Financial fraud detection based on the part-of-speech features of textual risk disclosures in financial reports

Abstract

The textual risk disclosures in the annual financial reports, which discuss the companies’ potential risks in the future, are rarely considered in financial fraud detection. The purpose of this study is to detect financial fraud by incorporating the textual risk disclosures. To quantify the linguistic features of textual risk disclosures, we analyze the part-of-speech (POS) for each word in the text and measure the percentage of different types of POS words. Based on the textual risk disclosures of 8999 firm-year financial reports for U.S. energy companies from 2006 to 2019, the empirical results corroborate that the POS features can provide significant detective ability for financial fraud, and can also improve the financial fraud detection performance based on commonly used financial variables. This study is helpful for investors and regulators to understand the role of textual risk disclosures in financial reports and provide theoretical guidance for the integration of textual information in financial fraud detection.