Some Further Developments in Cointegrating Regression Analysis: Theory for Applications

Abstrac:

The analysis of linear cointegrating relationships in a regression framework is typically carried out using modified least squares estimators that involve corrections for endogeneity and error serial correlation in order to obtain limiting distributions that allow for asymptotic standard inference. Several such estimation procedures are available in the literature. In this talk we survey extensions of two of these estimators – Fully Modified OLS and Integrated Modified OLS - along four dimensions: First, the considered functional forms are more general, with a focus on polynomial-type functions that occur in many applications (environmental Kuznets curves, Translog cost and production functions, etc.). With respect to functional form choice, some effort is placed on specification analysis and testing. Second, we add the cross-sectional dimension to the analysis and discuss appropriate modelling and estimation strategies for situation with small cross-sectional dimension and cross-sectional dependences, a typical situation in applied macro-econometric analysis. Third, we relax the assumption of stationarity of the underlying processes, thereby extending the cointegration concept to the integrated locally stationary framework. The talk throughout illustrates the different aspects of the research agenda with applications and concludes with a brief outline of the steps ahead. The talk gives a bird’s eyes view of research undertaken in Dortmund as well as in cooperation with Robert M. de Jong, Mathias Vetter and Timothy J. Vogelsang.