Conditional Moment Estimation of Stochastic Discount Factors: A GEL Approach

This paper re-examines empirical inference on stochastic discount factors employing recently developed conditional moment procedures (Kitamura, Tripathi and Ahn, Econometrica, 2004; Dominguez and Lobato, Econometrica, 2004, inter alia). Unlike unconditional estimation methods, this approach does not imply potential losses of information and is therefore more efficient. Generalized Empirical Likelihood estimation is particularly suitable for this setting, as it has relatively low bias when estimation requires many moment conditions. The procedures employed in this study also allow us to conduct weak identification-robust inference, which is a pervasive feature in this type of models. Our approach provides an alternative to the kernel-based non-parametric method of Wang (Journal of Finance, 2003).