Attention-related predictability in the stock market
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Business |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2024
Department
Major/Subject
Mcode
Degree programme
Finance
Language
en
Pages
55
Series
Abstract
The attention of investors has profound implications on their investment decisions and the aggregate movements of the markets. It underlies the learning of new economic information as well as the decisions to act. The prior body of academic literature has detailed the nature of attention both theoretically and empirically. Theoretical models consider attention to be a limited resource, which is allocated to reduce the uncertainty in portfolios and to generate higher economic gains. Empirical studies have validated these notions and provided further insights by employing investor attention proxies to measure it indirectly. Notably, the absence of attention has been associated with the prominence of post-earnings-announcement price drifts, in which the stock prices tend to continue drifting in the direction of the earnings surprise. The present study aims to determine whether the price drifts are influenced by the ownership structures of the companies, with a focus on the implications of institutional attention for the price efficiencies and the predictability of these drifts. In addition, the utility of investor attention proxies in combination with modern machine learning methods for predicting the price drifts is evaluated. The research methods employed are primarily based on regression analyses and the application of neural networks. Stocks classified as being institutionally owned displayed more efficient prices than the stocks with a more retail-oriented investor base. Conversely, the price drifts of retail-owned stocks appeared to be more predictable. Additionally, the combination of modern machine learning methods and investor attention proxies turned out to provide substantial practical utility in predicting the post-earnings-announcement price drifts and generating economic gains. In particular, it was found that the price drifts following negative earnings surprises were more predictable and greater in magnitude than those following positive earnings surprises, yielding information ratios twice as large in comparison. Overall, the conducted analyses provided further insights into the progressions of post-earnings-announcement price drifts and demonstrated the viability of using machine learning methods combined with investor attention proxies in forecasting them. Although the results of this study were compelling, numerous areas for improvement and further exploration were identified.Description
Thesis advisor
Suominen, MattiKeywords
investor attention, machine learning, neural networks, forecasting