| This dissertation is a theoretical study on management of audit evidence. An audit is a process of collecting, evaluating and aggregating evidence. This dissertation deals with these three aspects of the audit process. It employs two popular normative techniques, Bayesian and belief-function formalisms, to model uncertainty. Further, mathematical programming techniques are used for optimization. A brief discussion of the four parts of the dissertation follows.;The second part of the dissertation deals with belief revision in auditing. Belief revision among auditors has been the focus of many recent empirical studies. The findings of these studies are interesting and are not in complete compliance with the normative prescriptions of belief revision. In this part of the dissertation, the deviations from the normative prescriptions are analyzed and modelled them by relaxing some axioms of the theories.;In the third part of the dissertation an audit risk model is formulated in the Bayesian framework. First, a network of audit evidence is constructed. The network contains various accounts, objectives and items of evidence, and it also represents various inter-relationships between objectives and accounts. Next, the strength of evidence is propagated through the network using local computations. Finally, aggregation of evidence across different accounts are discussed.;The fourth and final part of the dissertation pertains to collection of evidence. Here the cost of collecting evidence is incorporated into the audit planning model. The cost aspect of the audit has been completely ignored in the auditing literature. It is desirable to obtain sufficient evidence at a minimum cost. The audit planning problem is formulated as a constrained optimization problem. Further, the solution is obtained using stochastic dynamic programming. Also, an alternative technique, one-step-look-ahead, is suggested to ease the computations in complex problems.;The first part of the dissertation deals with the evaluation of the strength of evidence. There are no clear guidelines, in the auditing literature, for classifying and comparing different items of evidence. Further, there is no consistent numerical measure of the strength of evidence in auditing. Here, a numerical measure of the strength of evidence and the related concepts are discussed under both the Bayesian and the belief-function formalisms. |
