Auditors' inherent risk assessments: The relationships among task experience, inferability of conditioning information, second-order uncertainty and extent of testing.

Comparisons between experienced and inexperienced auditors have indicated that differences in knowledge which interact with the environment result in performance differences (Frederick and Libby 1986; Libby and Luft forthcoming). Experienced auditors' recognition of and reaction to variations in relevant conditioning information for inherent risk assessments have important implications for the efficiency and effectiveness of an audit. In addition, previous research concerning subjective probability has indicated that, contrary to the theory's requirements, actual individuals have imprecise degrees of belief (Ellsberg 1961; Gardenfors and Sahlin 1982). This impression in degree of belief results in individuals having second-order uncertainty about first-order probability assessments. This study extends previous research by providing evidence concerning a knowledge effect in the inherent risk assessment task, by considering the effect of the inferability of the conditioning information and by providing exploratory evidence concerning the relationship of first- to second-order uncertainty. Second-order uncertainty, in this context, is an auditor's uncertainty about a first-order inherent risk assessment. Inferability is a dimension of audit evidence which depends on auditors' internal knowledge structures. Their knowledge enables them to infer alternative values from other information to replace values that would be inferred from conditioning information that is unavailable in the current audit evidence set. The current experiment involved responses of 210 experienced and inexperienced auditors from KPMG Peat Marwick who assessed the inherent risk of a continuing manufacturing client. Subjects were assigned to the experienced group if they had experience with five or more inherent risk assessments. Inexperienced subjects had no experience. The data were analyzed as a 2x3 analysis of covariance. The between-subject independent variables were experience and inferable conditioning information. The dependent variable was second-order uncertainty. Inherent risk assessment was a significant covariate. The results indicate that when inferable conditioning information was relatively complete, experienced auditors had significantly lower second-order uncertainty than inexperienced auditors. The interaction of inferability of conditioning information and experience was marginally significant. However, the first-order inherent risk assessments indicate that experienced auditors have higher assessments when second-order uncertainty is low