Determinants of client‐initiated and auditor‐initiated auditor changes

Abstract

Purpose Prior studies include auditor‐initiated (AI) and client‐initiated (CI) auditor changes together and conduct research as though they were the same. This paper aims to hypothesize that AI and CI changes are driven by different interests and to examine factors that could explain and predict them. Design/methodology/approach Based on the prior literature, the authors classified factors into AI and CI categories a priori, and then developed hypotheses that aligned with the classification. The data source is AuditAnalytics. Findings The univariate and multivariate analyses support the hypothesis that CI and AI separations are driven by different interests. The empirically determined prediction model (a classification and regression trees (CART)‐generated six‐step process), which performed better than other multivariate models, classifies AI and CI auditor separations with an accuracy rate of 68 percent. Research limitations/implications The findings suggest that future research should incorporate the differences between AI and CI separations into their studies. One intriguing issue is the lack of importance of the variable audit fees to total fees in the CART model. Though statistically significant in some of the multivariate analyses, the variable is not important in the CART model. This issue merits further investigation as it could provide important insight into dependency issues that are presumed by current audit regulation. Practical implications The paper has important implications for practice. It offers insight into factors that drive CI and AI auditor separations. The CART model may serve as a decision aid for both auditors and regulators as they contemplate and evaluate auditor change decisions. Originality/value The paper provides original insight into AI and CI auditor changes when considered as distinct issues that are driven by different interests.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 34 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.