Top Wealth in America: A Reexamination

Why it matters, in detail. Gap 2016 = 2.16pp. Established: Forbes replacement +0.90pp and the 3% boutique cap +0.23pp (52%). Claimed but uncomputed: interest +0.66pp and a pass-through residual of +0.60pp. SZ22's own pass-through factors deliver +0.19/+0.39/+0.66pp for a quarter/half/all of partnership wealth tiered, before FA rescaling. Corrected SZZ is 8.44%/8.72%/9.33% under SZZ-favourable/central/SZ-favourable settings, leaving 0.55-0.83pp (26-39% of the gap) except at the SZ-favourable corner. Against SZ's Feb-2022 series (9.99%) the residual is 0.65-1.55pp. Trends are not recomputed at all.

Evidence. 2016 gap = 9.275 - 7.116 = 2.16pp (USDINA Sept-2020 TE1; SZZ TotalWealthShare.xlsx). Established by published computations: SZZ's Forbes-400 replacement +0.900pp (SZZ Table III row 11; 42% of the gap) and SZZ's own 3% boutique-rate cap +0.232pp (row 15; 11%), together 1.13pp (52%). SZ22's interest closure of 0.66pp (30%) is not computed (F2). The residual SZ22 attribute to pass-throughs is 0.60pp. At SZ22's own factors (S-corp ceiling x1.236 from SOI; tiered-partnership x2.14) applied to SZZ's P99.99-99.999 pass-through wealth (24.5% x 3.9% = 0.96% of total wealth, SZZ Tables I-II), the effect is +0.19pp (a quarter of partnership wealth tiered, half of S-corp wealth in the $5-50M band), +0.39pp (half tiered, all S-corp wealth in band), +0.66pp (all tiered), or +1.09pp if all group pass-through wealth were tiered partnerships. These figures are before FA rescaling, which lowers them. Corrected SZZ top-0.01% share: 8.44% (SZZ-favourable: 3% cap, low pass-through), 8.72% (central: 2.5% scaled from row 15, central pass-through) or 9.33% (every correction at SZ's end), against 9.27% for SZ20. The residual is 0.83pp / 0.55pp / -0.06pp; against SZ's then-current Feb-2022 series (9.99%) it is 1.55 / 1.26 / 0.65pp. SZZ's Forbes row also lowers P99-99.9 by 0.2pp and P99.9-99.99 by 0.1pp. Adjudicator script: scripts/gap_accounting.py.

Why this severity. Reconciliation is the paper's headline conclusion. Only about half of the gap (1.13 of 2.16pp) is closed by computations anyone has run. Closing the rest needs both uncertain corrections at the SZ-favourable end of their ranges; under central assumptions about a quarter of the gap (~0.55pp) remains. The direction of the critique survives.

The authors' response. SZ22 (pp. 8-9) acknowledge that 'uncertainty remains about the true wealth of the top 0.01%' and argue that FA pass-through totals are conservative and that SZZ's bottom-up estimates, once fixed, would raise top-end wealth. No response by Saez and Zucman to SZZ's 28 Sep 2022 report was located.

Why it matters, in detail. SZZ's FA-normalized, re-ranked implementation of a 3% boutique cap raises top-0.01% fixed income x1.16 and the share +0.23pp. Scaled to 2.5% it gives about +0.32pp, against SZ22's x1.6 and +0.66pp. The estate-based x1.6 implies a top-0.01% rate near 1.9%, below the 2.95-4.48% CMD interval for the broader top 0.1%, which SZ22 do not mention. The hedge-fund benchmark depends on household ownership of domestic funds, put at 23-35% by Form PF. The direction (SZZ's 6% extrapolation may be too high) is plausible; the increment is uncertain by roughly a factor of 3.

