Structural breaks, unit roots and methods for removing the autocorrelation pattern

This paper examines the asymptotic and finite sample behaviour of the ADF and Phillips-Perron statistics under the presence of a break in the trend function. We prove that the break magnitude affects these two statistics in a similar way, drawing them closer the acceptance zone, the higher the break magnitude. By contrast, the influence of the lag truncation parameter does not have a similar effect on them, being very important for the ADF whilst negligible for the Phillips-Perron statistics.