The authors based their volatility returns primarily on two instruments: variance swaps and options. Their data series has an earliest start date of 1995, with starting dates varying depending on data availability.
To create series that were comparable across assets as well as time, the authors scaled each of the 34 volatility return series to target an annualized volatility of 1% each month at trade inception. The scaling was ex ante, using historical realized volatility calculated monthly with an expanding window and a minimum of 36 observations.
To facilitate comparison across asset classes, they then combined the scaled volatility return series on an equal-weighted basis by asset within each asset class. They chose an equal-weighting scheme because of its simplicity and transparency.
This approach resulted in four composite volatility return series—for equities, fixed income, currencies and commodities—each of which they then scaled to target 1% risk on an ex-ante basis using the methodology described previously.
Finally, to facilitate the statistical and economic evaluation of volatility as an asset class, they combined the four asset class return composites on an equal-weighted basis into a single composite scaled to target 1% annualized risk, again using the methodology previously described. They called this series the GVCP.
The Results
Following is a summary of their findings:
- Negative (short) volatility premiums are widespread, statistically significant and economically meaningful. There was a consistently positive mean for the spread between implied and realized volatility in all asset classes and components.
- Selling volatility is profitable in virtually all markets nearly all the time, including the five-year period surrounding September 2008, with a consistently positive mean for volatility returns (but with fat left tails).
- Adding the GVCP in small amounts to typical institutional portfolios would have enhanced substantially long-term returns (increasing the combined Sharpe ratio by as much as 12% in the authors’ sample) but at the cost of increased short-term tail risk.
- All 34 means were positive. The annual mean return to the 11 stock components ranged from 2.8% to 4.5%, and the annual standard deviation ranged from 5.4% to 7.9%. For the four bond components, the annual mean return ranged from 1.8% to 4.0%, and the annual standard deviation ranged from 4.6% to 14.1%. For the nine currencies, the annual mean return ranged from 0.6% to 1.3%, with the annual standard deviation ranging from 2.1% to 4.5%. For the 10 commodities, the annual mean return ranged from 0.7% to 3.5%, with the annual standard deviation ranging from 3.4% to 13.4%.
