The titular claim of a recent Bloomberg article, “VIX Fails to Forecast S&P 500 Drop,” only makes sense in a world where nothing occurs beyond one standard deviation of predicted outcomes. The fact that the October decline was more than one standard deviation from the mean as predicted by the VIX is just another way of saying that the October decline was the sort of thing that happens less than 33% of the time; which is an unremarkable point to be making. And yet:
The breakdown was unprecedented for the VIX, which had indicated ranges for the S&P 500 that were off by more than 10 percent on just five occasions.
The authors seem to assume that the VIX predicts the actual SPX trading range, full stop, rather than one standard deviation within that trading range. What would be a newsworthy event would be if SPX traded beyond one standard deviation as implied by its options over, say, 45% of the time over a sufficiently long period.
Just to review, the VIX measures the annualized expectation of volatility in the S&P 500 over the next 30 days, where “expectation of volatility” refers to outcomes one standard deviation from the mean. For example, at the time of writing the VIX is at 42.84. That means that prices of SPX options imply about a 68% chance that the index will trade anywhere between 767 and 984 over the next 30-day period. In other words, there’s about a 33% chance that SPX will trade somewhere lower than 767 or higher than 984 over the next month. If SPX collapses to 700 the day after Christmas, will the VIX have failed in some collosal way? Hardly: that fall will have been a statistically less likely occurrence, but not at all excluded from the range of outcomes.
So the only failure here is among any traders who think that 68=99, i.e., that an index that estimates outcomes one standard deviation from the mean is actually estimating all outcomes. By comparison, three standard deviations (99% of outcomes) for SPX as of this date would put the index anywhere between 630 and 1200. With SPX at 500, maybe it would be approprite to talk about options traders “failing” to estimate outcomes properly, although we would still point to the 1% of outcomes that the estimate accounted for. If you want to argue that the tails at the edges of any distribution curve are fatter in reality than they are in the models, we’ll agree heartily. But that’s not the claim being made here.
One additional problem is the tone of the article. The narrative structure is basically a “he-said, she-said” of assertions, with little authorial corralling of those opinions. And isn’t the imposition of some fact-based objectivity precisely what distinguishes journalism from stenography? Granted, the use of VIX as a trading tool is a complicated issue. But here are the positions taken by sources quoted in the article:
- Stu Rosenthal: the VIX’s predictive power has completely broken down
- Ben Londergan: the VIX remains one of the best guides for investors
- Ryan Caldwell: you would have gotten killed if you used it
- Peter Sorrentino: using a VIX signal to go all in would’ve been a career ender
- Fritz Meyer: the VIX is just a bunch of guys in the options pit guessing
- Bernie Madoff: the VIX didn’t give me a pony
Ok, that last one is a fake, but it’s not so far off from the others. It would’ve been helpful if the article’s authors had pointed out which of these remarks deserves any credence; #1 is false, #2 is true, #3 is trivially true, #4 is tautologous, and #5 is false. The closing quotes in the article are particularly risible:
“If you looked at the VIX spike and then volatility coming back down, back then I thought ‘Ring the bell, maybe the worst is over,’” said Peter Sorrentino, who helps manage $16 billion at Huntington Asset Advisors Inc. in Cincinnati, including $127 million in options. “Not only did the VIX hit a new high, but the market also hit a new low. If you had used that as a signal to go all in, that was almost a career ender.”
For Fritz Meyer at Invesco Aim Advisors Inc., the biggest problem with relying on the VIX as a forecasting tool is that the index is still a product of predictions about the future.
“It’s just a bunch of guys in the options pit guessing,” said Meyer, the Denver-based senior market strategist at Invesco Aim, which oversees about $358 billion. “I don’t believe it.”
The VIX didn’t hit a new high; if you include the “old” VIX data (now listed under VXO), 1987 stands out as the high. As for using the VIX as a standalone trading signal, in our ongoing VIX coverage, we’ve commented many times before that the VIX is just one indicator among many for analyzing volatility. We humbly submit that using any indicator in existence as a “signal to go all in” will be a career-ending decision sooner or later.
The remark about the VIX being “a bunch of guys in the options pit guessing” is surely a joke that didn’t come out quite right. We’re not fans of the efficient market hypothesis in any strong sense, and we’re huge fans of not making ill-conceived trades on the basis of some half-baked guess about VIX action. But that doesn’t mean that volatility estimates themselves are mistaken; if anything, the academic literature on volatility arbitrage shows that trading the spread between realized and implied volatility has only gotten more difficult over the past decade as options volume increases.