To provide further evidence on the effectiveness of analytical procedures (APs) used in auditing. Computer simulation experiments are used to examine the error detection ability of a set of APs. Two different types of errors are examined and compared on the basis of the Type I and Type II errors they produce. The results of the experiments support earlier performance assessments of APs based upon simulated data. Higher noise levels reduce performance but a more detailed modeling of the process generating the data appears to produce a compensatory increase in performance. Contrary to earlier findings, some annual APs performed better than their related monthly counterparts. Case study and experimental results are better reconciled than in previous studies. The findings are based upon simulated data and deal with two types of error only. The experiments model the data generating process underlying accounting numbers but are simplifications of the real situation. Future research based upon the same approach but using more sophisticated experimental models and dealing with a wider class of errors would be useful. The findings echo earlier recommendations that APs should not be relied upon as lone, substantive testing devices for error and fraud. The simulation experiments use Statistical Activity Cost Theory to generate accounting numbers from specified, underlying stochastic processes. This allows errors to be related to transactions, i.e. the level at which they typically occur, whereas in prior experimental work errors have only been related to accounts.
Boon Law, S. and Willett, R. (2004), "The ability of analytical procedures to signal transaction errors", Managerial Auditing Journal, Vol. 19 No. 7, pp. 869-888. https://doi.org/10.1108/02686900410549402Download as .RIS
Emerald Group Publishing Limited
Copyright © 2004, Emerald Group Publishing Limited