[A between-subjects experiment was used to examine the effects of a decision aid in the AICPA's Audit Sampling Audit Guide on the magnitude and variability of auditor sample size judgments. Audit seniors were given background case information for a hypothetical audit task and were randomly assigned to one of three experimental groups: (1) an intuitive judgment group, (2) a decision aid group who calculated sample sizes using the AICPA Guide formula, or (3) a group who provided only the formula parameters from which the researchers later computed implied sample sizes. It was hypothesized that the sample sizes of group 2 would be larger than those of group 1 (due to insensitivity to power considerations), and that the implied sample sizes for group 3 would exceed those of group 2 (due to a tendency for auditors to "work backwards" toward an intuitive sample size). Both of these hypotheses were supported by the data. However, a test for differences in variability indicated that the decision aid led to a greater degree of inconsistency in sample size judgments. Results were consistent over two levels of internal control.]

[Auditing has been characterized as a sequential process of obtaining and evaluating evidence (Gibbins 1984). During this process, auditors continually update their beliefs about the audit assertion being examined. Recently, Hogarth and Einhorn (1989) have posited a belief-adjustment model for updating beliefs. Based on a sequential anchoring and adjustment strategy, the model has important implications for auditors concerning the effects of the order in which evidence is evaluated. The model predicts that the order in which evidence is received has no effect on the belief-revision process when the evidence is consistent (either all positive or all negative). However, when the evidence is mixed (positive and negative), the model predicts that a recency effect will occur. Ashton and Ashton (1988) tested an earlier version of this model with experienced auditors in "simplified, well-defined settings" and found results consistent with the model's predictions. They suggested that their findings should be tested in more realistic audit contexts. Hogarth and Einhorn (1989) have suggested that findings of recency or primacy may be influenced by task complexity. Therefore, the current study tests the predictions of the model using content-rich audit scenarios. Four experiments were conducted involving 251 experienced auditors. The first experiment tested for a lack of order effect using consistent positive evidence, and the second experiment tested for a lack of order effect using consistent negative evidence. The third and fourth experiments tested for recency effects using mixed evidence with either two or four pieces of evidence. The results of the first two experiments indicated that order of evidence was not significant. Order of evidence was significant in experiment 3 in the step-by-step response condition but not in the end-of-sequence response condition. In experiment 4, order of evidence was significant in both response conditions. Therefore, the results of this empirical work support the predictions of the model, thus reinforcing the findings of Ashton and Ashton (1988).]

Tests the predictions of a descriptive model of sequential belief revision using content-rich audit scenarios. Characteristics of auditing as a sequential process; Prediction of the model in which an auditor obtains and evaluates evidence; Testing of lack of order effect.

This paper presents the results of an experiment that examined five hypotheses related to the information search and evidence evaluation that takes place as part of an auditor's sequential decision process. The results indicate the following. (1) When given a choice, auditors examined the most reliable type of evidence. (2) Auditors who stopped at intermediate stages in the sequential task had, on average, more extreme assessments of the outcome, although there were differential effects depending on whether the auditors based their decisions on positive or negative evidence. (3) Changes in the direction of the auditor's judgments were consistent with the signal in the audit evidence. (4) Negative evidence, in general, led to greater judgment revisions than positive evidence, and the most reliable negative evidence led to the greatest revisions. (5) Auditors searched for corroborating evidence, but their evaluation of that evidence did not impact the assessment of evidence received earlier in the sequential task. Résumé. Les auteurs présentent les résultats d'une expérience portant sur l'analyse de cinq hypothèses reliées à la recherche d'information et à l'évaluation de l'information probante qui s'inscrivent dans le processus de décision séquentiel du vérificateur. Les résultats obtenus indiquent ce qui suit: 1) lorsqu'ils ont le choix, les vérificateurs examinent l'information probante la plus fiable, 2) les vérificateurs qui s'arrêtent à des étapes intermédiaires de la tâche séquentielle produisent, en moyenne, des évaluations plus extrêmes des résultats, bien que l'on observe des incidences marginales selon que les vérificateurs ont fondé leurs décisions sur des informations probantes expresses ou tacites, 3) les modifications dans l'orientation des jugements posés par le verificateur sont conformes aux indications livrées par l'information probante, 4) l'information probante tacite mène, en général, à une révision plus importante des jugements que l'information expresse, et l'information tacite la plus fiable mène à la révision la plus approfondie, et 5) les vérificateurs cherchent à obtenir des preuves de corroboration mais leur évaluation de cette information probante n'a pas d'incidence sur lévaluation de l'information probante obtenue aux étapes antérieures de la tâche séquentielle.

A between-subjects experiment was used to examine the effects of a decision aid in the AICPA's Audit Sampling Audit Guide on the magnitude and variability of auditor sample size judgments. Audit seniors were given background case information for a hypothetical audit task and were randomly assigned to one of three experimental groups: (1) an intuitive judgment group, (2) a decision aid group who calculated sample sizes using the AICPA Guide formula, or (3) a group who provided only the formula parameters from which the researchers later computed implied sample sizes. It was hypothesized that the sample sizes of group 2 would be larger than those of group 1 (due to insensitivity to power considerations), and that the implied sample sizes for group 3 would exceed those of group 2 (due to a tendency for auditors to ‘work backwards’ toward an intuitive sample size). Both of these hypotheses were supported by the data. However, a test for differences in variability indicated that the decision aid led to a greater degree of inconsistency in sample size judgments. Results were consistent over two levels of internal control.