Peer-based approach for analytical procedures

This study extends the existing research in analytical procedures by allowing for learning from contemporaneous information transfers among peer companies. We introduce an approach for selecting peers for each client and perform tests to examine the contribution of peers' information to the performance of analytical procedures. We find that peer data are imperfect substitutions for contemporaneous firm-specific variables when such variables are not in error. However, we observe that contemporaneous peer specific data are especially beneficial when coordinated errors exist in multiple accounts. We demonstrate that when errors are seeded into two contemporaneous accounts, peer models perform better at detecting errors. We also find that fast-changing companies experience inferior prediction and error detection accuracy, and that larger companies experience more accurate prediction, lower Type 11 errors, and higher Type I errors. Additionally, we observe that significant improvements in the performance of analytical procedures are associated with larger clients indicating that auditors of larger companies can potentially benefit more from the use of peer data.