Abstract
The size properties of the recently advanced Lagrange Multiplier (LM) cointegration tests in the presence of structural breaks in time series are investigated. It is shown that misspecification of the types of breaks is liable to spurious rejection. In contrast, severe undersizing may result when ignoring the presence of any break.
Acknowledgement
Financial support from the Research Committee of the University of Macau under research grant no. RG001/08-09S/11T/TPS/FBA is gratefully acknowledged.
Notes
1 Westerlund and Edgerton (Citation2007) also introduced the normalized bias test, which performs less satisfactorily than the t-test. Hence, only the t-test is discussed here. Simulation results for the normalized bias test are qualitatively similar to those for the t-test and are available upon request.