Forecasting climate-sensitive industries' volatility: A regime-switching GARCH-MIDAS approach with multiple climate risk indicators

This study investigates the predictive power of multiple climate-related indicators in forecasting volatility across climate-sensitive industries using a regime-switching generalized autoregressive conditional heteroskedasticity mixed-data sampling (GARCH-MIDAS) model. We examine environmental, social, and governance (ESG) metrics, climate policy uncertainty (CPU), the Transition Risk Index (TRI), the Physical Risk Index (PRI), and economic policy uncertainty (EPU) to predict stock return volatility. Our analysis covers major indices including renewable energy, transportation, mining, aggregate energy, and the green economy across Asia, Europe, and the United States. Empirically, out-of-sample results reveal that the ESG and CPU indices are superior predictors of volatility for renewable energy, clean energy, and green economy indices, particularly in Asian and U.S. markets. PRI and EPU indicators demonstrate significant predictive power for volatility in the energy, mining, and transportation sectors. Incorporating uncertainty factors into the Markov regime-switching GARCH-MIDAS framework substantially improves forecast accuracy, as supported by both economic and statistical metrics. These improvements are validated through R2 direction of change and model confidence set tests. The findings carry important implications for climate policy development and implementation, offering critical insights for policymakers, investors, and industry stakeholders navigating the complexities of climate-sensitive sectors.