On 17-Aug-11, iPath issued a new ETF called S&P 500 Dynamic VIX ETN (XVZ) to help avoid the problems associated with rolling costs in volatility ETFs while maintaining a long exposure to volatility. This ETF maintains a varying long exposure to VXZ and either goes long or short VXX depending on the ratio of the VIX Index (30-day implied volatility of S&P 500) to the VXV Index (93-day implied volatility of S&P 500). XVZ is structured to increase during overall increases in volatility depending on the slope of the implied volatility curve while minimizing the holding costs of maintaining such a position. By actively managing between VXX and VXZ, this ETF would have returned 306% (as of 10-Aug-11) since 20-Dec-05 had it been in existence the entire time. A $1,000 investment would be worth $4,062.77.
Although this fund looks great, the ratio of VIX to VXV doesn’t actually reflect the rolling cost of VXX, because the futures’ term structure can differ from the shape of the implied volatility curve. VXX is based on the current and next month’s VIX future contract and VXZ is based on a portfolio of four to seven months VIX futures contracts. However, historically when VIX has been below XVX, VIX has also been below its mean-reverting level, which has historically resulted in the long investor paying a premium to the short investor for rolling future contracts (as shown in the table below). Although this has on average been true, the VIX futures’ curve can theoretically take any shape. If the second month’s future contract is lower than current month’s contract while VIX is less than XVX, an investor would be short VXX under the XVZ strategy when the rolling costs are discounted to a long investor (working against the short investor).
Click to enlarge
- Rolling costs are greater than zero (at a premium) if the next month future price is greater than the current month’s price (negative yield to a long investor rolling futures contracts)
- Rolling costs are less than zero (at a discount) if the next month future price is less than the current month’s price (positive yield to a long investor rolling futures contracts)
A better alternative would be to base the long or short amount of the VXX on the rolling cost premium or discount. As shown in the table below, since the rolling costs are sticky (premiums are typically preceded by premiums and discounts are preceded by discounts), investors can capture the premium (short VXX) or discount (long VXX) by continually rolling VIX future contracts.
The following table shows the number of months that the average rolling costs of VIX futures fell in certain ranges since 1-Jan-06. Additionally, the table shows the percentage of months where the average rolling costs were in the same range as the previous month.
Additionally, I ran an autoregression using the previous day’s premium/discount as the lagged variable. The output showed that 93% of the current day’s premium/discount is explained by the level of the prior day’s premium/discount. A description of my regression is as follows:
- N = 1460
- R-squared = .87
- t-stat = 98.97, suggesting statistical significance
Since I know that rolling costs tend to be persistent and can be substantial, it makes sense to try to capture them. To find the most efficient allocation strategy for capturing the rolling costs, I’ve conducted a Monte Carlo simulation based on a million iterations. Based on the simulation’s output (discussed in further detail later), I found a better allocation strategy by more aggressively exploiting the short-term volatility futures’ term structure and replacing VXZ with an exposure to long-term Treasuries (IEF). My strategy would have returned 2,216% (as of 10-Aug-11) since 20-Dec-05 had it been in existence. A $1,000 investment would be worth $23,159.49.
Rationale for IEF vs. VXZ - During recent spikes in volatility, long-term Treasuries have typically rallied as investors desired safe-haven assets. Therefore, an investment in long-term Treasuries can produce similar return directions (not magnitudes) during large volatility spikes as an investment in VXZ. Additionally, IEF provides current income while VXZ does not. Since my strategy more aggressively exploits the volatility term structure than XVZ, I wanted more stability in the strategy while maintaining similar directional exposures during volatility spikes.
Portfolio Weights – To find the portfolio weights that would maximize the portfolio’s ending value, I ran a Monte Carlo simulation at one million iterations by generating randomized weights at each level of the rolling cost premiums and discounts. From the output that the simulation produced, I analyzed the portfolio weights of the top 50 portfolios with the highest ending values. A consistent portfolio weighting scheme emerged which I have used to help create my strategy’s portfolio weights as shown in the table below. Note, as displayed below, my strategy more aggressively exploits the volatility term structure when rolling costs are at a significant premium or discount.
The table below shows my strategy’s portfolio allocation strategy at various ranges of premiums and discounts.
The table below shows the allocation strategy of XVZ, which less aggressively allocates to VXX during extremes in premiums and discounts.
Detailed Performance – As displayed below, my strategy would have outperformed XVZ by 5.7x over the same period.
The table below shows my strategy’s performance since 20-Dec-05 compared to the performance of XVZ through 10-Aug-11.
The table below shows my strategy’s quarterly performance since 1-Jan-06 (first full available quarter) as well as the annualized standard deviation and return (2011 performance is through 30-Sep-11).
Investors could actively manage this strategy with currently available volatility ETFs. However, the transaction costs might make this prohibitive if the costs comprise too much of the portfolio’s total worth. Ideally, one of the ETF companies will decide to issue my proposed ETF so that every investor for which this strategy is suitable for can participate. I would name the strategy “Short-Term Volatility Futures’ Term Structure Capture Strategy” and go by the ticker TCXX for Term Capture of VXX.
Disclaimer: Please consult your financial advisor before making investment decisions. Depending on your circumstances and risk tolerance, the strategy in this article may not be suitable for all investors.
Additional disclosure: I am also long XIV and XXV.