Need more proof that fading an overbought VIX works? Meet VXV
In our never-ending quest to bring you ways to analyze the CBOE Volatility Index (VIX), I bring you … the CBOE 3-Month Volatility Index (VXV)! It's VIX, only longer. While VIX proxies the implied volatility of a 30-day S&P 500 Index (SPX) option, VXV proxies implied vol on a 90-day option. As such, it moves less, as the longer you go out in time, the more mean reversion assumptions set in.
It moves almost in lockstep with VIX, only at a lesser magnitude. Here are the two pups together in 2015, for instance:
Not much there to the naked eye. But alas, the ratio between the two varies a bit. VIX averages about 0.93 of VXV. It does have its variation, though -- especially when VIX pops. VXV is much slower to assume the volatility pop will persist, identical to the way the VIX futures term structure acts.
My question is: Can we find much in the way of predictive value as to future market returns? On the surface, there's not much. The correlation between VIX/VXV ratio and one-month-forward SPDR S&P 500 ETF Trust (SPY) returns is an uninspiring 0.0324. The scatter plot is one big blob.
But alas, there's some signal in the extremes. The data only goes back to mid-2006, but that encompasses pretty much every sort of market backdrop. The VIX/VXV ratio closed at 1.10 or greater 113 times, or about 5% of all days. The average one-month forward return on those days is 0.22%; pretty lousy compared to a randomly timed one-month return of 0.75%, back to mid-2006. But the median tells a different story. The median return of owning SPY for one month forward when the ratio hits is 2.69% vs. a median of 1.49% for randomly timed one-month holds.
What we have here is a case of the data overlaps intersecting with 2008. What if instead, we do like we do with VIX vs. its simple moving average (SMA). That is, separate it into "incidents," so we're not reusing the same data. Here's the system: We buy SPY if the VIX/VXV ratio is 1.10 and hold for one month.
It looks pretty good. There are 25 "signals" over nine plus years, and the average return is 1.94%, median is 2.90%. The trade won 17 times. There was one huge draw-down of 16.37% in 2008.
It's a similar principle to the VIX 20% over the 10-day SMA, with moderately better results. And it reinforces the notion that fading an overbought VIX tends to work.
Disclaimer: Mr. Warner's opinions expressed above do not necessarily represent the views of Schaeffer's Investment Research.