ABSTRACT
Introduction Auditing is a standard and essential tool for assessing the validity and reliability of information and processes. Three of the most common forms of audit are financial, operational, and compliance. Financial audits are used to verify the accuracy of financial statements of governments and businesses. Operational audits are employed to assess managerial performance through an analysis of the effectiveness and efficiency of the operational structure, internal control procedures, and processes. Compliance audits are used to evaluate whether, and to what extent, policies, procedures, and other requirements for individuals, businesses, or organizations are being met. Compliance audits are frequently conducted by governments. Examples include examinations of tax returns; audits to assess compliance with regulatory policies, such as environmental regulations; and audits to evaluate whether reporting, spending, and other requirements are being met with respect to government-funded programs. A common feature of these various forms of audit is that they normally seek only to provide reasonable assurance. Due to practical constraints, it is often infeasible to exhaustively examine every detail or aspect of an operation, system, or report. Hence, audits normally rely on sampling and testing, either at random, or in areas deemed to be of greatest risk for substantial non-compliance with reporting, procedural, or other requirements. Moreover, even when an issue or process is evaluated, there is often potential for imperfect detection of non-compliance. For example, in tax audits, examiners are not always successful in uncovering certain forms of income that have been understated. Thus, audit findings are frequently subject not only to sampling errors, but also to errors in detection. In this chapter, we introduce some econometric methods for controlling for such errors when analyzing the results of audits, and we apply these methods to a sample of individual income tax audit results to develop estimates of detected and undetected tax non-compliance. Our approach is based on the detection controlled methodology introduced by Feinstein (1990, 1991), which we have adapted to account for the multi-stage nature of the tax return examination process.
