Efficacy of GARCH-EVT Model in Intraday Risk Management: Evidence from Severely Pandemic-affected Countries

Abstract

Countries around the world have experienced severe-to-moderate economic restrictions during the first two waves of COVID-19 pandemic. The present article captures the time frame of this unprecedented turmoil to test the efficacy of the conditional extreme value theory (EVT) model to forecast value at risk (VaR) and expected shortfall (ES). The article considers the eight most-affected countries and applies the generalized autoregressive conditional heteroskedasticity (GARCH) process in appropriate GARCH-EVT models assuming Student’s t distribution to accommodate the fat-tailed behaviour of financial return series. The study confirms that our model is substantially superior in forecasting intraday VaR and ES in comparison to other conditional EVT models with a classical assumption of normal distribution as well as an unconditional EVT model. Contrary to more commonly used 5-minute-frequency data in intraday risk management, we use hourly data, which is much less researched in the context of stock market volatility forecasting. Moreover, our study adds to the reliability of conditional EVT models because of its accuracy in predicting intraday risk during extreme economic uncertainty in severely pandemic-hit countries.