Explainable AI for Financial Distress: Evidence from Market Volatility and Regime Dynamics

This study investigates the role of market volatility, proxied by the CBOE Volatility Index (VIX), as a potential regime-dependent interaction of corporate leverage risk within the S&P 100. Addressing the limitations of traditional financial distress models in capturing non-linear and regime-dependent dynamics, we employ XGBoost combined with SHAP-based explainable AI (XAI) on a longitudinal dataset spanning 2000–2025. The results show that Total Debt remains the dominant predictor of financial distress, while the predictive contribution of risk-related variables such as the VIX and equity returns increases during crisis periods. Monetary policy indicators become more important during pandemic conditions, whereas inflation dominates in a stable environment. This finding highlights the regime-dependent nature of financial risk drivers and demonstrates the value of explainable machine learning in developing interpretable risk diagnostic frameworks. By integrating predictive accuracy with interpretability, this study provides new insights into the non-linear interaction between firm-level leverage and external market volatility.