Identifying accounting control issues from online employee reviews

Abstract

This paper presents and describes an approach to generate innovative labeled datasets that enable automated text classifiers to automatically detect online employee reviews referring to accounting control deficiencies, facilitating supplementary monitoring for auditors and management. Employees, who are on the front lines executing policies and procedures, play a critical role in a firm's control environment. Their feedback provides insights into how controls are functioning. Textual data were collected and manually coded using a structured coding scheme mapped to COSO internal control framework (2013) principles. The dataset is unique in that it provides a new source of data that has not been previously used in internal control research, offering new opportunities for exploring the relationship between employee feedback and control weaknesses, and shedding light on potential improvements in internal control practices. Downstream stakeholders (such as researchers, management, investors, and auditors) can benefit by having rapid, automated means for filtering and prioritizing employee reviews for further investigation, with respect to accounting control issue mentions.