Hype or Overhype? Effects of AI Orientation on Firm Performance and Market Valuation [PDF]
with Xuhui (Nick) Pan, Hua (Jonathan) Ye
Abstract: We measure firms' Artificial Intelligence (AI) orientation using their earnings conference calls. Our analysis reveals that a firm’s AI orientation negatively predicts its performance, but is positively received by the market, as indicated by a higher Tobin’s Q. These effects persist for at least eight quarters. The discrepancy between firm performance and market valuation is particularly pronounced among firms that are relatively underperforming compared to their industry peers. Our findings remain robust across instrumental variables regressions, different time periods, and alternative measures of AI orientation. We also find evidence that investors are, at least to some extent, aware of the (over)hype surrounding AI. Overall, our findings contribute to the ongoing debate about how AI orientation influences market dynamics and shapes investor expectations.
Keywords: AI orientation, firm profitability, market valuation, AI hype
The Sound of Sentiment: Vocal Cues in Conference Calls and Retail Trading (Job Market Paper) [PDF]
Abstract: This paper investigates the impact of vocal cues from social media on retail trading. Using audio recordings of conference calls from SeekingAlpha, I examine how vocal emotions and listenability, in addition to textual sentiment, influence future retail trading activity. I find that vocal emotions have short-term predictive power for retail stock and option trading beyond what is captured by textual sentiment, suggesting that vocal cues convey unique information. Unlike textual sentiment, the effect of vocal emotions varies by emotion type, as some emotions—such as surprised—cannot be simply interpreted as positive or negative signals. Finally, the findings reveal a disagreement channel, through which vocal emotion and listenability influence retail investor disagreement within a ten-day horizon, with effects varying across different emotion types. This study contributes to the literature by demonstrating how auditory information on social media shapes retail investor behavior in addition to textual information.
Keywords: Vocal Cues, Conference Calls, Retail Trading, Social Finance
Presentations: The 2nd Workshop on LLMs and Generative AI for Finance (Scheduled)
Cybersecurity Risk and Bank Lending (Second-year Paper) [PDF]
Abstract: I explore the intricate relationship between cyber risks faced by banks and their lending outcomes. By using a comprehensive cyber attack event dataset sourced from Reprisk, my research reveals a value destruction influence of cyber attacks on both banks (-0.7% on event day) and connected borrowers (-0.7% on event day for top 5 borrower and -0.2% for all borrowers) who initiated a syndicated loan with the banks prior to the attack date within 3 years. Additionally, my analysis shows a diminishing impact of cyber attacks on future lending amounts. Surprisingly, when examining ex ante cyber risks using a text-based measure derived from a BERT language model, no discernible effect is observed on bank lending at both the aggregate and loan-specific levels. I extend my investigation to the realm of cyber risk regulation in New York, where impacted banks exhibit a heightened awareness of borrowers’ cyber risks in determining loan spreads. My findings contribute to a nuanced understanding of the multifaceted interactions between cyber risks, banking lending, and regulatory responses.
Keywords: Cybersecurity risk, Bank Lending, BERT
None