Dynamic Heteroskedasticity and Taiwan Stock Return Nonlinearity

The results for the dynamic normal linear regression models fit in the previous chapter suggest that there are significant autocorrelation effects in the returns for all eighteen series examined and significant day-of-the-week effects, including a negative Monday effect, for four of the indices and two of the individual stocks. However, misspecification tests for these models reveal a number of potential complicating factors that raise questions about these results. A lack of normality, for example, calls into question the validity of the t-statistics used to determine which parameters were significant, as well as indicating the potential for more serious estimation problems.