Markets & Institutions · Valuation

Who Sets the Mark?

Valuation as a governance decision disguised as measurement — one discretion, three regimes, no common standard

Julian Gretzinger · November 4, 2026 · Substack
Abstract

Where no transaction supplies a price, somebody decides one. Accounting meets the decision in Level 3 fair value; banking in the measurement basis a classification selects; fund regulation in net asset value. These are not one legal mechanism. This essay argues they expose one governance problem under three names — how an institution assigns an economically consequential number to an asset whose observable price is incomplete, stale or absent — and that in each regime the pen is held, directly or at one remove, by a party whose compensation is a function of the number. That is not an accusation of fraud. It is an observation about architecture, and the architecture recurs wherever institutional and retail capital — the savings pool — is currently being redirected.

The honest core of the counterargument is conceded early: illiquid marks are genuinely indeterminate, and discretion is not corruption but the absence of a price. Precisely for that reason, the essay claims, valuation is pure governance — the interesting questions are who holds the pen, who can override it, what happened the last time a mark met a real transaction, and what changes when tokenisation turns a quarterly decision into a continuous feed with liquidation consequences.

A price is what someone paid. A mark is what someone decided — and the decision has an owner.

#valuation  #nav  #private-markets  #accounting  #governance

I

Three Regimes, One Discretion

The accounting version is the fair value hierarchy: Level 1 for quoted prices in active markets, Level 2 for observable inputs, Level 3 for unobservable ones — which is to say, for the entity's own assumptions, run through the entity's own model, disclosed in a note (IFRS 13; ASC 820). The banking version is classification: a qualifying debt security is carried at amortised cost rather than fair value when the business-model and cash-flow tests are met (IFRS 9; ASC 320). The bond's market value does not change; the measurement basis does — and the carrying amount can diverge from the observable exit price by enough that, as an earlier essay showed in detail, the divergence becomes institutionally fatal when it meets a deposit run. The fund version is net asset value: under AIFMD the manager must maintain a valuation function that is "functionally independent" from portfolio management, or appoint an external valuer, and the resulting number is the price at which investors enter, exit, and pay fees (Directive 2011/61/EU, Art. 19).

Strip the vocabulary and the three regimes face the same problem: where the market supplies no complete price, an institution must produce one, under rules that constrain the method and barely touch the incentive. They are not the same instrument — Level 3 is a measurement category, amortised cost a measurement basis, NAV a computation — but they pose the same governance question: who controls the consequential number the rest of the system consumes. That question — not the models — is the subject of this essay.

The discretion is largest where the market is silent, but it does not vanish where the market speaks. For a liquid book the choices shrink to source, timing, side and basis — and basis alone is enough: a constant-NAV money market fund holds Treasury bills with a price every second and publishes 1.00, carrying them at amortised cost while a shadow value moves beneath (Regulation (EU) 2017/1131, Art. 29–33). The rest of this essay dwells on the illiquid case because that is where the savings pool is going; the governance question is the same at both ends, differing only in how much room the pen has.

II

The Pen and the Paycheck

Follow the compensation. A private markets manager charges management fees on committed or net asset value and carries a performance share crystallising against the same figure; the mark is set by the manager's own valuation committee, reviewed by an auditor who audits the model rather than the asset. A bank's measurement choices feed the capital position that determines what it may distribute — and the choices are made, within the classification tests, by the management whose compensation those distributions include. A semi-liquid fund's inflows, the raw material of its fee base, track the smoothness of the very NAV series the manager produces. In each case the loop is closed: the mark-setter's payoff is a function of the mark.

Where a regime addresses the conflict explicitly, it does so through procedure — functional independence, external valuation, oversight, disclosure, audit — and each safeguard is thinner than its name. The "functionally independent" internal valuation function reports, ultimately, to the same management whose carry it computes. The external valuer, where appointed, is engaged and paid by the manager, and the directive routes its negligence liability to the manager; the liability that faces the fund and its investors is the manager's own, expressly non-delegable — held, that is, by the marked party itself (AIFMD Art. 19(10)). The auditor tests the estimate's assumptions and evidence against the reporting framework; what no audit does is certify an inherently uncertain estimate as the objectively correct price — a limitation that survives every cycle because it is, strictly, correct. Independence in valuation is real at the level of procedure and fictional at the level of economics, and the regimes' architects have chosen, again and again, to regulate the procedure.

The regimes know the mark-setter is paid on the mark. They regulate the procedure and leave the payoff alone.
III

The Honest Core

Now the counterargument, stated at full strength because most of it survives. Illiquid assets do not have a concealed true price that discretion is hiding. A loan to a mid-market company, a stake in a private business, a building — these trade rarely, in lumpy sizes, between counterparties whose circumstances contaminate every observed print. The indeterminacy is real. A regulator who abolished discretion tomorrow would not reveal prices; it would merely transfer the discretion to whoever wrote the replacement formula. Marking is not corruption. It is the unavoidable administration of an absence.

But concede that fully and the essay's claim gets stronger, not weaker. If no true mark exists, then valuation cannot be measurement even in principle — it is only governance. The number is the output of a process, and the process has an owner, a payoff, an appeal mechanism, and a track record. Those are governance facts, checkable today, and they matter more than the model precisely because the model is checking nothing external. The right question about any mark was never "is it accurate?" — there is no fact for it to be accurate about. The right question is: who set it, what do they earn from it, who can override it, and when did it last meet a transaction.

IV

When the Mark Meets a Transaction

Because transactions are the only test that exists, the places where marks and transactions collide are the empirical core of the subject — and the recurring pattern points one way. Secondary sales of private fund stakes can clear at material discounts to carried marks, with the discount widening exactly when liquidity is most demanded; listed vehicles holding private assets have traded below their own published net asset values through long stretches of the cycle. Premiums occur, and the gap has more inside it than staleness — liquidity, financing, costs, selection. None of that rescues the mark: it proves the smaller, sufficient point that a reported NAV is not an executable exit price. Open-ended funds holding unsellable assets meet redemption waves by gating — the United Kingdom's property funds after the 2016 referendum and again in 2020, an equity income fund whose illiquid tail forced suspension in 2019 — and every gate is evidence, filed in real time, of the gap between the liquidity the redemption mechanism promised and the liquidity the portfolio could supply — not proof, by itself, that the NAV was misstated, but proof that NAV and immediate cashability are different properties (FCA reviews of open-ended property funds, 2016–2020; FCA report on the Woodford suspension, 2019).

The gate deserves one more sentence than it usually gets, because its distributional logic is the essay's thesis in miniature. A fund redeeming at a stale mark is transferring wealth from the investors who stay to the investors who leave, in strict order of who read the situation first. Swing pricing, anti-dilution levies, and redemption fees are the industry's own acknowledgement of the point: mechanisms that adjust the "objective" NAV precisely because everyone operating the machinery knows the number bends under stress. The price of illiquidity is real; the mark's function is to decide, quietly, who pays it.

V

Volatility as an Output Choice

The most consequential product of the marking machinery is not any single number. It is a statistical property: smoothness. An appraisal-based or committee-based valuation series updates slowly and anchors on its own past — a phenomenon documented in the real estate literature for over three decades — so reported volatilities and correlations understate the economic ones by construction (Geltner, 1991, and the appraisal-smoothing literature). The understatement is then sold. "Private assets are less volatile" is, to a first approximation, a statement about measurement frequency wearing the costume of a statement about risk; Cliff Asness has spent years calling the phenomenon volatility laundering — a phrase adopted here because it captures how a smoother reported series makes risk look lower without making the exposure less risky, whether or not anyone intended the effect (Asness, AQR, 2023).

The laundering has buyers as well as sellers, which is why it persists. An allocator whose own performance is judged quarterly rationally prefers assets whose reported path is smooth — the smoothness launders the allocator's career risk along with the asset's volatility. An earlier essay argued that volatility was always a defective proxy for risk; the marking regimes weaponise the defect: they let an asset class purchase a better risk measurement without purchasing better risk. The Sharpe ratio of a smoothed appraisal series is partly a property of the measurement process — a design choice wearing the costume of a datum.

