Crypto Cost Basis Rules: Who Permits What

Introduction

A cost basis method answers one question: when you dispose of part of a holding built from several purchases, which units are treated as the ones that left? The answer sets the cost you relieve against the proceeds, and so sets the gain.

The question is universal. The answers are not, and the differences between jurisdictions are structural rather than cosmetic — one country pools by asset, another tracks by account, and software configured for one produces a materially wrong figure under the other.

This page covers what the rules actually say, with two specific corrections to claims that circulate widely. It does not tell you what to do: the honest output of comparing these systems is an understanding of which question your jurisdiction is asking, not a recommendation to pick a method.

Both corrections are about names rather than arithmetic. The first is that highest-in-first-out is not a method the United States regulations name. It is one way of exercising a power those regulations do grant — specific identification — and the distinction matters the moment a tool describes itself as HIFO compliant, because what has to be compliant is the identification and its timing, not the sorting rule that produced it.

The second is that the four-item list pairing it with first-in-first-out and last-in-first-out describes what tax software offers rather than what any tax authority has written. Neither of the two jurisdictions examined closely here publishes that list, and one of them has published nothing about method names at all. Once you know which of the two structures you are under, the list stops reading as a menu and starts reading as a description of a settings screen.

Everything below is read from primary text — the regulations, the manuals, and the revenue authorities' own pages — with access dates recorded in the sources section. Where a source has been superseded that is stated, along with how to tell: the single most-quoted United States page on this subject now carries a notice about itself, and a reader who has not seen that notice will go on citing a rule that changed in 2025.

Two things this page is not. It is not advice on which method to choose, because under a pooling system there is no choice to make, and under a selection system the choice is constrained by when you record it rather than by which rule you like. And it is not a guide to configuring any particular tool — the software question it does answer is narrower: what a method picker is for, and what it means when the figure it produces disagrees with the one a return requires.

What the different answers are worth in money, on one fixed set of trades, is shown in our method comparison.

The Question Every Method Answers

Buy the same token three times at three prices, sell part of the holding, and there is no observable fact about which units left. They are fungible and indistinguishable. The question has no answer you can look up in a transaction record; it has an answer a rule supplies.

Two families of answer

Broadly, systems answer in one of two ways. Selection systems pick particular units — the oldest, the newest, the most expensive, or ones you nominate. Pooling systems refuse the question: every unit carries the same averaged cost, so which one left does not matter.

Those are not variants of each other. Under selection, the order of your disposals changes your results, because spending cheap units early leaves expensive ones behind. Under pooling it does not, because the average is unchanged by a disposal.

The difference shows up most clearly in what each system needs you to know. A selection system has to keep track of individual acquisitions as distinguishable lots, because a disposal consumes particular ones. A pooling system deliberately destroys that information: once units enter the pool their separate acquisition prices are gone, merged into a single figure that no subsequent event can unpick. Neither approach is more accurate about the underlying reality, because the underlying reality has no answer to give — but they impose very different record-keeping obligations, and a history kept for one can be missing exactly what the other requires.

What no method changes

Whichever answer applies, the total cost relieved over the life of a holding is the same — every unit's cost is relieved exactly once, on the disposal that takes it. Methods move gains between tax years; they do not remove them. The method comparison demonstrates that on one ledger, with the arithmetic shown.

The United Kingdom: No Choice Is Offered

The UK answer is the shortest one on this page. A disposal is matched against same-day acquisitions, then against acquisitions in the following thirty days, and the remainder against the section 104 pool — in that order, with no selection anywhere in it.

What HMRC does and does not say about the method names

HMRC's Cryptoassets Manual does not name FIFO, LIFO or HIFO as options for individuals, and no page in it offers a choice of cost basis method.

That is worth stating carefully rather than dramatically. HMRC has not said it prohibits HIFO, and it has not said it does not permit FIFO — it has said nothing about those method names for cryptoassets at all. The defensible sentence is the structural one: the UK computation uses pooling, and specific-identification methods do not enter it.

