Gabriel Zucman · Paris: Le Seuil, October 2025; English edition Basic Books (UK), May 2026 (trans. Sara Freitas)

Les milliardaires ne paient pas d'impôt sur le revenu et nous allons y mettre fin / We Need to Tax Billionaires

12 problems found: 3 moderate · 9 minor

What the paper claims

Billionaires pay (almost) no personal income tax. In France the ~75 top households (about the billionaires) paid about EUR 300m of income tax, CEHR and CSG/CRDS in 2016, about 2% of their economic income and 0.1% of their wealth. If they all left for a tax haven, France would lose about 0.03% of its tax revenue. Comparable rates are 0-2% in the Netherlands, Italy, Brazil and Scandinavia, and 9% in the US. The mechanism is holding companies plus the 95% participation exemption (a 1.25% effective tax on intra-group dividends).
All taxes included, French billionaires pay about 25% of their income, half the 51% paid by the average French person ('deux fois moins'). All other social groups pay roughly 45-50%+, so the system is regressive only at the very top. Billionaires' contribution to French public finances is only 13% of their income, the other ~12% being CIT paid abroad.
US billionaires pay 24% of their income in taxes, against 30% for the average taxpayer. The 2019 Saez-Zucman estimate was later confirmed.
Top fortunes have exploded: the Challenges 500 families went from 6% of GDP (1996) to 12% (2010) and 42% (2024), growing 10%/yr against 4.5%/yr for average wealth. Low taxation 'powerfully contributed' to this ('a major contributory factor'). The EN edition adds UK and US analogues.
Remedy: a 2% minimum tax on wealth above EUR 100m (~1,800 French households) that credits personal taxes already paid. At a 6% average return it equals a 33% income tax and brings total rates to 50-55%, i.e. parity, and 'it is impossible to go below 2%'. Expected revenue is about EUR 20bn a year in France (15-25), USD 300-380bn worldwide and EUR 67bn in the EU, 80% of it from billionaires. The tax would be 'virtually impossible' to avoid.

Every problem found

1moderatelikely · previously known · methodology

The book says billionaires pay half the overall tax rate of the average French person, but the two figures do not count the same taxes: the billionaires' figure leaves out several taxes that are counted for everyone else, such as VAT on spending, some business and property taxes, and inheritance taxes. Measured the same way, billionaires in 2016 paid roughly half to two-thirds of the average rate, so the gap is real and large but probably smaller than 'half'. Cuts to company taxes since 2016 would lower billionaires' rate today, which partly offsets this.

Technical detail and evidence

Why it matters, in detail. Mechanism: a numerator-concept mismatch in C2's headline ratio. The billionaire rate (Bach et al. 2025; 25.1% for the top 75 households in 2016) counts PIT, CEHR, CSG/CRDS, flat withholding, wealth tax, non-contributory SSCs and CIT, but omits VAT and other product taxes, other production taxes, property and housing taxes, contributory SSCs and transfer taxes, all of which are in the 51% PO/NNI comparator (52.4% in 2016). Direction: the gap is overstated. Magnitude (adjudicator, 2016): adding non-payroll production taxes (Eurostat D29-D29C, EUR 66.0bn = 3.5% of NNI, allocated by factor income to the roughly half of billionaire income that is domestic: +1.0-1.7 points), VAT on consumption (+0.5-2) and the IPP transfer-tax simulation (+4.6) gives a like-for-like rate of 26.6-33.4%, i.e. 0.51-0.66 of the 51-52.4% average instead of 0.49, so the gap is up to about a third smaller than presented. Within the IPP concept, 25.1% against ~45% at the top-0.1% threshold gives 0.56 (P90-91 pay only 26.8% on that concept); the book's own alternative averages give 0.54-0.89. The sign of top-end regressivity survives every variant, and the unquantified post-2016 CIT cuts (large-firm implicit rate 19.3% -> 14.3%, 2016-2022) lower today's billionaire rate, partly offsetting the correction.

