Projecting sample misstatements to audit populations: theoretical, professional, and empirical considerations

This paper examines the auditing decision to isolate or project errors found in sampling an audit population. The topic is interesting from both practical and theoretical viewpoints because prior research suggests that theory and practice are juxtaposed. Prior research has suggested a stringent theoretical norm in which few errors should be isolated and also suggests that, in practice, auditors commonly isolate errors that they perceive to be confined to a well-defined subpopulation. We extend this research by using the professional auditing standards as a backdrop for the study of both theoretical and practical aspects of the projection-isolation decision. As to theory, we show that a biased estimator (i.e., one that excludes errors known not to recur) may be more precise than an unbiased estimator under certain assumptions about sample sizes and the frequencies of errors in accounting populations. Using simulation, we specify conditions under which a biased estimator is more precise than an unbiased one. As to practice, we show that auditors rely on containment information when choosing to project, and we use verbal reports to develop a description of auditors' strategies for using containment information. Considering both the theoretical and practical aspects in light of professional standards, we conclude that, under specific conditions, isolation of sample misstatements is appropriate. Further investigation of the topic appears warranted.