ECM-Cointegration test with GARCH(1,1) Errors
By Panagiotis Mantalos
By using Monte Carlo experiment, we study the robustness of the ECM-Cointegration tests when the first difference of the series follow a GARCH(1,1) process and the behaviour of the ARCH-tests of ECM residuals for variation of GARCH parameters. We found that if two series are individually I(1) and follow a GARCH (1,1), then ECM-Cointegration test tend to overreject for small samples but the problem is not very serious for large sample except when the errorsīs GARCH process is nearly integrated and the volatility parameter is not small. Finally the Wild Bootstrap is robust on GARCH(1,1) without lost of power.
ARCH, Bootstrap, Cointegration, GARCH, Wild Bootstrap
Renato Di Lorenzo
READING THE ARTICLE: You can read the article in
portable document (.pdf) format (333075 bytes.)
NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.
This page has been accessed 3005 times since July 24, 2006.
Return to the Home Page.