← All papers
David Gamage, Emmanuel Saez, Darien Shanske · Policy report on New York bill S8277B (2019-20), 1 September 2020; SSRN 3766547
The NY Billionaire Mark-to-Market Tax Act: Revenue, Economic, and Constitutional Analysis
10 problems found: 1 major · 4 moderate · 5 minor
What the paper claims
- The Act would raise $23.2 billion of additional New York State revenue for tax year 2020 from a one-time deemed sale (1 July 2020) taxing all unrealized gains of the state's billionaire residents.
- In subsequent years the annual mark-to-market tax adds about $1.2 billion a year over current revenue, and 'What is certain is that the subsequent revenue flow is going to be small relative to the initial tax on the stock, at least in the short run.'
- Economic effects are benign: current residents 'would not face any substantial new incentives to leave the state' because residence through 1 July 2020 locks in the 2020 tax; entrants are taxed only on post-entry gains; the ongoing tax (0.22% of wealth) is smaller than existing interstate income-tax differences, and 'In all cases ... the state tax rates are small enough to be below the revenue maximizing tax rate'; billionaires can borrow to pay without losing control of their businesses.
- 'The New York Constitution clearly and unambiguously permits New York to tax unrealized income through a mark-to-market regime'; the bill's credit and basis provisions 'eliminate any possible Constitutional concerns', so it 'is clearly authorized by both the New York and the Federal Constitution'.
Every problem found
1majorconfirmed · newly identified here · misrepresentation of source
The bill limits how much of a billionaire's gains can be taxed in one year to a quarter of their wealth above $1 billion, but the report read the limit as applying to the tax bill instead. As written, the bill would have raised about $9-10 billion for 2020, not $23 billion. Most of the rest would come due in 2021-22, and only from billionaires still living in New York, which undercuts the report's claim that the tax gives them no new reason to leave.
Technical detail and evidence
Why it matters, in detail. Under s.612-a(c) the 2020 inclusion is sum_i min(G_i, 0.25(W_i - 1)). With G_i = 0.5W_i this equals 0.25(sum W - N), so the liability is 8.82% x 0.25 x (530 - 120) = $9.04B (-61%), and the 2020 figure no longer depends on the 50% basis assumption. The remaining $162.5B of assumed built-in gains (tax value $14.3B) is recognised at the 31 Dec 2021 and 2022 deemed sales, under the same cap and only for taxpayers still resident with at least $1B. 2021 revenue (about $9.3B) is therefore about eight times the report's $1.2B flow. About 62% of 2020-22 revenue is exposed to departure, and the report's July-1 lock-in protects only about 39% of the stock tax.
Evidence. S8277B p. 1 ll. 10-14 and p. 2 ll. 21-43; A10414-A s.612-a(a)-(c); S4482 p. 2 ll. 39-42; S165 (2025); Saez web index entry '(official bill S8277B, 2019-20)'. Computations (final/scripts/headline_arithmetic.py, statutory_path.py; final/data/run_all_output.txt): 8.82% x 25% x ($530B - 120 x $1B) = $9.04B. That is $9.09B on the Forbes 29-Jul list without Bloomberg (the report's apparent list), $10.28B on the full 29-Jul list and $10.47B at the statutory 1-Jul date. Stock tax not due in 2020: $14.3B (61% of the uncapped $23.4B). With the report's other assumptions the simulated path is 2020 $9.1B, 2021 $9.3-9.4B, 2022 $5.4-5.9B, then about $1.7-2.3B a year. About 62% of 2020-22 revenue falls due after 2020. Static cumulative 2020-25 revenue is $29.4-30.2B, against $29.7-30.4B under the report's reading. Independently confirmed by Auditors A and B ($9.04B; $9.09B).
Why this severity. Cuts the 2020 headline by 55-61% and reverses the claimed revenue profile (large one-off, small flow). It also voids the July-1 lock-in for about 60% of the stock tax. Static cumulative revenue over 2020-25 is about unchanged and a large positive yield survives, so the core claim of substantial revenue is not eliminated.
