← All papers
Vicente Silva, Quentin Parrinello, Lamia Oualalou, Carmen Durrer de La Sota, Carlos Salgado, Juan Vergara, Gabriel Zucman · International Tax Observatory report commissioned by Brazil under the PTLAC Presidency, April 2026
Progressive Taxation and Inequality: Minimum Taxation of High-Net-Worth Individuals in Latin America and the Caribbean
19 problems found: 3 major · 7 moderate · 9 minor
What the paper claims
- A 2% minimum tax on the entire net wealth of individuals worth more than USD 100m in Argentina, Brazil, Chile, Colombia, Mexico, Peru and Uruguay would raise about USD 24bn a year, roughly 0.6% of GDP ('0.6 percentage points of regional GDP') and about 3% of tax revenue; a 3% rate would raise USD 36-37bn (0.9% of GDP); a 2% tax on billionaires only USD 13.6bn.
- Tax systems in the region do not correct inequality and are in general regressive: on average the bottom 50% pays about 30% of income in taxes, the middle 40% about 23%, the top 10% and top 1% about 22%.
- The ultra-wealthy pay much lower effective rates than the rest of the population: about 20% in Brazil and 11% in Chile, 'in both cases, half the population average'.
- Traditional reforms are insufficient and the minimum wealth tax is 'the most effective instrument': a 50% personal-income-tax increase barely moves top rates, whereas a 2% MWT reverses the decline in progressivity at the top.
- Wealth is extremely and increasingly concentrated: the top 10% hold up to 80% and the top 1% up to half of wealth; centimillionaires and billionaires hold between 35% and more than half of the top 1%'s wealth; LAC billionaire wealth rose sixfold from USD 117bn (2000) to USD 698bn (2026), about 10% of GDP, while bottom-50% wealth stagnated, and could approach USD 2tn by 2040.
- Latin America is the second most unequal region in the world (top 10% about 60% of national income, bottom 50% about 7%, top 1% about 25%), and administrative-data estimates of top income shares are at or above WID's.
Every problem found
1majorconfirmed · previously known · speculative assumption
The USD 24bn assumes that every very rich family pays in full and that nobody moves abroad, reorganises their holdings or gets their companies valued low. The authors' own earlier work and the studies they cite expect a fifth or more of such revenue to be lost, so on this count alone a realistic figure is nearer USD 19-20bn.
Technical detail and evidence
Why it matters, in detail. The published figure is the upper end of a static mechanical calculation. Uniform revenue losses give: 15% -> USD 20.5bn, 17% -> 20.0bn, 20% -> 19.3bn, 50% -> 12.1bn; migration alone at Jakobsen et al.'s ~2% fall in the stock of wealthy taxpayers per percentage point (~4% for a 2-point tax) gives ~23.2bn. Responses are likely larger in the USD 100m-1bn band, where most taxpayers sit close to an all-or-nothing notch. Direction: headline overstated.
Evidence. Report pp. 7-8, 25-27 (Table 6 and its note), p. 23 (Sec. 3.4.1); adjudicator scripts/a1_tables.py section C (every slope equals 1% of the Table 1 aggregate; implied c = 0.01-0.26% of W); Zucman (2024) G20 blueprint Sec. 2.3 pp. 24-25 (text checked); De Rosa & Vila (2024) IECON DT 14/24 Sec. 3.2 and Table 6 (17%/50% rows); Londono-Velez & Avila-Mahecha (2024) NBER w32134 p. 2.
Why this severity. Within the authors' own range (0-20% loss) and their source's (17-50%), the 2% headline falls by 15-50%; adding plausible valuation shortfalls on unlisted equity, a central loss of roughly 15-30% is defensible. That materially lowers the headline without eliminating substantial revenue.
The authors' response. No response to this report located. In the parallel French debate Zucman argued that a minimum tax with no exemptions leaves little room for avoidance (France 5, quoted by BFM TV, 16 Sep 2025; corpus entry offshore-havens-17), and the G20 blueprint (Sec. 2.3) shows both a 0% and a 20% loss scenario.
2majorconfirmed · newly identified here · internal inconsistency
The report says the new tax would push the super-rich's tax rate to 50% in Brazil and 34% in Chile, above what ordinary people pay. By the report's own revenue numbers, the tax does not raise enough money for that, even if every cent came from these few people. A plausible effect is a noticeable rise, to roughly 25-35% in Brazil and 16-25% in Chile, not a reversal.
