Globalisation, Taxation and Inequality

Why it matters, in detail. C2's headline rise 'from about 200 per cent in the 1950s to 600 per cent today in the US, in Western Europe, in China' applies to the US a trajectory that fits only continental Western Europe. In the cited Piketty-Zucman (PZ 2014) series the US private wealth/national income ratio averaged 356% over 1950-59 (range 339-375%), so the rise to 592% in 2020 (WIR 2022) is +236 pp (x1.66), not the implied +400 pp (x3): the US increase is overstated by ~70% (claimed/actual rise = 1.69), and the authors' own 2022 text describes a US doubling from 'less than three in the late 1970s', not a tripling since the 1950s. The tripling holds for Germany (166% -> 502%, x3.03) and France (213% -> 633%, x2.97) but not for the UK (313% -> 682%, x2.18); the unweighted four-country mean rises 262% -> 602% (+340 pp, ~15% below the claimed +400 pp). For China the cited sources contain no 1950s measurement (WIR 2022 starts at 120% in 1980; PZ's illustrative world table assumes 100% in 1950), so the 200% base is unsupported, although China's rise is if anything larger than stated. Direction: the magnitude of 'capital is back' is overstated for the first-named economy, but the qualitative finding - large rises to roughly 500-700% of national income in all the named economies - holds, and nothing downstream depends on the US magnitude, which keeps the issue at the upper end of moderate rather than major.

Evidence. Essay s.1, early-view p.2, fn 3. PZ 2014 appendix Table A5 (PikettyZucman2013AppendixTables.xls): USA 365.4% (1950), 1950-59 mean 356.2% (min 339.4, max 375.1); Germany 165.7%, France 213.3%, UK 312.8%, Europe (weighted) 233.2%. WIR 2022 Chapter 3 data-F3.2/F3.3, 2020: US 592.2%, Germany 501.8%, France 632.7%, UK 681.9%, China 526.0% (China 1980: 119.8%). Rise 1950s->2020: US +236.0 pp (x1.66), Germany +336.2 (x3.03), France +419.4 (x2.97), UK +369.1 (x2.18); unweighted four-country mean 262.0% -> 602.1% (+340 pp). PZ 2014 abstract: 'from about 200-300% in 1970 to 400-600% in 2010'. PZ Table A8 (illustrative world table): world 261% (1950) -> 440% (2010); North America 356% -> 410%; Western Europe 232% -> 545%; China 100% (assumption) in 1950. Saez & Zucman (2022, AEA P&P 112, p.58): US wealth/national income 'doubled from less than three in the late 1970s to over six in 2021'. Auditor B's Fed Financial Accounts/BEA check: US 1950s mean ~376-388%. WID 2026 vintage: US 1950s mean 344%, China 115%. Adjudicator script scripts/c2_wealth_income.py -> outputs/c2_wealth_income.txt.

Why this severity. For the US the stated rise is ~1.7 times the actual (+400 vs +236 pp), which on its own would meet the 'major' threshold for a US-specific trend. But the multi-economy statement is accurate for continental Western Europe (x3) and roughly right for the size of China's rise, the four-country mean rise is only ~15% below the claim, and 'capital is back' survives in every case; the figure motivates the argument and nothing downstream depends on the US magnitude. Upper end of moderate.

Why it matters, in detail. All Dubai figures in s.2.2 transcribe the May 2022 working paper exactly, but the authors' September 2025 revision of the same paper (Zucman co-author) estimates foreign non-resident holdings of Dubai residential real estate at $68bn in 2019 (30% of value), against the essay's $146bn of 'real estate owned in Dubai by non-residents ... in 2019': a 2025/2022 ratio of 0.47 (a further $30bn belongs to foreign nationals resident in the UAE, outside the essay's non-resident concept). Country holdings fall to 0.32-0.65 of the essay's values (India $29.8bn -> $9.7bn; UK $14.7bn -> $8.0bn; Pakistan $10.6bn -> $5.8bn; Saudi Arabia $9.8bn -> $6.4bn), owner-country GDP ratios to 0.30-0.68 of the essay's values (Jordan 12.1% -> 3.6%; Afghanistan 7.8% -> 2.8%; Syria 7.4% -> 5.0%; Sudan 3.0% -> 1.4%; Eritrea 4.5% -> 2.1%), and London ($143bn) and Singapore ($76bn) now exceed Dubai in absolute terms, so every derived comparison (France $140bn, London shell companies $66bn, Pakistan's foreign assets, GDP ratios) shrinks. Because the revision is residential-only and changes data and owner classification, the fall mixes the residence correction (F6) with scope and valuation changes and cannot be fully apportioned among them. Direction: C5's main quantitative illustration of the real-estate loophole is overstated about twofold on the essay's own stated concept; its qualitative core (an exceptional 30% non-resident share; ~80% non-reporting by Norwegian owners in the 2025 version) and the policy conclusion (exchange real-estate ownership information) are unaffected.

