Answers are grounded in this paper and cite their sources.
The Fundamental Review of the Trading Book (FRTB) replaces Value-at-Risk (VaR) with 97.5% Expected Shortfall (ES) as the benchmark for market-risk internal models. Features of China's A-share market—including price limits, T+1 settlement, and substantial retail participation—produce return tails that deviate markedly from normal and Student-t distributions. Standard models may therefore overestimate extreme risk. Using the CSI 300 Index, this study applies a filtered extreme value theory (EVT) framework. An EGARCH model with Student-t innovations first filters conditional heteroscedasticity; a peaks-over-threshold/generalized Pareto distribution (POT/GPD) model is then fitted to the tails of the standardized residuals. Rolling out-of-sample ES forecasts are subsequently evaluated by backtesting. Four findings emerge. First, EGARCH-t provides the best fit among the GARCH-family specifications and captures both asymmetric volatility and heavy tails. Second, filtered EVT passes the Acerbi-Székely backtest for 97.5% ES, whereas historical simulation and GARCH-t do not. Third, filtered EVT is the only model retained in the Model Confidence Set (MCS). Fourth, under the FRTB framework, filtered EVT requires approximately 16% less capital than historical simulation. A decomposition attributes this reduction to improved tail modeling rather than volatility modeling.
No takes yet. Share an insight, caveat, or question.
Manye Wang (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: