Allocating Internal Audit Resources for Fraud Risk Reduction

Abstract

Corporate fraud these days represents a huge cost to our economy. To counter this cost, organizations allocate lots of resources in terms of internal audit. Mostly, these audits are performed at a random sample of observations. This paper provides a methodology to help allocating efforts of internal audit more efficiently. Academic literature concerning fraud detection already concentrated on how data mining techniques can be of value in the fight against fraud. In this paper we discuss the use of a data mining approach to reduce the risk of internal fraud, more precisely by auditing specific observations instead of random sampled observations. Reducing fraud risk comprehends both detection and prevention, bearing close resemblance with the original aim of internal audit and control. The results of using a multivariate latent class clustering algorithm to a case company’s procurement data suggest that applying this technique in a descriptive data mining approach is useful in allocating efficiently audit resources and in assessing the current risk of internal fraud. The same results could not be obtained by applying a univariate analysis.