Predicting Voluntary Auditor Change Using the Water Cycle Algorithm

Independent auditing plays a fundamental role in improving users' decision-making and market efficiency through validating financial reports. The quality of the work of these auditors depends on several factors, one of the most important of which is their independence. Since the phenomenon of auditor change is related to independence, this phenomenon, which is considered one of the vital issues of every company, should be carefully examined. The phenomenon of auditor change creates a break between auditors and the employer. Predicting the continuation or termination of the relationship between the auditor and the client in the coming years is one of the challenging issues in the field of auditing. In this regard, the purpose of this research is to predict voluntary auditor change (non-continuation of the relationship with the current client) using a meta-heuristic algorithm (Water Cycle Algorithm, WCA) and to compare the results with the logistic regression method.

The statistical sample is 185 companies listed on the Tehran Stock Exchange, selected by the systematic elimination method from 2017 to 2023, and their information was collected. Years with mandatory auditor switching are excluded. Then, Excel and Matlab software were used for implementing methods and predicting auditor change. Thirteen financial and non-financial variables that were extracted from the literature were used to predict auditor change in this research. These independent variables include: current ratio, working capital, debt ratio, asset ratio, return on assets, earnings quality, firm size, audit firm size, auditor opinion type, management change, separation between CEO and the chair of the board of directors, accounting conservatism, and firm competition power. Moreover, using the confusion matrix, which includes 4 evaluation criteria: accuracy, precision, sensitivity, and specificity, the results obtained from the Water Cycle Algorithm were compared with the results of logistic regression, a prominent method for forecasting binary variables such as auditor change. Additionally, these algorithms were run 10 times to ensure the reliability of the results.

In almost all implementations and based on all 4 performance evaluation criteria, the Water Cycle Algorithm is more suitable than logistic regression for predicting auditor change. In general, the criteria of accuracy, precision, sensitivity, and specificity in predicting the change of auditor using the Water Cycle Algorithm were 89%, 75%, 2%, and 99.9%, respectively. In comparison, the logistic regression criteria were 67%, 74%, 1%, and 99.8%, respectively

The Water Cycle Algorithm can be useful for predicting voluntary auditor changes by users and auditing firms. Audit institutions can use this tool to predict the continuation or non-continuation of their relationship with clients in the coming years and better plan to maximize profitability. Audit clients and companies can also use this tool to forecast future relationships with auditors and plan and schedule more efficiently for selecting the next auditor.