Assessing and interpreting carbon market efficiency based on an interpretable machine learning.

The urgency of addressing climate change and reducing carbon emissions necessitates the establishment of effective carbon markets. Our paper focuses on assessing and interpreting the efficiency of carbon markets, which is a crucial aspect of their successful implementation. We selected Hubei, Shenzhen, and Guangdong carbon pilots as examples to measure their dynamic information efficiency by combining the rolling time window and wild-bootstrap variance ratio test. Furthermore, we combined SHapley Additive exPlanations (SHAP) and Extra Trees (ET) regressor to reveal the nonlinear relationship between market performance and efficiency. Our findings highlight the time-varying nature of carbon market efficiency and reveal that China's carbon market has not yet achieved stable weak-form efficiency. Additionally, we identified excessive trading volume volatility as a significant constraint on market efficiency, while the trading volume itself exhibits weak explanatory power. Our proposed hybrid model demonstrates superior predictive and explanatory capabilities, offering regulators a valuable tool for continuously monitoring and interpreting carbon market efficiency. Finally, we suggested that reducing trading volume volatility should be a major concern for regulators to optimize market operations at present. [ABSTRACT FROM AUTHOR]