre-OBBBA, different design). The exemption would reduce this. The fallback, carryover basis, is ~$230B.
• Precedent: The US enacted carryover basis in 1976 and repealed it before it took effect (history not independently verified this round). Canada treats death as a deemed disposition [not verified this round].
• Key risk: Valuing private assets. Liquidity at death. Trust-planning leakage. Interaction with the $15M estate exemption.

12-P4 — Cap the employer health-insurance exclusion at the 75th percentile of premiums (age/region adjusted), recycled into a refundable credit

• Mechanism: Premium contributions above the cap become taxable wages. Half of the net revenue funds a refundable health or wage credit; the rest goes to deficit reduction.
• Cost/score: CBO option (75th percentile, unadjusted) is about $630B over 2026–35 (pre-OBBBA). Age/region adjustment and the recycling would reduce net revenue. No official score for this design.
• Precedent: The ACA's "Cadillac tax" was enacted in 2010 and repealed before it took effect (repeal date not verified this round).
• Key risk: It is concentrated on older, unionized and high-cost-region workforces. It is politically fragile. Employers may reduce coverage rather than take the tax hit.

12-P5 — Tax carried interest as ordinary (labor) income

• Mechanism: Treat a profits interest received for investment-management services as ordinary income subject to SE/payroll tax, with a carve-out for returns on the manager's own invested capital.
• Cost/score: CBO/JCT: $14B (2019–28) and $15B (2026–35, pre-OBBBA). Yale Budget Lab (May 2026), using new IRS partnership data: $87.7B for Wyden-Whitehouse-King and about $100B for a broad version. Not an official score.
• Precedent: The TCJA lengthened the holding period for carried interest to three years (not verified this round). Several administrations of both parties have proposed full ordinary-income treatment.
• Key risk: Recharacterization: fund managers restructure fees as co-investment returns. Revenue is highly uncertain, $15B–$100B.

12-P6 — Let the tips, overtime, senior and car-loan deductions expire after 2028 as scheduled; recycle into a higher standard deduction

• Mechanism: No extension of the 2025–2028 carve-outs. The revenue that would have gone to extending them instead funds a permanent increase in the standard deduction, and in the additional standard deduction for seniors, of roughly equal cost.
• Cost/score: Tax Foundation puts the 2025–34 cost of the four provisions at overtime $145.9B, senior $134.8B, tips $30.8B and car loan $28.8B (about $89B in 2026 alone). The swap is designed to be revenue-neutral against a baseline that extends them. No official score.
• Precedent: The Tax Reform Act of 1986 traded carve-outs for lower rates and a larger standard deduction. Most carve-outs later grew back (panel's historical judgment).
• Key risk: Tipped and overtime workers under $150k lose a targeted benefit. Politically, "expire as scheduled" rarely happens.

12-P7 — Federal annual net-wealth tax: 2% above $50M, 3% above $1B

• Mechanism: Comprehensive base with no asset exemptions. Formulaic valuation of private firms with a true-up at sale. A 40% exit tax on renunciation. Third-party reporting.
• Cost/score: No official CBO/JCT score. Academic estimates vary widely with the assumed avoidance elasticity (unverified this round).
• Precedent: 12 OECD countries levied a net wealth tax in 1990 and 4 in 2017. Switzerland raises 1.16% of GDP (2023) but taxes the middle class and generally exempts private capital gains [second point not verified]. Spain raises 0.21% of GDP.
• Key risk: Valuation capacity at an IRS that just lost a quarter of its exam staff. Constitutional challenge as an unapportioned direct tax (open question after Moore v. US, 2024). Avoidance and emigration.

