Feunou, Bruno Fontaine, Jean-Sébastien Tedongap, Roméo
Year of Publication:
Bank of Canada Working Paper 2009,20
We introduce the Homoscedastic Gamma [HG] model where the distribution of returns is characterized by its mean, variance and an independent skewness parameter under both measures. The model predicts that the spread between historical and risk-neutral volatilities is a function of the risk premium and of skewness. In fact, the equity premium is twice the ratio of the volatility spread to skewness. We measure skewness from option prices and test these predictions. We find that conditioning on skewness increases the predictive power of the volatility spread and that coefficient estimates accord with theory. In short, the data do not reject the model's implications for the equity premium. We also check the model's implications for option pricing and show that the information content of skewness leads to improved in-sample and out-of-sample pricing performances as well as improved hedging performances. Our results imply that expanding around the Gaussian density is restrictive and does not offer sufficient flexibility to match the skewness and kurtosis implicit in option data. Finally, we document the term structure of option-implied volatility, skewness and kurtosis and find that time-dependence in returns has a greater impact on skewness.