Measuring Crypto Tail Risk under Extreme Shocks: An EVT-GARCH and CAViaR Comparative Study
TL;DR
The FTX window is associated with additional volatility in all four assets, including de-pegging-related risk in DAI, which supports the joint use of complementary models in cryptocurrency stress testing and risk monitoring.
Abstract
Cryptocurrency markets face frequent extreme price swings, flash crashes and liquidity shocks. The expansion of decentralized finance, stablecoins and interconnected trading venues further complicates tail-risk measurement. This study examines extreme downside risk in Bitcoin, Ethereum, Uniswap, and DAI, treating the 2022 FTX collapse as the primary exogenous shock event. Daily closing prices from the Kaggle dataset are aligned by calendar date and transformed into logarithmic returns. Descriptive statistics and Jarque-Bera tests are first conducted, followed by an EVT-GARCH model incorporating an FTX event dummy and an Asymmetric Slope CAViaR model for tail-risk estimation. The results reveal significant deviations from normality, heavy tails, and substantial differences across asset types. The FTX window is associated with additional volatility in all four assets, including de-pegging-related risk in DAI. Backtesting at the 99% and 95% VaR levels indicates that both models provide reliable risk forecasts. EVT-GARCH generates smoother risk estimates, while CAViaR responds more quickly to recent negative shocks and records smaller violation-count deviations for the benchmark assets. These findings support the joint use of complementary models in cryptocurrency stress testing and risk monitoring.