ERC-20 Token Valuation for Accounting: Which Price to Use

A market price is not automatically an accounting fair value

Crypto markets publish prices continuously, which can create the impression that valuation is trivial. It is not. ERC-20 token valuation changes the problem from finding a number to proving that the number is the one the accounting framework requires, at the right market, the right instant and with the right evidence.

The accounting question is not merely:

What was LINK trading at?

It is:

Which market is relevant under the applicable framework, what price in that market represents an orderly exit transaction at the measurement time, and what evidence supports that conclusion?

For a highly traded token, several venues show slightly different prices at the same second. For a thin token, the quoted price may come from stale trades, wash trading, a tiny liquidity pool, or an index assembled from markets the entity cannot access.

The number is only as defensible as the market-selection and evidence process behind it. Under US GAAP that process now runs at every reporting date for an in-scope crypto asset rather than only when an impairment indicator appears: ASC 350-60-65-1(a) makes fair value measurement effective for fiscal years beginning after 15 December 2024, including interim periods within those fiscal years, so for a calendar-year entity every question in this article is live from 2025 onwards.

Which standard applies to the token in the first place is settled earlier, by the ERC-20 scope test.

ERC-20 token valuation starts with the accounting framework, not the data vendor

IFRS 13 paragraph 9 defines fair value as "the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date". The AICPA renders the ASC 820 definition in the same words, as "the price that would be received to sell an asset ... in an orderly transaction between market participants at the measurement date".

A fair value measurement then assumes the transaction takes place in the principal market for the asset or, in the absence of a principal market, in the most advantageous market (IFRS 13 paragraph 16). Where a principal market exists, the measurement must represent the price in that market "even if the price in a different market is potentially more advantageous at the measurement date" (IFRS 13 paragraph 18). An entity does not survey venues after the period end and pick the most flattering quote.

The entity therefore should not begin with:

CoinMarketCap says $X

or:

our oracle says $Y

It should begin with:

What is our principal market for this asset?

Then select a price consistent with that market and the measurement rules.

Principal market is entity-specific

The principal market is not necessarily the largest crypto exchange in the world. The analysis considers the market with the greatest volume and level of activity for the asset that the entity can access, and the entity must have that access at the measurement date (IFRS 13 paragraph 19), which is why the same asset can have different principal markets for different entities.

An entity may have access to one exchange and not another because of:

• jurisdiction
• onboarding restrictions
• institutional account status
• custody integration
• sanctions controls
• trading permissions
• withdrawal restrictions
• credit or counterparty limits

A market the entity cannot access is not available for the principal-market analysis merely because its public price feed is visible.

US GAAP then adds a presumption the preparer cannot skip. Under ASC 820-10-35-5A, as the AICPA practice aid puts it at Response 16, "the market in which an entity normally transacts for the digital asset is presumed to be the principal market, unless contrary evidence exists", and "to overcome the presumption, an entity must obtain evidence that the market it normally transacts in is not the market with the greatest volume and level of activity for the digital asset". An entity that trades exclusively on one venue does not get to price its holding on a larger venue it has never used without that evidence.

Use bid, ask, last, midpoint, or close?

There is no universal crypto rule saying every token must be valued at the last trade, midpoint, or daily close. The policy has to be consistent with the fair-value standard, the market's structure, and the entity's facts.

Questions include:

• Is the quoted market active?
• Is the bid-ask spread narrow or wide?
• Is the last trade recent and representative?
• Is the measurement point an instant or a reporting-date close convention?
• Is the quoted price executable or merely indicative?

One question the preparer does not get to ask is whether the position is too large for the market. Fair value is measured for the unit of account, and for a crypto intangible asset that is generally the individual unit or a divisible fraction of one. A blockage factor, meaning a discount applied because the market's normal daily trading volume cannot absorb the quantity held, is not permitted in a fair value measurement (IFRS 13 paragraph 69), and where the asset is traded in an active market the measurement is the quoted price for one unit multiplied by the quantity held, "even if a market's normal daily trading volume is not sufficient to absorb the quantity held and placing orders to sell the position in a single transaction might affect the quoted price" (IFRS 13 paragraph 80). ASC 820-10-35-36B and ASC 820-10-35-44 carry the same two rules, and the AICPA answers the question for digital assets at Response 19 with one word, No, adding that the portfolio exception in ASC 820-10-35-18D "is not applicable to crypto intangible assets". The prohibition is categorical in both frameworks. There is no gap in which a size discount becomes available.

