Quentin Parrinello, Giulia Varaschin, Gabriel Zucman · EU Tax Observatory Policy Note, March 2025

Resources for a Safe and Resilient Europe: The Case for Minimum Taxation of Ultra-High-Net-Worth Individuals in the EU

9 problems found: 1 major · 3 moderate · 5 minor

What the paper claims

A 2% minimum tax on the wealth of people in the EU owning more than EUR 100 million would raise EUR 67 billion a year, 'a quarter' of the EUR 250 billion annual defence-spending increase estimated by Bruegel.
A 3% minimum tax would raise EUR 121 billion, 'close to 50%' of the Bruegel need.
A 2% minimum tax on centimillionaires 'would neutralize the regressivity of European tax systems'; a 3% minimum tax 'would make European tax systems slightly progressive'.
Member-State breakdown of the revenue (Table 1) for 23 EU countries.
Billionaires pay much lower all-in effective tax rates than other social groups in France, the Netherlands and Italy (tax systems are regressive at the top).
Supporting policy claims: a minimum tax 'addresses all potential forms of tax avoidance'; it 'would generally not raise liquidity issues' because UHNWI real returns exceeded 7% a year over four decades; it 'does not create any double taxation' because personal taxes are creditable and 'would only affect taxpayers who are both extremely wealthy and undertaxed'; 67% of Europeans support such taxes (Eurobarometer); France and the UK taxed the wealthy to fund WWII.

Every problem found

1majorconfirmed · newly identified here · overclaiming

The note says a 2% tax on fortunes above EUR 100 million would end the situation in which Europe's richest pay lower tax rates than ordinary people. Its own source says this is true only on a rough average dominated by the United States. It also warns that in the Netherlands and Italy people far below EUR 100 million already pay less than the middle class, which this tax would not change. The money estimates are not affected, but this fairness claim is overstated.

Technical detail and evidence

Why it matters, in detail. For C3 to hold, the all-in rates of the affected groups must reach those of lower groups. Under the source's App. B calibration, at 2% the remaining gap to the P50-P99 rate is 5-19pp for billionaires and 9-17pp for centimillionaires, and the dip in the Dutch and Italian P95-P99.99 groups (4-26pp below their middle classes) is not touched at all. The billionaire result is sensitive to the pre-tax-income/wealth ratio rho: break-even rho is 7.3% (FR), 6.5% (NL) and 4.6% (IT). France's own blueprint Fig. 2-3 values (individual tax about 0.12-0.13% of wealth and 1.7-1.9% of income) imply rho of about 6.3-7.6%, so French billionaires could reach rough parity. Dutch and Italian billionaires, centimillionaires and the sub-threshold top groups do not. The author's own argument that post-2016 corporate-tax cuts lowered billionaires' current rates (Le Grand Continent, 2 Jul 2025) would widen the gaps. '3% makes European systems slightly progressive' holds only for France (and Dutch billionaires).

Evidence. Note p. 3 (figure) and p. 4; Zucman (2024) pp. 11-12, 28-30, App. B pp. 48-49. The caveats are identical in the June 2024 original, whose App. B is on PDF p. 50. Adjudicator recomputation (scripts/digitize_figure.py, regressivity_check.py) applies App. B to the note's European profiles: individual tax is raised to the floor; pre-tax income equals 9% (billionaires) or 11% (other centimillionaires) of wealth; current rates come from blueprint Figs. 2-3. The method reproduces the blueprint's own 22% -> 39% result. At 2% with no avoidance, billionaires reach 47.9% (FR), 38.0% (NL) and 38.3% (IT), against population-weighted P50-P99 rates of 52.8%, 46.5% and 57.2%. Centimillionaires reach 44.0%, 32.7% and 40.0%. The P99-P99.99 groups (NL 26.3% and 20.6%; IT 45.1% and 36.9%) are untouched by a EUR 100m threshold. At 3%, billionaires reach 59.0%, 49.1% and 49.4%, and centimillionaires 53.1%, 41.8% and 49.0%. With the 10% avoidance used in App. B, all figures are 2-3pp lower. Auditor A's independent recomputation agrees within 0.4pp.

Why this severity. One of the note's headline claims is stated far more broadly than its own source allows. For the Netherlands and Italy it fails whatever the parameters, because their regressivity begins near P95, far below a EUR 100m threshold. The direction survives (a 2% floor sharply raises effective rates at the very top), and the revenue numbers are unaffected, so major rather than critical.

