Corporate social responsibility reports: topic analysis and big data approach

• Published in: The European journal of finance Vol. 25; no. 17; pp. 1637 - 1654
• Main Authors: Goloshchapova, Irina, Poon, Ser-Huang, Pritchard, Matthew, Reed, Phil
• Format: Journal Article
• Language: English
• Published: London Routledge 22.11.2019; Taylor & Francis LLC
• Subjects: Bias; Big Data; Corporate responsibility; Corporate social responsibility; Data mining; Drugs; Emissions; Employees; environment social and governance; Health care; Healthy food; Industrialized nations; latent dirichlet allocation; Packaging; Power; Securities markets; Social responsibility; topic modeling; Topics; Europe
• ISSN: 1351-847X; 1466-4364
• DOI: 10.1080/1351847X.2019.1572637

This paper performs topic modeling using all publicly available CSR (Corporate Social Responsibility) reports for all constituent firms of the major stock market indices of 15 industrialized countries included in MSCI Europe for the sample period from 1999 to 2016. Our text mining results and LDA analyses indicate that 'employees safety', 'employees training support', 'carbon emission', 'human right', 'efficient power', and 'healthcare medicines' are the common topics reported by publicly listed companies in Europe and the UK. There is a clear sector bias with industrial firms emphasizing 'employee safety', Utilities concentrating on 'efficient power' while consumer discretionary and consumer staples highlighting 'food waste' and 'food packaging.' To produce these results, we used a battery of python code to organize the hundreds of reports downloaded from Bloomberg and the internet, the latest R-algorithm to estimate LDA (Latent Dirichlet Allocation) model and the LDAvis interactive tool to visualize and refine the LDA model.