A new framework for selecting variables in fraud detection research

Cressey’s (1953) well-known fraud triangle states that instances of fraud share three common factors, an opportunity to commit fraud, a pressure to commit fraud and a rationalization for committing the fraud that is consistent with the perpetrator’s personal ethics. The fraud triangle is a useful conceptual model for studying and understanding the precursors to fraud. It has been explained in almost all industry and academic education on fraud, as well as being used in academic research and industry standards. Although it is has been widely used, the fraud triangle has been criticized for being inadequate, and additional extensions to the model have been proposed. The original fraud triangle and each of the well-known alternative models are analyzed in order to develop a New Fraud Detection Triangle framework. The new framework includes a new Suspicious Information category to acknowledge that unusual patterns in figures often occur as a consequence of fraud, as opposed to the precursors of fraud measured in the existing fraud triangle. Despite the fact that the selection of explanatory (independent) variables is crucial to developing a fraud detection model, the selection process in prior financial statement fraud detection studies is not standardized and the categories of variables vary between studies. The newly proposed framework can play a role as an overall theory to assist in guiding the selection of variables for future fraud detection research. While developed for financial statement fraud detection, the framework is more broadly applicable to fraud detection in general.