Detection of financial statements fraud using Beneish and Dechow models

Abstract

Fraudulent financial reporting is a big issue not only for investors but also for other stakeholders. This research uses two popular fraud detection models by Beneish (1997, 1999a) and Dechow et al. (2011). The main goal of this paper is to compare the precision of these two models for the prediction of fraud in the financial statements of Iranian companies. Firstly, we try to identify the statistical description related to the first and fourth quartiles of the Beneish and Dechow models. Then, we determine the models’ forecasting capabilities using SPSS software by t-test and variance analysis. We use the sample of 197 companies during the 11-years period from 2009 till 2019. The results indicate that the Beneish model has more precision and less error level in fraud detection in the financial statements than the Dechow model. The general precision of the Beneish model, with 83%, compared to the Dechow model, with general precision of 75%, demonstrates the volume of fraud in the company’s financial statements. According to the statistical results, the prediction precision of the Beneish model, compared to the Dechow model, is more, and its estimation error is less than the latter. Therefore, according to this hypothesis, the Beneish model enjoys a higher detection power in the probability of committing fraud in the financial statements than the Dechow model. Thus, in companies with a previous record of earnings management, there is the probability of committing fraud in the financial statements. It is possible to detect fraud more easily by the Beneish model. The findings of Beneish (1999b) research, Jones et al. (2008), Dechow et al. (2011), and Perols and Lougee (2011) confirm the result obtained from this hypothesis.