Introduction
A quick intro to VIX ETPs (some are ETFs, others are ETNs)1 before we get to the meat: the VIX itself is not tradable, only futures on the VIX are. These futures do not behave like equity index futures which move in lockstep with the underlying, for a variety of reasons. A way to get exposure to these futures without holding them directly, is by using one or more VIX futures-based ETPs. These come in many varieties (long/short, various target average times to expiration, various levels of leverage).
The problem with them, and the reason they fail so badly at mirroring movements in the VIX, is that they have to constantly roll over their futures holdings to maintain their target average time to expiration. A 1-month long ETP will be selling front month futures and buying 2nd month futures every day, for example. Front month futures are usually priced lower than 2nd month futures, which means that the ETP will be losing value as it makes these trades (it’s selling the cheap futures and buying the expensive ones), and vice versa for short ETPs. This amounts to a transfer of wealth from hedgers to speculators willing to take opposite positions; this transfer can be predicted and exploited.
I’ll be looking at two such VIX futures-based instruments in this post: VIXY (which is long the futures), and XIV (which is short the futures). As you can see in the chart below, while the VIX is nearly unchanged over the period, VIXY has lost over 80% of its value. At the same time, XIV is up 50% (though it did suffer a gigantic drawdown).
There are many different approaches to trading these ETPs (for example Mike Brill uses realized VIX volatility to time his trades). The returns are driven by the complex relationships between the value of the index, the value of the index in relation to its moving average, the value of the futures in relation to the index, and the value of various future maturities in relation to each other. These relationships give rise to many strategies, and I’m going to present two of them below.
I’ll be using different approaches for the long and short sides of the trades. Short based on the ratio between the front and 2nd month contract, and long using the basis. Here are the rules:
Go long XIV at close (“short”) when:
- 2nd month contract is between 5% and 20% higher than the front month contract.
Go long VIXY at close (“long”) when:
- Front month future is at least 2.5% below the index.
Results
First let’s have a look at how these strategies perform without the hedge. Using data from January 2011 to November 2012, here are the daily return stats for these two approaches individually and when combined:
Equity curves & drawdowns:
The biggest issues with VIX ETN strategies are large drawdowns, and large sudden losses when the VIX spikes up (and to a lesser extent when it spikes down; these tend to be less violent though). A spike in implied volatility is almost always caused by large movements in the underlying index, in this case the S&P 500. We can use this relationship in our favor by utilizing SPY as a hedge.
- When long XIV, short SPY in an equal dollar amount.
- When long VIXY, go long SPY in an equal dollar amount.
The stats:
And the equity curves & drawdowns:
The results are quite good. The bad news is that we have to give up about 40% of CAGR. On a risk-adjusted basis, however, returns are significantly improved.
- CAGR / St. Dev. goes from 38.7 to 45.9.
- CAGR / Max Drawdown goes from 4.5 to 4.9.
All risk measures show significant improvement:
- The worst day goes from a painful -12% to a manageable -9%.
- Maximum drawdown goes from -36.5% to -25.7%.
- Daily standard deviation goes from 4.28% to 2.76%.
Of course, just because risk-adjusted returns are improved does not mean it’s necessarily a good idea. Holding SPY results in both direct costs (commissions, slippage, shorting costs) as well as the opportunity cost of the capital’s next-best use. The improvement may not be enough to justify taking away capital from another system, for example.
Another possibility would be to implement this idea using options, which have the benefit of requiring a small outlay. Especially when holding XIV, SPY puts could be a good idea as both implied volatility and price would move in our direction when the VIX spikes up. However, this must be weighted against theta burn. I don’t have access to a dataset for SPY options to test this out, unfortunately (anyone know where I can get EOD options data that is not absurdly expensive?).
If you want to play around with the data yourself, you can download the spreadsheet here.
Footnotes- If you live in the U.S. there can be important differences in tax treatment depending on which one you trade, so do your research.[↩]
Christian Rolfsson says:
I have been thinking about testing a similar system with options but do not have historical options data. I was thinking that I could get results in the same ballpark by using Black Scholes to calculate a theoretical options price by using Futures price, VIX maybe combined with HV ( should have relatively high correlation with IV and used as a substitute for IV). Then simulate the system using the theoretical options price. This will of course not give the correct results but could validate an options system and give interesting insights of system sensitivity, robustness etc. not available anywhere else. Do you think this could work?
Qusma says:
That might be possible, I’m not sure how closely VIX would track implied volatility of the actual option contract that you’d use though. In any case I’ve gotten a couple of good suggestions regarding options data so I’ll be revisiting the issue “properly” in the future.
saltmine says:
Glad to have found this blog and thanks for sharing your research! I’ve traded different variants of this strategy for the last 2 years. The challenge for me has not only been finding an effective hedge with a high correlation, but a matter of scale as well.
I modeled the use of SPY as a hedge to the initial “short” strategy you outlined and discovered results similar to what you’ve posted here. I also modeled drawdowns over 4 specific time periods since VXX was created in 2009 and found the effectiveness of the hedge during those periods to be even less than the overall historical backtest would indicate. For the “short” strategy you tabulated, the maximum drawdown was only mitigated by 20% on a dollar for dollar basis, from -33.31% to -26.65%, which was similar to what I found.
