Time: 1:30 p.m., Tuesday, March 3, 2026

Venue: Room 104, Li Dak Sum Building

Host: Associate Professor Qinlin Zhong

Speaker: Da Xu (Tsinghua University)

Topic: Visualization and Mispricing

Abstract: Data visualization is now ubiquitous in modern business communication. Using customized CNN models, we identify data visualizations in more than 600,000 slide pages from earnings conference calls over the last decade. We find that the presence of visualizations diverts investors’ attention away from non-visualized slides, whose content can significantly predict future stock returns. A long-short strategy yields an annualized return of about 5.1%. This effect is more pronounced when investors’ attention is more constrained on the earnings call day. Further tests suggest that a top-down (goal-driven) attention process, particularly dual-coding synergy, is the primary mechanism through which visuals capture attention in earnings calls. Finally, although firms with weaker performance tend to use fewer visualizations, we find little evidence that managers strategically exploit this distraction channel to hide bad news in non-visualized slides. Overall, our findings caution that visualizations in corporate disclosures can distract attention in ways that potentially weaken market efficiency.