Market depth remains relevant evidence when evaluating whether observed transactions are orderly and whether a market is active. It is evidence about the market, not a deduction from the price.

Three prices can all be real at the same time

The research paper behind this series, ERC-20 Accounting: How to Account for a Plain Fungible Token, Allowances and Gas, read three sources for LINK inside a two-second window on 9 September 2026, from 21:52:22 to 21:52:24 UTC. The timestamps the three sources themselves carry are earlier, and by different amounts, which is part of the point.

Three publicly available, defensible numbers for the same token, read in the same two seconds, spanning 0.27 per cent. None of them is wrong. Each reflects a different venue or convention: a centralised order book's last executed trade, an automated market maker's current marginal rate, and an aggregated oracle answer that was already three and a half minutes old when it was read.

Applied to the 1,777.579813 LINK holding in the research paper's worked example, the highest and the lowest of those three prices are about $57 apart. On a position a hundred times larger the same 0.27 per cent is roughly $5,700 of reported assets, decided by which source the preparer happened to open. That is why the choice has to be made by policy in advance.

The defensible accounting price depends on:

• the principal market conclusion
• the measurement policy
• the timestamp convention
• the quality of the observation

This is why valuation workpapers should preserve the raw market evidence, not merely the final price copied into the journal entry.

Thin order books create a different problem from no market

A token can trade on a recognised venue and still have poor liquidity. Warning signs include:

• a wide bid-ask spread
• very small size at the best bid and ask
• infrequent trades
• large gaps between price levels
• abrupt prints that reverse immediately
• a large difference between spot venues
• a large difference between centralised and decentralised venues

Those facts do not invalidate the market. They increase the need for evidence that the observed price represents an orderly transaction and an active market where the applicable standard requires one.

A decentralised exchange price is not one simple number

Automated market makers such as Uniswap do not operate like a conventional order book. A pool contains reserves or concentrated liquidity positions, and the marginal exchange rate changes as a trade moves through the pool.

A displayed decentralised exchange price, commonly shortened to DEX price, can therefore mean several different things:

• the current marginal pool price
• a time-weighted average price
• the result of a small quote
• the result of a trade of a specified size
• an aggregator's routed price across multiple pools

For accounting, the preparer should record exactly which one was used, together with the pool identity, the token contract identities, the pool fee tier, the pool's liquidity, the timestamp or block, whether the pool was subject to manipulation or abnormal conditions, and whether the entity can access the venue.

Do not value by ticker

A market data vendor can return a price for "LINK". The accounting system must still prove that the vendor's asset identifier corresponds to the same chain and contract as the entity's holding. This matters because bridged versions trade separately, counterfeit contracts copy the symbol, tokens migrate contracts, wrapped versions use similar names, and a vendor may aggregate multiple markets or networks.

The valuation master should map the pair that actually identifies an ERC-20 token under the ERC-20 standard, which is the chain and the contract address:

chain id 1 (Ethereum mainnet) + 0x514910771AF9Ca656af840dff83E8264EcF986CA

to the vendor's instrument identifier, and retain evidence of that mapping.

Oracles are useful data, not automatic accounting authority

A decentralised oracle can be excellent evidence for smart-contract execution while still being unsuitable as the sole financial-reporting valuation source. An oracle may aggregate multiple exchanges, use delayed or time-weighted values, apply deviation thresholds, update only when certain conditions are met, or be designed for protocol safety rather than accounting cut-off precision. The Chainlink answer in the table above carried an updatedAt three and a half minutes before the read, which is normal behaviour for a deviation-threshold feed and a problem for a measurement fixed to an instant.

The question is not whether the oracle is reputable. It is whether its methodology produces the value required by the accounting framework at the measurement date. The same applies to market-data aggregators, benchmark administrators, pricing services and exchange indexes. The accountant needs to understand the methodology well enough to know what number is being imported.