Technical detail and evidence
Why it matters, in detail. Upper bounds, if all national MWT revenue were paid by the group: Brazil 33.4% (44.7% adding the omitted billionaire band) vs 50%; Chile 24.8% vs 33.7%. Using the authors' own 13.1% ratio for the band in which the implied average member falls: Brazil 33.0%, below the 42.5% average; Chile 16.0-25.4% depending on the credit rule. On the Chilean source's administrative (book-value) wealth, at most 20.7%. The two tables also appear to use different credit rules (PIT only for Brazil, all income taxes for Chile).
Evidence. Report Tables 4-5 pp. 24-25 and footnote 7; Spanish edition footnote 7 p. 27 (13.1%/6.1%); Palomo et al. (2025) Table A.1 and Sec. 5.2; Vergara et al. (2022) Tablas 1, 2 and 13 and summary pp. 4-5; adjudicator scripts/a6_tables45.py; Claude scripts/06_tables45_consistency.py; Codex scripts/audit_calculations.py and source_checks.py.
Why this severity. Eliminates the quantitative 'reversal' claim: Chile's ceiling is about 25%, roughly its 24% average, and Brazil's is about 33% with the printed revenue. The qualitative point that a wealth floor raises top effective rates, by roughly 5-14 points, survives.
The authors' response. No response located. The heuristic matches Zucman's published argument that billionaires earn about 6% a year, so a 2% floor is equivalent to a ~33% income tax (as quoted in corpus entry offshore-havens-22); the report applies it to income-ranked groups whose implied wealth falls in the band where its own footnote assumes 13.1%.
3majorlikely · previously known · speculative assumption
For four of the seven countries, nearly all the fortunes being taxed are not observed in any data. They are extrapolated from household surveys whose richest respondent had at most USD 47 million, then scaled up. For Colombia the method spreads the whole country's wealth over the households of three cities, which makes Colombia look richer at the top than Mexico or Chile; its USD 3.9 billion share of the total is probably several times too high.
Technical detail and evidence
Why it matters, in detail. Count-anchored Pareto bands (alpha 1.3-1.5 between USD 100m and 1bn) built on Forbes resident billionaires imply USD 100m-1bn wealth of about 104-156bn for Mexico (vs 207bn in Table 1), 22-71bn for Chile depending on the anchor year (vs 94bn), and 9-13bn for Colombia (vs 162bn). Revenue overstatement is therefore roughly MEX 1.0-2.6bn, CHL 0.4-1.4bn and COL 2.5-3.4bn. Claude's sensitivities: removing the post-Pareto scaling from the tail gives about 16.3bn; +/-0.05 in Mexico's alpha moves its >USD 100m base from 503bn to 319-837bn. Uruguay's three 'billionaires' (USD 6.5bn) have no Forbes counterpart. Direction: up, with large model uncertainty.
Evidence. Report Table 1 pp. 13-14 and source note; De Rosa & Vila (2024 WP) p. 4 (caution), Table 1 (IEFIC coverage), Table 2 (scaling factors, 'Assume Mex.'s'), Table 3 (Forbes counts and survey maxima), Table 6 (household totals; aggregate wealth by estimation stage), p. 16 ('implausible range'); DANE, CNPV 2018 household table (14,243,223 households); DIAN, Agregados tributarios del impuesto al patrimonio, ano gravable 2024, sheet 'Intervalos UVT' (32,397 declarants; patrimonio liquido ajustado COP 232.6tn); Forbes 2017/2019/2026 lists; adjudicator scripts/a1_tables.py, a3_forbes.py, a5_forbes_anchored_base.py; Claude scripts 05 and 07.
Why this severity. Against Forbes-anchored and administrative benchmarks, the modelled bases overstate revenue by roughly USD 3.9-7.4bn (about 16-31% of the headline), Colombia alone by about USD 2.5-3.4bn. F4 partly offsets this, but the headline's magnitude becomes materially uncertain.
4moderateconfirmed · previously known · internal inconsistency
In different places the report says the new tax would subtract income taxes only, income and wealth taxes, or even company taxes already paid. These give very different results: subtracting company taxes would roughly halve the extra tax at the top. The tables do not apply one rule consistently.
