On Sources of Risk Premia in Representative Agent Models

Abstract

We use options and return data to decompose unconditional risk premia into different parts of the return state space. In the data, the entire equity premium is attributable to monthly returns below -11.3%, but returns in the extreme left tail matter very little. In contrast, leading asset pricing models based on habits, long-run risks, and rare disasters attribute the premium almost exclusively to returns above -11.3%, or to the extreme left tail. We find that model extensions with a larger quantity of tail risk cannot account for the data, while models with a higher price of tail risk can.