Evidence. SZZ Table III row 15 (boutique rate capped at 3% in 2016, close to SZ22's 2.5% proposal): top 0.01% 7.116 -> 7.348% (+0.232pp, 11% of the gap); top-0.01% fixed income x1.16 (21.6% of 7.1% -> 24.2% of 7.3%). Scaling the boutique-asset increment by 1/r gives about +0.32pp at 2.5% (15% of the gap) and +0.46pp at 2.0% (21%). SZ's Jan-2022 note claimed a 3% rate 'adds around $0.5 trillion' to the top 0.1%; row 15 implies +0.32pp, about $0.25T. CMD: the 2016 rate for the top 0.1% by non-interest wealth is 3.7%, 95% CI 2.95-4.48% (SZZ QJE pp. 535-536; exhibits DataFig3CD). Both the ~1.9% top-0.01% rate implied by fn. 6 and the estate method, and SZ20's matched-estates rate for $20M+ estates (1.44% in 2016, SZ20 DataF4), lie below that interval. SZZ say the estate rate rests on ~700 estates with $117M of interest (App. I) and that estate/SCF denominators include dividend-paying bond and money-market funds (SCF top-0.1% rate 2.3% -> 3.9% once they are removed). SEC Form PF 2016Q3: US individuals own $319B, 11.7% of qualifying hedge-fund NAV; US-domiciled funds hold 33.7% of NAV. US individuals therefore own 23% (FA denominator, $1,396B) to 35% (SEC denominator) of domestic-fund NAV, and 59-90% only if all fund-of-funds holdings are attributed to them. Fn. 5's '42.5%-90%' is undocumented; its 90% matches the latter attribution. SZ22 contain no mention of 'boutique', 'minimum distance', CMD or SZZ's row 15. On Table 3's assumptions, hedge funds' bond and loan holdings yield 2.7%, so the qualitative point that they earn well below 6% stands. Scripts: szz_2016_values.py, fn6_interest_arithmetic.py, fed_hedge_funds.py.

Why this severity. x1.6 is a headline abstract number and supplies 30% of the claimed reconciliation. Under SZZ's implementation of nearly the same change the factor is about x1.16-1.22 and the gap closure 11-15%, a third to a half of the claim. The paper's figure sits at the top of a defensible range and is presented without that range.

The authors' response. SZ22 (pp. 6-7) argue that capitalizing pass-through interest separately 'is not a convincing methodology' because legal form does not determine a security's yield, and that their matched estates-income approach 'has a rigorous conceptual basis'. Fn. 7 acknowledges a post-Great-Recession interest premium for the rich. No response to the CMD estimates or to SZZ's row 15 was located.

Why it matters, in detail. Re-capitalizing pass-through interest from ~6% to 2.5% can raise 1/r by at most s x 23.3-23.9, where s <= 1 is the pass-through share of the group's taxable interest. Moving r from 3.79% to 1.9% requires +26.2, i.e. s = 1.02-1.12. At s = 0.80-0.85 the total multiplier is x1.42-1.46, so the increment is overstated by roughly 30-45% at the stated rate.

Evidence. Required pass-through share of taxable interest: 1.12 (pass-through rate 6.0%), 1.10 (6.2%), 1.02 (7.02%). Bank interest alone is 'just over 10%' of top-0.01% interest (SZZ p. 529). Even at a share of 1, the multipliers are x1.88-1.90 (taxable assets) and x1.53-1.54 (total, using the 40% muni share the footnote implies). At shares of 0.80-0.85 they are x1.71-1.77 and x1.42-1.46. The footnote's numbers fit a 2.0% rate (share 0.77-0.79), i.e. Table 3's hedge-fund yield, not the 2.5% stated in the text. Script: fn6_interest_arithmetic.py.

Why this severity. At the rate the text states, the rise in top-0.01% interest-bearing assets is overstated by about a third to a half (x1.42-1.46 instead of x1.6). The direction is unaffected, and x1.6 would follow from the paper's own 2.0% yield, so this reads as a 2.0%/2.5% slip.

Why it matters, in detail. SOI-based tiering factors are 1.93-2.09 (2.14 at 60%), and the S-corp factor is at most x1.236. Applied to SZZ's group pass-through wealth (0.96% of total), they deliver about +0.2-0.4pp under moderate tiering assumptions (maximum about 1.1pp), before FA rescaling. SZZ's unscaled bottom-up series moves the share -0.22pp.

