Thomas Piketty, Emmanuel Saez, Gabriel Zucman · WIL Technical Note 2024/04 (Sept 2024), revised August 2026; R&R Journal of Political Economy

Income Inequality in the United States: A Comment

17 problems found: 1 major · 5 moderate · 11 minor

What the paper claims

Eight methodological differences between PSZ and Auten-Splinter (AS) explain about three-quarters of the PSZ-AS gap in the 2019 top 1% income share and in its 1979-2019 rise, both pretax and posttax.
The largest pretax difference is partnership excess fiscal depreciation: slightly more than 100% of noncorporate excess depreciation belongs to partnerships, but AS allocate about 78% of it in proportion to depreciation and expensing of sole proprietorships, so AS give the top 1% only ~11% of it vs 65% in PSZ.
AS allocate S-corporation tax-exempt income to owners of C-corporations; together with partnership excess depreciation this explains 41% of the pretax trend gap.
Tax evasion creates mainly a level difference and explains only 5% of the 1979-2019 trend gap; contrary to a widely held view, it is not a key source of discrepancy in trends.
AS allocate retirement-account income partly to beneficiaries of unfunded pensions and allocate retained earnings mainly by dividends; together these explain 16% of the pretax trend gap.
AS add consumption taxes (sales taxes, excises, tariffs) to the pretax income of consumers, depressing pretax inequality and its rise; this explains 12% of the pretax trend gap.
For posttax income, AS's half-per-capita allocation of government consumption and tax-based allocation of the deficit reduce the level and rise of inequality, whereas PSZ treat both neutrally.
After the adjustments, the AS series becomes similar in level and trend to PSZ (pretax and posttax; also for the top 0.1% and 0.01%).
AS's assumptions imply an equalization of capital income since the 1960s that is at odds with the rise in wealth concentration in all available sources.

Every problem found

1majorlikely · previously known · speculative assumption

The paper's single biggest item is a large, fast-growing block of untaxed partnership income. The paper assigns it to the rich by assuming it is spread like the partnership income people report on their tax returns. The rival authors later traced who actually owns the partnerships that claim this depreciation, and found it much less concentrated at the top. The paper does not discuss that evidence. If the evidence is right, this item explains a small part of the disagreement rather than its largest part.

Technical detail and evidence

Why it matters, in detail. Substituting AS's linked-owner effects for 2019 (line 1 +0.16 to +0.28 pp; line 2 -0.11 pp; 1979 unchanged) reduces lines 1+2 from 1.67 pp to 0.05-0.17 pp. The six-item explained share then falls from 70.5% to 39-42% of the 2019 level gap and from 75.2% to 33-36% of the 1979-2019 trend gap; the posttax trend share falls from 74.6% to 43-46%. This is an illustration, not a corrected estimate, for four reasons. AS's effects are computed in AS's own microdata framework and baseline. They cover line-16c depreciation only (~$87bn linked to resident 1040s, about 15% of partnership fiscal depreciation plus amortization). They use total rather than excess depreciation. They rely on assumed -10% and -15% adjustments. PSZ's own 55-72% range implies 0.98-1.35 pp. A defensible range for line 1 is roughly 0.2-1.2 pp, and the paper reports only the top of it.

Evidence. PSZ (2026) pp. 7-9, Fig. 2, App. C.2 pp. 34-35, reference list pp. 21-22. AS (March 2025) 'Reply to PSZ (2024)' pp. 4-8 and workbook AS-March2025-reply.xlsx (checked cell by cell): sheet PartDist, linked depreciation $86.97bn, of which $33.74bn (38.8%) goes to the top 1% of returns by AGI and 23.1% to negative-AGI returns; first-tier recipients of ~$273bn line-16c depreciation: C-corporations $87.3bn (32%), other partnerships $111bn (41%), individuals $68.4bn (25%). Sheet F2: effects +0.250, +0.277 and +0.162 pp. Sheet ScorpDist: 39.4% to the top 1%, 84% of S-corp depreciation matched. Sheet F3: -0.108 and -0.102 pp. AS (Sept 2026) Table 1 repeats these (0.16; -0.11). SOI 2019: Form 1065 line 16c $276.0bn, out of $589.2bn partnership depreciation plus amortization; guaranteed payments (line 10) $68.8bn. Schedule E Part II partnership income (h)+(k) is $406.8bn against $557.1bn ordinary profit, so at most ~61% once guaranteed payments are removed. Cooper et al. (2016, p. 103): individuals directly receive 31.5% of partnership income allocations and other partnerships 26.3%.