The authors' response. None located. The authors' 9 Dec 2020 covering letter and the Feb 2021 Tax Notes State article by Avi-Yonah, Gamage, Shanske & Stark still describe the first year as taxing all accumulated gains ($23.3B). Later reintroductions (S4482 of 2021, S165 of 2025) keep the gain-cap wording.
2moderateconfirmed · newly identified here · arithmetic or transcription
The report put the 120 New York billionaires' combined wealth at $530 billion, but Forbes's own figures for the same 120 people that day add up to about $586 billion. The gap is almost exactly Michael Bloomberg's fortune, even though the report counts him and discusses his tax bill. This slip makes the estimate about a tenth too low, not too high.
Technical detail and evidence
Why it matters, in detail. Restoring Bloomberg raises the report-method score by 8.82% x 0.5 x $54.9B, about $2.4B (+10.4%, to $25.65B), and the flow by about $0.12B a year. Under the bill text it adds 8.82% x 0.25 x ($54.89B - $1B), about $1.19B (to $10.28B). Whether the cause was an omitted spreadsheet row or a deliberate, undisclosed exclusion cannot be established.
Evidence. final/scripts/forbes_ny_base.py and data/run_all_output.txt (NY totals for 18 Mar, 30 Jun, 1 Jul, 28 Jul and 29 Jul 2020); Forbes RTB Wayback captures (local copies in /mnt/zdata/downloads/ny-mtm-adjudicator/ and ny-mtm-claude/); scholars' letter of 9 Dec 2020 ('121 billionaires ... increased their collective net worth by 13.2% between March 18th and October 13th - or $68.7 billion').
Why this severity. A confirmed error of about 10% of the headline, in the conservative direction. It does not change the qualitative conclusion and is roughly offset by F3.
3moderateconfirmed · previously known · misrepresentation of source
The report says research shows state taxes on the very rich are always below the point where they start to lose money. The study it cites does not say that, and its main finding is that billionaires move readily to avoid state estate taxes. For billionaires the new yearly tax would also be about as large as the state income tax they already pay, so it is not obviously a small extra reason to move.
Technical detail and evidence
Why it matters, in detail. (a) No revenue-maximising rate is estimated by the source, which reports ratios above 1 for two states. (b) Applies to non-domiciled statutory residents only. (c) The 0.22% ongoing MTM is of the same order as the existing NY-vs-no-income-tax-state differential for billionaires (about 0.19% state, about 0.28% with NYC), roughly doubling it, so 'smaller than existing differences' does not follow. Induced departures would also cost existing PIT and estate-tax revenue, which the report does not net. Under the bill text (F1), about 62% of 2020-22 revenue depends on continued residence.
Evidence. Moretti & Wilson (2020 rev.) abstract, fn. 4, pp. 26 and 30-31; Young et al. (2016); NY Tax Law s.605(b)(1) and 2020 IT-203 instructions (via Auditor B); BSYZ (2025) pp. 2, 8-9.
Why this severity. Weakens the evidence offered for a secondary core claim (no new migration incentive). Under the report's own reading only the $1.2B-a-year flow is at stake; combined with F1, about $14B is.
4moderateconfirmed · previously known · overclaiming
The report says the tax is clearly constitutional. It leaves out the very next sentence of the state constitutional clause it relies on, a sentence critics had already used against the tax. It also skips the legal risk of taxing gains back to July 2020 and some double-tax cases. The tax might well survive in court, but the report's certainty is not earned.
Technical detail and evidence
Why it matters, in detail. Omitted text: the clause immediately after the one quoted. Retroactivity bears on the 1-Jul-2020 deemed sale, the basis of the July-1 lock-in. The outbound double-taxation gap is not covered by s.612-a's inbound credits. A judicial loss on any of these grounds would remove most of the scored 2020 revenue, but no probability can be assigned here.
