Capital is Back: Wealth-Income Ratios in Rich Countries, 1700-2010

Why it matters, in detail. Capital gains in the 1970-2010 decomposition are the residual after cumulating measured volume flows (saving plus other volume changes), so any R&D credited to saving is removed one-for-one from capital gains. With the A99 income weights (sum of income x rise = 1.299), each 0.1 of 2010 national income of uniform cumulated R&D adds 7.7 points to the additive saving share: 24.3% with none, 39.7% at 0.2 (the BEA figure PZ call an 'extreme lower bound'), 62.8% at the 0.5 behind the main-text 60/40 (multiplicative: -9.4% with none, 57.5% at +0.5%/yr). BEA-consistent U.S. allowances are only 0.08-0.18 of 2010 NI (perpetual-inventory run on NIPA R&D investment, 12-29% depreciation) or ~0.09 (PZ's own memo series of BEA net R&D stocks), so 0.5 is roughly 3-6 times what BEA data imply, and R&D intensity is lower in most other sample countries (scaling a 0.12 U.S. allowance by relative intensity gives a 33% saving share). At BEA-consistent values saving explains ~30-38% (additive) or roughly 0-10% (multiplicative, the version the authors say they prefer) of the income-weighted rise, so capital gains, mostly on housing and land, account for roughly 60-75% and over 80% respectively instead of 40%: the main-text attribution flips. Separately, other volume changes make up about 40% of the 62.8% 'saving' share (37.5% without them); in the U.S., Table V's gws of 2.12%/yr is 1.43%/yr saving plus 0.67 other volume changes, so U.S. wealth growth splits 49/23/28 (saving/other volume/capital gains), not 72/28. Direction: the main text (and its restatement in Sec. VI) overstates the saving channel of C3 and understates price effects; the measured rise (C1) and the qualitative presence of both channels are unaffected.

Evidence. Adjudicator recomputation from the country workbooks and App. Table A99 cells (scripts/reproduce_core.py, mult_rd_grid.py, us_rd_pim.py; logs/). A99 weights (income x rise; sum = 1.299) imply each 0.1 of uniform R&D allowance adds 7.7 points: additive saving share 24.3% (no R&D), 32.0% (0.1), 39.7% (0.2), 62.8% (0.5, = A99). Excluding other volume changes: -1.0% (0), 14.4% (0.2), 37.5% (0.5). Multiplicative (reconstructed, A99 = 57.2%): -9.4% (0), 4.0% (0.1%/yr = +0.13 of U.S. NI), 17.4% (0.2%/yr), 57.5% (0.5%/yr = +0.68 of U.S. NI). Benchmarks for the allowance: (a) PZ's own memo series of BEA net R&D stocks (USA.xlsx Table US.6b col. 10) is 15.0% of NI in 1970 and 13.9% in 2006, implying cumulated real net R&D investment of only ~0.09 of 2010 NI (0.14 - 0.15 x 0.33); (b) a perpetual-inventory run on current BEA NIPA R&D investment (private Y006RC + government Y057RC, ~3% of NI) gives 0.08-0.18 of 2010 NI for the 12-29% depreciation range PZ cite (App. p.18), 0.11-0.15 at 15-20%; (c) Bauluz (2017, rev. Aug. 2019, WID.world WP 2017/23, PDF pp.5-6, 11) finds that after SNA-2008 R&D capitalisation 'net capital formation ... is almost identical in the new and the old series' because R&D depreciates at ~20%; (d) R&D intensity is lower outside the U.S. (World Bank GB.XPD.RSDV.GD.ZS, 1996-2010 averages, % GDP: US 2.58, JP 3.00, DE 2.40, FR 2.13, UK 1.59, IT 1.06, CA 1.86, AU 1.91; contrast App. fn 42 'most countries ... about 2-3%'); scaling a BEA-consistent U.S. allowance (0.12) by relative intensity gives a 33% saving share. Auditor A's check (2026Q2 Fed Z.1 R&D-inclusive official U.S. net saving averaging 5.31% vs PZ's 5.22% over 1970-2009) points the same way but also absorbs other revisions.

