Gabriel Zucman · Chapter in an edited volume (Hermann, 2026), pp. 141-144

Inégalités

7 problems found: 3 moderate · 4 minor

What the paper claims

Billionaires pay proportionately less tax, all taxes included, than the rest of the population in every country with data (EU Tax Observatory, Global Tax Evasion Report 2024). The reason is that the richest can structure their wealth to generate little taxable income, so the income tax 'virtually collapses' at the top. ProPublica's examples: no income tax for Jeff Bezos in 2011, $8,410 for Elon Musk in 2018. The problem is probably worse in France and the EU than in the US, whose anti-abuse rules limit holding companies.
Billionaires' wealth is large and 'almost untaxed', a 'situation of virtual non-taxation' ('quasi-non-taxation'). About 3,000 billionaires (about 150 French) own about EUR 13tn (EUR 750bn for the French); this wealth 'has boomed since the 1980s'. The income tax they pay is about 0.3% of their wealth, and close to 0% for French ultra-rich.
A coordinated international standard requiring billionaires to pay at least 2% of their wealth in tax each year (an EU Tax Observatory proposal endorsed by Brazil) would raise about EUR 220bn a year worldwide and about EUR 15bn in France. It would be structured as an income-tax surcharge, with nothing due from those already paying 2% or more of their wealth in taxes.
The tax is feasible and has momentum. The technical obstacles are 'minor': it is 'quite simple to levy' given few taxpayers and highly visible, often listed wealth, and a wealth-referenced floor 'would combat all possible forms of tax optimization'. France, South Africa, Spain, Belgium, Colombia and the African Union support it; Biden's budget has a similar measure; the Brazilian G20 presidency can take a decisive step.

Every problem found

1moderateconfirmed · previously known · internal inconsistency

Billionaires pay very little personal income tax, but the companies they own pay corporate tax, which the author himself counts when he says billionaires pay less than everyone else. Counted that way, they already pay roughly 2% of their wealth in tax each year, not 'almost nothing'. The proposed 2% would come on top of that, which the article does not explain. The numbers are right, but the 'almost untaxed' message overstates the case.

Technical detail and evidence

Why it matters, in detail. The evidence for C1 allocates CIT to owners (G20 Fig. 1: FR 27%, NL 17%, US ~23% of pre-tax income; FR = 1.7% PIT + ~25% CIT, p. 15). Converted with the author's own income-to-wealth ratios (7.5%, p. 17; 9%, App. B), all-in taxes are 1.3-2.4% of wealth (mean 1.7-2.0%), against the 0.3% used for 'almost untaxed'. The ~EUR 220bn requires that only personal taxes be credited against the 2% floor (G20 §2.3). With the CIT borne credited, the average shortfall below 2% would be ~0-0.3% of wealth, removing most of the yield (critics' French estimates: -75% to -95%, offshore-havens-18). The author's benchmark is all-in parity (2% personal + CIT = ~50-55%), which is coherent but undisclosed in the op-ed. Direction: overstates the tax gap in wealth terms; revenue arithmetic unaffected under the stated design.

Evidence. Le Monde FR para 6 (visible): 'cette situation de quasi-non-taxation des milliardaires est difficilement justifiable'. Cahiers p. 25 ('virtual non-taxation of billionaires') and p. 26 ('this wealth is almost untaxed. The income tax paid by these billionaires represents around 0.3% of their assets'; 'For wealthy individuals who already pay the equivalent of 2% or more of their wealth in taxes each year, nothing more would be due'). G20 blueprint (25 Jun 2024): Fig. 1 p. 11; p. 15 (France: 1.7% individual income taxes plus about 25% corporate taxes; Netherlands 17% 'almost entirely due to the corporate tax'); p. 17 (R = 7.5%); p. 18 fn. 9 (business-level capital taxes 'close to 1% of wealth globally'); §2.3 p. 24 (standard defined on individual taxes); App. B p. 49 (pre-tax income/wealth = 9% for billionaires). GTER 2024 Finding #5 ('Personal taxes include all individual income taxes and wealth taxes'). Recomputation: scripts/reproduce.py, block I2 (1.68% / 2.01% of wealth at 7.5% / 9%). Critics' scoring of a CIT-credited floor for France: EUR 2-3bn (Levy, Senate report n° 689) and ~EUR 5.5bn central / ~EUR 1.25-1.5bn with full credit ('Trop c'est trop !' 2026), against EUR 20-24bn uncredited (corpus offshore-havens-18).

