Exploring the deep neural network model’s potential to estimate abnormal audit fees

Abstract

PurposeThe paper aims to estimate abnormal audit fees more precisely than the traditional audit fee model by applying an artificial intelligence (AI) method.Design/methodology/approachThe AI technique employed in this paper is the deep neural network (DNN) model, which has been successfully applied to a wide variety of decision-making tasks. The authors examine the ability of the classic ordinary least squares (OLS) and the DNN models to describe the effects of abnormal audit fees on accounting quality based on recent research that demonstrates a systematic link between accruals-based earnings management and abnormal audit fees. Thus, the authors seek to imply that their new method provides a more precise estimate of abnormal audit fees.FindingsThe findings indicate that abnormal audit fees projected using the DNN model are substantially more accurate than those estimated using the classic OLS model in terms of their association with earnings management. Specifically, when abnormal audit fees predicted using the DNN rather than the OLS are incorporated in the accruals-based earnings management model, the adjusted R2s are larger. Additionally, the DNN-estimated coefficient of abnormal audit fees is more favorably associated to earnings management than the classic OLS-estimated coefficient. Additionally, the authors demonstrate that the DNN outperforms OLS in terms of explanatory power in a negative discretionary accruals subsample and a Big 4 auditor subsample. Finally, abnormal audit fees projected using the DNN method provide a better explanation for audit hours than those estimated using the OLS model.Originality/valueThis is the first approach that utilized a machine learning technology to estimate abnormal audit fees. Because more precise measurement yields more credible research results, the findings of this study imply a significant advancement in calculating unusually higher audit fees.