Evidence. Book: FR PT14-PT16 ('s'effondre ... à 25 % tout compris'; 'Pour 1 euro de revenu national ... 25 centimes ... contre 51 centimes'; 'comparer rigoureusement'); fn 15 (FR PT38); EN PT11-PT13. IPP: Bach et al. (2025, RFBerlin DP 70/25) s.4.1-4.2 (list of taxes; 'notional rents, property taxes, and housing taxes are not taken into account'); Note IPP 92 (2023) p.4 ('nous n'incluons pas tous les impôts (ni la TVA, ni les impôts de production)'). Figure labels read from the PDF (scripts/fig_labels.py): Fig. 6 total rate P90-91 26.8%, peak ~44.8% near P99.9, top 75 25.1%; Fig. A4 bequest simulation 29.7%. Adjudicator computation (scripts/adjudicate_repro.py, section D): other production taxes excluding payroll taxes (Eurostat D29-D29C) EUR 66.0bn in 2016 = 3.5% of NNI. Allocating them by factor income, as INSEE WP 2023-07 p.18 does, to the roughly half of top-75 income that is domestic adds 1.0-1.7 points. VAT on consumption adds 0.5-2 points, and the IPP transfer-tax simulation 4.6 points. Like-for-like billionaire rate 2016: 26.6-33.4%, i.e. 0.51-0.66 of the 51-52.4% average instead of 0.49. The book's own alternative averages (41/32/46/28%) give ratios 0.61/0.78/0.54/0.89. Within the IPP framework, 25.1% is 0.56 of the ~45% paid at the top-0.1% threshold.

Why this severity. The qualitative finding survives every like-for-like variant: billionaires pay far less than the average and than the top 0.1%. But the 'half' magnitude is overstated by up to about a third for 2016 (true ratio ~0.5-0.65). The post-2016 CIT cuts, which the book invokes but does not quantify, would partly offset this.

The authors' response. Zucman, 'Comment taxer les ultrariches ?', Le Grand Continent (2 Jul 2025), reproduced in the book (FR PT13, PT24; EN PT10, PT20): excluding pension contributions (41%) or VAT (32%), or netting cash transfers (46%) or pensions (28%), still leaves the average above billionaires' 25-26%, and no statistical agency excludes these levies. VAT and property taxes are negligible for billionaires (FR PT16). An update for post-2016 CIT cuts would lower their rate (FR PT17).

2moderatelikely · previously known · speculative assumption

The book expects the tax to raise about €20 billion a year in France, but that figure simply applies the tax rate to a magazine's rich list and assumes that no one would move abroad, hide wealth or have their fortune valued below the rich-list figure. Using the author's own cautious assumption from earlier work, the estimate falls to about €15-17 billion, the bottom of the book's range, and some published estimates are far lower. The tax would still raise billions of euros a year, but the headline number is optimistic, and the book's claim that the tax would be almost impossible to avoid rests on arguments about its design rather than on evidence.

Technical detail and evidence

Why it matters, in detail. Mechanism: C5's French revenue is a static mechanical score, EUR 20.9bn = 1.7% (the 2% floor minus ~0.3% of wealth already paid in personal taxes) x EUR 1,228bn (Challenges 2024), with zero avoidance, evasion or emigration, rich-list valuations and no residence screening; the book attributes the 15-25bn range only to data uncertainty. Base errors roughly cancel: removing non-residents gives EUR 18.9bn (Wertheimer, 9.4% of the base) or 17.9bn (Levy's count of nine top-25 fortunes, ~14%), while the omitted EUR 100-245m tier (~EUR 90-190bn) adds EUR 1-3bn, so the static figure is not clearly biased. The upward bias lies in presenting it as the expected yield. The author's own G20 low-end 20% haircut gives EUR 16.7bn (14-15bn without non-residents); Advani and Tarrant's preferred UK range implies ~12-29% base erosion for a 1.7-point rise; the CAE/Jakobsen et al. Scandinavian retention ratio (0.26) gives ~EUR 5bn, although the CAE cautions against extrapolating it to this design. Valuation risk is material for unlisted holdings: gross estates of single Forbes decedents are 39% of Forbes wealth, Norwegian statutory values of unlisted shares a median 61% of sale prices, and in the French wealth tax responses to weak disclosure were ten times larger than rate responses (Garbinti et al. 2026). Concentration (Arnault and Hermès families 28% of the base, top 10 56%) amplifies these risks, and the Oct 2025 Conseil d'État opinion (ability to pay; 10-year exit tail vs Art. 49 TFEU) undercuts the anti-exile clause behind 'virtually impossible to avoid'. Net: under the author's own low-end assumption the expected yield sits at the bottom of the stated range (~EUR 15-17bn), revenue of ~0.5-0.7% of GDP survives, and the diagnosis (C1-C3) is unaffected.