Technical detail and evidence
Why it matters, in detail. Colombia's entire 2024 wealth-tax liability across all filers was about USD 0.28bn, which bounds the wealth-tax credit for Colombian centimillionaires at about USD 0.2-0.3bn; Argentina and Uruguay add perhaps USD 0.1-0.2bn. For Brazil's top 0.001%, CIT is 11.9% of income (Palomo); a PIT+CIT credit gives 38.6% instead of 50.5% in Table 4. A CIT-credited version of the tax would score far below USD 24bn (cf. the French debate, offshore-havens-18).
Evidence. Report pp. 7, 20, 21 (Table 3), 27 (Table 6 note), Tables 4-5; G20 blueprint Sec. 2.1 p. 20; DIAN 2024 wealth-tax aggregates (all filers: tax COP 1.14tn, about USD 0.28bn); OECD Revenue Statistics LAC (recurrent taxes on net wealth: URY 0.90%, ARG 0.61% of GDP); adjudicator scripts/a1_tables.py and a6_tables45.py.
Why this severity. Crediting actual wealth taxes has a small revenue effect (about USD 0.5bn or less, roughly 1-2%). But crediting corporate tax as Table 3 states would cut the top-up for Brazil's top 0.001% by about 40%, and the reader cannot tell which policy the published numbers describe.
The authors' response. No response to this report located; in the G20 blueprint (Sec. 2.1) and in the French debate Zucman treats the exclusion of corporate tax from the credit as a deliberate design choice (corpus entry offshore-havens-18), which contradicts Table 3.
5moderateconfirmed · newly identified here · methodology
The report's '0.6% of the region's GDP' is really a simple average of seven country percentages, computed with GDP from different years between 2017 and 2022. Measured against the seven countries' combined economy, the same USD 24bn is about 0.5% of their GDP in those years and 0.4% of today's.
Technical detail and evidence
Why it matters, in detail. Unweighted mean 0.576% (0.887% at 3%) vs revenue-weighted 0.516% (0.788%) of summed data-year GDP (USD 4.68tn); 0.40% of 2025 GDP. Tax-revenue denominators implied by Table 7: Chile 20.2% of GDP (OECD total including SSC); Mexico 13.7%, Brazil 24.5% and Peru 17.0% (total minus SSC); Colombia 18.1%, Uruguay 17.3%, Argentina 21.2% (neither). The '4.8%' of tax revenue at 3% is 4.99% as an unweighted mean.
Evidence. Report pp. 7-8, 27-28 (Table 7 and footnote 8); adjudicator scripts/a1_tables.py sections F-G; World Bank NY.GDP.MKTP.CD (Codex download); OECD Revenue Statistics in LAC extract (total tax and SSC, % of GDP); De Rosa & Vila (2024) Table 6 (households).
Why this severity. The stated 0.6% overstates the like-for-like aggregate ratio by about 10% (data-year GDP) to 45% (2025 GDP) before any substantive correction; the dollar totals are unaffected.
The authors' response. The Table 7 footnote acknowledges that wealth years differ by country and that comparisons 'deben interpretarse con cautela', and the body text says 'on average'; the executive summary's 'regional GDP' wording is not qualified.
6moderateconfirmed · newly identified here · overclaiming
The report says a wealth floor is the best tool, but it compares it with only one deliberately weak option: raising existing income taxes by half. It also ignores the new minimum tax on high incomes that Brazil, the country that commissioned the report, passed in 2025.
Technical detail and evidence
Why it matters, in detail. A 10% minimum on fiscal income, which is about 45% of economic income at Brazil's top 0.001% (Palomo Table A.1), would add only about 2-3 points to their rate. That supports the report's intuition that income-based minimums miss retained earnings, but the report neither shows it nor evaluates retained-earnings or look-through alternatives that would reach the same base.
Evidence. Report pp. 7, 19-20, 24-25, 29; Receita Federal, Perguntas e respostas on Lei 15.270/2025 (Dec 2025) pp. 4-6; full-text search of the EN, ES and PT editions (no mention of the law).
Why this severity. Overstates the comparative policy conclusion, which rests on one alternative that is weak by design; it does not affect the revenue or diagnostic findings.
7moderateconfirmed · previously known · misrepresentation of source
The report says studies show the rich rarely move or dodge wealth taxes. The studies it cites actually find noticeable departures in Scandinavia, large drops in declared wealth in Switzerland, and up to a fifth of revenue lost in Colombia, from a study that does not look at migration at all.
