Auditors' analytical review decision processes: The impact of hypothesis set size on decision accuracy and information search
Auditors may initially generate and test a small or a large number of hypotheses when performing various tasks. However, little is known about the relative efficiencies of testing a different number of hypotheses. This study uses an analytical review framework to examine the impact of testing different hypothesis set sizes on auditors' decision accuracy and information search.
Professional auditors were given information which indicated a fluctuation in a client's financial statements. The auditors were divided into four groups and asked to either generate a specific number of hypotheses (one, three, or six) that may explain the deviations, or to test any number desired (no-restriction group). They then analyzed a computerized detailed information set which provided data that could rule out all but a specific error seeded in the financial statements. The auditors either decided that one of the hypothesized causes explain the deviation, or continued generating and testing additional hypotheses. The specific variables examined were the time taken to perform the task, decision accuracy, and the amount and type of information searched.
The results indicated that auditors testing three hypotheses at a time spent less time on the decision task than subjects in the other groups. In addition, the three hypotheses group was as accurate as the one hypothesis group, and more accurate than the six hypotheses and the no-restriction groups. As compared to the three hypotheses group, the one hypothesis group spent significantly more time on the evaluation of each hypothesis, and on the generation of subsequent hypotheses. To minimize their effort, auditors chose to test the smallest total number of hypotheses in the single hypothesis group, which would likely have reduced their decision accuracy. The six hypotheses group spent more time generating the initial set of hypotheses, and considered the largest total number of hypotheses. However, these auditors spent the same amount of time actually evaluating information as the one and three hypothesis size groups and therefore they may have failed to conclusively test the large hypothesis size. The lack of constraints on the hypothesis set size did not benefit the time efficiency or the accuracy of the no-restriction group.
These results suggest that, in tasks similar to the one investigated here, a moderate hypothesis set size (e.g., three hypotheses size) appears to enhance the time efficiency of hypothesis testing, while modestly improving its accuracy. Implications for audit judgment and decision making research are discussed.