Paper Type
Short
Paper Number
PACIS2026-2138
Description
Amid the digital transformation of capital markets, managerial strategic evasion on investor-listed company interactive platforms poses latent market risks. Based on 5.8 million question-answer pairs of Chinese A-share listed firms (2010–2024), this study constructs a refined managerial evasive behavior index via LLM prompt engineering, empirically investigating its impact on stock price crash risk and the mediating effect of large shareholder shareholding reduction. Results show managerial evasion significantly elevates crash risk, with robustness verified by excluding the 2015 stock market crash sample and replacing the core variable with cosine similarity. Mechanism tests confirm large shareholder shareholding reduction exerts a partial mediating role: managerial evasion exacerbates information asymmetry, enabling large shareholders’ opportunistic share reduction, which sends negative market signals and amplifies crash risk. This study innovatively applies LLM prompt engineering to quantify managerial strategic information behavior in financial text analysis, deepens insights into the economic consequences of digital interactive disclosure.
Recommended Citation
Huang, Jinshui and Wang, Jun, "LLM-Driven Evasion Index and Stock Crash Risk: The Mediating Role of Large Shareholder Sell-Offs" (2026). PACIS 2026 Proceedings. 11.
https://aisel.aisnet.org/pacis2026/data_analtyics/data_anltics/11
LLM-Driven Evasion Index and Stock Crash Risk: The Mediating Role of Large Shareholder Sell-Offs
Amid the digital transformation of capital markets, managerial strategic evasion on investor-listed company interactive platforms poses latent market risks. Based on 5.8 million question-answer pairs of Chinese A-share listed firms (2010–2024), this study constructs a refined managerial evasive behavior index via LLM prompt engineering, empirically investigating its impact on stock price crash risk and the mediating effect of large shareholder shareholding reduction. Results show managerial evasion significantly elevates crash risk, with robustness verified by excluding the 2015 stock market crash sample and replacing the core variable with cosine similarity. Mechanism tests confirm large shareholder shareholding reduction exerts a partial mediating role: managerial evasion exacerbates information asymmetry, enabling large shareholders’ opportunistic share reduction, which sends negative market signals and amplifies crash risk. This study innovatively applies LLM prompt engineering to quantify managerial strategic information behavior in financial text analysis, deepens insights into the economic consequences of digital interactive disclosure.
Comments
05-DataAnalytics