Bootstrapping cointegration tests under structural co-breaks: A robust extended ECM test

The aim of the paper is the analysis of ECM (Error Correction Model) bootstrap cointegration tests under structural breaks. Classical ECM tests depend on some nuisance parameters, which is an undesirable feature for empirical applications. This problem is overcome by using the bootstrap ECM test, which shows good size and power properties when there are no breaks. In this paper we study the small sample properties of alternative bootstrap ECM tests under different co-breaking situations. ECM test statistics are made robust to partial co-breaking by using extended error correction models or by imposing a common factor restriction.