Evidence. Book: FR PT27, PT31-PT32; EN PT23, PT26-PT27. Senate report 689 (2024-25), sole article IV, 'Un dispositif trop sensible au risque d'exil fiscal' (1,600-1,800 households; 0.3% -> 1.7% x 1,228bn ~ 20bn; range 'compte tenu d'incertitudes dans la méthodologie du magazine Challenges'). G20 blueprint s.2.3 p.25 (low-end 20% loss). Base: Bach et al. (2025) s.2.1.4 warn that Challenges 'includes a large number of corporate owners who are French citizens but not French residents' and aggregates extended families. Challenges 2024 (AFP, 18 Jul 2024) puts the Wertheimer family (Chanel) at EUR 115bn, 9.4% of the base; Challenges (29 May 2026) says the Wertheimers, Castel and Edmond de Rothschild are no longer French tax residents. The same article reports A. Levy's count of nine top-25 fortunes domiciled abroad (EUR 173bn in 2025, ~14%). Excluding them cuts the static score to EUR 18.9bn / 17.9bn. The Senate computation also omits the ~EUR 90-190bn held between EUR 100m and the EUR 245m Challenges threshold (+EUR 1-3bn), so the static figure is roughly unbiased. Behaviour: G20's own 20% gives EUR 16.7bn (EUR 14-15bn with non-residents excluded). Advani and Tarrant (2021) give a preferred UK range of 7-17% base erosion per point of tax, i.e. ~12-29% for 1.7 points (corpus optimal-tax-behavior-20). CAE Focus 118 pp.16-17 / Jakobsen et al. (2024) find Scandinavian responses cost 74% of mechanical revenue, i.e. ~EUR 5bn here, but warn against extrapolating to this design. Garbinti et al. (2026, Zucman co-author) find tiny rate elasticities in the French wealth tax but responses to weak disclosure ten times larger. Valuation: Balkir et al. (2025) find gross estates of single Forbes decedents at 39% of Forbes wealth; Skar et al. (2025) find Norwegian statutory values of unlisted shares at a median 61% (mean 109%) of sale prices. Concentration: the Arnault and Hermès families hold 28% of the base and the top 10 hold 56%. Legal: on 31 Oct 2025 the government read a Conseil d'État opinion that the contribution fails the ability-to-pay requirement and that a 10-year post-departure tail is incompatible with Art. 49 TFEU (Public Sénat, 3 Nov 2025). Computations: scripts/adjudicate_repro.py section F.

Why this severity. The static arithmetic is about right and the order of magnitude (~0.5-0.7% of GDP) survives. Under the author's own low-end assumption the expected yield is ~EUR 15-17bn, the bottom of the stated range, and pessimistic but published scenarios are far lower. This weakens a headline policy number and the 'no escape' claim, not the diagnosis.

The authors' response. Zucman (France 5, quoted by BFM TV, 16 Sep 2025): past wealth taxes were riddled with exemptions, while a no-exemption tax from EUR 100m leaves no way to escape. To the Senate rapporteur he described exile effects as small (Senate report 689: a 1-point tax rise cuts the number of taxpayers by ~2% in the long run). The G20 blueprint allows 0-20% losses. C. Landais (as reported by France 24) said the Danish case is not comparable to a minimum tax on total wealth at market value.

3moderatelikely · previously known · overclaiming

The book says a 2% yearly tax on large fortunes is exactly the rate at which billionaires would pay the same overall share of their income in tax as the average person, but that only works if their fortunes produce income equal to about 6% of their value each year. For the two largest French family fortunes, which produce less, the new tax plus company taxes would take roughly 70% to 100% of the income their fortunes generated in 2024, which invites a legal objection (French constitutional rules forbid confiscatory taxes) that the book does not discuss; for fortunes that produce more, 2% would not be enough to reach equality. A tax of about this size would make the system fairer on average, but 2% is not the precise 'equal treatment' rate the book presents it as.