2moderateconfirmed · newly identified here · methodology

The note counts every billionaire with an EU passport as taxable in their home country. Many of them actually live in Switzerland, London, New York or Monaco, and taxes like the French bill the note cites apply to residents. About a fifth of the wealth in the note's base is therefore out of reach unless special rules are added. Correcting this lowers the EU totals by up to about a fifth and cuts some countries' figures by half or more.

Technical detail and evidence

Why it matters, in detail. Allocating by Forbes residence gives EUR 54.8bn (2%) and EUR 98.4bn (3%), -18.5%. On the more complete Forbes 2025 annual list (EU-citizen wealth +11.8%), residence-based revenue is EUR 61.9bn and EUR 111.2bn (-7.9%). Country rows at 2%: IE 1.43 -> 0.26; CY 1.20 -> 0.33; GR and NL about -47%; IT 8.34 -> 5.10; SE 4.62 -> 3.27; DE 16.9 -> 14.3; FR 19.4 -> 17.0; BE 1.01 -> 1.82; BG and EE -> 0. Forbes city is not legal tax residence. Source taxation of EU-situated assets and trailing or exit rules could recapture part of this, subject to tax treaties and EU free-movement law. About 19% of EU-citizen billionaires live abroad, against about 9% globally (blueprint p. 31), which also shows that relocation is a live margin for a regional tax.

Evidence. scripts/forbes_residence.py on the Wayback capture of the Forbes RTB API (3 Mar 2025, 22:46 UTC). It uses an independent city-to-country mapping and reproduces Auditor A's figures exactly; Auditor B's stricter city screen flags a subset (72 entries, $426.0bn). Examples: Kuehne ($36.6bn, Schindellegi CH); G. and A. Wertheimer ($70.4bn, New York); Pignataro ($32.7bn, St. Moritz); Bechtolsheim ($18.3bn, Palo Alto); Puech ($17.5bn, Martigny); Fredriksen ($15.6bn, London, counted in Cyprus); V. Adani ($13.8bn, Dubai, Cyprus); the Collisons ($20.2bn, San Francisco, Ireland); the Mistrys ($15.8bn, Mumbai, Ireland). Luxembourg's only listed billionaire lives in Baar (CH). The residence-based base including the French +$150bn is $2,062bn, vs $2,530bn in the note.

Why this severity. On the note's own snapshot the EU headline falls by 18.5%. That is below the quarter threshold, and a separate conservative input choice (an incomplete Forbes snapshot) offsets more than half of it. Several Member-State rows are wrong by 30-100%.

The authors' response. No direct response located. The blueprint (p. 37) proposes trailing taxation of long-term residents, and the authors' later FES brief (Varaschin, Parrinello & Zucman, 'Wealth taxes and high-net-worth individuals in Europe: Five Lessons', Dec 2025) recommends exit taxes and trailing-residence rules. If adopted, these would reach recent emigrants but not people who left long ago or never lived in the EU.

3moderateconfirmed · previously known · selective presentation

The note presents its totals as what the tax 'would raise'. In fact they assume nobody avoids or evades the tax, under-values assets or moves abroad. Its own source gives a range up to 20% lower and calls a large part of the estimate tentative. Real collections would very likely be lower; how much lower is uncertain.

Technical detail and evidence

Why it matters, in detail. With the source's 20% allowance the totals are EUR 53.8bn and EUR 96.6bn, or 21.5% and 38.6% of Bruegel's EUR 250bn against 'a quarter' and 'close to 50%'. Elasticities of 7-17 would imply 14-34% erosion at 2% over 4-8 years, and more at 3%. The critics' 74% loss comes from Scandinavian broad-base reforms and is not transferable: the CAE Focus itself warns against extrapolating it, and Landais, a co-author of that evidence, says so. Combined with F2 and F4, the source's low end gives about EUR 48bn and EUR 78bn. The note never labels its figures as mechanical or first-year.

