The tax treatment of cryptocurrency forks presents four unique challenges: (1) parent/child designation, (2) new token access, (3) assessment of fair market value, and (4) assessment of comparable contemporaneous fair market values. We provide evidence that each issue complicates the determination of income realization, or basis apportionment. We compare three existing approaches for assets acquired without a purchase. We conclude that the least problematic approach (adopted by Japan) assigns zero tax basis to the new coin and taxes the proceeds upon sale. Treating the new coins as realized income (as recently ruled in the US) is the most problematic approach.
The last few years have seen an explosion of interest in, and activity around, distributed ledger technologies. Since the introduction of Bitcoin, many other cryptocurrencies have been launched,1 and businesses have raced to find ways to incorporate blockchain technology in their operations. Tradable and storable tokens associated with blockchains are just one example of a growing class of digital assets that exist within and depend on community-wide protocols.2 A well-known example of such a protocol is the World Wide Web, whose domain names are regularly bought, traded, and commonly understood as “property.”
The rise of digital assets creates interesting tax issues. The most visible and vibrantly debated tax issue specific to blockchain-based currencies is whether they should be treated as currency or property. The first published Internal Revenue Service (“IRS”) guidance on the topic maintains the position that they should be taxed as property,3 while in other jurisdictions (e.g., Germany) cryptocurrency can be used as normal currency, i.e., spent to purchase goods without incurring gains and losses.
The focus of this paper is a less-obvious but equally important issue: the tax implications of token issuances after chain splits, one of the most unique technological aspects of a purely digital asset class. Chain splits are also the subject of Revenue Ruling 19-24 of 9 October 2019,4 the second and latest installment of IRS guidance on cryptocurrencies. These events consist of a duplication of the ledger (the blockchain) on which token transactions are recorded following a so-called fork. Chain splits, whose full implications are discussed in Section 2, result in the allocation of additional tokens to existing (and possibly unwitting) token holders. The tax treatment of such a token acquisition and of a potential subsequent sale poses important questions for a regulator.
Chain splits are frequent. Some, like the fork that created Bitcoin Cash, are of significant economic importance. Others, like the fork that created Bitcoin Pizza, are at best highly speculative.5 Prior to Revenue Ruling 19-24, the lack of guidelines had caused uncertainty for taxpayers.6 The new ruling does answer some of the taxpayers’ most frequent questions, but it appears to apply only to those taxpayers who hold cryptocurrencies indirectly via an exchange, and not those who hold them directly. Regardless of applicability, we also argue that the approach chosen by the ruling is not sustainable.
In Section 3, we identify four key issues for taxation authorities when dealing with chain splits: (1) parent/child designation, (2) the assumption that the taxpayer has access to the new token, (3) assessment of fair market value, and (4) assessment of comparable contemporaneous fair market values. We provide empirical evidence that each of these issues is a hurdle in determining whether income has been realized, or in apportioning the basis between the two split tokens, or both.
In Section 4, we then examine the shortcomings and benefits of three existing tax treatments that could be applied to cryptocurrency splits: (i) treating the new coin as an independent accession to wealth by the taxpayer (a “treasure trove,” the approach adopted by Revenue Ruling 19-24 in the USA); (ii) considering the chain split as a stock spinoff or similar event (an “asset split,” the approach adopted by the UK); and (iii) treating the new coin as the offspring of the existing coin (“calving,” the approach adopted by Japan). In light of the four issues we identified above and our empirical evidence, we conclude that the third approach is the least problematic, and the first approach the most problematic. Section 5 concludes with this and a few additional suggestions.
For the purpose of this paper, the reader can simply consider a blockchain to be a history of transactions and a software program (i.e., a set of rules) to process and verify these transactions shared across multiple individuals, or more specifically, nodes.7
A cryptocurrency is a distinct form of digital token whose ownership history is recorded in a blockchain. As a purely digital asset, a cryptocurrency is defined by the transaction history recorded on its associated blockchain. Each blockchain can be associated with one or more tokens. For example, the Bitcoin blockchain records Bitcoin (BTC) transactions, the NEO blockchain records both NEO and NeoGAS (GAS), and so on. Tokens can be traded directly between users, or can be traded using an intermediary, like a cryptocurrency exchange. Most trades occur via cryptocurrency exchanges.
As with all software programs, the software code of a blockchain can fork, i.e., be updated. In practice, the term “fork” is used to indicate both the block number (the height) at which the updated software is allowed to operate and the time at which that block is mined. A fork can be successful, if all nodes update, and unsuccessful, if all nodes reject it. Such successful and unsuccessful forks do not cause the creation of any new asset.
There is, however, a third possibility. If the blockchain is not centrally managed, each node must independently decide whether to update their version of the code. It is therefore possible that some nodes incorporate the update, and others do not. A chain split occurs when both the original and the updated software are used simultaneously by a sufficient number of nodes. Even though all nodes recognize a common transaction history up to the fork block, updated and non-updated nodes can create mutually incompatible blocks of transaction records. The addition of incompatible blocks to the original blockchain results in divergent histories and, therefore, non-identical blockchains. Given that the token is inherently defined by its blockchain, this means that the token itself splits into multiple instances as well. It is only when the fork results in a chain split that owners of a token on the parent chain effectively acquire a new asset.
