A Hurdle Approach to Modeling Audit Samples with Partial Misstatements

A well-known problem in statistical audit sampling is inferring an upper bound for the population misstatement from a sample with partial misstatements. Typically, auditors use the Stringer bound for this purpose. However, this method is often too conservative and lacks a solid statistical foundation. Despite the development of better alternatives, almost none of these methods have been widely adopted, partly because they are impractical. To address these issues, we propose an intuitive approach that combines Bayesian statistics with hurdle modeling. We show that this approach is reliable, has a stronger statistical foundation than the Stringer bound while also being more efficient, and is more practical than alternative methods. Additionally, it enables auditors to incorporate multiple sources of pre-existing audit information into the statistical analysis in a straightforward manner, thereby further increasing efficiency. We argue that the proposed method is an attractive solution for situations where the Stringer bound creates challenges.