Using structural topic modeling to gain insight into challenges faced by leaders

Abstract

This paper leverages technological and methodological advances in natural language processing to advance our understanding and approaches to leadership research by introducing structural topic models (STM) to researchers wanting to inductively code massive amounts of unstructured texts. Specifically, we illustrate the application of STM applied to a large corpus (N ≈ 8000) of unstructured text responses from a diverse sample of leaders to inductively generate a classification system of leader challenges and simultaneously examine whether the challenges being experienced by leaders covary with leader characteristics. Overall, we identify nine central leader challenges. Results indicate that certain leader challenges are more prevalent depending on the leader’s characteristics (e.g., gender), and that two challenges, Daily Management Activities and Communication, were significantly related to boss’ ratings of performance. We also highlight additional applications of this technique to aid leadership researchers who wish to inductively derive meaning from large amounts of unstructured texts.