Contrary to intuition, first digits of randomly selected data are not uniformly distributed but follow a logarithmically declining pattern, known as Benford’s law. This law is increasingly used as a ‘doping check’ for detecting fraudulent data in business and administration. Benford’s law also applies to regression coefficients and standard errors in empirical economics. This article reviews Benford’s law and examines its potential as an indicator of fraud in economic research. Evidence from a sample of recently published articles shows that a surprisingly large proportion of first digits, but not of second digits, contradicts Benford’s law.
© 2019 by Walter de Gruyter Berlin/Boston