Štefan Lyócsa: Investor attention and asset price fluctuations

• 2 October 2024
  10:00 AM – 11:00 AM

  Add to calendar

  Add to Google calendar

  Add to Microsoft 365 calendar
• Large meeting room 215, ECON MUNI, Lipová 41a, Brno

A public lecture of prof. Ing. Štefan Lyócsa, PhD., within his professorship procedure in Economics.

Lecture topic: Investor attention and asset price fluctuations

Lecture annotation: Investor attention is a limited cognitive resource and plays a key role in asset pricing models, as information can only be priced in if attention is paid to it. However, in an environment with abundant information, investors face the challenge of selecting the most relevant information. The literature suggests that investors tend to focus on "attention-grabbing" information. Building on this, we review several in-sample studies linking attention to major events with future asset price variation. However, using attention in an out-of-sample (predictive) context is more difficult, even with scheduled macroeconomic news announcements. Although these announcements are of interest to investors, their role in predicting price fluctuations is underexplored for several reasons: i) macroeconomic announcements are scheduled but infrequent; ii) announcement timing may shift; iii) market expectations of the variables change over time; iv) news may be priced in over an extended period; and v) the relevance of a given macroeconomic variable can vary over time. To address these issues, we estimate investor attention and sentiment towards ten scheduled macroeconomic variables, using data from social media, news articles, information consumption, and search engines. By applying standard and machine-learning methods, we improve volatility forecasts for nearly all 404 major U.S. stocks in our sample. Sentiment-based models consistently enhance volatility forecasts across all economic sectors, with an average improvement of 14.99% over the benchmark method—especially on days of extreme price variation, when forecast accuracy is most critical. The magnitude of improvements varies by data source used to estimate attention and sentiment and is strongest in data-driven machine-learning models.

Wednesday 2 October, ECON MUNI large meeting room, 10:00-11:00 AM.

You can also join on MS Teams: https://muni.cz/go/6e3baf

Organized by

Research and Projects Office (Faculty of Economics and Administration)

Responsibility

Mgr. Blanka Šustrová