Prominent voices and prevalent discourses: A corporate social responsibility application

In this study we develop a simplified technique for identifying prominent voices (and characterizing prevalent discourses) using Text Data Mining around Corporate Social Responsibility (CSR) issues or topics. We do this by analyzing a corpus of CSR reports produced by 7 US firms (Citi, Coca-Cola, Exxon-Mobil, General Motors, Intel, McDonald's and Microsoft) in 2004, 2008 and 2012, and focusing on a reduced set of vectors — or Singular Vector Decompositions (SVDs)-derived from these CSR reports while exploring term associations (Text Topics or Term Clusters). Specifically, we use centroid clustering on these SVDs to identify centroid-guiding-CSR-report-components (or firms with prominent voices and prevalent discourses around a CSR topic). The analysis is performed by year in order to discern the way in which prominent voices and prevalent discourses (around CSR topics) have evolved through time. Results indicate that it is difficult for firms to maintain a prominent voice around CSR issues through time, and that when they manage to do so it is because the prevalent discourse has direct business implications.