Application of Statistical Methods to Discovery of Anomalies in Accounting Data

• Activity Number: 463 - SPEED: Methodological Advances in Time Series: BandE Speed Session, Part 1
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Eugene Yankovsky* and Ana Yankovsky and Loren Williams
• Companies: EY and Intuitive and EY
• Keywords: anomaly; accounting; auditor; detecting; principal component; discordency

Abstract

The paper compares efficiency of applying several statistical methods to discovery of accounting anomalies in an anonymized data sample. Distinguishing features of accounting data are a) collinearity caused by applying double entry rule that imposes equality of amounts accounted in debit and in credit lags of a transaction, b) sparsity of the data when non-existing records are substituted with zeros in data aggregation for the analysis. The following methods were applied to the transactional records where debit and credit lags of the same account were treated as separate accounts: A. Normal probability charts of projections of observations on leading principal components; B. Gamma probability charts of the sum of squares of the principal components’ residuals. Both methods are applied separately to daily aggregations of the transactions and to the lowest level transactions in clusters of the accounts to minimize effect of zeros entered for null values on relations between the accounts. The method with higher efficiency of detecting likely anomalies can be included into auditors’ test procedures and improve quality of the accounting data.