Technical detail and evidence

Why it matters, in detail. Mechanism: the parity claim is an identity, floor/income = 2%/y, and because only personal taxes are credited the total rate is ~2%/y + CIT/income. At the assumed average yield y = 6% this gives 33% + CIT = 50-55%, but the result is very sensitive to y and to firm-level CIT. On the book's own 2016 numbers (y ~5.4%, CIT 23% of income) a 2% floor gives a 60% total rate (54% with CIT at ~17% after the cuts); parity at 51% would need a floor of only ~1.5-1.85%. For the dominant low-yield fortunes at 2024 valuations the overshoot is large: Hermès family (y = 2.8%) 72% of pre-tax income, ~101% with CIT; Arnault/LVMH (y = 4.7%) 43% and ~71%; together 28% of the base. At the G20 simulation's 9% billionaire yield the floor is 22% of income and the post-floor rate ~39%, below the French average. Plausible parity floors therefore range from under 1% to over 2.5%, so neither 'impossible to go below 2%' nor 'une simple mise en conformité' follows, and the 70-100% burdens expose the confiscatory-tax objection (Conseil constitutionnel 2012-662 DC, raised in the June 2025 Senate report) that the book does not discuss. Direction: the average effect of the floor (removing top-end regressivity) survives; what fails is the claimed precision and uniformity of 2% as the parity rate (C5).

Evidence. Book: FR PT28 ('le taux de rendement sur la fortune s'élève en moyenne à 6 % ... 33 % ... 50 % - 55 % ... impossible d'aller en dessous de 2 %'); EN PT24; EN PT29 (payment in shares). Hermès FY2024 release: profit before tax EUR 6,432m, income tax EUR 1,845m; family stake ~2/3 (Hermès URD 2024 as cited by Auditor A; not re-read). LVMH 2024 financial documents: operating profit 18,907, net financial expense 792, income taxes 5,157 (EUR m); Arnault family group 49.00% of capital. Challenges 2024: Arnault EUR 190bn, Hermès family EUR 155bn. G20 blueprint App. B (9% / 11% income-to-wealth). Senate report 689 IV.A-B (case law 2012-662 DC; AFEP: over half of SBF 120 firms pay dividends below 3%). Public Sénat (3 Nov 2025) on the Conseil d'État opinion. Computations: scripts/adjudicate_repro.py section E.

Why this severity. The proposal's logic survives: a floor of roughly this size would remove the top-end regressivity on average. But the claim that exactly 2% is the parity rate is not robust. Parity floors range from under 1% (low-yield giants) to over 2.5% (high-yield fortunes), and the largest fortunes would face 70-100% total rates, which exposes the omitted confiscatory-tax objection.

The authors' response. The book says 'en moyenne' and allows payment in shares for owners of unprofitable firms (EN PT29). In the Aghion debate (Challenges, Sep 2025) Zucman conceded that edge cases need special payment arrangements. The G20 blueprint (p.29) says a common floor does not guarantee progressivity in every country. The Assemblée rapporteur E. Sas (cited in Senate report 689) argued that 7-10% returns distinguish the floor from the ISF precedents.

4minorconfirmed · newly identified here · methodology

A section heading in the book says billionaires contribute only 13% of their income to French public finances, but the calculation leaves out the taxes their companies pay abroad while still counting the money used to pay those taxes as part of their income. Calculated consistently, the figure is about 15%, still far below the roughly 45-50% paid by other groups, so the book's point stands.

Technical detail and evidence

Why it matters, in detail. Mechanism: the 13% is France-only taxes (the 25% total minus ~11.6 points of CIT paid abroad = 13.6%) over worldwide pre-tax income, so foreign profits stay in the denominator gross of foreign tax, whereas the book frames rates per euro of national income, in which foreign profits enter net of foreign tax. On that basis the consistent ratio is 13.6/(100 - 11.6) = ~15.3%. Direction and magnitude: the book understates billionaires' French contribution by ~2 points; against ~45-50% for other groups the contrast is unchanged. The IPP domestic-only variant (33.8%) drops foreign income and re-ranks households, so it answers a different question and is not the relevant correction. Affects a secondary C2 headline only.

Evidence. FR PT17 ('non pas à 25 % de leur revenu, mais à 13 % seulement – les 12 % restant étant payés à d'autres pays'); EN PT14. Bach et al. (2025) s.5.1 ('almost half of overall CIT among the top 75' paid abroad). Computation: 13.6 / (100 - 11.6) = 15.3% (scripts/adjudicate_repro.py section C).

Why this severity. The figure moves from 13% to ~15%, and its comparison with other groups (~45-50%) is unaffected. It is a framing slip in a secondary headline, not a change in the conclusion.

The authors' response. The book itself notes that it is not shocking in itself that half of billionaires' taxes are paid abroad (FR PT17); no response on the denominator.

5minorconfirmed · newly identified here · arithmetic or transcription

In one sentence the English translation says billionaires provide 0.3% of France's tax revenue, when the French original, and the English edition's own introduction, say 0.03%, ten times less. It is a printing or translation slip and does not affect the book's conclusions.