Evidence. Zucman (2024) pp. 25-27 (Table 1: >$1bn $193-242bn and >$100m $302-377bn at 2%; >$100m $551-688bn at 3%); note pp. 3-4; scripts/sensitivity_combined.py; Forbes RTB privateAssetsWorth field ($1,307bn of $2,379bn). Literature: Advani & Tarrant (2021, Fiscal Studies 42(3-4): 509-537) put the elasticity at 7-17, about 14-34% base erosion at 2%. Bruelhart et al. (2022, AEJ: Economic Policy 14(4): 111-150) find large Swiss responses. Aghion et al. (Le Monde, 9 Sep 2025, via CAE Focus n. 118) apply a contested 0.26 retention ratio.

Why this severity. The headline is the top of the source's own range, not a central estimate. The source's low end is 20% lower, and the true long-run loss is unknown and could be larger for a regional tax. This is a material undisclosed qualification rather than a demonstrated large error.

The authors' response. Zucman (France 5, as quoted by BFM TV, 16 Sep 2025): a tax starting at EUR 100m with no exemptions leaves no possibility of escape. Landais (as reported by France 24, 27 Sep 2025): the Danish evidence 'has nothing to do with' the Zucman tax. Blueprint p. 27: short-run responses other than avoidance and evasion are negligible because wealth is a stock. These are design arguments, not estimates.

4moderateplausible · newly identified here · speculative assumption

To include people worth EUR 100 million to EUR 1 billion, the note multiplies each country's billionaire figure by one worldwide factor. The data behind that factor suggest the EU has less wealth in this bracket, relative to its billionaires, than the world as a whole, although these Europeans also pay less tax today. The 2% total may be roughly right, but the 3% total is probably too high by around a tenth, and small countries' figures are guesses.

Technical detail and evidence

Why it matters, in detail. Using m(t) = 1 + R(t - tau_c)/(t - tau_b) with the note's tau_b = 0.2%: R = 0.55 and tau_c = 1.2% give EUR 54.0bn and EUR 91.3bn; R = 0.66 and tau_c = 0.61% give EUR 65.5bn and EUR 105.5bn; the QE aggregate (R = 1.065) gives EUR 63.9bn and EUR 113.7bn. The lower European wealth ratio and the lower European tax rate offset each other at 2% but not at 3%. WID country tails above $100m are modelled (Pareto-interpolated and anchored on Forbes), so the EU ratio itself is uncertain, hence the 'plausible' rating.

Evidence. scripts/wid_centimillionaire_ratio.py and sensitivity_combined.py; Zucman (2024) pp. 25-27 and fn. 17; blueprint Fig. 2 (P99.99-99.999 and P99.999-99.9999: FR about 10.3% and 5.3%, NL 4.4% and 2.1% of income). The multipliers reproduce exactly as 1 + R(t - 1.2%)/(t - 0.3%) with R = 16.9/14.2.

Why this severity. EU-specific variants move the 2% headline by -20% to +10% (central -3%) and the 3% headline by -24% to -1% (central -13%). The 3% figure is therefore likely overstated by about a tenth, and the country rows are not credible for states with few billionaires.

5minorconfirmed · newly identified here · documentation gap

For France the note mixes two rich lists and adds $150 billion, apparently the Hermes family that Forbes leaves out, without explaining how. The adjustment is small and, if anything, cautious, so it does not inflate the totals.

Technical detail and evidence

Why it matters, in detail. Using Challenges' billionaire total would raise France's 2% figure to roughly EUR 24-32bn. Family-level totals overstate an individual-level base only where members fall below EUR 100m, which is unlikely for most of these families. Part of the French base is non-resident (the Wertheimers, Puech, Drahi, Perrodo; see F2).

Evidence. Note p. 4, Table 1 notes; Challenges 2024 ranking (public totals and top 10); scripts/forbes_residence.py (France: 49 Forbes entries, $579.8bn) and table1_recompute.py.

Why this severity. The adjustment is disclosed, small (5.9% of the headline) and conservative relative to the ranking it cites; only its documentation and the count are inconsistent.

6minorconfirmed · newly identified here · internal inconsistency

A few small slips in the table's notes and inputs change the totals by about 1% at most.

Technical detail and evidence

Why it matters, in detail. The largest is (a), -1.3% to -1.4% on both headlines; the others are below EUR 0.3bn.

Evidence. scripts/table1_recompute.py, wid_centimillionaire_ratio.py, sensitivity_combined.py; Forbes RTB of 3 Mar 2025.

