Brian Galle, David Gamage, Emmanuel Saez, Darien Shanske · Expert memo, 5 February 2026; SSRN 6073728

Expert Report on Valuation of Controlling Shares of Publicly Traded Companies Under the California Billionaire Tax Act (CBTA)

9 problems found: 2 major · 2 moderate · 5 minor

What the paper claims

Founders' super-voting/control shares in publicly traded companies (e.g., Alphabet, Meta) are outside Sec. 50303(c)(3): the voting presumption of (c)(3)(C) 'does not apply', because (c)(3) 'says expressly' that it excludes 'assets and entities governed by paragraph (1)', i.e. 'shares in publicly traded entities'; a contrary reading 'would not be viable'. Backstop: the Sec. 50308(j) catch-all 'would lead to the same result', so such shares are 'valued only at their market value'.
Taxpayers will not be taxed on value exceeding the market value of their holdings: publicly traded assets are valued at fair market value on the valuation date (Sec. 50303(c)(1)); 'the plain text of the CBTA shows that the claim that holders of voting or control shares in public companies, or any others, will be taxed beyond the fair market value of their shares is simply false.'
Even for private entities the voting presumption is only a default: a taxpayer 'need only demonstrate that the presumption would overstate' value 'by submitting a certified appraisal'; taxpayers 'can submit an appraisal' or rely on the default; an appraisal 'should rebut the presumptive value easily' and 'can replace any and all formulaic presumptions'. Worked example: founder with 1,000 nine-vote shares (10% of equity, 50% of votes) faces a $50m default value, replaced by a $15m appraisal, tax $750,000.
Interpretive support: the FTB can issue regulations, canons of interpretation favour market-value treatment, and 'The revenue scoring for the CBTA is based on the publicly traded value for such shares' (authors' analysis, LAO fiscal analysis, Forbes real-time list); Forbes values unlisted super-voting shares like the listed class; 'agencies and the courts should look to how this proposal was scored'.
Claims that founders would be obliged to sell large portions of their companies to pay a 5% tax (in 5 annual instalments) are 'absurd'.

Every problem found

1majorconfirmed · misrepresentation of source

The memo says anyone who thinks the default formula overvalues their stake can hand in an appraisal and pay tax on that number. The law is stricter. The owner must prove with 'clear and convincing evidence' that the formula 'substantially' overstates the value, and the tax agency can use the same route to push values up. Small overvaluations would stick, and private-business owners could find the escape hatch harder to use than the memo suggests.

Technical detail and evidence

Why it matters, in detail. The statute sets a burden-shifting gate. The taxpayer must first show, by clear and convincing evidence, a 'substantial' misstatement; only then can a certified appraisal replace the presumed-percentage method. That burden is heavier than the preponderance standard in the federal estate/gift valuation disputes the memo cites as its analogue, and the Board holds the same power upward. Consequences: (i) non-substantial overstatements cannot be rebutted. 'Substantially' is undefined; purely as an illustration, if a 20% overstatement survived (the size of the Act's Sec. 50312(a) penalty trigger), the effective rate would rise from 5% to 6%. (ii) For hard-to-value private firms, appraisal evidence may fail the burden. (iii) Contesting carries penalty exposure under Secs. 50312 and 50305(c), mitigated by Sec. 50312(h). The memo's own example (default 3.3x the appraisal and 5x pro rata) would clear the bar. But portraying the presumption as easily defeated sits oddly with the memo's own account of its purpose, an anti-avoidance rule against trust-based planning; that purpose is why the statute sets a heightened bar. For the public-company headline, the correct standard weakens the fallback but probably does not change the outcome.

Evidence. scripts/quote_check.py section B (all three qualifiers present in the statute, absent from both memo versions). Initiative 25-0024A1 Sec. 50303(c)(3)(F), AG PDF p.15; same wording in the Prop 40 ballot text p.18. The authors' Dec 2025 expert report fn. 5 states the test correctly: 'rebut that presumption through clear-and-convincing evidence and a certified appraisal where it would substantially misstate actual value or ownership'. So does the drafters' Sept 2026 summary, pp. 8-9: 'demonstrate with clear and convincing evidence that any aspect of the Presumed Percentage Method ... would substantially overstate or substantially understate'; 'The Board is similarly empowered'.