Technical detail and evidence

Why it matters, in detail. Mechanism: a factor-of-ten transcription slip in the English body text (EN PT15: '0.3 per cent' of tax revenue), contradicting the French original (0,03 %), the English Introduction (p.3) and the same sentence's 0.016% of national income. Recomputed: EUR 300m = 0.029% of 2016 taxes and compulsory SSC and 0.016% of 2016 NNI. Direction: the slip inflates billionaires' stated contribution tenfold in one sentence; no effect on any estimate or on C1, since the correct figure appears elsewhere in the same edition.

Evidence. EN PT15; FR PT18 ('0,03 % des recettes fiscales de la France, 0,016 % de son revenu national'); EN Intro p.3. Recomputed: EUR 300m = 0.029% of 2016 taxes and compulsory SSC, 0.016% of 2016 NNI.

Why this severity. Isolated translation/typesetting error; the correct figure appears elsewhere in the same edition.

6minorconfirmed · newly identified here · data error

The book says France's former wealth tax applied to fortunes above €1.2 million in 2017; the actual threshold was €1.3 million. This small factual error in background material does not affect any of the book's findings.

Technical detail and evidence

Why it matters, in detail. Mechanism: a contextual statutory error. Both editions give the 2017 ISF entry threshold as EUR 1.2m of net wealth; liability applied to net taxable wealth above EUR 1.3m, with the schedule starting at EUR 0.8m (BOFiP BOI-PAT-ISF-40-10). Magnitude: EUR 0.1m in a historical comparison; no bearing on any estimate or core finding.

Evidence. FR PT30; EN PT25. BOFiP BOI-PAT-ISF-40-10 (version in force through 2017): 'patrimoine net taxable ... supérieur à 1,3 M€'.

Why this severity. Contextual factual slip with no bearing on the conclusions.

7minorconfirmed · newly identified here · internal inconsistency

Some key figures in the book differ, without explanation, from those in the author's other reports and presentations: for example, billionaires' overall tax rate (25% in the book, 27% elsewhere), the income their wealth is assumed to produce each year (6% of its value versus 9%), and the expected revenue from a UK version of the tax (£15 billion in one presentation, £10 billion in a later report). None of these changes the French conclusions much, but together they suggest the policy numbers are less settled than the book's precise figures imply.

Technical detail and evidence

Why it matters, in detail. Mechanism: unreconciled parameters across the book and the author's companion documents. French billionaires' total rate: 25% in the book (Bach et al. 2025) vs 27% in the G20 blueprint and 27.4% on the May 2026 LSE slide (older IPP vintage), ~2 points. Income/wealth yield: 6% in the book vs 9% in the G20 simulation, which moves the 2% floor from 33% to 22% of income and drives the parity result (F3). UK: ~20% of GDP and ~0.4% of GDP of revenue in the book vs ~18% and GBP 15bn on the slide vs 16% for the top 300 resident households and GBP 10bn in Tippet and Zucman (2026). The EU note's French billionaire wealth (EUR 695bn for 147 billionaires) equals Challenges' top 10 alone (EUR 693bn); its EUR 19.4bn France row matches the Senate-method EUR 20.9bn only because the global 1.55 multiplier offsets the understated base. Direction: none systematic; each gap is small for the French headlines (C2, C5), but together they show that the policy parameters are less settled than the book's point estimates suggest.

Evidence. FR PT16, PT28; EN PT17, PT24. G20 blueprint p.11 and App. A-B. LSE slides 3 and 8 (20 May 2026). Tippet and Zucman (2026) pp.1-3. Parrinello, Varaschin and Zucman (2025) Table 1 and notes. Challenges/AFP 18 Jul 2024.

Why this severity. Each discrepancy is small for the French headlines or concerns companion materials. Together they show that the policy numbers are less settled than the book's precise figures suggest.

8minorconfirmed · newly identified here · selective presentation

The English edition, published in May 2026, still mentions only that the French National Assembly approved the tax in February 2025 and that the Senate blocked it; it leaves out that the Assembly rejected it in a budget vote in October 2025 and that the government's top legal adviser, the Conseil d'État, had found legal problems with it. This does not affect the book's evidence on how little billionaires pay, but English readers are not told about setbacks that bear on whether the proposal can be put into practice.

