Auditor Response to Estimated Misstatement Risk: A Machine Learning Approach
研究了用机器学习估计的错报风险是否接近审计师的实际风险评估，发现审计师会对此风险定价，且大审计师更敏感，但审计质量与风险的关系在不同审计师间差异不大。
SYNOPSIS We investigate whether misstatement risk estimated using advanced machine learning techniques—hereafter, estimated misstatement risk (EMR)—approximates auditors' risk assessments in practice. We find that auditors price EMR and auditor turnover is more likely to occur when EMR increases, indicating that EMR is associated with auditors' risk assessment. We also find evidence that EMR is positively and significantly associated with audit fees and auditor switching for companies with Big N auditors but not for other companies, suggesting that Big N auditors are more responsive to risks captured by EMR. Additional analyses reveal that companies switching auditors when EMR increases are more likely to engage non-Big N auditors. Surprisingly, we find little evidence that the association between audit quality and EMR differs by auditor type. Our findings are consistent with the notion that the documented association between audit fees and EMR primarily reflects a risk premium in our setting.