Hypothesis revision strategies in conducting analytical procedures

A number of accounting tasks require an evaluation of competing hypotheses. For instance, a management accountant considers alternative reasons for a significant standard cost variance while an auditor must determine whether an unexpected fluctuation in a client's account balance was caused by an error, irregularity or a change in economic conditions. In spite of the prevalence of hypothesis evaluation in practice, little is known about the process. In this paper we provide evidence on two issues that could compromise effective and efficient hypothesis evaluation: (1) the revision of beliefs over competing hypotheses given diagnostic evidence; and (2) whether an eliminated hypothesis is Subsequently resuscitated (i.e. reconsidered as plausible). To address these issues, 55 auditors were provided audit test results and were asked to evaluate five potential hypotheses that may have caused a material fluctuation in the gross margin ratio of a client. The findings indicate that, when presented with diagnostic evidence about a target hypothesis, auditors were prone to adjusting the likelihood of the target hypothesis but not revising the likelihoods of the competing hypotheses. This non-Bayesian revision strategy, which we label the ''one-hypothesis syndrome'', appears to represent a trade-off reflecting competing forces of cognitive strain, efficiency and accountability. While our preliminary results indicate high performance (in identifying the actual cause), additional research in less controlled settings is needed to fully evaluate the effect of this trade-off on efficiency and effectiveness. Finally, contrary to anecdotal evidence from an SEC release, auditors were reluctant to eliminate hypotheses and those who did, did not regard their eliminations as permanent. (C) 1997 Elsevier Science Ltd.