Perhaps because a chain split can only happen after a fork, this scenario is often referred to informally as a “hard fork.” 8 In this paper we use the terms “fork” and “split” in a strict sense, as defined above. For clarity and simplicity, in the remainder of the paper we will also assume that a chain split results in exactly two blockchains, and a user that owned one coin associated with the original blockchain prior to a split now owns two coins, one on each chain.9
Figure 1 shows a stylized fork timeline. In the typical sequence of events, at some date an entity announces the intention to fork at some future block X. When block X is mined, a replica of the original (“parent”) chain up to that point is made, determining the ownership of tokens on the new (“child”) chain. As the child chain “goes live,” i.e., as its nodes enter operation and start adding blocks to it, the child chain diverges from the parent, effecting the split. Finally, if demand for trading the child token exists, one or more exchanges can choose to list it. Note that block X functions like stocks’ ex-dividend date: the initial allocation of tokens is determined at the fork date, after which time there is no need to hold the parent token to receive the child token.
It is crucial to understand that Figure 1 shows the typical order of events, but the time between successive steps, their order, and their existence may vary. The date associated with block X need not be the date when the child chain nodes go live, or the date that trading begins. On occasion, a fork may be announced, but the developers may be unable to launch their network at the time block X is mined (in fact, as we show below in Section 3.2, this is common). The new chain may never be launched, and even if it is launched, it may never get listed. Conversely, a token may get listed and traded before the chain goes live because the existence of a fork is enough for exchanges to determine ownership of the new token.
A clear understanding of fork mechanics is necessary to understand market prices. A price may exist because the child token is traded on a live blockchain. However, a price may exist because the child token is already listed and traded after the fork, but prior to the child chain going live, not unlike “when-issued” trading in traditional securities markets. Finally, a price may exist even before the fork because of the existence of forward or futures contracts, if exchanges feel sufficiently confident that the fork will happen when announced.
A chain split causes the owner of one token on one blockchain to become the owner of two tokens on two blockchains, thus acquiring a new asset. A framework for the taxation of cryptocurrency events should provide guidance with respect to two main policy decisions: the timing of income realization, and the tax basis assigned to the parent and child tokens.
To determine the timing of income realization, a framework for the taxation of cryptocurrency events must take a stand on whether the receipt of new cryptocurrency tokens per se constitutes gross income. Under US case law, “gross income” is defined as “instances of undeniable accessions to wealth, clearly realized, and over which the taxpayers have complete dominion.”10 In practice, in the case of an asset acquired without a purchase transaction, income could be deemed to be realized at the time of the acquisition or deferred until the time of a subsequent sale.11 Deferral reflects concerns about the difficulty of valuing assets in the absence of transactions as well as the tax authority’s reluctance to force taxpayers to liquidate assets simply for the purpose of paying taxes.
In turn, the problem of determining the tax basis depends on the choice made with respect to the timing of income realization. If income is recognized at the time of the acquisition, the framework must provide a way to value the asset at the time of the acquisition; this will then be its tax basis. If income is recognized at the time of a subsequent sale, the framework must provide a way to determine the tax basis of the newly acquired asset, i.e., to apportion the original acquisition cost between the original token and the new token. In this second case, valuing the new token may or may not be necessary. This decision is important because an incorrect basis allocation could create tax-arbitrage opportunities for the investor. It may also generate trading distortions by incentivizing a taxpayer to hold on to the asset whose basis has been set too low in order to avoid capital gains tax (the well-known “lock-in effect”) or to sell the asset whose basis has been set too high (generating a tax deduction due to capital loss).
Chain splits present technical and economic aspects that create four unique difficulties with respect to these policy decisions:
Parent/child designation: Difficulty in identifying which token generated from a split represents the continuation of the original token.
Token access: One or both of the tokens generated by the split may not be accessible, or easy to sell.
Fair market value (“FMV”): The value of a token generated by a split may not be easy to access or quantify.
Comparable contemporaneous FMVs: Difficulty in establishing the value of the two split chains relative to each other at the same point in time.
Issues 1, 3, and 4 pose a challenge in determining the tax basis, and 2 and 3 in determining the timing of the realization of income. In the rest of this section we establish these facts empirically.
While our focus is on policy decisions that are economic in nature, there are other important ones outside the scope of our paper. One such decision concerns the characterization of any income as ordinary or capital gain. Another concerns reporting requirements, i.e., if and when an investor should report a split. Because of the difficulty of tracing cryptocurrencies, various existing policy proposals12 all stress the need for incentivizing early reporting of the existence of a new asset, regardless of the timing of income realization. For a deeper discussion of these topics, see Xu13 or Chason.14
3.1. Parent/Child Designation
The first difficulty is designating one of the two chains as the parent. The fact that both chains are subjectively correct continuations of the original chain poses a new, unique challenge: which token represents the “true” continuation of the originally purchased token? The continuity question is not as simple as it may appear. Consider the two most visible and economically important splits to date, Bitcoin in August 2017 and Ethereum in 2016.
