Unit root tests for time series with level shifts: A comparison of different proposals

A number of unit root tests which accommodate a deterministic level shift at a known point in time are compared in a Monte Carlo study. The tests differ in the way they treat the deterministic term of the DGP. It turns out that Phillips-Perron type tests have very poor small sample properties and cannot be recommended for applied work. Moreover, tests which estimate the deterministic term by a GLS procedure under the unit root null hypothesis are superior in terms of size and power properties relative to tests which estimate the deterministic term by OLS procedures.