![US Bitcoin ETFs: Understanding fair value (1) US Bitcoin ETFs: Understanding fair value (1)](https://i0.wp.com/questdb.io/img/blog/2024-01-16/banner.webp)
QuestDB is a high performance time-series database with SQL analytics that can power through data ingestion and analysis. It's open source and integrates with many tools and languages. Give us a try!
Bitcoin Exchange-Traded Funds (ETFs) are live in the United States! Asanticipated, the Security and Exchange Commission approved 11 products to starttrading on regulated exchanges. Investors are excited! But are they going to geta fair price? ETF launches have interesting dynamics, and in this post we'lllook at how some of them materialized.
ETF trading dynamics, the secondary market
A primary market is when investors buy securities directly from the issuer, andnot from other investors. An Initial Public Offering (IPO) is one such example,where the funds raised go to the company in exchange for initial share. Thereare then secondary markets.
The secondary market is the most intuitive. Shares of a stock like AAPL
typically trade only on the secondary market. After shares are issued, theychange hands on a secondary market such as a stock exchange. Just like stocks,ETFs trade on the secondary market. Unlike stocks, ETF shares also trade on the'primary' market, which we'll explain further into this post.
ETF issuers typically contract market-makers to enrich the secondary market.Market-makers post executable quotes most of the time in the ETF orderbook, sothat it is liquid.
This liquidity makes the product more attractive to investors as they know theycan execute against these quotes to enter and exit a position. In other words:They will avoid a situation where there is no one to buy their ETF if they needto liquidate for whatever reason.
Considering only the existence of a secondary market for the moment, we can seethat a market-maker is constantly posting quotes with a fixed spread. But whatif everyone was selling to the market-maker?
In that case, they would sometimes offset/shift/fade (terminology varies bydesk) their prices. The more they buy, the lower their bid and ask will be. Thisis both as a measure to mitigate concentration risk, or buying a lot of the samething, and to mitigate inventory.
Consequently, the ETF may trade away from its fair value as a result of marketpressure. In the case of Bitcoin ETFs, we can imagine that most people would bebuying on the first day, and so the question becomes: Is there any particular,observable skew on launch day?
To determine this, we first need to calculate the theoretical ETF value.
Pricing fair value of an ETF
At the core, most ETFs represent a basket of assets. The price of such an ETFshould be equal to the cost of replicating the basket. For example, buying theindividual constituents should be equal to the cost of buying the index.
In this case, the basket forIBITaccording to the issuer's website on Jan 11 close is worth 24.94 (Net AssetValue) with a spot on BTC-USD of 43,770.04. Further, the issuer posted that thefund contains 99.99% of BTC, and 0.01% of cash.
Therefore, we can price the ETF synthetically as follows:
SELECT
timestamp,
price * 24.94 * (1 / 43770.04 * 0.9999 + 0.0001) AS IBIT_IMPLIED_FROM_BTCUSD
FROM
trades
WHERE
$__timeFilter(timestamp)
AND symbol = 'BTCUSD'
This gives us the following fair value as implied from the price on BTC-USD, astraded on Coinbase. We can also superimpose the price of IBIT directly and seethat they are quite close and correlated, as expected:
We can generalize this concept to ETFs with more constituents. Their fair valuecan be calculated as a sum-product of the quantities + weights and the realtimeasset prices.
Explaining the concepts of Premium and Discount
Shares of stock exist in a relatively static amount. While companies can issuemore stock, and some events such as convertible bonds or warrants can result instock issuance, external participants don't have control on the number ofshares. After all, the number of shares is not meant to change every day.
In practical terms, this means that if someone sells a share of Company XYZ,then they need to source this physical share. They do so either by buying itfrom someone else or by borrowing it. Their ability to do so depends on:
- The general availability of shares - are there many in existence?
- The float - are there enough people willing to sell their shares, or are theholders keeping them?
This is also the case for many funds such as mutual funds. Since the number ofexisting shares for Fund ABC is not totally variable, one must find a willingseller to be able to purchase such share. And of course, a willing buyer to sellit. Consequently, shares of a mutual fund can sometimes trade far off of theirfair value.
If there are more buyers than sellers, then the market price can be higher thanthe fair value. In this case, the fund is said to trade at a 'Premium'. If thereare more sellers than buyers, the market may trade at a 'Discount' to the fairvalue.
The level of Premium and Discount can be calculated by comparing the level atwhich shares of the fund are trading to the official Net Asset Value which iscalculated based on the assets it holds:
The Grayscale Bitcoin Trust, before it recently became an ETF, was subject tosuch fluctuations. In 2020, many market participants wanted to buy the fund togain exposure to Bitcoin. However, there was a long delay in creating new sharesof the fund. The Premium reached nearly 40%, which meant that one would buy thefund at a price 40% higher than the value of the assets it contained. Not a gooddeal.
Following the lending bankruptcies in 2021, the fund started trading at a deepdiscount reaching nearly 50%. This was mainly because selling the fund requiredfinding a buyer, and there was no way to 'turn the fund back into Bitcoin' andsell the Bitcoin. In other words, once they bought the fund, investors werelocked in until they could find a buyer at an acceptable price, or until thefund was approved to trade as an ETF.
As the ETF conversion became more likely, the discount converged back to zero:
ETF trading dynamics, exploring the primary market
Unlike shares of stocks and mutual funds, ETFs have a Creation/Redemptionmechanism. Such a mechanism is designed to make it easy to issue and destroyshares of the fund. As a result, this greatly reduces the Premium and Discount.
