Kevin M. Murphy, Emmanuel Saez · In T. Church, C. Miller and J. B. Taylor (eds.), Inequality and Economic Policy: Essays in Memory of Gary Becker, Stanford: Hoover Institution Press, 2015, ch. 6, pp. 111-146

Income and Wealth in America

14 problems found: 1 major · 4 moderate · 9 minor

What the paper claims

The US top-10% pre-tax (fiscal) income share follows a century-long U: about 45% before WWII, a big fall during the war, about 33% in the following decades, then a rise to 50% (a 17-point increase), so that recent concentration is 'as high as it's been over the last hundred years'.
The rise is concentrated at the very top: the top 1% share went from slightly below 10% to above 20% and accounts for most of the 17 points, with some gain for P95-99 and little for P90-95.
Top-income concentration fell in all rich countries in the first half of the 20th century but rebounded since the 1970s only in some (much less in continental Europe and Japan), so globalization cannot be the sole explanation.
Surges in pre-tax top income shares are strongly correlated with cuts in top marginal tax rates, across 18 OECD countries ('1960-4 to 2005-9'; the US and UK had the highest rates and cut them most) and in the US time series (top rates 'in excess of 70 percent from 1933 to 1981'); supply-side and rent-extraction readings are 'very hard to distinguish'.
Real income per adult of the top 1% and the bottom 99% each grew roughly fourfold over 1913-2012 but with opposite timing (bottom-99% growth strong when top tax rates were high, top-1% growth after the 1970s), and only the top 1% recovered quickly after the Great Recession.
Wealth (Saez-Zucman 2014 capitalized income tax data): the bottom-90% share rose to a peak of 36% in the mid-1980s and has fallen to the low 20s (an 'erosion of middle-class wealth'); the top-1% share rose from the low 20s in the late 1970s to 42%; bottom-90% real average wealth in 2012 is back at its 1980s level while top-1% wealth resumed its rise.
The bottom-90% wealth erosion 'has to be explained by a very low savings rate' while the top 1% saves at very high rates; combined with surging top incomes this created new (younger, labor-income-rich) fortunes; low middle-class saving reflects stagnating incomes, easier credit and 'present-biased consumption behavior', motivating default-based saving policies.
Murphy: the college premium fell in the 1970s and roughly tripled after 1980; wage inequality rose throughout the distribution from 1970-72 to 2010-12 while women gained relative to men; a Katz-Murphy supply-and-demand model fits the premium; better (especially K-12) human-capital formation is the remedy.
Supporting statements in the policy section and Q&A: over 50% of Harvard/Stanford students come from top-10% families and over 20% from top-1% families; realized gains are included and top-1% charitable giving has stayed around 4% of reported income, so the rise is not a tax-avoidance artifact; munis cannot matter much; US taxes reached 35% of GDP by mid-century; student loans are a big reason for middle-class wealth erosion.

Every problem found

1majorlikely · previously known · methodology

The chapter says the richest 1% went from owning under a quarter of US household wealth in the late 1970s to 42% in 2012. That figure came from a then-new method that is very sensitive to how much interest rich people earn on their savings; when this is measured better, including by the authors themselves, the rise is roughly a fifth to two-fifths smaller. Wealth did become much more concentrated, but by less than the chapter's figure implies.

Technical detail and evidence

Why it matters, in detail. With within-class yields assumed uniform, fixed-income wealth = interest / average yield; if top yields exceed the average (BHH; SZ's own revised series uses ~1.4x for the top 1% after 2008; SZZ more), top fixed-income wealth and its post-2008 growth are overstated, increasingly so as the capitalization factor rises (24 -> 97). Corrections lower the 2012 level by 2.7 points (authors' revision) to ~8 points (SZZ; BHH 2011: -6.6) and the 1978-2012 rise by 20-43%. The bottom-90% decline is more robust (-10 to -12 points in SZ vintages; about -8 in SCF/DFA household data over 1989-2013). Remaining disagreement is concentrated in the top 0.01%.

