Abstract: This paper takes the 2021 China Life employee whistleblower incident as a case. Based on the idea of active learning, this paper uses the two-layer corpus constructed by a human-machine co-working method to classify public comments at the sentimental level and synthesize public sentiments into a curve to analyze the impact of reputation risk events on the stock price of the involved company and the spillover effect on non-involved companies. The research discovers that public emotions have a significant impact on the abnormal return of China Life. Even after controlling for regulatory variables, the significance of its impact on abnormal returns remains unchanged. As important stakeholders of the company, the public may get emotional in reputation events and the emotion can influence their decision-making behavior, thereby changing the market’s expectation of the involved company’s future cash flow stability and subsequently impacting the company’s stock price. This incident also displays a spillover effect mainly driven by competitive effects. The objective and subjective conditions required to generate such spillover effects are: the involved company and non-involved companies must have a certain degree of similarity, and this similarity must be perceived by stakeholder groups. The less similarity and perception, the more prominent spillover effect and particularly competitive effects.

Key words: active learning, corpus, social media, reputational risk, spillover effect