Order effects in auditors' internal control judgments: Belief perseverance versus the contrast effect.

Abstract

Prior research suggests that beliefs are affected by the order in which information is processed. However, the empirical evidence with respect to this "order effect" is itself anomalous in that the direction of the bias is inconsistent across studies. This study examines and attempts to reconcile the theory of belief perseverance, which predicts primacy, and the contrast-effect theory, which predicts recency. An "integrated" theory of belief revision is presented which proposes that belief perseverance (the contrast effect) is an increasing (decreasing) function of confidence in beliefs. The theory is formalized by expanding Hogarth and Einhorn's (1992) belief-adjustment model to include the proposed effect of confidence on belief revision. The model's predictions were tested in a field experiment in which subjects received a series of information regarding a hypothetical audit client's internal controls and were asked to assess the likelihood that controls would prevent or detect material misstatement. Half of the information in each series was "positive" (describing internal control strengths), while the other half was "negative" (describing internal control weaknesses). The order of information presentation (positive-negative versus negative-positive) was manipulated between subjects. Confidence was manipulated by varying two factors: task experience and amount of information. The experiment was administered to 50 undergraduate auditing students at The University of Arizona and 85 experienced auditors from two of the Big-Six accounting firms. Half of the subjects in each experience-level group received a "short" series of information, while the other half received a "long" series. Confidence assessments were elicited from subjects in each of the four (experience-level/amount-of-information) subject groups and used to form predictions about order-effect differences between groups. The model's predictions were supported for inexperienced subjects but not for experienced subjects, suggesting that confidence affects the direction of order effects only when task experience is low. Furthermore, when order effects were present for experienced-subject groups, they were in the direction of recency. This suggests that increased task experience may lead to a decrease in belief-perseverance proneness.