There is a mathematical term used to describe a time series' propensity to mean-revert or not. Autocorrelation measures the tendency for today's price direction to be in the same direction as yesterday's. In a period of negative autocorrelation (such as today) when the market sells off one day it is much more likely to rebound the next. As Artemis Capital's Chris Cole notes, the current level of negative auto-correlation (often associated with positive for 'buy-the-dip' strategies in an upward trending market) has never been higher. Mean reversion and negative autocorrelations are one reason why many pure 'portfolio insurance' strategies are struggling with losses. If you are constantly shorting volatility this trend toward powerful mean reversion is your best friend. However, empirically, this high mean reversion is unsustainable.
Via Artemis Capital,
Today, mean reversion of daily volatility returns is at the highest average levels in the history of modern derivatives markets. Autocorrelations of daily returns measure the propensity for today’s price direction to be in the same direction as tomorrows. For example in a period of negative autocorrelation (such as today) when the market sells off one day it is much more likely to rebound the next. An environment of negative autocorrelation is often associated with positive performance for “buy the dip” strategies in an upwardly trending market (correlations measure direction but not trend).
In April we experienced some of the highest levels of daily mean reversion for volatility in history. The chart at the top shows the negative autocorrelations of 3-month rolling VIX index returns (please note y-axis has been reversed to show rising levels of mean reversion or negative autocorrelation). The second chart tracks mean reversion regimes measured over a longer 12-month lagged period demonstrating the most extreme autocorrelations since just prior to the crash in 2008 (but not the highest ever). Mean reversion and negative autocorrelations are one reason why many pure ‘portfolio insurance’ strategies are struggling with losses. If you are constantly shorting volatility this trend toward powerful mean reversion is your best friend.
High Mean Reversion is Unsustainable. Historically when autocorrelations in daily volatility drop below -0.30 the VIX index has closed higher over the next month 76% of the time with an average gain of 19.11% (still putting us at a mild 16 level on the VIX from today). The last peak in mean reversion occurred in early 2008 just prior to the onset of the financial crisis. This is not meant to imply that I expect a great crash but I do think it gives a powerful technical signal that we are long overdue for some kind of meaningful pullback in equities (even if that is just 5-10%).
The phenomenon of increasing autocorrelations is often seen in stock market crashes whereby one bad day is likely to be followed by another bad day. We saw rising autocorrelations during periods of financial contagion including 1987 Black Monday, the 1998 Russian Default and LTCM crash, and the 2008 meltdown.
The potential for mean reversion regimes to ‘shift’ is driven by increasing leverage and interconnectedness in the system. It’s hard to predict when these shifts occur! When they do periods of high autocorrelations can be very damaging for equity portfolios but good for the Artemis Vega Fund LP.
In this environment fading fear and buying equity on every dip has made money. This cannot continue forever and presents a powerful asymmetric opportunity. At the same time it is dangerous for any long volatility player to underestimate the extent to which these cycles will persist. Eventually the rebuild of leverage in the system will provide an opportunity when the mean reversion dynamic breaks.
The fact that NYSE Margin Debt has been moving in lock-step with negative autocorrelations and volatility is not a coincidence.