On August 1, 2017, Bitcoin underwent a contentious fork that resulted in a chain split. The source of disagreement was a code update that allowed a larger block size. The non-updated half continued with the name of Bitcoin, while the updated half took on the name of Bitcoin Cash. This outcome is effectively the opposite of the Ethereum chain split in July 2016. In Ethereum’s case, both chains continued with identical technology—the update only affected the transaction history by reverting a small number of transactions. The updated blockchain continued with the name of Ethereum. The non-updated chain took the new name Ethereum Classic,15 while nonetheless claiming to be “a continuation of the original Ethereum blockchain—the classic version preserving untampered history.”16 These examples show that objective criteria such as technology, naming, and ledger integrity have limited discriminatory power.
This is not to say that technological criteria have no power, but simply that they are not universal: for example, if one of the chains is taken offline at the time of the fork, that chain may be reasonably deemed to be the child chain. However, in both the Bitcoin and Ethereum forks the child networks went live on the day of the fork. Similarly, using measures of community support such as network size or cryptographic strength (“hashrate”) could yield incorrect results if the fork changes the technological requirements of the network.
In spite of the lack of objective universal criteria, an informal community consensus has usually emerged prior to the split as to which chain is the parent and should retain the ticker symbol, with the minority choosing a different ticker they would prefer to use after the split. This consensus has become a practical necessity given the increasing complexity of trading in the cryptocurrency market and increasing use of intermediaries such as exchanges: in order to be able to offer uninterrupted trading (including spot, margin, and future trades), upon a split cryptocurrency exchanges must associate the original ticker with a specific blockchain.17 However, this consensus is neither formal, nor guaranteed, as decentralized cryptocurrency markets are not designed to enforce consensus.18
Empirical Observation 1: There is no formal way to establish which chain is the “parent” (i.e., the original) and which is the “child” that will work in all cases.
The Ethereum example above, where the split involved a rollback of a subset of transactions, shows clearly that a split may generate new tokens not only due to the creation of the child chain, but also due to changes in the parent chain.
An additional complication is that most traders never interact with the underlying blockchain, instead trading on exchanges. Upon a split, exchanges—which themselves have no obligation to be associated with the blockchain—independently decide which of the two tokens is the parent and which is the child for the purpose of assigning the original ticker symbol and fulfilling outstanding contracts and obligations.19 Although historically exchanges have always followed the informal community consensus, and have usually respected the desired ticker requested by the minority,20 the coin names, tickers, prices, and ownership experienced by an individual taxpayer ultimately depend on their exchange’s individual decisions. Exchange-level decision-making thus adds an additional layer of market uncertainty atop the technological uncertainty arising from the split itself.
A vivid example of this problem is represented by the following announcement regarding the distribution of Bitcoin Cash tokens to existing exchange users with open Bitcoin positions:
Although several months ago, we developed a methodology for handling contentious hardforks, we will be applying a different methodology [...] Due to the net amount of BTC committed in margin positions at the time of the fork, the above methodology may result in Bitfinex seeing a surplus or deficit of BCH. As such, we will be resolving this discrepancy in the form of a socialized distribution coefficient. For example, currently, there are more longs than shorts on the platform, causing a distribution coefficient of ~1.091 (Meaning that for each qualifying BTC a user will receive 1.091 BCH). The actual coefficient will be calculated at the moment of the distribution. [...] We recognize that this model is not perfect, but given the short time frames imposed upon us, better choices are not available.21
Empirical Observation 2: There may be variation across taxpayers in the number and nature of tokens acquired on either the parent or the child chain.
Empirical Observation 1 implies that an ideal framework for the taxation of cryptocurrency events would be agnostic as to which chain is the parent, bypassing the need for a determination in order to assign the tax basis. Empirical Observation 2 implies that an ideal framework should be flexible enough to handle heterogeneity of outcomes across taxpayers and should take into account that this heterogeneity may make auditing labor-intensive.
3.2. Token Access
The second issue is that of token access: some users may not receive the token associated with a split until months after the event, if ever. A framework that assumes the constructive receipt of split tokens places an undue burden on taxpayers to keep track of any announced forks, to determine which forks actually result in a split, to ensure that their tokens are stored in a location that grants them access to the generated token if and when the split happens, and to undertake any additional actions required to claim the split token, however worthless.
These concerns are not isolated to forks originating from obscure coins. Bitcoin had, to the best of our understanding, forty-eight announced forks in the last three years,22 the most famous of which was Bitcoin Cash on August 1, 2017.23
Of the identified 48 announced forks, only 32 ever launched an active network and created a split. Of those 32 forks that implemented chain splits, we could find trade data between the fork date and July 31, 2019 for 17 tokens using CoinGecko.24 In other words, only two-thirds of announced Bitcoin forks resulted in chain splits (32 of 48), and only a third of announced Bitcoin forks resulted in tokens that were ever openly traded (17 out of 48). We will refer to announced forks that did not result in tradable tokens as failed forks. On the basis of this evidence, we state the following empirical observation.
