| We introduce learning into a general equilibrium model with Epstein and Zin preferences and study the impact on equilibrium equity premium, risk-free rates and their volatilities. We consider two types of dynamics for the consumption and dividend growth rates: Gaussian noise process, and vector autoregressive process, with one in common---there exists parameter uncertainty that is being resolved through Bayesian learning. We assume that all the parameters of the process driving the fundamentals are unknown, and using Bayesian learning, naturally arrive at the time-varying expected consumption and dividend growth rates and fluctuating economic uncertainty. We show that learning and the choice of beliefs about the structure and parameters of the process driving the fundamentals are as important for asset pricing as the choice of preferences characterized by risk aversion and intertemporal elasticity of substitution. Introducing learning into the general equilibrium model with Epstein and Zin preferences allows us to generate substantial values for the equity premium with low degree of risk aversion. Moreover, we show that in the environment with higher estimation risk, a higher equity premium is generated. We show that it matters for asset pricing which process is believed to drive the economy. |
