Relative efficiency of Component GARCH-EVT approach in managing intraday market risk

Abstract

The purpose of this study is to estimate intraday Value-at-Risk (VaR) and Expected Shortfall (ES) of high frequency stock price indices taken from select markets of the world. The stylized properties indicate that the return series exhibit skewed and leptokurtic distributions, volatility clustering, periodicity of volatility and long memory process in volatility, all of which together suggest the usage of Component GARCH- EVT combined approach on periodicity adjusted return series to forecast accurate intraday VaR and ES. Hence the study estimates intraday VaR and ES using Component GARCH-EVT combined approach with different innovation distributions such as normal, student-t and skewed student-t and compares its relative accuracy with the benchmark GARCH-EVT model with different distributions. The Component GARCH-EVT models in general perform better than GARCH-EVT models and the model with skewed student-t innovations forecasts more accurately. The study is useful for market participants involved in frequent intraday trading in such markets. (JEL: G10, G15, G17, G19)