Why this severity. No number falls: the 0.3% personal-tax figure and the revenue arithmetic are right under the author's stated design. But the central characterization ('virtually untaxed') holds only for personal income tax. On the article's own all-taxes basis the burden is ~6-7 times larger, and the revenue depends on a design choice (no credit for corporate tax) that readers are not told about. This is an important undisclosed caveat, not a refutation.

The authors' response. Zucman, 'Comment taxer les ultrariches ? Cadrer le débat sur l'impôt plancher', Le Grand Continent, 2 Jul 2025. He concedes 'l'essentiel des 26 % provient de l'impôt sur les sociétés', but argues the 2016 rate overstates today's, since the CIT rate fell from 33% to 25% and much of it is paid abroad. The 2% floor is deliberately set on personal taxes so that, together with CIT, billionaires reach total-tax parity (G20 blueprint §1.3, p. 15; 2025 book, per corpus offshore-havens-18).

2moderateconfirmed · previously known · selective presentation

The EUR 220 billion assumes that nobody avoids the tax and that every country with billionaires takes part. The author's own report, out a few weeks later, put the low end at about EUR 178 billion. Standard assumptions about avoidance give roughly EUR 160-195 billion, and far less if only the countries then backing the idea joined. It is still a lot of money, but the article presents the best case as if it were the expected result.

Technical detail and evidence

Why it matters, in detail. Upward bias in the expected-revenue reading of C3. Relative to the EUR 222.6bn mechanical figure: - -20% at the author's own low end; - -12% to -29% for semi-elasticities of 7-17 applied to the 1.7%-of-wealth increase (Saez-Zucman 2019; PWBM 2019; Advani-Tarrant 2021); - -35% to -59% if statutory valuations were 70-50% of Forbes values (estate-tax evidence, contested: us-wealth-taxes-11). Coverage is the largest lever: the named supporters plus Brazil hold ~8.7% of Forbes 2024 wealth (~EUR 20bn mechanical) unless untested backstop rules work. Offsetting factors: heterogeneity in current rates raises minimum-tax revenue above the average-based figure (G20 fn. 15), and Forbes wealth rose to $16.1tn (2025) and $20.1tn (2026).

Evidence. Cahiers p. 26. G20 blueprint §2.3 pp. 24-25 and Table 1 p. 27 ('$242 billion' well enforced; '$193 billion' with 20% avoidance/evasion). GTER 2024: 'nearly $250 billion' (text) vs Table 2 total $214bn. Recomputation in scripts/reproduce.py (blocks R4 and R4_sensitivity): - mechanical EUR 222.6bn; - 15% haircut (Saez-Zucman 2019 practice): EUR 189bn; - G20 low end: EUR 178bn; - semi-elasticities 8 / 13 / 7-17: EUR 192bn / EUR 173bn / EUR 196-158bn; - tax-assessed values at 70% / 50% of Forbes: EUR 144bn / EUR 92bn (Auditor A's EUR 111bn halves the net yield rather than the base); - Aghion et al.'s 0.26 retention (French context, contested): EUR 58bn. Coalition share 8.7% of Forbes 2024 wealth: scripts/coalition_share.py. VOA, 20 May 2024; Reuters via swissinfo, 23 May 2024.

Why this severity. The arithmetic is right, but the headline is the top of the author's own range. Under defensible avoidance assumptions the yield is roughly 10-30% lower (~EUR 158-196bn), and lower still if tax-assessed values fall short of Forbes values or coverage is partial. 'Significant returns' survives qualitatively; the magnitude is an upper bound.

The authors' response. The G20 blueprint reports $193-242bn (Table 1). It argues that a coordinated standard without exemptions, backed by exit taxes and 'tax collector of last resort' rules, limits avoidance and can work without universal participation (executive summary; §2.3; §4.3). On the French variant, Zucman argued that a tax without exemptions leaves no way to escape (France 5, quoted by BFM TV, 16 Sep 2025, per corpus offshore-havens-17).

3moderatelikely · previously known · overclaiming

The article says the practical hurdles are small and support is building. The author's own detailed report lists unresolved problems: valuing private companies (about half of billionaire wealth), getting countries to share information, and dealing with countries that refuse to join. The article also leaves out that the United States had publicly rejected a global deal 11 days earlier. This doesn't change the numbers, but it makes the plan look easier and closer than it was.

