|dc.description.abstract||Value at Risk using GARCH models remains the workhorse model in risk management as financial markets are becoming more complex followed by increased uncertainty worldwide. This research attempts to explore the comparative ability of the GARCH models in modeling and forecasting Value at Risk of the Zimbabwe Stock Exchange. The study used daily data of ZSE Industrial index for a period from 2010 to 2016. Data analysis was done using R version 3.5. The symmetric ARIMA-GARCH (1, 1) and asymmetrical ARIMA-EGARCH (1, 1) models were compared on the basis of AIC, BIC and log likehood values. The results shows that asymmetric ARIMA-EGARCH (1, 1) models outperforms the symmetric ARIMA-GARCH (1, 1). The Kupiec and the Christoffersen test were used for evaluation of VaR calculated using the different GARCH models. The most adequate GARCH family models for estimating VaR in the Zimbabwe Stock Exchange are the asymmetric ARIMA-EGARCH model with student-t distribution at 5% level of significance. Basing on the research findings a recommendation was also made to the investors and the stakeholders of the stock market to use an asymmetric GARCH model especially ARIMA-EGARCH model with student-t distribution that can capture leverage effect and risk factors in predicting VaR for the stock market. A recommendation was made for further studies that comparison of ARIMA-GARCH and other models like Multifactor Risk Approach, Risk metrics and Monte Carlo Simulation would help in the search of appropriate model for VaR estimation.
Keywords: Value at Risk, backtesting, GARCH models, ZSE||en_US