A1150
Title: Modeling and forecasting financial tail risk: Methods and applications
Authors: Cathy W-S Chen - Feng Chia University (Taiwan) [presenting]
Abstract: A unified Bayesian framework is presented for modeling and forecasting financial tail risk, with a focus on value-at-risk (VaR) and expected shortfall (ES). The first part develops nonlinear threshold extensions of ESCAViaR-type models, allowing for regime-dependent tail behavior and improved characterization of extreme risks in cryptocurrency markets. The second part introduces a semi-parametric approach that incorporates geopolitical narratives and range-based volatility measures to capture forward-looking uncertainty and high-frequency market dynamics. The third part considers a flexible distributional framework based on skewed and heavy-tailed specifications within a GARCH-type structure, enabling coherent joint modeling of VaR and ES while accounting for asymmetry and tail thickness. Across all three parts, the analysis employs Bayesian inference to accommodate parameter uncertainty and enhance predictive performance. Empirical studies on major financial assets, including cryptocurrencies and crude oil markets, demonstrate that incorporating nonlinear dynamics, distributional flexibility, and external information leads to substantial improvements in tail risk forecasting. The discussion includes evaluation using scoring rules and backtesting procedures, with emphasis on joint assessment of accuracy and calibration.
