Facundo Alvaredo, Lucas Chancel, Thomas Piketty, Emmanuel Saez, Gabriel Zucman · In J.E. Stiglitz, J.-P. Fitoussi and M. Durand (eds.), For Good Measure: Advancing Research on Well-being Metrics Beyond GDP, ch. 6, OECD Publishing 2018; The New Press 2019 (pp. 181-207)

Distributional National Accounts

15 problems found: 1 major · 6 moderate · 8 minor

What the paper claims

Programme and methods: WID.world/DINA aims to produce annual distributions of income and wealth consistent with national accounts (adults 20+, equal split within couples, SNA-2008 national income, five income concepts) and synthetic micro-files for all countries; the series are 'imperfect, fragile and subject to revision'.
United States: 'complete collapse' of the bottom-50% pre-tax income share and a rise of the top-1% share; 'no growth at all for the bottom 50%' against strong growth at the top.
China: combining tax and survey data revises inequality sharply upward (estimates are 'lower bounds'); very low inequality in the late 1970s, now approaching US levels; very strong growth for all groups including the bottom half.
France ('broadly representative of the West European pattern'): rising inequality confined to numerically negligible top groups; bottom-50% growth equal to average growth.
General rise of net private wealth relative to national income, 'largely unaffected by the 2008 financial crisis', and a decline of public property.
Wealth concentration: a large rise of top-1% wealth shares in the US and China, a more moderate rise in France and the UK.
Conclusion: rising top income and wealth shares in nearly all countries; the varying magnitudes suggest that country-specific policies and institutions matter.

Every problem found

1majorplausible · previously known · external inconsistency

To estimate who gets the income that never shows up on tax forms, the authors had to make assumptions, and other researchers using different assumptions find that the rise in the share going to the richest 1% was only about half as big. Who is right is still debated; either way, the gap between the top and the bottom in the US widened.

Technical detail and evidence

Why it matters, in detail. Core finding C2 rests on allocating untaxed income in proportion to reported income. Under Auten-Splinter's audit- and source-based keys, the 1979-2014 rise of the equal-split top-1% pre-tax share falls from 9.0 to ~4.7 pp, and bottom-half per-capita pre-tax income grows ~40% (1979-2019), against ~0% in PSZ. Neutral evidence suggests PSZ overstate the misreporting component of the trend by at most ~1 pp (Iselin-Reck), and Gale et al. side mostly with PSZ on underreporting. The defensible range for the US top-1% rise therefore runs from ~4.7 to ~9 pp. The direction (rising top share, falling bottom-half share in all estimates) is unaffected.

Evidence. PSZ Nov-2017 data (PSZ2018MainData.xlsx, sheet Data): US top-1% pre-tax share 11.15% (1979) -> 20.20% (2014), +9.04 pp (outputs/psz_t1_1979_2014.txt; scripts/misc_checks.py (b)). Auten & Splinter (2024), JPE 132(7): 2179-2227, Table 4 and pp.2209-2211: AS 2014 top-1% share 14.2%, 6.0 pp below PSZ; PSZ minus AS is 1.7 pp in 1979, so AS's 1979-2014 rise is ~4.7 pp (52% of PSZ's). The 4.3 pp trend gap splits into underreported income 0.7, retirement income 1.2, other taxes 0.5, non-retirement corporate income 0.4, tax-definition corrections 0.5, non-dependent US residents 0.4, inflation correction 0.4 and smaller items. AS (2024, pp.2182, 2204-2205): bottom-50% pre-tax share -5.1 pp and real per-capita pre-tax income +40% over 1979-2019 (individuals ranked by size-adjusted income, PCE deflator), versus PSZ's near-zero growth. Neutral readings: Iselin & Reck (2025, us-income-15) put the misreporting effect on the 1962-2019 change in the pre-tax top-1% share at +0.2 to +0.8 pp for plausible constant relative misreporting rates, against +1.0 pp under PSZ and -0.7 pp under AS. Gale, Sabelhaus & Thorpe (2023, us-income-21) side mostly with PSZ on underreporting. PSZ (2026) attribute only ~5% of the 1979-2019 trend gap to evasion (corpus us-income-10). Auditor B examined the underreporting dispute and declined to score it as an error (codex report s.6).

Why this severity. If the critics' allocations are right, the US top-1% rise is about half the chapter's (+4.7 vs +9.0 pp over 1979-2014). The 'no growth at all' result would also fail under their units and concepts, which materially changes a core finding. The qualitative US divergence survives every estimate, and the critique is disputed, so confidence that an overstatement of this size is real is only 'plausible'.

The authors' response. Saez & Zucman (2020, NBER WP 27921, s.3.1) and Piketty, Saez & Zucman (2024; 2026 comments) contest the audit-based allocation. They argue that random audits miss sophisticated evasion, that Auten-Splinter mix legally exempt depreciation with evasion, and that ranking by true income matters; they now attribute only ~5% of the 1979-2019 trend gap to evasion. The chapter itself warns that its series are 'imperfect, fragile and subject to revision' (p.148).