Why this severity. It directly undercuts the memo's second core claim: that holders of private-entity interests (and founders, if the presumption applied) can simply substitute an appraisal. It also undercuts the 'or any others' assurance, because overstatements that are not 'substantial', or cannot be proven to a clear-and-convincing standard, stand. It is major rather than critical because for listed-company founders the correct test would very probably be met: the presumption overstates their holdings about 2x or more, and the listed price is strong evidence.

The authors' response. The drafters' Sept 2026 summary states the full test and expects appraisals based on recent arm's-length transactions to be 'relatively predictable and uncontroversial'. The memo was not corrected in either version obtained (v1, Jan 2026; v2 dated 5 Feb 2026, still the version on Saez's site).

2majorlikely · previously known · overclaiming

The law gives market-price treatment to 'publicly traded assets', and founders' special voting shares are not themselves traded on any exchange. So whether they are valued at the stock price depends on a decision by California's tax agency, or on how courts read the law; the text does not settle it as the memo claims. If the less favourable reading won and no challenge succeeded, Mark Zuckerberg's bill could roughly double (about 10% of his wealth instead of 5%), although that outcome looks unlikely.

Technical detail and evidence

Why it matters, in detail. Classification decides the valuation branch. Under (c)(1), Zuckerberg's Class B is valued at the listed price: tax = 5.0% of his $220.7bn Forbes wealth (25 Oct 2026). Under (c)(3), (C) sets his ownership at no less than his 60.6% of votes (vs 13.3% of equity), and (E) values Meta at GAAP book value plus 7.5 times average 2024-26 GAAP profits. That is about $711-768bn (book value $272.4bn at 30 Sep 2026; FY2026 profit estimated from 9 months), giving a base of $434-468bn and an effective rate of 9.8-10.6% before any rebuttal. Codex's end-2025 snapshot for the whole Class B pool gives 1.68x the parity value (8.39% rate). If the Sec. 50303(c)(10) 'sale of equity' floor is read to cover exchange trading, the entity value becomes market capitalisation (about $1.64tn) and the rate about 22.6%. Documented unlisted Class B is $464bn, 20.0% of the authors' own scored base (12.4% excluding Page and Brin, whom Forbes now lists as Florida/Nevada residents). Mitigants: 1:1 convertibility and identical economics give a strong catch-all and rebuttal case; Sec. 50309(b)(2) frees FTB guidance from the APA until 2028; ballot materials describe '5 percent of their net worth'; and under the alternative reading a founder whose tax exceeds his publicly traded assets could defer through an Optional Deferral Account (Sec. 50304) rather than sell. The critics' own figures (Tan: 50%; Tax Foundation: 22-25% of the stake in wealth tax alone, e.g. 22.4% for Zuckerberg, about 36% including capital-gains tax on the sales) multiply voting share by market cap, and overstate even the alternative reading under the statutory formula.

Evidence. Memo v2 p.1 para. 2, p.3 paras. 2-3, p.4; identical wording in v1 (Missouri repository PDF created 16 Jan 2026; scripts/diff_versions.py). Initiative 25-0024A1 Secs. 50303(c)(1), (c)(3), (c)(3)(E), 50308(i)-(j) (AG PDF pp. 13-15, 25), checked visually against the Prop 40 ballot text (SOS public-display PDF pp. 17-18, 28). Meta FY2025 10-K Item 5 and Note 12; Alphabet 2025 Annual Report Item 5 and Note 11. Galle et al. Expert Report dated 31 Dec 2025, fn. 5: 'If such an interest is valued under the non-public business-entity rules ... then Section 50303(c)(3)(C) presumes the taxpayer's ownership percentage is at least their voting/control percentage'. That footnote is absent from the 16 Feb, 25 Apr and 20 Jul 2026 versions. scripts/quote_check.py phrase counts (statute vs memo): 'shares in publicly traded' 0 vs 3; 'publicly traded compan' 0 vs 4; 'publicly traded entit' 1 vs 5. Stakes: scripts/reproduce.py.