Technical detail and evidence

Why it matters, in detail. Mechanism: version currency. The May 2026 English edition reports only the February 2025 Assemblée adoption and the Senate block, omitting the Assemblée's 31 Oct 2025 rejection (228-172) of the 2% floor amendment to the 2026 budget and the Conseil d'État opinion read in that debate (contrary to the ability-to-pay principle; 10-year post-departure clause contrary to EU freedom of establishment). No effect on the empirical estimates (C1-C4); it bears on the feasibility of the remedy (C5), see F2 on the anti-exile clause and F3 on the confiscatory-tax objection.

Evidence. EN PT23. Public Sénat, 3 Nov 2025 (vote 228-172; quotations of the opinion).

Why this severity. Context only, with no effect on the empirical claims; it bears on the feasibility of the remedy (see F2, F3).

The authors' response. The French edition went to press (24 Oct 2025) before the 31 Oct vote; no response located for the English edition.

9minorconfirmed · newly identified here · documentation gap

The book does not explain how it arrived at several of its own numbers, such as the €15-25 billion revenue range, the 1,800 households that would pay the tax, or the claim that 80% of the money would come from billionaires. This makes the book harder to check, but no error in its conclusions was traced to it; the 80% figure, for instance, looks plausible when rebuilt from the rich list.

Technical detail and evidence

Why it matters, in detail. Mechanism: verifiability. Several book-specific figures (the EUR 15-25bn range and its method, the 1,800 households, the 80% billionaire share of revenue, the 46/30/28% net average rates, the UK series and the 13%) have no appendix or single cited source and must be reconstructed from the G20 blueprint, the EU note, the Senate hearing and Le Grand Continent. Of these, the 80% is plausible for France (~81-88% on a Pareto interpolation of Challenges 2024, vs 64% from the EU note's global multiplier); the 46/30/28% rates are only approximately consistent with INSEE WP 2023-07; and the 1,800 count, the 15-25bn method and the UK series could not be reproduced. No demonstrated bias in C2 or C5; the cost is that the policy numbers cannot be independently verified.

Evidence. FR notes PT37-PT38; EN notes PT31-PT35; no appendix or replication link on the author's book page; scripts/adjudicate_repro.py section I.

Why this severity. Limits verification; no demonstrated effect on conclusions. This is typical for a 60-page essay whose underlying studies are public.

10minorconfirmed · newly identified here · overclaiming

The title says billionaires pay no income tax, but the book itself explains that they do pay a little: about 2% of their full income in France (counting profits kept inside their companies) and about 9% in the United States. The title exaggerates for effect, but the underlying point, that billionaires pay very little income tax compared with what they really earn, is well supported.

Technical detail and evidence

Why it matters, in detail. Mechanism: rhetorical overstatement of C1. The title ('ne paient pas d'impôt sur le revenu') and the EN Introduction ('never really been subject to income tax') imply a zero, while the body concedes 'pas, à strictement parler, zéro': personal taxes of ~2% of economic income in France (EUR 300m on ~EUR 16.3bn, 1.84%) and ~9% individual income tax in the US (Balkir et al. 2025: income and payroll taxes 8.0% for the top 100 and 11.0% for the top 400) under the personal holding company tax in force since the 1930s. Rates are near zero only relative to economic income. Direction: overstates the level, not the finding; the substantive C1 estimates are unaffected.

Evidence. Title; EN Intro p.3; FR PT18, PT25-PT26; Balkir et al. (2025) Table 2 (individual income and payroll taxes 11.0% of economic income for the top 400, 8.0% for the top 100).

Why this severity. This is rhetoric that the text itself qualifies. The substantive claim (personal taxes ~2% of economic income in France and similar countries) is well supported.

The authors' response. The book's own qualification (FR PT18: 'Commençons par lever toute ambiguïté : ces derniers ne paient pas, à strictement parler, zéro').

11minorlikely · newly identified here · overclaiming

The book says low taxes were a major reason why the wealth of France's 500 richest families grew more than twice as fast as average wealth between 1996 and 2024. Using the book's own tax figures, lower taxes account for only about a sixth to a third of that faster growth; the rest reflects other factors, such as rising values of luxury-goods companies and changes in who is on the list. Taxes still explain a meaningful part, and the figures on how much top fortunes grew are correct.