2moderateconfirmed · previously known · methodology

The authors later found and fixed a mistake in how they assigned retirement income, and with that and other updates their own current figures show the richest 1% gaining a little less and the poorer half gaining a little (about 7% over 37 years) rather than nothing. The overall story does not change.

Technical detail and evidence

Why it matters, in detail. Allocating private pension flows partly by rollover-inflated non-taxable distributions raised top-1% pre-tax income by ~0.6 pp (2016) and top-1% wealth by 1-1.5 pp. In the authors' current series the 1978-2015 top-1% rise is 8.5 pp (vs 9.4) and bottom-half real growth is +6.9% (~0.2%/yr). That is still near-stagnation, against +64% average growth.

Evidence. IRS SOI 2014 Table 1.4 (irs-14in14ar.xls, TBL14 AK24:AK28, AM24:AM28): for AGI >= $1m, pension and annuity amounts $21.06bn, of which taxable $4.36bn, so 79.3% non-taxable (adjudicator re-run; matches Auten & Splinter 2024 fn 33). Saez & Zucman (2020, NBER WP 27921, s.3.3, p.31): PSZ (2018) 'put too high a weight on non-taxable distributions ... inflated by rollovers'. PSZ Updates (s.2.1.7, 2.2.1): revised weights of 60% taxable distributions, 30% wages, 10% non-taxable distributions lower the top-1% pre-tax share by about 0.6 point in 2016 and the top-1% wealth share by 1-1.5 points. Current WID.world (2026-10-15 extract, sptincj992 x anninci992; scripts/wid2026_compare.py): US top-1% share 10.35% (1978) -> 18.89% (2015) vs 10.76% -> 20.19% in the chapter; bottom-50% share 20.24% -> 13.21% vs 19.99% -> 12.53%; average growth +63.8% vs +58.7%; bottom-50% growth +6.9% vs -0.5% (printed -1%). About half of the bottom-50% change comes from revised macro aggregates and half from revised shares.

Why this severity. A genuine, acknowledged error and later corrections shift the US magnitudes by roughly a tenth (top-1% rise 8.5 vs 9.4 pp; bottom-half growth +7% instead of -1% over 37 years). That is noticeable but does not change the qualitative US result.

The authors' response. Accepted: Saez & Zucman (2020, NBER WP 27921, s.3.3, p.31): 'AS raise a valuable point ... We are grateful to AS for bringing this issue to our attention'; the fix 'closes about 10% of the gap' with Auten-Splinter. They argue Auten-Splinter in turn put too little pension income at the top.

3moderateconfirmed · newly identified here · documentation gap

To turn China's money incomes into 'real' growth, the authors used their own estimate of price increases instead of the official one, without saying so in the chapter; with the official figure China's growth would look almost twice as large, and with another respected estimate about a third smaller. Either way Chinese incomes grew enormously for everyone, so the main conclusion stands.

Technical detail and evidence

Why it matters, in detail. All China rows of Table 6.1 scale by a common factor: (1+g_avg) = nominal x90.64 / price ratio. Average growth is +570% (Maddison), +811% (PYZ/Young), or +1,366-1,391% (World Bank/NBS); bottom-50% growth is +268% to +720%. Relative growth across groups and the US-China-France ranking are unchanged. The cross-country comparison silently mixes a researcher-corrected deflator for China with official deflators elsewhere.

Evidence. PYZ (June 2017) national-accounts workbook (scripts/china_deflator.py). Sheet A10: per-adult national income 637.8 yuan (1978) -> 57,806.6 yuan (2015) current, x90.64 nominal. Sheet AP11: price ratio 2015/1978 = 9.955 (PYZ/Young), 6.079 (NBS implicit GDP deflator), 6.182 (World Bank), 13.535 (Maddison). Implied average real growth: +811% (PYZ), +1,391% (NBS), +1,366% (World Bank), +570% (Maddison). Rescaled bottom 50%: +401% / +720% / +707% / +268%; top 1%: +1,898% / +3,171% / +3,117% / +1,369%. PYZ 2017 WP pp.12-13 and AER 2019 p.2474: 'All the real values we report ... use this revised GDP deflator'. The authors' current WID series implies a x13.55 price rise and +538% average growth (scripts/wid2026_compare.py). The chapter's own source workbook (ACPSZ2017.xlsx Data1, '2013 yuan (GDP deflator)') holds another per-adult series rising x11.8 (+1,081%). The Young deflator rises 6.0x over 1978-2000 vs 3.6x officially, but slightly less than official prices over 2000-15.

Why this severity. A headline magnitude, China's growth, ranges from about -30% to +70% of the printed value across defensible deflators, and the choice is not disclosed. Shares and the within-China pattern are unaffected. The PYZ choice lowers growth relative to official data, so it does not favour the chapter's narrative.

The authors' response. None specific to the chapter. PYZ (2017 WP pp.12-13; AER 2019 p.2474) justify the Young (2003)-method deflator as a correction for overstated official growth.

4moderateconfirmed · previously known · overclaiming

The chapter says its new estimate for China's top 1% (about 13% of income) is a minimum, but the underlying study itself points to reasons it could be too high as well as too low, and the tax data behind the correction exist only for 2006-2010 and were stretched to cover 1978 and 2015. The finding that China became much more unequal is solid; the exact levels are less certain than presented.

Technical detail and evidence

Why it matters, in detail. The 2015 corrected top-1% share embeds a 2006-10 tax/survey upgrade profile (f(p) from 1 at P80 to 1.3-1.6 at P99.5) transplanted to 2015 and 1978, plus a ~2.5 pp national-accounts imputation. Offsetting unit biases are of unknown net sign: individual tax incomes overstate equal-split top incomes, while household splitting understates them. A 1978 urban/rural ratio of 2.57 instead of 1.94 lowers the 1978 bottom-50% share by ~2.4-2.8 pp (the authors' current WID 1978 value is 24.7% vs 27.5%), shrinking its 1978-2015 fall by roughly a fifth (from 12.4 pp to ~9.6-10.6 pp). Later vintages raised the 2015 level (14.0-14.5%), consistent with, but not proof of, the lower-bound reading.

Evidence. Chapter p.153 and note 10. PYZ 2017 WP (WID WP 2017/6): fn 19 p.18 (2006-2010 average upgrade profile applied 'to all years of the period 1978-2005' and 2011-2015); fn 20 p.18 (individual tax data make 'our estimates tend to over-estimate inequality as compared to the equal-split benchmark'); fn 14 (household-level split understates inequality); fn 33 (2011-15 extrapolation makes a rebound undetectable); pp.28-29 (2015 top 1%: 6.5% raw -> 11.5% with tax data -> 14% with undistributed profits). ACPSZ2017.xlsx Data3 row 43: raw fiscal top 1% 6.565% vs corrected pre-tax national income 12.682%. PYZ2017DistributionalData.xlsx (2017 and 2018 files) TB2!P6:P8 = 2 typed constants, M6 = AVERAGE(M7:M8) = 1.940 (same 1978 ratio as ACPSZ Data1 col I). The official 1978 per-capita figures (urban 343.4, rural 133.6 yuan; ratio 2.57) are taken from this project's PYZ (2019) adjudication, citing China Statistical Yearbook 2009 Table 9-2. Current WID 1978 China: bottom 50% 24.73%, top 10% 27.98%, top 1% 6.54% (chapter 27.50%, 26.26%, 5.78%). With those 1978 shares and the chapter's 2015 shares and averages, bottom-50% growth would be +457% (vs +401%) and top-1% growth +1,666% (vs +1,898%). Later vintages give a 2015 top-1% share of 14% (June 2017), 14.5% (AER 2019) and 14.04% (WID 2026).

Why this severity. The upward revision is robust in direction (tax data alone give ~11.5%), and 1978 China stays a low-inequality economy under any baseline. But the level is assumption-driven and the claimed one-sidedness is not shown. The 1978-2015 changes shrink by roughly 10-20% under the corrected baseline that the authors' own current data use.

The authors' response. PYZ (2017 WP fn 19) justify the constant correction profile by the absence of a trend in Piketty & Qian's (2009) simulated/observed tax revenues (1986-2003) and by provincial data for 2011-15. Fn 20 acknowledges the opposite-direction unit bias; fn 33 warns that the post-2006 stabilisation 'could partly reflect data limitations'.

5moderateconfirmed · previously known · methodology

The table gives growth figures for tiny groups at the very top (a few hundred to a few thousand people) as if they were precise, but these numbers are mostly modelled rather than observed and have changed a great deal between versions, for France from +158% to +47%. The broader point that top incomes grew fastest still holds.

Technical detail and evidence

Why it matters, in detail. Growth rates for fractiles above the last observed tax threshold are driven by Pareto-tail assumptions and by constructed base years. China's top-0.001% gain relative to the average ranges from 2.9x to 4.3x across vintages, and France's from 1.9x (chapter) to 1.0x (current WID). Integer precision without error bands conveys false precision; only the US very-top rows are stable.

Evidence. China top 0.001%: 3,111% (chapter), 3,817% (cited PYZ June 2017 Table 3), 2,529% (PYZ AER 2019 Table 2, per Auditor A), 2,630% (current WID). Top 0.01%: 2,685%, 3,113%, 2,766%, 2,091%. France top 0.001%: 158% (chapter); ~87% (64-110% given rounding) using the chapter's 1978 shares with GGP's own 2014 shares (Table 1 of the Jan-2018 revision: T1 10.8%, T0.01 1.3%, T0.001 0.4%); 47% in current WID (equal to average growth). France top 0.01%: 93% vs ~67% vs 71%. GGP 2017 s.3.2 and fn 7: before 1988 micro-files exist only for 1970, 1975, 1979 and 1984, and 2013-14 are imputed. US very-top rows are stable (685% vs 694% in current WID). Scripts: scripts/wid2026_compare.py; scripts/misc_checks.py (e).

Why this severity. These are secondary rows that move by up to a factor of three across vintages. The qualitative statements built on them survive using the top-1% rows: the top grew faster than average in all three countries, and in current data France's top 1%-0.01% grew +71% to +93% against +47% on average.

