Audit Market Measures in Audit Pricing Studies: The Issue of Mechanical Correlation

Abstract

Mechanical correlation bias is inherent in audit pricing studies when independent variables (X) are derived from firm level audit fees (Y). Such variables are endogenous by construction leading to biased estimates, since (mechanically) X determines Y and Y determines X. After reviewing the extant auditing/accounting literature where mechanical correlation obtains we employ mathematical derivations and simulations to quantify the bias associated with a range of mechanically correlated market competition and industry specialist variables. Since auditor market competition variables are important to regulators and antitrust authorities, we analyze the mechanical correlation issue with regard to an extant study which introduces a novel measure of audit market competition (derived from audit fees). The study provides evidence that smaller incumbent auditors are pressured into offering lower fees when competing against a large local audit firm. However, when the current client’s audit fee is ‘decoupled’ from this new competition measure to mitigate bias, it is statistically insignificant in our multivariate regression analysis. Additionally, we employ auditee sales and total assets to construct proxies for competition variables (which are not mechanically correlated) and find them to be statistically insignificant. We conclude with suggestions of how to address the issue of mechanical correlation in future studies.