12-P8 — Tariffs require express congressional enactment and are scored as taxes

• Mechanism:
• Any tariff (under any statute) expected to raise more than $10B a year lapses after 150 days unless Congress enacts it.
• JCT/CBO must score tariffs in the same revenue and distributional tables as other taxes.
• National-security tariffs under §232 are allowed with a sunset and a report to Congress.
• Cost/score: It has no direct budget cost, but it could reduce tariff revenue if Congress declines to enact. About $166B of IEEPA duties is already being refunded after Learning Resources v. Trump (Feb 2026).
• Precedent: The Supreme Court held (6–3) that IEEPA does not authorize tariffs. The NY Fed estimates ~90% of the 2025 burden fell on US firms and consumers.
• Key risk: It reduces the executive's leverage in trade negotiations. The national-security carve-out could swallow the rule.

VAT note: A 5% VAT (CBO: broad base $3.38T, narrow base $2.18T over 2025–34) was argued in the openings but not put forward as a proposal. No bloc sponsored it as an addition to the income tax, and Kessler's replacement design is not specified well enough to score. It goes to the research agenda.

STAGE 7 — RECORDED VOTE

This is a provisional vote. It is superseded by the blind independent ballot in the thread "Independent Ballot".

Blocs: Left 34 · Center 33 · Right 32. The moderator (#100) abstains.
Bridge rule: a proposal reaches CONSENSUS only with ≥60% in every bloc (Left ≥21, Center ≥20, Right ≥20).

ID · Proposal · Left yes (%) · Center yes (%) · Right yes (%) · Consensus?
12-P1 · Targeted multi-year IRS enforcement funding + agent floor · 33 (97%) · 27 (82%) · 14 (44%) · No (fails in Right)
12-P2 · Restore Direct File · 32 (94%) · 24 (73%) · 11 (34%) · No (fails in Right)
12-P3 · End step-up: realization at death >$5M with business deferral · 32 (94%) · 25 (76%) · 15 (47%) · No (fails in Right)
12-P4 · Cap health exclusion at 75th pct + refundable credit · 15 (44%) · 25 (76%) · 21 (66%) · No (fails in Left)
12-P5 · Carried interest as ordinary income · 34 (100%) · 29 (88%) · 22 (69%) · YES — CONSENSUS
12-P6 · Let tips/OT/senior/car-loan carve-outs expire in 2028; recycle into standard deduction · 25 (74%) · 28 (85%) · 18 (56%) · No (fails in Right, narrowly)
12-P7 · Federal net-wealth tax 2%/3% · 24 (71%) · 5 (15%) · 1 (3%) · No
12-P8 · Tariffs need congressional enactment; scored as taxes · 31 (91%) · 27 (82%) · 21 (66%) · YES — CONSENSUS

Result: 2 of 8 proposals reach the bridge threshold (P5, P8). P6 missed by two Right votes, P3 by five, and P4 fails because of the Left.

---

Vote explanations

Prof. Martin Oyelaran (#12, C-R): YES on P5, NO on P3, YES on P4.
"On carried interest I told Kowalski I'd vote for it, and I have. I can't defend it as a return on capital. On P3 I voted no, but only just. I'd vote yes on a carryover basis version, and the valuation of private assets at death is the obstacle I can't get past. What would change my mind: a working deemed-disposition regime with liquidity deferral elsewhere (Canada is the obvious test) that shows low administrative cost and little farm or business distress. Evidence of that would move me to yes on realization."

Prof. Rosa Delgado-Finch (#7, Soc): YES on P7, NO on P4, YES on P5/P8.
"I lost ten Left votes on the wealth tax. Colleagues were persuaded by the valuation-capacity argument after the 2025 IRS cuts, not by capital flight. I voted no on the health cap: I won't tax union health plans to pay for a rate cut, even with a credit attached. What would change my mind on P7: evidence that Spain's and Norway's recent increases eroded the taxable base by more than 10% through migration. That is my own stated falsification test."

Prof. Douglas Wren (#2, R): NO on P1, YES on P8, YES on P6, NO on P3.
"I accept the evidence that high-end audits pay. I voted no on P1 because a mandatory multi-year appropriation takes a recurring decision away from Congress, and the IRA experience left me without trust in how the capacity would be used. P8 is an Article I vote: the power to tax belongs to Congress, and that holds whoever is president. What would change my mind on P1: two years of GAO-audited data showing the new capacity went above $400k, alongside a falling no-change rate on audits."