What that means in practice

A UK individual has no election to make and no setting that legitimately changes the answer. Software offering a method choice is offering a choice of which country's rules to compute, and for a UK return only one of those options is correct.

The full mechanics — the matching order, the pool, what enters it as cost — are on our UK crypto tax reporting rules.

The United States: Two Regimes in One Regulation

The United States does offer selection, and the first thing to understand is that it offers it twice, under different conditions, depending on where the units are held.

Units not held in a broker's custody

For digital assets you hold yourself, basis is determined by specific identification of the units in that wallet. Where no specific identification is made, the units are treated as disposed of in order of time from the earliest date on which units of the same digital asset not held in a broker's custody were acquired.

Note the boundary carefully: the ordering runs from the earliest acquisition of units in that wallet, not across your whole portfolio. Calling this "FIFO across your holdings" describes something wider than the regulation does.

Units held in a broker's custody

For broker-held units the requirement is different in a way that matters enormously: the taxpayer must specify the units to the broker, no later than the date and time of the disposal. A standing order or instruction for specific identification counts as an identification made at the time of disposal.

Where no adequate identification is given to the broker, the units are treated as disposed of in order of time from the earliest date on which units of that same digital asset held in the custody of a broker were acquired — first in, first out, within that account.

Why keeping the two apart matters

Broker-held and self-custodied units do not follow the same rule. For broker-held units the identification must reach the broker; for self-custodied units your own books and records are the place it lives. A holder with assets in both places is operating under both rules simultaneously.

A temporary relief altered where the broker-held identification had to be recorded, and it has a defined end date — the detail is on our Form 1099-DA guide.

Specific Identification, and Its Timing Rule

Specific identification is the United States mechanism that everything else on the US side is built from. Its definition is precise, and one clause in it is responsible for more incorrect advice than any other sentence in this area.

The definition

An identification is made if, no later than the date and time of the sale, disposition or transfer, the taxpayer identifies on its books and records the particular units to be sold, disposed of or transferred — by reference to any identifier, such as the purchase date and time, or the purchase price for the unit.

"Any identifier" is genuinely permissive: the regulation does not prescribe a format. What it does prescribe, without flexibility, is when.

No later than the date and time of the disposal

This is the clause that most summaries omit and most software implies away. A taxpayer cannot choose lots at filing time, or when preparing the return, or in the spring after the year ends. The identification must exist no later than the moment of the disposal.

The rule is a rule against hindsight. Choosing which units to treat as sold once you know the prices, and know which choice produces the better outcome, is a different act from identifying them in advance — and the regulation's timing requirement is what separates the two.

What this implies about software that offers a method picker

A tool that lets you switch between FIFO and highest-basis at the end of the year, and recomputes, is showing you what the answer would have been had a contemporaneous identification been made. It is not making one retrospectively, and it cannot.

That does not make such tools useless — modelling the difference is genuinely informative, and the method comparison exists for exactly that. It makes the difference between modelling and complying worth keeping clear.

Standing Orders: How an Identification Persists

The timing requirement raises an obvious operational problem. If an identification must exist at or before every disposal, how does an active trader comply without performing a separate act hundreds of times a year?

The regulation answers it. For units held in a broker's custody, a standing order or instruction for the specific identification of digital assets is treated as an adequate identification made at the time of the disposal.

What that buys, and what it costs

It converts a repeated act into a single one. A rule set in advance — take the highest basis first, or the latest acquired — governs every subsequent disposal without further intervention, and each disposal is treated as identified at the moment it happens.

The cost is the symmetrical one. A standing order applies whether or not it suits a particular disposal, and it cannot be revisited after the fact for a trade you would rather have treated differently. Changing it changes what happens next, not what already happened.

The trade-off, stated plainly

Per-disposal identification is more precise and more demanding. A standing order is less precise and far more likely to actually be complied with. For an investment portfolio with a handful of disposals a year, the first is workable; for an active trading pattern, the second is often the only one that survives contact with reality.

Neither is better in the abstract, and neither is available to a United Kingdom taxpayer, where the question does not arise at all.

What a standing order needs to contain

The regulation does not prescribe a form of words, which leaves the content to the holder. What it has to do is determine an outcome without further input: a rule that requires a judgement at the moment of each disposal is not standing, it is a policy of deciding later.

So "take the highest basis first" determines an outcome; "take whichever is most advantageous" does not, because advantage is only knowable after the fact. The difference between those two phrasings is the difference between an instruction and an intention, and it is the same line the timing rule draws everywhere else in this area.

It also matters where the instruction sits. For units held in a broker's custody the standing order is something the broker holds, not something you have written in a private document — and a broker unable to accept such an instruction cannot hold one, which is the practical gap that made temporary relief necessary in the first place.

The Timing Rule, Illustrated

Because the timing requirement is where most of the practical difficulty lives, it is worth seeing the difference between two records that look equally tidy and are not equivalent.

A record made at the time

A holder disposes of units on a Tuesday. Before or at the moment of that disposal, their records identify which units are being disposed of — by purchase date and time, or by purchase price, or by any other identifier that picks them out.

Whatever happens to prices afterwards, the identification stands. It was made without knowing the outcome, which is precisely the condition the rule is written to require.

A record made in March

The same holder, same disposal, no contemporaneous record. In March, preparing the return, they run their software with several selection rules, observe which produces the lowest gain, and adopt it.

The resulting spreadsheet may be immaculate and internally consistent. It documents a choice made with hindsight, which is not an identification made no later than the date and time of the disposal. The quality of the documentation is not what the rule asks about.

Modelling is not the problem

None of this makes it improper to compute what different rules would produce. Understanding the spread is useful, and our method comparison exists to show it on real arithmetic.

The distinction is between using that knowledge to decide how to operate going forward — including whether to put a standing order in place — and using it to select an answer for disposals already made. The first is planning. The second is the thing the timing rule exists to exclude.

Holding in Both Places at Once

A United States holder with assets at a custodial venue and assets in self-custody is not operating under one regime with an exception. They are operating under two rules simultaneously, and the rules differ in the part that is easiest to get wrong.

Two identifications, two destinations

For the broker-held units, an identification must reach the broker — by instruction or by standing order. For the self-custodied units, it lives in the holder's own books and records. The same intention, expressed in the same words, satisfies one requirement and not the other depending on where it was recorded.

That is a genuinely unusual structure, and it is the reason a single well-kept spreadsheet is not automatically sufficient. It may be exactly right for half the portfolio and beside the point for the other half.

Two defaults, each with its own boundary

The defaults differ too, and neither spans the whole holding. Broker-held units without an identification run earliest-first within that broker account. Self-custodied units without one run earliest-first within that wallet.

So a holder with three accounts and two wallets has five separate default orderings, not one. Any summary that reduces this to "the default is FIFO" has dropped the boundary that makes the default computable.

And transfers move the question

Moving units from self-custody to a broker does not create a disposal, but it does move those units from one rule to the other — and from one default ordering to a different one. A transfer is therefore a bookkeeping event with tax consequences that arrive later, which is the pattern this cluster keeps returning to.

The reporting side of the broker relationship, including what the broker files about you and why its figures may disagree with yours, is covered in our Form 1099-DA guide.

HIFO Is Not a Method the Rules Name

This is the correction that most changes how the subject should be described, and it comes straight from the regulation's own words.

What the text actually says

The regulation refers to "a method of specifically identifying the units of a digital asset sold, disposed of, or transferred... for example, by the earliest acquired, the latest acquired, or the highest basis".

Read the structure of that sentence. Earliest-acquired, latest-acquired and highest-basis are given as examples of ways to exercise specific identification. They are not a menu of separate methods sitting alongside it. The acronym HIFO appears nowhere in the provision.

The four-item list that should not exist

So the familiar formulation — "the IRS allows FIFO, LIFO, HIFO and specific identification" — misdescribes the structure it is summarising. There is specific identification, which you may exercise by any consistent rule including highest-basis selection; and there is a default that applies when you have not made one.