Why this severity. Under defensible (BEA-consistent) R&D allowances the attribution in the main text flips: capital gains, mostly on housing and land, account for roughly 60-75% of the income-weighted 1970-2010 rise in the additive decomposition and over 80% in the multiplicative one the authors say they prefer (p.1274), versus the 40% claimed. The measured rise (C1) is untouched and both channels named in the abstract survive qualitatively, so this is not critical.

The authors' response. Online Appendix A.5.2 (p.36): the computations are 'merely illustrative'; saving explains 'at least 40%, and more probably around 60%'; the BEA 20% is 'an extreme lower bound' because BEA may overstate depreciation or understate gross R&D (App. A.2.3, p.18). No later published response located.

Why it matters, in detail. In the accumulation identity beta rises through the volume channel only if saving-induced wealth growth gws exceeds income growth g, so Table V's 'share of wealth growth' explained by saving (U.S. 72%) is not a share of the rise in beta. Over 1970-2010 gws (including other volume changes, before any R&D) was below g in the U.S. (2.12 vs 2.81%/yr), UK (1.45 vs 2.20) and Australia (2.55 vs 3.19), and 1970-2010 s/g (1.86, 2.42, 2.80) lay well below the 1970 ratios (4.04, 3.65, 3.91): volume flows alone (q = 0) would have taken beta to 327%, 293% and 365% by 2010 against 431%, 527% and 584% observed, so the entire rise in these countries is a valuation effect; even with PZ's generous R&D the A99 saving shares are -100%, -14% and +12%. The long-run s/g target fell because net saving halved, not because growth slowed (U.S. 1950-80 -> 1980-2010: 3.08 -> 1.60, s 10.9% -> 4.4%, g 3.53% -> 2.77%; UK 4.66 -> 1.77 with g rising from 2.11% to 2.55%; Australia 1970-90 -> 1990-2010: 4.02 -> 2.15 with g unchanged at 3.19%). These three countries carry 54% of the A99 income weights, so the growth-slowdown channel of C3/C4 operates in five of eight countries (Germany, France, Japan, Italy, Canada): the claim is narrowed, not overturned, and the authors partly concede the U.S. case (no rise absent net foreign gains, p.1291).

Evidence. Adjudicator recomputation from Table XX.5b of each country workbook (logs/reproduce_core.txt). gws(incl. other volume) vs g, 1970-2010: U.S. 2.12 vs 2.81, UK 1.45 vs 2.20, Australia 2.55 vs 3.19 %/yr; s/g (1970-2010) vs beta(1970): U.S. 1.86 vs 4.04, UK 2.42 vs 3.65 (corrected series), Australia 2.80 vs 3.91; q = 0 predicted 2010 beta: 327%, 293%, 365% vs observed 431%, 527%, 584%. Even with PZ's generous R&D, A99 saving shares are U.S. -100%, UK -14%, Australia +12%. These three countries carry 54% of the A99 income weights. The mechanism does operate in Germany (s/g 3.56 -> 5.32), France (3.47 -> 4.41), Japan (4.86 -> 16.1), Italy (3.87 -> 7.84) and Canada (3.50 -> 3.87).

Why this severity. The paper's signature mechanism fits five of eight countries but not the U.S., UK or Australia, where the rise is wholly a valuation effect; the paper partly concedes this (U.S. foreign gains, large UK gains), so the claim is narrowed rather than overturned.

The authors' response. PZ restrict 'capital is back because low growth is back' to Europe and Japan, explain lower U.S. ratios by population growth, and state that 'absent net foreign gains, the U.S. wealth-income ratio would not have increased at all since 1970' (p.1291) and that capital gains are larger in the UK, Italy and Australia (p.1286).

