FORECASTING VOLATILITY OF STOCK MARKET INDICES WITH GARCH MODELS

ABSTRACT

This study examined the volatility forecast of stock indices in some selected stock markets using data from World Federation of Exchange. The efficacy of the volatility measures in forecasting stock indices are examined in the study using symmetrical and asymmetrical models. This study employed six volatility models to examine their forecasting accuracy and forecasting power of predicting stock market index volatility. The data covers specific stock markets in Africa but used an aggregate value of broad index for America, Asia – Pacific and also Europe – Africa – Middle East. The study used descriptive statistics for examining the normality characteristic of the series, while the Augmented Dickey – Fuller test was employed to examine the stationarity features of the series. The study employed the Breusch-Godfrey Serial Correlation LM test to examine whether there is ARCH effect in the series used for the study. The distributional assumption for the model were examined, the lag selection criteria was also employed in the study. In conducting the volatility test, optimal order selection was determined using model selection criteria. The volatility models were estimated in symmetric and asymmetric models. The ARCH-LM test was conducted to check whether there remains any ARCH effect in the series for the volatility model adopted. Then the forecast evaluation test was conducted to obtain the forecast evaluation statistics and lastly the variance ratio test was conducted on the model series to ensure appropriateness of the best model chosen by the model selection criteria and forecast evaluation. The statistical software employed in the conducting the statistical estimation is the EView 10 software.