Evidence. SOI Partnership Returns TY2016, Table 1: ordinary business income $354.7B, interest paid $96.8B, depreciation $187.5B, so I+D is 44.5% of EBITDA. Adding Form 8825 rental real estate from Table 4 (net $57.0B, interest $119.0B, depreciation $131.7B) gives 56.5%. The implied tiering factors are 1.93 / 2.09, and 2.14 at 60%. SOI Form 1120S Table 2.4: S-corp I+D is 23.8% of EBITDA, so x1.236 is the maximum. 'Ordinary income from other partnerships and fiduciaries' is $146.6B, 22% of positive ordinary business income, so tiering is quantitatively relevant. SZ20 (NBER w27921, pp. 37-38, Fig. 20) compare capital shares by asset bin: 'The gap is small for firms with a few million in assets'. SZZ (QJE p. 542) motivate the $50M hybrid by citing Smith et al. (2022). SYZZ (NBER w25442, p. 26) report an 87.1% dollar-weighted owner-death effect for firms with less than $50M of profits, which earn 72% of top-1% pass-through profits. SZZ Table III row 5 (unscaled pass-through): top 0.01% 6.89% vs 7.12%; per-capita wealth $222.2M vs $227.7M. Auditor B's sensitivity: with sales and EBITDA each 10% of value the tiering factor is 1.19; uniform corrections cancel under FA rescaling. SZZ's report (sec. 3) adds that many tiered owners are institutions, that C-corp owners raise double-counting issues, and that financial flows are capitalized elsewhere. SZ22's claim that trust-held businesses get zero value was not verified either way. Script: soi_partnership_check.py.

Why this severity. C5 is hedged in the abstract ('suggest', 'up to'), and the mechanism is credible. The defects make the factors ceilings and leave the top-share effect unmeasured. That matters mainly for C6, which is counted under F1.

The authors' response. SZ22 present these as mechanisms, propose linking partnerships as in Cooper et al. (2016) (p. 7), and argue that SZ (2016, 2020), which capitalize business income, do not suffer from these issues (p. 8).

Why it matters, in detail. The 2016 P90-99 level gap (2.2pp) equals the top-0.01% gap. The 1978-2016 rise in the top-10% share is +8.7 vs +4.5pp. SZ22's corrections act only on the top 0.01% (+1.1 to +2.2pp in 2016), so the top-10% trend would remain about 30% larger in SZ20. The claim does hold within P99-99.99 (+6.1 vs +7.0pp).

Evidence. SZZ Table I (2016): top 10% 68.6% vs 73.4%; P90-99 34.9% vs 37.1%; bottom 90% 31.4% vs 26.6% (average wealth $112k vs $95k). By contrast P99-99.9 is 18.0 vs 17.8 and P99.9-99.99 8.6 vs 9.1. Changes 1978-2016: top 10% +4.5pp (SZZ) vs +8.7pp (SZ20); bottom 90% -4.5pp vs -8.7pp; P90-99.99 -0.4pp vs +1.8pp; P90-99 -6.5pp vs -5.2pp; P99-99.99 +6.1pp vs +7.0pp. Script: gap_accounting.py.

Why this severity. The top-0.01% analysis is unaffected. But the framing claim, and 'reconciled', overstate agreement on a standard concentration measure: SZ20's rise in the top-10% share since 1978 is about twice SZZ's, and would still be about 2pp larger after SZ22's top-0.01% corrections.

Why it matters, in detail. Relative to SZ's then-current series the gap is 2.87pp and the post-Forbes top-0.001% residual is 0.85pp. The +0.88pp revision to the top 0.001% (+0.71pp to the top 0.01%) came from SZ's own top-end procedure, not from the newer FA vintage, which lowers concentration. So the benchmark's very top is uncertain by roughly the size of the gaps being 'reconciled'.

Evidence. USDINA TE1 for 2016, Sept-2020 vs Feb-2022: top 0.01% 9.275 vs 9.988; top 0.001% 4.072 vs 4.954; top 0.1% 18.42 vs 18.53; top 1% 36.27 vs 35.52. PSZ 'Log of updates' (Feb 2022, sec. 1.1) describes the Forbes-matching change and the split of a top record with weight above 400. The Feb-2022 FA vintage has about $1T fewer bonds and deposits, which 'slightly reduces' concentration, so the increase does not come from the aggregates. Wayback capture of gabriel-zucman.eu/usdina/ (16 Jun 2022) lists 'Current annual series ... (February 2022)' and 'Archived annual series: September 2020 vintage'. Against Feb-2022 the top-0.01% gap is 2.87pp (not 2.16), the top-0.001% gap is 1.75pp, and SZZ's Forbes replacement closes 51% of the latter rather than 'all the gap'. The +0.88pp revision to SZ's top 0.001% equals the whole top-0.001% gap SZ22 attribute to SZZ. Counterpoint (Auditor B): SZZ matched the Sept-2020 aggregates component by component, so that vintage is the like-for-like comparator. Script: usdina_vintages.py.