Technical detail and evidence
Why it matters, in detail. The cited literature implies non-trivial base erosion: about a 4-6% fall in the stock of taxpayers from migration for a 2-3-point tax (Jakobsen et al.); immediate revenue losses of up to 20% in Colombia; and large reported-wealth elasticities where third-party reporting is weak (Switzerland). These are exactly the margins the score sets to zero.
Evidence. Report Sec. 3.3.3 p. 23; Spanish edition p. 25; Jakobsen et al., NBER w32153 abstract; Brulhart et al., AEJ: Economic Policy 14(4) abstract; Londono-Velez & Avila-Mahecha, NBER w32134 pp. 1-2 and full-text search.
Why this severity. Misstates the evidence used to justify setting behavioural responses to zero (F1), which bears directly on the headline, but does not by itself change any number.
8moderateconfirmed · newly identified here · documentation gap
The report does not explain how its key numbers were produced, and the English version leaves out the assumption that drives its most striking result. Several of the errors found here would have been caught if the method had been written down.
Technical detail and evidence
Why it matters, in detail. Replacing the implied credits with the G20 blueprint's 0.3%-of-wealth average would lower the headline by about 10% (to about USD 21.6bn), but that average (US, French and Dutch billionaires) is not shown to apply in LAC. Brazil's implied credits (0.12-0.16% of wealth) are consistent with Palomo's 2.0% PIT at plausible returns, whereas Mexico's (0.01-0.02%) look low.
Evidence. Report Secs. 3.5-3.6, Tables 1 and 4-7; Spanish edition footnote 7 p. 27; Portuguese edition footnote 7; English 14 Apr and 23 Apr editions (text search); adjudicator scripts/a1_tables.py section C (implied credits).
Why this severity. Prevents independent verification of the headline and let errors F3, F4 and F14 go unnoticed; not a bias in itself.
9moderatelikely · newly identified here · internal inconsistency
Because of how the wealth table was put together, Brazil's 70 billionaires (and those of Argentina and Peru) seem to have been dropped from the main estimate. Putting them back raises the figure by about a fifth, so this mistake makes the headline too low rather than too high.
Technical detail and evidence
Why it matters, in detail. Adding the three billionaire bands gives 13.0bn at 1%, 29.6bn at 2% (+22.7%) and 45.3bn at 3%; counting only Forbes residents, 2% gives about 28.0bn. The most conservative repair keeps the counts and gives each of Brazil's 1,290 and Argentina's 133 non-billionaire centimillionaires the USD 100m minimum. That adds USD 70.5bn of wealth and 1.41bn of revenue (+5.8%), but it is a floor, not an estimate.
Evidence. Report Table 1 p. 14; Table 6 pp. 26-27 and note; adjudicator scripts/a1_tables.py sections B, C and H; scripts/a2_wid_brazil_tail.py and a4_wid_brazil_bands.py (WID Brazil bands); scripts/a3_forbes.py (Forbes 2026 totals).
Why this severity. Biases the headline down by about 16-23% (USD 3.8-5.5bn), and by at least 6% under the most conservative repair. It runs against the authors' conclusion and partly offsets F1-F2, so it does not undermine the core finding.
10moderatelikely · previously known · data error
About a third of the wealth counted for Brazil's billionaires, and two thirds for Colombia's, belongs to people who live in other countries and could not be taxed as residents.
Technical detail and evidence
Why it matters, in detail. The billionaire-only 2% scenario falls from 13.55bn to about 11.3bn (-16%); the headline corrected for F4 falls from 29.6bn to about 28.0bn. Forbes' city field is only a proxy for tax residence; exit taxes and extended residency could reach recent emigrants but not long-standing non-residents.
Evidence. Forbes 2026 list, 'city' field (adjudicator scripts/a3_forbes.py; Claude scripts/05 section C); De Rosa & Vila (2024) Table 3 (16 vs 17 Mexican and 2 vs 3 Colombian Forbes billionaires).
Why this severity. Affects the reported billionaire-only scenario (about -16%) and any corrected headline (about -USD 1.6bn); the printed USD 24bn is affected only indirectly, because of F4.
11minorconfirmed · newly identified here · arithmetic or transcription
The report counts Mexico's billionaires twice when describing how much of the top 1%'s wealth the very richest hold. The correct figure is about a third, not more than half, which is still a very high concentration.
Technical detail and evidence
Why it matters, in detail. Correct values from Table 1: Mexico 933 households, 0.25% of the top 1%, 33.6% of its wealth; Chile 37.0%, Colombia 45.4%, Uruguay 24.3%, i.e. a range of 24-45%, not 35% to more than half.
