Market Risk Modeling and Stress Testing: Volatility Spillover, Asymmetric Leverage Effect and Two-Stage GARCH-DID Empirical Framework

Abstract

Against the backdrop of deepening global financial integration, cross-border liquidity shifts trigger volatility swings and asymmetric tail risks in China’s A-share and Hong Kong markets. Traditional single-stage GARCH models fail to identify causal policy spillovers, while static stress testing cannot capture time-varying risk structures. This paper constructs a two-stage GARCH-DID quasi-natural experiment framework, embedding DID interactions into the EGARCH conditional variance equation, and develops a fully reproducible RStudio rolling-window dynamic stress testing pipeline. Taking the 2020 Fed unlimited QE as a single-period exogenous shock, this study adopts a PSM-DID design with high-foreign-ownership firms as the treatment group. To isolate monetary spillovers from pandemic interference, it employs event-study parallel tests, double-layer placebo randomization and lagged grouping variables. Results show loose global liquidity exerts volatility smoothing effects and mitigates asymmetric leverage risk from negative news. Dynamic rolling-window EGARCH stress testing outperforms static historical approaches in out-of-sample tail risk forecasting, passing Kupiec and Christoffersen backtesting. Robustness tests-including double-layer placebo trials, two-way clustered standard error correction and alternative rolling windows—consistently validate core conclusions. This paper provides a standardized open-source technical paradigm for cross-border market risk supervision under capital account liberalization.