Behavioral-economics approach to auditors' risk assessments;

              6
A Behavioral-Economics Approach to Auditors' Risk Assessments
William S. Waller
University of Arizona
Strict Bayesians are legitimately challenged to tell us where they get their numbers.
I. Levi
To establish a sound basis for a decision about the audit report, auditors process a variety of information which, in light of prior knowledge, sufficiently limits their uncertainty about misstatements in auditee assertions. As a frame for research on the problem of limiting such uncertainty, three general aspects may be distinguished: normative, descriptive, and prescriptive (Ashton et al. 1988). The normative aspect concerns the manner in which auditors, as unboundedly rational economic agents, should structure and solve the problem. The descriptive aspect concerns the manner in which auditors, as boundedly rational economic agents with limited cognitive capacity, structure and solve the problem in actuality. The prescriptive aspect concerns the ways in which boundedly rational auditors might improve on their current solution to the problem. This distinction is both important and problematic. It is important, because it reflects the dual role of auditing research, which is to understand and improve behavior in practical settings. It is problematic, because it raises difficult issues about how the three aspects relate to each other. Central to this relationship, in auditing as well as other areas of judgment and decision making, is the rational choice model, i.e., expected utility maximization under the subjective or Bayesian interpreta­tion
of probability (Savage 1954).
Applications of the rational choice model in specific economic domains tend to use one of four approaches: positivistic, decision-analytic, heuristic-and-bias, generalized. The positivistic approach adopts the view that understanding an economic agent's behavior requires the assumption that the individual is acting rationally with respect to his or her opportunities, beliefs, and desires (Schoemaker 1982). Economic agents by assumption are Bayesian expected-utility maximizers, and a goal of research is to explain behavior by inferring agents' utilities and subjective probabilities, without regard to the psychological reality of these constructs. Given this assumption, the positivistic approach effectively rules out the possibility of agent error (Einhorn and Hogarth 1981a). Apparent inconsistencies between agent behavior and the rational choice model are handled by re-specifying the model (e.g., adding arguments to the utility function) or the assumed conditions of the setting to which the model is applied (e.g., viewing the setting as strategic rather than parametric), not by relaxing the assumption of rationality. The decision-analytic approach uses the rational choice model prescriptively as a means for structuring a real agent's problem (Raiffa 1968). Unlike the positivistic approach's goal of inferring an agent's utilities and subjective probabilities, the decision-analytic approach seeks to construct these numbers so as to facilitate choice, with no attempt to describe or explain how the agent might otherwise
115