Forecasting Value-at-Risk of Asian Stock Markets Using the RDCC-GARCH Model Under Different Distributional Assumptions

The aim of this paper was to accurately and efficiently forecast from multivariate generalized autoregressive conditional heteroscedastic models. The Rotated Dynamic Conditional Correlation (RDCC) model with the Normal, Student's-t and Multivariate Exponential Power distributions for errors were used to account for heavy tails commonly observed in financial time series data. The daily stock price data of Karachi, Bombay, Kuala Lumpur and Singapore stock exchanges from January 2008 to December 2017 were used. The predictive capability of RDCC models, with various error distributions, in forecasting one-day-ahead Value-at-Risk (VaR) was assessed by several back-testing procedures. The empirical results of the study revealed that the RDCC model with Student's-t distribution produced more accurate and reliable risk forecasts than other competing models.