About Those Proposed Bitcoin ETFs
How Will They Really Work?
The crew was complete: it included a Boots—
A maker of Bonnets and Hoods—
A Barrister, brought to arrange their disputes—
And a Broker, to value their goods.
A Billiard-marker, whose skill was immense,
Might perhaps have won more than his share—
But a Banker, engaged at enormous expense,
Had the whole of their cash in his care.
— Lewis Carroll, The Hunting of the Snark (1876)
Bitcoin ETFs
A couple of years ago I wrote a note about the institutional structures necessary if cryptocurrencies are to become a meaningful feature of our economic lives. The note is here: https://tiemanninvestmentadvisors.com/portfolio-items/historical-perspectives-on-bitcoin/ My goal was not to argue against cryptocurrencies, but to argue that to overcome the inefficiency of the blockchain technology behind Bitcoin and other cryptocurrencies, a crypto-economy would require an extensive set of intermediaries, substitute instruments representing crypto-denominated claims, and off-blockchain book entry transaction systems to support any meaningful volume of commerce. Several sponsors now have proposals before the US Securities and Exchange Commission (SEC) seeking authorization to launch cryptocurrency-based exchange-traded funds (ETFs). At first glance, Bitcoin ETFs (we could talk about any cryptocurrency ETF, but I’ll stick to Bitcoin for convenience) would seem to fill at least some of these needs. But would they?
The basic argument for a Bitcoin ETF is that it would offer investors a convenient, liquid way to participate in the Bitcoin economy, and in a form – an exchange-traded fund – with nearly thirty years of successful market application and a strong recent increase in market acceptance. Unlike the most important existing investment vehicle, the Grayscale Bitcoin Trust (GBTC), which has often traded in the market at a substantial discount to its Net Asset Value (NAV, the value of its underlying holdings), investors hope that like most ETFs, Bitcoin ETFs would trade in the market at prices near NAV, tracking closely the fluctuations in the dollar price of Bitcoin.
The planned operations of the proposed Bitcoin ETFs may vary from offering to offering, but we know enough about the mechanics of Bitcoin and the mechanics of ETFs that we can make a reasonable sketch of how these investments would generally work. In addition, recent regulatory filings, such as the 19b-4 filing by NASDAQ regarding operational plans for BlackRock’s iShares Bitcoin Trust https://listingcenter.nasdaq.com/assets/rulebook/nasdaq/filings/SR-NASDAQ-2023-016_Amendment_1.pdf , provide important details.
We can expect that fluctuations in the share price of a Bitcoin ETF will mirror some sort of market price of Bitcoin, so ETF buyers can expect to participate in the ups and downs of the cryptocurrency. But ETF buyers will not be holders of Bitcoin itself, or of anything they can convert directly into Bitcoin. That means that Bitcoin ETFs might facilitate speculation in Bitcoin, but will do nothing to make Bitcoin easier to use. William Cohan has also made this point recently in the Financial Times. https://www.ft.com/content/2356d32d-f5d6-4017-9179-5787ced68118
The new ETFs may create increased speculative interest in Bitcoin, which may in turn generate increased trading activity. But the Bitcoin network, which clears primary, direct Bitcoin transactions on the blockchain (on which more later), probably does not have spare capacity. In any case, as the BlackRock-related filing indicates, ETF plans call for relying on Bitcoin custodians (Coinbase in the iShares case) to clear their trades – a remarkable feature of these securities, given the malfeasance that has come to light in the past year or so in the world of cryptocurrency exchanges.
To see why Bitcoin ETFs can only facilitate speculation in, but not use of, Bitcoin, and to see why custodians, exchanges, and even other Wall Street firms are crucial to their functioning, we need to take a look at the mechanics of both ETFs and Bitcoin.
Exchange-Traded Funds.
