Specification Tests in the Efficient Method of Moments Framework with Application to the Stochastic Volatility Models

Ming Liu - The Chinese University of Hong Kong and Harold H. Zhang - Carnegie Mellon University

The efficient method of moments (EMM) estimation bases its inference on the versatile auxiliary model which is selected by a sieve seminonparametric (SNP) method. As the auxiliary model becomes rich enough, the EMM estimator is asymptotically efficient and the EMM criterion can be used as a specification test. While theoretically it is possible for the auxiliary model to become very large, in practice, the expansion of the model is often limited by the sample size. In this case, the approximation error of the auxiliary model could lead to serious inference bias. In this paper, we propose a new specification test which could minimize the inference bias caused by the approximation error in the auxiliary model. Empirical implementation of our new test to the stochastic volatility model fitted to the daily stock price index shows a sharp contrast to the existing test results based on the EMM criterion.

Scheduled for Session 5.2 Asset Markets