Title: On the estimation of value-at-risk and expected shortfall at extreme levels
Authors: Emese Lazar - University of Reading (United Kingdom)
Shixuan Wang - University of Reading (United Kingdom)
Jingqi Pan - University of Reading (United Kingdom) [presenting]
Abstract: Two dynamic semi-parametric models that estimate Value-at-Risk (VaR) and Expected Shortfall (ES) are generalized, specifically the one-factor GAS model and the Hybrid GAS/GARCH model, to enhance them for risk estimation at extreme levels (corresponding to very low values of alpha). We achieve this by simultaneously estimating VaR and ES for multiple levels of alpha. We found that this approach improves on the risk estimation for low alpha values by having a unique hidden process that drives risk estimates for multiple levels of alpha. The simulation results indicate that both generalized models are better than their corresponding benchmarks in terms of estimated loss, forecast loss and the percentage backtest rejections for extreme values of alpha.