Video-Based Deception Detection and Financial Fraud

We examine whether deception indicators based on video analyses predict fraud. Applying machine learning algorithms to firms’ IPO roadshow videos that are required to disclose in🔸快乐10分玩法说明, we construct a deception score that quantifies the probability of managers deceiving in the video. We show that the video-based deception score predicts fraud, after controlling for financial, textual, and audio fraud indicators. The result is stronger for IPO firms backed by VCs, with greater management shareholdings, and with lower profitability. In addition, firms with higher deception scores experience higher managers’ stock sales and lower returns during a 30-day, 60-day and 120-day window immediately after the lock-up period expires. The evidence suggests that firms with higher deception scores during roadshows are able to achieve higher profit for their offered shares until insiders (pre-IPO shareholders/managers) are allowed to exit. Moreover, the predictive power of the video-based deception score comes from the deceptive information instead of managers’ untruthful personality. Facial cues in discussing financial topics are especially useful for fraud detection.
Contact Emails: