Discussion of Error Rates, Detection Rates, and Payoff Functions in Auditing.

Abstract In this article, the author presents his views on error rates, detection rates and payoff functions in auditing. It has long been recognized that auditing problems differ from single-person statistical decision problems. The actions of the auditee depend upon anticipations of the auditor's actions. With an even mildly rational auditee, these anticipations must bear some systematic relation to the auditor's actions. This ability to influence the auditee's actions and, ultimately, the accounting populations, represents an important departure from single-person statistical decision theory, which studies games against nature. There are at least two ways to include the reported balance, both of which complicate the analysis. One alternative is to view the report as mechanistically transmitted from a reporting system, the error tolerances of which are subject to auditee discretion. Another alternative is to make the reported balance directly controllable by the auditee. Under this approach, the auditee, with full knowledge of the underlying population from which the auditor will sample, chooses a balance to report.