This paper reports an empirical study of 281 errors requiring financial statement adjustments on 152 audits. It reports, among other things, the audit areas in which the errors occurred, the audit procedure, circumstance or other event that initially signaled that an error had occurred, and the apparent causes of the errors, including whether they appeared to have been caused intentionally or unintentionally. The results suggest that client personnel problems, such as inexperience and insufficient knowledge of accounting, and cut-off or accrual problems, are important causes of errors. The results also suggest that analytical review procedures and discussions with the client signal a large proportion of errors. Various other results are presented, and some potential implications for audit planning are discussed.

The study reported here provides evidence on one assumption that is made routinely in the conduct of an audit -- that the U. S. Postal Service will return to the auditor confirmation requests that are difficult or impossible to deliver. Six hundred confirmation requests were sent to non-existent addressees and/or non-existent street addresses in 95 cities, and the percentage and timing of the returned requests were analyzed. The use of a complete 2x2x2x3 factorial design enabled the impact of four independent variables--size of city, type of recipient, type of postage, and type of "problem" or error--to be assessed for both the percentage of requests returned and the timing of the returns. In this study, 100 percent of the confirmation requests mailed to non-existent names at non-existent addresses were returned as undeliverable by the Postal Service. However, when an existing name (address) was used, some of the requests were delivered, despite the non-existent address (name). Overall, the results suggest that auditors are justified in relying on the U. S. Postal Service to return "problem" confirmation requests.