Does Systematic Selection Lead to Unreliable Risk Assessments in Monetary-Unit Sampling Applications?

Monetary-unit sampling (MUS) applications using systematic selection are evaluated via the use of a statistical function that describes the distributional properties of simple random samples. Because systematic selection produces a significantly smaller set of potential samples, its distributional properties differ from those of simple random selection. Whether these distributional differences lead to unreliable MUS risk assessments is the focus of our study. Our findings indicate that risk assessments of MUS applications using systematic selection exhibit material error at a nontrivial rate. We also find that risk assessment reliability declines as sampling interval width decreases, error tainting magnitudes increase, and errors are increasingly concentrated in population members with larger recorded values. Given the availability of alternative sample selection methods, our findings suggest that auditors should avoid the use of systematic selection in MUS applications.