Continuous tests of details and analytical procedures in continuous auditing

In a Continuous auditing (CA) environment audits are performed on a more frequent and timely basis than traditional auditing. CA is a great leap forward in both audit depth and audit breadth. Thanks to fast advances in information technologies, the implementation of CA has become technologically feasible. Moreover, the recent spate of corporate scandals and related auditing failures drives the demand for better quality audits. New regulations such as Sarbanes-Oxley Act require better corporate internal control and shortened reporting lags. These factors have created an amenable environment for CA development. Especially in the past few years CA has attracted the attention of more and more academic researchers, audit professionals, and software developers. The research on CA has been continuously flourishing. This study extends the prior research by using a real-world case to discuss two essential procedures in CA---tests of details and analytical procedures.; This dissertation consists of two primary parts. First, it discusses how to apply continuous test of details to detect errors in the business processes of the company's procurement cycle. Second, it proposes new auditing protocols, tests various expectation models, and compares anomaly detection performance using aggregated and disaggregated data for the analytical procedures in CA. The online learning protocol and error correction protocol are introduced for expectation models to improve anomaly detection performance. Four different expectation models are analyzed in terms of the prediction accuracy and detection performance, which include Linear Regression Models (LRM), Simultaneous Equation Models (SEM), Subset Vector Autoregressive (SVAR) models and Bayesian Vector Autoregressive (BVAR) models. The results indicate that error correction protocol generally improves the detection performance. Using disaggregated data can lead to better anomaly detection when the entire error concentrates on a single day or a single location. However, the detection performance would deteriorate when error is dispersed evenly on every day or each location.