Generalized Method of Moments versus Generalized Quasilikelihood Inferences in Binary Panel Data Models

Generalized method of moments (GMM) estimation approach has a long history in the econometrics literature. Since the seminal paper of Hansen (1982, Econometrica), this GMM approach has been widely used mainly by econometricians to obtain consistent and efficient estimates for regression parameters in the class of linear dynamic dependence models. To reflect its importance on both theoretical and applied econometrics, the Journal of Business and Economic Statistics (JBES) has recently launched an interesting special issue (JBES, 2002, Vol. 20, No. 4) on this GMM estimation approach. In this paper, we propose a generalized quasilikelihood (GQL) approach, which, as compared to the GMM approach, appears to produce more efficient estimates for the regression and the dynamic dependence parameters in the non-linear regression set up. We examine this superior efficiency performance of the GQL approach in the context of binary panel data analysis, through both asymptotic and simulation studies. The GMM and GQL estimation approaches are illustrated by re-analyzing the Survey of Labour and Income Dynamics (SLID) data from Statistics Canada.