Wash trading and disorderly transactions matter

Stakeholders told the FASB during the ASU 2023-08 project that because crypto asset markets are largely unregulated, "certain transactions that are not orderly (for example, wash trades that are intended to manipulate prices) may appear as orderly transactions and could distort an entity's fair value measurement". Those stakeholders asked the Board to preclude fair value measurement of crypto assets without existing active markets, and the Board rejected that alternative and pointed preparers back to Topic 820 (ASU 2023-08, BC37 and BC40). The concern is a stakeholder concern that the Board recorded and did not adopt, not a standard-setter finding, and it is not authority for departing from an observed price. It remains a real risk a valuation policy has to address.

Reported crypto volume is not automatically evidence of genuine market activity. A valuation policy should therefore include escalation criteria such as:

• extraordinary volume with little price movement
• repeated self-trading patterns where data is available
• sudden divergence from major venues
• obvious manipulation events
• exchange outages
• withdrawal freezes
• extreme spreads
• market suspensions

The objective is not to become a market-surveillance firm. It is to avoid treating any visible print as unquestionably representative.

Fair-value hierarchy classification needs evidence

Both IFRS 13 and ASC 820 sort the inputs into three levels. Level 1 inputs are quoted prices, unadjusted, in active markets for identical assets that the entity can access at the measurement date (IFRS 13 paragraph 76). Level 2 inputs are other inputs that are observable, directly or indirectly (paragraph 81), and Level 3 inputs are unobservable (paragraph 86). The measurement is categorised in its entirety at the lowest level of input that is significant to it (paragraph 73).

A directly quoted, unadjusted price in an active market for an identical asset that the entity can access is a Level 1 input, and it is not an option. IFRS 13 paragraph 77 requires a quoted price in an active market to be "used without adjustment to measure fair value whenever available", subject only to the narrow exception in paragraph 79. ASC 820-10-35-44 then measures the position as that quoted price multiplied by the quantity held, and the AICPA practice aid at Response 18 states that, except in the circumstances identified in ASC 820-10-35-41C, "there should be no adjustment to Level 1 inputs, and the fair value of the digital asset should be determined based on price times quantity". The same response carries the note that "it is not appropriate to use a volume-weighted average price (VWAP) or other types of aggregated pricing when measuring the fair value of digital assets traded in active markets".

That matters for the inputs preparers reach for when the direct quote is inconvenient. A price taken from a market the entity cannot access, an index, a blended vendor price, an adjusted exchange price or a DEX-derived model is not an unadjusted quote in the principal market. For a token that trades in an active market, an aggregated price is not simply a lower level in the hierarchy, it is the wrong input. Where no active market exists, the measurement generally falls to Level 2, and to Level 3 where an unobservable adjustment is significant to it.

What if there is no active market?

Many ERC-20 tokens have little or no meaningful trading activity. The preparer should resist the temptation to manufacture a precise number from a stale website quote.

One fallback ladder must be defined in advance, and one only. The research paper's is:

1. Use the principal market's observation for the event minute.
2. If absent, use the most recent populated minute at or before the target on the same product and venue, subject to a documented maximum look-back.
3. If still unavailable, use a second accessible venue identified in advance, while documenting that this is no longer the principal market's direct price.
4. For a reporting-date measurement, consider an independently administered benchmark with the appropriate fair-value-hierarchy analysis.
5. If measurement cannot be supported reliably, escalate under the applicable accounting framework rather than invent a price.

Silent interpolation and a silent switch to a coarser candle are both excluded, because both change the nature of the input, and a change in the nature of the input is a change in the hierarchy level.

The important part is not that every entity must use this exact ladder. It is that the entity defines its ladder before a problematic reporting date rather than selecting the most convenient price afterwards.

A real missing-minute example

The research paper tested several products on one venue for the minute beginning 31 December 2025 at 23:59:00 UTC. ETH-USD and LINK-USD each returned a populated one-minute candle. ETH-GBP and LINK-GBP each returned no candle at all for that minute.

A product can exist, be online, and still have no trade in the exact reporting minute. A price policy tested only on a deep USD pair may fail when applied to a thinner GBP or EUR pair, which is exactly the case an entity with a non-USD functional currency is in.

The opposite failure is worse, because it looks like data. For the same minute, Bitstamp's ETH-USD one-minute row reports a close of 2,966.62 on a volume of 0.000000, carrying the previous price forward as a flat candle, while the Coinbase Exchange candle for the same minute closes at 2,966.84 on a volume of 27.14663102 ETH. Coinbase omits a minute entirely when nothing trades. An accountant comparing the two venues without noticing this would conclude that Bitstamp was the more liquid venue, which is the reverse of the truth, and would report a price at which nothing changed hands. A zero-volume price is not a trade. It is the interface saying that nothing happened, and a parser that does not read the volume field cannot tell the difference.

