Simulation Based Inference for Auditing

Keywords: Auditing; Bayesian inference; rare errors; simulation

Abstract: In carrying out a financial audit, the auditor attempts to assess the size of the errors in the account of an organisation. An error occurs when the organisation records or makes a payment that is incorrect in some way, for example, in the context of government welfare provision, an error is made when a payment is made to a person who does not qualify for such a payment. In most situations the number of transactions carried out by the organisation is too large for the auditor to look at them all, and so a sample of the transactions is audited. This sample is then used to make inferences about the population of transactions and the associated financial errors in the recorded amounts. To describe the correctness of the whole account, inference statements are then required for the sum of the transaction errors, that is, the total financial error in the account.

In this paper we describe some of the modelling problems associated with audit populations, and give possible solutions. The nature of the models proposed make computational issues very important, and we apply some of the recently developed simulation techniques, such as MCMC, to make inferences from our models.