Making differences work: Financial fraud detection based on multi-subject perceptions.

This study examines whether internal and external perceptions and perception differences can significantly improve the effectiveness of financial fraud detection. Specifically, we construct and validate management and media perceptions using various state-of-the-art methods. The empirical research is based on data from 2017–2021 from the Chinese market. The results reveal that the enhancement effect is more significant when considering multi-subject perceptions. A cross-sectional analysis and application of real scenarios further support the conclusions. The predictive direction and implications of perception indicators are fully discussed. The results imply that the media can perceive firms' potentially fraudulent behaviour and play a regulatory role. • This paper uses multi-source text data to detect financial fraud. • Fraud detection efficiency is significantly improved with multi-subject perceptions and perception differences. • Cross-sectional analysis confirms industry heterogeneity in fraud detection. • Predictive power analysis provides reliable results. • We confirm the important role of the media in fraud regulation. [ABSTRACT FROM AUTHOR]