Evidence. Adjudicator scripts a03/a02/a12 on the authors' own workbooks. Top-1% share 1978 -> 2012: chapter vintage 22.9 -> 41.8 (+18.9); SZ revised Sept-2020, tax units 23.9 -> 39.0 (+15.1, 20% smaller); WID 2026 tax units 23.9 -> 39.1 (+15.2), equal-split adults 21.8 -> 36.6; BPEA-2019 'capitalization revised' 25.5 -> 36.3 (+10.8, 43% smaller); Smith-Zidar-Zwick Apr-2020, tax units 22.6 -> 33.8 (+11.3, 40% smaller). 1989 -> 2012: chapter +14.0 vs SZ revised +8.7, DFA adjusted to tax units +8.0, SZZ +9.0 (36-43% smaller). In the SZ WP file itself: SCF top 1% 30.1% (1989) -> 35.8% (2013); estate multiplier 19-23% over 1976-2000. Bricker, Henriques & Hansen (2018, FEDS 2018-024, abstract): 2011 share 40.5% -> 33.9% with heterogeneous interest returns (33.9-37% across variants). SZ (2016) appendix Table B40/B41b higher-yield variant: 39.97% in 2012. Kopczuk (2015, JEP, p. 61): capitalization factor 24 -> 96.6 over 2000-2012. Counterpoint: SZ's harmonized SCF (tax units, Forbes 400, FA totals) gives 41-43% in 2013, and harmonized DFA and SZ-revised give 38.1% and 37.9% in 2018, so the level is less fragile than the trend.

Why this severity. Every series shows a large rise, so the direction of C6 survives, but the headline rise since the late 1970s is 20% smaller in the authors' own current series and about 40% smaller in heterogeneous-return estimates, and the 1989-2012 rise is 36-43% smaller across SZ-revised, DFA and SZZ, beyond the rubric's quarter-of-the-trend threshold.

The authors' response. Saez & Zucman added a top interest-rate premium (BPEA 2019; NBER w27921, 2020: about 1.4 times the average for the top 1% since 2008), argue that SZZ's top yields exceed those observed in any data with both income and wealth, and show that harmonized DFA and their revised series both give about 38% in 2018 (JEP 2020 data file: 38.1% and 37.9%). The chapter itself predates these exchanges.

2moderateconfirmed · previously known · methodology

The chapter's income numbers come from what people report on tax returns. When the authors later measured all national income, the rise in the top 10%'s share since the 1970s was still large but about a fifth to a third smaller, and whether today's concentration matches the 1920s-30s depends on how married couples are counted. Part of the late-1980s jump reflects business income moving onto personal tax returns after the 1986 tax reform rather than people becoming richer.

Technical detail and evidence

Why it matters, in detail. Fiscal income omits roughly 40% of national income (employer benefits, imputed rent, retained earnings, etc.) and responds to tax-base reclassification (post-1986 C-to-S conversions move retained profits onto Form 1040). On the same tax-unit basis, pre-tax national income lowers the 1970s-2012 top-10% rise from 17.0 to 13.2 points (-22%) and the top-1% rise from 13.3 to 9.8 (-26%); equal-split adults lower the top-10% rise to 11.5 (-32%) and put 2012 below the 1925-40 average (45.6 vs 47.6). The 1986-88 jump is 23-36% of the 1980-2012 fiscal rise; DINA neutralizes most of it by imputing retained earnings. The same reporting response contaminates part of the US time-series link between top rates and top shares (C4).

Evidence. a01 (TabFig2012prel): top 10% 1970s avg 33.4 -> 2012 50.4 (+17.0); top 1% excl. gains 9.13 (1986) -> 13.17 (1988), +4.0 = 36% of the 1980-2012 rise of +11.2; incl. gains 12.67 (1985) -> 15.49 (1988) = 23% of +12.4. a02 (WID sptinct992 / sptincj992): tax units 1925-40 avg 50.2, 1928 50.8, 1934 51.4, 1970s avg 37.1, 2012 50.3; equal-split 47.6, 48.2, 49.05, 34.1, 45.6; top 1% tax units 1970s 11.6 -> 2012 21.4 (+9.8 vs +13.3 fiscal). Auten & Splinter (2024, JPE 132(7), p. 2182): top-1% pre-tax national-income share +4.4 points over 1979-2019, 'about half the increase in fiscal income' (disputed by PSZ). Burkhauser et al. (2012, REStat): the 1986-88 jump is primarily reporting.

Why this severity. The U-shape and the dominance of the top 1% survive in every concept; what changes is the size of the rise (about a fifth to a third smaller in the authors' own DINA, more under the disputed Auten-Splinter adjustments) and the century-high comparison, which roughly holds for tax units but not for equal-split adults. The chapter discloses its concept, so the qualification is noticeable rather than transformative.

The authors' response. In the chapter (p. 139) Saez notes that realized gains are included and that income shifts between corporate and individual form after tax episodes. Piketty & Saez (2003, pp. 3-4, 31) attribute the 1986-88 jump partly to fiscal manipulation, and PSZ (2018) DINA adds retained earnings; their 2006 reply to Reynolds notes that the gains-inclusive series and the top-1% wage share also doubled (as summarized in the critique corpus).

3moderateconfirmed · newly identified here · overclaiming

The chapter's saving rates were not measured from household budgets; they were worked out backwards from the wealth estimates, so they cannot confirm those estimates, and any error in the wealth figures passes straight into them. The authors' own paper warns about this, but the chapter does not, and it blames short-sighted behavior when ordinary explanations such as easier credit, rising house values or growing pension entitlements fit the same numbers.

Technical detail and evidence

Why it matters, in detail. S_p = W_p(t+1)/(1+q_p) - W_p(t) absorbs (i) errors in the capitalized stocks (F1 passes almost one-for-one into the top-1% rate), (ii) within-class capital-gain heterogeneity, (iii) mobility across fractiles. With W/Y of about 10 for the top 1%, a 1-point-a-year price misstatement shifts its rate by about 9 points. Negative bottom-90% saving in 1998-2008 is equally consistent with rational wealth effects of the housing boom, relaxed borrowing constraints or Social Security crowd-out, so present bias is not required by the data.

Evidence. WP 20625 pp. 28-31 (text checked). Auditor B rebuilt the annual rates (1986-2011 mean: bottom 90% -0.25%, top 1% 36.1%; max deviation 0.014 point) and showed that a 1% change in the gross price factor moves them to -2.1/+1.6% and 27.1/45.3%; these magnitudes match the groups' wealth/income ratios (about 2 and 10). Rough adjudicator approximation: the BPEA-revised top-1% path (25.8 -> 36.3, 1986-2012) instead of the chapter's (25.1 -> 41.8) implies about 0.65 point a year slower relative wealth growth, i.e. a top-1% synthetic rate about 6 points lower (still high). Alternatives recorded in the chapter: relaxed credit constraints and human wealth (Murphy, pp. 133-134); Social Security/Medicare wealth crowding out saving (questioner, p. 143). External: Fagereng, Holm, Moll & Natvik (2019, NBER w26588, Norway): saving rates net of capital gains are roughly flat across wealth, gross rates rise with wealth ('saving by holding').

Why this severity. The accounting decomposition is internally consistent and reproduces, and the qualitative pattern (very low bottom-90% and high top-1% residual saving) survives plausible revisions, but the behavioral and policy conclusions are not identified by it; C7 is secondary to the measured income and wealth trends.

The authors' response. Within the chapter Saez introduces the rates as 'consistent with the patterns that we find' and 'at least that's what you need to explain how their wealth has increased' (p. 129); the source WP states the synthetic definition and its limits.

4moderatelikely · previously known · methodology

Tax records from before the Second World War cover only a small, rich slice of the population, so turning them into national shares requires judgment calls. A 2022 re-analysis of the same records finds that the top 10% held closer to 39% of income before the war, not 45%, and that most of the decline happened in the Depression rather than during the war. This does not affect the recent rise, but it makes the historical 'U' shallower.

Technical detail and evidence

Why it matters, in detail. Upward-biased numerator harmonization and a low pre-1944 denominator inflate pre-war shares; the biases fade as filing becomes near-universal in WWII, exaggerating the 1940-44 drop. GMMS cut the 1925-40 top-10% level by 6.5 points and the 1939-44 fall by 7.8 points (60%); the depth of the U (pre-war average minus the 1950s) shrinks from about 12 to about 8 points. Post-1950 values barely change (1950-60 average 33.7 vs 31.1).