Empirical Observation 3: Not all announced forks result in chain splits, and many chain splits do not result in traded tokens.
This is not to imply that a child token with trading activity should be construed as a success. It is possible that while a child token may not have strictly speaking zero value, it may also not be sufficiently valuable and thus represent a nuisance token that a taxpayer may wish to ignore or leave unclaimed instead of expending effort to acquire. Our approach is based on market volume data: we identify a token as a nuisance token if it does not sustain a significant volume for a sufficient amount of time. This approach may fail to identify some nuisance tokens given the rampant market manipulation of lower value and volume coins by both traders and exchanges attempting to inflate their trade volumes to attract more consumers as shown by Hougan et al.25 and Gandal et al.26 An alternative approach to the identification of nuisance tokens would be based on technology data, such as node counts or chain hashrate. However, this approach is even less desirable. First, technology-based metrics may not be directly comparable across chains if the technology itself has changed as a result of the fork, and they may be inflated by zombie nodes.27 Second, and most important, this approach suffers from crippling data limitations as historical series of node counts and hashrates are not immediately available.
Table 1 presents the data of the 17 chain splits prior to July 31, 2019 for which we could find trade data. The data presented, in addition to name and ticker symbol, includes the number of days between the fork date and the start of trading data (Days to Trade). We also present the highest ratio of trading volume and price normalized by the respective Bitcoin values to adjust for fluctuations in the price and trading volumes in cryptocurrency markets, and the July 31, 2019 trading volume and price for each coin. Note that some tokens become tradable on the day of the fork, while for others it took up to 530 days (almost a year and a half). This adds further nuance to Empirical Observation 3: even among the one-third of announced forks that resulted in a traded token, significant time may have elapsed between the fork date and feasible trading—a time during which the taxpayer had no assurance that the token would ever be traded.
Of the 17 splits with trading data, listed in Table 1, only 7 have ever reported even a single day of trade volume that exceeded 1% of their parent chain (Bitcoin) in their entire trading history: Bitcoin Cash, Bitcoin Diamond, Bitcoin File, Bitcoin Gold, BitcoinX, Classic Bitcoin, and Micro Bitcoin. Because, as described above, reported high volumes may be caused by manipulation, and more significantly biased for lower volume coins, our finding that 7 coins have achieved at least 1% of the parent’s trade volume is itself an overestimate. Given this finding, we label nearly two-thirds of Bitcoin chain splits with data (10 out of 17) as nuisance forks. Thus, failed and nuisance forks represent 84% of the original sample (41 out of 48).
Empirical Observation 4: Most chain splits are nuisance forks.
Empirical Observation 3 suggests that the mere announcement of a fork does not imply that a taxpayer who owned a token in the parent chain has acceded to a new token (by any reasonable definition of “accession”). Empirical Observation 4 suggests that even if the taxpayer is technically able to claim possession of a split token and to sell it, he or she may not want to take those steps because the likelihood of obtaining sufficient value to compensate for the effort is low. Taken together, these two observations suggest that an ideal framework for the taxation of cryptocurrency events should not assume positive income is realized at the time actual trading of the new token starts, let alone at the time a fork is announced.
A further complication is that most traders trade and store tokens via exchanges, and can therefore only own tokens their exchange chooses to list. The majority of large exchanges are “custodial,”28 i.e., traders actually hold exchange IOUs in lieu of the actual tokens, which are recorded on the blockchain as belonging to the exchange. Upon a chain split, the exchange is not obligated to issue, let alone list, IOUs denominated in the child token. Its sole incentive to do so is reputation, a concern that must be weighed against the cost and security risk of listing a new token, a decision that can take months or even years. Exchanges may also choose at their discretion to delist previously listed tokens, resulting in users losing access. Finally, even listed tokens may be so worthless that they are impossible to sell, and even traders who own tokens directly may find themselves having no practical access to them if such tokens are not listed on any exchange.
3.3. Fair Market Value
The next issue is that of assigning a token its Fair Market Value (“FMV”) without the taxpayer engaging in a trade of the token. An incorrect FMV will cause taxpayers to have excessive taxable gains or losses upon accession or sale, creating a lock-in effect or encouraging the sale of tokens regardless of market price and subjective valuation.
Generally, the existence of a traded market price is understood to eliminate most or all uncertainty about the true value of an asset. In this subsection we argue that this is not the case for the price of a split token that has recently (or has yet to) come into existence. Our argument is based on a series of items of circumstantial evidence that together support the interpretation that split tokens are severely overpriced in the immediate aftermath of the fork. We call this evidence “circumstantial” because the economic fundamentals that determine the true value of a blockchain token are poorly understood, both theoretically and empirically, and therefore attempting to rigorously prove that a certain token is overpriced is a quixotic exercise. We encourage the reader to critically assess the evidence we present and to consider alternative explanations.