Why this severity. Comparing like with like is defensible, so this is not an error of comparison. It is an undisclosed caveat on the headline gap (2.2 vs 2.9pp) and on 'closes all the gap' at the top 0.001%. It also shows the benchmark moves by about a third of the gap under changes to top-end calibration.

Why it matters, in detail. The Forbes route (+0.90pp) depends on Forbes's non-listed valuations (~64% of Forbes-400 wealth) and on the one-entry-one-unit assumption. Estate filings value single decedents' gross estates at 39% of Forbes, which SZ attribute to undervaluation for estate tax. The truth is unidentified, so up to about 0.4pp of the 0.9pp is assumption-dependent.

Evidence. SZZ (2023) pp. 551-552, fn. 39-42: the top 0.001% (2,400 adults) holds $2.5T vs $2.4T for the Forbes 400. Balkir, Saez, Yagan & Zucman (2025, NBER w34170, pp. 3, 10): gross estates of single Forbes decedents are 39% of Forbes estimates, and estate values about 30%. Illustrative adjudicator sensitivity: if Forbes overstated non-listed wealth by 25%, the Forbes-400 total would be about $2.0T and replacement would add about $0.35T instead of $0.73T, roughly halving the +0.9pp. The private C-corp values in Table 2 exceed Forbes net worth (F15).

Why this severity. This governs the largest established piece (~40%) of the reconciliation. A plausible overstatement of private valuations would cut that piece materially, but the SEC-validated listed portion (~0.45-0.6pp) is robust.

The authors' response. SZ22 (p. 4): Forbes estimates, 'although imperfect, ... are more accurate than those implied by capitalizing reported individual income'. Balkir, Saez, Yagan & Zucman (2025, p. 10) attribute the estate-Forbes gap to 'techniques available to high-net-worth individuals to undervalue assets in the context of the estate tax'.

Why it matters, in detail. The 1981->1982 jump in SZ20 (+0.58pp) is about ten times SZZ's unbenchmarked change and matches SZZ's own Forbes increment for 1982 (+0.38pp). About 6-7% of SZ20's 1978-2016 rise therefore reflects the start of benchmarking, not the economy. Pre-TRA86 dividend avoidance could bias early capitalized levels down, and SZZ's unscaled bottom-up series gives -0.22pp. None of this is weighed against the upward corrections.

Evidence. SZ20 (Sept-2020 TE1): top 0.01% 2.93% (1981) -> 3.51% (1982), +0.58pp; top 0.001% 0.98% -> 1.42% (+44%). Over the same year SZZ's unbenchmarked baseline moves 2.90 -> 2.96% (+0.06pp) and SZZ's equal-returns series +0.12pp, while SZZ's Forbes replacement adds +0.38pp in 1982 (Spec 11). PSZ log of updates: 'We make no correction before 1982'. SZZ Table III row 5: 6.89% vs 7.12%. Kopczuk (2019, BPEA) criticises SZ's 'optimistic empirical assumptions' with uncertainty 'likely to run mostly in one direction'. Scripts: usdina_vintages.py, gap_accounting.py.

Why this severity. The 2016 diagnosis is not overturned. But the one-sided 'conservative' conclusion is undercut, and so is the trend comparison: SZ20's 1978-2016 top-0.01% rise of 6.9pp contains a 0.4-0.5pp method break.

The authors' response. SZ22 (p. 8) argue that FA pass-through aggregates are conservative and that bottom-up estimates, once fixed, would raise top-end wealth. The PSZ log of updates notes that the Forbes 400 held under 1% of aggregate wealth in 1982.

Why it matters, in detail. The 6% is the rate SZZ apply to top-AGI owners, not the funds' asset-weighted average (about 2-5%). Implied fund assets are about $120-200B rather than $100B, and the household share of the $6B is a guess applied to a selected subset. The 2%-vs-6% contrast that drives capitalization is unaffected.

Evidence. SZ's own Jan-2022 note (App. B) puts the sample's average rate at 'about 3.25%' and says 6% arises when funds are matched to top individuals. SZZ's July 2022 reply and App. I give asset-weighted rates of 3-5% for funds with fewer than 10 partners ($3.5B of the $6B) and about 2% for large funds, with 6% near the 75th percentile. At 3-5%, the subset holds roughly $120-200B, so domestic hedge funds' $1,032B is about 5-8x rather than 10x. The currently posted SZZ appendix (Sept-2024 file, App. I) says the full sample of 18,758 funds distributes $31B (about 20% of partnership interest), and that the $6B is the subset paying individuals. The Aug-2022 Wayback copy is truncated, so whether this text was available to SZ22 could not be confirmed. Households' K-1 interest is $21B + $7B + $3B = $31B, so the $6B subset can be up to about 19% of it.

