Robust-H-infinity Forecasting and Asset Pricing Anomalies

We present an alternative expectation formation mechanism that helps rationalize well known asset pricing anomalies, such as the predictability of excess returns, excess volatility, and the equity-premium puzzle. As with rational expectations (RE), the expectation formation mechanism we consider is based on a rigorous optimization algorithm that does not presume misperceptions - it simply departs from some of the implicit assumptions that underlie RE. Agents fear that existence of misspecifications and design strategies that will be robust against a very large class of misspecifications. The new element is that uncertainty cannot be modeled via probability distributions. We consider an asset pricing model where uncertainty is represented by unknown disturbance sequences, as in the H-infinity-control literature. Agents must filter the persistent' and transitory' components of a sequence of observations in order to make consumption and portfolio decisions. We find that H-infinity forecasts are more sensitive to news than RE forecasts and equilibrium prices exhibit the anomalies previously mentioned.