Technical detail and evidence

Why it matters, in detail. C4 asserts low administrative cost and broad support. The author's blueprint names two first-order open problems: (i) cross-border beneficial-ownership and information exchange; (ii) non-participation, to be handled by untested 'tax collector of last resort' rules. Valuing private holdings (~50% of the base) relies on peer multiples, and wealth-tax practice shows undervaluation (Skar et al. 2025: median statutory value 61% of sale price under Norway's method). 'All possible forms of tax optimization' ignores the valuation, expatriation and concealment margins the blueprint itself discusses. The momentum paragraph omits the US position (40% of billionaire wealth). No numerical effect, but the adoption-probability framing is biased upward; subsequent events (Rio 2024, US 2025, France 2025) went against it.

Evidence. Cahiers pp. 25-26 (paras 7-8, 11-12). G20 blueprint executive summary p. 7; §4.1 p. 34; §4.3 p. 37. VOA (R. Garver), 'US will not support global tax on billionaires', 20 May 2024. Reuters via swissinfo, 23 May 2024. Treasury FY2025 Greenbook (11 Mar 2024) p. 83 (25% minimum tax on total income incl. unrealized gains, wealth > $100m). Later events: - Rio G20 Leaders' Declaration (18 Nov 2024) only commits to 'engage cooperatively to ensure that ultra-high-net-worth individuals are effectively taxed'; - US memorandum of 20 Jan 2025 (GovInfo DCPD-202500150) gives the OECD Global Tax Deal 'no force or effect'; - the French domestic variant was rejected by the Senate on 12 Jun 2025, with a Conseil d'État opinion of Oct 2025 (offshore-havens-21); - valuation errors for unlisted shares (Skar et al. 2025, offshore-havens-23).

Why this severity. Feasibility is one of the op-ed's main arguments ('Some may call this utopian ...'), but it is qualitative and leaves the diagnosis and the arithmetic intact. The overstatement is clear relative to the author's own report; how large the technical obstacles really are is contested rather than shown.

The authors' response. G20 blueprint §4 (pp. 33-39) proposes valuing private firms with listed-peer multiples and reportable transactions, adding beneficial-ownership data to country-by-country reports, and strengthening exit taxes plus 'tax collector of last resort' rules. Zucman told Reuters (23 May 2024) he hoped the US would come around, since Yellen had rejected 'global redistribution', which is not central to the proposal.

4minorconfirmed · newly identified here · internal inconsistency

The French numbers (about 150 billionaires with EUR 750 billion) don't come from the same list as the world numbers and don't match either of the two usual rich lists. The resulting estimate of EUR 15 billion for France is still in the right range (about EUR 12-15 billion or a bit more), so this is a small slip.

Technical detail and evidence

Why it matters, in detail. Mixed bases: a Forbes count of individuals/families for the world, a Challenges count of family fortunes for France. A consistent basis gives France 53 billionaires and EUR 621bn (Forbes) or 141-147 fortunes and more than EUR 811bn (Challenges). The EUR 15bn (2% x EUR 750bn) compares with EUR 11.6bn (Forbes, net of the 0.13% current personal tax) and at least EUR 15.2bn (Challenges lower bound, net). Bias of uncertain sign, within about 25%; no effect on C1 or the global totals.

Evidence. Cahiers p. 26. Forbes 2024 API data (scripts/reproduce.py: France n = 53, $673.5bn = EUR 620.6bn at the ECB rate of 31 May 2024). Challenges 2023 via the Centre Presse Aveyron / La Dépêche relay of 5 Jul 2023 (141 billionaires; EUR 1,170bn; top ten EUR 203 + 137.8 + 100 + 77.2 + 39 + 32 + 31.2 + 20 + 14 + 13.5bn). France yield: EUR 12.4bn gross / EUR 11.6bn net of the 0.13% current rate (Forbes basis); at least EUR 15.2bn net (Challenges lower bound). Auditor A's lower bound of EUR 798.7bn is tightened here to EUR 811bn.

Why this severity. A documentation and consistency slip. The EUR 15bn France yield stays within the EUR 11.6-15.2bn+ range spanned by the two sources (-23% to +1%); the global figures are unaffected.

5minorconfirmed · newly identified here · speculative assumption

The claim that billionaires pay 'about 0.3% of their wealth' in income tax comes from studies of just three countries (the US, France and the Netherlands) and is assumed to hold everywhere. That is a reasonable guess. But for more than half of billionaire wealth there is no direct evidence, so the revenue estimate could be off by about a fifth in either direction.