The authors' response. The chapter repeatedly describes its series as fragile, exploratory and subject to revision (pp.148, 158).

6moderatelikely · previously known · methodology

The US wealth figure relies on estimating fortunes from the income people report, assuming everyone earns the same rate of return; since the rich earn more on their savings, this overstates how much they own, and other methods show a smaller rise or, in one case, none. The rich still own a much larger share than in 1980 by most measures, but by less than the chart suggests.

Technical detail and evidence

Why it matters, in detail. Equal within-class yields push top-1% fixed-income wealth up as rates fall (capitalization factors near 100). Heterogeneous yields lower the 2011-2016 top-1% level by ~2-7 pp, depending on unit and source. The authors' current series cuts the 1978-2014 rise from 15.6 to 14.0 pp. The estate-multiplier series is flat (authors attribute this to mortality differentials), and a Social Security-inclusive concept shrinks the rise by about two-thirds (SCF basis). So C6's 'large rise' is robust only in direction, and only for marketable wealth.

Evidence. Fig 6.4 StatLink chart cache: US 21.66% (1978) -> 37.24% (2014), +15.6 pp (= PSZ Nov-2017 TE2b col N, all 102 points). Current WID (shwealj992): 21.79% -> 35.83% (2014), +14.0 pp. Auditor B: SZ (2016) workbook TableB5b vs TableB41b (ten-year bond yield for the top 1%): 2012 tax-unit share 41.82% vs 39.97%. Frozen-group illustration on the chapter's 2014 adult series, s' = s - f(1-1/r) with fixed-income component f = 12.06%: 35.67% (r = 1.15) and 33.80% (r = 1.4), recomputed by the adjudicator. Bricker, Henriques & Hansen (2018): 2011 top-1% share 40.5% -> 33.9% with heterogeneous interest returns. Smith, Zidar & Zwick (2023, Table I): 2016 top-1% 33.7% vs 38.9% under equal returns. Fed DFA ~31% vs ~37% in 2014 (Batty et al. 2019), though Saez & Zucman (2020 JEP) reconcile both at ~38% for 2018 after harmonising units and concepts. ACPSZ2017.xlsx Data5 (adjudicator, scripts/wealth_checks.py): US estate-multiplier top-1% share 19.3% (1976), 21.5% (1995), 20.1% (2007), 20.2% (2008), against 23.5% -> 38.1% for the capitalization series. Catherine, Miller & Sarin (2025): SCF top-1% share +6.4 pp (1989-2019) for marketable wealth vs +1.9 pp including Social Security wealth. This project's adjudication of Saez & Zucman (2016) found that defensible corrections cut the 1978-2012 rise by roughly 15-40%.

Why this severity. Under every marketable-wealth estimate except estate data, the US rise remains large. Its magnitude is 10% (authors' own revised series) to roughly 40% (heterogeneous-return corrections) smaller. The method dependence is not disclosed beside the 'robust conclusions' claim.

The authors' response. Saez & Zucman adopted a top interest-rate premium (about 1.15 in 2003-07 and 1.4 from 2008) and split bond funds from interest-bearing assets (PSZ Updates s.2.1.1). They argue that, after harmonising units and concepts, the DFA and their series both give ~38% for 2018 (SZ 2020 JEP). They attribute the estate-series divergence to rising mortality differentials and estate-tax avoidance (SZ 2016 QJE s.VII.B), and exclude Social Security from wealth for consistency with national accounts (SZ 2020 JEP p.8). The chapter calls wealth statistics 'highly imperfect' (p.157).

7moderatelikely · newly identified here · methodology

The rise in the richest Chinese 1%'s share of wealth is measured by joining two different surveys that capture the rich differently, and the in-between years were filled in with adjustments the paper does not describe. Wealth concentration in China clearly rose, but probably by less than the chart shows.

Technical detail and evidence

Why it matters, in detail. The China top-1% wealth trend (+13.8 pp, 1995-2015) is identified mainly across the CHIP->CFPS switch; CFPS measures more top wealth than CHIP (wealth Gini 0.74 vs 0.62 around 2012-13, per Luo & Chen 2021 as cited in the PYZ adjudication). In years without survey data (2003-09) the published path is driven by a ramped upgrade factor (1.6->2.1); this project's PYZ adjudication attributes ~90% of the 2003-09 top-decile increase to the ramp. A like-for-like CHIP comparison implies a top-decile rise about half as large over 2002-2013.