Why it matters, in detail. C5 asserts that relative asset prices are roughly trendless over 1870-2010 (q ~ 0). The one-good (q = 0) counterfactual built from PZ's own saving, war-destruction and other-volume flows reaches only 71% (U.S., 308% vs 431%), 58% (UK, 308% vs 527%) and 74% (France, 446% vs 605%) of observed 2010 national wealth/NI, versus 113% in Germany (470% vs 416%); the implied cumulative relative-price factors are x3.14, x1.51, x1.27 and x0.45, and App. Table A108 attributes 29%, 41%, 26% and -13% of 2010 levels to capital gains. The U.S. 1870-2010 q (0.8%/yr) equals the 1970-2010 U.S. q that Sec. V calls 'a clear pattern of positive capital gains', and the 1910-2010 'slight overshooting' (+0.24-0.25%/yr in the UK and France) compounds to +27-29%. The European-level wash-out therefore rests on Germany's negative q, itself sensitive to the contested pre-1914 German level (F5). Direction: long-run price effects are larger than the paper says, which weakens reading the 1970-2010 gains as mere recovery of 1910-1950 losses; the 'capital policies' explanation is argued, not tested, and Knoll, Schularick & Steger (2017) find flat real house prices until the mid-20th century and land explaining ~80% of the post-war boom, more consistent with a structural rise. Saving still accounts for the majority of long-run accumulation everywhere, so C5 survives qualitatively.

Evidence. Adjudicator recomputation (logs/reproduce_core.txt): predicted vs observed 2010 national wealth/NI from 1870 with q = 0: U.S. 308% vs 431% (0.71), UK 308% vs 527% (0.58), France 446% vs 605% (0.74), Germany 470% vs 416% (1.13); cumulative (1+q) 1870-2010: U.S. x3.14, UK x1.51, France x1.27, Germany x0.45. Online App. Table A108 (additive): capital gains = 29% (U.S.), 41% (UK), 26% (France), -13% (Germany) of 2010 levels. 1910-2010 additive: capital gains 39% (UK, corrected series) and 27% (France) of 2010 levels. Knoll, Schularick & Steger (2017), AER 107(2): 331-353, abstract (Crossref).

Why this severity. Saving still accounts for the majority of long-run accumulation everywhere, so the core long-run claim survives qualitatively, but 'wash out' and 'no significant divergence' do not describe the U.S., UK or France, which weakens reading the 1970-2010 gains as mere recovery.

The authors' response. p.1293: 'the imperfections of our data do not allow us to put a precise number on asset over- or undervaluation in 2010' and residuals 'might also reflect measurement issues'; p.1260: 'It is only in the very long run or at a very aggregate level (i.e., at a European rather than country level) that relative price effects seem to wash out.'

Why it matters, in detail. The sigma > 1 inference compares the proportional rise in the capital share alpha with that in k. The measured rise in alpha depends on the base year: the cross-country mean moves from 25.0% (1970), 20.5% (1975) or 22.4% (1970s average) to 27.2% in 2010, i.e. +2.3, +6.7 or +4.9 pp, so Fig. XII's 1975 start roughly triples the rise relative to 1970; from 1970 Japan (36.1% -> 27.0%), France (23.6% -> 22.8%) and Italy (28.7% -> 28.6%) show no rise, and the 2010 range is 22.8-31.0% rather than 25-35%. Implied sigmas from 1970 scatter widely (Japan 0.60, France 0.95, Italy 1.00, UK 1.53, Australia 1.54, Canada 4.74), and in the U.S. and Germany alpha rose proportionally more than k, which no fixed-technology CES can produce. More fundamentally, identifying sigma from alpha-k co-movement requires k to rise through volumes, whereas most of the 1970-2010 rise in market-value k is a relative-price effect (F1); the sigma involved is a net elasticity, and micro and meta-analytic estimates lie below 1. C8 is therefore weakly supported; it is explicitly exploratory, and the core wealth findings are unaffected.