Why this severity. These two lines are the paper's centerpiece and 41% of the pretax trend gap. Replacing them with the only direct owner-linked evidence, imperfect as it is, would cut the explained pretax trend from 75% to about a third. The paper presents only variants of its own proxy.

The authors' response. None in the audited version, which does not cite AS (2025a). AS (2026, Section I) note that the comment 'again calls for linked data but does not discuss this evidence'. The paper's general defence (pp. 7-9) is that allocating to losses first targets capital-intensive firms, and that Balkir et al. (2025) find Forbes-400 partnership losses coincide with positive book income.

2moderateconfirmed · newly identified here · methodology

How much of the gap is credited to each disagreement depends on a bookkeeping choice, and the paper makes that choice in the way most favourable to its 'three-quarters' figure. Under equally reasonable choices, the explained share is closer to two-thirds before taxes and between a half and two-thirds after taxes. Most of the gap is still explained.

Technical detail and evidence

Why it matters, in detail. Pretax trend explained: 75.2% (paper), 68.9% (AS-world switching for lines 6-8), 67.4% (PSZ-world, using Auditor A's micro-file values for lines 6 and 8). The 2019 pretax level share falls from 70.5% to 62-65%. Posttax trend falls from 74.6% to 53-67%, and the posttax level from 82.7% to 60-73%. Correcting only the posttax carry-over of line 6 gives 74.5% (level) and 68.6% (trend). The 0.8 pp of posttax trend attributed to AS's per-capita lump sum arises only when the rule is evaluated at PSZ's level of inequality: 0.72 pp PSZ-world, about 0 AS-world.

Evidence. PSZ2026.xlsx DataTab2: AT22/AT62 = 0.1077/0.1900 and AZ/BE = 0.0855/0.1514 are hard-coded PSZ shares; BH = B + BA + BF, so the posttax total includes all pretax corrections. The adjudicator recomputed from AS's posted Table C1 (scripts/as_stages.py): AS-world line 6 = 0.283/0.494 pp (1979/2019; trend 0.211), line 7 = 0.558/0.550 (trend -0.008), line 8 = -0.067/0.301 (trend 0.368), posttax line 6 ≈ 0.16/0.12. From PSZ's public Feb-2022 DINA micro-files (scripts/psz_world_line7.py): PSZ-world line 7 = 0.649/1.366 pp (trend 0.717), reproducing Auditor A. PSZ 'US DINA: Updates' (Feb 2022) p. 2 describes the posttax treatment of indirect taxes. AS (2024, JPE) Table 4 shows an aftertax offset of -0.3 pp for 'other taxes' in 2014.

Why this severity. The qualitative claim that these items explain most of the gap holds under every convention. But 'three-quarters' is the top of a range of roughly 53-75%, depending on the income concept and convention, and the 15% posttax-trend attribution to government consumption is not robust.

The authors' response. Footnotes 14 and 50 acknowledge that ss depends on the order of additions within a framework but call the effect small. The cross-framework interaction and the posttax carry-over are not discussed.

3moderateconfirmed · newly identified here · misrepresentation of source

What the paper calls 'consumption taxes' also includes items that are not taxes, such as income of charities and payments from the Federal Reserve. Counting only the taxes, this item explains less of the gap, and the taxes did not grow as a share of national income as the paper says.

