Firm capabilities versus regional factors: A random forest approach for predicting toxic chemical releases among US industrial facilities

Predicting toxic releases remains a challenge for environmental protection, as effective regulation and mitigation depend on understanding the key drivers behind firms’ pollution behavior. We examine the factors that influence firm environmental pollution by determining whether corporate capabilities or regional factors better predict toxic chemical emissions. We combine toxic releases data across 1976 industrial facilities in the United States with firm-level financial performance indicators and regional socio-demography variables to identify the most significant predictors of environmental outcomes. We use a random forest model due to its ability to capture complex, nonlinear interactions among predictors and provide robust variable importance rankings. The results show that: (1) corporate characteristics, particularly net debt, net income, and operational size, are strong predictors of toxic chemical releases than regional factors, (2) among regional factors, population density emerged as the strongest predictor, and (3) larger firms and mid-sized firms with lower profitability release a similar predicted share of toxic chemicals (26–27 %), while mid-sized profitable firms release significantly more (46 %). The findings suggest that targeting corporate characteristics, particularly financial health and operational size, through policy interventions could be more effective in reducing toxic chemical releases, offering valuable insights for improving environmental performance at the firm level. • A random forest model predicts toxic chemical releases by US industrial facilities. • Financial performance is the top predictor of toxic chemical releases. • Operational size is also an important predictor. • Corporate characteristics are vital for predicting environmental performance.