Basic finance theory posits that the price of an asset should only change when new information gets incorporated into prices. When a fork is announced, the price of the original token should change to reflect both the expected value of the child and any implications for the value of the parent. When, subsequently, the fork is executed, the price of the parent should drop by the value of the child (in the same way as a stock price should drop by the amount of the dividend on the ex-dividend day), but the expected combined value of parent and child should not exceed the value of the parent token immediately prior. If such a scenario arose, market participants would buy the parent token just prior to the split and sell both the parent and the child immediately afterwards for a quick profit. The demand for the parent token would cause its price to rise until there is no longer an expected profit.
Our first observation is that, in practice, such profit opportunities appear to exist, i.e., basic finance theory does not appear to hold around forks. On August 1, 2017, Bitcoin Cash simultaneously forked, triggered a chain split, and began trading at a significant fraction (roughly $400 vs. $2700) of the value of Bitcoin, its parent token. On the day of the fork, however, Bitcoin’s price did not fall by an appreciable amount, bestowing a daily return of roughly 10% on the lucky Bitcoin holders of record (or at least those of them who managed to quickly access and sell their newly acquired Bitcoin Cash holdings). This return ranks #5 out of 1686 daily returns measured from the beginning of 2015 to the present day.29
It is important to note that this exceptional daily return might simply have been due to exceptional news being impounded into prices. However, this is unlikely, because the fork had already been announced, and in fact there had been a forward market price for Bitcoin Cash since at least a week prior (July 23, 2017).30 To the extent that the successful roll-out of the split reduced uncertainty about the viability of Bitcoin Cash, it would have caused its price to rise, something we do not observe.
This first observation shows that the market price may not reflect all available information, but not that it is irrational. The “quick profit opportunity” described above is not a riskless arbitrage, requiring instead to hold an unhedged Bitcoin position at fork time. In the case of the Bitcoin Cash fork, a trader could not have (i) bought one Bitcoin, (ii) sold one Bitcoin forward, and (iii) sold a Bitcoin Cash forward, because there were no Bitcoin forward or futures contracts at the time.31 In the presence of limits to arbitrage, even assuming market participants are perfectly rational, relative mispricing can persist for a long time in the presence of, e.g., slow information diffusion or market segmentation.
Our second observation is that indeed, mispricing appears to be persistent. To show this, Figure 2 plots the token price as a ratio of BTC price for the first 365 days of trading of six tokens. These tokens are selected based on three criteria: (i) they have at least 365 days of trading data; (ii) they began trading within two weeks of the fork; and (iii) they eventually achieved near-zero price ratio.32 While some child tokens (Bitcoin Atom) slide almost immediately to near-zero price, others (Bitcoin X, Bitcoin Diamond) hover over 2% for several months before permanently declining.
It is important to note that this evidence does not imply that any one of these six tokens was initially overpriced. However, if one believes that they were overpriced to begin with, this evidence does show that mispricing is persistent.
Our third observation is that, at least based on the historical record, Bitcoin child tokens yield starkly negative expected returns. Standard asset pricing theory posits that assets should either have positive expected returns or provide some form of insurance, or else market participants would not want to hold them. Therefore, systematically negative observed returns are suggestive of initial overpricing.
For this analysis, we use only the 12 tokens that have at least 365 days of trading history. The results are shown in Figure 3 and Table 2. At the 90-day mark, the average return is -64%. At the one-year mark, 7 out of 12 tokens are already worthless (-98% or worse).
Our results are not unique to Bitcoin forks. Evans studies a smaller sample of only five forks across four parent chains.33 Like us, he observes that most child forks have depreciated substantially relative to their parent chain. In addition, he also argues that by one specific objective metric child tokens are overvalued relative to the parent token.34
It is important to note that the observed negative expected returns may simply be the product of a highly skewed return distribution. The expected return may still be positive in spite of our small-sample measurement if, for instance, 99 out of 100 tokens are expected to become worthless and one to experience a 20,000% return. Nonetheless, the observed return distribution shows that, at a minimum, child tokens are highly speculative investments.
Our fourth and last observation is that the price of child tokens is subject to large positive spikes. Figure 3 displays sharp spikes in the cumulative returns of child tokens, reflecting similarly sizable underlying price spikes. Each spike is driven by an individual token increasing in price and then reverting to the long-run trend. Similar spikes are also visible for some of the cryptocurrencies in Figure 2. The speed of increase, its reversion, and the absence of negative spikes all suggest that these phenomena are not driven by news about fundamental value of the payment network. Given the extent of market manipulation of low volume coins (see Section 3.2), it is not impossible that these spikes are caused by wash-traders engaging in pump-and-dump schemes, rather than by sudden increases in rational agents’ perceived usefulness of the network followed by equally sudden changes of mind.
Empirical Observation 5: Newly split tokens are likely overpriced. The overpricing is highly persistent, taking up to a year to decay, and resulting in large negative average returns. Their market price is highly unstable and possibly subject to intermittent manipulation.
Thus, an ideal framework for taxation of cryptocurrency events would minimize reliance on difficult determinations of what tokens are worth until they are sold by the taxpayer. Tax policies that set the FMV of a child token based on its price at trading launch—or even based on the average price over some period after trading launch, as in Xu’s proposal35—run a severe risk of setting an unduly high tax basis, encouraging taxpayers to dump any child tokens they are awarded to generate a tax loss deduction. Such policies could also create an additional incentive to inflate token prices at launch to later generate tax losses.
