Title: Efficient estimation by fully modified GLS with an application to the environmental Kuznets curve
Authors: Yicong Lin - Maastricht University (Netherlands) [presenting]
Hanno Reuvers - Erasmus University Rotterdam (Netherlands)
Abstract: The aim is to develop the asymptotic theory for a Fully Modified Generalized Least Squares (FMGLS) estimator for multivariate cointegrating polynomial regressions. These regressions allow both deterministic and stochastic trends and their integer powers to enter the cointegrating relation and therefore form natural extensions of the linear cointegration framework. The FMGLS estimator relies on the inverse autocovariance matrix of the multidimensional errors and requires bias correction terms for endogeneity and serial correlation. These quantities are unknown in practice. To obtain a feasible FMGLS estimator, we propose a consistent estimator of the inverse matrix. The bias correction terms for endogeneity and serial correlation are conveniently obtained as byproducts from our estimation framework. The resulting feasible FMGLS estimator allows for standard asymptotic inference. Extending earlier work, we elaborate on the conditions that make this FMGLS estimator asymptotically equivalent to its ordinary least squares counterpart. Both finite sample and asymptotic efficiency gains can be substantial if the least squares estimators have different limiting distributions. A comprehensive simulation study supports our theoretical outcomes. As a practical illustration, we test for the Environmental Kuznets Curve (EKC) hypothesis in a panel of early industrialized countries.