Abstract: This paper investigates the issue of accounting manipulation. Specifically, it aims to identify quantitative and qualitative variables that allow the distinction between companies that engage in earnings manipulation and those that do not. To achieve this, the Random Forest methodology is utilized. Based on a balanced sample of 80 companies—half of which have been identified as accounting manipulators and the other half as likely not having manipulated their financial statements—a total of 47 warning signals, both quantitative and qualitative, are analyzed. In this way, the most relevant factors for differentiating between the two groups are determined. The results provide greater insight into the warning signals associated with accounting manipulation. This research may be of significant interest to users of corporate financial information, including executives, shareholders, investors, creditors, competitors, banking analysts, investment analysts, auditors, and regulatory bodies.
Keywords: Accounting makeup, accounting manipulation, financial information, machine learning, Random Forest, predictive model, discriminant variables.
Full Document
