With all the talk of the current low volatility environment and a large portion of people implying mean reversion is imminent, we thought we would weigh in further on the topic. Volatility is both positively serially correlated and mean-reverting. That might seem counterintuitive, but the way we look at it is that volatility is mean-reverting within regimes and serially correlated between regimes. Even more confused? Stay with us as we dive into more detail with the hope of providing some clarification on a potentially difficult topic to grasp.
Data disclaimer: Monthly VIX Index data from 1/1/1990 to 12/31/2017 (Got that out of the way… Now to the good stuff.)
The serial correlation on end of month VIX levels is 84%, meaning that the current VIX level is highly predictive of next month’s VIX level. On the other hand, the VIX flipped directions about 59% of the time; meaning if the VIX was up last month compared to 2 months ago then it was more likely to be lower this month, this is the mean-reversion aspect of volatility.
Table 1: Current VIX vs. Next Month VIX
|Lower Percentile||Upper Percentile||Lower VIX||Upper VIX||# of Oberservations||Average VIX||Average VIX
Table 1 shows different VIX “regimes” and how the VIX is fairly stable within regimes. The first row shows that the bottom 10% of end of month VIX levels during the time period (34 observations) ranged from 9.51 to 11.98 and the average level within this range was 11.24. (Here’s the kicker…) The average VIX level next month, when the current month’s VIX is within this range, was 12.41. As you can see, the average VIX level next month is very similar to the average VIX level for all 6 regimes.
Figure 1: Mean-reverting VIX from 3/31/2016 to 3/31/2017
From March 2016 to March 2017 VIX ranged from 12 to 17 but every month it flipped directions. If VIX was down last month compared to 2 months ago then it was up this month. This is a good example of the reverting nature of VIX within regimes. (We wouldn’t have shown it to you if it didn’t prove our point. On a related, and very important, note… beware of the publishing bias.)
In (our short and unscientific) conclusion, we consider VIX to be both positively serially correlated and mean-reverting. So if we are in a low volatility environment our expectation is to remain in a low volatility environment (at least for next month). And if we could predict black swans (by definition no one can) and the corresponding spike in volatility then this would be a completely different blog.
(If volatility does happen to significantly rise after publishing this blog, we take no responsibility.)
Read more about A (Mostly) Overlooked Cause of Low Volatility >