ABSTRACT
The (Newcomb-)Benford Law has been widely used to detect fraud in data from accounting and finance, or in economic, survey and scientific data. Many empirical studies rely on the outcomes of two particular statistical tests. Our power investigation shows that these tests are weak in terms of power under specific fraudulent pattern. Much more powerful criteria are identified, and in particular, a simple, one-sided mean test is recommended.
Acknowledgments
The authors thank an anonymous referee and Jan Reitz for helpful comments on an earlier version. They are particularly grateful to Karl-Heinz Tödter for comments that improved the paper a lot.
Disclosure statement
No potential conflict of interest was reported by the authors.