In our previous post: Where Will This VIX Spike Top Out? we showed that, by one measure, this month’s VIX spike was 9th in the entire history of the VIX in terms of how quickly it rose from recent levels and then bucked mean-reversion and continued even higher. Specifically, we used Adam Warner’s criterion for calling the start of a spike as being when the VIX closed at least 20% above its 10-day average, and then measured how far the spike continued beyond that point.
Now our current spike has accomplished another historically significant feat: on Tuesday it closed more than 20% below its 10-day average. That’s only the 13th time that’s happened since 1986 (1).
To illustrate what we’re talking about here, the graph below shows the VIX index since 03/2004, when VIX futures began trading, with “plunges” of at least 15% below the 10-day average marked in orange (2). Note that I reduced the threshold to 15% to give us more observations to consider.
As you would expect, all of the “orange circles” occur after a sharp decline from a near-term top.
The important question is what tends to happen immediately after one of these plunges? To answer that, the following table shows the average change in the VIX in the days and week following one of our “orange circles”.
Historically, the day immediately following a plunge has been mixed, but in 72% of cases, the VIX has been higher a week out, with an average change of +6.2%.
The usual warnings about small sample size definitely apply here, and these stats are in no way conclusive on their own, but they do point to the tendency for VIX mean-reversion to cut both ways.
As we showed in our last post, when the VIX spikes too far too fast, there’s a strong tendency for the VIX to fall, and conversely, when the VIX plunges too far too fast, there’s an opposing (though not quite as strong) tendency for it to rise.
Click to see Volatility Made Simple’s own elegant solution to the VIX ETP puzzle.
Volatility Made Simple
- For dates prior to 1990, I used the VXO index in place of the VIX. It’s an imperfect comparison, but close enough I think for the purposes of this illustration (and it’s useful to be able to capture the 1987 crash).
- To control for overlapping VIX plunges, I required that after the VIX fell at least 15% below its 10-day average, it had to then rise back above its 10-day average before it could then register a new plunge.