Evidence. PYZ 2017 WP p.20 and fn 25-26: CHIP 2002 -> 2013 top-10% wealth share 37.1% -> 48.4% (+11.3 pp, Knight & Li 2016) vs CHIP 2002 -> CFPS 2010 38.7% -> 64.2% (+25.5 pp, Li & Wan 2015). PYZ BenchmarkWealthTabulations.xlsx (2018 files; scripts/pyz_wealth_factors.py) survey-top upgrade factor by year: 1.3 (1995), 1.5 (2002-03), 1.6, 1.7, 1.8, 1.9, 2.0, 2.1 (2004-09), 1.3 (2010), 1.5 (2011-15). The 2003-09 sheets are the CHIP-2002 tabulation rescaled by aggregate means (e.g. '2006'!H5 = '2002'!H5*T0!L26/T0!L22), and 2011 is the CFPS-2012 sheet rescaled. Fig 6.4 StatLink, China top-1% share: 15.8% (1995), 20.4% (2002), 20.5% (2003), 22.5%, 23.7%, 26.2%, 28.5%, 29.3%, 31.2% (2004-09), 30.5% (2010), 27.3% (2012), 29.6% (2015). The 2015 value equals PYZ AER 2019 Table 3 and current WID (29.63%). This project's PYZ (2019) adjudication (F1, major/likely) puts the defensible 1995-2015 top-10% rise at ~18-32 pp vs 26.6 pp published.

Why this severity. Wealth concentration in China rises in every survey and rich-list source, so the direction of C6 for China survives. The magnitude of the rise, and the 2003-09 path, are partly constructed; the rise is plausibly up to ~30% smaller.

The authors' response. None located. PYZ (2017 WP fn 25) justify not using CHIP 2008 (sampling problems) and CHIP 2013 (no housing values); PYZ (AER 2019 fn 28) describe the upgrade factor's effect as second order.

8minorconfirmed · newly identified here · overclaiming

A footnote predicts that counting taxes and government benefits would make the US look even worse relative to France, but the authors' own figures show that after taxes and benefits the gap in levels narrows somewhat (though the trend is similar). The main text is clear that it measures income before taxes and benefits, so this is a small overstatement.

Technical detail and evidence

Why it matters, in detail. Post-tax, the US-specific level contrast (bottom-50% share < top-1% share) disappears in PSZ's own series. Bottom-50% post-tax national income grows ~19-21% (1978/80-2014), though cash disposable income grows only ~9-11%. US redistribution narrows rather than widens the US-France level gap (T10/B50 ratio 2.55 -> 2.25), while the trend divergence is preserved.

Evidence. PSZ (2016, NBER WP 22945) Table 1 (2014): bottom-50% vs top-1% shares 12.5% vs 20.2% pre-tax, 18.1% vs 17.0% post-tax disposable, 19.4% vs 15.6% post-tax national income. Table 2: bottom-50% growth 1980-2014 +1% pre-tax vs +21% post-tax. PSZ Nov-2017 data (scripts/psz_posttax.py): bottom-50% growth 1978-2014 -1.6% pre-tax, +18.8% post-tax national income, +5.7% excluding health transfers, +8.6% cash disposable income. PSZ (2016) fn 61 made the same 'even greater after taxes and transfers' prediction. Bozio, Garbinti, Goupille-Lebret, Guillot & Piketty (2018, WID WP 2018/10, pp.17-19, Table 2): taxes and transfers cut T10/B50 by 34% in the US vs 23% in France; T10/B50 1990 -> 2015 is 11.4 -> 18.9 (US pre-tax), 6.2 -> 7.4 (France pre-tax), 8.1 -> 11.5 (US disposable), 5.2 -> 5.1 (France disposable). The US/France ratio thus falls from 2.55 to 2.25 in levels, while the 1990-2015 trend gap is ~0.33 log points pre-tax and ~0.37 post-tax.

Why this severity. The chapter clearly labels its series as pre-tax, the trend contrast survives post-tax, and the policy emphasis on pre-tax inequality is supported by Bozio et al. Only the footnote's level prediction is wrong, which does not affect the core findings.

The authors' response. None specific. PSZ (2016, pp.20-21) stress that bottom-50% post-tax gains are almost entirely in-kind health transfers and that cash disposable income stagnated. Bozio et al. (2018) conclude that the US-France gap is 'entirely due to differences in pretax inequality'.

9minorconfirmed · newly identified here · documentation gap

Some of the '2015' numbers are really earlier years' figures carried forward, and the chapter cites newer versions of its sources than the ones its numbers actually come from, which report somewhat different figures. This makes the numbers harder to check but does not change the conclusions.

Technical detail and evidence

Why it matters, in detail. The '1978-2015' endpoints are partly distribution-neutral nowcasts (US, France) of earlier observed or imputed years. The cited 2017 sources report different very-top numbers and a higher 2015 China top-1% share (~14%). Fig 6.2/Table 6.1 and Fig 6.4 use different PSZ vintages. None of this alters average, bottom-50% or top-1% conclusions.