Evidence. Report Table 1 p. 13; text pp. 14-15; adjudicator scripts/a1_tables.py section I.
Why this severity. Wrong descriptive statistic; the conclusion that wealth is very concentrated within the top 1% is unaffected.
12minorconfirmed · newly identified here · misrepresentation of source
The text misreads the report's own chart for Brazil and overstates how fast tax rates fall within the richest 1%. The chart itself, and the main contrast of about 20% at the very top against 42.5% on average, are correct.
Technical detail and evidence
Why it matters, in detail. Within-top-1% decline is about 30% -> 19.7% (10 points), not more than 20; the 22-point fall starts at P95-P99, below the top 1%.
Evidence. Report pp. 7, 17-19, Figure 6; Palomo et al. (2025) Figure 10, abstract and Sec. 5.2 (20.6% for dollar millionaires, about the top 0.01%; 26.2% for the top 1%).
Why this severity. Narrative misreading of a correct figure; the central contrast (19.7% at the very top vs 42.5% on average) is unaffected.
13minorconfirmed · newly identified here · arithmetic or transcription
A few numbers in the tables and text do not match each other, including Mexico's figures for a 1% tax. None of this changes the main 2% estimate.
Technical detail and evidence
Why it matters, in detail. Correcting Mexico's 1% cells raises the regional 1% totals by USD 959m and 310m (10.46bn -> 11.42bn; 6.24bn -> 6.55bn).
Evidence. Report pp. 8, 25, Tables 6-7; adjudicator scripts/a1_tables.py sections C-E and G.
Why this severity. Affects the 1% scenario and rounding only; the 2% headline is unaffected.
14minorconfirmed · newly identified here · misrepresentation of source
Some sentences describing the charts overstate them slightly; for example, Latin America ranks second, not first, on the richest 1%'s share of income. The charts themselves match their sources.
Technical detail and evidence
Why it matters, in detail. WID 2023-24 (MER): LAC top 10% 58.6%, bottom 50% 7.6%, top 1% about 25%; MENA top 10% about 64%, top 1% about 28%. LAC is second on each measure, narrowly ahead of South and South-East Asia on the top-10% share.
Evidence. Report pp. 10-12, Figures 1-3; WID market-exchange-rate regional aggregates (XL-MER, XN-MER, XS-MER); Palomo et al. (2025) pp. 3, 27-29; Vergara et al. (2022) Sec. 2.
Why this severity. Wording errors around figures that are themselves correct; 'second most unequal' holds on all three measures.
15minorconfirmed · newly identified here · documentation gap
Several references are wrong, and one cited paper, used to support the claim that the rich avoid taxes rather than emigrate, does not appear to exist.
Technical detail and evidence
Why it matters, in detail. One unverifiable reference supports the behavioural claim criticised in F8; the other errors are bibliographic.
Evidence. Report bibliography pp. 31-33; Crossref record 10.1257/aer.20241596; NBER w32153 title page; AEJ:EP article header (10.1257/pol.20200258); WIL Activity Report 2020 p. 16; OpenAlex search.
Why this severity. Documentation and credibility only; no number depends on these references.
16minorconfirmed · previously known · other
A large part of the money would come from a few families, and their fortunes swing by a third or more from one year to the next with share prices. The report gives one precise number as if it were stable.
Technical detail and evidence
Why it matters, in detail. Year-to-year swings of 25-50% in the billionaire tier translate one-for-one into the billionaire component of the yield; revenue concentration in about five families makes the estimate sensitive to individual migration or valuation outcomes (cf. offshore-havens-22).
Evidence. Adjudicator scripts/a5_forbes_anchored_base.py (volatility and concentration); Forbes lists 2016-2026; report Figure 4 (about 405bn in 2025 vs 698bn in 2026).
Why this severity. A robustness and precision caveat with no demonstrated bias in the printed headline.
17minorlikely · newly identified here · misrepresentation of source
To compute what ordinary Chileans pay, the report counts some payments to private health insurers and similar schemes as taxes, which makes their tax rate look higher. The very rich still pay much less than average, but the gap is somewhat smaller than 'half'.
Technical detail and evidence
Why it matters, in detail. Removing non-tax items would lower Chile's bottom-50% (about 25.5%) and average (24%) rates by several points, so the top 0.01%'s 11.7% would be about 55-65% of the average rather than 'half'. Health contributions to the public FONASA are legitimately tax-like for most low-income workers.