Why this severity. This affects the rhetoric about sample size and coverage. The capitalization comparison (2% vs the 6% SZZ apply) is untouched, and the extrapolation point remains qualitatively valid.

The authors' response. SZ22 fn. 3 labels the $1-2B 'Our $1-2 billion guess' and calls it 'a key statistic we asked Smith et al. (2022) to report'.

Why it matters, in detail. The benchmark yield ranges from about 1.96% to 2.79% across defensible assumption sets. The change between drafts favours SZ's argument but is arguably more accurate (G10 sovereigns yielded about 0.3% in 2016).

Evidence. Jan-2022 note fn. 9: cash 1%, Treasuries 1.8%, G10 sovereigns 2.5%, MBS 3.5%, corporates 4.5%, other 5.0%, giving 2.5%. Aug-2022 Table 3: 0.1-0.5%, 1.2%, 0.3%, 2.2%, 3.2%, 6.1% and 12.7%. On the same balance sheet the yield is 1.96% with the August yields and 2.79% with the January yields (Auditor A; loans kept at the August yields, about 2.6% if loans are also priced at 5%). It is 2.14% excluding MMF shares and FX, and 2.67% on debt securities and loans only. Script: fed_hedge_funds.py.

Why this severity. Every variant (2.0-2.8%) is far below SZZ's 6.2-7.0% boutique rates, so the qualitative contrast survives. Only the exact counterfactual rate is affected.

Why it matters, in detail. Zuckerberg -$6.1B and Adelson -$8.8B; Walton +$9.5B on true value and about +$9.1B on the SZZ value, since Walmart pays dividends. Net: factor 2.13 -> ~2.07, gap -$14B.

Evidence. Zuckerberg: $53.4B = 463.8M shares x $115.05 = his own 410.4M shares (Form 4, 21 Dec 2016) + 4.45M Chan Zuckerberg Foundation shares + 48.9M Moskovitz shares 'subject to voting proxy' (FB 2017 proxy, fn. 3 and 15). His own stake is $47.2B ($47.7B including the foundation). Adelson: rows of $19.3B + $12.6B = $31.9B, about 597M LVS shares, vs the whole group's 431,987,723 shares (54.5%; LVS 2017 proxy fn. 3), worth $23.1B at $53.41. Overlapping beneficial ownership (e.g. trust shares listed under both Miriam Adelson and Irwin Chafetz) is the likely cause. Walton: rows total $97.4B vs Walton Enterprises (1,415.9M shares) + Walton Family Holdings Trust (131.2M) = $106.9B at $69.12 (WMT 2017 proxy). Corrected totals are about $860B true and $415B SZZ, a factor of about 2.07. Script: sec_stake_checks.py.

Why this severity. Net effect: the 2.13 factor becomes about 2.07 and the $459B gap shrinks by about 3%. Immaterial to C2.

Why it matters, in detail. The documented minimum is 9 of 20, not 13. That weakens the breadth claim for low-dividend C-corps among large private firms, not the Table 2 quantification.

Evidence. Table A2 (printed p. 13). Even counting the two 'unknown' firms whose notes lean towards C-corp status (Pilot Flying J, Fidelity) gives 11. Of the 9, Albertsons is owned by a private-equity fund and Publix pays out one-third of net earnings.

Why this severity. A supporting statistic only; the three firms in Table 2 are individually documented.

Why it matters, in detail. Each discrepancy is below $1.2B per row, and the net effect on the $459B listed-equity gap is a few billion dollars.

Evidence. Introduction: SZZ values of $1.3B (Zuckerberg) and $5.7B (Bezos) vs 1.2 and 5.3 in Tables 1/A1, implying a ~4.0 rather than 3.7 gains multiplier. Dell revenue: 54 (Table A2), '$59B' (A2 note) and 50.9 (Table 2). Cargill: 120.4 vs 109.7. The A1 note gives a 1.4% dividend yield for the 'top 120 excluding top 10', but the listed rows give 1.57%; the next-280 extrapolation also uses a 3.3% gains/value ratio vs 2.0% in the listed rows (net effect on the gap about $3B). Ballmer: 383 x 33.4 = 12.8 vs 11.6 shown; the row uses 2014 data, so a 2014 factor plausibly explains it. Script: tableA1_recompute.py.

