Statistical Sampling in Auditing: An Auditor's Viewpoint

Abstract

The use of statistical sampling techniques to assist the auditor in arriving at his audit opinion poses a number of statistical problems. Accounting populations are frequently very highly skewed and the error rates which the auditor is seeking to detect are often extremely low. This paper seeks to review the various techniques currently being used by auditors and to consider the practical problems involved in relating statistical sampling to the auditor's objectives. Over the years many articles and books on applying statistical techniques to auditing procedures have been published. Initially, there was little interest within the auditing profession as statistical techniques were generally difficult to apply manually and were often insufficiently related to audit objectives. The former objection began to dissipate in the early 1960s with the introduction of computerized audit software for the processing of accounting information. A major step forward came in 1963 with the publication of Arkin's Handbook of Sampling for Auditinzg and Accounting, a book giving comprehensive and not too technical details of how to apply statistical techniques in auditing situations. The techniques discussed included various methods of attribute and variables The auditor would use an attribute sampling tecnhique when his objective was to determine the occurrence rate of some population characteristic or he would use a variables method when his objective was to estimate the value of the particular population. At about the same time as Arkin's book was published, a major United States' accounting firm introduced an audit sampling plan which was fundamentally different to the sampling plans discussed by Arkin. The great claim of this audit sampling plan was that it used the simple statistical theory of attribute sampling to give a result in terms of value rather than as an error rate. The plan is based on the deceptively simple concept of defining the sampling unit as each individual pound (or dollar) rather than defining the sampling unit as the more usual item or balance. Hence, the term monetary unit sampling. However, the mathematical principles underlying unit sampling were not immediately published and it was some years before the principles were published by Anderson and Teitlebaum (1973).