Evidence. NY Constitution (Dept. of State, 1 Jan 2025), art. XVI s.3; Empire Center (McMahon 2020); Tax Foundation (Walczak 2020); S8277B s.612-a(a), (b), (g); Ordower, Tax Notes State 99 (22 Feb 2021), pp. 800-801; Avi-Yonah et al., Tax Notes State 99 (8 Feb 2021), pp. 541-542.
Why this severity. Overstates the certainty of a core (legal) conclusion and leaves undisclosed a risk to the whole revenue stream. No judgement is made on the legal merits.
The authors' response. Avi-Yonah, Gamage, Shanske & Stark (2021), 'Is New York's Mark-to-Market Act Unconstitutionally Retroactive?', Tax Notes State 99: 541-548, argue it is not, under rational-basis review. No response to the undistributed-profits or later-state double-taxation points was located.
5moderateplausible · previously known · speculative assumption
The estimate assumes New York would value billionaires' mostly private companies at Forbes's figures. In earlier work the same economist assumed billionaires would report only 85% of Forbes values, and tax records show rich people's estates valued far below Forbes. Because the tax falls only on gains above what the assets originally cost, a modest undervaluation cuts revenue a lot, and the report shows no range.
Technical detail and evidence
Why it matters, in detail. Basis is historical, so if reported value is k x Forbes while basis stays at 0.5 x Forbes, the gain is (k - 0.5)W. Under the report's reading, k = 0.9 cuts revenue 21% and k = 0.8 cuts it 41%. If Forbes instead overstates true value (basis proportional), the cuts are 11% and 22%. Under the binding gain cap the effect is roughly proportional (-12.5% and -25%). The Warren-letter convention (k = 0.85) gives $16.2-19.7B instead of $23.2B (report reading) and $7.3B instead of $9.04B (bill text).
Evidence. final/scripts/base_composition.py and haircut_and_families.py; Saez & Zucman letter to Sen. Warren (18 Jan 2019), item 4 and fn. 2; Balkir, Saez, Yagan & Zucman (2025), NBER w34170, pp. 3 and 10; Walczak (Tax Foundation, 8 Dec 2020): publicly traded assets are about one-fifth of top-1% assets.
Why this severity. With the authors' own 15% haircut, the report-method headline falls 15-30% (to about $16-20B) and the bill-text 2020 figure 19% (to $7.3B). A shortfall of 20% or more would make this major, but its size under this bill's anti-discount and asset-reporting rules is not demonstrated.
The authors' response. Within the report (p. 2): Forbes is 'accurate on average' and more likely to miss fortunes than to exaggerate them, and enforcement and the small number of taxpayers limit undervaluation. Elsewhere Saez & Zucman read the estate-Forbes gap as estate-tax avoidance (corpus us-wealth-taxes-11).
6minorconfirmed · newly identified here · internal inconsistency
The bill values assets as of 1 July 2020, but the report used Forbes values from 29 July. The difference is about 1% and, if anything, makes the estimate slightly too low.
Technical detail and evidence
Why it matters, in detail. Valuing at 1 July adds about $0.33B (+1.3%) under the report's method and about $0.19B under the bill text.
Evidence. Forbes NY totals: $590.95B (30 Jun close), $593.98B (1 Jul, 119 people), $586.06B (29 Jul, 120 people) (final/scripts/forbes_ny_base.py).
Why this severity. About 1% of the headline.
7minorconfirmed · newly identified here · documentation gap
A journal page range is mistyped, and one statistic is compared with the wrong benchmark, though the point still holds. The report also never says which version of the bill it scored and does not publish its list of billionaires, which made the errors above hard to spot.
Technical detail and evidence
Why it matters, in detail. Bibliographic and labelling errors only; the growth premise of fn. 2 is, if anything, understated.