Why this severity. The headline claim is a claim of legal certainty, and that certainty is not supported. The predicted result depends on a discretionary Board determination, a purposive reading of 'entities', or a successful heightened-standard rebuttal. Under the defensible asset-level reading, with no rebuttal, the paradigm founder's taxable base would be about 2x (statutory formula) to 4.5x (market-cap floor) his market holdings, which would substantially weaken the core finding. It is not critical because the memo's practical conclusion is still the more likely outcome.

The authors' response. The drafters' Sept 2026 Section-by-Section Summary (pp. 6-8) restates the entity-level reading: the (c)(3) presumptions 'do not apply to interests in publicly traded companies, whether or not those interests themselves are publicly traded'. It says courts reject 'strained, unreasonable arguments to conjure up ambiguity'. It adds that such shares 'would be taxed on the appraised value of those shares', worth 'not less than' the listed price. The Feb/Apr/Jul 2026 expert-report updates call the opposing claim a 'distorted misreading' by 'bad-faith actors'. No direct reply to Walczak or Hoffmann was located.

3moderateconfirmed · previously known · overclaiming

The law blocks the usual discounts for owning a small, hard-to-sell slice of a business. It also never lets a company be valued below a recent fundraising price unless the owner proves a big gap. So some private-business owners could be taxed on more than a buyer would actually pay, which contradicts the memo's blanket promise, though it matters little for founders of big listed companies.

Technical detail and evidence

Why it matters, in detail. With a genuine 20% minority/marketability discount, Sec. 50303(b) taxes the pro-rata value, so the effective rate on sale value is 6.25% (5.56%, 7.14% and 8.33% at 10%, 30% and 40% discounts). The funding-round floor binds whenever value has fallen within two years, or preferred-stock terms inflate the post-money valuation, unless the taxpayer clears a clear-and-convincing 'significant overstatement' bar. These are deliberate anti-avoidance choices, but they make taxable value a statutory construct that can exceed an interest's arm's-length price. The control-premium channel (Treas. Reg. Sec. 20.2031-2(e)) carries little practical weight for transfer-converting dual-class stock.

Evidence. Initiative 25-0024A1 Sec. 50303(a)-(b) (AG PDF p.13), (c)(3)(D)-(E) (pp.14-15), (c)(10) (p.17). Codex reproduction 'pro_rata_floor_sensitivity', reproduced in scripts/reproduce.py section 6. Drafters' Sept 2026 summary p.6 ('not less than'; subdivision (b) described as 'Anti-abuse rules') and p.13 (floors). Treas. Reg. Sec. 20.2031-2(e)-(f) (controlling blocks; 'degree of control' factor), which the drafters name as a 'reasonable reference point'.

Why this severity. This affects the memo's secondary, universal assurance rather than its headline about listed-company founders. Their Class B shares convert to Class A on transfer, so a buyer cannot acquire the control rights and the willing-buyer value is essentially the listed price. For holders of private-company interests, though, these rules can push taxable value noticeably above sale value.

The authors' response. The drafters' Sept 2026 summary presents subdivision (b) and the floors as anti-abuse rules and expects 'a certified appraisal will routinely determine the valuation of an asset when the appraisal is based on evidence of a recent arms' length transaction'. It does not reconcile them with the memo's 'any others' assurance.

4moderateplausible · newly identified here · methodology

Mark Zuckerberg holds almost all of his special Meta shares through private holding companies, not in his own name. Read literally, the law values what he owns, stakes in those private companies, with a formula rather than the stock price. The tax agency would probably look through them, but the memo's 'plain text' argument does not cover this common arrangement.

Technical detail and evidence

Why it matters, in detail. If the LLC interest is valued under (E), value equals the LLC's GAAP book value plus 7.5 times its average book profits. With fair-value accounting of the Meta stake, unrealised gains enter 'profits' and could push the formula well above market; with cost accounting the formula would understate, and the Board would rebut upward under (F). The private-equity-entity floor ('the value (as determined under this subdivision) of each entity owned') brings the Class B classification question back in. The effect cannot be computed without the LLCs' accounts. Sec. 50312(l)(2) lets the Board disregard entities lacking economic substance, and Sec. 50312(m) imports income-tax doctrines under which single-member LLCs are disregarded, so an eventual look-through is likely but discretionary.

