An Investigation of Auditor Decision Processes in the Evaluation of Internal Controls and Audit Scope Decisions 1. Introduction Over the past decade numerous studies have focused on auditors’ judgments. With few exceptions these studies have found that when auditors are given the same task, they frequently make different decisions. ‘ This result has been obtained for relatively simple tasks (e. g. , Joyce ) and relatively complex tasks (e. g. , Reckers and Taylor  and Mock and Turner ).
While consensus of decisions is not * Visiting Assistant Professor and Professor, University of Southern California. We would like to thank William Waller and Paul Watkins for their comments on an early draft of this paper. The paper has also benefited from suggestions made by the referee. Financial support from Peat, Marwick, Mitchell and Co. , the Center for Accounting Research, University of Southern California, and the Graduate School, University of Wisconsin- Madison is gratefully acknowledged. [Accepted for publication August 1982. ‘ Studies indicating lack of agreement include: Joyce , Reckers and Taylor , Mock and Turner , and Gaumnitz et al. . In addition, Ashton  and Libby and Lewis  cite other studies where a lack of agreement has been found. The primary exception to the lack of judgment consensus has been Ashton , which has been replicated by Hamilton and Wright , Ashton and Brown , and Ashton and Kramer . 234 Copyright (C, Institute of Professional Accounting 1983 AUDITOR DECISION PROCESSES 235 he only criterion by which auditors’ judgments should be evaluated, it has been of continuing concern to practicing accountants (e. g. , Elliott [1980, p. 10]) and accounting researchers who continue to assess it in various settings. This concern is justified since consensus can be viewed as one of several necessary, although not sufficient, conditions of expert judgment (Einhorn ) and since a lack of consensus may result in excessive audit costs (see Joyce [1976, p. 31], Mock and Turner , and Elliott [1980, p. 10]).
What is surprising, however, is that there is virtually no evidence which can explain why a lack of consensus is often observed. 2 The research reported here was aimed at this question, using verbal protocol analysis to determine how four experienced auditors made different internal control evaluations and audit scope decisions in a relatively complex and comprehensive experimental task developed by Mock and Turner . The research is a follow-up to the experimental phase of the Mock and Turner study where approximately 200 auditors performed the same experimental task.
Mock and Turner found considerable variability in auditors’ decisions, but their experimental treatments were unable to explain more than 24 percent of the variability. 3 It is hoped that verbal protocol analysis can be used to obtain evidence on the variables which lead to decision divergence by examining the details of subjects’ step-by-step information-processing behavior. When considered with Mock and Turner , this research can be viewed as a multimethod investigation of auditor information-processing behavior.
Such an approach has been recommended as a particularly useful way to employ verbal protocol analysis. (See Einhorn, Kleinmuntz, and Kleinmuntz  and Payne, Braunstein, and Carroll . ) The overall purpose of the protocol methodology used in this research was to provide detailed evidence of information-processing and choice behavior. Such evidence should help explain any observed differences in decisions, thereby providing knowledge of auditor behavior not available from preceding experimental studies.
The focus of the analysis was on the identification of the elements of each subject’s “problem space. ” A problem space is a model of an individual’s cognitive representation of a task and is a major construct in Newell and Simon’s  theory of human problem solving. Einhorn and Hogarth [1981, p. 57] have recently emphasized the importance of the problem space in understanding human behavior since different representations can lead to different decisions. Most attempts to explain differences in auditor judgments have concentrated on differences in experience or differences in audit firm association (see Joyce , Hamilton and Wright , and Ashton and Kramer ). In general, these two factors have not provided satisfactory explanations of judgment differences. ‘ Subsequent analysis by Mock and Watkins  using a multivariate approach improved the amount of variance explained. However, the variance explained was less than 50 percent in all cases. 236 JOURNAL OF ACCOUNTING RESEARCH, SPRING 1983
A key issue, however, is how such evidence can lead to improved judgments. To address this issue two types of situations need to be distinguished. The first consists of situations where optimal solutions are available. Here, research can identify suboptimal behavior and detailed descriptive evidence of auditor decision processes can suggest how to improve subsequent decisions. The second type involves situations in which optimal solutions are not evident. 4 This may be a result of task complexity, incompatible objectives, or a number of other factors.
In such cases, the ultimate goal of improving decision making becomes much more difficult to achieve. One approach is to stress consensus in evaluating decisions. While this approach has some merit, there are situations where a consensus cannot be reached, and even if is reached, it may not provide conclusive evidence that a correct decision was made. An alternative approach, which seems to be developing in the research literature, is to gather and evaluate evidence on the components of individuals’ decision behavior (Hogarth and Makridakis  and Einhorn and Hogarth ).
The assumption here is that while it may be difficult to assess the correctness of the final decision outcome, it may be possible to evaluate the decision’s individual components. Both satisficing rules and normative evidence from other fields may provide evidence useful in evaluating individual components. For instance, during information acquisition, selective perception (Egeth ) may lead the auditor to disregard conflicting evidence concerning the quality of internal accounting controls, resulting in overreliance on the control and excess audit risk.
In complex situations, a multitude of other “biases” could also occur (Hogarth and Makridakis ) making it difficult to specify, a priori, which of the relevant variables, biases, and heuristics to test. The type of evidence obtainable in a protocol study could be an important first step in developing knowledge which will result in better focused nomothetic audit research and ultimately in improved auditor judgment.