Evidence. Young et al. (2016) first page; Warren letter item 6; auditors' recomputations (A: 5.15 points; B: 4.85 points).
Why this severity. No effect on conclusions, though (c) is why F1 and F2 went unnoticed.
8minorconfirmed · newly identified here · internal inconsistency
The letter that carried this report to New York's leaders quoted slightly different numbers and named an earlier version of the bill that did not contain the disputed limit. The differences are small, but they show that the version of the bill being scored was never clearly fixed.
Technical detail and evidence
Why it matters, in detail. $23.3B vs $23.2B and 121 vs 120. The bill version cited in the covering letter (S8277A) is not the one whose provisions the analysis describes (S8277B), consistent with the scored design never having been pinned to the printed text (see F1).
Evidence. Letter PDF (DocDroid file linked by NY1, 11 Dec 2020; Wayback capture 20230329132157); Politico, 11 Dec 2020 (Wayback 20201212000242); Avi-Yonah et al. (2021) p. 542 and n. 4; Ordower (2021) n. 44; Wayback CDX for the report URL and Saez's home page.
Why this severity. 0.4% numerical difference; a documentation problem only.
9minorconfirmed · newly identified here · internal inconsistency
The report says newcomers would be taxed only on gains made after they move to New York. The bill actually starts the clock on 1 January of the year they move, which the report itself states correctly on its last page. This does not change the revenue estimate.
Technical detail and evidence
Why it matters, in detail. No entrant revenue is scored. The mismatch bears only on the claim that the Act creates no new disincentive to become a resident.
Evidence. Report pp. 1, 2 and 4; S8277B p. 2 ll. 32-39.
Why this severity. No effect on the revenue estimate; affects only the wording of a secondary claim.
10minorlikely · newly identified here · data error
One person counted as a New Yorker appears, by Forbes's own description, to live in Switzerland. Another of the largest fortunes was inherited in 2019, so most of its past gains had already been wiped out for tax purposes. These errors push the estimate up by about as much as the Bloomberg error pushes it down.
Technical detail and evidence
Why it matters, in detail. Wertheimer: -8.82% x 0.5 x $24.62B, about -$1.09B under the report's method (-$0.52B under the bill text). Koch: -8.82% x ($23.8B - $5.1B), about -$1.65B (report method), or about -$0.58B (bill text). Combined, about -12% (report) or -11% (bill), roughly cancelling F2. The corrected bill-text 2020 figure on the 1-Jul list is about $9.4B.
Evidence. Forbes RTB records ('bios' field) in the 29 Jul 2020 and 15 Aug 2019 Wayback captures; Auditor A's corrected_scoring.py re-checked by the adjudicator.
Why this severity. Each is 5-7% of the headline, they offset F2, and they fall within the averaging the report states. No change to conclusions.
Claims examined and rejected
Problems one auditor raised that did not survive checking.
- I8: '$23.2 billion ... for 2020' is a tax-year liability; with the 10-year installment option, 2020-21 cash receipts could be a fraction of it (minor). — The installment option and its 7.5% charge are disclosed in the report's first paragraph, and tax-year liability is the standard scoring concept. At a non-deductible 7.5% deferral charge, against the low borrowing rates the report cites, paying up front (or borrowing to pay) beats deferring, so cash and liability are likely to coincide. Any deferral raises the present value the state receives. Auditor B reached the same view.
- I2: The annual incremental-revenue scenario lacks a validated stock-flow and transition model (moderate). — The $1.2B formula is the steady-state identity dU = A - R with U/W = 0.5 held constant, as Auditor B itself shows. Its inputs are disclosed and flagged as uncertain (p. 1), and they are plausibly conservative: BSYZ (2025) find top-400 taxable income equal to about one-third of their 2010-19 wealth rise, and Forbes-400 wealth grew far faster than 5%. Sensitivity to disclosed parameters is not a defect. The real transition problem, the carry-forward of capped gains, is F1.