Evidence. ACPSZ2017.xlsx Data3 formulas (scripts/data3_endpoints.py): US 2015 shares =AA42:AD42 (2014). France 2014 =W41 (2013) and 2015 =W42; the 2013 values are themselves imputed (GGP fn 7), and the '2015' French average (EUR 34,578) is GGP's 2014 average. Fig 6.3 chart cache: Japan, UK and Germany 2014-15 equal 2013; France 2015 equals 2014. Fig 6.4: US ends 2014, France and UK 2012, despite the title '1890-2015'. Data3 headers: 'China PYZ 2116 [sic]', 'DINA France (GGP 2016)', 'DINA USA (PSZ 2016)'; ACPSZ (2017) cites PYZ (2016) and GGP (2016), while the chapter cites PYZ (2017) WP 2017/6 and GGP (2017) WP 2017/4. Cited PYZ (June 2017): top 1% 'around 14%' (p.6) vs 'around 13%'; its Table 3 differs from Table 6.1 in 6 of 8 China rows (middle 40% 768 vs 779; top 10% 1,289 vs 1,294; top 0.1% 2,405 vs 2,261; top 0.01% 3,113 vs 2,685; top 0.001% 3,817 vs 3,111). Fig 6.4 US = PSZ Nov-2017 TE2b col N (102/102 points); Data3 col AL (Dec-2016) differs by up to 0.08 pp in 1978-86. Auditor B, verified by the adjudicator: the 2015 US adult count (236.73m) is extrapolated and 0.88% below the Census 2016-vintage count (238.81m); using Census adults gives average growth +57.3% and bottom-50% growth -1.4%.

Why this severity. Documentation and vintage problems; the effect on the main rows is negligible. US bottom-50% growth is -0.5% with the carry-forward vs -1.6% for 1978-2014, and average growth +58.7% vs +57.3% with Census adults. They do complicate replication from the citations alone.

The authors' response. None specific; distribution-neutral carry-forward is standard WID practice, and the OECD StatLinks preserve the plotted vintage.

10minorconfirmed · previously known · version instability

The authors' database has been updated since 2018, and a few details in the chapter no longer hold, for example that France's poorer half gained exactly as much as the average and that Japan's government has negative net wealth. The main patterns described remain.

Technical detail and evidence

Why it matters, in detail. In current data, French bottom-half growth is 85% of average growth, not equal to it; the China bottom-50%/top-1% ordering in 2015 becomes a tie; Japan's net public wealth sign flips. The China wealth-aggregate revisions enlarge the documented privatisation (public share -48 pp rather than -38 pp).

Evidence. Current WID (2026-10-15 extract; scripts/wid2026_compare.py, scripts/wid2026_wealthagg.py). France 1978-2015: average +47.4%, bottom 50% +40.4% (chapter: both +39%). China 2015 bottom-50% vs top-1% shares 14.12% vs 14.04% (chapter 15.13% vs 12.68%). Japan's 2015 net public wealth +6.6% of national wealth (chapter data -0.9%; text 'negative'). China 1978 public share 82.1% (chapter 69.2%) and private wealth 73% of national income (chapter 115%), so the decline of public property and the rise of private wealth are even larger in current data. US 2015 public share -6.6% and UK -9.2% (still negative).

Why this severity. These revisions post-date the chapter, which warned that its series are 'subject to revision'. They change secondary statements but reverse no core finding.

The authors' response. The chapter warns that its series are 'subject to revision' (p.148); the US revisions are documented in PSZ Updates (2020-2022).

11minorconfirmed · newly identified here · internal inconsistency

The text gives a figure for China's government-owned share of wealth in 2015 (35%) that does not match the chart printed right below it (about 32%). The error, if anything, understates the shift from public to private ownership.

Technical detail and evidence

Why it matters, in detail. The stated endpoint overstates the 2015 public share by 3.3 pp; the 1978-2015 decline is 37.5 pp, not ~35 pp.

Evidence. Fig 6.3 StatLink chart cache and ACPSZ2017.xlsx Data2 col W: 69.17% (1978), 31.66% (2015). Printed Fig 6.3 (p.156, rendered by the adjudicator) shows the China line ending just above 30%. PYZ alternative variants (Data2 cols AF/AG) give 28.9% and 38.6%. PYZ 2017 WP (pp.1, 6) and AER 2019 (p.2469): 'about 70% ... to about 30%'. Current WID: 33.8% (2015).

Why this severity. A 3.3-point text/figure inconsistency that understates, rather than manufactures, the decline; no conclusion is affected.

12minorconfirmed · newly identified here · misrepresentation of source

The chart credits the US wealth numbers to the wrong paper; the plotted figures come from a related study by the same authors that counts individuals rather than tax households. This matters only for readers trying to trace the numbers.

Technical detail and evidence

Why it matters, in detail. Wrong source label for the plotted series (PSZ equal-split adults vs SZ tax units); the 2012 levels differ by ~3 pp because of the unit.

Evidence. Fig 6.4 chart cache US points match PSZ2017AppendixTablesII(Distrib).xlsx TE2b col N exactly for all 102 years (adjudicator check; Auditor B found the same mapping). 1978 21.66%, 2012 38.85%, 2014 37.24%. Saez-Zucman (2016) capitalization series (ACPSZ2017.xlsx Data5 col P): 22.94% (1978), 41.82% (2012).

Why this severity. A citation error: the PSZ series builds on the SZ method, and the plotted data are correctly described as equal-split adults. A reader checking SZ (2016) will find 41.8% for 2012, not the 38.8% plotted.

