Disentangling and hedging global warming risk: A machine learning approach

As global warming provokes increasing attention from investors, this study disentangles global warming risk (GWR) for investors by leveraging energy futures volatilities. This study derives GWR from energy futures using an extreme gradient boosting (XGB)-genetic programming (GP) framework. Our XGB-GP framework develops volatility forecasting models for GWR from selected energy futures markets identified by XGB as key contributors to global warming, surpassing traditional models in forecasting accuracy. The originality of the study rests on the pioneering integration of the XGB-GP framework in predicting climate risk, linking energy futures markets with climate risk management and enabling feasible climate-featured portfolio hedging. Our study also sheds new insights for policymakers to design carbon trading systems and carbon pricing mechanisms, as they can use relevant energy futures prices as a basis for carbon trading calibration. • Global warming risk (GWR) significantly affects the valuation of listed companies. • An extreme gradient boosting-genetic programming (XGB-GP) model is used to predict GWR from energy futures volatilities. • The model identifies major contributors to GWR within energy futures markets. • Risk hedging strategies are proposed to help investors mitigate GWR using energy futures.