Equally, "the IRS permits HIFO" and "HIFO is an IRS-approved method" both claim an approval of a named method that the text does not contain. The accurate statement is that highest-basis selection is one way of exercising specific identification — which is a narrower and more useful thing to know, because it carries the timing requirement with it.

Why this is practical rather than pedantic

Because the requirements travel with the method, not with the label. Anyone who believes they have "elected HIFO" may believe they have done something that persists and applies automatically. What the regulation contemplates is an identification made at or before each disposal, by whatever rule you are applying — and a standing order is how that is made to persist for broker-held units.

It Is Not a Method of Accounting Either

A second widespread claim concerns what happens when you change approach, and the regulation addresses it directly.

A method of specifically identifying digital asset units "is not a method of accounting", and a change — "for example, from the earliest acquired to the latest acquired" — "is not a change in method of accounting" to which the relevant Code sections apply.

What that rules out

Three familiar statements do not survive it. There is no election to make. There is no form to file to change approach. And there is no requirement to continue with whatever you did last year merely because you did it.

That is genuinely freeing relative to how the subject is usually described — and it does nothing whatever to relax the timing requirement, which is where the actual constraint lives. You may apply a different identification rule to a disposal in March than you applied in February. You may not apply either of them in the following December.

Average Cost, and Where It Actually Applies

Average cost is the most commonly mis-stated entry on the standard list, because the same idea appears in several systems in forms that are not interchangeable.

In the United States, not for ordinary crypto

The average basis method is available to shares in a regulated investment company and to stock acquired after 31 December 2010 in connection with a dividend reinvestment plan. The digital asset rules reach average basis only through a narrow case involving an asset that is simultaneously a digital asset and a security eligible for it.

So the claim that ordinary cryptocurrency can use average basis in the United States is not supported. A passing mention of average basis inside the digital asset provision is confined to that narrow overlap and does not open the method generally.

In the United Kingdom, an average that is not simply an average

The section 104 pool computes a pooled average, and in the absence of the matching rules it behaves exactly like an average-cost method. But it is preceded by two mandatory stages that take units out before averaging begins, and a description that omits them is describing a simpler system than the UK has.

In Canada, the adjusted cost base

The Canada Revenue Agency describes the amount subtracted from proceeds on a crypto-asset disposition as the adjusted cost base — usually the weighted average cost of a crypto-asset.

That is genuinely an averaging system, and it is tempting to equate it with the UK pool. The equation does not hold: the UK pool is preceded by the same-day and thirty-day matching rules, which take units out of the pool before any average is applied. Same arithmetic at the centre, different machinery around it.

Why Selection Systems Exist at All

It is reasonable to ask why any system allows a taxpayer to choose which units left, given that the choice moves money. Pooling looks tidier, and the United Kingdom manages without selection entirely.

The case selection was built for

The rules come from securities, where a holder genuinely could own identifiable lots — certificates with numbers, holdings with acquisition records held by a custodian. In that setting, identifying which particular shares were sold was not a fiction but a description of a real fact about real objects.

Extending that to fungible digital units keeps the mechanism while weakening the thing it described. The units are indistinguishable, so what the identification records is an intention rather than an observation — which is why the timing requirement carries so much weight. It is the only thing standing between an intention and a preference expressed later.

The case pooling answers

Pooling takes the opposite view: if the units are genuinely indistinguishable, the question of which one left is not a question at all, and a system that pretends otherwise invites exactly the hindsight the other approach must legislate against.

The cost is flexibility. A pooling taxpayer cannot manage the timing of gains through selection, because there is nothing to select. Whether that is a loss depends on whether you regard the flexibility as a legitimate planning tool or as an artefact, and reasonable systems have landed on both answers.

What this explains about the rules themselves

Seen this way, several features that look arbitrary become legible. The United States timing requirement is strict because the identification is an assertion rather than an observation. The United Kingdom thirty-day stage exists because a pooling system still has to stop a holder from manufacturing a loss and immediately buying back in.

Each system is defending against the specific way its own structure could be gamed, which is usually the fastest way to understand why a tax rule has the shape it has.

