Predicting Auditor Dismissals with Shareholder Votes on Auditor Ratification Using Data Mining Tools
Abstract
In this paper, we revisit the association between shareholder voting on auditor ratification and subsequent auditor changes using data mining approaches as compared to the traditional logit model. Our logit model shows that shareholder voting on auditor ratification is not associated with subsequent auditor dismissals. Although the logit model is widely used to study the relationship between dichotomous observations and influencing factors, machine learning models simulate the structure of the human brain and generate predictions through training. Our data mining tools show that the Support Vector Machine (SVM) model shows the best Area Under ROC (AUC) curve, considering AUC is particularly useful in handling imbalanced classifications. In terms of predictive accuracy, the Random Forest (RF) and the Classification and Regression Tree (DT) models outperform the traditional logit model. The predictive accuracy of the RF and DT models is above 95%, with AUC values among the top 3. Overall, the findings of this study suggest that the integration of machine learning may assist researchers in determining the prediction of auditor dismissals using shareholder voting on audit ratification and other financial variables.
Recommended Citation
Cheng, Xiaoyan; Kwak, Wikil; Shi, Yong; and Qu, Yi
(2024)
"Predicting Auditor Dismissals with Shareholder Votes on Auditor Ratification Using Data Mining Tools,"
Journal of Business, Industry, and Economics: Vol. 29, Article 5.
Available at:
https://roar.una.edu/jobie/vol29/iss1/5