Predicting corporate financial performance in the FMCG industry through machine learning: comparing mandatory and voluntary ESG disclosures
研究了强制与自愿ESG披露对快消品公司财务绩效预测的影响，发现强制披露下ESG更能提升预测准确性，且资产回报率和营业收入受益更稳定。
Numerous global agreements and protocols urge industries and businesses to address the challenges of global warming and climate change. Emerging concepts, such as environmental, social, and governance (ESG) scores, have been introduced to measure and encourage corporate responsibility in tackling these issues. This study investigates whether the predictive role of ESG and its pillars on corporate financial performance (CFP) in the fast-moving consumer gods (FMCG) industry varies between mandatory and non-mandatory ESG disclosure using machine learning (ML) techniques. In this context, the study analyzes data from over 174 FMCG firms in Western Europe and North America from 2013 to 2020, corresponding to the second commitment period of the Kyoto Protocol (SCKP). Methodologically, the proposed ML framework, which combines principal component analysis (PCA) with a multi-output gradient boosting model (GBM), demonstrates superior predictive performance compared to traditional models. The findings reveal that ESG improves the predictability of CFP more under mandatory disclosure regimes in Europe than under voluntary disclosure in North America, highlighting the value of regulated ESG transparency. In terms of financial metrics, return on assets (ROA) and operating income (OI) consistently benefit from ESG integration, while return on equity (ROE) appears more leverage-sensitive, reflecting differences in capital structures across regions. Regarding SHAP values, higher ESG and ENV generally enhance model predictions across CFP, while SOC and GOV demonstrate more variable and context-dependent impacts. Overall, the results provide evidence that mandatory ESG disclosure, combined with advanced ML models, strengthens the link between sustainability practices and corporate performance.