Evidence. Essay s.2.2, early-view p.4. APZO 2022 pp.9-11 and public workbook APZO2022Data.xlsx (aggregated_info; Last-Modified 2 June 2022), reproduced exactly: foreign-owned $146.17bn of $532.6bn (27.4%); India $29.83bn; UK $14.67bn; Pakistan $10.63bn; Saudi Arabia $9.78bn; Jordan 12.11%, Syria 7.37%, Afghanistan 7.81%, Sudan 3.04%, Eritrea 4.53% of GDP. ACPZO 2025 (16 Sept 2025) abstract ('30% ($68 billion) in 2019') and Introduction pp.1-2 ('more than $68 billion is invested in the Dubai housing market by foreign non-residents ... only rivaled by London ($143 billion) and Singapore (76 billion)'); Table H2 (offshore USD m / % of GDP): India 9,692.9/0.3; UK 7,995.6/0.3; Pakistan 5,814.6/1.8; Saudi Arabia 6,367.2/0.8; Jordan 1,580.3/3.6; Syria 1,511.7/5.0; Afghanistan 519.0/2.8; Sudan 454.8/1.4; Eritrea 42.5/2.1; Appendix F (residential units only). Script scripts/c5_dubai_workbook.py -> outputs/c5_dubai_workbook.txt.

Why this severity. On the essay's own stated concept (non-residents, 2019) its main illustration of the real-estate loophole is overstated by roughly a factor of two according to the authors' later estimate, and every derived comparison (France, London, Pakistan, GDP ratios) shrinks. The qualitative point - Dubai offshore real estate is large relative to the city (30% non-resident share) and largely undeclared by Norwegian owners - survives, and the policy conclusion (exchange real-estate ownership information) is unaffected.

The authors' response. No reply referring to the essay; the Sept 2025 revision (Zucman co-author) is itself the authors' updated estimate and supersedes the May 2022 figures the essay used.

Why it matters, in detail. C6's 'lessons from history' attribute the failure of past European wealth taxes partly to self-reported rather than pre-populated returns, implying that modern information reporting is the missing fix. That fits France, Spain, Germany and Switzerland (OECD 2018: filing 'generally based on self-assessment') but not the Scandinavian taxes in the essay's own list: Denmark's was mostly third-party reported (Jakobsen et al. 2020, cited in fn 15; only cash, durables, unlisted equity, noncorporate business assets and assets abroad were self-reported), Sweden had information returns for financial wealth from 1986 and pre-populated wealth returns from 1997 until repeal in 2007 (Saez & Seim 2025), and Norway's is third-party reported and pre-filled for domestic wealth. These best-administered regimes still showed long-run elasticities of 0.77-1.15 (Denmark) or were repealed (Sweden), so the lesson that pre-population is the missing fix is weaker than presented. Direction: weakens the historical premise that carries the 'fixable flaws' feasibility argument, though not the narrower claim that information reporting improves compliance (Saez and Seim find that pre-population raised it).

Evidence. Essay s.3.2, early-view p.6 ('They were not based on pre-populated returns like the income tax, but on self-reported information'; adapted from Saez & Zucman 2022, AEA P&P 112:58-62). JJKZ (2020, QJE 135(1)) pp.332-333 and 336 (third-party reported: deposits, listed securities, mortgages, real estate registries; self-reported: cash, durables, unlisted equity, noncorporate business assets, assets abroad). Saez & Seim (2025, NBER w33419) abstract ('highly-developed third-party reporting through information returns'; pre-population from 1997; information returns for financial wealth 1986). APZO 2022 s.5.1 p.16 ('Domestic income and wealth are typically third-party reported and pre-filled'). OECD (2018) The Role and Design of Net Wealth Taxes in the OECD, 'Tax filing and payment procedures': 'Wealth tax filing is generally based on self-assessment'.

Why this severity. An overgeneralisation in the 'lessons from history' that carries the feasibility argument: the counterexamples are the best-administered wealth taxes (one studied by the author), where responses remained substantial. It does not refute that information reporting matters (Saez-Seim show pre-population raised compliance), so the argument is weakened, not overturned.

