Forecasting stock prices on the Zimbabwe Stock Exchange (ZSE) using Arima and Arch/Garch models

The main thrust of this study is to find out whether the stock prices on the ZSE can be predicted using ARIMA and ARCH/GARCH models. The ZSE currently does not have a model that predicts stock price movements. Thus this study attempts to explore econometrics models to predict future stock prices on the Zimbabwe Stock Exchange (ZSE) selected counters. Stock price data is differenced and tested for stationarity using KPSS test and the Augmented Dickey Fuller test. The final models are found to be Econet Wireless, ARIMA(1,1,0), Dairiboard, ARIMA(1,1,0), Delta, ARIMA(1,1,1), SeedCo, ARIMA(1,1,1) and Old Mutual, ARIMA(1,1,0). The GARCH(1,1 model for all the counters forecast better than ARIMA models considering the minimum deviations of the forecasted values from the actual ones. This is because the ARCH/GARCH models incorporate new information and analyses the series based on conditional variances where users can forecast future values with up to date information. Old Mutual had the best ARIMA model with the lowest error where as Dairiboard had the best GARCH model as shown by the minimum Schwarz criterion value of 1.365. We conclude that GARCH(1, 1) model outperforms ARIMA models in modeling stock prices in this study.