Conference presentation at MBC2 2022 Models and Learning for Clustering and Classification 6th International Workshop, Catania.
ISSN/ISBN: Not available at this time. DOI: Not available at this time.
Abstract: There is a rapidly increasing literature on the statistical detection of frauds through Benford’s law (Berger and Hill, 1995), whose applications span over diverse fields such as accounting and finance (Nigrini, 2012), electoral processes (Pericchi and Torres, 2011) and international trade (Cerioli et al. 2019). Checking confor- mance to Benford’s law requires the availability of appropriate test statistics, which can be obtained through different characterizations of the law (Barabesi et al., 2018, 2021, 2022). In this work we address the tricky situation where the fraudster can anticipate the existence of statistical checks based on conformance tests of the Benford hypothesis. Manipulation that makes the data follow Benford’s law precisely is difficult due to administrative and accounting constraints, and also because statistical checks are often based on a battery of models that include Benford’s law as one among several instances. However, it may realistically happen that the Benford-savvy fraudster is able to ensure compliance of the first digit of his numbers to Benford’s law. We then develop new tests tailored to this situation and study their empirical performance, both through simulations and numerical examples taken from concrete scenarios.
Bibtex:
@inproceedings{,
author = {Andrea Cerioli and Lucio Barabesi and Andrea Cerasa and Domenico Perrotta},
title = {Who is afraid of the probability-savvy fraudster?},
year = {2022},
booktitle = {Proceedings of the MBC2 2022 Models and Learning for Clustering and Classification 6th International Workshop},
address = {Catania, Italy},
}
Reference Type: Conference Paper
Subject Area(s): Accounting