Technical detail and evidence

Why it matters, in detail. Keeping the paper's ss values but restricting S to taxes: the 2019 level is 0.65 pp (AS otx) or ~0.57 pp (NIPA or PSZ wedge), against 0.80; the 1979-2019 trend is 0.30-0.35 pp, against 0.455. That is 8-9% of the pretax trend gap instead of 12%. Reassigning the non-tax part to the residual lowers the explained pretax trend from 75.2% to ~72.6%. Combined with F2, the posttax contribution of consumption-tax treatment proper is ~0.1-0.3 pp.

Evidence. AS workbook INPUTS (re-read by the adjudicator): 2019 otx 968.3 + subtr 89.3 + npgv 89.6 + fed 54.9 = 1,202.1bn = C1 GJ-FZ; 1979: 115.7 + 4.2 + 7.5 + 9.3 = 136.7bn (non-tax share 15.3%). AS online appendix Section 11 (pp. 24-25): indirect taxes by disposable income less saving; transfers, subsidies and nonprofit/government income half per capita and half by wages; Fed payments by absolute improved fiscal income. BEA NIPA Table 3.5 (2019): federal taxes on products $173.6bn + state and local $691.6bn. Workbook column DataTab2!AS2 is labelled 'Sales taxes & other'.

Why this severity. Overstates a secondary result, the consumption-tax attribution, by about a quarter to a third, and misstates a size and trend in the text. The total explained share changes by only ~3 pp.

4moderateconfirmed · previously known · misrepresentation of source

The paper says the rival study hands investment income to public employees whose pensions have no assets behind them. That is not how the rival study actually allocates this income. By the paper's own benchmark the effect is about half what it claims, and the amounts are small either way.

Technical detail and evidence

Why it matters, in detail. Line 4 = (10.25% - 6.30%) x 6.74% = 0.27 pp in 2019 (trend 0.26). With PSZ's own funded-DFA benchmark, (8.3 - 6.3) x 6.74% ≈ 0.13 pp, about half. AS's illustrative alternative is 0.0-0.1 pp. The line is ~7% of the pretax trend gap. The substantive concern, that DB income is allocated by earnings regardless of plan funding, is plausible but unquantified.

Evidence. AS online appendix pp. 12-13; AS (2026) Section II.2 and fn 10 (FL593064105, FL593064205). DataTab2!AG62 = 0.0630, AI62 = 0.0674, AJ62 = 0.1025, AK62 = 0.0027. PSZ App. E.2 pp. 38-39 (7.0% in 1989, 8.3% in 2019). AS (2025a) fn 24-25 reconstruct the 8.3% as $1.7tn/($27.2tn - $6.5tn).

Why this severity. Roughly halves a secondary result and misdescribes the source behind one of the four AS choices the conclusion calls 'indefensible'; but the line is small (0.27 pp).

5moderateconfirmed · previously known · overclaiming

The paper argues that the rival study's capital-income numbers cannot be right because wealth became more concentrated. But it compares numbers ranked in different ways, and the paper's own capital-income series shows a similar decline since the 1960s. The argument therefore does not single out the rival study.

Technical detail and evidence

Why it matters, in detail. 1989-2019 like-for-like (income-ranked): DFA wealth +8.0 pp vs AS capital income +2.9 pp (plotted series; +3.6 pp including taxes), and PSZ's own capital income -1.1 pp. 1962-2019: AS -7.0 pp vs PSZ -5.0 pp (no-tax series), while PSZ wealth +5.8 pp. Declining interest rates, composition shifts toward pensions and housing, and rank differences mean capital-income and wealth shares need not move together.

Evidence. PSZ2026.xlsx sheet capital (read by the adjudicator), columns AY/AZ (AS), BI/BJ (PSZ by pretax income), BC (PSZ wealth), BL/BM (DFA by wealth/by income), rows 62-100. Hybrid formula verified by Auditor B: 0.4491 x (1-0.8385) + 0.1641 x 0.8385 = 0.2101 (= AZ100). AS (2025a, 2026) Section III; Splinter (2023) short reply.

Why this severity. The supporting argument (C9) does not survive a like-for-like check, but it does not enter the Table 2 reconciliation.