- I3 (part): the $530B base and its 120-person membership cannot be independently reproduced from a dated Forbes snapshot. — Superseded. Archived Forbes real-time data for 29 July 2020 (Wayback, digest-verified) reproduce the count of 120 exactly and show a total of $586.1B. The base is reproducible, and it is in error, not merely undocumented (F2).
- Checks passed: 'Only one version exists online, and the SSRN version is the same document. No undisclosed revisions were found.' — Partly incorrect. The report text is stable: the analysis appended to the 9 Dec 2020 scholars' letter matches the SSRN copy word for word. But the covering letter, led by the three authors, states $23.3B, 121 billionaires and 'S8277A/A10414', and the PDF's printed date (1 Sept 2020) predates every located copy, the earliest being 11 Dec 2020 (F9).
- I4 (part): the 1 July residence lock-in argument is categorically wrong because 1 Jan-1 Jul 2020 is only 183 days. — Partly rejected. The day count matters only for statutory (non-domiciled) residents. A domiciliary resident on 1 July is subject to the 1 July deemed sale even after a later move. The report's conclusion therefore holds for the typical NY billionaire, so the point is kept only as a minor sub-point of F5(b).
- I5: the inbound credits leave cross-year interstate realization cases unresolved, making the constitutional assurance a confirmed overclaim (moderate). — Partly rejected. The design gap is real: there is no relief when another state later taxes a gain NY has already marked. But neither Ordower nor the auditors show it is unconstitutional, and Ordower frames it as a double-taxation and harmonisation risk. It is kept as one sub-point of F6, not as a separate confirmed constitutional defect.
Audit notes
What reproduced. Reproduced exactly from the stated inputs: $23.373B ($23.4B), $23.2B after relief, 4.4% of wealth, and the $1.169B (0.22%) flow. The $0.2B relief reproduces under the report's tax-cap reading. The count of 120 NY billionaires matches Forbes real-time data for 29 Jul 2020. The $23.2B itself ($23.23B) is reproduced only on the Forbes list without Bloomberg. Not reproduced: the $530B base (Forbes: $586.06B for the same 120 people; $531.17B without Bloomberg), and the 2020 headline under the bill's text: $9.04B on the report's aggregates, $9.09B on its apparent list, $10.28B on the full 29-Jul list and $10.47B at the statutory 1-Jul date. Statutory path with the report's other assumptions: 2020 $9.1B, 2021 $9.3-9.4B, 2022 $5.4-5.9B, then about $2B a year; static cumulative 2020-25 revenue is $29.4-30.2B, against $29.7-30.4B under the report's reading. SCF 2016 unrealized-gain shares replicate Auditor B (top 0.1%: 47.9%). The Forbes Wayback captures were independently re-downloaded and checked against the CDX digests. Code: final/scripts/run_all.sh; output: final/data/run_all_output.txt.
Limits of this audit. No author workfile or list exists, so the mechanism of the Bloomberg omission is inferred. Individual billionaires' residence, basis and taxable values are unobservable, and Forbes residence coding may contain further errors. The multi-year statutory path keeps the report's stylised assumptions (5% growth, half of accruals realized, uniform 50% basis). Its 2021-22 values depend on whether the first period is treated as 12 or 18 months. It ignores: the broad 'net assets' definition in (d) (foundation contributions, trusts and gifts), which may raise the cap for large donors; NYC conformity; New York's April 2021 rate increases (top rate 10.9% above $25M for 2021-27), which would raise bill-text revenue from 2021; loss carry-forwards; and all behavioural responses. Legal points are assessed only for completeness and accuracy of presentation; no view is offered on how a court would rule. The 1938 Revised Record quotation could not be checked (the Internet Archive copy is access-restricted and HathiTrust was blocked). The SSRN landing page was not accessible, and no copy dated 1 September 2020 was located. The drafters' intent (tax cap or gain cap) is unknown; the finding is that the report's score does not match the text it names.