Forecasting volatility in the Chinese stock market: Comparative performance of carbon transition risk indicators

This paper develops a series of carbon transition risk concern indices derived from multi-source textual data to examine financial market reactions within the context of carbon transition. Leveraging natural language processing techniques to extract sentiment and concern indicators, and using the monthly volatility of the Chinese stock market as the forecasting target, the study systematically evaluates the predictive performance of different information sources. The results demonstrate that the carbon transition risk concern index based on analysts’ reports exhibits a significant and robust advantage in forecasting market volatility. Furthermore, a comparative analysis of various information fusion strategies reveals that shrinkage modeling approaches, which effectively identify key predictors, substantially enhance prediction accuracy and robustness. The robustness of these findings is confirmed through model confidence set methods and analyses across different forecasting windows. Overall, this study highlights the heterogeneity of information sources in shaping market expectations, contributes new insights into the financial transmission mechanisms of carbon risk, and offers valuable implications for investors and policymakers.