Order and level effects of negative evidence regarding internal accounting controls: An examination of alternative models of auditor judgment

Audit risk is the risk that an auditor may unknowingly fail to modify appropriately the auditor's report on financial statements that contain a material misstatement. However, because an indisputable value for audit risk cannot be determined, the auditor must gather evidence about the financial statements and, based on his or her professional judgment, develop an estimate of true audit risk.; Because the estimation of audit risk depends on an auditor's professional judgment, it is of interest to describe how auditors evaluate and aggregate audit evidence. A number of models have been proposed to accomplish this. Although each model to some degree may be descriptive of the judgment process, the underlying basis of each model is quite different. Depending on the model, the auditor may aggregate evidence differently and make different evaluations. Therefore, an estimate of audit risk calculated using one model may be affected by influences to which other models are not sensitive.; This study examines in a quasi-experimental setting three models descriptive of the audit judgment process: the Audit Risk model (ARM), the Dempster-Shafer Belief Functions model (DSBF), and the Hogarth and Einhorn Belief Adjustment model (HEBA). Specifically, the sensitivity of each model to differences in the order in which internal control information is obtained, and the effects of different levels of change in internal control effectiveness are examined. Additional analysis examines the degree to which each model is descriptive of the judgment process and a comparison of models is made.; The responses of eighty-five subjects indicate that the HEBA model is sensitive to the order in which evidence is obtained whereas the ARM and the DSBF models are not. None of the models displayed sensitivity to different levels of change in control effectivness. For the ARM and the HEBA models, the data indicates a linear relationship between subjects' response and corresponding values predicted by the model. While the mean level of estimated audit risk was significantly different for the three models, the percentage difference between subjects' responses and corresponding values predicted by each model were not significantly different.