 |
Machine Learning Forecasts of Tail Risk Spillovers in Carbon and Energy Markets
1 : Nanjing University of Finance & Economics
2 : Massey Institute at Nanjing University of Finance&Economics,
3 : McDonough School of Business, Georgetown University
This study investigates tail risk dependencies between the global carbon market and major energy markets using the multivariate multi-quantile conditional autoregressive value-at-risk model. To enhance predictive performance, we develop a bidirectional forecasting framework that integrates machine learning approaches including quantile random forests, gradient boosting, and quantile regression neural networks. These models significantly improve the accuracy of tail risk forecasts by capturing nonlinearities and complex cross-market interactions. Our results reveal substantial spillover effects, particularly strong bidirectional tail risk transmission between the carbon and natural gas markets. Importantly, incorporating cross-market tail risk indicators into the forecasting models improves early warning capabilities, demonstrating the predictive value of carbon market signals for broader energy market risks. These findings underscore the practical utility of machine learning-based forecasting for systemic risk monitoring and support the development of more responsive and data-driven risk management and regulatory policies in an era of heightened market volatility, environmental transition, and geopolitical uncertainty.