Why this severity. Each item moves the headline by 1.4% or less.

7minorconfirmed · newly identified here · misrepresentation of source

The survey cited asked whether governments should tax the rich to help the poor, not about this particular tax, and the historical source mentions only Britain. Neither affects the numbers.

Technical detail and evidence

Why it matters, in detail. This affects only the political-support and historical-precedent framing of the case, not any estimate.

Evidence. ebs529_vol_A.xlsx, sheet QC16_1 (EU27 column); UK Parliament, 'The cost of war' (Wayback capture, 2025); Burilkov & Wolff (Bruegel, 21 Feb 2025), section 'The fiscal aspect'; Draghi report Part A.

Why this severity. Context only; no effect on the estimates or on C1-C5.

8minorconfirmed · previously known · documentation gap

The chart of who pays what comes from studies of single years (2015-2016), and the Italian billionaire figure is based on one man. Experts also disagree about which taxes should be counted. The basic pattern is well documented, but the chart is less precise than it looks.

Technical detail and evidence

Why it matters, in detail. Under narrow conventions (about 32% for the average French person), a 2% floor would more than close the French gap. Under the authors' all-tax convention it does not close the Dutch and Italian gaps (F1). By the author's own argument, post-2016 corporate-tax cuts mean the 2016 billionaire rates overstate current ones. The steepest drop in the figure rests on a single Italian taxpayer.

Evidence. Note p. 3; Zucman (2024) pp. 10-11, 15, 47-48; corpus offshore-havens-19. Zucman's reply, 'Comment taxer les ultrariches ?' (Le Grand Continent, 2 Jul 2025, read here), gives the average French rate as 52%, about 41% excluding pension contributions, about 32% also excluding VAT, and 45% net of transfers, against 26% for billionaires.

Why this severity. The core empirical pattern (personal taxes collapse at the very top) is documented in published studies and is not contested. The omitted caveats mainly affect how C3 should be calibrated, and the dispute over conventions cuts both ways.

The authors' response. Zucman (Le Grand Continent, 2 Jul 2025): the critics' arguments are 'unfounded' and the gap persists under alternative conventions; the 26% is if anything too high, because the French corporate-tax rate fell from 33% to 25% after 2016.

9minorlikely · previously known · overclaiming

The note reassures readers that the tax would not force anyone to sell assets and would never tax the same income twice. Those reassurances rest on long-run averages and on counting only personal taxes. Some owners of fast-growing but cash-poor firms could face real difficulties, as the author has acknowledged.

Technical detail and evidence

Why it matters, in detail. Owners whose cash yield is below 2% (low-payout or growth firms) would need to sell, borrow or pay in kind. Payment in kind trades liquidity against revenue timing and raises governance issues. Crediting corporate tax would be a different policy that cuts revenue sharply (critics' estimates for France: EUR 1.25-5.5bn, offshore-havens-18); not crediting it is a deliberate design choice, not an arithmetic error.

Evidence. Note p. 3; Zucman (2024) pp. 15, 17-18, 21; corpus offshore-havens-18, -20, -22.

Why this severity. These are supporting policy arguments with published author replies; they enter neither the revenue arithmetic nor C3.

The authors' response. Zucman (Challenges transcript of the 17 Sep 2025 debate with Aghion): the tax can be paid in shares, and edge cases need special payment arrangements. Zucman (Le Grand Continent, 2 Jul 2025; his 2025 book as quoted by Allegre & Timbeau 2026): not crediting corporate tax is deliberate, aiming at parity of total taxes. Blueprint p. 21: access to borrowing.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