Why this severity. Presentation slips with no effect on conclusions.

Why it matters, in detail. The 'factor of 2' for listed partnerships is an input (from the stylized Section 4 mechanism), not an output, so it cannot corroborate that mechanism. It covers $13.4-17.3B of the $459B.

Evidence. SZ22 fn. 8 (p. 7) vs the Table A1 notes (p. 12). These rows contribute $13.4B of assumed, not measured, gap ($17.3B including Blackstone's K-1-based row). Script: tableA1_recompute.py.

Why this severity. The illustration restates an assumption, but the amount is about 3% of the listed-equity gap.

Why it matters, in detail. Implied stakes are 13-38% above Forbes's total Koch net worth and about 50% above Dell's, so the private-firm leg is not conservative relative to the Forbes benchmark SZ22 otherwise favour. The magnitude is about $20-40B, or 0.03-0.05pp of wealth.

Evidence. Reproduced exactly as revenue x 0.75 x Damodaran US price/sales: Koch $129.7B and Cargill $98.5B using the mean of the 5-Jan-2016 and 5-Jan-2017 files; Dell $61.2B using Jan-2016 only; total $289.4B. The Koch brothers' 84% = $108.9B vs Forbes $79.2B (Mar 2016) or $96.6B (Mar 2017). Michael Dell's ~50% = $30.6B vs $19.8-20.4B. Using Forbes values would trim the $258B private-firm gap by roughly $20-40B. Script: damodaran_table2.py.

Why this severity. It affects a component worth about a third of the ~$700B equity correction by tens of billions of dollars. The Forbes route (0.9pp) is unaffected.

Why it matters, in detail. Under the FA constraint the SEC-route effect is +0.67 to +0.84pp, depending on whether the extra equity is rescaled or added to the total. The Forbes route's +0.90pp is unaffected.

Evidence. SZZ 2016: C-corp equity is 14.1% of wealth, and the top 0.01% holds 3.12% of wealth in C-corp equity (43.9% x 7.116%). Rescaling all C-corp equity proportionally to the FA total gives +0.45pp for $459B and +0.67pp for $700B. Adding the $700B to total wealth instead gives +0.84pp. So the listed-stake gap is 21-27% of the 2.16pp gap rather than 'nearly a third'. SZZ's own Forbes row, which raises the total by $0.6T, gives +0.90pp.

Why this severity. The order of magnitude stands; the SEC-route effect is about 0.67-0.84pp rather than 0.9pp.

The authors' response. SZ22 (p. 8) argue the FA aggregates are conservative, which would favour adding the corrections to the total (the +0.84pp case).

Why it matters, in detail. The most consequential undocumented numbers are the estate-based interest correction and the pass-through effects, which together supply about 60% of the claimed reconciliation. The equity results are verifiable.

Evidence. Not reproducible: the next-280 extrapolation ($182B true / $106B SZZ, 21% of the $865B); the realized-gain construction from Form 4 sales (basis, timing); fn. 5's 42.5% lower bound; the statement that about 70% of top-50 Forbes members' private non-financial business wealth was in low-dividend C-corps; fn. 2's 'about half of the $289 billion ... belongs to members of the Forbes 400'; the estate-based x1.6 and '30% of the gap'; any pass-through effect size. Reproducible after re-collection: Tables 1-3, fn. 3-5 ($181B, $31B), fn. 4's '$4 trillion' and fn. 9.

Why this severity. Graded minor only because its substantive consequences for C4-C6 are counted under F1 and F2. For C2-C3 the main results can be rebuilt from public sources and largely reproduce.

Why it matters, in detail. The 2.1 factor is a statement about listed stakes, about a third of Forbes-400 wealth. Its robustness (2.07-2.28) is not in question.

Evidence. Table A1: $865B true (about 36% of the $2.4T Forbes-400 total), of which $182B (21%) is extrapolated; the observed top-120 factor is 2.28. Private equity is assessed separately by valuation (Table 2). SZZ's report (Fig. 1) shows the SZ16 and PSZ18 capitalizations are also an order of magnitude off for the Table 1 individuals.

Why this severity. Wording only; the factor itself is robust (about 2.07-2.28).

The authors' response. SZ22 (p. 3) acknowledge that SZ (2020) and the DFA handle the very top by benchmarking to Forbes rather than through capitalization.