Why it matters, in detail. C6 dismisses existing property taxes as 'archaic and regressive' and 'trivial for the rich' because the share of real estate in assets 'converges to nearly 0' at the top. That holds for residential property tax (0.04% of net wealth for the top 0.01% vs 0.51% for P50-90 in the Piketty-Saez-Zucman DINA 2014 micro-file), but business property taxes, 52.5% of US property taxes, are assigned to owners of equity and business assets in the authors' own distributional national accounts, the framework behind fn 16. On that incidence total property taxes take 0.63% of net wealth from the top 0.01%, close to the 0.82% paid by P50-90 (0.54-0.59% across P90-99.99), and 7.24% of pre-tax income at the top 0.01%, the highest of any group (2.59% for P50-90); the top 1% pays 28.7% of all property taxes, and property taxes make up 18.6% of the top 0.01%'s total taxes (1980 file: 0.87% vs 1.04% of net wealth). Direction: under the authors' own conventions the premise that existing property taxes spare the rich and are regressive is undermined, an important undisclosed caveat; the case for taxing financial assets net of debts is unaffected, and the result itself rests on an assumed incidence of business property taxes.

Evidence. Essay s.3, early-view p.5. PSZ DINA micro-file usdina2014.dta (proprestax; propbustax labelled 'Business property tax (prop. to equity & business assets)'; adults ranked by net wealth hweal): business share 52.5%; all property taxes / net wealth P50-90 0.82%, P90-99 0.59%, P99-99.9 0.55%, P99.9-99.99 0.54%, top 0.01% 0.63%; / pre-tax income 2.59% (P50-90) vs 7.24% (top 0.01%); residential only 0.51% vs 0.04%; housing share of gross assets 42.1% (P50-90), 18.3% (top 1%), 5.4% (top 0.01%); property taxes = 18.6% of the top 0.01%'s total taxes; 1980 file: top 0.01% 0.87% vs P50-90 1.04% of net wealth. PSZ (2022) appendix TG2 column 'Corporate (& business property) tax'. Script scripts/c6_property_tax_dina.py (reproduces Auditor A to the decimal).

Why this severity. Undermines a secondary premise of C6 (that existing property taxes spare the rich and are regressive) under the authors' own conventions - an important undisclosed caveat - but does not affect the case for including financial assets and debts in the base, and the incidence of business property taxes is itself an assumption.

Why it matters, in detail. The feasibility half of C6 requires behavioural responses to a well-designed wealth tax to be small. The essay asserts that 'with proper tax design, elasticities can be reduced, possibly to very low levels', but the cited Garbinti et al. (2023) shows that removing reporting requirements raised evasion and made the base more elastic (with zero bunching at pure rate kinks): evidence of design-dependence, not of attainable low long-run elasticities. The best long-run estimates, which the essay does not engage, are substantial: 0.77 (moderately wealthy) and 1.15 (very wealthy) with respect to the net-of-tax return in the author's own Danish study under mostly third-party reporting (Jakobsen et al. 2020); a semi-elasticity of at least 43 in Switzerland (Brulhart et al. 2022); a 7-17% base response to a well-designed 1% UK tax (Advani & Tarrant 2021). Market valuation of unlisted business wealth and liquidity (a 2% tax absorbs 100/50/25% of a 2/4/8% return) are not addressed. Direction: the essay overstates how far avoidance can be designed away, and so the ease of implementation; but the revenue case survives at elasticities in the published range (a 7-17% base loss at a 1% rate still leaves 83-93% of the mechanical base) and the claim is hedged ('possibly'), so the premise is weakened rather than refuted.

Evidence. Essay s.3.2, early-view p.6. Garbinti, Goupille-Lebret, Munoz, Stantcheva & Zucman (2023, NBER w31333 rev0) abstract and Conclusion p.33 ('zero bunching responses to changes in the wealth tax rates alone (kinks)'; their review: Seim 2017 0.1-0.3; Jakobsen et al. 2020 'long-run elasticities of taxable wealth around 1'; Londono-Velez & Avila-Mahecha 0.3-4.4). Jakobsen, Jakobsen, Kleven & Zucman (2020, QJE 135(1), p.332): long-run elasticity w.r.t. the net-of-tax return 0.77 (moderately wealthy) and 1.15 (very wealthy). Brulhart, Gruber, Krapf & Schmidheiny (2022, AEJ:EP 14(4)): semi-elasticity >= 43 (corpus optimal-tax-behavior-19). Advani & Tarrant (2021, Fiscal Studies 42(3-4)): 7-17% base response to a well-designed UK 1% tax (optimal-tax-behavior-20). Kopczuk (2019, BPEA comment): bunching 'a severe understatement'; valuation without arm's-length transactions (optimal-tax-behavior-16, us-wealth-taxes-18). Scheuer & Slemrod (2021, JEP 35(1)) (us-wealth-taxes-21). Auditor B's arithmetic: a 2% tax absorbs 100/50/25% of a 2/4/8% return.

