Nonlinear Insights Into the ESG–Crash Risk Dynamics From a Machine Learning Perspective: Evidence From an Emerging-Transition Market

ABSTRACT

  We examine how environmental, social, and governance (ESG) engagement influences stock price crash risk using a random forest model with Shapley Additive Explanations (SHAP) on KOSPI-listed firms (2012–2021). Large firms gain greater risk-reducing benefits, but high leverage or simultaneous ESG and R&D spending can heighten crash risk. When goodwill and return on assets (ROA) are both high, even low ESG engagement aligns with reduced crash risk, suggesting profitability and intangibles act as stabilizing buffers. Component analysis shows high environmental scores may raise crash risk, whereas governance provides a conventional mitigating effect. A two-stage, non-parametric mediation framework links ESG to an organizational resilience (OR) index capturing firm size, market value, liquidity, leverage, R&D intensity, board structure, and other fundamentals, which in turn suppresses crash risk. Results indicate ESG enhances OR, thereby curbing extreme crashes. Findings underscore the value of interpretable machine learning for revealing interaction and mediation effects beyond linear models.

2026년 1월
저널 Corporate Social Responsibility and Environmental Management (SSCI, JCR 랭킹 Q1)
논문 타이틀: Nonlinear Insights into the ESG–Crash Risk Dynamics from a Machine Learning Perspective: Evidence from an Emerging-Transition Market