The approximate slopes and the power of the GMM overidentifying restrictions test

Abstract

The concept of approximate slopes is employed to study the power properties of the Generalized Method of Moments (GMM) overidentifying restrictions test. Two versions of the overidentifying restrictions test are compared globally by means of the approximate slopes approach. With no autocorrelation in the moment functions, it is found that the GMM overidentifying test with the mean deviation covariance matrix is more powerful than the test using the conventional non-mean deviation one.

Acknowledgement

The author would like to thank Professor Bruce Hansen and Professor Yuichi Kitamura for their valuable comments and helpful discussions. All errors which remain are mine.