IPO Breaking Anomaly in China: A Prospectus‐Based Textual Analysis

We explore the determinants behind China's Initial Public Offering (IPO) ‘breaking’ anomaly, assessing the predictability of both break issues and the initial performance of IPOs. By integrating textual analysis with an ensemble of advanced machine learning techniques, our findings reveal that a higher degree of similarity, diminished readability and a more pronounced negative sentiment within the prospectus are associated with an increased likelihood of break issues and diminished initial returns. The text features, particularly similarity and sentiment, demonstrate substantial predictive power. Furthermore, our research highlights the superiority of certain sophisticated machine learning methodologies over traditional parametric models when dealing with high‐dimensional data.