A Multilayered Perceptron Approach to the Prediction of the SEC's Investigation Targets

Abstract

This paper predates the Sarbanes-Oxley Act of 2002 (Section 704: Study of Enforcement Actions) by six years. We used seven red flags which are composed of four financial red flags and three non-financial turnover red flags in order to predict the targets of the SEC's investigation of fraudulent financial reporting. The red flags are computed over 70 firms spread among various industrial sectors, and form the base data that is used used for developing the computational prediction model. Multilayered perceptron (MLP) computation of this data was able to predict the targets of the SEC investigation with an average of 88% accuracy in the cross validation test.

Another interesting finding is that the non-financial data such as the frequency of turnovers of CEOs, CFOs, and auditors of a firm can be more informative than the financial data alone in predicting which firms are likely to engage in fraudulent financial reporting. However both types of information are necessary to attain high levels of prediction accuracy.

Keywords: SEC, Financial Reporting Fraud, Red Flags, Sarbanes Oxley Act of 2002, Financial Engineering, Neural Networks, Political Economy

JEL Classification: C45, G18, G28, G38, K22, K23, L51, M41, P16