An investigation of the impact of financial viability model selection on audit costs: logit, multivariate discriminant analysis and artificial neural networks

The purpose of this paper is twofold. Firstly, we provide evidence that relying on Type I, Type II and overall error rates to select a model for analysing the financial health of audit clients can result in greater costs than using our alternative approach. Secondly, we show that auditors who use an artificial neural network (ANN) as a tool to analyse the financial viability of audit clients need to consider the underlying ANN paradigm before developing a model in order to minimise audit costs. Our results show that a categorical learning neural network (CLN) minimises the overall cost associated with the auditor examination of audit client financial health. This ANN outperforms both statistical techniques and other ANN paradigms. Consequently, auditors who wish to minimise the total costs associated with their audits should use a CLN or similar type of ANN when assessing audit client financial health.