Evidence. a10 (GMMS DataAppendix-6-7-21.xlsx, sheet TopIncomeShares; the PS column matches TabFig2012prel): top 10% 1925-40 average PS 45.7% vs GMMS 39.2%; 1928 49.3 vs 42.5; 1939 45.5 vs 36.5; 1944 32.5 vs 31.2. Fall 1939-44: 13.0 vs 5.2 points (60% smaller); 1928-44: 16.8 vs 11.2 (33% smaller). Top 1%: 1928 23.9 vs 19.5; 1939-44 fall 4.9 vs 1.7. No reply by Piketty or Saez among the 29 works citing GMMS in OpenAlex.

Why this severity. It changes the left arm of the U (pre-war level about 39% rather than 45%) and the size and timing of the wartime compression that underpins the visual link with top tax rates in Fig. 6.9, but leaves the post-1970s rise untouched; under GMMS 2012 is even further above pre-war levels.

5moderateplausible · previously known · methodology

The wealth figures count homes, savings and shares, but not the value of future Social Security pensions, which are a large part of what ordinary families can count on. Some economists find that counting them makes the rise in wealth concentration much smaller, while others dispute how to value them; the chapter does not discuss this even when asked.

Technical detail and evidence

Why it matters, in detail. SSW is a long-duration, non-marketable claim concentrated below the top decile whose present value rises when real rates fall, in parallel with the revaluation of the rich's assets that marketable series capture. Including it cuts the 1989-2019 rise in the SCF top-1% share from 6.4 to about 1.9 points (CMS; the NPV variant gives -0.2). It leaves the marketable-wealth facts intact but weakens their welfare interpretation and offers a crowd-out explanation for low measured saving.

Evidence. Catherine, Miller & Sarin (2025, Journal of Finance 80(3); corpus us-wealth-taxes-09): Social Security wealth $7.2tn (1989) -> $40.6tn (2019); SCF top-1% share +6.4 points for marketable wealth vs +1.9 including SSW (risk-adjusted), robustness range -0.2 to +3.7; Sabelhaus & Henriques Volz (2020) find SSW lowers inequality without reversing the trend (both as summarized in the corpus, not re-read here). Chapter pp. 127, 131, 143-144; Murphy (p. 133) on human wealth.

Why this severity. Marketable wealth is the standard concept and the chapter's statements are correct for it, but the 'erosion of middle-class wealth' reading and the low-saving story depend on excluding the largest asset of the middle class, whose effect on trends is large but disputed and discount-rate sensitive.

The authors' response. Saez & Zucman (2020, JEP, p. 8) argue that promises of future government benefits are excluded from wealth by the Fed, by themselves and by other countries, for conceptual consistency (as summarized in the corpus).

6minorconfirmed · newly identified here · internal inconsistency

The chart of saving rates shows the richest groups saving a few points more than the authors' own data tables say, by up to 5 points in the 1980s. It does not change the picture.

Technical detail and evidence

Why it matters, in detail. A systematic 0-5 point (about 5-10% proportional) upward offset in top-group decade means, largest in the 1980s (50.9 vs 45.9), probably from a pre-release data iteration; it slightly exaggerates the 1980s peak in top-1% saving.

Evidence. a04 (vector extraction of chapter p. 129 vs the Oct-2014 Table B33): top 1% plotted/table 1917-19 35.0/32.9; 1970s 38.6/35.8; 1980s 50.9/45.9; 1990s 32.6/31.0; 2000s 35.0/34.6; 2010-12 38.6/38.6. Top 10-1%: +0.1 to +2.4 points. Bottom 90%: within 0.4 point. Vector data of WP 20625 (NBER PDF p. 61) also give about 51% for the 1980s top 1%.

Why this severity. At most 5 points in one decade; the pattern (top saving far above bottom-90% saving) is unchanged.

7minorconfirmed · newly identified here · misrepresentation of source

Answering a question about tax-free municipal bonds, Saez overstated how much of them the top 1% owned in his own data (about 65%, not 75%). His conclusion, that these bonds are too small to distort the figures, is still correct.

