Testing for a Unit Root in the Presence of a Jump Diffusion Process with GARCH Errors.
Melbourne Institute Working Paper No. 28/06
Date: December 2006
Author(s):
Abstract
The use of GARCH and jump models to capture asset price dynamics is ubiquitous in economics and finance literature. We show that the size of Breitung (2002) nonparametric unit root test is robust to the presence of jump and GARCH errors but not for the other standard unit root tests. The power performance of all tests, except for Phillips (1987) test, is fairly robust provided that the mean process is not nearly integrated.