Size Distortions of the Wild Bootstrapped HCCME-Based LM Test for Serial Correlation in the Presence of Asymmetric Conditional Heteroskedasticity

This paper investigates the size distortions of HCCME-based tests for serial correlation and the wild bootstrapped counterparts in the presence of asymmetric conditional heteroskedasticity. Thereby, asymmetric effects are allowed to enter the residual process of the dynamic regression model in both the GARCH parameterization and the innovation process. Monte Carlo evidence reported in this paper indicates that wild bootstrap versions of the LM test for serial correlation tend to overreject the null hypothesis, but the problem is generally not very serious