The Most-Quoted US Source Is Superseded

If you have read about United States crypto cost basis anywhere, there is a good chance you have read a paraphrase of a particular set of Internal Revenue Service frequently asked questions — the long-quoted answers on specific identification and first-in-first-out ordering.

What that page now says about itself

It carries a statement that those questions and answers generally apply to transactions involving digital assets completed before 1 January 2025.

That is the pre-2025 position — the universal-wallet approach in which a holding was treated as one undifferentiated stock across every venue — and it was replaced by the regulation described above. The IRS says so on the page itself.

How to use that when reading anything else

It gives you a quick and unusually reliable test of whether a source is current. Any article still describing the universal method as how United States crypto basis works, or still citing those questions as the governing answer, is describing the position before 2025 — whatever its publication date says.

The test is worth applying to advice you have already acted on, not only to advice you are reading now. A basis position adopted on the old footing and never revisited is a position built on a rule that no longer applies, and the transition arrangements that existed to move between them were keyed to dates that have passed.

What Happens When a System Changes Underneath You

The United States move from a universal approach to per-account tracking is a useful case study, because changing a cost basis system is not like changing a rate. It raises a question no ongoing rule answers: what happens to positions that were built under the old one?

The question a new rule leaves open

A holder whose basis had been pooled across every venue arrives at the changeover with a single aggregate figure and no allocation of it to particular accounts. The new rules ask a question about each account that the old records were never built to answer.

Transition arrangements exist precisely for that gap, and they typically require an allocation to be made and recorded by a deadline keyed to the changeover rather than to a convenient date later.

Why these are the arrangements most often missed

Because nothing prompts them. A rate change is visible in every subsequent calculation; a basis transition is invisible until a disposal happens in an account whose allocation was never made, which can be years later.

And by then the remedy is harder in a specific way: an allocation made with hindsight is not the same act as one made at the time, which is the same distinction the timing rule draws elsewhere in this area.

The general lesson for any system change

Two questions are worth asking whenever a jurisdiction changes how basis is determined. What does the new rule require of positions that already exist, and by when? And is there a transition arrangement whose window is defined by events rather than by a calendar date, so that it can close without an announcement?

Both are questions about your own history rather than about the new rule, which is why they are easy to read past when the change is described.

Elsewhere, Briefly and Carefully

Two further jurisdictions are worth a short entry each, because both are frequently summarised in ways their own sources do not support.

Ireland

Revenue's manual on the taxation of crypto-asset transactions states that no special tax rules for crypto-asset transactions are required, and that the amount subject to Capital Gains Tax must be calculated separately on each asset on a disposal-by-disposal basis.

What that manual does not do is mention first-in-first-out, pooling, or the four-week rule anywhere. So "Ireland uses FIFO for crypto", cited to Revenue, attributes to the crypto guidance something the crypto guidance does not contain — the identification provisions people have in mind sit in general legislation headed for shares and securities, and whether and how they reach cryptoassets is a question the manual does not answer in those words.

Germany, and the limit of what we can say

The German Federal Ministry of Finance issued a letter dated 6 March 2025 on the income-tax treatment of crypto assets, replacing its letter of 10 May 2022. That much is established.

The letter is reported to set out a hierarchy of individual identification followed by first-in-first-out, with an averaging method permitted for valuation, applied wallet by wallet. We are not stating that as a finding, and we are not quoting a paragraph number or a sentence from it, because the primary document could not be read directly when this page was written.

That is a deliberately unsatisfying entry. It is also the honest one: a precise-looking hierarchy quoted from a source nobody on our side has opened is exactly the kind of statement that spreads and becomes hard to correct. If German treatment matters to you, the letter itself is the thing to obtain.

The pattern across all four

Read together, these entries show why a comparison table with one method name per country is worse than useless. Two of the four jurisdictions here do not describe their treatment in method names at all; one of them says explicitly that no special rules are required; and the fourth reaches an averaging answer by a different route from the pooling system it superficially resembles. A table would have to invent a label for at least half of them, and the invented labels would then be quoted as though they had come from the tax authority.