Technical detail and evidence

Why it matters, in detail. The gap to 100% is 35 points, not 25, so the extreme-allocation bound is 1.07 points (1986) rather than about 0.77, and at most 0.6 point before 1982. For the income series, tax-exempt interest is excluded throughout, and rising muni holdings in the 1980s would, if anything, bias measured top-income growth down.

Evidence. SZ Oct-2014 AppendixTables(Distributions), sheet 'wealth_detail' (group muni holdings as a share of total household wealth divided by aggregate munis/wealth): top-1% share of munis 65.0% (1962-86), 64.6% (1987), 63.2% (1988); munis 1.2-1.6% of household wealth in 1962-81, 3.1% (1986), 3.5% (1988). Auditor B's bound (0.77 point) assumed the stated 75%.

Why this severity. The claim that munis cannot materially affect the series is confirmed; only the stated share and the implied bound (understated by about 40%) are wrong.

8minorconfirmed · newly identified here · documentation gap

The chart comparing countries says it covers 1960-64 to 2005-09, but for most countries the dates differ. The link it shows between cuts in top tax rates and rising top incomes is unaffected.

Technical detail and evidence

Why it matters, in detail. The caption implies a common 45-year window; actual windows differ by up to two decades. Correct windows leave the slope within 0.44-0.52.

Evidence. a05: own translation of PSS international.do reproduces 0.470 (s.e. 0.114), n = 18; exact 1960-64/2005-09 windows only for Canada, Finland, New Zealand and the UK. Robustness: excluding the US and UK 0.492 (0.180); excluding all six English-speaking countries 0.440 (0.183); only true 1960-64 starts 0.516 (0.121); excluding Portugal 0.458 (0.120). PSS NBER WP 17616 lists the late starts (Denmark 1980, Ireland 1975, Italy 1974, Portugal 1976, Spain 1981) and Switzerland's 1995 end.

Why this severity. A labeling problem only; the elasticity is unchanged and robust to every subsample examined, including dropping all English-speaking countries.

The authors' response. PSS (2014; NBER WP 17616) disclose the country-specific windows in their own text and figure notes.

9minorconfirmed · newly identified here · external inconsistency

Saez said US governments were collecting 35% of the economy in taxes by mid-century; the real figure was closer to 22-24%. This aside does not affect the chapter's inequality figures.

Technical detail and evidence

Why it matters, in detail. It overstates the US tax/GDP ratio by about 11-14 points of GDP (about 10 points on a national-income base).

Evidence. a07 (BEA NIPA Table 3.1, 2026 vintage): taxes + contributions 21.7% of GDP (1950), 22.4% (1955), 23.8% (1960); 1945-75 average 23.4% of GDP and 26.6% of national income; maximum total current receipts 27.8% of GDP before 1976 (1974) and 30.6% in 2000. Auditor B (OMB Historical Table 14.1): 20.2% (FY1950), 22.3% (1955), 24.4% (1960).

Why this severity. An off-the-cuff historical aside in the Q&A that feeds no income or wealth estimate; it does not affect any core finding, so the rubric places it at minor.

10minorconfirmed · newly identified here · overclaiming

Student debt grew fast, but it is far too small to explain much of the fall in middle-class wealth: at most about a tenth of it.

Technical detail and evidence

Why it matters, in detail. Upper bound of about 10% of the share decline and about 14% of the 2006-12 dollar decline, assuming all student debt is held by the bottom 90% and none existed in 1986.

Evidence. a08: student loans outstanding (FRED SLOAS) $1.03tn in 2012 = 1.9% of SZ household wealth; with no student debt the 2012 bottom-90% share would be 24.2% instead of 22.8%, i.e. at most 1.4 of the 13.6-point fall since 1986 (10%). 2006-12: +$0.53tn of student debt vs a $3.7tn fall in bottom-90% nominal wealth (14%). WP 20625 pp. 3 and 25 list student loans among the debts behind the decline.

Why this severity. A Q&A remark; the chapter's main account of the decline (low saving, mortgage debt) is unaffected.

11minorconfirmed · newly identified here · internal inconsistency

The dates given for the era of very high top tax rates are slightly off; the chart itself is right.

