Abstract: Under the registration system with information disclosure at its core, the quality of information disclosure by listed companies and its economic consequences have attracted significant attention. On the basis of using deep learning models to construct interactive Q&A indicators for measuring IPO online roadshow strategies, this study empirically tests the impact and mechanism of template based Q&A and non-question based strategic interactive Q&A in the interactive communication section of IPO online roadshows on stock returns on the first day of listing. The findings show that strategic interactive Q&A has a negative impact on stock returns on the first day of listing. The mechanism analysis reveals that strategic interactive Q&A can generate negative emotions among investors, thus reducing stock returns on the first day of listing. Further analysis reveals that strategic interactive Q&A sessions related to the issuance overview, basic information of the issuer, corporate governance, and investor protection themes are more likely to generate negative emotions among investors, thereby having a negative impact on stock returns on the first day of listing. There is no significant difference in market reactions caused by strategic interactive Q&A of listed companies on the main board and the science and technology innovation board, indicating that compared to the main board market that implemented the approval system at the same time, the quality of information disclosure for IPO online roadshows of listed companies on the STAR market that implemented the registration system has not improved. To this end, the regulatory authorities need to carry out multi link and full chain supervision before, during, and after the event. The companies planning to go public need to establish a new type of investor relationship.

Key words: strategic interactive Q&A, IPO online roadshow, STAR Market, IPO returns, deep learning