Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
Please note that all times are shown in the time zone of the conference. The current conference time is: 21st Apr 2025, 08:48:50am CEST
Washington University in St. Louis, United States of America
Discussant: Romain Boulland (ESSEC)
In this study, I examine the value of information from sell-side analysts by analyzing a large corpus of their written reports. Using embeddings from state-of-the-art large language models, I show that textual information in analyst reports explains 10.19% of contemporaneous stock returns out-of-sample, a value that is economically more significant than quantitative forecasts. I then perform a Shapley value decomposition to assess how much each topic within the reports contributes to explaining stock returns. The results show that analysts' income statement analyses account for more than half of the reports' explanatory power. Expressing these findings in economic terms, I estimate that early acquisition of analysts' reports can yield significant profits. Analysts' information value peeks in the first week following earnings announcements, highlighting their vital role in interpreting new financial data.
Mental Models and Financial Forecasts
Prof. Francesca Bastianello1, Prof. Paul Decaire2, Prof. Marius Guenzel3
1University of Chicago; 2W.P. Carey School of Business, Arizona State University, United States of America; 3University of Pennsylvania
Discussant: Diego Garcia (University of Colorado Boulder)
We uncover financial professionals’ mental models—the narratives they use to explain their subjective beliefs. Using 82,000 equity reports, we prompt large language models (LLMs) to extract 3.5 million narratives, each combining a topic, valuation channel, sentiment, and time outlook. To validate the reliability of our output, we introduce a multi-step LLM-based approach and new diagnostic tools. We establish three sets of findings. First, narratives are centered around a limited set of topics, primarily focused on top-line items, with variation in topic focus over time and across industries. Narratives are mostly forward-looking, with three times as many arguments focusing on the future as on the past. Second, differences in topic focus, sentiment, and time outlook across forecasters strongly predict the extent of disagreement in their subjective quantitative forecasts. Lastly, time-series variation in the average narrative’s sentiment and in the average narrative’s focus on top-line items closely track Shiller’s CAPE ratio (ρ = 0.84 and ρ = 0.42), and the cross-sectional variation in narratives predicts key asset pricing patterns. Narratives associated with growth stocks are more optimistic and forward-looking than those for value stocks, consistent with forecasters (mis)perceiving growth stocks as having above-average growth potential. Overall, this paper helps bridge the gap between ‘what forecasters believe’ and ‘why they believe it.’