13minorconfirmed · newly identified here · overclaiming

A few phrases go further than the chapter's own figures: the poorer half of Americans saw their incomes stall, not fall; US household wealth did dip sharply after 2008 before recovering; and the UK's top wealth share in 2012 was below its 1978 level, rising only from a low point in the late 1980s.

Technical detail and evidence

Why it matters, in detail. Stagnation is described as a fall; a temporary 18% drop and recovery of the US wealth-income ratio is described as 'largely unaffected' (true for the other six countries, apart from one-year dips of ~4% in the UK and ~9% in Norway); the UK 'rise' holds only from 1988 (+4.7 pp), not from 1978 (-1.3 pp).

Evidence. (i) ACPSZ2017.xlsx Data4: US bottom-50% real mean $16,486 (1978) -> $16,402 (2015), -0.51%, while the share fell 19.99% -> 12.53%. (ii) Data2 net private wealth / national income, US: 499% (2007), 444% (2008), 411% (2009), 416% (2010-11), 420% (2012), 464% (2013), 492% (2014), 500% (2015); UK 541% -> 521% (2008) and Norway 221% -> 201% (2008), both recovering the next year; France, Japan, Germany and China show no fall (outputs/private_wealth_ratios.txt). (iii) Fig 6.4 chart cache, UK top-1% share: 21.2% (1978), 18.5% (1979), 15.2% (1988 trough), 16.2% (1995), 19.9% (2012) (outputs/wealth_checks.txt).

Why this severity. Rhetorical overstatements: the correct numbers are printed next to them, and none alters a conclusion.

14minorconfirmed · newly identified here · internal inconsistency

The table and chart notes say incomes are shared equally between spouses in all three countries, but for China they are shared across everyone in the household, which is not quite the same and makes the comparison less exact.

Technical detail and evidence

Why it matters, in detail. China's series is a hybrid of a broad (household) equal split in the survey part and individual incomes in the tax-corrected tail, not the narrow (couple) split the captions state for all countries.

Evidence. Chapter p.149, p.151, notes to Fig 6.2 (p.154) and Table 6.1 (p.155). PYZ 2017 WP fn 14 (household-level split; 'we probably somewhat underestimate inequality in China as compared to the United States and France') and fn 20 (individual tax incomes).

Why this severity. A labelling inconsistency; the comparability caveat appears elsewhere in the chapter, and the net bias is unknown (see F5).

The authors' response. PYZ (2017 WP fn 14) acknowledge the household-level split; the chapter flags broad vs narrow splitting for China on p.151.

15minorconfirmed · newly identified here · arithmetic or transcription

A few typos and labelling mistakes (a wrong year for an international statistics manual, mixed-up titles in the downloadable data files). They do not affect any finding.

Technical detail and evidence

Why it matters, in detail. Editorial; the 0.5 pp rounding slightly overstates the fall in the US bottom-50% share in the text (7.5 pp actual).

Evidence. Chapter p.148, p.147, p.153. Table 6.1 StatLink Sheet1!B2 ('Table 6.1. Ratio of household adjusted disposable ...'), verified by the adjudicator; Figure 6.1 StatLink B2 carries Table 6.1's title (Auditor B). 2008 SNA Preface (predecessor = 1993 SNA); Piketty & Zucman (2014, QJE p.1262): 'only in 1993' (Auditor A).

Why this severity. Presentation slips with no effect on any result.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