Technical detail and evidence

Why it matters, in detail. Start date off by three years (63% in 1933-35); the strict inequality fails for 1965-67 and 1971-81.

Evidence. a06 (PSS usdata.dta, topmtr): 1932-35 63%; 1936-39 79%; 1940-63 77-94%; 1964 77%; 1965-67 70%; 1968 75.25%; 1969 77%; 1970 71.75%; 1971-80 70%; 1981 69.13%; 1982 50%. Auditor B: IRS histab23 gives 63% (1933-35), 79% (1936), 69.125% (1981).

Why this severity. A slip in dates and in the strict inequality; the high-rate era (at least 70% from 1936 to 1980) and the plotted series are correct.

12minorconfirmed · newly identified here · internal inconsistency

Murphy's own chart shows women's wages growing about 30 percentage points more than men's, more than the 'almost 20' he mentions.

Technical detail and evidence

Why it matters, in detail. About 10 log points of understatement at the median relative to the plotted 1970-72 to 2010-12 changes.

Evidence. a09 (vector extraction, chapter p. 114): gap 31.3 (P20), 29.9 (P50), 25.4 (P80), 27.0 (P90); range 24.5-42.4, mean 30.5 over 36 plotted percentiles. Auditor B: 31.9 / 29.9 / 24.8 at P20 / P50 / P80. Men's growth is about 0.21 at the median and 0.49 at P90, matching the text ('about 0.2', 'almost 0.5').

Why this severity. The understatement runs against the speaker's own point (convergence is larger than stated); no conclusion changes.

13minorconfirmed · newly identified here · arithmetic or transcription

A handful of typos in captions, dates and references; none changes the findings.

Technical detail and evidence

Why it matters, in detail. None on estimates; the omitted point (about 4.5% in B33) is in line with neighboring decades.

Evidence. Chapter pp. 115, 127-129, 146; Crossref and OpenAlex for DOI 10.1162/00335530360535135 give pp. 1-41; a04 shows the bottom-90% polyline has 10 of 11 decades.

Why this severity. Presentation errors that can be corrected from the chapter itself; no effect on any conclusion.

14minorlikely · newly identified here · documentation gap

The chapter says over a fifth of Harvard and Stanford students come from the richest 1% of families, citing a paper that was never published. When the research did come out, the figure was about 15%, still striking but lower.

Technical detail and evidence

Why it matters, in detail. The top-1% share is overstated by about 5-6 points (35-40% relative) against the published estimates; preliminary 2014 figures may have used other cohorts or rank definitions, which cannot be checked because the cited paper never appeared.

Evidence. Crossref: the only Chetty item in AER 105(5) (May 2015) is 'Behavioral Economics and Public Policy: A Pragmatic Perspective' (pp. 1-33); no title match in Crossref or OpenAlex. Opportunity Insights mrc_table2.csv: Harvard top-10% parents 57.0%, top-1% 15.4%; Stanford 55.8% and 14.5%; Princeton 20.1%, Duke 19.2%, Brown 18.9%, Yale 17.6%.

Why this severity. A supporting fact in a policy argument; the qualitative point (very strong over-representation of top-income families) is confirmed by the published data.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