Why the Four-Item List Persists Anyway

Given how poorly the standard list matches either jurisdiction, it is worth asking why it is everywhere.

It describes software, not law

Tax software genuinely does offer FIFO, LIFO, HIFO and average cost as selectable settings, because it serves users in many countries and because modelling alternatives is a useful feature. A list of settings is a real thing that exists.

What it is not is a list of methods a given taxpayer may choose between. The settings enumerate what the tool can compute; the law determines which of those computations is the one your return requires.

It was once closer to true in the United States

Under the pre-2025 position, with a universal-wallet approach and guidance that spoke in terms of specific identification and first-in-first-out ordering, a four-item framing was a rougher approximation of a looser regime. The regulation that replaced it is more structured, and the old framing survived the change it no longer fits.

The better question

Rather than "which method should I use", the question that produces a correct answer in both jurisdictions covered here is: does my jurisdiction let me select units at all, and if so, what must I do and when?

In the United Kingdom the answer to the first half is no, and the second half does not arise. In the United States the answer is yes, with a timing requirement that is the real constraint and a distinction between broker-held and self-custodied units that decides where the record lives.

What to Record, Under Each System

The records a system needs follow from the question it asks, and the two jurisdictions here ask different questions — so a record built for one can be perfectly adequate and still leave the other uncomputable.

For a pooling system

A pool needs a complete running history of acquisitions and disposals for each token, across every venue, with acquisition costs and allowable expenses attached. What it does not need is any record of which particular units you intended to sell, because the question never arises.

What it does need, and what is most often missing, is the venue-spanning completeness. A pool is per asset rather than per account, so a history assembled venue by venue and never merged is not a pool history at all.

For a selection system

A selection system needs everything above, plus two things a pooling system never asks for: the account or wallet each unit sits in at any moment, and the identification made for each disposal together with the time it was made.

That second item is the one with no natural home in any exchange export. It is not a transaction, it is a decision about a transaction, and unless it is recorded deliberately it leaves no trace at all.

For a holder subject to both

Someone with exposure in both jurisdictions needs the union rather than the intersection, and the union is not onerous once it is understood: a complete per-asset history that also knows where each unit sat, plus a contemporaneous note of any identification made. Written out as fields, that is four things.

• Every acquisition and disposal of each token, across every venue, merged into one history per asset. A pool cannot be computed from venue-by-venue records that were never combined, and merging them later is the step most often skipped.
• Acquisition cost and allowable expenses on each acquisition, converted into the currency the return is filed in.
• Where each unit sat at each moment. A selection system's answer depends on the account or wallet, and a transfer changes it without being a disposal.
• The identification, and the time it was made, for any disposal where one was made. This is the field no exchange export contains, because it records a decision rather than a transaction.

Building that once is considerably cheaper than discovering in March that the record supports one computation and not the other. The practical mechanics of assembling it are covered in our records and reconstruction guide.

What a Misconfigured Tool Actually Produces

Because the method setting is the most visible control in most crypto tax software, it is worth being concrete about what happens when it is set to the wrong system — which is not simply "a slightly different number".

A UK position computed by selection

A tool applying first-in-first-out to a United Kingdom holding produces a gain figure derived from a rule the UK computation does not use. It will also miss the same-day and thirty-day stages entirely, since a selection system has no equivalent of them.

The error is not confined to the year computed. Because the stages change which units left and therefore what cost remains, the position carried forward is wrong too — and every subsequent year inherits it.

A US position computed by pooling

In the other direction, a tool averaging cost across a holder's entire portfolio reproduces the pre-2025 universal approach rather than the current per-account rules. It blends units that the current rules keep in separate accounts and wallets, and it produces no identification record at all.

That output is not a conservative approximation of the right answer. It is an answer to a different question, and the fact that it looks orderly is part of why it goes unchallenged.

The check worth running once

Take a small number of disposals — half a dozen is enough — and compute them by hand under the rules your jurisdiction actually applies. Compare with what your tool reports for the same disposals.

