Tuesday, July 09, 2019

Looking at the Market Through Different Lenses - 3: Volatility

When I model the market, I typically include three variables:  1) a measure of short-term performance; 2) a measure of longer-term performance; and 3) volatility.  I think of volatility as a regime measure, so that--for a given vol regime--I want to see what happens when the market has been short-term and longer-term overbought or oversold.  The most common measure for volatility is VIX, which captures the movement anticipated in the options market (implied volatility).  Another measure would be "realized" volatility, which I measure by a moving average of the average daily true range.

I find value, however, in tracking unique volatility measures.  One, for instance, looks at volume bars in the ES futures and tracks *their* realized volatility.  That measure, which I dub "pure volatility", represents the amount of movement we get per unit of trading volume.  That has been quite helpful as a regime measure, but also helps in terms of sizing positions, as it captures the degree of movement one can expect for a given trading volume.

Yet another measure was one that I created in 2012, a regression equation designed to predict the VIX from two variables:  the net directional movement over a several week period and the realized volatility of the daily bars during that period.  These two variables were very significantly predictive of VIX.  When I looked at the residuals, however, I noticed additional value.  The "excess" VIX--the amount that VIX exceeds or falls short of what it "should" be given the regression formula--tells something about the psychology of the market.  It is the degree to which options participants are anticipating unusually high or low movement.

The excess VIX is plotted against SPY (blue line) above.  Note that spikes in excess VIX have been associated with intermediate-term market bottoms.  When I performed a quartile split on the daily data, an interesting pattern emerged.  Returns have been significantly better than average when excess VIX has been in its lowest and highest quartiles.  Specifically, over the next 20 trading sessions, SPY has averaged a gain of almost +1.2% when excess VIX has been in the extreme quartiles and only +.23% in the middle quartiles.  

What that suggests is that there is a momentum effect when options players price in very low volatility relative to what would be expected and a value effect when they price in too much volatility.  Much of the market's uptrend over this period came from occasions when VIX was priced "too low" or "too high" relative to its norm.  This is an interesting finding, given that the results are entirely out of sample.

This is a good example of the value of assessing the psychology of the market, not just our own psychology.  It's also a good example of the value of going beyond traditional measures to create unique indicators that capture what others miss.  (Note:  the recent Excess VIX has been in its third quartile, where VIX is relatively fairly priced.  There has been less upside edge in the market over the next 20 trading sessions at such times.)

Further Reading: