Estimating accounting errors in audit sampling: Extensions and empirical tests of a decomposition approach

Abstract

A decomposition approach was presented by Chan (1988) to estimate accounting errors based on a ratio estimator in audit sampling. That approach decomposed the tainting distribution into several distinct components according to the characteristics of typical accounting populations, and modeled each component separately. A simulation procedure was then used to combine the probabilistic components to determine the error bounds. A deficiency with this approach is its reliance on the central limit theorem (CLT) to model an important component of the tainting distribution when the error rate is low. This paper proposes two alternatives, the chi-square and the exponential distributions to replace the use of the central limit theorem in the model. Empirical tests confirm that the proposed alternatives improve on the overall reliability of the original CLT method for low-error-rate populations. For populations with high error rates of 10 percent or more, however, the performance of all three methods is similar.