Question: When a market is stressed, do you expect its subsequent behavior to be more volatile than normal, or less volatile, or the same? And is your return expectation higher, lower, or the same?
That’s a question that Mike Sylvanus, Mike Thomas and I looked at a couple of years ago in a paper on the subject of volatility-responsive asset allocation. We found that volatility in the equity market tended to persist but was not associated with higher returns. In other words, in times of stress, volatility is sticky, but there doesn’t seem to be any persistent return premium attached to that volatility. This led us to conclude that there was a case for varying asset allocation in response to market volatility, an idea which has been incorporated into some volatility management strategies.
In a working paper titled The Origin of Fat Tails, Martin Gremm of Pivot Point Advisors has looked in more depth at the same question. Looking separately at the equity and bond markets, and using the implied volatility in options pricing as a proxy for market stress, he too found that market stress has tended to bring higher volatility, but not higher returns. These results are shown below.
Stress-ordered sample estimates¹
VXO is a measure of the equity market volatility that is implied by options prices. The downward slope in the chart “Stock Mean” shows that the average return on stocks tends to be no higher, and perhaps even slightly lower, during times of unusually high market volatility or stress. The chart “Stock StDev” shows, however, that when implied volatility is high, actual volatility does tend to turn out higher. On both of those charts we observe that once implied volatility rises above 70, behavior can become extreme. On that point, we should note that there are not many data points at those levels, and the data points that do exist represent highly unusual market conditions. It probably does not make sense to look for “normal” patterns of what to expect when implied volatility is that extreme.
There’s a similar pattern in the bond market results (MOVE being a measure of implied bond market volatility), although with perhaps less evidence of a negative link to expected returns in this case.
The focus of Martin’s paper is on how these relationships can be used to build better models of market behavior, and readers who would like to look more closely at that angle should refer to the paper. But as regards the question of a dynamic response to the market environment, these results seem to offer further evidence that a strategic asset allocation that is conditioned on market volatility may be worthwhile.