Why this severity. This is the key enabling premise of the feasibility argument, but it is hedged ('possibly'), and the broader conclusion that a well-designed wealth tax can raise substantial revenue survives at elasticities within the published range (e.g. a 7-17% base loss per percentage point); what is weakened is the suggestion that behavioural responses can largely be designed away.

The authors' response. Saez & Zucman (Feb 2019 note 'How would a progressive wealth tax work?'; BPEA 2019) argue that avoidance is a policy choice (comprehensive base, third-party reporting, audits, exit tax, payment in kind) and attribute large Swiss responses to bank secrecy and cantonal mobility; no point-by-point rejoinder to Kopczuk located (per corpus entries).

Why it matters, in detail. The $146bn is a stock classified by owner nationality rather than residence, dated 2020 (prices January 2018-November 2020) rather than 2019, and covering all property types, with commercial property, offices, hotels and labour camps valued at residential square-metre prices; the source itself warns that 'nationality is not always the same as residency' and that taxation 'is based on residency, not nationality'. Because the essay's comparisons are residence-based (owner countries' official foreign assets and GDP), nationality-based figures overstate offshore holdings: in the 2025 revision 31% of foreign-national residential value belongs to UAE residents (India 44%, Pakistan 27%). The 'universe of all properties (about 800,000)' is ~883,000 full or partial ownership interests and excludes some property classes. Direction: inflates C5's magnitudes, but the quantitative consequences are counted under F2 (headline and country totals) and F8 (Pakistan), so on its own this is a labelling error.

Evidence. Essay s.2.2, early-view p.4. APZO 2022 s.3.2 pp.7-8 ('we focus on classifying property owners by nationality, while noting that nationality is not always the same as residency'; 'With few exceptions (e.g., the United States), taxation is based on residency, not nationality'); s.3.3 pp.9-10 ('USD 533 billion in 2020'; fn 7: residential m2 prices used for all properties incl. commercial). ACPZO 2025 abstract ($68bn non-resident + $30bn UAE-resident foreign nationals) and Table H2 (India offshore 9,692.9 vs total 17,266.1; Pakistan 5,814.6 vs 8,012.4).

Why this severity. A labelling/concept error whose quantitative consequences are counted under F2 and F8; the source itself loosely calls the estimate 'offshore real estate'.

Why it matters, in detail. The essay presents 371, 227 and 66 as Norwegian nationals, Norwegian tax residents and owners who 'duly reported', established by matching owners to their wealth-tax returns. In the source these are property counts (owners: 259, of whom 172 tax-resident); the 66 is the Norwegian Tax Administration's aggregate count of UAE-wide properties declared by persons and businesses, and compliance was assessed by comparing aggregates (also in the 2025 revision), not by owner-level matching; 66/227 is 29%, not 'about a quarter', and the source says unreported holdings 'likely correspond to tax evasion' where the essay says evasion 'was involved'. Because the 66 covers the whole UAE and businesses, it is an upper bound on Dubai compliance, so the aggregate design is conservative: non-reporting is at least 70.9% of properties and ~76.8% of value ($15.2m declared vs $65.4m owned), and 'nearly 80%' in 2025. Direction: overstates the precision and individual-level nature of the evidence, not the 'about three-quarters' evasion magnitude in C5.

Evidence. Essay s.2.2, early-view p.4. APZO 2022 s.5.1 p.16 (259 owners; 172 tax residents); s.5.2 pp.17-18 and Figure 10(a) (371, 227, 66 properties; 'at least 70% of Dubai properties were unreported'); value $65.4m owned vs $15.2m declared (76.8% unreported). ACPZO 2025 pp.30-31 (aggregate statistics from the Norwegian Tax Administration; 'nearly 80%' not declared).

Why this severity. The non-reporting magnitude (>=70.9% of properties, ~77% of value; ~80% in 2025) survives and the aggregate design is conservative; the error overstates the precision and individual-level nature of the evidence, not the conclusion.

Why it matters, in detail. The essay says Pakistan's private-sector foreign assets are 'about $10 billion' (unsourced) and then that Pakistanis' Dubai holdings are 'as large as the officially reported total foreign assets of Pakistan'. State Bank of Pakistan data put total external assets at $26.9bn at end-2019 (reserves $16.4bn, non-reserve $10.5bn), so the nationality-based Dubai figure ($10.6bn) equals ~101% of non-reserve assets but 39% of total assets, and the comparison also sets a nationality-based holding against residence-based external-asset statistics. On the authors' 2025 non-resident basis ($5.8bn) it is 22% of total assets, matching the revision's own statement that Dubai real estate would raise Pakistan's recorded gross foreign assets 'by almost 25 percent'. Direction: the comparison is overstated about 2.5-fold against total external assets (about 4.6-fold on the 2025 basis); it is purely illustrative and has no effect on C5's argument.

Evidence. Essay s.2.2, early-view p.4 (no source given for the $10bn). State Bank of Pakistan, International Investment Position of Pakistan 2019 (review), Table 1: total assets $26,944.4m, reserve assets $16,442.7m (end-2019). APZO2022Data.xlsx: Pakistan $10.63bn. ACPZO 2025 p.26 and Table H2 (Pakistan offshore $5,814.6m). Script scripts/c5_dubai_workbook.py.

Why this severity. An illustrative comparison overstated about 2.5-fold relative to total external assets (about 4.6-fold on the 2025 basis); no effect on the essay's argument.

Why it matters, in detail. The 3,748x multiple is average total Dubai holdings per Sierra Leonean owner ($2.0m; 13 owners holding 68 properties worth $26.0m) relative to an implied GDP per capita of ~$534, not the value of 'the apartment or villa'; per property ($382,659) the multiple is ~717x, so the statistic as described is overstated about five-fold. It also rests on 13 owners, and the 2025 revision puts Sierra Leonean non-resident holdings at $10.9m (vs $26.0m). Direction: exaggerates an illustrative statistic; either figure shows that owners are extremely rich relative to average income at home, so C5 is unaffected.

Evidence. Essay s.2.2, early-view p.4. APZO 2022 s.4.2 p.13 ('owners from Sierra Leone on average hold real estate in Dubai worth $2 million, which is equal to 3,748 times the GDP per capita'). APZO2022Data.xlsx row Sierra Leone: 13 owners, 68 properties, $26,020,787; per owner $2,001,599; per property $382,659 (= workbook mean); implied GDP per capita $534 -> 716.5x per property. ACPZO 2025 Table H2: Sierra Leone offshore $10.9m (0.3% of GDP).

Why this severity. The illustrative statistic is overstated about five-fold, but either version illustrates that owners are extremely rich relative to average income at home.

Why it matters, in detail. Read literally, VAT plus social contributions and payroll taxes is about 38-42% of global tax revenue (OECD Global Revenue Statistics, 143 countries, revenue-weighted, 2015-22). The 'more than half' figure holds only for the cited source's broad 'indirect taxes' category (domestic consumption taxes, excises and tariffs) plus payroll taxes: 52-56% in the OECD data (all goods-and-services taxes) and 53.8-54.9% in the Bachas et al. (BFJZ 2022) public file (2015-18). The essay inherits the source's loose 'VAT and tariffs' label. Direction: overstates the weight of VAT specifically, but C1's argument concerns taxes on consumption and labour that exempt saving and capital, for which the broad category is the relevant one and does exceed half, so the finding stands.

Evidence. Essay s.1, early-view p.1, fn 1. BFJZ (2022, NBER w29819 rev0) p.14 ('indirect taxes (VAT and tariffs)') vs Figure 1 note ('indirect taxes (including domestic consumption taxes, excises, and tariffs)'). OECD Global Revenue Statistics (143 countries, general government, USD, revenue-weighted), 2015-2022: VAT+SSC+payroll 38.1-42.4% (2019 unweighted country mean 39.4%); general consumption taxes+SSC+payroll 42.7-45.6%; all goods-and-services taxes+SSC+payroll 52.4-56.2% (India, Russia and Saudi Arabia not covered). BFJZ public CSV (NDP-weighted): SSC+indirect 54.7/54.9/54.9/53.8% (2015-2018); even counting every non-US indirect tax as VAT, 50.6-51.0% in 2015-17 (49.6% in 2018, a year depressed by a zero entry for Japan's social contributions in the public file). Script scripts/c1_revenue_composition.py.

