Abstract

Introduction: Stakeholder engagement with Pillar 3 disclosures is key to understanding financial and regulatory data interaction. These disclosures provide critical insights into banks’ risk profiles, capital adequacy, and liquidity, ultimately informing about banks’ financial health. The interest in mandatory published information has been the focus of several studies, which have measured engagement through visits to bank web sections related to Pillar 3. These studies commonly relied on log file-based methodologies, which, while effective, are demanding in terms of data acquisition and preparation and are limited to analyzing data for individual banks. The present study aims to propose a methodology for investigating interest in mandatory disclosure using publicly available data, making it applicable to broader regions.

Methods: This study examines the temporal dynamics of user interest in information related to Basel III Pillar 3, analyzing variations across different quarters, subcategories, and languages within the Visegrad Four countries (Slovakia, Hungary, Poland, and the Czech Republic) from 2007 to 2022. Using Google Trends data and natural language processing (NLP) methods, we analyzed information-seeking behavior through keyword frequency (w_count), keyword specificity (w_specificity), and keyword density (w_density).

Results: The results reveal statistically significant differences in user interest across all examined factors. To maintain the power of significance tests for testing global null hypotheses, we employed modified tests for repeated measurements, due to the violation of the assumption of sphericity of the covariance matrix (e.g., w_density: W = 0.008, Chi-Sqr. = 647700.00, p < 0.0001)​. Adjusted univariate tests further indicated significant variations in search interest by quarter, subcategory, and language (p < 0.05)​. A multi-stage comparison identified distinct language homogeneous groups, with notable differences in the weight w_density. Each language formed its own group, highlighting diverse levels of interest – Poland, for instance, exhibited the highest search rates compared to other countries​. Seasonal trends were also evident, with increased search activity in the first and fourth quarters, coinciding with key financial periods such as annual reporting and audits.

Conclusion: The impact of Pillar 3 disclosures on information-seeking behavior varies by time period, information type, and geographical context. Our findings align with previous log file-based studies, confirming that user interest peaks during the first quarter and that semi-structured documents, such as annual reports, garner sustained attention. This research demonstrates the effectiveness of combining NLP techniques with publicly available data sources as a viable alternative when direct log files are inaccessible, offering valuable insights into stakeholder behavior patterns regarding financial disclosures.