siniša slijepčević, branimir blašković: statistical detection of fraud in the reporting of croatian public companies financial theory and practice 38 (1) 81-96 (2014)
introductionThe usage of statistical methods in the analysis and detection of fraud in financial reporting is becoming widespread and necessary. Perhaps the simplest and best known, but still effective test is based on Benford's law. Newcomb (1881) and later Benford (1938) noted that in a sufficiently large collection of numerical data expressed in a decimal form, the distribution of occurrences of first digits is not uniform (we explain Benford's distribution in more detail in section 2). Many authors, including Carslaw (1988), Guan et al. (2006), Kinnunen and Koskela (2003), Nigrini (2005), Niskanen and Keloharju (2000), Skousen et al. (2004), Thomas (1989) and Van Caneghem (2002, 2004 investigated the applicability of this fact in accounting and auditing, and specifically in the detection of "cosmetic earning management". It has been shown that the distribution of first digits in financial reports normally complies with Benford's law. If, however, there have been a-posteriori "cosmetic interventions", then for both statistical and psychological reasons, the distribution of first digits changes. As a result, the likelihood of "cosmetic earning management" can be statistically verified, and at least theoretically, the reliability of a specific financial report can be quantified.