A Unifying Framework for Analysing Common Cyclical Features in Cointegrated Time Series

Abstract

This paper provides a unifying framework in which the coexistence of different form of common cyclical features can be tested and imposed to a cointegrated VAR model. This goal is reached by introducing a new notion of common cyclical features, namely the weak form of polynomial serial correlation common features, which encompasses most of the previous ones. Statistical inference is obtained by means of reduced-rank regression, and alternative forms of common cyclical features are detected by means of tests for over-identifying restrictions on the parameters of the new model. Some iterative estimation procedures are then proposed for simultaneously modelling different forms of common features. Concepts and methods are illustrated by an empirical investigation of the US business cycle indicators.