Volatility Modeling for Forecasting Stock Index with Fixed Parameter Distributional Assumption

The aim of this paper is to empirically investigate the in sample and out of sample forecasting performance of several GARCH-type models such as GARCH, EGARCH and APARCH model with Gaussian, student-t, Generalized error distribution (GED), student-t with fixed DOF 10 and GED with fixed parameter 1.5 distributional assumption in case of Colombo Stock Exchange (CSE), Sri Lanka. The daily All Share Price Index (ASPI) of CSE from January 02, 1998 to December 29, 2006 for a total number of 2150 observations is used for empirical analysis. We consider first 1950 observations for in sample estimation and last 200 observations for out of sample forecasting evaluation. Our empirical study showed that fixed DOF 10 of student-t density and fixed parameter 1.5 of GED density fail to improve the in sample estimation performance compared to student-t and GED distributional assumption. Among all of these models, APARCH model with student-t density give better in sample estimation results. In case of out-of-sample forecasting performance we found that APARCH model with all distributional assumption give lower value of Mean Squared Error (MSE) and Mean Absolute Error (MAE). According to the densities student-t distribution with fixed DOF 10, student-t and Gaussian distributional assumptions give better results in case of GARCH, EGARCH and APARCH model respectively. The estimation results of SPA test suggest that APARCH model with Gaussian distributional assumption give better forecasting performance in case of all share price index of CSE, Sri Lanka.