Model Uncertainty in the Cross Section

Abstract

We develop a transparent Bayesian approach to quantify uncertainty about linear stochastic discount factor models. Under our framework, posterior model probabilities increase with maximum in-sample Sharpe ratios and decrease with model dimensions. The entropy of posterior probabilities represents model uncertainty. We show that model uncertainty displays massive independent variations from popular uncertainty proxies, and episodes of heightened model uncertainty coincide with major market events. These observations hold not only in US markets but also in European and Asian Pacific equity markets. Moreover, positive model uncertainty shocks relate to sharp outflows from US equity mutual funds but significant inflows to government bond funds, with effects persisting for three years. Finally, the survey data suggests that investors tend to be more pessimistic about equity performance during periods of higher model uncertainty.

Keywords: Model Uncertainty, Linear Stochastic Discount Factor, Bayesian Inference

JEL Classification: C11, G11, G12.