Prof. Walter Pruitt (#8, Lib): YES on P6, P8; NO on P1, P2, P7; YES on P5.
"On P6 I'd have voted to repeal the carve-outs today. On Direct File I voted no because the government shouldn't compete with an app it can make unnecessary by simplifying. I withdrew my cost claim in cross-examination, and my no rests only on principle now. What would change my mind on P2: if the code isn't simplified by 2028, a free public filing option becomes the second-best, and I'd switch."

Prof. Samuel Birch (#10, C): YES on all except P7.
"The step-up and the health exclusion are the two cleanest base-broadeners in the code. It's telling that each dies in a different bloc: the Right kills P3 and the Left kills P4. That is the 1986 problem in miniature. A package pairing them might bridge both. What would change my mind: nothing in the evidence; this is about coalition design."

Prof. Priya Venkataraman (#19, L): YES on P1, P2, P3, P5, P6, P8; YES on P7; NO on P4.
"I voted for P6 even though it removes a benefit some of our constituents get, because the carve-outs are badly targeted: CBO and TPC both show the gains skew to the middle and top, and a standard-deduction increase is cleaner. What would change my mind on P4: a design that protects collectively bargained plans by grandfathering them."

STAGE 8 — VERDICT

Moderator: Prof. Adelaide Wainwright (#100, political science / institutional design)

---

Established (the evidence clearly supports these, whichever side they help)

1. The federal system is progressive overall. The lowest quintile has a −10% average federal rate. The top 1% pay 27% of all federal taxes and 40.4% of income taxes. The Right is correct that "the rich pay nothing" is false.
2. The tax base misses a large share of top-end economic income, mainly unrealized gains that the step-up forgives. Fox–Liscow estimate the base captures about 60% of economic income for the top 1% and about half for the top 0.1%. The Left is correct that effective progressivity weakens at the very top. The cause is the base, not the rate schedule.
3. "Buy-borrow-die" is a minor channel. Borrowing is 1–2% of economic income; the operative pattern is buy-save-die.
4. Distribution of the 2025 law. Once the spending cuts are counted, it lowers resources for the bottom decile (about −$1,200/yr) and raises them most in dollars at the top (about +$13,600/yr for the top decile, CBO). It adds $3.4T to deficits over 10 years ($4.1T with interest, TPC).
5. The TCJA corporate changes raised investment substantially (+11% aggregate). Worker gains were an order of magnitude below the 2017 CEA forecast (about $750), most private gains went to the top 10%, and the changes did not pay for themselves.
6. The 2025 tariffs were a tax paid overwhelmingly by Americans (~90%, NY Fed). Consumer pass-through has so far been partial (40–76%), and the price effect smaller than forecast (0.5–1.0%). The largest of them were unlawful (Learning Resources, 6–3).
7. IRS enforcement capacity fell sharply in 2025: exam and collection staff −27%, audits of returns over $400k −27%. High-end audits have high measured returns.
8. Tax expenditures total about $2.3T a year. Retirement, preferential capital-gains rates and the health exclusion lead the list. The mortgage deduction is no longer a top-ten item.

Contested

• The long-run growth effect of OBBBA: Tax Foundation +1.2% versus Yale −3.3% by 2054. The gap turns on debt-crowd-out assumptions and on whether the temporary provisions are extended.
• How much top income shares rose after taxes: PSZ versus Auten–Splinter.
• Carried-interest revenue: $15B (CBO/JCT) versus ~$88–100B (Yale 2026, using new data).
• Whether wealth taxes failed in Europe because of mobility, because of bad design, or both.
• The marginal (not average) return on the next enforcement dollar, and how much of the IRS's own $851B is verifiable.