The authors' response. Not addressed in the 2026 revision; the figure notes describe the DFA adjustments.

6moderatelikely · methodology

To measure how the rival study treats untaxed depreciation, the paper uses an average that also covers tax evasion. The rival's own figures show it gives the rich far more of the depreciation than that average implies. So the paper probably overstates its biggest item and understates the role of tax evasion. Its claim that evasion barely matters for the trend is not secure.

Technical detail and evidence

Why it matters, in detail. Holding the pool fixed, an AS ss for excess depreciation of 16%, 22% or 29% gives line 1 = 1.10, 0.97 or 0.81 pp (trend 1.02, 0.89, 0.73), against 1.20 (1.13) in the paper. Line 3 becomes 0.89, 1.04 or 1.22 pp (trend 0.31, 0.46, 0.64), against 0.78 (0.20). Evasion's share of the pretax trend gap rises from 5% to 8-17%, and in 2019 levels evasion can exceed excess depreciation. Lines 1+3 together (~2.0 pp) and the total explained share do not change. Iselin & Reck's independent estimate (~1.1 pp of 1962-2019 trend, about 30% of the AS-PSZ divergence) points the same way.

Evidence. DataTab2!V62 '=Q62', V22 '=Q22'; L62 = G62 = capital!BS100. Q62 = ('C1-Incomes'!EQ67-EG67)/(EL67-EB67) = 0.11418. AS INPUTS 2019 (re-read by the adjudicator): wgu 159.3, farmu 30.3, nfu 532.5, nfcc 381.4, nfres -91.0, rentu 18.4, scpu 53.6, summing to 1,084.5 = the Ed1 stage. AS SAS code: UNDER = UNDER + 0.85*PropExpn + (nfcc - 0.85*totexp)*(PropDepr/tPropDepr) + nfres*(pPARTSCP+pBUSN)/(...). AS-March2025-reply.xlsx sheet F3 rows 42-45: EXP 32.04/66.97; DEPR 107.42/314.39; RESID -38.53/-90.96; UNDER 200.21/1,084.49. AS online appendix p. 21; INPUTS nfcc2017 = $311.9bn. PSZ App. D.2 discloses the pooling. Iselin & Reck (Dec 2025) Table 1 and pp. 3-5.

Why this severity. The total explained share is unchanged, but 0.1-0.4 pp moves from the paper's headline item to evasion. This undermines the highlighted claim that evasion explains only 5% of the trend.

The authors' response. App. D.2 states that AS's Table C1 only provides the pooled stage and that the computation follows AS's ordering. App. E.1 argues that partnerships are barely examined in random audits.

7minorconfirmed · newly identified here · misrepresentation of source

The paper says the rival study gives this income mostly to small self-employed people. In fact its formula also uses depreciation claimed by landlords and farmers. The broader point, that the rival's formula is not based on who owns the partnerships, still stands.

Technical detail and evidence

Why it matters, in detail. Composition of the allocation key in 2019: about one third Schedule C, more than half individual rental depreciation, plus farm depreciation and pass-through Section 179. No direct effect on line 1, since the AS ss comes from Table C1 (see F3 for the share actually allocated to the top).

Evidence. AS SAS code lines ~3380-3381; SOI Pub. 4801 (2019 line items) pp. 103-104 (Form 4562), Schedule C, E, F line items; PSZ fn 5-6 and pp. 2, 5-7.

Why this severity. A factual misdescription that appears in the abstract, but Table 2 uses AS's computed ss, and the key's concentration is about what the paper implies.

8minorconfirmed · previously known · version instability

For one item the paper uses older official estimates that make the difference look bigger. Newer estimates, which the paper itself uses elsewhere and used in its previous version, cut the item by about 40%.

Technical detail and evidence

Why it matters, in detail. With BEA-2024 aggregates (0.967% of NI), line 2 = (55.8 - 28.2) x 0.967% = 0.27 pp instead of 0.46 pp. Lines 1+2 fall from 41% to ~36% of the pretax trend gap, and total explained pretax trend from 75.2% to 70.1%.