Why this severity. A labelling problem inherited from the cited source's loose wording; the argument concerns taxes that exempt saving and capital, for which the broad category is the relevant one and does exceed half of revenue.

Why it matters, in detail. The framing that VAT- and payroll-heavy systems are 'largely creations of the 1950s', designed for post-war conditions, misdates them for most countries: VAT originated in France in the 1950s, but in the BFJZ data only 1.1% of countries had one in 1960 and 11.7% in 1980, against 62.2% in 2000 (81.7% in 2018), and 69 of 147 adopters introduced it in the 1990s (40 more from 2000 on), when wealth-income ratios and inequality were already rising. The social-contribution share of global revenue grew mainly between 1965 (19.2%) and 1980 (26.1%). Direction: the historical narrative overstates how far these systems reflect 1950s conditions, but C1's argument that they are ill-suited to today's high-wealth, high-inequality economies does not depend on adoption dates, so the effect is on framing only.

Evidence. Abstract; s.1, early-view p.1. BFJZ repository vat.dta (180 countries): share with a VAT 0.6% (1955), 1.1% (1960), 5.6% (1970), 11.7% (1980), 23.3% (1990), 62.2% (2000), 81.7% (2018); first adoption by decade: 1950s 1, 1960s 8, 1970s 11, 1980s 18, 1990s 69, 2000s 34, 2010s 6. BFJZ CSV: social contributions 19.2% (1965) -> 26.1% (1980) of global revenue. Script scripts/c1_vat_adoption.py.

Why this severity. Framing only: the instruments' post-war European origin is accurate and the argument about their unsuitability today does not depend on adoption dates.

Why it matters, in detail. C4 presents Pillar 2 as 'the first time that countries have agreed on a minimum tax rate', with 0% corporate rates previously compatible with 'all international economic agreements'. Counterexamples: the eight-member West African UEMOA's binding 25-30% corporate-tax rate band (Directive 08/2008/CM/UEMOA, adopted against 'dumping fiscal') and the EU's 15% minimum standard VAT rate (Art. 97, Directive 2006/112/EC). Pillar 2 remains the first near-global minimum on corporate profits, so the error is one of absolute wording; C4's substantive point, that tax competition is a policy choice that coordination can change, is unaffected.

Evidence. Essay s.1, early-view p.3. Directive n. 08/2008/CM/UEMOA (Bamako, 26 Sept 2008) 'portant harmonisation des taux de l'impot assis sur les benefices des personnes morales', Art. 2: 'les Etats membres fixent un taux compris entre 25 % et 30%'; preamble: 'Soucieux de lutter contre le dumping fiscal au sein de l'Union, entretenu par un recours excessif a la baisse des taux de l'impot sur les benefices' (scanned text read visually). European Commission VAT-rates page: standard rate 'must be no less than 15%' (Art. 97, Council Directive 2006/112/EC).

Why this severity. The absolute wording is wrong, but the substantive point (the first near-global minimum on corporate profits) stands.

Why it matters, in detail. The essay describes automatic exchange of bank information as 'in place globally since 2017/18' and lists its limits (developing-country non-participation, unverifiable reporting, under-use of the data) without noting that the United States, the largest financial market, is the one jurisdiction asked to commit to the CRS that has not (OECD 2018 AEOI report, para. 8). The US exchanges under Model 1 FATCA agreements with an acknowledged reciprocity gap, and the CRS literature, including the essay's own fn 10 source (Bomare & Le Guern Herry 2022, citing Casi et al. 2020), notes relocation of unreported assets to the US. Direction: the omission understates the gaps in current cooperation (C4) and in AEOI coverage (C5), which if anything strengthens the essay's case for deeper cooperation; it does not affect any conclusion.

Evidence. Essay s.1 p.2 and s.2.1 p.3. OECD Global Forum, AEOI implementation report (2018) para. 8: 'All jurisdictions asked to commit to the Global Forum's AEOI Standard have now done so, except the United States. As of 2015, the United States exchanges certain information automatically pursuant to its various Model 1 FATCA intergovernmental agreements, which includes recognition by the government of the United States of the need to achieve full reciprocity.' Bomare & Le Guern Herry (2022, Sciences Po DP 2022-10) cite 'relocation of unreported assets to the US (Casi et al., 2020)'.

Why this severity. An omission that, if anything, strengthens the essay's case for more cooperation; it does not affect any conclusion.