Evidence. Report footnote 6 p. 18, Figure 7, Table 5; OECD Revenue Statistics in LAC, Chile 2017 (SSC CLP 2,628bn; employer SSC CLP 73.6bn); Claude scripts/07 section 3; Codex calculations.json (health sensitivity).
Why this severity. Inflates Chile's bottom and average rates (a secondary comparison); the top 0.01% rate itself is unaffected and remains well below average.
18minorlikely · newly identified here · external inconsistency
The report says the poorer half's wealth has barely changed. The database it uses elsewhere shows it rose by about half since 2000, from a very low level. The chart of billionaire wealth uses inflation-adjusted figures without saying so, and one year looks wrong.
Technical detail and evidence
Why it matters, in detail. The 'sixfold' (5.97x), '10% of regional GDP' (about 9.4% of 2025 LAC GDP per Claude) and the conditional 7.1%-a-year extrapolation to about USD 1.8tn by 2040 are arithmetically correct for the series shown; the contrast with the bottom 50% is not.
Evidence. Report p. 7 and Figure 4 p. 15; WID API pull (lac_wealth.json: ahweal/shweal p0p50 and p99p100, XL-MER); adjudicator scripts/a3_forbes.py (LAC totals by list year); Claude scripts/02.
Why this severity. Affects a rhetorical contrast, not any number used in the revenue estimate; billionaire wealth relative to GDP did rise.
19minorplausible · previously known · methodology
The finding that the poorest half pays a larger share of its income in taxes than the richest 1% broadly matches the underlying studies. But its size depends on accounting choices, it is not true in every country shown, and it leaves out public spending and cash benefits.
Technical detail and evidence
Why it matters, in detail. Excluding social contributions, DFM's own rates (household unit, Claude's recomputation) are bottom 50% about 27.7%, top 1% about 17.2%, so the bottom-above-top pattern survives. Netting out transfer-financed consumption would trim bottom rates by perhaps a few points (the mechanism of us-wealth-taxes-12); the report's 'do not correct inequality' refers to taxes only.
Evidence. Report p. 7, Figure 5 and footnote 3 p. 16; De Rosa, Flores & Morgan (WIL WP 2022/13, 2023 update) Sec. 1.3 (CEQ incidence for consumption taxes); Palomo et al. (2025) p. 3 (consumption taxes paid out of transfers excluded); figure readings by both auditors (bars average about 29/24/22/23).
Why this severity. The broad pattern is consistent with the sources, but its size depends on conventions that are not discussed; no demonstrated reversal.
Claims examined and rejected
Problems one auditor raised that did not survive checking.
- The Brazil/Argentina/Peru inconsistency cannot be resolved by reading rows as exclusive bands (Colombia's centimillionaire mean exceeds USD 1bn; Table 6 defines minimum thresholds), so only a conditional minimum repair of +USD 1.41bn (+5.8%) is identified. — Partly rejected. Table 1 uses cumulative rows for the four De Rosa-Vila countries and band rows for Brazil, Argentina and Peru. Under that mixed reading every BRA/ARG/PER row is coherent, Brazil's band averages match WID's, and the billionaire rows equal Forbes 2026 totals. The +5.8% figure assumes that 1,290 Brazilian centimillionaires each own exactly USD 100m, which no distribution supports, so it is a floor rather than an estimate. The inconsistency itself is retained as F4, with a +16-23% central range.
- Crediting existing wealth taxes (Colombia, Argentina, Uruguay) would lower the headline by 3.7-8.3%, with Colombia falling from USD 3.87bn to 2.0-3.05bn. — Magnitude overstated. Colombia's entire 2024 wealth-tax liability across all 32,397 filers was COP 1.14tn, about USD 0.28bn (DIAN), so credits for its centimillionaires cannot exceed about USD 0.2-0.3bn; with Argentina and Uruguay the effect is roughly 1-2% of the headline. The design-versus-simulation inconsistency is retained as F5.
- Anchoring Mexican and Chilean billionaire wealth to Forbes in the survey years cuts the headline to USD 20.4bn (-16%), showing the base is inflated. — Not a valid measure of bias for a tax presented as a current annual yield. Against the 2026 Forbes list, the Table 1 billionaire totals for Mexico (USD 296bn vs 267bn) and Chile (63bn vs 63bn for residents) are close. The inflation lies in the modelled USD 100m-1bn band and in Colombia, which F2 retains.
- The 'USD 2 trillion by 2040' projection is unreliable because it depends on ending the extrapolation at the 2026 peak (about USD 0.85-1.6tn with 2025 or 2024 end points). — The report states the projection conditionally ('If this trajectory ... continues'), and the arithmetic is correct (7.1% a year gives about USD 1.8tn). End-point sensitivity is a caveat, not an error.
- The 'regressive' label does not hold for every country: Ecuador, El Salvador, Colombia and Argentina show top-1% rates at or above the top-10% rate. — The report says systems are 'in general' regressive and 'flat or slightly regressive' above the bottom half. In Figure 5 the bottom 50% pays more than the top 1% in 7 of 10 countries, and a top-1% rate above the top-10% rate does not contradict that wording. The exceptions are noted under F16 but are not a misstatement.
- Using the G20 blueprint's 0.3%-of-wealth average for taxes already paid would lower the headline by 10.4%. — Not a demonstrated bias. The 0.3% is an average for US, French and Dutch billionaires, and Brazil's implied credits (0.12-0.16% of wealth) are consistent with Palomo et al.'s 2.0% PIT at plausible returns. Retained only as an undocumented assumption under F9, noting that Mexico's implied credits look low.
Audit notes
What reproduced. Reproduced: (i) Table 1 internal identities (count x average = aggregate, 34/34 rows); (ii) Table 6 column totals (USD 10,457 / 24,143 / 36,869m; billionaire totals within USD 1m) and 40 of 42 cells as R(t) = t*W - c, with slopes exactly 1% of the Table 1 aggregates and implied credits of 0.01-0.26% of wealth; (iii) Table 7's implied GDP years (MEX 2019, CHL 2017, COL 2018, BRA 2022, PER 2022, URY about 2019, ARG about 2021), unweighted means of 0.576% of GDP and 3.25% of tax revenue, against aggregates of 0.516% and 2.59%; (iv) Brazil's 42.5%, 26.2%, 20.6% and 19.7% (PIT 2.0%, CIT 11.9%) from Palomo et al., and Chile's 11.7% from Vergara et al.; (v) the Tables 4-5 values of 50.5% and 33.5-33.7% via ETR + 2%/6.1% - credit, using the return assumption disclosed only in the Spanish and Portuguese editions; (vi) the Forbes 2026 totals behind the Brazil, Argentina and Colombia billionaire rows; De Rosa & Vila's household totals, scaling factors and Forbes counts; Figure 1 against WID regional aggregates. Not reproduced: the WIL+Forbes construction for Brazil, Argentina and Peru; the 'income taxes already paid' credits; the Figure 5 merge of social contributions; any internally consistent version of Tables 4-5. Independent plausibility check (not a replacement estimate): Forbes 2026 resident billionaires (97, USD 547bn) plus a count-anchored Pareto band for USD 100m-1bn (alpha 1.2-1.5) give a mechanical 2% revenue of USD 15-22bn (about 23.8bn on a citizenship basis with alpha 1.4). That compares with 24.1bn printed and 29.6bn if the omitted billionaire rows are restored. After a 15-20% avoidance allowance, the range is roughly USD 12-20bn, about 0.2-0.33% of the seven countries' 2025 GDP (USD 6.03tn).
Limits of this audit. No author materials (code, spreadsheets, technical annex) are public, so the Brazil/Argentina/Peru reading and the Tables 4-5 reverse-engineering are inferences from published tables and cited sources; authors were not contacted and no published response to this report was found. The published De Rosa & Vila (2026) article was not accessible; the 2024 working paper, whose household totals and survey years match Table 1, was used, and the final version may differ. Forbes' 'city' field is used as a proxy for tax residence. The Forbes-anchored Pareto benchmark depends on the tail parameter (alpha 1.2-1.6) and on Forbes coverage, which may miss Latin American fortunes held through family groups or offshore; it is a plausibility check, not a replacement estimate. WID Brazil comparisons use the October 2026 bulk vintage, which may differ from the authors'. DIAN figures are at fiscal values, which understate market values. Figure readings are approximate (about +/-0.5 points). No behavioural elasticity specific to a Latin American minimum wealth tax exists, so loss scenarios are illustrative. Historical claims (dates of past LAC wealth taxes), the Latinobarometro figure and the Friedman et al. (2024) UK evidence were not independently audited.