Does consensus imply accuracy in accounting studies of decision making?: Alison Hubbard Ashton, Accounting Review 60 (185) 173–85

Abstract

The empirical relationship between consensus and accuracy is examined in two prediction tasks of interest to accountants-a managerial accounting task and an auditing task. The relationship between these two attributes is particularly important in auditing contexts in which consensus is sometimes accepted as a practical goal. The results of this study indicate a consistent, highly positive relationship between consensus and accuracy. The mean Pearson (Spearman) correlation between the two attributes was .84 (.82) across (1) the two tasks, which featured both continuous and dichotomous outcomes and which were performed by different sets of experts, (2) both individual and pairwise measures of subjects' consensus and accuracy, and (3) both correlational and absolute measures of performance in the managerial accounting task. AN important goal of research in individual decision making is improved decisions. Improvement implies some standard to which actual decisions can be compared. In most research, this standard has been correctness or accuracy. However, a problem in most accounting research contexts is the absence of an objective external criterion. The criterion problem exists, for example, in performance evaluation tasks, but is particularly acute for various types of audit judgments, such as internal control evaluation and probability assessment. The response to the criterion problem in a large body of accounting research has been to study consensus, or agreement among decision makers, rather than accuracy. Einhorn [1974] describes consensus as a necessary, but not sufficient, condition for accuracy among a group of expert decision makers. While lack of consensus among a group of experts implies that at least some of the experts are not accurate, strong consensus does not necessarily imply accuracy. Einhorn argues, however, that actions and decisions may have to be based on the imperfect consensus criterion when use of the preferable accuracy criterion is not feasible. In accounting and auditing decision contexts for which accuracy cannot be used, it would be useful to know the empirical relationship between consensus and accuracy. One could then assess the extent to which the two attributes are related in those contexts, and consensus could be evaluated as a surrogate for accuracy. One way to proceed toward this end is I am grateful to Thomas Kida for providing the data on which part of this paper is based and to Vijay Karan and Greg Tully for helpful comments on an earlier draft. Alison Hubbard Ashton is Visiting Associate Professor ofAccounting, University of Alberta, on leave from New York University. Manuscript received October 1983. Revision received April 1984. Accepted June 1984.