Why it matters, in detail. The claim that 'the countries that used to have wealth taxes did not tax their expatriates', which frames trailing taxation of emigrants as an untried remedy for tax competition, is contradicted by Germany's 1972 Aussensteuergesetz (ss.2-3: ten years of extended limited wealth-tax liability, beyond ordinary domestic assets, for emigrants to low-tax territories who kept substantial German economic interests) and by Norway's rule keeping residents of ten years or more tax-resident, and so liable to worldwide wealth tax, until the end of the third income year after emigration. Both are narrower than the essay's proposal (Germany: low-tax destinations and German-source-type assets only; Norway: three years), so the broader point that past regimes did little against emigration stands. Direction: an overgeneralisation in C6's historical argument that makes the proposal look less tried than it is; no effect on the proposal itself.

Evidence. Essay s.3.2, early-view p.6. Aussensteuergesetz of 8 Sept 1972 (BGBl. I 1972, p.1713), s.2(1) (extended limited income-tax liability 'bis zum Ablauf von zehn Jahren') and s.3(1)-(2) ('Vermoegensteuer': limited wealth-tax liability extended beyond Inlandsvermoegen; DM 60,000 allowance). Norwegian skatteloven s.2-1(3)(b) (current text from lovdata.no; date of introduction not verified).

Why this severity. Factual overgeneralisation in the historical argument; the broader point that past regimes did little against emigration stands (the German rule covered German-source-type assets and low-tax destinations only; Norway's lasts three years).

Why it matters, in detail. The rise in private wealth/national income that motivates C2 and the case for taxing capital in C6 is, in Western Europe, mostly housing: gross housing accounts for 54-83% of the 1970-2020 rise (Germany 64.4%, Spain 82.9%, France 79.0%, UK 65.7%, Italy 53.6%; 41-69% net of the rise in household debt), but only 29.6% in the US and 23.8% in Japan (WID 2026 vintage). Housing wealth is broadly held, already the base of property taxes and mostly below the proposed exemption thresholds, and the measured ratio excludes pension and social-security entitlements. Direction: no number in the essay changes, but in Europe much of the 'return of capital' is broadly held housing that sits mostly below the thresholds of the proposed tax, so the trend is a narrower argument for a high-threshold wealth tax than presented; an interpretive caveat only, since market wealth is a legitimate tax base and in the US most of the rise is non-housing.

Evidence. WID 2026 vintage (wpwhou vs wpweal), 1970-2020: gross housing = 64.4% (DE), 82.9% (ES), 79.0% (FR), 65.7% (GB), 53.6% (IT) of the rise, but 29.6% (US) and 23.8% (JP); net of the rise in household debt 41-69% in the five European countries. PZ 2014 Table 7 (Auditor B): housing = 71% (US), 96% (FR), 107% (UK) of the 1970-2010 rise in domestic capital/national income. Script scripts/c2_housing_share.py.

Why this severity. Interpretive context only; market wealth is a legitimate tax base and in the US most of the rise is non-housing.

The authors' response. Piketty & Zucman (2014, s.IV.C, Table II) acknowledge housing's role and argue that excluding housing would be inappropriate; Piketty (2015, AER P&P) accepts Weil's point on social security (per corpus).

Why it matters, in detail. (i) The statement that the 2019 Sanders and Warren proposals had thresholds of $32 million and $50 million 'and the rates increased to up to 8 per cent for net wealth above $10 billion' reads as describing both plans, but the 8% bracket belongs to Sanders (1% from $32m rising to 8% above $10bn); Warren's plan was 2% above $50m plus 1% above $1bn, i.e. a 3% top rate. (ii) The capital-share rise, the cross-country inequality ranking and Pakistan's '$10 billion' of private foreign assets are unsourced; the first two are broadly consistent with the cited WID/BFJZ data, and the third is discussed under F8. Direction: documentation and precision only; no effect on any core finding.

Evidence. Essay s.3.1, early-view p.5 (fn 11) and s.1 p.2. Saez & Zucman letters: Sanders (2019) 1% from $32m rising to 8% above $10bn; Warren (18 Jan 2019) 2% above $50m plus 1% above $1bn.

Why this severity. No effect on conclusions.

Why it matters, in detail. The claim that offshore real estate is 'extremely concentrated at the top of the distribution' is inferred from the Sierra Leone per-owner multiple, which measures owners' wealth relative to average income at home, not how holdings are distributed. The cited 2022 paper says only that offshore real estate 'appears to be concentrated at the top' (ownership probabilities rising with wealth among Norwegians); the 2025 revision, the only value-based distributional evidence, finds it 'concentrated among the upper end of the wealth distribution, but less so than offshore financial wealth': among Norwegian owners the top 0.1% hold 4% of Dubai real-estate value and the rest of the top 1% hold 30%, versus top-0.1% shares of 98% (HSBC Switzerland) and 77% (Norwegian tax amnesty) for offshore financial wealth. Direction: overstates how far offshore real estate is a very-top phenomenon, slightly weakening the framing of Dubai properties as 'the new Swiss bank accounts', but not the case for exchanging ownership information.