The redemption mechanism works as follows:
Someone sells 1 share of the ETF to a market-maker
The market maker is now long 1 ETF share. Since they don't want the marketrisk, they sell 1 BTC against it, and are then 'hedged'
The market-maker is an "Authorized Participant" (AP), which means they areallowed to create and redeem shares of the ETF directly with the ETF issuer
The AP contacts the issuer to redeem shares of the ETF as follows:
- AP gives 1 share of the ETF to the issuer
- Issuer gives the underlying assets to the AP (e.g. 1BTC) and 'destroys' theETF share
This is a "risk-less'"transaction. The issuer and the AP exchange 1 sharerepresenting 1 BTC against 1 BTC at the same time. Since the AP received 1 BTCfrom the issuer, they covered the short they initiated to hedge their positioninitially.
The flows are highlighted below, and all cash flows equalize. Though keep inmind that generally the ETF seller would receive a little bit less than X
toaccount for the spread:
This is Redemption. The Creation process is analogous but reverses the sides ofall transactions. In practice, the AP window to create and redeem has a littlemore friction than described above. There are also fees associated.
Sometimes there is a fixed fee per creation/redemption, sometimes it is avariable fee as a percentage of the transaction value. Sometimes it's acombination of both. There is also a minimum size which means you can't do itfor one share but rather, as an example, for 10,000 shares.
Calculating Bitcoin ETF Premium and Discount
We can use the above prices to derive the intra-day Premium and Discount with anASOF JOIN
of our theoretical price against the observed market price:
WITH
implied AS (
SELECT
timestamp,
price * 0.000569739 + 0.002494 AS IBIT_IMPLIED_FROM_BTCUSD
FROM
BTCUSD
WHERE
$__timeFilter(timestamp)
),
actual AS (
SELECT
timestamp,
price AS IBIT
FROM
IBITETF
WHERE
$__timeFilter(timestamp)
)
SELECT
implied.timestamp,
last((IBIT / IBIT_IMPLIED_FROM_BTCUSD - 1) * 10000) AS premium
FROM
implied ASOF JOIN actual SAMPLE BY 30s
The result on the 12th of Jan is as follows:
It appears that there was a significant premium at the start of day (~1%) whichthen gradually reduced as the day went on. Part of this could be because BTC-USDdropped by more than 6% during the day. We can generalize this approach andapply it to other funds such as BITC:
Explaining the Premium and Discount
As previously mentioned, market-makers are paid to provide executable quoteswith a maximum spread, over a minimum size. For example, this could be a minimumsize of $100,000 for 95% of the time, with a maximum spread on 0.5%.
On launch day, you can witness very directional flows. It's more likely thatinvestors will buy the fund than that they will sell it. Given this, amarket-maker can anticipate that, say, 90% of trades will be investors buying,and 10% selling. One can skew their prices to maximize value out of this:
Skew | Bid | Ask | Spread | Buy trades (10%) | Sell trades (90%) | Spread earned (Buy) | Spread earned (Sell) | Total | |
---|---|---|---|---|---|---|---|---|---|
Scenario 1 | 0 | 99.75 | 100.25 | 0.5 | 100 | 900 | $125 | $125 | $250 |
Scenario 2 | 0.25 | 100 | 100.50 | 0.5 | 100 | 900 | 0 | $450 | $450 |
On a day where flow are anticipated to be very imbalanced, there is a strongincentive to skew prices while keeping the spread constant. This is because itmaximizes market-making earnings.
We suspect this could be why we saw Premiums as high as 1.4% on the day. Here ishow skewed and unskewed quotes would look relative to fair value, resulting intoa Premium:
Although we said that the Creation/Redemption arbitrage should prevent largePremiums and Discounts, the mechanism is not completely perfect. For example:
- The minimum size for Creation and Redemption is generally quite large
- The Creation/Redemption can have costs. The Premium or Discount needs to besignificant enough to cover such costs
- While the mid is significantly higher than fair value, an arbitrageur wouldneed to sell the ETF into the market at the Bid price to complete thearbitrage. The bid is lower than the mid because of the spread, and may notbe far enough above fair value to make the arbitrage viable
There are other situations where such Discounts and Premiums can occur as aresult of market-makers shifting their quotes:
Existing risk: If a market-maker is short an ETF, or other ETFs with thesame risk profile, they may choose to bid higher. As a result, they may offerhigher as they maintain the spread constant to encourage people to sell tothem and discourage them from buying to keep their risk under control.
Inventory management: If the market-maker is short an ETF for a smallsize, they may skew their prices to incentivize participants to sell to themto close their short. They do so because it would be cheaper and morepractical than having to go through the primary market.
Flow management: A burst of buy orders after the fund appeared ontelevision and a popular stream.
Summary, check Premiums before you trade!
Hopefully with this post you have a solid grasp on the concepts ofCreation/Redemption and Premium/Discount for ETFs. You can use this knowledgecombined with your own market data and the holdings published by the issuers tocalculate your own fair values.
Most ETFs should trade close to such fair value. However, transient situationscan occur and result in a significant deviation. By arming yourself with yourown tools, you can avoid losing on trades by buying / selling 1% away from the'true' price.
While some ETFs can trade at Premium and Discount for a sustained period,chances are the deviation is temporary and will be reduced shortly. Either way,when buying or selling an ETF, it should be your right to know if you're gettinga fair price and you have all the information and tools at your fingertips!
Download QuestDB Open source under Apache 2.0. Blazing fast ingest. SQL analytics.