If the figures agree, the configuration is probably right and you need not repeat the exercise. If they do not, you have found it cheaply, in an afternoon, rather than through a correspondence that begins years later.

• 1. Pick disposals that draw on more than one acquisition. A disposal that can only have come from a single lot cannot disagree with anything, so it tests nothing.
• 2. Compute each by hand under the rules your jurisdiction applies, writing down the cost relieved and which acquisitions it came from.
• 3. Compare the cost relieved rather than the gain. Two offsetting errors can leave a gain looking right while the basis behind it is wrong, and the basis is what carries forward.
• 4. If they disagree, check the jurisdiction setting first, then whether the tool is pooling across venues or tracking per account — those two settings produce most of the divergence.

Reading a Method Claim Critically

Given how much circulating material on this subject describes software settings as though they were entitlements, a short set of tests is more durable than any list of methods.

• Does it name a jurisdiction in the same sentence? A claim about cost basis with no country attached is not a claim about anything — the answers differ structurally, and an unattributed one will be right somewhere and wrong where you are.
• For the United States, does it predate 2025? Sources describing a universal approach across all your holdings are describing the position the current regulation replaced, and the IRS says as much on its own superseded page.
• Does it mention when an identification must be made? A description of selection that omits the timing requirement has omitted the binding constraint, and is describing something easier than the rule.
• Does it distinguish broker-held from self-custodied units? A single United States answer covering both is covering two different requirements with one sentence.
• Does it attribute a prohibition to a tax authority? "HMRC bans X" and "the IRS approves Y" are strong claims. Where the authority has simply not addressed a method name, both are inventions — and the honest description is structural rather than permissive.

Why this is worth the trouble

Because the cost of getting it wrong is not a single wrong figure. It is a cost basis position carried forward, compounding through every subsequent disposal, and discoverable only by recomputing the history it came from.

Five questions applied once to a source is cheap by comparison, and it is the same discipline the rest of this cluster runs on: prefer what a rule says to what a summary of it says, and notice when a summary has quietly widened.

Conclusion

Two jurisdictions, two structurally different answers, and one widely repeated list that fits neither.

The United Kingdom does not offer a choice: a disposal is matched same-day, then against the following thirty days, then against a pooled cost, and no selection enters it. Saying HMRC bans or permits particular method names overstates what it has written, which is nothing about them at all.

The United States does offer selection, through specific identification, with highest-basis being one way of exercising it rather than a separate approved method. The binding constraint is not which rule you apply but when you record it — no later than the date and time of the disposal — and the place the record must live depends on whether the units sit with a broker or with you.

Two habits follow. Check whether a source predates 2025 before relying on it for the United States, using the test above. And treat a method picker in software as a modelling tool rather than as a menu of entitlements — what it computes is a possibility, and what your return needs is the one your jurisdiction's rules produce.

There is also a broader point worth keeping once the specifics fade. Every system on this page is defending against the particular weakness its own structure creates. A selection system must legislate against hindsight, because an identification of indistinguishable units is an assertion rather than an observation. A pooling system must legislate against a sale-and-repurchase, because averaging alone would not notice one. Reading any unfamiliar jurisdiction's rules with that question in mind — what could a taxpayer do here that the structure would not otherwise catch — usually explains the provisions that look arbitrary on first reading, and it explains them faster than memorising which acronym each country accepts.

One last practical note. If you hold in more than one place, or under more than one set of rules, this page on its own is not enough: it has to be read alongside the records that would let you demonstrate any of it. An identification you cannot evidence is indistinguishable, to anyone reading your return, from one you never made.

Sources

Cited by identifier and the date read. HMRC manual sections carry no claimed update date, because the stamps shown on a manual page apply to the whole manual rather than the section.

