Fighting Accounting Fraud through Forensic Analytics

Accounting Fraud is one of the most harmful financial crimes as it often results in massive corporate collapses, commonly silenced by powerful high-status executives and managers. Accounting fraud represents a significant threat to the financial system stability due to the resulting diminishing of the market confidence and trust of regulatory authorities. Its catastrophic consequences expose how vulnerable and unprotected the community is in regards to this matter, since most damage is inflicted to investors, employees, customers and government. Accounting fraud is defined as the calculated misrepresentation of the financial statement information disclosed by a company in order to mislead stakeholders regarding the firm’s true financial position. Different fraudulent tricks can be used to commit accounting fraud, either direct manipulation of financial items or creative methods of accounting, hence the need for non-static regulatory interventions that take into account different fraudulent patterns. Accordingly, this study aims to identify signs of accounting fraud occurrence to be used to, first, identify companies that are more likely to be manipulating financial statement reports, and second, assist the task of examination within the riskier firms by evaluating relevant financial red-flags, as to efficiently recognise irregular accounting malpractices. To achieve this, a thorough forensic data analytic approach is proposed that includes all pertinent steps of a data-driven methodology. First, data collection and preparation is required to present pertinent information related to fraud offences and financial statements. The compiled sample of known fraudulent companies is identified considering all Accounting Series Releases and Accounting and Auditing Enforcement Releases issued by the U.S. Securities and Exchange Commission between 1990 and 2012, procedure that resulted in 1,594 fraud-year observations. Then, an in-depth financial ratio analysis is performed in order to evaluate publicly available financial statement data and to preserve only meaningful predictors of accounting fraud. In particular, two commonly used statistical approaches, including non-parametric hypothesis testing and correlation analysis, are proposed to assess significant differences between corrupted and genuine reports as well as to identify associations between the considered ratios. The selection of a smaller subset of explanatory variables is later reinforced by the implementation of a complete subset logistic regression methodology. Finally, statistical modelling of fraudulent and non-fraudulent instances is performed by implementing several machine learning methods. Classical classifiers are considered first as benchmark frameworks, including logistic regression and discriminant analysis. More complex techniques are implemented next based on decision trees bagging and boosting, including bagged trees, AdaBoost and random forests. In general, it can be said that a clear enhancement in the understanding of the fraud phenomenon is achieved by the implementation of financial ratio analysis, mainly due to the interesting exposure of distinctive characteristics of falsified reporting and the selection of meaningful ratios as predictors of accounting fraud, later validated using a combination of logistic regression models. Interestingly, using only significant explanatory variables leads to similar results obtained when no selection is performed. Furthermore, better performance is accomplished in some cases, which strongly evidences the convenience of employing less but significant information when detecting accounting fraud offences. Moreover, out-of-sample results suggest there is a great potential in detecting falsified accounting records through statistical modelling and analysis of publicly available accounting information. It has been shown good performance of classic models used as benchmark and better performance of more advanced methods, which supports the usefulness of machine learning models as they appropriately meet the criteria of accuracy, interpretability and cost-efficiency required for a successful detection methodology. This study contributes in the improvement of accounting fraud detection in several ways, including the collection of a comprehensive sample of fraud and non-fraud firms concerning all financial industries, an extensive analysis of financial information and significant differences between genuine and fraudulent reporting, selection of relevant predictors of accounting fraud, contingent analytical modelling for better differentiate between non-fraud and fraud cases, and identification of industry-specific indicators of falsified records. The proposed methodology can be easily used by public auditors and regulatory agencies in order to assess the likelihood of accounting fraud and to be adopted in combination with the experience and instinct of experts to lead to better examination of accounting reports. In addition, the proposed methodological framework could be of assistance to many other interested parties, such as investors, creditors, financial and economic analysts, the stock exchange, law firms and to the banking system, amongst others