Understanding and Detecting Criminal Firms

This dissertation addresses a recent and poorly covered field of knowledge in the accounting literature: criminal firms. We define a “criminal firm” as an apparently legal business controlled by a person connected to an organized criminal association. Criminal firms represent a phenomenon of utmost interest in accounting studies for two main reasons. First, because of their social and economic relevance. Their strategic importance in illegal schemes, the magnitude of the proceeds they manage, and the detrimental consequences they bring to the environment where they operate have alerted a vast range of actors, such as enforcement authorities, policymakers, academics, and legal competitors. Second, they make us look at the role of accounting in our society from a different angle. As the articles in this thesis will prove, criminal financial statements are different from legal financial statements, since the underlying businesses are far from the common market logic. As the economic literature produced a fair coverage of the topic, the accounting literature lagged despite its privileged position coming from its ability to understand an essentially firm-based phenomenon. Accounting is instrumental for criminal owners for a variety of purposes, ranging from covering offenses to laundering money. We delve into the criminal firms’ peculiarity by trying to explain what may drive their internal heterogeneity and, eventually, by exploiting the intrinsic differences between legal and criminal financial statements by implementing a state-of-the-art machine learning detection model. As the title suggests, this thesis follows a logical path that starts with the improvement of our understanding of the phenomenon and ends with the operationalization of academic knowledge into a predictive model that may be used by a wide range of users.