Modeling Heteroscedasticity in the Presence of Outliers in Discrete-Time Stochastic Series

The aim of this study is to examine the effects of outliers on the specification and efficiency of heteroscedastic models fitted to the daily closing share price returns series of two outstanding banks in Nigeria from January 3, 2006 to November 24, 2016. The series consists of 2690 observations for each bank. The data were obtained from the Nigerian Stock Exchange. GARCH(2,0) model with respect to student-t error distribution and GARCH(1,1) model under normal error distribution were successfully fitted to the outlier contaminated series of Diamond bank and United bank for Africa accordingly. On the contrary, EGARCH(1,1) model with respect to student-t error distribution adequately captured the changing variance in the outlier adjusted series of the two banks considered. Substantial evidence revealed that the presence of outliers in returns series leads to model misspecification and adjusting for such outliers ensures model efficiency.