Can Linguistic Predictors Detect Fraudulent Financial Filings?

Abstract

Extensive research has been done on the analytical and empirical examination of financial data in annual reports to detect fraud; however, there is scant research on the analysis of text in annual reports to detect fraud. The basic premise of this research is that there are clues hidden in the text that can be detected to determine the likelihood of fraud. In this research, we examine both the verbal content and the presentation style of the qualitative portion of the annual reports using natural language processing tools and explore linguistic features that distinguish fraudulent annual reports from nonfraudulent annual reports. Our results indicate that employment of linguistic features is an effective means for detecting fraud. We were able to improve the prediction accuracy of our fraud detection model from initial baseline results of 56.75 percent accuracy, using a “bag of words” approach, to 89.51 percent accuracy when we incorporated linguistically motivated features inspired by our informed reasoning and domain knowledge.

Predicting Fraud from Quarterly Conference Calls: A Small-Sample Study of Scripted Language

“…While financial indicators remain a popular and powerful way to detect fraud, new research has suggested that linguistic analysis of the management discussion and analysis (MD&A) section of 10-K or 10-Q financial statements contains meaningful information to market participants. This type of analysis also has the ability to predict fraud (Cecchini et al 2010a;Glancy and Yadav 2011;Goel and Gangolly 2012;Goel et al 2010;Humpherys et al 2011). There is also meaningful information in news coverage about firms (Tetlock et al 2008).…”

“…Prior work in this area of fraud detection has explored the language of financial statements (Cecchini et al 2010a;Goel and Gangolly 2012;Goel et al 2010;Humpherys et al 2011) and the verbal (Larcker and Zakolyukina 2012) and non-verbal (Hobson et al 2012;Mayew and Venkatachalam 2012) behavior of executives in the conference calls. This paper is interested in integrating these two data sources to uncover instances of financial fraud.…”

Text mining to detect indications of fraud in annual reports worldwide

“…The deceptive messages showed less lexical diversity (Zhou et al, 2004a). In the research of Goel et al (2010), the lexical diversity is one of the predictive features for fraud in 10-K annual reports. Fuller et al (2011) applied the number of words and sentences and lexical diversity in their machine learning model to detect deceptive utterances in the statements of persons of interest.…”

“…Several mathematical implementations exist that differ in how they learn the characteristics and construct the model. The Naïve Bayes classifier (NB) and Support Vector Machine (SVM) have been proven successful in text classification tasks in several domains (Cecchini et al, 2010;Conway et al, 2009;Glancy and Yadav, 2011;Goel et al, 2010;He and Veldkamp, 2012;Joachims, 1998;Manning and Schütze, 1999;Metsis et al, 2006;Purda and Skillicorn, 2015). Therefore, this research uses these two types of machine learning approaches to develop text mining models that can detect indications of fraud in the management discussion and analysis section of annual reports.…”

“…Even for this limited data set they were able to assign the majority of the reports to the correct cluster with an accuracy that is well above chance level. Goel et al (2010) exploited the text of the entire 10-K reports for the data set consisting of 405 fraud reports of 126 companies and 622 non fraudulent reports of 622 companies. They extract word counts and several linguistic features such as the average word length, average sentence length and the frequency of proper nouns.…”