The phenomenon of herding behavior can be observed throughout human societies and has been discussed from different perspectives in research areas such as information systems or finance. In this context, financial analysts have been found to exhibit a disposition towards herding behavior. We identify the topic selection within analyst reports as a potential source for the detection of herding behavior among financial analysts. In this paper, we utilize a qualitative dataset of about 140,000 analyst reports and 1,500 conference call transcripts. Building on the topic modeling technique latent Dirichlet allocation, we calculate similarity scores between the respective documents. Potential herding behavior is observable by revealing exceptional topic structures within groups of analyst reports. In this study, these are found especially during a phase of economic recession: the financial crisis spanning from 2008 to 2010. Viewing prospects, this approach might complement existing research in herding behavior by quantifying qualitative data not only within the financial domain but also in other research areas.
Palmer, Matthias; Eickhoff, Matthias; and Muntermann, Jan, "DETECTING HERDING BEHAVIOR USING TOPIC MINING: THE CASE OF FINANCIAL ANALYSTS" (2018). Research Papers. 97.