On the Robustness of Theoretical Asset Pricing Models

We derive a parsimonious returns-based stochastic discount factor that is robust to model misspecification. We consider a general equilibrium model with heterogeneous agents who can invest their wealth in many assets. As long as (i) agents have (individual-, time-, and state-dependent) recursive preferences that are homothetic in current consumption and continuation value with a common relative risk aversion coefficient γ and (ii) asset returns and individual state variables are conditionally independent (e.g., GARCH processes), we prove that the (−γ)-th power of market return is a valid stochastic discount factor. Within this class of models, asset prices are determined by relative risk aversion and technology alone, and “returns-based asset pricing” is robust to model misspecification as opposed to the consumption-based approach. We recast the equity premium puzzle as a consumption/saving puzzle, not as an asset pricing puzzle.