Timestamp policy matters in a 24/7 market

Crypto does not close at 4 p.m. on a national exchange. The entity should define a reporting-time policy such as:

23:59:59 UTC on the reporting date

or another time tied to the reporting framework and ledger cut-off. The policy should address the time zone, daylight-saving changes where local time is used, exchange candle conventions, whether a "daily close" represents the beginning or the end of a minute or a day, and how missing observations are handled.

A one-minute mismatch can matter during volatile periods. The same convention problem on native ETH is worked through in Ethereum Valuation for Accounting.

Transaction costs are not the same as fair value

The price in the principal market shall not be adjusted for transaction costs, because transaction costs "are not a characteristic of an asset or a liability" but are specific to a transaction (IFRS 13 paragraph 25). Transport costs are treated differently: where location is a characteristic of the asset, as it may be for a physical commodity, the price is adjusted for the cost of transporting the asset to that market (paragraph 26). No specific guidance was identified applying that limb to a token held in a self-custody address rather than an exchange-controlled one. Searched: IFRS 13 in full, ASU 2023-08 in full, and the AICPA practice aid in full.

For a crypto transaction, trading fees, withdrawal fees, network gas and slippage each need separate accounting analysis, and none of them reduces the measured fair value.

The valuation team should therefore avoid embedding entity-specific transaction costs in the quoted fair value unless the applicable standard requires a different measurement basis. The one that does is IAS 2 paragraph 3(b), fair value less costs to sell, and it is narrower than it sounds. It applies to commodity broker-traders, who are "those who buy or sell commodities for others or on their own account", holding inventories "principally acquired with the purpose of selling in the near future and generating a profit from fluctuations in price or broker-traders' margin", wording the June 2019 IFRS Interpretations Committee agenda decision quotes in full from IAS 2 paragraph 5. An entity holding tokens for gain over an indefinite horizon is not that, and a treasury that reads "qualifying broker-trader" as describing itself has over-claimed the exclusion.

A reproducible price convention

The research paper chooses a concrete convention rather than writing "use Coinbase" and leaving the rest implicit.

For LINK-USD it uses the Coinbase Exchange one-minute candle close for the UTC minute in which the event occurred. For the reporting date, it uses the close of the minute beginning 31 December 2025 at 23:59:00 UTC. The observation is the fifth field of that candle, requested as GET https://api.exchange.coinbase.com/products/LINK-USD/candles?granularity=60&start=<minute>&end=<minute>, and Coinbase documents that the granularity value "must be one of the following 'second' values: {60, 300, 900, 3600, 21600, 86400}" (Coinbase Exchange, Get product candles).

That convention fixes four variables:

venue       = Coinbase Exchange
product     = LINK-USD
interval    = 1 minute
time basis  = UTC
observation = candle close

An ERC-20 token valuation policy should be that specific. "Use market price" is not reproducible.

The research paper also validates the candle-field order against the published schema rather than assuming it. Coinbase documents the candle as [timestamp, price_low, price_high, price_open, price_close], so a parser that assumes the conventional open-high-low-close order silently reads the low as the close. That kind of implementation check matters because the error is invisible in the output: every number is a real price from the right minute, and every one of them is the wrong field.

Retain the raw response, not only the resulting spreadsheet cell

A public API can change, delist a market, alter history retention, or disappear. The workpaper should preserve:

• the request parameters
• the raw response
• the retrieval timestamp
• the parser version where automated
• validation that the returned timestamp equals the intended period
• the transformation from raw source fields to the final valuation input

That lets a later reviewer reproduce the number without relying on a third-party service still behaving exactly as it did at the close.

A defensible valuation workpaper

An ERC-20 token valuation file is judged by whether someone else can reperform it. For each material token, retain:

• Asset identity. Chain ID, contract address, symbol and name, vendor instrument ID.
• Market selection. Markets accessible to the entity, volume and activity evidence, the principal-market conclusion and how the ASC 820-10-35-5A presumption was applied, and any change from the prior period.
• Price observation. Venue, exact timestamp, the bid, ask, last, midpoint or close convention, raw market data, source URL or data extract, and the conversion rate where the quote currency is not the reporting currency.
• Market-quality review. Spread, trading volume, depth where relevant, exchange status, abnormal events, manipulation concerns.
• Accounting conclusion. Fair value amount, hierarchy level if required, fallback method used if any, review and approval.

