Bitcoin VIX Signals Red Alert for Variance Swaps
Bitcoin VIX indices recently hit all-time highs. What sort of sums will the issuers of bitcoin variance swaps be paying to the lucky few who purchased them recently?
The greatest bitcoin bashing so far occurred in the early hours of 12 March during a coordinated sell-off on bitcoin futures and perpetuals exchanges, Huobi and BitMeX in particular (see our article on this). Options followed suit, as we can see from the CryptoMarketRisk ticker x.VXBT shown above, which represents the bitcoin VIX of maturity x days for x = 7, 14, 21 and 28.[1]
These indices are derived from an implied volatility smile surface, such as that shown in Figure 2, which is another way to visualise the traded prices of options at a particular moment in time.[2] Typically, the option’s moneyness (measured as K/S, where K is the option strike and S is the current bitcoin price) ranges from 0.75 (OTM puts) to 1.4 (OTM calls) and the maturity is up to 6 months (T = 0.5).[3]
Returning to Figure 1, from 1 Jan until 12 March the indices displayed a typical contango term structure, with only very brief periods of backwardation, fluctuating around the 60% level. For instance, the 28-day index at 60% is equivalent to monthly return standard deviation of a bit less than 17% — or (very roughly) an expectation of +/– 3 x 17% = +/– 50% returns on bitcoin over the next 28 days. With 99% confidence.
But at 10:30 UTC 12 March the bitcoin price fell by more than 30% in less than 60 minutes, from $7500 to $5500 (see our article on this). This precipitated a jump in the VXBT indices, first from 60% to 80% and then up to 140% within 2 hours. The further recent falls in bitcoin prices have sent the 7.VXBT to 250%!
Keep calm and carry on….breathe….OK?
First, the CBOE methodology is flawed. It is heavily tilted towards very low strike puts. These options, if written at all, can reach ridiculous prices when prices take a nose dive. In fact, the VIX index reflects nothing more than a widespread market panic at such times. That’s why its called the ‘fear gauge’ (see our article on this).
The 28.VXBT index is more presentative of real expectations about the evolution of volatility. Currently at 145%, that’s a monthly standard deviation of 40%, or (very roughly) +/– 120% returns on bitcoin over the next 28 days. With 99% confidence.
Besides the information about volatility and expectations of returns, the VXBT indices can be used as fair-value strikes for bitcoin variance swaps, such as those offered by boutique structured product companies such as GSR.[4] These swaps have pay-offs defined by the difference between a floating ‘realised variance’ leg and the fixed swap rate. Issuing variance swaps is a risk business. The variance risk premium is typically negative because bitcoin investors are happy to pay for the insurance these swaps provide.
Most of the time, with VXBT remaining stable (at around 60% in the figure above) the variance-swap issuers will pocket a nice premium. When the swap terminates the settlement will be from the buyer of realised variance to the seller, e.g. GSR. However, when bitcoin markets nose dive, as they have just recently, the issuer have massive pay-outs to their customers.
For example, consider this 28-day variance swap, hypothetically agreed on 18 February as follows:
- The 28.VXBT index was at about 60% then, so let’s assume the swap rate agreed was 60%
- Supposed the agreed notional was 250 USD vega
In other words, the pay-off from the issuer to the buyer, made when the swap terminates on 17 March, will be 250 [RV — 60²] where RV is the realised variance, which is derived from the sum of squared daily log returns on bitcoin between 18 February and 17 March.
On 18 March, the swap settles. The RV turned out to be 200.[7] So the issuer of this hypothetical variance-swap has to pay the buyer:
250 [200² — 60²] = 250 [40000–3600] = 9.1 million USD
One heck of a big pay out for a vega of just 250 USD!
Issuing variance swaps is similar to writing OTM puts. But even more risky. Most of the time, the market maker makes a profit.[8] But when the underlying crashes they become liable for some huge insurance pay-outs.[9]
No wonder the market for US equity variance swaps dried up during the 2008–9 financial crisis.[10]
Carol Alexander & Arben Imeraj.
@CryptoMarketRisk, QFIN, University of Sussex
[1] These volatility indices mimic the CBOE VIX methodology. The VIX is widely accepted to represent S&P 500 option trader’s expectations of volatility over the next 30 days. [The CBOE seem to have taken that interpretation off their website. But a google search on the term VIX brings it back everywhere else.]
[2] Put the traded price on the left of the Black-Scholes formula, use the option characteristics K (strike) and T (maturity) and forget interest rates, these are zero now anyway ….then back-out the volatility that is implied by that market price.
[3] Options (typically deep OTM calls) with very ow trading volume are excluded. Some OTM puts had over 2000 contracts traded on this day. The options used need sufficient volume on Deribit to avoid stale or artificial prices.
[4] See https://medium.com/gsr-trading/introducing-cryptocurrency-variance-swaps-293724914c2b
[5] Market convention is to quote RV as its square root in an annualised form, in percentage points. Put another way, the realised volatility was 200%.
[6]For instance, a swap with the same terms but agreed on 1 August with swap rate 100% (the 28.VXBT was 100.1% on that day) would have returned a profit of 1.34 million USD to the issuer, because the RV between 1 Aug and 29 Aug was 68%, so the pay-off was 250 [68² — 100²] USD.
[7] Unless they are hedged. But that’s another story.
[8] But they have come back, albeit with limited liability e.g. in the form of ‘corridor swaps’.
