Further results on the detection of changes in persistence in linear time series

Abstract

Recent research concerning the properties of ‘change in persistence’ tests in the presence of structural change is extended. It is found that the recently proposed tests of Leybourne, Kim and Taylor (2004) are subject to severe size distortion in the presence of both breaks in level and drift under the unit root null hypothesis. The present analysis suggests that any results obtained from these tests should be treated with caution to avoid drawing a spurious inference of a change in persistence.

Notes

¹ The superscript ‘f ’ is used to denote the use of the forward realizations of the series of interest. Later, reversed realizations of the series will be employed, with the superscript ‘r’ used.

² The following DGP was also employed by Leybourne et al . (1998) and Leybourne and Newbold (2000) to analyse the properties of the Dickey–Fuller test and weighted symmetric Dickey–Fuller test of Park and Fuller (1995) in the presence of breaks under the null.

³ Results for this single sample size are reported in the interests of brevity. Further similar results for alternative sample sizes are available from the author upon request.

⁴ The treatment of initial conditions, method of random number generation, sample size, and number of replications and discards for the break in drift experiments are the same as for the earlier level break experiments.