Exchange-Traded Funds (ETFs) have been a feature of the investment landscape for a quarter-century. Over that time, they have become increasingly familiar and earned widespread investor confidence and acceptance. ETFs are popular for good reason. Like mutual funds, they offer small investors convenient access to the benefits of professional investment management, usually at reasonable – and often low – cost. But traditional mutual funds are cumbersome to trade. Investors can only buy or redeem their shares at the end of the trading day, transacting at a price calculated from the Net Asset Value (NAV) of the fund’s portfolio. An investor buying into a mutual fund receives newly-issued shares. When a fund holder makes a withdrawal, the fund extinguishes those shares. For some classes of funds, investors buying shares may also incur sales charges, and sponsors may also restrict the frequency with which investors can move in an out of their funds – or close the funds to new investors altogether.
ETFs, in contrast, trade like stocks throughout the day on a stock exchange (hence the name), and at market prices, rather than at NAV. While brokers may charge commissions to handle ETF trades, several discount brokers now charge no commission at all. But there is a wrinkle. Ordinary investors buying and selling ETF shares trade existing shares with one another. The market price at which they trade can drift away from the fund’s NAV, to either a premium or a discount to NAV. The key feature of ETFs is the mechanism designed to tie their market prices close to NAV. This is the role of the Authorized Participants (APs), Wall Street firms with the contractual privilege of engaging in special transactions, generating what the ETF world calls creation units and redemption units.
When buying interest in an ETF is high, the natural course of this demand would be to create inflows into the fund, but so long as the number of shares remains fixed, the buying pressure tends to raise the fund’s market price above its NAV. Conversely, when more investors are interested in leaving the fund, if the number of shares remains fixed, the market price will tend to fall to a discount to NAV. The role of the APs is to provide the mechanism for increasing the ETF’s share count when buying interest drives inflows, and for decreasing the share count when selling interest produces outflows. The APs’ incentive for doing so is their privileged opportunity to capture an arbitrage profit between the ETF’s market price and its NAV. Here is how it works:
APs have the contractual right to add or withdraw shares from an ETF at any time, although only in sizeable blocks. Presumably they will only do this when it is advantageous. If buying pressure raises an ETF’s share price to a premium over NAV, APs become interested in generating creation units. For most ETFs, the AP can deliver into the fund a portfolio of securities closely matching the fund’s holdings corresponding to a block of ETF shares. This is a creation unit. In exchange, the ETF issues new shares to the AP, effectively at NAV. The AP is then free to sell the new ETF shares in the open market at the elevated market price, earning an arbitrage profit and in the process driving the market price back toward NAV. In the opposite case, where the ETF price has fallen to a discount, the APs become interested in generating redemption units (the industry somehow failed to call them “destruction units”). The AP buys up a block of ETF shares in the open market at the discounted price, and then delivers those shares (the redemption unit) to the fund in exchange for (in most cases) a proportional slice of the ETF’s holdings, which it is again free to sell in the open market. The ETF cancels the shares the AP delivered in.
The creation/redemption unit process first proved its effectiveness with the early S&P 500 index ETFs. By the time those funds appeared on the market, large brokers and investment managers were already accustomed to trading baskets of securities matching the S&P 500. Delivering a basket of 500 stocks for a creation unit, or receiving that basket in exchange for a redemption unit, was a simple application of this “program trading,” which has now become so commonplace that it seldom receives any mention in business reporting. One of the side benefits for the ETF manager is that this scheme outsources most of the fund’s trading to the APs.
The Proposed Bitcoin ETFs.
NASDAQ’s filing relative to the proposed iShares Bitcoin Trust indicates that the Bitcoin ETFs would make use of the Authorized Participant and creation/redemption units concept, but it would work differently from an S&P 500 index ETF. (Other Bitcoin ETFs may not operate quite the same way, but I suspect the iShares plan is typical. If I’m incorrect, please let me know.)
