When Charts Deceive: Misleading Visualizations in Earnings Conference Calls
Release time：02 September 2025
Sep
05
|
Time & Date
|
10:30 am
-
12:00 pm,
September
05,
2025
(Friday)
|
|
Venue
|
Room 103, Conference Complex Ⅱ
|
| TOPIC | When Charts Deceive: Misleading Visualizations in Earnings Conference Calls |
| TIME&DATE | 10:30 am - 12:00 pm, September 5, 2025 (Friday) |
| Venue | Room 103, Conference Complex Ⅱ |
| Speaker |
Yang Cao The Hong Kong Polytechnic University |
| Abstract | Earnings conference call slides often employ visual tricks that can mislead investors, yet little is known about their prevalence or impact. We assemble 432,003 presentation slides containing 153,331 charts from 2003–2023 and leverage GPT 4o to flag five misleading features: missing Y-axis, truncated Y-axis, dual Y axes, 3D effects, and omitted data labels. We document that over 10% of slides, and more than half of calls, include at least one such feature. Misleading visuals are concentrated in non-financial disclosures and correlate systematically with firm traits: smaller, more profitable firms use them most, while volatile firms, those under CIO oversight, or with high retail investor engagement use them least. Importantly, calls featuring misleading charts earn positive short term abnormal returns that subsequently reverse, consistent with temporary investor misperception—an effect strongest for missing Y axes in non-financial slides. These tactics are persistent over time and appear to reflect deliberate disclosure strategies rather than one‐off lapses. Our findings underscore a regulatory blind spot: visual elements in corporate presentations deserve the same scrutiny as narrative and numerical disclosures. |
| Biography | Yang Cao is an Assistant Professor in the School of Accounting and Finance at The Hong Kong Polytechnic University (PolyU). Prior to joining PolyU, he earned his Ph.D. in Accounting from Boston College. His research primarily explores how emerging technologies—such as artificial intelligence and machine learning—impact information processing. |
