Human belief revision, a process in which people update their initial beliefs when new pieces of evidence are presented, is ubiquitous in everyday life. Empirical research in psychology shows that humans display order effects in belief revision when evaluating conflicting pieces of evidence in a sequential mode. Auditors are no exception. The order effects phenomenon is especially important to auditors because many judgments in auditing involve sequential processing of evidence.
Rather than making a judgment after all evidence is collected, auditors normally make judgments in a sequential manner as each piece of audit evidence is received. Although the sequential processing of evidence promotes cognitive economy, recent findings in auditing indicate that it can lead to non-normative responses. Consequently, order effects are generally classified as biases and considered as indications of irrational decision-making.
Several prior studies in auditing and accounting have examined the impact of task variables (including 'temporal order' and 'mode of processing') and other situational variables (experience, hypothesis frames, accountability, task complexity, review process) on order effects and auditor judgments using the Belief-Adjustment Model. None of these studies, however, examine the order effects of contrary and supporting evidence with varying strengths.
The objective of this research is to investigate the relation between audit evidence characteristics and the order effect phenomenon. This study investigates the effects of conflicting audit evidence and source reliability within a contradictory evidence paradigm. The research question is:
Do auditors display order effects in their belief revisions when positive and negative (mixed) audit evidence of varying reliability is presented in a sequential mode?
Four independent experiments were conducted to investigate the effects of four source reliability factors (source competence, nature of evidence, source objectivity, source mode) and conflicting audit evidence on auditor belief revision. A 23 mixed factorial design was employed to test the effects of positive audit evidence (high versus low reliability of source), negative audit evidence (high versus low reliability of source), and order of presentation (positive followed by negative and vice versa) on auditor judgment. Eighty-eight auditors from the three of Big Five CPA firms in Fiji participated in the experiment. The data were analysed using a 3-way ANOVA model.
The results show three major findings. First, auditors only display significant order effects when they evaluate conflicting audit evidence from a source with high reliability in a sequential mode. Where either positive or negative audit evidence or both are received from a source with low reliability no significant order effects were observed. Second, the results show that source reliability is an important variable where conflicting audit evidence is evaluated sequentially. Application of a linear model also provides evidence on how auditors weigh conflicting pieces of audit evidence of varying source reliability. The linear model computed weights show that audit evidence from a source of high reliability is weighed more than evidence from a source with low reliability. Third, additional analysis shows that when auditors are presented with a first set of conflicting audit evidence of varying reliability, they pay less attention to source reliability when audit evidence is negative (i.e., without order effects). However, when presented with a second set of conflicting audit evidence with varying reliability, auditors do not weigh the pieces of evidence in the same manner. Order effect biases appear to affect the way auditors weigh conflicting pieces of audit evidence from sources with varying reliability.
The results of this study further improve our understanding of the relation between conflicting audit evidence and source reliability. The results are likely to be helpful in validating and testing theories of evidence. The findings have important implications for audit engagements. For example, the weights computed using a linear model are likely to be useful in setting benchmarks for the validation of cue weights in expert systems. The results of this study have possible implications for regulators of auditing practice and for corporate governance. While the present audit practice statements emphasize source reliability factors, no mention is made of the direction of audit evidence. Regulators may need to provide guidelines for evaluating conflicting audit evidence.