VI

The Mark Becomes Infrastructure

Everything above is the settled critique. What changes now — and the reason to write the essay in 2026 — is that the mark is being promoted from a reporting artefact to a piece of live infrastructure. Tokenised money market funds publish NAV as an on-chain feed; tokenised feeders on private credit and semi-liquid vehicles are marketed with continuous or daily pricing; and once a token has a machine-readable value, it gets used as collateral — pledged, margined, and liquidated against that value, automatically. The shortening of money examined what continuous settlement does to funding; this is its valuation counterpart: the quarterly governance decision becomes an oracle, and the oracle acquires settlement consequences.

The promotion changes the stakes without, by itself, changing the epistemology. Where the feed is a signed republication of the manager's own estimate — the prevailing design — a mark published every block is the same committee judgement, resampled. Where tokenisation brings genuinely independent pricing sources or multiple valuation agents into the loop, something does improve, which is why the operative questions are who originates the value, who can amend it, how staleness is handled, and which automated actions consume it. Either way, the number can now trigger liquidations, margin calls, and cascade dynamics that quarterly marks never could. Higher frequency also manufactures false precision: a smooth daily series looks like information where a quarterly one looked like an estimate, and the collateral machinery downstream cannot tell the difference. The same discretion, wired into automated enforcement, is discretion with a detonator. The regimes built their procedural safeguards for a tempo at which a bad mark worked slowly — through reports, covenants, redemption cycles. Tokenisation moves the same judgement to machine speed, and nothing in the safeguards was built for the speed. The change is arriving by product design.

VII

Who Checks

Survey the checkers and a pattern emerges. The external valuer is paid by the manager and liable to the manager. The auditor attests process, disclaims the number, and is paid by the entity whose number it is. The depositary, under the fund directives, oversees that valuation procedures exist and are applied — a control on form. The administrator computes NAV from prices the manager's hierarchy supplies. And the newest checker, the on-chain oracle, is in most current designs a signed republication of the issuer's own figure — the manager's mark with a cryptographic costume, verifying provenance and nothing else. At every station, the checker is either paid by the marked party or is checking the model rather than the number. The liability that does face the holder is the manager's own non-delegable responsibility for the valuation — the investor-facing station sits inside the marked party. What the stack lacks is an economically independent checker, with liability to the holder, confirming the value itself — the station the apparatus is assumed, by its users, to contain.

VIII

Who Benefits From the Gap

The incentive channels run the length of the chain, roughly in order of proximity to the pen. Managers: fees on the number, carry against the number, fundraising on the smoothness of the number. Allocators: reported stability that launders career risk, purchased with their beneficiaries' actual risk. Platforms, newest: a continuous NAV feed is the raw material for trading, lending, and collateral products, and the platform's revenue scales with activity against the mark, not with the mark's fidelity. The holder pays at the two moments the machinery cannot smooth — entry and exit against a stale number, and forced liquidation against an automated one.

The diagnostic, as ever, fits in three questions. For any unlisted or semi-liquid instrument: who sets the mark, and what does that party earn as a function of it. Who can override it, with what liability, owed to whom. And when did the mark last meet a transaction — a secondary sale, a redemption paid in full, a liquidation — and on which side of the mark did the transaction print. No model review required. The answers are governance facts, and they are all, today, either in the documents or conspicuously absent from them.

—

The Number Has an Owner

A growing share of financial claims is priced by decision rather than by transaction — that is, mechanically, what the migration of capital into private credit, semi-liquid vehicles, and tokenised feeders implies. The decision is not going away, because the indeterminacy it administers is not going away. What can change is whether the decision is treated as what it is. Measurement invites model review. Governance invites harder questions: interest, override, track record.

A price is what someone paid. A mark is what someone decided. The instruments now absorbing that capital are, to a growing degree, marked rather than priced — and the useful discipline, for anyone holding them, is to stop asking whether the number is accurate and start asking who owns it.


Sources

Secondary-market discount and gating patterns are characterised qualitatively from the public record; no individual fund's marks are alleged to be misstated.

Written in a personal capacity. Analytical views only — not legal or investment advice.

Julian Gretzinger — Investor and writer on monetary history, real wealth mechanics, and financial markets. substack.com/@juliangretzinger