Evidence. Adjudicator computation from App. Tables A48 and A51 (logs/capital_share_check.txt): mean alpha 25.0% (1970) / 20.5% (1975) / 22.4% (1970s avg) -> 27.2% (2010), i.e. +2.3 / +6.7 / +4.9 pp; from 1970 Japan 36.1 -> 27.0, France 23.6 -> 22.8, Italy 28.7 -> 28.6; 2010 range 22.8-31.0%. Implied sigma from 1970: Japan 0.60, France 0.95, Italy 1.00, UK 1.53, Australia 1.54, Canada 4.74; U.S. and Germany: alpha up more than k (no CES); from 1975 the U.S., Germany, UK and Australia fit no CES. The UK A48 column is also lagged one year (F9).

Why this severity. Affects a secondary, explicitly exploratory result that the abstract nevertheless highlights ('the shape of the production function, and the rise of capital shares'); the core wealth findings are unaffected.

The authors' response. p.1306: the discussion 'should be viewed as merely exploratory and illustrative'; p.1303 lists bargaining-power and three-factor alternatives. Piketty (2015, JEP p.81; AER P&P Sec. IV) moves to multi-sector models with higher inter-sector elasticities.

Why it matters, in detail. beta = s/g implies that halving g doubles beta only if the net saving rate s is invariant to g. With a constant gross saving rate s_G and depreciation delta (as a share of wealth), beta = s_G/[g + delta(1 - s_G)]; calibrated to beta = 333% at g = 3%, halving g raises beta by 38% (delta = 3%), 32% (4%) or 23% (7%) rather than 100% (Auditor B: +32/+23/+21% for delta = 4/7/8%), so the abstract's illustration (300% -> 600%) overstates the growth effect by a factor of about three to five. Net saving was not invariant in PZ's own data: income-weighted net national saving roughly halved between 1950-80 and 1980-2010 (U.S. 10.9% -> 4.4%, UK 9.8% -> 4.5%, Germany 17.6% -> 9.2%, France 15.6% -> 7.9%). The forward-looking claims (low-growth countries 'bound to have high beta'; long-run beta 'can naturally be very high - 600-700%, or even more') inherit the assumption, so the quantitative content of C4 and of the growth channel in C3 is overstated; the sign (lower g raises beta) survives in most models and the descriptive results are unaffected.

Evidence. Adjudicator derivation and arithmetic (matches Auditor B's computation: +32.1/+22.7/+20.9% for delta 4/7/8%); income-weighted saving rates by sub-period from the country workbooks (logs/reproduce_core.txt). Krusell & Smith (2015, JPE 123(4)) and Jones (2015, JEP 29(1)) make the same point.

Why this severity. The quantitative strength of one of the two mechanisms in the abstract is overstated by a factor of about three to five under standard saving behaviour; the sign (lower g raises beta) survives in most models and the paper mentions endogenous-saving models (fn 23), so the descriptive results are unaffected.

The authors' response. Piketty (2015), 'Putting Distribution Back at the Center of Economics', JEP 29(1): 67-88, pp.81-82: 'there is, of course, no reason why the net-of-depreciation saving rates should be viewed as a constant'; beta 'will not go to infinity' because, as Jones notes, 'depreciation would then become enormous'. He maintains that beta falls with g.

Why it matters, in detail. Two effects. (i) For 1870-1913 German private wealth is Hoffmann's net fixed assets, built by cumulating net investment, plus land, public debt and net foreign assets, while saving comes from the same Hoffmann accounts; for reproducible capital the Table VII 1870-1910 result (saving 107%, q = -0.1%/yr) is therefore largely an accounting consequence rather than a test. (ii) The level is revised down: Albers, Bartels & Schularick (2022), mainly through lower land values, give private wealth of 476% vs PZ's 608% of NI in 1910 (477% vs 623% in 1913), and WID, which now uses their series, gives 456% vs 608% (private) and 484% vs 637% (national) in 1910 and 485% vs 745% (national) in 1870; even Helfferich's contemporary estimate is 575% vs 636%. Germany's 2010 ratio then stands at ~90% of its pre-1914 level instead of ~65-68%, so its U-shape becomes much shallower, and its long-run and 1910-1950 decompositions (Tables VII-VIII) and the negative German q that underpins the European wash-out (F4) are affected. Because WID raises the UK pre-1914 level (private 767% vs 703% in 1910) and France is within ~5%, the unweighted UK-France-Germany 1910 average stays ~640-660% and C2's 600-700% European benchmark survives.