I then modeled the idea you suggested at the end of your post, which was to use SPY options as an alternative method to hedge. I found historical options data at, appropriately enough, http://www.historicaloptiondata.com/
Several variables had to be taken into account. I modeled purchasing SPY put options 3, 6, 9 and 12 months out, with strikes ATM, 10% OTM and 20% OTM. I compiled the prices surrounding two drawdown periods, specifically July 2011 and May 2012, with underwhelming results. The low deltas and high extrinsic value of options more than 3 months out renders the effectiveness of the hedge no higher than using the underlying SPY. Rolling the hedge position every 3 months or less creates exceedingly high costs and the prices of OTM options 3 months out or less creates high slippage costs and raises liquidity issues. I’m putting on a fairly large position in VXX and XIV for a retail trader. What’s the market impact of trying to acquire a couple thousand SPY contracts on the open market? I’m not an options trader so I don’t know the answer to that. ATM, I think I would be fine, but the liquidity drops off rapidly on short dated options for 10% and 20% OTM strikes. The implementation would have to be spread out and ended up creating enough ambiguity in my mind combined with the low degree of effectiveness that I decided it was not viable for me.
Post earnings announcement drift (PEAD) is something I’ve daytraded consistently over several years. I’m currently in the process of correlating the returns in the databases I’ve compiled with the VIX ETP strategies we’ve discussed here. I look forward to comparing results here as well as investigating some of the compelling mean reversion ideas you’ve introduced in other posts.
Regards,
saltmine
qusma says:
What a fantastic comment saltmine, thanks a lot!
Interesting to hear about the options results, I was worried that trading cost issues would turn out to be too much, compared to the effectiveness of the hedge. There’s plenty of liquidity in the SPY options, but it tends to be centered at ATM options and can fall off pretty hard the further away you get…could be a very significant problem when it comes to rebalancing/rolling the position.
As for the PEAD I think it’s very interesting, but haven’t had a look at it beyond the standard academic papers…doing some tests myself is on my “todo” list, but pretty far down…my impression is that there isn’t a whole lot of alpha there, at least for multi-day swing approaches. You mention that you daytrade the PEAD, are always flat at the end of the day when doing so?
saltmine says:
My previous comment is confusing…I’ve traded earnings releases on the open of the next session and closed out after 3:00 and/or MOC the same day for about 10 years. Diminishing returns and capacity issues led me to investigate pre and post announcement returns a few years ago. You are correct in that the average daily return for swing approaches is not very impressive at first glance. However, the correlation is powerful nonetheless with much more capacity/liquidity than same day trades.
I’ve compiled a running database of all earnings releases on briefing.com since 2007. I always try to trade on the opening cross whenever possible due to slippage and easier balancing of long/short allocations. The cumulative 4 day return, open to open, for all stocks on NYSE >$5, >100K average volume is +0.3% for the 4 days prior and -0.4% for the 4 days after the release. Nasdaq returns are very similar. That’s about 4000 data points both long and short on NYSE and another 2000 or so for Nasdaq. Sorting by parameters queried from Yahoo the day of the release augments those returns significantly. 1% a week compounds to 64% annualized. Square that if you use 2x leverage and you get a pretty big number. Not that I’ve realized anything close to that, but you can see where the motivation came from.
The problem, as it relates to hedging the VIX strategies outlined here, is the highly nonlinear flow of the volume of earnings releases every quarter. There is an infinite amount of liquidity, for my intents and purposes, about 4 weeks a quarter, trailing off over the next 4 weeks and the final 3-4 weeks render very few available trades. During peak earnings season, I could effectively hedge as large a position in VIX ETPs as I would be willing to put on, but the availability to hedge with earnings trades would dissolve about halfway through the quarter unless I was willing to enter appalingly large position sizes. Even then, the last 4 weeks of each quarter simply do not have nearly enough trade candidates. The drawdown in XIV over the last 2 weeks of 2012 provided a perfect example of that. SPY barely moved so no permutaions of market ETFs or derivatives would’ve worked there either.
Hope your trading in 2013 is off to a great start! Comments would be greatly appreciated. Regards,
– saltmine
@ContangoMojo says:
Thanks for this post. I am a small player but have been working out some trading strategies just as you described. I am looking at the futures term structure and the futures premium over the VIX index just as you described. Certainly “short” vix trades have done well in the last couple of years. Any system that avoided the massive spikes around August 2011 would have been a great winner. Even a simple moving average on VXX would be a good tell for when to go to cash.
As a small retail player, I find that “long” volatility trades are really tough to put on. If your timing is off, you can miss the spike and get whipped out pretty quick. I would rather sit in cash and wait for the next “short” entry.
Would UVXY or SVXY options be a viable tool for the hedge? Are they too costly relative to SPX options? They might be more effective since SPY and VIX sometimes go in the same direction.
qusma says:
I definitely agree that the “long” trades are tougher to time. The “long” moves are generally fewer and rarer as well. On the other hand they can be very explosive and result in great returns in very little time if you manage to capture the move early on. I find the 1st month future : spot VIX ratio to be extremely useful for that purpose.
I don’t think UVXY or SVXY options would be viable compared to SPY ones. The issue of SPY and VIX moving in the same direction is generally not problematic. It does happen, but what we’re really trying to avoid with the hedge are the catastrophic days. And you’ll never see big movements in the same direction.
shahyad says:
very good. thanks. کرکره برقی