Title: Flexible nonlinear trend specifications and cointegration
Authors: Hanno Reuvers - Erasmus University Rotterdam (Netherlands) [presenting]
Yicong Lin - Maastricht University (Netherlands)
Abstract: Inference is developed for a model that combines a power law trend specification with a polynomial cointegration framework. We provide the asymptotic distribution of the nonlinear least squares (NLS) estimator when the regression errors and integrated regressors are serially dependent and cross correlated. The limiting distribution is nonstandard thereby complicating inference. We discuss two methods to conduct inference: (1) fully modified estimation, and (2) simulated estimation of the null distribution. An extensive simulation study illustrates and compares the finite sample performance of these methods. Our approach is illustrated with a detailed study on the existence of the Environmental Kuznets Curve (EKC) for Belgium, Denmark, France, Netherlands, UK and USA over the period 1870-2014. The main question at hand is whether stochastic trends or deterministic trends are responsible for the nonlinearities observed in the data.