Evidence. ABS replication file Wealth_totals_Kaiserreich.xlsx, sheet 'Comparison to PikettyZucman' (PZ vs ABS private wealth, bn marks): 1910 277.2 vs 217.1 (608% vs 476% of NI); 1913 324.7 vs 248.4 (623% vs 477%). WID bulk file (15 Oct 2026), adjudicator check (logs/wid_check.txt): Germany private 1910 456% (PZ 608%), national 1910 484% (PZ 637%), national 1870 485% (PZ 745%); UK private 1910 767% (PZ 703%), 2010 462% (PZ 522%); France within ~5%. With ABS/WID, Germany's 2010 ratio is ~90% of its pre-1914 level instead of ~65-68%; the unweighted UK-France-Germany 1910 average stays ~640-660%.

Why this severity. Germany's U-shape becomes much shallower and its long-run and 1910-1950 decompositions (Tables VII-VIII) are affected, but France is unchanged and the UK is revised up, so the European level stays within 600-700% and C2 survives.

The authors' response. App. D.4.1 cross-checks the 1909-1911 level against contemporary estimates (Steinmann-Bucher, Helfferich) and acknowledges known problems with Hoffmann's series; no response to ABS located.

Why it matters, in detail. The main text (p.1275) rules out measurement as an explanation of Germany's low wealth-income ratio, which underlies its German 'missing wealth' and its 'stakeholder' reading of low Tobin's Q, while App. D.1 names two candidate downward biases: building-land values that are 'largely based on estimates' and that the Bundesbank suggests 'may be downwards biased', and 'very tentative' other-equity values that published financial accounts may 'substantially under-estimate'. Both would understate German wealth, so correcting them would raise Germany's level - Albers, Bartels & Schularick (2022) put the 2010 level ~29 points (~7%) above PZ - narrowing the German puzzle and, if anything, strengthening C1. The effect is confined to a secondary result, and the main text itself later allows that German statisticians 'might ... underestimate the current stock of wealth' (p.1290).

Evidence. QJE p.1275 and fn 27 vs Online App. pp.71-72 (fns 174-179); Deutsche Bundesbank, Monthly Report, January 2008, p.41 (quoted in App. fn 176). Later evidence: Albers, Bartels & Schularick (2022) raise German household wealth well above the official balance sheets (about 120% of GDP for recent years; their 2010 level is ~29 points, ~7%, above PZ per Auditor B).

Why this severity. Concerns a secondary result (the German level puzzle, German 'missing wealth' and the 'stakeholder' reading of low Tobin's Q); the main text itself later allows that German statisticians 'might ... underestimate the current stock of wealth' (p.1290), and correcting it would if anything strengthen C1.

The authors' response. App. D.1: 'We do not believe that the bias is large, as survey data ... also indicate that the real estate capitalization is relatively low in Germany.'