Evidence. DataFig3 D23-D28 (45.83% x 10.15% - 2.52% - 0.45% = 1.68% NI); DataFig3 E11-E15 (BEA 2024: excess depreciation $147.7bn, untaxed $258.7bn incl. $82.0bn evasion); WIL TN 2024/04 fn 48 ($720bn - $461bn - $54bn = $205bn) and Table 2 (0.3); PSZ (2026) fn 15, 17; AS (2026) p. 8.

Why this severity. Disclosed in footnotes and worth only ~0.2 pp, but it pushes a headline number up through a vintage reversal.

The authors' response. Fn 15: the older aggregates 'underpin the February 2022 PSZ vintage, which we use for our comparison with AS'; fn 17 reports the 0.3 pp alternative.

9minorconfirmed · newly identified here · internal inconsistency

The paper splits hidden business income between two kinds of businesses in two different ways in two places. This changes the evasion item somewhat but not the main conclusions.

Technical detail and evidence

Why it matters, in detail. Line 3 (2019) falls from 0.78 to 0.52 pp under the Figure-3 split. The 1979 split (17.7% partnerships) is close to Figure 3's, so the rising partnership weight in 2019 drives part of PSZ's evasion trend.

Evidence. business!AC64/AD64 = 185.8/418.5; DataFig3 C12/D12 = 482.6/89.1; DataTab2!S62 formula. Adjudicator recomputation: with the Figure-3 split, PSZ evasion ss falls from 28.3% to 22.7%.

Why this severity. An internal inconsistency affecting one line by ~0.26 pp; it cuts against F3, lowering evasion's contribution.

10minorconfirmed · newly identified here · internal inconsistency

The paper says its own method spreads public spending on things like defence and roads in a way that leaves inequality unchanged. In fact it spreads it according to cash income, which tilts it slightly toward the rich.

Technical detail and evidence

Why it matters, in detail. PSZ's cash-income key adds ~0.25 pp (2019) and ~0.05 pp (1979) to PSZ's own posttax top-1% share relative to a truly neutral allocation, about +0.2 pp of the posttax rise (micro-file basis). Measured against PSZ's actual allocation rather than the neutral benchmark, line 7 would be ~1.65 pp in 2019 instead of 1.46.

Evidence. scripts/psz_microfile_keys.py on usdina2019.dta and usdina1979.dta: colexp/dicsh constant; ss of colexp on base (poinc - colexp) = 16.35% vs base share 14.83% (2019); 7.98% vs 7.69% (1979). PSZ (2018, QJE) Section on collective expenditure. The deficit is close to neutral, as the comment says (prisupgov ratio to posttax income between -0.035 and -0.039 across records), consistent with PSZ Updates (Feb 2022).

Why this severity. A small misdescription of PSZ's own method (~0.25 pp of PSZ's 2019 posttax top share); in the decomposition it slightly understates line 7 relative to PSZ's actual practice.

11minorconfirmed · newly identified here · documentation gap

The benchmark series called 'pretax' is a slightly different published series. It makes essentially no difference to the results.

Technical detail and evidence

Why it matters, in detail. Using TB1 instead would raise the 2019 pretax gap from 5.21 to 5.29 pp; the trend gap changes by 0.002 pp; the largest annual difference is 0.47 pp (1989).

Evidence. PSZ2022AppendixTablesII(Distrib).xlsx: TA1 1979/2019 = 10.7737%/18.9954% = DataTab2!AT22/AT62; TB1 = 10.8541%/19.0778%; TC1 = 8.5477%/15.1381% = DataTab2!AZ. Workbook header 'PSZ top 1% factor income'.

Why this severity. A labelling/documentation problem worth ~0.08 pp at each endpoint and ~0 on the trend.

12minorconfirmed · newly identified here · internal inconsistency

A chart includes one small extra correction that its caption does not mention. It does not change any conclusion.

Technical detail and evidence

Why it matters, in detail. Adjusted AS posttax rise 5.33 pp in the figure vs 5.28 pp from Table 2's eight items; both round to 5.3.

