Predicting volatility sounds dubious, but when you break down the numbers, it might not be so complicated
So, if you were vacationing on Mars for the last year or so, and just came back and flipped on the TV, you'd probably have a few thoughts: There are a lot of "-gates" associated with the Patriots, is one; DraftKings must have about a $4 billion ad budget, is another; and most importantly, increased volatility is here to stay!
Yes, that's pretty much the consensus these days. But, as we mention periodically, the market ALWAYS expects higher volatility, or at least folks that comment on the market always expect higher volatility. That's especially true when the CBOE Volatility Index (VIX) is low, but it persists even in times of higher VIX, like now. It's just that instead of predicting a VIX rise, it's more like predicting VIX-stasis. It inspired me to have a little fun with VIX numbers.
The median VIX going back to 1990 is exactly 18. We sometimes hear a number of "20," but that's really the average VIX. Median is a better descriptor, in my humble opinion, seeing as how VIX has lifted exponentially at times, and those outliers skew the average a bit. Median, in this case, equals "typical" -- and that's a better tell. There are 6,475 trading days in the data set, so it's quite expansive.
VIX at 24.50 doesn't sound all that exceptional, but in fact it's in the top 21% of all closes; thus, it's actually on the high end. All those recent calls that volatility is "here to stay" sound silly on the surface, but in reality, it's an intelligent call. In fact, if someone ever asks you, "Where is the VIX going to be a month from now?", the best answer is "right here." And it almost doesn't matter where "here" is. There's a 0.982 correlation between VIX now and VIX one month from now.
That makes sense, of course. It's not some sort of random number fixing; rather, it's a continuous process. It takes time for volatility assumptions to move.
OK, fine, so let's give it more time. How about three months from now?
Well, there's still a 0.968 correlation. So, if you answer "right here" to the "Where's VIX three months from now?" question, good chance you'll do well again.
All of that begs an obvious question: Why does the futures term structure almost always slope upward? That is, why does it always expect VIX to lift? Well, that's not so much the case now; we're reasonably close to parity and the curve is relatively flat. But, in all those years when VIX was "low," why did the curve behave that way?
The answer is: It shouldn't have. Let's say we look at only the bottom 50% of VIX closes -- all the ones under 18. They have a median of 14.07. And the median VIX close one month later? It's still 14.07. Three months out, that median jumps all the way to ... 14.09.
In English, low volatility predicts more low volatility. Playing for a VIX pop at some indeterminate time in the future doesn't make enormous sense. The same is true in the "high" VIX part of the sample. Median "high" VIX is 23.3, one month out from that has a median of 23.25, and three months out has a median of 23.27.
You have to really start moving to extremes to see VIX out in time deviate from VIX now. And in those cases, it's just a tendency toward the overall VIX median. Really low VIX predicts modestly higher VIX, really high VIX predicts modestly lower VIX.
My conclusions? Don't ever buy a future at a sizable premium. And redefine the "mean" in mean reversion. So long as VIX isn't at an extreme level, "right here, right now" VIX is as good an estimate for "mean" as anything.
Disclaimer: Mr. Warner's opinions expressed above do not necessarily represent the views of Schaeffer's Investment Research