3.4. Comparable Contemporaneous FMVs
The final issue is the potential difficulty in establishing comparable contemporaneous fair market values. While related to the prior difficulty of establishing the value of a token, this empirical fact is specific to tax law applications in which both tokens need to have an established value relative to each other at the same point in time.36 The prices of tokens can vary substantially, and inconsistently, across different exchanges based upon location, financial regulations, volume, and number of coins they accept, making the establishment of a uniform market price challenging for smaller, less traded tokens. This phenomenon was first documented and examined by Pieters and Vivanco.37 Unsurprisingly, the arbitrage window exists for tokens arising from a chain split as well.
Figure 4 plots the price difference of Bitcoin Gold on the exchanges Bitfinex and Exmo. The price of Bitcoin Gold on Bitfinex can be as much as 8% higher, or 4% lower, than the price on Exmo on any given day, with an average variation of 0.38%. Given that prices can be higher or lower, we also calculate the average of the absolute value of the difference, which is 1.27%. A taxpayer whose cost basis is set using Bitfinex, or using Exmo, may be subject to a price that is different from the one they actually receive in the market where they trade.
Empirical Observation 6: The price of tokens may vary substantially across markets.
There are two scenarios in which the relative fair market values of the split tokens may be needed for tax purposes: if the tax basis needs to be apportioned at the time of the split, or if apportionment is delayed until sale, but only one coin is sold. This difficulty arises because split tokens may have lower liquidity or be traded on markets other than the parents’, and transactions reported on these markets may be different in nature or even completely fabricated. This empirical observation suggests that an ideal framework for taxation of cryptocurrency events would minimize reliance on the relative FMVs of new and old tokens.
Taxpayers frequently acquire financial or real assets without a purchase transaction, and the tax code already contains numerous provisions to establish (i) whether the acquisition itself constituted accession to wealth, and (ii) a putative acquisition cost (or tax basis) relative to which subsequent accessions to wealth should be measured.
In order to avoid distorting taxpayers’ incentives to sell or retain assets, these rules are usually designed case by case to reflect the underlying economic reality to the extent possible. Rather than deriving the theoretically optimal tax treatment from an economic equilibrium model, in this section we define three existing approaches (labeled treasure trove, asset split, and calving) that could be applied to the taxation of split tokens. In light of the empirical facts established in the previous section, we discuss their underlying economic and financial assumptions and examine their specific advantages and disadvantages, summarized in Table 3.
To show what happens under each approach, we apply the respective rules to the same specific example. A user, “Mina,” purchases a token for $100 that later splits into two coins, A and B. In order to focus on the mechanics of the tax rules, we assume that only the tax basis needs to be determined, and none of the four key issues of the prior section apply. Specifically: market consensus clearly designates A as the original token, Mina has immediate access to both tokens, and tokens have clear fair market values at all times. Namely, at the time of the split, the market prices of A and B are available and can be determined to be $1000. Later, when Mina considers a sale, she can sell either token for $150.38
The tax consequences of different approaches are summarized in Table 4, which shows that Mina’s total taxable income remains the same regardless of the timing of the sale.39 Since Mina initially pays $100 to acquire the original coin, and eventually owns $300 worth of coins, her total (realized and unrealized) income must be $200. What changes across different approaches is the mechanism that determines the allocation of gain across the two tokens. This, in turn, creates different trading incentives, which we also discuss.
4.1. The “Treasure Trove” Approach
The reception of new cryptocurrency could be treated as a lucky find of an asset with no readily ascertainable owner. (An oft-cited example is finding cash hidden in a piano previously bought at auction). The intuitive appeal of this analogy derives from the fact that taxpayers appear to unexpectedly obtain something in exchange for nothing.40
Applying this analogy to chain splits presents serious difficulties. A treasure trove is ex ante unexpected and ex post known and desirable, while a chain split may fail to be any one of these things. Even if the taxpayer was not originally aware that the chain would split when they conducted their original purchase, chain splits are often announced in advance. If market participants are aware of a future chain split, the future expected value of the child token may already be included in the current price. In this case the analogy with the piano example breaks down, as if the auction participants had known about the cash in the piano all along. At the other extreme, it is problematic to assume ex post that the taxpayer knows about the split and benefits from it unless he or she has claimed the split tokens: as we showed in Section 3, announced forks are often not implemented, and many forks are a nuisance for the investor. Furthermore, there are no reporting requirements for those forks that are implemented.41
If a cryptocurrency event is treated as a treasure trove, the tax basis of the token deemed to belong to the parent chain remains unvaried. The taxpayer immediately realizes taxable income equal to the fair market value (however determined) of the child token, whose tax basis is set equal to the amount of income realized. This is the approach chosen by the new IRS guidance.