Technical detail and evidence

Why it matters, in detail. Mechanism: C4 attributes much of the Challenges top-500's excess growth (10.25%/yr vs 4.5%/yr for average wealth, 1996-2024, a ~5.7-point annual gap) to low taxation ('puissamment contribué'; EN 'a main driver'), citing a G20 accounting that is global and concerns a much smaller excess return. Treating the book's own tax differentials as an annual drag on wealth, the personal-tax gap (~1% vs 0.1% of wealth, 0.9 points a year) explains ~16% of the gap; paying the average total rate (51% vs 25% at a 5.4% yield, 1.41 points) ~25%; and the full 2% floor every year ~35%. Direction: the causal role of taxation in France is overstated; the remaining ~65-85% reflects valuation gains (notably in luxury) and the changing composition of a fixed-size list. Minor because the claim is hedged ('pas entièrement'), taxes still explain a non-trivial share, and the descriptive series (6% -> 12% -> 42% of GDP) reproduce.

Evidence. FR PT21 ('celle-ci y a puissamment contribué', fn 16); EN PT18 ('a main driver ... a major contributory factor'). G20 blueprint p.18 ('about half of the excess net return of billionaires owes to their relatively low individual capital tax rates'), a global accounting with a much smaller excess return. Computation: scripts/adjudicate_repro.py section H.

Why this severity. The statement is hedged ('pas entièrement') and taxes do explain a non-trivial share. The descriptive wealth-concentration figures are correct.

12minorlikely · newly identified here · misrepresentation of source

The book lists Italy among countries where billionaires pay almost no income tax, but according to the author's own earlier report that figure comes from a single person, Silvio Berlusconi, and the versions of the cited study that could be checked contain no figure for billionaires. This matters little, because the broader finding rests on studies of official tax records in France, the Netherlands, Scandinavia, Brazil and the United States.

Technical detail and evidence

Why it matters, in detail. Mechanism: Italy is one of the countries behind C1's international 0-2% pattern, but by the author's own G20 blueprint (fn 2, fn 21) the Italian billionaire estimate rests on a single observation (Silvio Berlusconi, 2020), and the cited Guzzardi et al. (JEEA 2024) shows no billionaire estimate in its working-paper version or published abstract (the JEEA full text was not accessible). Direction: Italy should carry essentially no evidential weight for C1. Magnitude: negligible for the pattern, because France, the Netherlands, Scandinavia, Brazil and the US rest on administrative studies.

Evidence. FR PT25; EN PT21 and note 3 (PT31). G20 blueprint fn 2 (p.10) and fn 21 (p.47), App. A. Guzzardi et al., LEM WP 2022/06 (no 'billionaire'/'Berlusconi'); JEEA 22(5) abstract (BOA Unimib record). JEEA full text not accessible.

Why this severity. France, the Netherlands, Scandinavia, Brazil and the US rest on administrative studies, so the international pattern survives.

The authors' response. G20 blueprint fn 2: Berlusconi's structures are available to and used by European billionaires generally.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

I3: the 0.03% figure omits its denominator, which is GDP (per Zucman's Kessel post), so reading it as a share of tax receipts understates the loss by a factor of 2.24-2.33. — The French body text states the denominator explicitly: '0,03 % des recettes fiscales de la France, 0,016 % de son revenu national' (FR PT18). The arithmetic matches it: EUR 300m is 0.029-0.030% of 2016 taxes and compulsory SSC and 0.016% of 2016 NNI. 0.03% of 2016 GDP would be ~EUR 670m, more than twice the EUR 300m the book reports. The Introduction's 'loss of tax revenue ... around 0.03 per cent' reads naturally as a share of tax revenue. The inconsistent formulation is the Feb 2025 Kessel post ('0,03 % du PIB'), not the book. Codex saw only pp.1-5 of the book.
I1 (magnitude): the 13% understates billionaires' French contribution by up to ~21 points, since the IPP domestic-only rate is 33.8%. — The domestic-only variant (Bach et al. Fig. A6) removes foreign income and foreign CIT and re-ranks households, so it answers a different question. The consistent correction to the book's statistic is ~2 points (13% to ~15%). The issue is kept as F4 but graded minor rather than moderate.
I9: billionaires' current personal taxes are given inconsistently as 0.1% (book), 0.2% (EU note) and 0.3% (Senate, G20) of wealth. — These rates refer to different populations: the French top 75 in 2016, all EU billionaires, and the French top 500 and global billionaires. Using 0.3% for the whole Challenges 500 is consistent with lower-ranked fortunes paying more, and it is conservative for revenue.
I9 / reproduction: the book's '80% of revenue from billionaires' is discrepant with the 64-65% implied by the G20 Table 1 and the EU note. — The 64-65% shares come from applying the global billionaire-to-centimillionaire multiplier (1.55) mechanically to France. A Pareto interpolation of the Challenges 2024 distribution (147 billionaire families, top 10 = EUR 693bn, entry threshold EUR 245m) gives a French billionaire share of ~81-88%. The figure is plausible; its lack of documentation is kept under F11.
I3 evidence: 70% of the Challenges base is held by the top 10. — The 70% comes from the Senate report. Challenges' own 2024 release (AFP, 18 Jul 2024) gives EUR 693bn of EUR 1,228bn, i.e. 56%. Concentration is still high (the top three hold 37%), but the figure is overstated.
I3: excluding non-resident wealth (Wertheimer, 9.4%) lowers the French estimate to EUR 18.9bn, a net upward bias of the book's figure. — Narrowed. The non-resident inclusion is real, and larger if A. Levy's count of nine top-25 fortunes is used (~14%). But the Senate-described computation also omits households between EUR 100m and EUR 245m (~EUR 90-190bn, +EUR 1-3bn), so the static ~EUR 20bn is not clearly biased. The substantive problem is the missing behavioural and valuation allowance, kept as F2.
I4: for most of the expected revenue the floor would far exceed parity (Hermès 101%, Arnault 73%). — Narrowed. The overshoot is demonstrated for families holding ~28% of the base (~35% with Bettencourt/L'Oréal), not for 'most' of the revenue. Yields of other large fortunes (e.g. CMA CGM, retail groups) were not computed and may be higher. Kept as F3 with this scope.

