Identification of Heteroscedasticity in the Presence of Outliers in Discrete-Time Series

Abstract

This study considered the effects of outliers on the identification of heteroscedasticity in the daily closing share price returns series of Diamond Bank, Fidelity Bank and Skye bank using correlogram, Ljung-Box test and Lagrange Multiplier test. The data were obtained from Nigerian Stock Exchange from January 3, 2006, to November 24, 2016, and comprises 2690 observations. About Seventeen outliers were detected in the return series of Diamond bank, sixteen outliers identified in the return series of Fidelity bank and twenty-six outliers found in Skye bank, and their effects were removed to achieve an outlier adjusted series for respective banks under study. Meanwhile, heteroscedasticity was found to exist in the two (the outlier contaminated and the outlier-adjusted) series. However, the results of our findings indicated that outliers could hide significant heteroscedasticity in correlogram, minimize the power of Ljung-Box test and amplify the power of Lagrange Multiplier test. The implication is that failure to account for outliers would result in impaired or spurious heteroscedasticity detection in discrete-time series. Thus, the strength of this study is in highlighting the undesirable effects of outliers on heteroscedasticity detection.