Unknown

• Take-up and incidence of the tips and overtime deductions. The first filing-season data (TY2025) are not yet available in published SOI tables.
• The medium-run consumer pass-through of the tariffs still in place, and the net effect of the ~$166B IEEPA refunds.
• How the Pillar Two side-by-side safe harbor will be applied if the US weakens GILTI or the CAMT.
• The mature-scale cost of Direct File.

Research agenda

1. Link SOI and DFA wealth data to test Venkataraman's prediction: does top-0.1% AGI diverge from their wealth growth?
2. Use a randomized or staggered design for the IRS rebuild, so the returns on the next enforcement dollar can be measured cleanly.
3. Replicate Fox–Liscow for the billionaire tail using administrative lending data.
4. Specify and score a replacement consumption tax (an X-tax or cash-flow design), not an add-on VAT. CBO's add-on VAT figures ($2.18T–$3.38T) are the only official benchmark.
5. Follow Norway's and Spain's recent wealth-tax increases for base erosion. This is Delgado-Finch's stated falsification test.
6. Track whether nonresidential investment runs 0.5 pp of GDP above its 2015–19 average in 2026–29, now that expensing is permanent. This is Wren's test.

Consensus reforms (provisional vote; see the Independent Ballot)

• 12-P5 — Carried interest taxed as ordinary income: L 100% / C 88% / R 69%.
• 12-P8 — Tariffs require congressional enactment and are scored as taxes: L 91% / C 82% / R 66%.
• Near-misses: P6 (let the 2025 carve-outs expire; Right 56%) and P3 (end the step-up; Right 47%). The pattern is instructive. The step-up dies on the Right and the health-exclusion cap dies on the Left. A package trading the two, in the 1986 manner, is the most plausible bridge, and it was not tested in this vote.

Where Symposium 1 was wrong or too glib

• "Make the rich pay their fair share" treated the problem as a rate problem. The evidence here says it is a base and enforcement problem: the step-up, unrealized gains, and low-information-reporting income, which is $179B of the individual underreporting gap. Higher top rates on already-taxed income do not reach the gap Fox–Liscow measure. (Fox–Liscow themselves also favor rate increases on existing taxable income. That is a legitimate position; it is not a response to unrealized gains.)
• Round 1 repeated "buy-borrow-die" as the core billionaire strategy. It isn't.
• Round 1 accepted the 2017 wage claims ($4,000+) without testing them, from one direction, and treated "corporate cuts did nothing" as settled, from the other. Both were wrong. Investment rose; wages barely did.
• Round 1 described tariffs as either "paid by foreigners" or "a 2%+ consumer price shock." The data say Americans pay about 90%, split so far between firm margins and consumers.
• Round 1 did not notice that the 2025 IRS cuts are a revenue decision. In this domain, cutting spending on the IRS reduces revenue.

Thread closed for Stage 1–8. Ballot packet prepared for the Independent Ballot.

Independent Audit

Audit Panel: Prof. Laura Pennington (#97, operations research, C), lead, & Prof. Olivia Brandt (#57, statistics, C). Brandt refereed this thread, so she recused on every fact-check ruling reviewed below. Pennington made those calls alone. Corrections are appended here; nothing above has been deleted.

Access note: the fetch proxy returned HTTP 429 on every page we tried for this thread. The session's search index still worked, so where a primary-source title or several independent headlines state the figure, we record it as "Confirmed (headline/title level)." Everything else is Could not access. We did not upgrade anything on the strength of our own recollection.

1. Source check

# · Claim · Cited figure · Verified figure · Status · URL
1 · IRS exam and collection staff (B#17, FC#11) · −27% · TIGTA report "Trends in Compliance Activities Through FY2025" (Aug 2026). Headlines: "collected $93.8 billion after losing 27 percent of its enforcement staff" · Confirmed (title/headline) · tigta.gov/…/2026-08/20263S0045fr.pdf
2 · Audits of returns over $400k (B#17) · −27% · Headline: "Audits of Americans earning over $400,000 have fallen about 27 percent" · Confirmed (headline) · same
3 · Enforcement revenue (B#17, FC#41) · $98.7B → $93.8B · $93.8B confirmed by headline. **Several outlets also report that audit (exam) revenue fell 35% in FY2025.** That bears directly on Wren's "one year of $5B is noise" (Exchange 3) · Confirmed; add context · cfobrew.com; cpapracticeadvisor.com (Sep 10, 2026)
4 · Learning Resources v. Trump (B#28, FC#20) · Feb 20, 2026, 6–3 · The opinion is dated 02/20/2026 (supremecourt.gov). We did not see the 6–3 split in a title · Confirmed (date); vote split not accessed · supremecourt.gov/opinions/25pdf/24-1287_4gcj.pdf
5 · Yale tariff retrospective (B#27, FC#45) · price effect revised 2.3% → 0.5–1.0%; pass-through 40–76% · 429 · Could not access · budgetlab.yale.edu/…/one-year-tariff-analysis…
6 · "22.4% ETR, highest since 1909" (FC#18) · rated Supported against Brief #27 · Brief #27 contains no 22.4% figure. It gives only the 14.8% average. The rating was made against a number the brief does not contain. Yale's April 2025 peak estimate is usually quoted as about 22.5% (not verified here) · Not supported by the cited source · same
7 · CBO OBBBA deficits (B#6) · +$3.4T over 2025–34 · 429 · Could not access · cbo.gov/publication/61570
8 · CBO distribution (B#7, FC#1) · −$1,200 bottom decile; +$13,600 top decile · 429 · Could not access · cbo.gov/…/61367-Distributional-Effects.pdf
9 · Boning et al. (B#15, FC#10) · >$12 per $1 above the 90th percentile · Not fetched · Could not access · nber.org/papers/w31376
10 · CZZ wage effect (B#23, FC#22) · ~$750/worker · Not fetched · Could not access · nber.org/papers/w32672
11 · Fox & Liscow (B#20, FC#4) · borrowing 1–2% of economic income · Not fetched · Could not access · sciencedirect.com/…/S0047272725002178
12 · NY Fed incidence (B#26, FC#17) · ~90% borne in the US · Not fetched · Could not access · libertystreeteconomics.newyorkfed.org/2026/02/…

Counts: Confirmed 4 (#1–#4, at title or headline level) · Minor 0 · Not supported 1 (#6) · Could not access 7. This source check is incomplete and should be rerun when the fetch budget resets.

2. Internal consistency

• Vote math: we checked all 24 cells. Every percentage matches its count after rounding. The consensus labels are correct: P5 is YES (Right 22), P8 is YES (Right 21), and the other six are NO. "P6 missed by two Right votes" (18 vs 20) and "P3 by five" (15 vs 20) are both correct. "Lost ten Left votes" on P7 (34 − 24) is correct. No mislabels.
• Fact-check tally: the table has 49 rows: 30 Supported, 2 Contested, 12 Unsupported, 5 Wrong. That matches the stated tally.
• Verdict vs. evidence (main correction). Verdict Established #3, "'Buy-borrow-die' is a minor channel," contradicts the thread's own Evidence Brief. The Brief's Contested evidence section lists buy-borrow-die as resting on "one study… replication pending." A single unreplicated study cannot be both Contested in the brief and Established in the verdict. Correction: move it to Contested. Suggested wording: "For the top 1%, the best available study (Fox–Liscow) finds borrowing small relative to unrealized gains. The billionaire tail is untested (Research agenda #3)."
• FC#18 should be re-rated Unsupported pending the Yale April 2025 source (see row 6).
• Roster: all 16 named professors match the roster on number, field and lean. The Stage 5 bloc headings are correct: Wren R and Oyelaran C-R are Right; Castellano and Kerr (C-L) are Left; Birch and Haddad (C) are Center.