Note 11 is 'contradicted' by the authors' own post-tax series and the issue warrants moderate severity (Auditor A, I2). — Partly rejected (downgraded to minor, F8). The level part is right: post-tax, the US bottom-50% share exceeds the top-1% share, and redistribution is larger in the US than in France. But the US-France trend contrast is unchanged or slightly stronger post-tax (T10/B50 1990-2015: US +66% pre-tax / +42% disposable; France +19% / -2%; Bozio et al. 2018). The chapter labels its series as pre-tax, and the policy emphasis on pre-tax inequality is supported by the authors' French post-tax evidence.
Later WID revisions constitute a moderate version-instability issue in which several core numbers move by more than 25% (Auditor A, I6). — Partly rejected. Revisions after publication are not errors in the 2018 chapter, which warned its series were 'subject to revision'. The US and China headline revisions are scored where their causes lie (F2 pension and other corrections; F4 deflator). The remaining changes are graded minor (F10), and two of them (China 1978 public share 82%, private wealth 73% of national income) strengthen the chapter's narrative.
China's raw-survey 2015 values equal the 2014 values (part of Auditor A, I9). — Not exact in the chapter's workbook: ACPSZ2017.xlsx Data3 raw top-1% share is 6.5651% (2014) vs 6.5653% (2015), and other shares differ in the fourth decimal. The 2015 survey shape is essentially the 2014 one reweighted (as the PYZ files show), so the substance stands, but the 'bit-identical' description is wrong; it is folded into F9 without this claim.
An apparent discrepancy between the chapter's US wealth line and the source (an initial comparison against a tax-unit column). — Withdrawn by Auditor B itself and confirmed spurious: the US line matches PSZ Nov-2017 TE2b col N at all 102 points (adjudicator check). What remains is only the citation error recorded in F12.
China's 120,000 -> 150,000 yuan fiscal-threshold conversion may be an error (examined by Auditor B). — Auditor B did not establish an error, and the adjudicator agrees it is not a demonstrable error. The 80% taxable/fiscal ratio is a documented assumption with 70%/90% variants. It is part of the two-sided uncertainty described in F5, not a separate finding.
Corpus critique international-global-03 (Galbraith 2019): the US bottom-50% 'collapse' is a 'nonsensical' artefact of misunderstood low-income filers (examined by Auditor A, who did not uphold it). — Not upheld for this chapter. PSZ distribute all national income and impute non-filers, and a large fall of the bottom-50% share appears in every DINA-type US series, including Auten & Splinter's (-5.1 pp, 1979-2019). The TRA86 reporting break concerns fiscal-income series; DINA adds retained earnings. The magnitude debate is covered in F1.
Corpus critiques us-income-09 and us-income-05: tax-unit and marriage-decline bias inflates the chapter's US top shares (examined and dismissed by both auditors). — The chapter uses equal-split adults, not tax units. In Auten & Splinter's Table 4 the unit/grouping difference ('groups by individuals/size-adjusted incomes') accounts for only ~0.1 pp of the 2014 top-1% gap.
Corpus critiques us-income-13 and us-wealth-taxes-12/13: public-goods, tax-incidence and deficit-allocation conventions bias the chapter's results (examined and dismissed by Auditor B). — The chapter presents no post-tax series, so these conventions do not affect its numbers (Auditor B reached the same conclusion).
Corpus critique us-income-19 (Splinter & Auten): cross-sectional growth incidence ignores mobility, so Table 6.1 misstates who gained (examined and dismissed by Auditor B). — Table 6.1 is standard anonymous growth incidence, and the chapter makes no claim about the same individuals over time (Auditor B's assumption inventory). The panel approach answers a different question and mechanically embeds mean reversion.
Corpus critiques capital21-05 (Rognlie 2015) and capital21-06 (Bonnet et al. 2014): the rise in wealth-income ratios is a housing-price effect that the chapter misreads (examined and dismissed by Auditor A). — Accommodated: the chapter explicitly lists rising real-estate prices among the drivers of the rise (pp.155-156) and makes no claim about productive capital.
The HAL record's four-author listing (omitting Saez) indicates an authorship problem (examined by Auditor B, who did not score it). — Metadata only: the chapter's title page (p.143) lists all five authors.

Audit notes

What reproduced. Reproduced exactly from the authors' own files: all 24 cells of Table 6.1 (ACPSZ2017.xlsx Data4); the Fig 6.2-6.4 series (StatLink chart caches vs ACPSZ Data2-Data5 and PSZ Nov-2017 TE2b); the p.153 text numbers; China's +811% from PYZ's national accounts; the private wealth-income ratios and public-wealth signs on pp.155-156; and the Box 6.1 ratios. Not matching: the text's 35% China public share (data 31.7%); the cited PYZ June-2017 Table 3 and GGP Jan-2018 2014 shares (the chapter uses Dec-2016 vintages); and 2015 endpoints that are copies of 2014 (US) or 2013 (France). Sensitivities re-run by the adjudicator: China growth under the NBS/World Bank/Maddison deflators (+1,391%/+1,366%/+570% vs +811%); current-WID versions of every Table 6.1 row; PSZ pre- vs post-tax bottom-50% growth; US CPI-U-RS deflation; the PSZ 1979-2014 top-1% rise vs Auten-Splinter's Table 4 gap; the IRS 2014 non-taxable pension share (79.3%); the PYZ wealth upgrade factors and 1978 urban/rural constant. Not reproduced: the underlying PSZ/PYZ/GGP microsimulations, Auten-Splinter's or Smith-Zidar-Zwick's re-estimates (taken from the published record), and any reranked counterfactual series; corrections in different issues are not additive.

Limits of this audit. The New Press 2019 edition (pp.181-207) could not be obtained by either auditor (publisher bot protection), and HAL holds no full text, so cross-edition differences are unverified; the OECD 2018 text was audited. The adjudicator worked from the authors' summary files (ACPSZ2017 workbook, PSZ Nov-2017 files, PYZ 2017/2018 workbooks, GGP Jan-2018 paper) and did not re-run the PSZ, PYZ or GGP microsimulations. The Auten-Splinter and Smith-Zidar-Zwick magnitudes are taken from their published tables, not re-estimated. The China wealth upgrade-factor evidence comes from PYZ's 2018-vintage files; the chapter's Dec-2016 vintage path is consistent with it but its factor file was not inspected. The official 1978 Chinese urban/rural income figures are taken from this project's PYZ adjudication (China Statistical Yearbook 2009), not re-downloaded. Current-WID comparisons use the shared 2026-10-15 filtered extracts. Box 6.1 was checked against its OECD source only by Auditor A (Wayback copy). Social Security wealth effects are from Catherine, Miller & Sarin's SCF-based estimates, not a DINA re-estimate. Sensitivities in different issues are not additive.