Accounting and tax can legitimately use different price conventions

Fair-value accounting asks for the price required by the applicable accounting framework, tied to the entity's principal market.

A tax authority may instead ask for a reasonable method applied consistently. The Canada Revenue Agency puts it in those terms, requiring a reasonable method for determining the value of a crypto-asset even where a direct value is not readily available, and asking that whichever method is chosen be used consistently from year to year with a record of how it produced the figure. The governing rule differs by jurisdiction and by transaction, and ERC-20 token tax: lots, gas and character takes the tax side further.

A single enterprise can therefore have one supported financial-reporting price and a different supported tax valuation for the same timestamp. That is not an inconsistency. It reflects two different legal measurement questions.

The accounting takeaway

The existence of a crypto price feed does not remove the ERC-20 token valuation problem. It changes the problem from "can we find a number?" to "can we prove this is the number the accounting framework requires?"

A sound process identifies the exact token by chain and contract, determines the relevant market and records why, defines the observation convention to the minute and the field, evaluates market quality, documents one fallback ladder in advance, and retains enough evidence for reperformance.

That is the difference between a dashboard price and an accounting valuation.

How Tokenbooks helps with valuing an ERC-20 token

ERC-20 token valuation in Tokenbooks resolves through a fixed, ordered chain, and the first match wins: a redemption rate for vault and share tokens so an accrual is not erased by a peg, then the underlying of a token that only represents another, then a native fiat rate, then a stablecoin peg where the portfolio treats stablecoins as base currency, then a market day close, then a pool-derived day close for long-tail tokens, and last the value implied by the other legs of the same transaction.

That last step is deliberately asymmetric. A value is inferred for an acquisition, never for a disposal, which has to stand on a market price or on historical basis. When nothing in the chain returns a price the engine does not invent one: it opens the lot at zero and flags the transaction, so an unpriced asset is a visible exception.

Every quote records which source produced it and at which valuation instant, stored with the transfer. A fair value can also be pinned per transfer, as a price per unit or as a total, and a hand-entered number is refused as a contributor to any other leg's derived price.

Start with the Ethereum integration page.

More in this series

Accounting Token Anatomy: ERC-20 Tokens. This article stands alone, but the series builds in order.

Previous (02.1.4): ERC-20 Token Scope: ASC 350-60, IAS 38 and Carrying Amount

Next (02.1.6): ERC-20 Token Journal Entries: A Complete Worked Example

Sources and further reading

• IFRS 13 Fair Value Measurement, IFRS Foundation 2025 annotated text, for paragraphs 9, 16, 18, 19, 25, 26, 69, 73, 76, 77, 80, 81 and 86. ifrs.org
• IAS 2 Inventories, IFRS Foundation 2025 annotated text, for paragraphs 3(b) and 5. ifrs.org
• IFRS Interpretations Committee, Holdings of Cryptocurrencies, June 2019. ifrs.org
• FASB ASU 2023-08, for ASC 350-60-65-1 and Basis for Conclusions paragraphs BC37 and BC40. storage.fasb.org
• AICPA, Accounting for and Auditing of Digital Assets, practice aid as of 30 September 2025, Fair value measurement Responses 16, 18 and 19 on principal market, price times quantity and blockage. assets.ctfassets.net
• Coinbase Exchange REST API, Get product candles, for the granularity values and the candle field order. docs.cdp.coinbase.com
• Coinbase Exchange candles endpoint, ETH-USD one-minute candle for 31 December 2025 23:59:00 UTC. api.exchange.coinbase.com
• Bitstamp OHLC endpoint, ETH-USD one-minute rows around 31 December 2025 23:59:00 UTC. bitstamp.net
• Canada Revenue Agency, Determining the value of crypto-assets for tax filing. canada.ca
• Ethereum.org, ERC-20 standard. ethereum.org This article is educational material. Fair-value conclusions require the full requirements of the applicable standard and entity-specific market-access facts.