In the authors' own DINA series the 2012 top-10% share is below its 1920s-1940 levels, so the 'as high as it's been over the last hundred years' comparison reverses. — Overstated: it holds only for equal-split adults. On the like-for-like tax-unit basis the chapter uses, DINA gives 50.3% in 2012 vs a 1925-40 average of 50.2% (peaks 50.8% in 1928 and 51.4% in 1934), so the level claim roughly survives; Geloso et al.'s pre-war corrections would strengthen it. The robust part of the point, a 22-32% smaller rise, is retained in F2.
The p. 124 remark that pre-tax shares have 'no mechanical relationship' with tax rates ignores the reporting/avoidance elasticity that PSS model. — 'Mechanical' denotes accounting effects (taxes do not reduce pre-tax income by construction), which is correct; behavioral avoidance is a separate channel, which the chapter takes up in the Q&A (pp. 138-141). The tax-reform reporting break is retained in F2.
The saving-rate argument is circular (category 'circularity'). — The chapter presents the rates as inferred from the wealth stocks ('consistent with the patterns that we find'; 'at least that's what you need'), and the decomposition is transparent accounting, so it does not assume its conclusion. The real problem, the behavioral and policy over-interpretation, is retained as overclaiming in F4.
'Governments in advanced economies take 35 percent to 50 percent of market incomes' (p. 145) is an overstatement. — Not adjudicated as an error: the sentence describes advanced economies in general, not the US. The US sits below the range (taxes plus contributions about 28-30% of national income in 2012-14, NIPA); other countries were not recomputed here. Only the mid-century US claim on p. 137 is clearly wrong (F10).
The '35 percent of GDP' misstatement is a moderate issue. — The error is confirmed, but the severity is lowered to minor: it is a Q&A aside that feeds no estimate, and Auditor B itself lists no affected core finding.
A two-period log-utility example (income 100 then 120, zero interest) disproves Saez's statement that the borrowing story requires present bias. — Partly rejected. The example assumes rising income, whereas Saez's story is about stagnating incomes, so it does not by itself rebut the stated mechanism. The conclusion that present bias is not required is retained in F4 on other grounds: rational wealth effects of the housing boom, relaxed credit constraints (Murphy, pp. 133-134) and Social Security crowd-out.
The elite-college statistic is an unresolved reproduction gap rather than a finding, because the reference population of the 2015 source is unknown. — Upgraded to a minor finding (F9): the cited AER P&P paper verifiably never appeared (Crossref), and the same team's published estimates give about 15% for Harvard and Stanford. Confidence is set to 'likely' to reflect the unknown preliminary definitions.
No separate finding is warranted for the exclusion of Social Security/transfer wealth because human and transfer wealth are openly discussed in the chapter. — Not accepted. The questioner raises Social Security wealth (p. 143), but Saez's answer concerns only future income series, and the exclusion matters for the 'erosion of middle-class wealth' reading. Retained as a moderate, plausible issue (F5).

Audit notes

What reproduced. Reproduced with the adjudicator's own scripts (final/scripts/a01-a12, run_all.sh) from the authors' files and official data: all of Saez's income-share numbers and thresholds; the 0.470 (0.114) cross-country elasticity and new robustness checks; the US top-rate history; Fig. 6.10 growth factors to 2010; the Oct-2014 wealth shares and real means; the later-vintage and cross-method wealth comparisons; Fig. 6.13 against Table B33 (vector extraction); the muni holdings; NIPA tax ratios; the student-loan bound; the Chetty et al. college shares; the Geloso et al. pre-war arithmetic; and Murphy's Figs. 6.1, 6.2 and 6.4 as plotted. Not reproduced: Murphy's underlying Katz-Murphy CPS updates (unpublished); the exact 2012 per-adult extension behind Fig. 6.10 (Auditor B's extension gives x4.84 / x3.66 for 1913-2012); the 2014-15 preliminary Chetty et al. statistic (the paper never appeared); Duquette's charitable-giving microdata; any re-estimation of capitalized wealth with heterogeneous returns, which needs confidential IRS data.

Limits of this audit. Confidential IRS microdata were not available, so no capitalized-wealth series with heterogeneous returns was re-estimated; the BHH, SZZ, DFA and Catherine-Miller-Sarin magnitudes come from published files, papers or the critique corpus. The Geloso et al. corrections were reproduced arithmetically from their workbook, not rebuilt from SOI tables. Murphy's figures could only be read from the chapter's vector graphics. The 2014 conference slides and any preliminary Chetty et al. draft were not found, so differences between the spoken talk and the edited transcript cannot be assessed. Murphy's financial-aid remark ('about 40 percent of your savings') was not adjudicated: the federal formula assesses parental assets at up to 5.64% a year, so 40% is reachable only on a multi-year, cumulative reading. Auditor A read Duquette's (2018) charitable-giving rates from his figures; they were not recomputed. The RePEc record listed as the version URL points to a different chapter (a cataloguing error, not a defect of the work). Some author responses are cited as summarized in the critique corpus. Auditor A's status file still read 'running' when this adjudication started, but its report and findings were complete.