The iShares Trust plans to hold Bitcoin in a custodial account at Bitcoin exchange Coinbase. APs wishing to generate creation units will deliver cash, rather than Bitcoin, and the Trust will arrange to use that cash to buy Bitcoin, either from Coinbase’s own Bitcoin brokerage or other parties, but in any case, in trades cleared through Coinbase. The Trust will then issue new shares to the AP against the Bitcoin they buy in those trades, with the risk in the price the Trust pays for the Bitcoin falling on the AP. APs may also put in orders for redemption units. In that case, the Trust will sell Bitcoin from its holdings (again in trades cleared through Coinbase), cancel the shares corresponding to the redemption unit, and deliver the cash proceeds to the AP –the price risk again falling on the AP.
It's worth looking more closely to see why Coinbase (or some other custodian) is crucial to the Bitcoin ETFs. (And to contemplate what the consequences might have been had a Bitcoin ETF launched in 2021, with Sam Bankman-Fried’s FTX as custodian.) For that, a review of how Bitcoin works will be helpful. If you’ve clicked through to read my earlier paper, you can skip the next few paragraphs, which I’ve lifted from it intact.
The paper[1] outlining the design proposal for Bitcoin describes it not as a currency, but a mode of electronic payment not reliant on financial institutions as intermediaries. Bitcoin itself is just an arbitrary unit of account. Rather than requiring users to trust either a governmental issuer of currency or a bank, it relies on a peer-to-peer network. Anyone willing to download and run Bitcoin software can participate.
Bitcoins exist entirely in a great general ledger called the blockchain, and bitcoin holdings at any point in time are nothing but the end result of a calculation tracing all blockchain transactions since Bitcoin Day 1, early in 2009. This calculation associates each bitcoin position with a specific address, or “wallet.” Each holder owns one or more wallets.
The Bitcoin network and blockchain have no central, authoritative record-keeper. Any time a bitcoin transaction occurs, the relevant data emanate throughout the Bitcoin peer-to-peer network from the transaction’s point of origin to all users. The basic data for the transaction include the address (the wallet) of the sender, an address for the recipient, the number of bitcoins in the transaction, and authenticating information. The authentication, both of individual transactions and the blockchain, relies on data cryptography (hence the term “cryptocurrency.”) Each transaction record contains a digital signature derived from a password called a private key, which, hopefully, only the sender knows.[2] Encryption preserves the privacy of this key while authenticating the signature.
The system has to keep users from spending the same bitcoin several times. This is the most ingenious part of the design. Users — not a centralized party — validate transactions and commit them to the permanent general ledger, the blockchain. Users performing this function are bitcoin “miners.” Any user with a suitable hardware rig, the right software, and willingness to pay for the power to run them can become a miner. As transactions occur and users broadcast them to the network, miners pick them up and assemble them into blocks. The miners verify the transactions by comparing them with the existing account (address) balances implicit in the blockchain history, and then apply a cryptographic algorithm in an effort to be first to create an encoding of a new block of transactions that meets a specific set of requirements. Cryptographers call this encoding a hash, and the encoding algorithm the hash function.
Running the hash function isn’t very hard, so the Bitcoin protocol adds a “proof of work” test to increase the effort the miners have to expend. To validate a block, the encoding must meet an arbitrary format requirement — it has to start with a minimum number of zeroes. The raw data of the block won’t produce that result, so miners have to find a string of nonsense data, called a nonce, which, when added to the block, yields a hash meeting this requirement. The way to generate this result is by the brute-force method of guessing a nonce, calculating the encoding, seeing whether it works, and if not, going back and guessing another one. This is the “complex mathematical puzzle” news reports sometimes say miners must solve.[3]
Eventually, a miner will hit on a nonce that produces the required encoding. That miner broadcasts the result to the network, adding the newly mined block to the end of the blockchain. The successful miner receives a bounty, or reward – the privilege of beginning the next block with a transaction awarding itself a specified number of newly created bitcoins, along with any fees that users have offered miners to prioritize their transactions. (Miners do not try to validate all outstanding transactions at once, because the computing power and time necessary to calculate a hash for a block increase with its size. Bitcoin traders offer fees to induce miners to prioritize their transactions.) Currently, the bounty stands at 6.25 bitcoins, worth around $250,000 at prices exchanges[4] are quoting. The Bitcoin protocol includes a rule that cuts the reward in half from time to time. The next “halving” is likely to occur soon (early in 2024); it will cut the reward to 3.125 Bitcoins. The reward will fall to zero when about 21 million Bitcoins exist, after which miners will have to rely entirely on transaction fees.
