GMM ESTIMATION OF A PRODUCTION FUNCTION WITH PANEL DATA: AN APPLICATION TO SPANISH MANUFACTURING FIRMS

In this paper we consider the estimation of a Cobb-Douglas production function using a panel dataset of Spanish manufacturing firms. As it is stressed in the econometric literature, the use of standard GMM first differences estimators to eliminate the unobserved firm-specific effects may yield imprecise estimates, particularly in the case of the estimation of the production function. The reason is that the high persistence of output and inputs involved in the estimation of production functions make that their lagged levels to be weak instruments for the first differences of these series. The extended GMM estimator proposed by Arellano and Bover (1995) considers further orthogonality conditions based on lagged differences as instruments for the equation in levels. This approach has been applied to the estimation of technological parameters by Blundell and Bond (1999). Our estimation results, based on this approach, confirm the better performance of the extended GMM estimator compared to the standard first-differenced GMM estimator