Evidence. Meta 2026 DEF 14A beneficial-ownership table, fn. 2 (as of 1 Apr 2026: 341,823,978 Class B, 60.8% of votes); Meta PRE 14C (SEC accession 0001628280-26-066710, filed 21 Oct 2026) table as of 1 Oct 2026 (341,215,148 Class B, 60.6%). Initiative 25-0024A1 Secs. 50302(c), 50303(c)(2)-(3)(E), 50308(i), 50312(l)-(m). Drafters' Sept 2026 summary p.1.

Why this severity. This is a second, independent route by which the paradigm holder's shares could fall outside (c)(1), and the memo does not address it. Its impact is undemonstrated and could run either way, and the FTB's look-through and substance-over-form powers would probably resolve it, so it does not reach major.

The authors' response. Drafters' Sept 2026 summary, introduction: 'some highly technical issues, such as the treatment of property held through single-member LLCs, may be resolved through further guidance from California tax regulators.'

5minorconfirmed · newly identified here · internal inconsistency

A few weeks before the memo, the same authors had written that founders' special shares might fall under the private-company rules. They later dropped that sentence without saying so. Their later guide says the shares will be worth 'not less than' the stock price, which is not the same as the memo's 'only' the stock price.

Technical detail and evidence

Why it matters, in detail. The authors move from conditional acknowledgement (Dec 2025) to categorical denial (Jan/Feb 2026) and then to a hedged position (Sept 2026: appraised value with a floor at the listed price), and no document acknowledges the change. 'Not less than', together with the Sec. 50303(b) asymmetry (discounts barred, premiums not), means the authors' later position does not guarantee the memo's 'only at their market value', although a premium for non-transferable super-voting rights would probably be small.

Evidence. Expert report versions: 31 Dec 2025 (MD5 ae15844a..., Last-Modified 2 Jan 2026), fn. 5, p.4; 16 Feb, 25 Apr and 20 Jul 2026, Q&A 'If a founder has special control shares...'. Drafters' Sept 2026 summary pp. 6-8. In the same Q&A the Page/Brin stake changed from 'about 3%' (16 Feb) to 'about 6%' (25 Apr update), again without notice.

Why this severity. Its substantive weight is already counted in F1 (the authors themselves treated the classification as open) and F3 (a floor, not a cap, at the listed price). As a separate problem it concerns transparency across documents, and any implied premium for transfer-converting shares is small.

6minorconfirmed · newly identified here · misrepresentation of source

The memo says the state's official fiscal analyst assumed founders' special shares would be valued at the stock price, but that analysis says nothing on the point. The authors' own estimate does, but courts give little weight to what a law's drafters intended unless voters were told.

Technical detail and evidence

Why it matters, in detail. The LAO estimate is a range with no disclosed model, so it cannot show that super-voting shares were scored at listed prices; only the authors' own Forbes-based scoring shows that (Forbes values Class B at the listed-class price). In California initiative construction, drafters' undisclosed intent and non-ballot materials carry little weight. Ballot materials do carry weight, and they support a '5 percent' reading in general but are silent on share classes.

Evidence. LAO/DOF letter to the AG, 11 Dec 2025 (2 pp.), and lao.ca.gov/BallotAnalysis/Initiative/2025-024 (the memo's link); SOS Official Voter Information Guide, Prop 40 analysis and summary pages; Taxpayers to Limit Campaign Spending v. FPPC, 51 Cal.3d 744, fn. 10 (text verified); the authors' expert report (Dec 2025: Forbes real-time list as of 17 Oct 2025; '72% of the wealth of CA billionaires is in the form of publicly traded stock (e.g. Meta for Mark Zuckerberg)').

Why this severity. It concerns a supporting interpretive argument, not the operative analysis, and the memo's statements about its own scoring and about Forbes are accurate.

7minorconfirmed · newly identified here · version instability

There are at least two versions of the memo online with different dates, and the earlier one, still posted, exaggerates what the critics argued. The legal reasoning is the same in both, so this mostly matters for accurate citation.