Why it matters, in detail. Five UK spreadsheet problems, none of which changes a conclusion. (a) Undisclosed switching between national-wealth variants: Tables V and VIII use Method 1 (1970 314%, 1950 208%), Tables II, VI, VII and App. A99/A101 the corrected Method 2 (365%, 241%), and A107/Fig. VII compares Method-1 observed values with a Method-2 prediction; with Method 2, Table VIII's UK 1910-1950 split (insufficient saving/war destruction/capital losses) becomes 50/4/46 instead of 46/4/50 - the largest effect, about four points - while Table V's UK shares are unchanged (42/58). (b) Table VII prints the 1855 value (694%) as the UK 1910 endpoint instead of 719%; growth rates use 719%, but the additive UK.4c 1870-1910 capital-gain residual is understated by 24 points of NI. (c) The Method-2 saving rate is shifted one year for 1855-1918; correcting it moves the UK 1870-2010 saving share from 83.4% to 82.6%. (d) The UK column of App. Table A48 (capital shares, Fig. XII) is lagged one year (1975 printed 17.1% vs 13.7% same-year), which also enters F6. (e) The 1920 Irish boundary change (3.2% of NI) sits in the 1910-1950 capital-loss residual, about -0.08 pp/yr of the -1.9%/yr q (~13 of Table VIII's -256 points).

Evidence. UK.xlsx TableUK5b, TableUK4c, TableUK11a; Tables.xls; App. Tables A1, A48, A99-A101, A107. Adjudicator: Table VIII UK with Method 2 gives insufficient saving / war / capital losses = 50/4/46% (published 46/4/50%); Table V UK shares unchanged (42/58 either way); 1920 boundary effect ~ -0.08 pp/yr of the -1.9%/yr 1910-1950 q (~13 points of Table VIII's -256). Codex: correcting (c) moves the UK 1870-2010 saving share 83.4% -> 82.6%.

Why this severity. None changes a conclusion; the largest effect shifts the UK 1910-1950 attribution by about four points.

The authors' response. Both UK variants and the boundary treatment are documented in App. F (pp.106-107, fn 287, fn 389); no response to the errors located.

Why it matters, in detail. Presentation and transcription slips that leave every reproduced total and conclusion unchanged: fn 53 reverses U.S. gross foreign positions (2010 data: assets 102%, liabilities 128%, net -25% of NI, vs ~120%/~100% in the text); App. K.1.2 misprints the growth factor ((Y_t+n - Y_t)/Y_t for Y_t+n/Y_t) and reverses the inequality for q > 0; Table II's German 2010 housing/other capital (241%/136%) differs from the country sheet and A16 (235.7%/140.9%) with the total unchanged; the abstract's ranges are exceeded at the top (1970 private ratios 225-342%, four countries above 300%; 2010 410-676%, Italy above 600%); and several text-table mismatches (French population growth 0.5% vs 0.6% in Table III; disposable income '75-80%' of NI vs 91% for the U.S.; fn 15's '97-103%' vs ~95% for Australia; Fig. XIII's title 'Average Return on Private Wealth' vs the text's return to domestic capital, alpha/k). App. fn 42's statement that most countries spend 'about 2-3%' of GDP on R&D (World Bank 1996-2010: UK 1.6%, Canada 1.9%, Australia 1.9%, Italy 1.1%) matters only through F1's uniform R&D allowance.

Evidence. USA.xlsx TableUS6b (2010 row: FA 1.0219, FL 1.2761, NFA -0.2543); Online App. pp.145-146; Tables.xls 'Table 7' cell D12 = 2.4069 vs Germany.xls TableDE6c 2010 = 2.3573 and A16; App. Tables A5, A40, A46; Tables.xls 'Table 2'; World Bank API indicator GB.XPD.RSDV.GD.ZS.

Why this severity. Presentation and transcription slips; none affects the reproduced totals or conclusions (the R&D-intensity slip matters only through F1).

Why it matters, in detail. Table VIII splits the 1910-1950 fall in beta into insufficient saving, war destruction and capital losses, with capital losses computed as the residual. The war-destruction inputs rest on rough, disclosed assumptions - French destruction is Piketty's (2010) estimate 'simply divided ... by 2' to separate out foreign-asset losses, and German WWII destruction is set at 25% of the 1939 stock of 'other domestic capital' with assumed annual timing - so any error in them moves the war-destruction and capital-loss shares one-for-one in opposite directions (published insufficient saving/war/capital losses: France 38/27/35%, Germany 31/29/40%). The size and direction of any such error were not quantified; the total 1910-1950 fall, which reproduces within 1-3 points from the annual series, and the conclusion that all three factors mattered (C6) are unaffected, and the authors judge the consequences 'minimal'.

