We develop a joint model for the physical and risk neutral variance implying variance risk premia that are reasonably smooth, appropriately reacting to changes in level and variability of the variance and naturally satisfying the sign constraint. The model, straightforward to estimate using option market data and high frequency returns, allows identifying agents’ sentiment indicators. Our results show that excess returns are to a large extent explained by fear or optimism towards future extreme variance events and only marginally, if at all, by the premium associated with normal price fluctuations. We also find evidence of asymmetry in the response of future excess returns to fear and optimism with respect to the state of the market at the time expectations are formed. Robustness checks show that the documented predictive power is distinct from that contained in several well-known economic predictors and that the general findings extend internationally to eight major markets.