Evidence. DataTab2!B62 = T62+O62+Y62+AD62+J62+AK62+AU62; AD22 = 0.000143, AD62 = 0.000673.

Why this severity. Undisclosed but tiny (0.05 pp on the trend).

13minorconfirmed · previously known · overclaiming

The paper describes the rival's choices as uniformly biased and the reconciliation as closer than the numbers show. Its own judgment calls also tend to lean one way.

Technical detail and evidence

Why it matters, in detail. Annual adjusted pretax gap (PSZ factor income minus AS-adjusted): min 0.33 pp (1980), max 1.91 pp (2016); posttax -0.36 to 1.26 pp.

Evidence. AS (2024 JPE) Table 1 p. 2194 and Table 4 p. 2210; PSZ (2026) pp. 19-20; DataTab2 columns C-E and BI-BK, 1979-2019 (scripts/table2_check.py); AS (2026) p. 5; Splinter (2023).

Why this severity. Rhetorical overstatement; it does not change the computed contributions.

14minorconfirmed · version instability

A comparison with an official study of tax evasion uses numbers that measure different things, and it was quietly swapped for a different comparison between versions.

Technical detail and evidence

Why it matters, in detail. Removing the CCAdj from AS's 2001 pool would probably lower AS's evasion ss, which favours PSZ's point. The units and ranking differ, so the 5.3 pp contrast is not a like-for-like test.

Evidence. WIL TN 2024/04 p. 10; PSZ (2026) App. E.1 pp. 37-38; AS INPUTS nfcc2001 = $71,141m, nfres2001 = $19,950m; AS (2025a) p. 10.

Why this severity. Weakens a supporting comparison; the direction of any bias is ambiguous.

15minorconfirmed · newly identified here · misrepresentation of source

A footnote misreads an official table. It does not change the main depreciation estimate.

Technical detail and evidence

Why it matters, in detail. The negative sole-proprietor excess depreciation (-$43.6bn) comes from BEA's legal-form CFC split (nonfarm sole-proprietor CFC $100.2bn vs tax depreciation $56.5bn). Its explanation is open, but the partnership estimate stands on its own inputs.

Evidence. BEA FA Section 5 (Sept 2026 file): FAAt501 line 5 = $1,921,850m; FAAt504 line 5 = $55,770m; PSZ (2026) fn 3, p. 4.

Why this severity. Misread source in a footnote; no demonstrated effect on Table 1 or Table 2.

16minorconfirmed · newly identified here · arithmetic or transcription

A few typos in a technical appendix; the numbers that matter are right.

Technical detail and evidence

Why it matters, in detail. Exposition only: Table 1 computes $641.7bn - $277.9bn = $363.8bn and $589.2bn - $178.1bn = $411.1bn correctly; read literally, the text's 'sum' would give $388bn.

Evidence. PSZ (2026) App. A.1 pp. 29-30, fn 31 and 33; Table 1.

Why this severity. Exposition errors; Table 1 and the computations are correct.

17minorconfirmed · newly identified here · arithmetic or transcription

Several small numbers in the text are slightly off. None changes the findings.

Technical detail and evidence

Why it matters, in detail. Each is at most 0.1 pp or purely descriptive; the rising education share slightly weakens PSZ's argument that only a small, constant share of government consumption can be allocated per capita.

Evidence. AS C1 LA->LK stage: bottom-50% ss 0.4019 (1979), 0.3716 (2019); half of $2,868.8bn / 328.24m = $4,370; DataTab2 formulas reference 'C1-Incomes'; Table A1; DINA micro-files educ/colexp 0.294 (1979), 0.345 (2019).

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

Claims examined and rejected

Problems one auditor raised that did not survive checking.

If sole-proprietor excess depreciation were zero, partnership excess depreciation would be ~$367bn and line 1 would fall by ~0.13 pp (Auditor A, I7). — The $411bn partnership figure is not a residual of the noncorporate total. It is computed directly as partnership Form 4562 depreciation plus amortization ($589.2bn) minus BEA's partnership CFC ($178bn). The counterfactual requires BEA's legal-form split of CFC to be wrong by ~$44bn, and no evidence is offered for that. The footnote-3 misreading itself is retained as F15.
The landlord and farm depreciation in AS's allocator implies AS's own share of excess depreciation is likely well above the pooled 11.4% (Auditor A, I4 'implication'). — The allocator's top-1% share is ~16% (AS 2025a, p. 5), close to the 15% of sole-proprietor income that the paper cites, so the composition error alone does not imply a much more concentrated key. The evidence that AS's excess-depreciation share exceeds 11.4% comes instead from AS's final allocation shares (F3). The misdescription is retained as a minor issue (F7).
p. 2's 'remaining 95% overwhelmingly come from sole proprietorships' contradicts footnote 6 (17% from S-corporations) (Auditor A, I13f). — Not an inconsistency: 95% = 78% (individuals' own depreciation and expensing) + 17% (S-corp Section 179). 'Overwhelmingly' is loose wording; the substantive problem, what the 78% contains, is covered by F7.
The DFA 'funded pension wealth by income' figures (7.0% in 1989, 8.3% in 2019) cannot be traced (Auditor A, I9). — Partly rejected. They are undocumented in the workbook, but AS (2025a, fn 24) reconstruct the 8.3% from DFA totals as $1.7tn/($27.2tn - $6.5tn), i.e. by assuming all unfunded DB claims sit below the top 1%. F5 retains this as 'undocumented and assumption-dependent'.
The posttax 'eight components' total is a 'double count' of the consumption-tax line (Auditor A, I1). — Partly rejected as to the label only. No item is summed twice; the problem is that the pretax contribution is carried into posttax income, where both methods take consumption taxes out again. The substance is retained in F2.

Audit notes

What reproduced. Reproduced exactly by the adjudicator: Table 2 and Table A1 contributions and all percentages quoted in the text, from PSZ2026.xlsx; AS stage ss values and top shares, from AS's posted workbook (independently of PSZ's formulas); the PSZ benchmark shares (TA1, TC1); footnote 1 (TabFig2022); and the composition of AS's Ed6 and Ed1 stages (INPUTS). Also reproduced: SOI 2019 line items (Form 4562, Schedules C/E/F, Form 1065); the linked-owner tabulations in AS's March-2025 workbook; AS-world switching for lines 6-8; the PSZ-world government-consumption line and PSZ's allocation keys (public DINA micro-files 1979, 2019); the 1979-2019 annual adjusted series; and NIPA noncorporate CCAdj 2019-2025. Relied on auditors without re-running: partnership Form 4562 totals and BEA CFC behind $411bn (both auditors); PSZ-world lines 6 and 8 (Auditor A's micro-file runs); within-PSZ ordering effects (Auditor A). Not reproducible: PSZ's comment-specific microdata programs (ss_psz sheets are pasted outputs); AS's confidential linked K-1 microdata; NIPA Table 7.13 historical vintage; BEA S-corp prototype tables (read only through the authors' and AS's figures).

Limits of this audit. AS's confidential microdata and linked K-1 files are not accessible, so AS's linked-owner effects are taken as reported and cannot be placed on PSZ's basis. PSZ's comment-specific programs are not public, so the PSZ-side ss values were checked against pasted outputs and approximate micro-file re-computations only. PSZ-world switching for lines 6 and 8 relies on Auditor A's micro-file runs (line 7 was re-run). The public 1979 micro-file does not reproduce PSZ's published 1979 posttax share (7.7% vs 8.55%), so 1979 PSZ-world legs are approximate. Iselin and Reck's estimates cover 1962-2019, not 1979-2019. The adjudication does not re-estimate the true distribution of untaxed income or take a position on the conceptual incidence debates (government consumption, deficits, price basis of consumption taxes), beyond noting that both sides' choices are conventions. AS's September-2026 reply postdates the audited version and is used only as evidence, not as a standard the authors could have met.