Technical detail and evidence

Why it matters, in detail. This creates citation ambiguity only. Every passage criticised in F1-F3 and F8 appears verbatim in both v1 and v2. The 16 Feb 2026 SSRN revision could not be inspected.

Evidence. v1: Wayback raw capture 20261012042342 of scholarship.law.missouri.edu/cgi/viewcontent.cgi?article=2351&context=facpubs (MD5 be67ecfb94197014db97dda6a180865b; captures of 7 and 12 Oct 2026 share the same digest). v2: eml.berkeley.edu/~saez/galle-gamage-saez-shanskeCBTAcontrollingshares.pdf (MD5 7ca39eb1463631c8231f479fa655e6cb). scripts/diff_versions.py output. SSRN abstract page, Wayback 5 Sep 2026.

Why this severity. The analysis is the same in every version obtained, so conclusions do not depend on which text is cited. The problem is documentation, plus a straw-man framing that v1 and the SSRN abstract still carry.

The authors' response. SSRN abstract: 'This memo was slightly revised on 2/16/26.' No change log.

8minorconfirmed · newly identified here · arithmetic or transcription

A few small slips: one quotation of the law is inaccurate, one sentence about the example's tax is confusingly worded, a name is misspelled, and the memo does not mention the extra charge for paying in instalments. None of them changes the memo's argument.

Technical detail and evidence

Why it matters, in detail. No effect on the legal argument. The instalment charge raises nominal payments by 15% and is present-value neutral at the 7.5% statutory rate; the example's arithmetic is exact.

Evidence. scripts/quote_check.py section A (p.2 fragment not in statute); scripts/reproduce.py sections 1-2; Initiative 25-0024A1 Secs. 50303(c)(3), (c)(3)(E), 50301(c); memo_v1 date line.

Why this severity. Presentation-level; none changes a conclusion.

9minorconfirmed · newly identified here · other

The 'expert report' was written by the people who wrote the ballot measure, and the memo does not say so. Its reading of the law is the sponsors' view, not an independent assessment.

Technical detail and evidence

Why it matters, in detail. For literature classification this is an advocacy document by the measure's drafters. Under Taxpayers to Limit Campaign Spending v. FPPC fn. 10, its statements of drafting intent carry little interpretive weight unless they were communicated to voters.

Evidence. Sacramento Bee, 12 Oct 2026 (Wayback capture 20261016164625); drafters' Sept 2026 Section-by-Section Summary p.1 and fns. 2 and 4; memo p.3.

Why this severity. This affects how readers should weigh the document (a proponent's legal position, not an independent expert valuation), not the correctness of its reading.

The authors' response. The Sept 2026 summary is openly 'prepared by the drafters of Proposition 40'; the memo itself contains no such statement.

Claims examined and rejected

Problems one auditor raised that did not survive checking.

I7: v1-v2 differences are unknown, so substantive changes may separate the January and February texts and 'readers citing Galle et al. (2026) may be citing different texts' with different content. — Narrowed. The adjudicator obtained v1 (Wayback raw capture of the Missouri repository PDF, 12 Oct 2026). A word-level diff shows v1 and v2 are 97.8% identical. The only substantive change is the summary's description of the critics' claim, plus the date line; every passage criticised in F1-F3 and F8 appears in both. The surviving documentation point is F7 (minor). The 16 Feb SSRN revision remains unseen.
I3: the memo silently 'reverses' a published position of the authors (Dec 2025 fn. 5); rated moderate. — Partly rejected. Fn. 5 was conditional ('If such an interest is valued under the non-public business-entity rules (and the proposal defines such assets broadly - See Section 50308(j)) ...'): it flagged the possibility rather than adopting the opposing reading. The shift is an unacknowledged hardening, kept as F5 (minor), and its substantive weight is counted in F1 and F3. Factual correction: the Page/Brin stake changed from 'about 3%' to 'about 6%' in the 25 Apr 2026 update, not on 20 Jul.
I5(1): because Sec. 50303(c)(1) only 'presumes' the trading price and Sec. 50303(b) bars discounts but not premiums, founders' controlling blocks could be assessed at a control premium above market (Treas. Reg. Sec. 20.2031-2(e)). — Largely rejected for the memo's actual subject. Meta and Alphabet Class B convert to Class A on any transfer, so a willing buyer (Sec. 50303(a)) cannot acquire the super-voting rights, and the hypothetical-sale value is essentially the listed price. Kept only as a theoretical caveat within F3.
I1 magnitude: un-rebuttable overstatements 'of up to roughly a fifth' would raise the effective rate from 5% to about 6%. — Not adopted as a finding. 'Substantially' is undefined, and the 20% figure borrows the Sec. 50312(a) penalty trigger, which serves a different function. The arithmetic is right (5% x 1.2 = 6%), but the threshold is an assumption; F2 reports it only as an illustration.
No version-instability finding is warranted. — Partly rejected. Codex is right that the substance is stable (confirmed here by obtaining v1), but undocumented versions with different dates circulate, and v1 and the SSRN abstract still mischaracterise the critics' claim, so a minor documentation issue survives (F7).