Technical detail and evidence

Why it matters, in detail. tau enters revenue as (2% - tau) x W. With W = EUR 13.09tn: - tau = 0.29-0.34%: EUR 218-224bn; - tau at the US level for all non-US billionaires (0.68%): EUR 173bn; - tau = 0: EUR 262bn. The inputs are for 2016 (FR, NL) and 2018 (US, a 'triangulated' Saez-Zucman estimate: us-wealth-taxes-14), applied to 2024 wealth. Bias of unknown sign.

Evidence. Cahiers p. 26. G20 blueprint §2.3 p. 24; App. B pp. 48-49; Fig. 3 p. 17 (bars read from the chart, +/-0.02pp). scripts/reproduce.py: - R3: 0.29% arithmetic; 0.33% wealth-weighted (US = 40.1% of wealth); 0.34% from GTER Table 2 (44/12,916); - R4 sensitivity: tau = 0 gives EUR 262bn; tau = 0.68% gives EUR 173bn. Auditor A's '~57%' outside the three countries recomputes to 54.8%.

Why this severity. The central value is a reasonable assumption, consistent with GTER and Nordic estimates, and the uncertainty runs both ways. Revenue moves by about EUR 13bn per 0.1 point of tau (about -22% / +18% at the extremes), which does not threaten the qualitative claim.

The authors' response. G20 blueprint §2.3 (pp. 24-25) states the assumption and calls it 'plausible given the widespread availability of the income tax avoidance mechanisms', cites Swedish and Norwegian estimates (Fig. 3) as support, and says 'additional studies in more countries ... would be valuable'.

6minorconfirmed · newly identified here · documentation gap

The official online record lists the chapter under the book's title instead of its real one ('A 2% tax for the planet's 3,000 billionaires'). This is a cataloguing slip by the publisher's platform, not the author's mistake, and it doesn't affect what the chapter says.

Technical detail and evidence

Why it matters, in detail. Metadata only: the DOI resolves correctly and author, pages and date are consistent across records. The cost is discoverability: title-based searches miss the chapter's billionaire-tax topic and its 2024 versions. No effect on C1-C4.

Evidence. api.crossref.org/works/10.3917/herm.plane.2026.01.0142 (title = container-title = 'Inégalités'); Cairn OAI GetRecord oai:cairn.info:HERM_PLANE_2026_01_0142 (dc:title); Cairn METS HERM_PLANE_2026_01 (structMap label, pp. 141-144).

Why this severity. A labelling error in the registration metadata, attributable to the registrant (Cairn/Hermann), not the author. It has no bearing on any claim in the chapter.

7minorspeculative · newly identified here · other

The 2026 book chapter looks like a reprint of a 2024 newspaper article. If it was not updated or dated, readers in 2026 would get old numbers (by then there were more billionaires with more money) and an outdated picture of political support, which had since faded. We could not check whether it was updated.

Technical detail and evidence

Why it matters, in detail. If unchanged: the wealth base is understated by ~29% against Forbes 2026 ($14.2tn vs $20.05tn), so revenue is understated in levels, and the adoption narrative is stale (no G20 standard; US withdrawal from the OECD deal; French rejections). No bias against the author's quantitative claims; the risk is that the momentum framing (C4) misleads 2026 readers.

Evidence. Cairn METS record HERM_PLANE_2026_01 (table of contents) vs the Cahiers 32 table of contents (orbita PDF p. 6). Forbes 2026 data (scripts/reproduce.py, block I7). G20 Rio Leaders' Declaration, 18 Nov 2024. GovInfo DCPD-202500150. Corpus offshore-havens-21 (French votes).

Why this severity. Speculative, since the 2026 text was not seen. Even if true, updated figures would raise rather than lower the revenue claim; only the political framing would be stale.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