• United States Treasury regulations at section 1.1012-1(j) — specific identification, the default ordering rules, and the treatment of a change of identification rule. Read 21 September 2026. federalregister.gov
• IRS frequently asked questions on virtual currency transactions — carrying the statement that they generally apply to transactions completed before 1 January 2025. Read 21 September 2026. irs.gov
• HMRC Cryptoassets Manual, CRYPTO22200 — the UK matching order and the section 104 pool. Read 21 September 2026. gov.uk
• Canada Revenue Agency guidance on reporting capital gains as a crypto-asset user — the adjusted cost base description. Read 21 September 2026. canada.ca
• Irish Revenue Tax and Duty Manual, Part 02-01-03, on the taxation of crypto-asset transactions. Read 21 September 2026. revenue.ie
• German Federal Ministry of Finance letter of 6 March 2025 on the income-tax treatment of crypto assets — named but not quoted; the primary could not be read directly when this page was written. bundesfinanzministerium.de
• Because a selection system tracks basis per account, which venue holds which units is part of the computation rather than an administrative detail. The account side of that is covered in our guide to the first thirty days on an exchange.
• Where cost basis sits in the wider picture — what is taxable at all, what gets reported about you, and which records make any of it computable — is mapped in our complete crypto tax guide.

Frequently Asked Questions

Can I choose FIFO, LIFO or HIFO for my UK crypto tax?

No. The United Kingdom computation matches a disposal against same-day acquisitions, then against acquisitions in the following thirty days, then against the section 104 pool, in that order — the UK computation uses pooling and specific-identification methods do not enter it. HMRC's Cryptoassets Manual does not name FIFO, LIFO or HIFO as options for individuals, and it has not said it prohibits or permits any of them: it says nothing about those names at all.

Does the IRS permit HIFO for crypto?

That is not how the United States rules are structured. The regulation refers to a method of specifically identifying units "for example, by the earliest acquired, the latest acquired, or the highest basis" — so highest-basis selection is one way of exercising specific identification rather than a separately approved method, and the acronym HIFO appears nowhere in the provision. The familiar list of "FIFO, LIFO, HIFO and specific identification" as four alternatives misdescribes the structure it is summarising.

Can I pick which lots I sold when I prepare my US return?

No. Under the United States regulation a specific identification is made only if, no later than the date and time of the sale, disposition or transfer, the taxpayer identifies the particular units on its books and records. Choosing lots at filing time, once prices and outcomes are known, does not meet that timing requirement — and it is the single most common error in summaries of this area.

What happens in the US if I do not identify any units?

A default ordering applies, and its boundary matters. For units held in a broker's custody without an adequate identification given to the broker, they are treated as disposed of from the earliest acquired within that broker account. For units not held in a broker's custody, the ordering runs from the earliest acquisition of units in that wallet. Neither default runs across a taxpayer's whole portfolio, so "FIFO is the IRS default for crypto" is incomplete without the account or wallet boundary.

Do I have to elect a method or file a form to change it?

No. The United States regulation states that a method of specifically identifying digital asset units is not a method of accounting, and that a change — for example from the earliest acquired to the latest acquired — is not a change in method of accounting. There is no election to make, no form to file to change approach, and no requirement to continue with last year's rule. The timing requirement on each identification is unaffected by any of that.

Can I use average cost for cryptocurrency in the United States?

Not for ordinary cryptocurrency. The average basis method is available to shares in a regulated investment company and to stock acquired after 31 December 2010 under a dividend reinvestment plan, and the digital asset rules reach it only through a narrow case involving an asset that is simultaneously a digital asset and a security eligible for average basis. A passing mention inside the digital asset provision does not open the method generally.

Is the UK section 104 pool the same as Canada's adjusted cost base?

They share their arithmetic and differ in their machinery. The Canada Revenue Agency describes the amount subtracted from proceeds as the adjusted cost base, usually the weighted average cost. The United Kingdom pool also averages — but it is preceded by two mandatory matching stages, same-day and then the following thirty days, which remove units from the computation before any averaging applies. Treating the two as equivalent skips the stages that most often change a UK answer.

Financial Disclaimer

This content is not financial advice. All information provided is for educational purposes only. Cryptocurrency investments carry significant investment risk, and past performance does not guarantee future results. Always do your own research and consult a qualified financial advisor before making investment decisions.