Evidence. Essay s.2.2, early-view p.4. APZO 2022 s.5.2 p.18 ('appears to be concentrated at the top'), Figure 9. ACPZO 2025 Introduction pp.3-4 and pp.29-30 ('Only 4% of Dubai real estate is owned by the top 0.1%, while the rest of the top 1% owns 30%. This is in contrast to ... HSBC bank accounts and the Norwegian tax amnesty, where the top 0.1% owned 98% and 77% respectively').

Why this severity. A rhetorical overstatement about who holds offshore real estate; it slightly weakens the framing of Dubai property as 'the new Swiss bank accounts' of the very top, but not the case for exchanging ownership information.

The authors' response. The authors' own 2025 revision supplies the qualification ('but less so than offshore financial wealth').

Why it matters, in detail. The proposal to tax all net wealth at market value above a high threshold with no ceiling relief, and to keep taxing former residents for '10 or 15 years after you left', together with the conclusion that 'the forces of tax competition can be tamed by unilateral action', omits three legal constraints: treaties following the OECD Model make 'all other elements of capital' of a resident taxable only in the residence state (Art. 22(4)), which after emigration is the new country; EU free-movement case law restricts exit taxation (CJEU C-9/02 de Lasteyrie du Saillant, 2004); and French constitutional ability-to-pay case law has been read as requiring a ceiling on high-rate wealth taxes. Direction: the unilateral route is less readily available than presented and would require treaty renegotiation or design changes; this is a feasibility caveat for a sketched proposal, which the essay itself ranks below international agreements, and treaties can be renegotiated.

Evidence. Essay s.3.2, early-view p.6. OECD Model Tax Convention on Income and on Capital (2010) Art. 22(4): 'All other elements of capital of a resident of a Contracting State shall be taxable only in that State.' CJEU Case C-9/02 de Lasteyrie du Saillant (press release 13/04, 11 March 2004): French exit tax on unrealised gains infringes freedom of establishment. Senate reading of Conseil constitutionnel case law (corpus offshore-havens-21).

Why this severity. A feasibility caveat for a sketched proposal; the essay itself says international agreements are preferable, and treaties can be renegotiated.

The authors' response. On the related 2025 French proposal, Zucman argued that a minimum wealth tax would bring French law into line with constitutional equality before taxation (as reported in corpus offshore-havens-21); no reply on treaty constraints located.

Why it matters, in detail. Fn 16's claim that very wealthy individuals 'have seen their taxes also fall the most' holds within the Piketty-Saez-Zucman distributional accounts (top 0.01% total tax rate 63.1% in 1960, 45.6% in 1980 and 35.1% in 2019, against 22.1%, 25.1% and 26.0% for the bottom 50%), but the long-run fall at the top depends on assigning the corporate tax to shareholders and on treating refundable credits as transfers rather than negative taxes. CBO-based estimates instead show federal progressivity rising since 1979 (Kakwani index +46%; Splinter 2020), mainly through refundable credits. Direction: alternative conventions shrink the measured decline at the top rather than reversing the essay's argument, and the claim is a closing rhetorical point, so the effect on C6 is small; the authors defend their conventions (Saez & Zucman, NBER w31912: top 1% rate ~50% in the early 1950s vs 32% in 2021).

Evidence. Essay s.3.2, early-view p.6, fn 16 (Saez & Zucman 2019a). PSZ (2022) appendix TG2/TG2c (reproduced): top 0.01% 63.1% (1960), 45.6% (1980), 35.1% (2019); bottom 50% 22.1%, 25.1%, 26.0%. Splinter (2020, NTJ 73(4):1005-1024): CBO-based Kakwani index of federal progressivity +46% since 1979.

Why this severity. The direction at the very top holds under the authors' conventions and the claim is a closing rhetorical point; alternative conventions shrink the decline rather than reverse the essay's argument.

The authors' response. Saez & Zucman (NBER w31912, 2023, rev. 2026) defend current-incidence conventions and find the top 1% effective rate fell from ~50% in the early 1950s to 32% in 2021 (per corpus).