BTC ETFs: The SEC’s Approval of Index Providers
The ten listed Bitcoin ETFs utilize four index providers with four different calculation algorithms. The Securities and Exchange Commission (SEC) has approved all of them, indicating that the specific benchmark is not a primary concern for the regulatory body as long as it fairly represents the market it is intended to measure.
This article examines the theoretical and empirical differences between these reference rates. The following abbreviations will be used:
- VWAP: Volume-weighted average price
- VWMP: Volume-weighted median price
- TWAP: Time-weighted average price
- ETFs: Exchange-traded funds
The four reference rates used in the spot bitcoin ETFs are:
- A. Spot price at 4 PM on the most liquid exchange, determined according to rules. This exchange selection typically remains consistent month-to-month.
- B. VWMP for the entire hour.
- C. One-hour TWAP of twelve 5-minute VWMPs.
- D. One-hour TWAP of twenty 3-minute VWMPs.
From the constructions, we can see that A≠B and A≠C but C≈D.
Comparisons and conclusions
A vs B: Large differences
A vs C: Medium differences
C vs D: Small differences
A vs B: Large differences
The reference prices for A and B can vary significantly. A may use a transaction occurring as early as 3:01 PM, while B consistently references 4 PM transactions. BTC can move greatly in an hour. Looking at the last three months, hourly price moves of -4.5% and +2.5% have happened. On ¼ of the hours, the price decreased 15bps to 492bps. On ¼ of the hours, the price increased 18bps to 266 bps. The proof is in Appendix 2. (Although the prices can be very different, the prices can also be very close. A may select a price at 3:59 pm, in which case A≈B.)
A vs C: Medium differences
Algorithm A is straightforward, utilizing the 4 PM price on their selected exchange. The numerical value will differ from C and D on many days because if the price trends upwards from 3 to 4 pm, the TWAP will be much lower than the 4 pm price. If the prices move linearly from $100 to $106 during the hour, the TWAP is $103, whereas the 4 pm price is $106, which is a whopping 3% difference.
C vs D: Small differences
C and B both deploy a TWAP over one hour. They differ in exchange selection and the length of each time slice.
Slicing the hour into 3 — or 5-minute slots has an effect, but it will be small because using 12 or 20 data points in the average will only slightly impact the daily benchmark value. Note that 20*3=60 and 12*5=60. The more you slice the hour, the thinner each slice becomes.
While the two providers use different sets of exchanges, this has a limited impact on the final price. Bitcoin arbitrageurs actively trade to exploit price discrepancies, ensuring prices remain relatively aligned across exchanges. A utilizes Coinbase, Bitstamp, ItBit, LMAX, Gemini, and Kraken. B uses the top 5 exchanges according to CCData, which currently are Coinbase, Bitstamp, Kraken, Bybit, and Bullish.
The prices will be very close: C≈D.
Vinter algorithms
Vinter offers three different calculation algorithms: VFIX, VAP, and BRR.
- VFIX: Vinter’s Fixing (VFIX) is the real-time value at 4 pm. It’s the median spot price at 4 pm.
- BRR: The Benchmark Reference Rate (BRR) by Vinter is a TWAP from 3 to 4 pm of volume-weighted median prices.
- VAP: Vinter’s Average Price (VAP) is a TWAP from 3 to 4 pm of median prices.
Read more on the website https://vinter.co/reference-rate or the methodology document https://methodology.vinter.co/vinter/reference-rates#daily to learn all details.
We conclude the following by comparing these three algorithms (1–3) with the previous four (A-D).
- VFIX is similar to A.
- BRR is similar to C and D.
- VAP is a simpler algorithm than C, but since both are one-hour TWAPs, their numerical values are close.*
*On average, 0.008 bps which is less than one-hundredth of a basis point (using daily data from Feb 2023 to Feb 2024). Since the average daily diff is small, the reference rates are numerically similar. For details, see Appendix 1.
The numerical can be summarized as follows: 1≈2 and 2=C and 3≈C. A precise numerical comparison which shows that 2=C is a blog post for another day.
Appendix 1: VAP vs C
Detail for math/stats geeks below: percentiles of the difference in basis points.
count 273
mean 0.0080
std 0.3718
1% -0.7970
25% -0.1349
50% -0.0221
75% 0.1296
99% 1.0582Appendix 2: Hourly BTC returns
Detail for math/stats geeks to back up the previous sentence, here is a numerical boxplot for 83 days of hourly BTC returns
min -4.92%
25% -0.15%
50% 0.005%
75% 0.18%
max 2.66%The boxplot of absolute difference shows that ¼ of the hours the price moved more than 32 bps, all the way up to a 492 bps move. BTC is volatile so a lot can happen over an hour.
min 0.00012%
25% 0.0707%
50% 0.1706%
75% 0.3250%
max 4.9160%Appendix 3: How the VWMP works
The VWMP is the price for the transaction whose volume is the median volume. In the illustrative table below, the median volume is 12, and the price for that transaction is $102, so the VWMP is $102. One way to understand the VWMP is to imagine a long table of all transactions with two columns (volume, price) and then select the price for the transaction in the middle.
To learn how the VWMP compares to the VWAP, read this post:
