An explainable machine learning framework linking ESG performance and supply chain resilience prediction: evidence from Chinese listed companies

Supply Chain Resilience (SCR) has become a strategic capability for firms facing increasing uncertainty. However, the vast majority of existing research falls into an “ex-post explanation” paradigm, which focuses on retrospective attribution analysis of realized shock outcomes or treats resilience merely as a static outcome variable. This paradigm fails to meet the urgent needs of firms for ex-ante risk early warning and capability assessment in uncertain environments. To fill this gap, this study develops an explainable machine-learning framework to predict resilience using downside-risk-based labels and comprehensive firm characteristics. “Using a comprehensive panel dataset of Chinese A-share listed companies spanning raw operational data from 2010 to 2023 (with predictive modeling evaluated on 2014–2023 labels), this study integrates financial flexibility, supply-chain structure, corporate governance, and ESG performance into multiple machine-learning models, among which XGBoost demonstrates the strongest predictive accuracy (AUC = 0.868). SHAP analysis further reveals heterogeneous and nonlinear mechanisms underlying resilience formation. Crucially, the model uncovers an internal hierarchy within the ESG framework: Governance (G) explicitly dominates Environmental (E) and Social (S) dimensions as the primary predictor of crisis survival. Specifically, governance quality and institutional ownership are associated with strong monotonic increases in predicted resilience, acting as immediate risk-governance mechanisms, whereas E and S factors serve as latent reputational buffers. Furthermore, liquidity shows diminishing marginal predictive contributions, and customer concentration has a U-shaped relationship driven by strategic dependence versus vulnerability. These findings broaden theoretical insights into the ESG–resilience nexus and highlight the necessity of balancing governance discipline, financial adaptability, and relational coordination to enhance organizational resilience. The research contributes a transparent and forward-looking decision-support tool for firms and policymakers who seek to improve resilience under evolving supply-chain risks.