Assessing Volatility Modelling using three Error Distributions

Abstract

The current study focuses on estimating the volatility of stock returns in the presence of flat tails error distribution (i.e. asymmetry of the distribution) which a traditional generalized auto-regressive conditional heteroscedasticity GARCH model usually fails to explain. The study, unlike the previous studies, compares three sets of error distributions for GARCH (1, 1) model of stock returns.  The three sets of error distributions used for comparing the predictive ability of GARCH (1, 1) model are –Gaussian (normal distribution), student’s t and generalized error distribution (GED). Eviews software was used for analyzing a time series data of Flying cement stock shares consisting of 245 days of in sample and 15 days of out-of-sample data. To compare the forecasting capability of three models root mean square (RMSE) and Theil’s Inequality Coefficient (TIC) were used. Akaike information criterion (AIC), the Schwarz information criterion (SIC), Hannan, and Quin (HQ) information criteria were examined for selection of a suitable model for capturing volatility of stock returns in the presence of symmetrical and asymmetrical distributions. Results of the study revealed that GARCH (1, 1) with GED is the best model for capturing the volatility of stock returns of Flying Cement Industry. Results of the present study will provide a stimulus to academia and practitioners for incorporating asymmetry aspect of the distribution in future prediction and capturing volatility of stock returns.