Classifying Restatements: An Application of Machine Learning and Textual Analytics

• In Event: SET Learning

Abstract

Restatements are used for evaluating reporting quality, audit quality and for other evaluative purposes. It is crucial to differentiate between restatements due to unintentional errors and those due to intentional misstatements because these different causes have different implications for financial statement preparers, users, auditors, regulators and researchers. Manual content analysis of restatements is tedious, resource consuming and error prone. Using machine learning, we constructed and validated a model that classified restatements as either correcting unintentional error or intentional misstatement. Our method does not require a dictionary of words associated with management intent, is automated, easily replicated and scalable and may be used to classify restatements without Form 8-K Item 4.02 disclosure by management (that reliance should not be placed on previously issued financial statements). This study also demonstrates that the language in restatement disclosures is a useful indicator of reporting and audit quality.

Authors

• Beverly Louise Hayes, University of Guelph

• Jefim Boritz, University of Waterloo