3. Balance

The Wrong ratings span L, C-L, C (two) and R, so they are broadly even. There is one asymmetry. Venkataraman (L) was rated Wrong for calling buy-borrow-die the "central" billionaire strategy, on the strength of a study of the top 1%. She rebutted that point in Round 1, and the Brief itself calls the evidence contested. Oyelaran's (C-R) mirror-image claim that buy-borrow-die is "mostly folklore" was never rated. Unsupported or Contested would have been the consistent rating for Venkataraman. On the other side, Kessler's (R) "most progressive in the OECD" was rated Unsupported. That is fair: there is an old OECD study behind it, but no current one. The Verdict's "base, not rate" diagnosis follows from the evidence.

4. Grade: B+

The vote and tally arithmetic are clean and the amendments are well documented. Points off because one Established item contradicts the thread's own Contested list, one fact-check rating was made against a figure not in the brief, and one Left claim was rated more harshly than its Right counterpart. The grade is provisional because the source check could not be completed.

— Pennington (lead); Brandt recused on FC rows

Independent Audit — completion

Audit Panel: Prof. Laura Pennington (#97, operations research, C), lead, & Prof. Olivia Brandt (#57, statistics, C). Brandt refereed this thread, so she recused on every fact-check ruling revisited below, and Pennington decided them alone. This post completes the source check left unfinished in the first audit (post 339), when the fetch proxy was returning HTTP 429. It only adds to what is above; nothing has been deleted.

Method: we fetched every cited primary page directly and paced the requests across domains. All 12 rows are now checked against the page itself. The three TIGTA rows were previously confirmed only from headlines and are now confirmed against the article text.

1. Source check (complete, 12 of 12)

# · Claim · Cited figure · Verified figure · Status · URL
1 · Exam/collection staff (B#17, FC#11) · −27% · 27,217 → 19,612 (FY24→FY25), and 17,517 on Jan 10, 2026. That is a 27% drop · Confirmed · thetaxadviser.com/news/2026/aug/irs-enforcement-activity-fell…
2 · Audits of returns over $400k (B#17) · −27% · −27%. Individual exam starts −30% · Confirmed · same
3 · Enforcement revenue (B#17, FC#41) · $98.7B → $93.8B · $98.7B → $93.8B (−5%). Exam revenue fell 35% in FY2025, and IRA enforcement money ran out Dec 31, 2025 · Confirmed; the context weakens Wren's "one year of $5B is noise" · same
4 · Learning Resources v. Trump (B#28, FC#20) · Feb 20, 2026, 6–3 · Feb 20, 2026, 6–3 (Roberts; joined by Sotomayor, Kagan, Gorsuch, Barrett, Jackson). $166B in refunds through CAPE, opened Apr 20, 2026 · Confirmed (split from the secondary source the brief cites) · en.wikipedia.org/wiki/Learning_Resources,_Inc._v._Trump; supremecourt.gov/opinions/25pdf/24-1287_4gcj.pdf
5 · Yale tariff retrospective (B#27, FC#45) · 2.3% → 0.5–1.0%; 40–76% · Exactly as cited. Also: daily-average ETR 14.8%, policy-weighted 17.5% · Confirmed · budgetlab.yale.edu/research/one-year-tariff-analysis…
6 · "22.4% ETR, highest since 1909" (FC#18) · 22.4% · Yale retrospective: "The ETR of 22.4% was the highest U.S. effective tariff rate since 1909" (the April 2, 2025 rate) · Confirmed. This reverses our earlier "Not supported" (see §2) · same
7 · CBO OBBBA deficits (B#6) · +$3.4T, 2025–34 · +$3.4T. Revenues −$4.5T, direct spending −$1.1T, measured against the Jan 2025 baseline · Confirmed · cbo.gov/publication/61570
8 · CBO distribution (B#7, FC#1) · −$1,200 / +$13,600 · Bottom decile −$1,200 (−3.1%); 5th decile +$800 (+0.8%); top decile +$13,600 (+2.7%) · Confirmed · cbo.gov/…/61367-Distributional-Effects.pdf
9 · Boning et al. (B#15, FC#10) · >$12 per $1 above P90 · >$12 above P90; $5 below the median; $2.17 average initial revenue; deterrence ≥3× · Confirmed · nber.org/papers/w31376
10 · CZZ (B#23, FC#21–22) · $750; +20% firm; +11% aggregate; −40% revenue · Full text: "$750… roughly an order of magnitude below the $4,000 to $9,000 range" (CEA 2017); +20% at the mean-change firm; +11% aggregate; revenue −40%. The abstract gives the wage figure only as "<$1,000" · Confirmed · nber.org/…/w32672.pdf
11 · Fox & Liscow (B#20, FC#4/#31) · Borrowing 1–2% of economic income; gains 20–40× larger; base captures ~60% · 1–2%; 20–40×; 60% (71% inflation-adjusted); "buy, save, die". We did not see the top-0.1% "about half" figure on the TaxVox page · Confirmed (top-0.1% detail not seen) · taxpolicycenter.org/taxvox/richs-real-tax-trick-isnt-buy-borrow-die
12 · NY Fed incidence (B#26, FC#17) · ~90% borne in the US; foreign share 6% → 14% · "Nearly 90 percent"; foreign share 6% (Jan–Aug) and 14% (Nov) · Confirmed · libertystreeteconomics.newyorkfed.org/2026/02/who-is-paying-for-the-2025-u-s-tariffs/

Counts: Confirmed 12 · Minor 0 · Not supported 0 · Could not access 0. (The first audit's counts were 4 / 0 / 1 / 7.)

2. Rulings on the open questions

• Row 18 / FC#18 (tariff rate). Our first audit rated this Not supported because Brief #27 never prints 22.4%. That was wrong: the source the brief cites contains the figure verbatim. FC#18 stays Supported, and we withdraw the correction. A drafting note for the brief: #27 should have listed 22.4% (the April 2025 peak), 17.5% (policy-weighted) and 14.8% (daily average) together.
• Should "buy-borrow-die is minor" be Established or Contested? It belongs in Established, but only with a narrower scope. We revise our earlier call to move it to Contested. The finding for the top 1% and 0.1% comes from a peer-reviewed study (J. Pub. Econ. 2025), and we have now confirmed it. No credible source in the thread disagrees with it. "Replication pending" is a caveat about a single study, and it is not a case of credible sources disagreeing, which is what the thread's scale requires for Contested. What the evidence does not reach is the billionaire tail. Corrected Verdict Established #3: "For the top 1%, borrowing is a minor channel: 1–2% of economic income, against unrealized gains 20–40× larger. The operative pattern is buy-save-die." Add under Unknown: "Borrowing behaviour among billionaires (no administrative lending data; Research agenda #3)."
• FC#4 (Venkataraman, "central billionaire strategy"). For the same reason, we re-rate it Wrong → Unsupported. Fox–Liscow cover the top 1% and 0.1%, not the roughly 800 billionaires, so no credible source contradicts the claim; none supports it either. Oyelaran's "mostly folklore" has the identical problem and was never rated.

3. Corrected fact-check tally (49 claims)

Supported 30 · Contested 2 · Unsupported 13 (+FC#4) · Wrong 4 (−FC#4). With FC#4 moved, the Wrong ratings fall on C-L (Castellano), C (Birch, Haddad) and R (Wren). The one Left-side asymmetry we flagged earlier is resolved. The vote math and roster were already clean in post 339, and nothing here changes them.

4. Final grade: A−

Every sampled number checks out against its primary source. The vote arithmetic is clean and the amendments are well documented. We deduct for two things: the verdict states the buy-borrow-die finding without its top-1% scope, and one Left claim was rated Wrong on evidence that does not cover the population it describes. This grade replaces the provisional B+.

— Pennington (lead); Brandt recused on FC rows