Evidence. Online App. pp.86, 97; Table VIII reproduces within 1-3 points from the annual series (logs/reproduce_core.txt), with capital losses as the residual.

Why this severity. Shifts weight between two components of the 1910-1950 fall; the total fall and the conclusion that all three factors mattered are unaffected.

The authors' response. App. fn 248: 'The consequences for our decomposition results are minimal.'

Why it matters, in detail. C2's claim that today's ratios are returning to 'the high values observed in Europe in the eighteenth and nineteenth centuries (600-700%)', and the 'rich countries' framing, generalise from the UK, France and Germany. Waldenstrom (2017, JEH) finds Sweden's 19th-century ratio only about one-third to one-half of theirs, on a par with the U.S.; Swedish private wealth was ~200% of NI in the 1980s-90s and ~300% in the 2000s versus 400-600% elsewhere, with the rise coming later (late 1990s); before 1910 capital gains were the most important source of Swedish real private wealth growth, and public pension wealth played a larger role. This narrows the scope of C2 and of the saving-driven account - the 600-700% historical level and saving-led accumulation are not general European facts - rather than contradicting PZ's results for their own sample, and Sweden also shows a later rise.

Evidence. Waldenstrom, D. (2017), 'Wealth-Income Ratios in a Small, Developing Economy: Sweden, 1810-2014', Journal of Economic History 77(1): 285-313 (author version, IFN, PDF pp.4, 9, 12), https://doi.org/10.1017/S0022050717000080.

Why this severity. A scope qualification: PZ's claims are made for their sample, and Sweden also shows a later rise, but the 600-700% historical level and the saving-driven account are not general European facts.

Why it matters, in detail. Under market valuation, changes in house prices relative to rents enter the 1970-2010 change in beta; revaluing 2010 household housing at the base-year net rental yield (a rough version of Bonnet, Bono, Chapelle & Wasmer 2014, using App. Tables A16 and A144) strips that component out. The rise in national wealth/NI (in multiples of NI) then falls by more than half in France (1979-2010: 2.30 -> 0.95) and Canada (1.28 -> 0.57) and disappears in Germany (1990-2010: 0.58 -> 0.01), is roughly unchanged in the UK (1.63 -> 1.54), and increases in the U.S. (0.27 -> 0.56), Japan (2.56 -> 3.09), Italy (1990-2010: 1.99 -> 3.29) and Australia (1.94 -> 2.05). The direction is country-specific: the rent-valuation critique qualifies C1 and C7 for France, Canada and Germany but cannot overturn the eight-country rise. The estimates are rough because the A144 yields are noisy (France 3.6% in 1979, 4.4% in 2000, 2.3% in 2010), and market valuation is a disclosed, standard choice.

Evidence. Auditor A's computation from App. Tables A16 (housing/NI), A144 (net housing yields) and A1; adjudicator re-derived the France case (housing 3.71 x 2.30%/3.61% = 2.36 of NI). Rise in national wealth/NI, observed -> constant yield: France 2.30 -> 0.95, Canada 1.28 -> 0.57, Germany (1990-) 0.58 -> 0.01, UK 1.63 -> 1.54, U.S. 0.27 -> 0.56, Japan 2.56 -> 3.09, Italy (1990-) 1.99 -> 3.29, Australia 1.94 -> 2.05. A144 yields are noisy (France 3.6% in 1979, 4.4% in 2000, 2.3% in 2010).

Why this severity. Market valuation is a disclosed, standard choice; the alternative changes the picture materially only for France, Canada and Germany and does not overturn the eight-country rise.

The authors' response. fn 17: excluding housing 'strikes us as inappropriate'; Table II discloses housing's role; p.1289 notes possible overshooting and 'abnormally high real estate prices in 2010'. Piketty (2015, AER P&P) stresses real-estate price movements.