基于跳跃风险的中国碳市场波动率预测研究

A Study on Volatility Prediction of China’s Carbon Market with Jump Risk Considered

摘要: 全球气候变化背景下，碳市场成为调控温室气体排放的核心机制。中国碳排放权交易体系(CCETS)自2021年启动以来，其价格受政策冲击、极端事件的影响出现跳跃波动，但现有研究普遍忽视这一关键因素。本研究基于2021~2024年中国碳市场日度交易数据，构建自回归条件跳跃强度(ARJI)模型与机器学习算法融合的预测框架，探究跳跃风险对碳市场波动率预测的影响。研究发现，ARJI-GARCH与 ARJI-GARCH-h_t模型均能有效捕捉跳跃风险，其中ARJI-GARCH-h_t模型通过引入时变跳跃幅度方差与非对称参数，显著提升了对突发波动的拟合能力。将两类模型测算的跳跃强度引入偏最小二乘(PLS)、径向基神经网络(RBFNN)等机器学习模型后，中国碳市场波动率预测精度显著提升，其中PLS-ARJI-GARCH-h_t模型表现更优。本研究为政策制定者优化碳配额分配、交易员设计抗风险策略提供了理论依据，也为新兴碳市场的波动率预测提供了方法借鉴。

Abstract: Against the backdrop of global climate change, carbon markets have emerged as a core mechanism for regulating greenhouse gas emissions. Since its launch in 2021, the price of the China’s Carbon Emissions Trading System (CCETS) has been presenting jumping fluctuations which are triggered by factors such as policy shocks and extreme events. However, existing studies have generally overlooked this critical factor. Based on daily trading data from the national carbon market from 2021 to 2024, this study constructs a forecasting framework integrating the Autoregressive Conditional Jump Intensity (ARJI) model with machine learning algorithms to investigate the impact of jump risks on carbon market volatility prediction. The study finds that both the ARJI-GARCH and ARJI-GARCH-h_t models effectively capture jump risks. Notably, the ARJI-GARCH-h_t model significantly enhances the fitting capability for abrupt volatility by introducing time-varying jump amplitude variance and asymmetric parameters. When the jump intensities calculated by the two models are incorporated into machine learning models such as Partial Least Squares (PLS) and Radial Basis Function Neural Network (RBFNN), the forecasting accuracy of China’s carbon market volatility improves significantly. Within those models, the PLS-ARJI-GARCH-h_t model represents the optimal performance. This study provides a theoretical basis for policymakers to optimize carbon quota allocation and for traders to design risk-resistant strategies. It also offers methodological insights for volatility forecasting in emerging carbon markets.