Performance of LM-type unit root tests with trend break: A bootstrap approach

Using the bootstrap approach, we study the finite-sample properties of the Lagrange Multiplier (LM) unit root tests when level shifts are allowed under the null hypothesis. Bootstrapped critical values support the invariance property of the LM tests. Applying two LM-type tests to the Nelson-Plosser data, we find less evidence against the unit root null than that given by Zivot and Andrews [Zivot, E. and Andrews, D.W.K. (1992), “Further Evidence of the Great Crash, the Oil Price Shock and the Unit Root Hypothesis,” Journal of Business and Economic Statistics 10, 251-270.] when level shifts are allowed under the null.