Evaluating stochastic discount factors from term structure models

Abstract

This paper introduces a new approach to testing continuous time models which can be applied to almost any pricing model and does not rely on the pricing equation having a closed form solution. It also allows use of more primitive assets then factors without having to specify an error distribution. This is achieved by examining the model in terms of the Stochastic Discount Factor (SDF) implied by the model. In this paper I examine the performance of asset pricing models from the term structure literature to illustrate the approach. I propose an observable proxy for the SDF from continuous time models and demonstrate via monte carlo methods that this proxy is an extremely good approximation to the true, unobservable SDF. I show how to use the SDF to value a series of returns. I find that the most popular one-factor parametric models can be rejected by the data. I also examine the most popular two-factor models which have been proposed in the literature.