In Mina’s example (first column of Table 2), the tax basis of B is set to its fair market value ($1000) at the time of the split, with a simultaneous realization in the split year of taxable ordinary income of $1000. Token A, deemed to be the “original” token, retains the cost basis of $100. If and when Mina sells the two tokens, she realizes a capital gain of 50 (= 150 – 100) for A and a capital loss of 850 (= 150 – 1000) for B, respectively. Total realized income is therefore 1000 + 50 – 850 = 200.
The benefit of this approach in the context of cryptocurrency splits is that because the original cost basis continues with the parent chain, a simultaneous value relation between the parent and child chains need not be determined. In the rare instances of forks in which the child token has significant value, we find no evidence that the parent’s market price is affected.
A shortcoming of this approach is that, as discussed in Section 3.3, a strong relation should exist—namely, the combined value of the parent and the child at the time of the fork should be roughly the same as the value of the original token one instant prior to the fork, because unlike an actual treasure trove, the execution of an announced fork does not constitute a surprise. While the market price may appear to behave as if the fork is a surprise due to frictions such as market segmentation and imperfect information, adopting this approach means betting that these frictions will continue to exist in their current form. Furthermore, since our evidence of Section 3 suggests that split tokens are likely to be systematically overvalued and subject to large negative expected returns, taxing the receipt of new cryptocurrency as current ordinary income is likely to result in substantial upfront ordinary income (with the associated taxes) followed by substantial capital losses.
Other shortcomings include the need for the identification of a parent chain, a fair market value assessment for the new token at the time of the split, and an assumption that the taxpayer has immediate access to every token. In particular, the last assumption is highly problematic because of the overwhelming prevalence of nuisance forks, and the mutual disconnect between the announcement of a fork, the tradability of the token on exchanges, and the actual existence of a new live blockchain. Finally, the treasure trove approach may also induce downward price pressure in the market by incentivizing the sale of the split tokens—in order both to raise cash to pay tax bills, and to harvest losses if the tax basis of the split token is set too high.
4.2. The “Asset Split” Approach
Another analogy is to represent a chain split as an asset split such as a land subdivision or a corporate spinoff. Under this approach, a blockchain can be seen as a business that produces payment services, the miners (i.e., the nodes that provide verification services in exchange for some form of compensation) as its workers, and the blockchain tokens as shares (both stock shares and cryptocurrency tokens are abstract, proportional claims on the underlying business). Upon a stock spinoff, the investor retains the shares of the parent company and receives some new shares of the formerly private subsidiary. In a split, the investor receives tokens in a new blockchain in addition to the existing one.
One important way in which this analogy fails is that stock spinoffs are conscious decisions made by the parent firm, and therefore the subsidiary is rarely in direct competition with the parent. Indeed, often the newly-public, spun-off firm remains controlled by its former parent. In contrast, chain splits create a completely independent entity, not unlike a disgruntled employee opening a competing firm. Typical chain splits in which both branches of the chain have meaningful value arise from disagreements between developers or users and therefore pitch the two post-split chains in direct competition with one another. In general, this approach assumes that the fork does not, at a first approximation, create or destroy value, but is instead a mere rearrangement of the ownership structure. In practice, we do not fully understand the factors that determine the value of a cryptocurrency.42
If a cryptocurrency event is treated as a spinoff, the tax basis should be apportioned between the parent and the subsidiary (child) based on their contemporaneous fair market values at the time of the split. This model is the closest to the tax treatment in the United Kingdom, although the relevant regulations do not provide guidance on how exactly to allocate the basis.43
In Mina’s example (second column of Table 2), the acquisition cost of the original asset is apportioned based on fair market value at the time of the sale, leading to a 50/50 split of the original $100 cost basis. If and when Mina sells the two tokens, she realizes a capital gain of 100 (= 150 – 50) for each token. Total realized income is therefore 100 + 100 = 200.
An advantage of this approach is that it does not rely on a violation of finance theory because it does not assume that wealth was created, let alone accessed, at the time of the fork or the split. An additional advantage of this treatment is that it does not require a determination of which chain is the parent. Finally, attempting to apportion the tax basis based on fair market values equalizes (at least, in expectation) the tax cost of selling each asset, and therefore minimizes the resulting trading distortions.
The disadvantages are two. Unlike in a traditional stock split, taxpayers holding their coins at a custodial exchange may receive the child token with arbitrary delay, if at all. In the interim, a taxpayer seeking to sell the parent token would need to apportion part of the basis to the child without knowing if and when the child will become accessible. Finally, apportionment of the original cost basis is likely to be challenging because (i) the child token is likely to be overvalued, and (ii) comparable market values may not exist if parent and child are sold on different exchanges.
4.3. The “Calving” Approach
One last possible analogy is to represent the split as the birth of new livestock. If the livestock is held by the individual taxpayer as an asset, the tax consequences of the birth of new offspring materialize only when the offspring are sold, and the entire sale proceeds are treated as a capital gain for the taxpayer (i.e., the basis is set to zero). This approach, too, has intuitive appeal. The blockchain’s data and rules can be thought of as its genetic material. Further, unlike in the case of asset splits, the post-split chains are not “pieces” of the original chain—just as the calf is not a “piece” of the cow. Like livestock, whose ability to bear offspring is not a function of the number of past births, a chain’s ability to split is not a function of its past splits. Because of this ex ante uncertainty on the total number of offspring, it is not obvious how to apportion the acquisition cost upon successive births.44
One important way in which this analogy fails is that in the case of livestock there is no doubt about the identity of the parent, whereas for cryptocurrency splits the identity is not obvious, and at times there is open disagreement.
If a cryptocurrency event is treated as calving, the cost basis of the tokens on the child chain should be set to zero. This model is the closest to the tax treatment in Japan, where the new cryptocurrency is written on the books with a tax basis of zero.45 This model is also essentially consistent with existing U.S. proposals.46
In Mina’s example (third column of Table 2), the tax basis of token B is simply set to zero, and the acquisition cost of A is left unchanged at 100. If and when Mina sells the two tokens, she realizes a capital gain of 50 (= 150 – 100) for A and a capital gain of 150 (= 150 – 0) for B, respectively. Total realized income is therefore 50 + 150 = 200.
An advantage of this approach is that it is agnostic about whether value was created, and it assumes that no income was realized upon the fork or the split, sidestepping the need for valuation or other difficult determinations. Another advantage is its simplicity. The acquisition costs of each token can be determined independently of one another and of any market information, and therefore concerns about token access, token value measurement, and comparable value do not apply.
This treatment has two disadvantages. First, the identification of parent and child chains could be challenging for an honest taxpayer and a boon for a malicious one, who could name as parent any worthless token, sell it for near-zero, and take an immediate tax loss for the entire amount of the basis. The tax authority would then have the unrealistic burden of auditing and possibly challenging individual parent/child determinations. Note that the treasure trove approach also requires a parent/child determination, but it does not suffer from this problem because it is in the interest of the taxpayer to label as “child” the token with the lower market value.
Second, if both chains have substantial economic value, the chain that has been assigned a tax basis of zero will be more costly to sell, from a tax perspective, than the chain whose tax basis remains the original acquisition cost. The resulting trading distortion will increase the price of the child token (likely already overpriced) as taxpayers become locked-in to avoid realizing the gain, constricting the supply of the token and increasing its price.
The IRS’s recent Revenue Ruling 19-24 establishes that the receipt of new cryptocurrency following a fork is deemed taxable income, whose amount is to be determined based on the newly minted cryptocurrency’s fair market value. In this paper we have raised a number of serious questions about the applicability of this treatment. These issues go beyond those raised by existing policy proposals47 and academic studies48 that were published prior to or around the ruling. The evidence we present paints a pessimistic picture: it may not always be possible to determine (i) which of the two split tokens is the “new” one that has been “received”; (ii) whether the taxpayer’s wealth has increased or has merely been rearranged; (iii) even if taxpayer’s wealth has increased, whether the additional wealth is accessible, and (iv) if it is accessible, whether the taxpayer has accessed a non-negligible amount of wealth, and if so how much. In particular, our evidence suggests that newly minted cryptocurrencies are likely to be systematically overvalued. Using observable market value at the time of the split to determine an immediate tax liability is therefore likely to result in a large initial current income followed by large capital losses.
In light of these considerations we recommend, in line with existing policy proposals, that new cryptocurrency received upon a fork should not be taxed at receipt. Instead, its basis should be set to zero, and it should be taxed at sale. Regardless of whether realized income is deemed ordinary or capital gains, this treatment would help mitigate the observed overpricing and problems stemming from price deviations across markets. Additionally, given that we document a high failure rate for forks, this treatment also reduces the burden on the taxpayer to claim tokens that are not sufficiently valuable to compensate for the effort of claiming.
Unlike existing proposals, however, we highlight that even such an approach is not a panacea because of the first problem we identify: there are no universal objective criteria to identify the parent, and malicious taxpayers could exploit the inherent uncertainty and manufacture large loss deductions by labeling a worthless coin as “parent” and selling it. To mitigate this problem, guidance should be provided as to objective criteria on which the parent/child determination could be based (and later possibly audited), with the understanding that these criteria cannot be universal.
We also recommend allowing voluntary reporting of the acquisition of new tokens as soon as the taxpayer believes he or she has acquired them and exercises “dominion” (i.e., control) over them. Such voluntary reporting would be in the interest of the taxpayer because it would result in the longest possible holding period, increasing the likelihood of favorable long-term treatment of any subsequent gains.
One issue our paper has not covered is that of consistency between the tax treatment of futures contracts and that of the underlying child tokens. While the cryptocurrency futures trading market is still small, futures trading for announced tokens has been available in anticipation of many forks. A taxpayer expecting to receive a token from an announced fork can either wait to receive the token and sell it, or short a future contract before the fork and have the token delivered to the counterparty when issued.49 The tax consequences of these two actions cannot be too different without creating perverse incentives.
While the taxation of cryptocurrency forks may seem a niche technical topic, increasingly many intangible assets exist and are defined within distributed protocols, whose recordkeeping devices (blockchain, the Internet’s Domain Name System, etc.) possess the ability to split. Many of the issues we raise are likely to exist for many of these assets, and our analysis has much broader applicability.