By design, the average time interval between successive blocks is about ten minutes. As miners apply more computing power, they verify blocks more quickly. If successes become too frequent, the protocol increases the number of initial zeroes required for a valid hash, increasing the difficulty of mining.
As appealing as the idea of a distributed ledger with no central point of control may be, the overhead and inefficiency of operating the Bitcoin network are staggering. The network processes only a few hundred thousand transactions per day, whose average size appears to be on the order of one Bitcoin – not a quantity suitable for everyday commerce. Bitcoin users, too, face an administrative task in tracking their holdings, safeguarding their private keys, and verifying that their transactions have made it to the blockchain, a process that may take multiple Bitcoin mining cycles.
BlackRock presumably has the resources to build a system for holding the iShares Trust’s Bitcoins directly, but they have chosen instead to rely on Coinbase to act as custodian. Using a custodian does more than just relieve BlackRock of an administrative burden. Since the Trust will require all trading counterparties to trade through Coinbase, I would expect that a large fraction of these trades will not occur on the blockchain at all, but instead clear through journal transfers of wallets Coinbase holds, much like shares that a brokerage house holds for customers in “street name.” Otherwise, large flows into or out of Bitcoin ETFs could swamp the Bitcoin network. It may or may not be of significance that a search of the word “audit” in the iShares-related 19b-4 filing produces no matches.
Issues and Conclusions.
The design of Bitcoin ETFs relies heavily on both custodians and Authorized Participants. Coinbase has a solid track record as a Bitcoin custodian, but it remains a locus of operational risk, and as the experience of FTX has shown, cryptocurrency custodians and exchanges are not foolproof. Still, BlackRock, which has its own reputation to protect, has evidently evaluated those risks and decided to proceed.
Any ETF also depends on its APs. APs will only generate creation and redemption units if they can expect to profit. For the iShares Bitcoin Trust, APs will deliver or receive cash, but bear the price risk in the purchase and sale of Bitcoin for their creations and redemptions, and they will have to rely on the Trust to execute those trades for them. Especially when the price of Bitcoin is volatile, APs may choose to wait until the ETF’s premium or discount has grown large before they are willing to take the risk of generating creation or redemption units. Alternatively, the Trust may find some way to subsidize the APs’ risk, either at BlackRock’s expense or that of the ETF’s shareholders – although the SEC might take a dim view of a hidden cost to shareholders.
The main issue with Bitcoin ETFs, though, is the one that William Cohan identified: They may facilitate Bitcoin speculation, but they do nothing to promote or facilitate the use of cryptocurrencies in economic real life.
[1] Satoshi Nakamoto, “Bitcoin: A Peer-to-Peer Electronic Cash System” available at https://bitcoin.org/bitcoin.pdf. “Satoshi Nakamoto” is a pseud- onym, and the true identity of “Satoshi” – whether an individual or a group –remains a closely-guarded secret. The paper looks like an academic working paper, such as a researcher might circulate professionally before submitting it for publication in an academic journal.
[2] A more complete description of bitcoin transactions is at https://www.coindesk.com/learn/bitcoin-101/how-do-bitcoin-transactions-work. The Bitcoin Foundation’s site, bitcoin.org, also gives a good deal of background.
[3] If we think of the hash as an essentially random string of zeroes and ones, then each bit has a 50-50 chance of being a zero. If the solution requires 20 leading zeroes in its hexadecimal (base 16) representation, for instance, then on average the proof-of-work would require 280 = 1.2 septillion (1.2 x 1024) trials.
[4] Bitcoins, of course, aren’t convertible into conventional currency unless someone is willing to trade one for the other. Quoted prices vary rather widely. The price is also volatile, so this estimate will be wrong in a short time. For this note, I am using a price of $40,000, a level typical for early 2024.