Audit notes

What reproduced. Reproduced exactly: the worked example (50% votes; $50m default; $15m appraisal; $750,000 tax); the Forbes decomposition of Zuckerberg's $220.73bn (unlisted Class B valued at the META price); the instalment arithmetic (nominal 1.15x, PV 1.00 at 7.5%); codex's Meta end-2025 diagnostic (formula value $622.0bn; Class B pool 61.06% of votes; formula value 1.678x listed-price parity; 8.39% rate); codex's pro-rata-floor sensitivities (5.56%-8.33%); the Tax Foundation Table 1 arithmetic (5% x voting share x market cap; totals / (1-0.371)); and auditor A's base share (documented unlisted Class B $464bn = 20.0% of the authors' scored base; 12.4% excluding Page and Brin). Reproduced closely: auditor A's Zuckerberg alternative-reading rates, 9.8-10.6% under the (c)(3)(E) formula (depending on the FY2026 profit estimate) and 22.6% with the market-cap floor (A: 10.6% and 22.5%); and the PRE 14C dilution algebra, where '53%' implies selling about 81.6m Class B (23.9%), $52.5bn gross, about 13-15% of wealth after capital-gains tax (A: 84.1m, 13.4-24.5%), while the Special Committee's ~$20bn matches the memo-reading liability ($20.2bn gross). New: the January v1 text was obtained and diffed against v2 (97.8% identical); the official Prop 40 ballot text was checked against Amdt. 1 for the provisions at issue; the authors' drafting role was confirmed from a public source. Not reproducible: the SSRN 16 Feb 2026 revision and the Yale document cited for 'canons of statutory interpretation' (HTTP 403, not archived). The memo has no data or estimates to replicate beyond its hypothetical example.

Limits of this audit. This is a legal-interpretation adjudication by a non-lawyer. How the FTB or courts would rule is a prediction, not a finding; California doctrine was checked only against the text of Taxpayers to Limit Campaign Spending v. FPPC fn. 10, and the drafters' reliance on Rossi v. Brown was not verified. The SSRN 16 Feb 2026 revision and the Yale source cited for canons could not be obtained. The stakes figures are conditional illustrations, not predictions. They use FY2026 profits partly estimated from nine months, Q3-2026 book value, 25 Oct 2026 Forbes prices instead of the 31 Dec 2026 valuation date, and voting shares as of 1 Oct 2026, and they ignore a (c)(3)(F) rebuttal, Optional Deferral Account deferral (whose interaction with the voting presumption is itself contested) and penalties. Page's and Brin's residency on 1 Jan 2026 is not established. The founders' LLCs' accounting is unknown, so F4's consequences are reasoned, not computed. The Prop 40 ballot text is image-only and was checked visually for the operative provisions only. The accuracy of the (c)(3)(E) formula as a valuation method (e.g., relative to the Swiss practitioners' method the authors cite elsewhere, which weights capitalised earnings and net assets 2:1 rather than adding full book value) lies outside this memo's claims and belongs to the Prop 40 expert-report record.