Jun 27, 2011

I received a number of follow-up question to the XIV post. While I cannot share the data for the underlying index (per data subscription agreement) I'll explain the methodology in detail. Here is the sample spreadsheet that you should reference.

Sheet1 has four numerical columns, everything is done in google docs - so all this is very basic. The first contains values for SPVXSP index, which is the underlying index for VXX but without the costs and interest. Their effect is minimal in comparison to volatility of underlying index.

The values for SPVXSP index are in 90K range, that is because the starting value of the index is 100K. It is just a number and really does not matter that is it unusually high. In the second column I calculate simple returns for the index, third I calculate the inverse return by multiplying return by negative one. In the last column I apply inverse returns to get prices, with starting value equals \$100. Again, all is very basic.

The table of monthly returns for the simulated index are on Sheet2. You may notice a large difference between simulation and actual XIV performance in the first month - that is only because the dates are different, and actual XIV has one less day than simulated version.

How risky is XIV? I apply bootstrap to monthly returns to find that there is about 60% chance that XIV will be up after one year, and 20% chance that it will more than double in price. This is of course contingent on previous return distribution to be true in the future, which we have no way of knowing. These results are of course excellent. On the downside there is about 12% chance of losing more than 50%, and expected maximum drawdown is 37%.

What is surprising to me is that return distribution is clearly positively skewed despite being a short-vol strategy. If you have any thoughts on this, please share or leave a comment.

Now let me clarify even though I think that XIV is a good product I don't plan on investing in it, because I think for me as a full-time trader it is far better to trade and actively manage risk than it is to buy and hold. You should decide for yourself if the product fits your investment objectives and risk tolerance.

1. Hi ~,
My assumption is that the positive skew is due to contango. Since the 2008/2009 crash I think it has been eroding the SPVXSTR at an average rate of 5% to 10% per month. I would expect that to skew the distribution.

-- Vance

2. Hi,
Thanks for the in-depth analysis. The positively skewed distribution is expected because it is a short-vol strategy. In the short run, long XIV is like shorting a 60-day term straddle on SPX while dynamically maintaining delta-neutrality. It's a non-directional short gamma strategy. It is a business of selling insurance. It is risky and deserves a premium.

3. @VanceH - I agree on positive return due to contango, but contango causing positive skew seems like a stretch.

@Kenny Yu - short vol strategies have negative skew, delta-hedging short vol has negative skew. Selling insurance is also negative skew. Maybe you mean something difference by negative skew that I do; please clarify.

4. Does anyone know why the XIV and VXX have decoupled from the VIX since about the end of August? The XIV and VXX typically are very well correlated with the VIX, although their amplitude is only about 0.46 x the VIX.

Since Jan 30, 2009 there has only been a handful of instances where the daily VXX has varied by more than 2 standard deviations from the 0.46 rule. This trend was broken starting 7/29 when 10 of 16 trading days experienced high VXX volatility. Said differently, the VXX is not as well correlated with the VIX since 7/29/2011.

5. Anonymous8/24/2011

I would venture to that that it is due to the fact that each futures price is determined by a market (albeit based on the vix index) but also on expectations. When the vix is in a "normal" range than the normal rules of ~.46 move on the front month applies. When all of a sudden traders larger expectations are thrown off, then that rule shouldnt hold. Its most notable when looking at similar levels of the vix and the difference in the first month (or xiv/vxx (granted backwardian added some difference). To give a guess and narrate the first spike of the vix was written off by traders more so as a vix spike that will quickly subside. When the spx fell back down and the vix took a second spike, traders probably though... oh sh*t this volatility is here to stay and adjusted their estimation (and the`.46 general rule) in the vix futures.

6. You could be right, but I'm not sure that XIV/VIX is useful at all due to contango/backwardation as you've mentioned. The daily percentage change in XIV/VIX reains very well correlated at -1.

If I look at the XIV, it would be priced at \$15.65 as of yesterday's (8/25/2011) close if the XIV dialy percentage change was -.46 x daily VIX percentage change. It actually closed at \$7.41! I'm wondering if there's something systemic/predictable quantitatively and if this could reverse leading to a significant run up in the XIV when volatility subsides.

7. I did not have the time to look into it, but I just saw that Jev K published a post on it http://matlab-trading.blogspot.com/2011/08/xiv-is-stealing-my-money.html