Unit root tests based on IV estimators for time series with multiple breaks

Abstract

Unit root tests are developed for multiple break time series. The tests, being based on an instrumental variable estimator and recursive mean adjustment, have standard Gaussian null asymptotics regardless of the number of breaks, which is not shared by other existing tests. A Monte-Carlo experiment shows that the proposed tests have stable sizes and reasonable powers. © 2008 Elsevier Ltd. All rights reserved.