How to value a carbon credit you cannot trust

Start with the position the controller is now in. Under ASC Topic 818, created by FASB's ASU 2026-02 on 19 May 2026, voluntary carbon credits are carried at cost less impairment at each reporting date, with a fair-value election that runs changes through earnings, and the impairment is irreversible. Whatever else that means, it means a number has to be produced, defended to an auditor, and stood behind on the face of the accounts. So the practical question is no longer whether to value the book. It is how, and with what input.

The honest starting point is that the asset is one you cannot trust at face value. The grade behind it was frequently commissioned by the seller. The dispersion within the book is enormous, with high-quality removal credits above twenty-six dollars and the weakest tier near three. And the thing being valued is not a financial claim with a contractual cashflow but an environmental assertion about a physical process that may or may not still be happening. Valuing it properly is not a matter of finding the right price feed. It is a matter of understanding what fair value and impairment actually demand, and then noticing that none of the inputs the market reaches for first will satisfy them.

What the standard actually asks for

01 · Impairment is a recoverability test, not a price lookup

Impairment is not the same as marking to the last screen price. It is a judgement about recoverable value: what the asset can actually deliver to its holder, tested against its carrying amount, at each reporting date, irreversibly once taken. For a carbon credit, recoverable value is a function of whether the underlying tonne still exists and is still stored, whether the credit can still be used for the purpose it was bought for, and whether the market will still accept it at all. A price observed in a trade tells you what one counterparty paid on one day for one parcel. It does not tell you whether the project behind your specific vintage has since suffered a reversal, been reclassified out of CORSIA eligibility, or lost the claim it was bought to support. The standard asks a recoverability question. A price is an answer to a different question.

02 · Fair value is an exit-price concept, and exit is conditioned

Fair value, where elected, is an exit price: what you would receive to sell the asset in an orderly transaction at the measurement date. For carbon that exit is heavily conditioned. The same nominal tonne is worth radically different amounts depending on its rating, its vintage, its method, its geography and its claim status, and those conditions move independently of any headline index. An exit-price estimate that ignores the credit-specific conditions is not conservative. It is wrong, and it is wrong in a direction the auditor is now obliged to probe, because gross presentation puts the asset next to the obligation and a fair-value election runs the difference through earnings.

Why the obvious inputs fail

03 · A broker quote is a transaction, not a valuation

The instinct of a treasury team under deadline is to call the broker who sold the credits and ask what the book is worth. The problem is structural and it should be familiar from every other asset class where this mistake has been made. A broker quote is a price at which a specific intermediary might transact a specific parcel, supplied by a party with a commercial interest in the relationship continuing. It is the carbon equivalent of valuing an illiquid bond book at the originating dealer's mark. It may be a useful market data point. It is not an independent, recoverability-tested, audit-defensible input, and an audit committee that has read the FASB analysis will recognise it as the sell-side mark it is. The lesson the financial system already paid to learn is that you do not let the party that sold you the asset tell you what it is now worth.

04 · A registry serial proves existence, not value

The next instinct is to lean on the registry. The serial number proves the credit was issued, identifies the project and vintage, and confirms the credit has not yet been retired or transferred. All of that is necessary. None of it is valuation. A registry serial is a record of issuance and chain of custody. It is silent on whether the project has since suffered a fire, a reversal, a permanence failure or a methodology challenge. It is silent on whether the vintage has been reclassified by ICVCM or dropped from CORSIA eligibility. A serial number tells you the asset exists and is yours. It tells you nothing about what it is worth, and the gap between those two facts is precisely where the impairment lives.

A serial number tells you the asset exists and is yours. It tells you nothing about what it is worth. The gap between those two facts is where the impairment lives.

05 · A project rating does not aggregate to a book number

The most sophisticated instinct, and the one most likely to mislead a numerate team, is to take the project-level ratings and roll them up. Independent ratings (BeZero, Sylvera, MSCI Carbon Markets, Calyx Global) are genuinely useful and a buyer should interoperate with them rather than fight them. Seventy-nine per cent of buyers already require a BBB-equivalent grade or better. But a project rating is an opinion about a project's likelihood of representing what it claims. It is not a price, it is not denominated in money, and it does not aggregate to a book-level fair value or a recoverable amount.

The aggregation problem is real and it is the kind of thing finance professionals are trained to spot. Ratings are ordinal opinions on heterogeneous criteria; a book is a money number. Translating a distribution of ordinal grades across different vintages, methods and geographies into a single defensible carrying amount requires a model that maps quality to recoverable value, that handles vintage-specific and method-specific dispersion, that reflects claim and target eligibility under SBTi V2 and the EU claims regime, and that updates when a reclassification event occurs. The ratings are an input to that model. They are not the output, and presenting a weighted average of letter grades to an auditor as though it were a valuation is a category error.

Why a one-off number is the wrong shape

There is a deeper problem than which input to use, and it is about timing rather than source. The events that destroy recoverable value (a reversal at the project, a permanence failure, a CORSIA delisting, an ICVCM reclassification, a successful greenwashing challenge to the claim the credit supported) do not occur on reporting dates. They occur continuously and arrive without a watchlist. Under irreversible impairment tested at each reporting date, the question is not what the book was worth at acquisition. It is what has happened to every position since, and whether any of it has crossed the line from a risk into a loss.

A valuation produced once, by hand, at year-end, from stale inputs, cannot answer that. It is a photograph of an asset that is moving. What the standard implicitly demands, even though it does not say so in these words, is a continuously maintained view of each position's condition, so that when the reporting date arrives the impairment question can be answered from a record that has been tracking the relevant events all along rather than reconstructed in a panic from broker emails and registry screenshots.

What an audit-grade mark actually looks like

Assemble the requirements and the shape of the right input becomes clear. It has four properties, and each one is a direct answer to a failure above.

First, it is independent: produced by a party that does not sell the credits, does not rate for both sides, does not transact, and is paid by the buyer rather than the seller. This is the 2008 lesson applied. The party relying on the number must be able to obtain it from someone with no interest in it being high.

Second, it is evidence-pinned: every position's mark traces to specific, citable evidence about that specific position's condition, its rating, its vintage, its method, its geography, its claim and target eligibility, and the state of the underlying project. Not a headline index applied uniformly, but a position-level recoverability view an auditor can follow back to its sources.

Third, it is continuously updated: the underlying condition is monitored between reporting dates so that reversals, permanence failures and reclassifications are captured as they happen, not discovered at year-end. The mark is a live field on each unit, not an annual verdict.

Fourth, it is aggregation-correct: it carries a model that maps the distribution of position-level quality and condition into a defensible book-level carrying amount and impairment view, presented gross, aligned to ASU 2026-02, and structured so the audit committee sees the asset and the obligation as the standard requires.

The mark is a live field on each unit, not an annual verdict.

Notice what this is not. It is not another carbon data platform, and it is not a ratings agency competing with the grades a buyer should keep using. It interoperates with those grades and turns them, together with monitoring and the accounting structure, into the one thing the controller actually has to produce and the auditor actually has to test: a recoverable, defensible, position-level mark on the book. That product does not exist in the market as a product today. The Big Four and the larger mid-tier firms have all published interpretation of ASU 2026-02, which means they own the advisory, but advisory is a memo at a point in time. None of them sells the continuous, independent, evidence-pinned mark the standard now requires every reporting date.