Part of A's I1: at 2% every European profile stays regressive, with French billionaires about 5pp below the middle class. — Rejected in part. The French figure depends on the blueprint's 9% pre-tax-income/wealth assumption. France's own blueprint Fig. 2-3 values (individual tax about 0.12-0.13% of wealth and 1.7-1.9% of income) imply rho of about 6.3-7.6%, around the 7.3% break-even, so French billionaires could reach rough parity (49-57%). The overclaim (F1) stands on the Netherlands, Italy, centimillionaires, the sub-threshold top groups and the source's own caveats.
I6: the French reconciliation is a moderate issue. — Downgraded to minor (F5). The upgrade is disclosed, accounts for 5.9% of the headline, and is conservative relative to the Challenges billionaire total (at least EUR 875bn vs EUR 695bn). Forbes lists only one Hermes heir, so the $150bn plausibly stands for the missing Hermes family stake.
I5: liquidity and individual under-taxation is a moderate issue. — Downgraded to minor (F9). These are supporting policy arguments with published author responses; they do not affect the revenue figures or C3.
I4: the global-multiplier issue is 'confirmed'. — Confidence lowered to plausible (F4). The uniform multiplier is confirmed, but its bias at 2% is not demonstrated. The EU ratio is 0.55 or about 1.06 depending on the aggregation, lower European tax rates offset it, and WID country tails are modelled.
I3: volatility, because the Forbes snapshot was taken near a market high. — Not substantiated as a bias. EU-citizen Forbes wealth was $2,373-2,379bn on 1-4 Mar 2025 and higher on 8 Apr 2025 ($2,460bn) and 1 Sep 2025 ($2,797bn), partly because Forbes added people. Valuation volatility is a generic caveat that cuts both ways.
I3: the French critics' 0.26 retention ratio (-74%, EUR 17.5bn/31.4bn) as a sensitivity. — Not adopted as a plausible scenario. The ratio comes from broad-based Scandinavian reforms affecting the top 1-2%. The CAE Focus n. 118 itself warns against extrapolating it to a narrow tax on business assets, and Allegre & Timbeau (OFCE WP 10/2026) judge such extrapolation 'simply not possible'. Auditor A also flagged it as contested.
Considered and dismissed by Auditor B: a possible double count of Draghi's EUR 750-800bn with Bruegel's EUR 250bn. — Concur with B: the note never adds the two figures.
Considered and dismissed by Auditor B: legal infeasibility (ability-to-pay case law, EU limits on trailing or exit taxes; offshore-havens-21, -24) as a finding. — Concur with B. It is a real design constraint relevant to closing the residence gap (F2), but not an error in a conditional revenue estimate, and its revenue effect cannot be quantified here.

Audit notes

What reproduced. Reproduced exactly: all 120 computed cells of Table 1 and the EU totals (EUR 67.23bn / 120.77bn); the 'quarter' and 'close to 50%' shares (26.9%, 48.3%); and the 1.55/1.79 multipliers (Zucman 2024 Table 1, re-derivable from R = 16.9/14.2, tau_c = 1.2%, tau_b = 0.3%). Reproduced closely: Table 1 columns 1-2 from the Wayback capture of the Forbes Real-Time list of 3 Mar 2025 (identical counts for 22 non-French countries; wealth within +/-$1.5bn; total $2,528.6bn vs $2,530bn including the French +$150bn); and the blueprint's own progressivity simulation (22% -> 39%). Not reproducible or not supported: the composition of the French +$150bn (no files published); the authors' exact retrieval time; national centimillionaire bases (none observed); and C3, for which the note offers no calculation and which the source's method, applied to the note's own European profiles, does not support. EU-consistent static variants (Forbes 2025 list, residence allocation, EU centimillionaire ratio 0.66, Europe-only centimillionaire tax 0.61%) give about EUR 60bn (range about EUR 50-68bn) at 2% and about EUR 97bn (about EUR 84-110bn) at 3%. With the source's own 20% avoidance allowance these become about EUR 48bn and EUR 78bn (scripts/sensitivity_combined.py).

Limits of this audit. No authors' spreadsheet or code exists, so the reconstruction relies on the stated formula, and the French +$150bn cannot be decomposed. Residence is inferred from Forbes city fields, not tax records; people with no city were treated as EU residents, which favours the note. WID wealth above $100m by country is modelled and ends in 2021, and the EU ratio depends on the aggregation (0.55 as the EU-27 sum vs 1.065 for the Europe aggregate). Figures were digitized (about +/-0.5pp; only the top groups of blueprint Fig. 2 are reliable). Italy's current centimillionaire tax rate was proxied by the France-Netherlands average, and the App. B parameters (9%/11% income/wealth, 10% avoidance) are the source's, not validated here. Only the totals and top 10 of the Challenges ranking are public, so its billionaire subtotal is a bound. Behavioural responses, valuation shortfalls for unlisted assets and legal feasibility were not estimated. The Le Monde op-ed behind offshore-havens-17 and the Landais interview were not read directly; they are cited via the corpus and press quotations.