Applying topic modeling to Management Science: empirical results and implications

Topic modeling is a statistical method to study unstructured text collections. It is used to find the underlying classification structures of a corpus when the number of documents renders an exhaustive analysis implausible. In this study, we empirically test the applicability of topic models in the context of management science. We apply the technique to the articles published in the Management Science between 1998 and 2011, and compare the results with the 50th anniversary review done by the editorial team of the journal, for the overlapping period (1998-2004). This setting is particularly valuable as it encompasses distinguishing features of text analysis in social science such as the study of documents that are both, very diverse and highly interrelated. Methodological limitations, contributions and future avenues for research are discussed.