A Robust Variance Bound on Pricing Kernels
53 Pages Posted: 29 Nov 2014 Last revised: 4 Dec 2014
Date Written: November 22, 2014
Abstract
This paper proposes a data-based measure of model performance to discriminate among competing asset pricing models of return predictability. I form a set of variance bounds on pricing kernels based on different systems for predicting asset returns. For a given asset pricing model, I define the robust variance bound to be the tightest variance bound that this model-implied pricing kernel is able to satisfy. Using the diagnostic results of the robust variance bounds, I then construct a model performance index. This index quantifies the degree of return predictability which a given asset pricing model is able to obtain. I apply this method to examine the performance of three leading classes of asset pricing models: long run risk, external habit and rare disasters. The long run risk type of rare disaster model of Nakamura et al. (2013) performs best.
Keywords: returns predictability, robust variance bound on pricing kernels, asset pricing models
JEL Classification: G12, G11, E44, E37, C11
Suggested Citation: Suggested Citation