Transition to mandatory corporate social responsibility (CSR) reporting in Europe following Directive 2014/95/EU in 2014 opens new possibilities for comprehensive analysis of non-financial reporting in an environment of regulatory change. A significant fraction of CSR information exists in the form of text, and the number of entities required to publish this information includes all large European companies. Textual nature of data and lack of uniform presentation call for non-traditional tools to measure the change in reporting quality and its capital market effects. We address this issue using textual analysis, which seems to be a promising channel to capture regulatory impact. Since firms are free to choose whether to publish their CSR information in a stand-alone report or to incorporate it into the annual report, and many chose the latter, we are faced with a first formidable task. While isolating CSR discussion and analysis in a dedicated report is not an issue, its extraction from the annual report requires developing an adapted algorithm. The first chapter of the thesis describes such an algorithm developed for that purpose.
In our second chapter, we analyze CSR reporting by large European listed companies from 2013 to 2017, which is pre and post-adoption of Directive 2014/95/EU mandating non-financial reporting in the European Union. We assess the Directive’s impact on quantitative and qualitative dimensions of CSR disclosures published in annual and stand-alone reports. A significant change in the number of reports is the first indication of the Directive’s impact. We then operationalize qualitative guidelines of the EU Commission using Natural Language Processing metrics. Our findings confirm expectations of a quantity and quality change despite weak enforcement of Directive.
In our final chapter, our primary focus is to analyze the capital market conse- quences of mandatory CSR reporting around the new Directive. Early studies report that firms providing more public information can reduce information asymmetry proxied by bid-ask spread (Leuz and Verrecchia (2000)). We provide a comprehensive investigation of the link between textual attributes and bid-ask spread in CSR. Based on stand-alone CSR reports and CSR disclosures extracted from annual reports, we build an index consisting of textual attributes. We establish a significant negative relation between our index and bid-ask spread for the period from 2011 to 2017. Analysis of index components reveals that its main drivers are similarity and com- plexity, which increase transparency captured by the bid-ask spread significantly. We further apply our index to assess the economic role of changes in reporting initiated by Directive. We find a significant shift in the index implying economic consequences around 2017 - the year of the actual implementation of Directive. Our additional analysis shows the importance of both sources of reports: annual and stand-alone CSR reports. However, disclosures do not improve financial users’ informational environment when published in both sources.
Overall, this study contributes to CSR literature investigating economic conse- quences of disclosure regulations (Ioannou and Serafeim (2019)), summarizing CSR narratives with alternative measures besides non-transparent ratings (Muslu et al. (2017)) and extending scope of CSR disclosure analysis by including annual reports (Mittelbach-Hoermanseder et al. (2019)).