Audit notes

What reproduced. Reproduced (scripts/adjudicate_repro.py; log in work/repro_log.txt): - Macro and IPP-based numbers: 51.2% PO/NNI; EUR 300m = 1.84% of income, 0.029% of tax revenue, 0.016% of NNI; implied 5.4% yield. - The 13% arithmetic (13.6%; ~15.3% on a national-accounts basis); the Challenges ratios and 10.25% CAGR. - The G20 global range (USD 301-377bn), EU EUR 67.2bn, the France EU-note row EUR 19.4bn and the Senate-method EUR 20.9bn. - 2/6 = 33%; US BSYZ 23.8% vs 30.2%. - IPP figure labels read from the PDF: 25.1% top 75, 26.8% P90-91, 21.8% majority control, 29.7% bequest, 16.7% incidence, 33.8% domestic. Sensitivities computed: - Like-for-like PO-concept billionaire rate 26.6-33.4% (ratio 0.51-0.66 vs 0.49). - Parity: 2% floor gives 60% (54%) total on the book's own 2016 numbers; parity floor 1.5-1.85%. Arnault/LVMH 43%/71% and Hermès 72%/101% of 2024 pre-tax income. - French revenue: EUR 18.9bn / 17.9bn excluding non-residents, +EUR 1-3bn for the omitted EUR 100-245m tier, 16.7bn with 20% avoidance, ~5.4bn with a 0.26 retention ratio. - Growth gap: taxes explain 16-35%. Billionaire share of French revenue ~81-88%. Not reproducible: - Confidential IPP and BSYZ microdata; the 1,800-household count; the method behind the EUR 15-25bn range; the UK series and GBP 15bn. - The 46/30/28% net rates, beyond approximate consistency with INSEE WP 2023-07. - Firm-level yields beyond the two families examined.

Limits of this audit. Neither edition was available in full. The French and English texts were read through Google Books search-inside snippets (248 FR and 212 EN cached queries; 31 and 33 ebook pages touched) and the author-hosted English Introduction (pp.1-5). Passages not reached by any query (parts of the conclusion and notes, the gap in fn 15) were not audited, and 'no hit' only means no match in Google's index. Print page numbers are mostly unavailable, so locations are ebook page ids (PTn). The IPP and BSYZ microdata are confidential, so checks rely on published tables and figure labels. Challenges' raw lists are paywalled: totals come from AFP/press reports, the non-resident count is a critic's (A. Levy) as reported by Challenges, and the EUR 100-245m tier and billionaire wealth are Pareto interpolations. The Hermès family stake (~2/3) was not re-read, and firm yields use one year of profits and June-2024 valuations. Behavioural responses to a tax that has never existed cannot be estimated here; the haircuts are scenarios, not forecasts. Legal statements report the Senate report and the Conseil d'État opinion as quoted in the press; no legal determination is made. The JEEA version of Guzzardi et al. and the German edition (table of contents only) were not read.