Non-linear Spillover of External EPU on Macau Gaming Stock Volatility: Micro-foundations using TVP-VAR and ML Attribution

This paper examines the non-linear transmission of external Economic Policy Uncertainty to the volatility of Macau gaming stocks. By combining a TVP-VAR-DY framework with XGBoost and SHAP machine learning attribution, we map the dynamic risk spillovers originating from Mainland China, Hong Kong, and the United States. The empirical results reveal a dual risk topology: regional regulatory shocks manifest as transient, high-intensity pulses (episodic), whereas global systemic uncertainty exhibits a chronic structural persistence that determines the baseline volatility regime. Furthermore, our machine learning analysis reveals that firm resilience depends on specific micro-financial foundations. We find that operating profitability effectively buffers external shocks, whereas financial leverage significantly amplifies them. The study uncovers precise structural tipping points, identifying a sharp increase in risk sensitivity when the debt-to-asset ratio exceeds 80% and determining an optimal liquidity buffer at approximately 15% cash-to-assets. These findings challenge traditional linear assumptions and support the adoption of threshold-based macro-prudential regulations for the sector.