I5 (minor, misrepresentation_of_source): '$8,410 for Elon Musk in 2018' is attributed to ProPublica, which reports that Musk paid no federal income tax in 2018; the figure traces only to a deleted Zucman tweet. — The figure is in the cited ProPublica article (Eisinger, Ernsthausen & Kiel, 'The Secret IRS Files', 8 Jun 2021). Its embedded chart 'The Ultrawealthy by the Numbers' gives Musk's yearly 'Total Taxes Paid' as $30.4m (2014), $78,500 (2015), $424m (2016), $73,700 (2017) and $8,410 (2018). The text's 'no federal income tax' refers to income tax only ($68,000 in 2015, $65,000 in 2017, $0 in 2018). The op-ed presents the examples as 'meager contributions ... to the Treasury' (French: 'maigres contributions ... aux recettes du Trésor public'), which matches ProPublica's total-taxes measure. At most a trivial labelling nuance, and the income-tax figure ($0) would make the point more strongly. Verified in Auditor A's own cached copy of the article (scripts/reproduce.py, block I5).
Report §7 (noted, not counted as issues): the Cahiers translation calls the GTER the 'Global Tax Avoidance Report', renders 'une dizaine' as 'a dozen' and the child tax credit as 'family allowances'. — Correct for the 2024 English translation but not attributable to the target 2026 French chapter. The French original gives the correct title ('Rapport mondial sur l'évasion fiscale 2024') and 'une dizaine'. 'Allocations familiales' is the author's own French wording, a loose but not misleading rendering of ProPublica's '$4,000 tax credit for his children'. Not counted.
I1 sensitivity: valuation shortfalls like those in estate-tax data (~50% of Forbes) 'would roughly halve the yield (EUR 111bn)'. — Arithmetic corrected rather than accepted. Applying a 50% valuation ratio to the base, with current taxes unchanged, gives (2% x 0.5 - 0.3%) x EUR 13.09tn = EUR 92bn, not EUR 111bn (A halved the net yield). Either way it is a contested scenario, since Saez-Zucman read the estate-Forbes gap as avoidance; it is reported as such under F2.
No body-level core finding can be authenticated; the 2024 Cahiers Sens public article is only a 'candidate antecedent' and must not be used as a substitute version, so no substantive assessment is possible. — Partly rejected: the identification is strongly supported. (1) The Cairn record gives the 2026 chapter the op-ed's title. (2) The Hermann volume Inégalités. Que faire ? (Planel & Wormser) is the French edition of Cahiers Sens public 32, Inequality: What is to be done (curated by Planel, directed by Wormser), with the same contributors plus Rodrik and a chapter of the same four-page length. (3) Le Monde's lede reads 'une proposition du Brésil de taxation à 2 % des 3 000 milliardaires de la planète'. Auditing the full 2024 text is therefore informative. We agree that edits made for 2026 cannot be excluded, and the verdict is explicitly conditional on this.

Audit notes

What reproduced. Reproduced from Forbes 2024 list data (API JSON, valuation date 8 Mar 2024) and ECB reference rates: - 2,781 billionaires (claim ~3,000); - $14,209.6bn = EUR 13,094bn at 31 May 2024 (claim ~EUR 13tn); - personal-tax rate 0.29% (G20 three-country mean), 0.33% (wealth-weighted), 0.34% (GTER Table 2) (claim ~0.3%); - 1.7% x EUR 13.09tn = EUR 222.6bn ($241.6bn), identical to the G20 blueprint's perfect-enforcement $242bn (claim ~EUR 220bn); - EUR 15bn = 2% x EUR 750bn (arithmetic only). Not reproduced from any single source: the France base and count (Forbes 2024: 53 billionaires, EUR 621bn; Challenges 2023: 141 fortunes, at least EUR 811bn). The ProPublica figures reproduce exactly, Bezos 2011 income tax $0 and Musk 2018 total taxes $8,410 (article chart data). The all-in-tax-to-wealth conversion (1.7-2.0% of wealth) and the revenue sensitivities were recomputed from the author's published parameters. Country effective tax rates (IPP/CASD, CBS, IRS) rest on confidential microdata and were checked only against published figures (G20 Figs. 1-3 read from charts). Not done: reproduction from the 2026 chapter itself, which could not be obtained. Scripts: final/scripts/reproduce.py (output reproduce_output.json) and final/scripts/coalition_share.py.

Limits of this audit. The 2026 Hermann chapter (pp. 141-144) could not be read. Cairn returns HTTP 403 (DataDome), the Numilog ebook page is behind Cloudflare, the Google Books API quota was exhausted, and there is no Wayback capture; access controls were not circumvented. The audit therefore relies on the 2024 text: the full English translation (Cahiers Sens public 32, orbita PDF archived 17 Jan 2025) and ~44% of the French original (Le Monde Wayback captures of 31 May 2024). The numbers paragraph is available only in English. Identification rests on the Cairn title, the volume's one-to-one correspondence with Cahiers 32 and the Le Monde lede; edits for 2026 are unknown, and Hermann page references cannot be given. Other caveats: - G20 Figure 1-3 values were read off rendered charts (about +/-0.5pp for Fig. 1, +/-0.02pp for Fig. 3); - Challenges figures come from press relays, and the source of the France figures could not be identified; - country effective-rate estimates rest on confidential microdata not accessible here; - sensitivity scenarios borrow elasticities and valuation ratios from other settings and illustrate ranges; they are not estimates; - corpus critique entries were used as documented there (some critic texts were not re-read).