Full population testing: Applying multidimensional audit data sampling (MADS) to general ledger data auditing

Abstract

Changes to the General Ledger (GL) represent a link between transactional business events from Journal Entries and prepared financial statements. Errors in these very large datasets can result in material misstatements or account misbalance. Unfortunately, a plethora of conditions renders traditional statistical and non-statistical sampling less effective. As a full-population examination procedure, Multidimensional Audit Data Sampling (MADS) mitigates these issues. In conjunction with top practitioners, we utilize a design science approach in applying the full-population MADS methodology to a real dataset of GL account balance changes. Issues such as the effectiveness of internal controls, detection of low-frequency high-risk errors, and earnings management concerns are addressed. This paper demonstrates how vital insights can be gained using MADS. More importantly, this approach also highlights the exact portion of the population that is error-free with respect to the auditors' tests.