Green bonds, pivotal to green finance, draw growing attention from scholars and investors. Social media’s proliferation has amplified the influence of investor sentiment, necessitating robust analysis of its market impact. However, general sentiment lexicons often fail to capture domain-specific slang and nuanced expressions unique to China’s bond market, leading to inaccuracies in sentiment analysis. Thus, this study constructs a specialized sentiment lexicon for the green bond market, namely the COBC (Chinese online bond comments sentiment lexicon), to dissect bond market slang and investor remarks. Compared to three general lexicons (Textbook, SnowNLP, and VADER), it improves the average prediction accuracy by approximately 87.2% in sentiment analysis of Chinese online language within the green bond domain. Sentiment scores derived from COBC-based dictionary analysis are systematically integrated as predictive features into a two-stage hybrid predictive model is proposed integrating Support Vector Machine (SVM), Auto-Regressive Moving Average (ARMA), Bidirectional Long Short-Term Memory Networks (BiLSTM), and Attention Mechanisms to forecast China's green bond market, represented by the China Bond 45 Green Bond Index. First, ARMA-SVR is employed to extract residuals and statistical features from the green bond index. Then, the BiLSTM-Attention model is applied to assess the impact of investor sentiment on the index. Empirical results show that incorporating investor sentiment significantly enhances the predictive accuracy of the green bond index, achieving an average of 67.5% reduction in Mean Squared Error (MSE), and providing valuable insights for market participants and policymakers.

We use supervised machine learning to develop a Chinese language financial sentiment dictionary from 3.1 million financial news articles. Our dictionary maps semantically similar words to a subset of human-expert generated financial sentiment words. In article-level validation tests, our dictionary scores the sentiment of articles consistently with a human reading of full articles. In return validation tests, our dictionary outperforms and subsumes previous Chinese financial sentiment dictionaries such as direct translations of Loughran and McDonald’s (2011) financial words. We also generate a list of politically-related positive words that is unique to China; this list has a weaker association with returns than does the list of otherwise positive words. We demonstrate that state media exhibits a sentiment bias by using more politically-related positive and fewer negative words, and this bias renders state media’s sentiment less return-informative. Our findings demonstrate that dictionary-based sentiment analysis exhibits strong language and domain specificity.