A0437
Title: Transformed cointegration models with partially linear additivity
Authors: Yingqian Lin - Shanghai University of Finance and Economics (China) [presenting]
Yundong Tu - Peking University (China)
Abstract: Two general classes of transformed cointegration models are considered. In the first class of models, the dependent variable, after a parametric monotonic transformation, is cointegrated with nonlinear parametric functions of unit root regressors and stationary regressors in a linear form. The second class augment the first class with unknown integrable functions of the unit root regressors in an additive way. Extremum estimators for the parameters in the transformation function, plug-in estimators for parameters in the linear components, and sieve estimators for the unknown functions are presented. Asymptotic properties of the proposed estimators are developed, which are shown to depend on the transformation, the functions and model parameters. The theory is further extended to allow for both endogeneity of the nonstationary regressors and serially dependent errors. Numerical results demonstrate the nice performance of the estimators, corroborate the theoretical development and illustrate the practical merits of the proposed models.
