Properties of Subjective Beliefs Estimators

Abstract

Information-theoretic methods have recently been proposed for the simultaneous recovery of investors’ beliefs about future macroeconomic and financial outcomes and their risk preferences from observed asset prices. These methods estimate beliefs and preferences to minimize the statistical discrepancy between the recovered beliefs and the true data generating process (DGP), subject to asset pricing Euler equation constraints. This paper develops the asymptotic properties of these subjective beliefs estimators. We compare empirically the beliefs recovered with alternative estimators in this class, that differ on the basis of the statistical divergence functions used to characterize the discrepancy between the beliefs and the DGP.