Causal Discovery in Strategic Management Research

In recent years, management scholars have developed a keen interest in machine learning as a tool for pattern discovery in empirical research. A drawback of these methods, however, is that they are generally not able to infer causal relationships from data, which poses severe limitations for theory building and testing. In this paper, we introduce a novel class of machine learning methods that are able to overcome this limitation. These causal discovery algorithms leverage testable constraints imposed by the data generating process to infer causal models that are compatible with the observed conditional independence relationships in the data. We discuss the strengths as well as limitations of this approach and present applications of different causal discovery algorithms to research questions in strategic management.