7. DGM (2004) include tax overpayments from the tax return to control for late-year
events such as tax planning that reduce taxes payable; however, we believe that tax fees
paid to auditors represents an incremental contribution to this indirect control for tax
planning.
We conduct a Durbin-Wu-Hausman test to ensure that our tax fee variable
(Tax_Fees) and third-to-fourth-quarter ETR change variable (ETR4_ETR3) are not
endogenous. Consistent with Omer et al. 2006, we use two-stage least squares
regression, regressing Tax_Fees on audit fees (scaled by selling, general, and
administrative expense), auditor tenure, size (natural log of total assets), leverage,
return on assets, nonaudit fees (scaled by total fees), nonaudit fees (scaled by pre-tax
income), and categorical variables for Big 4 auditors and auditor change in the first
stage. We then include the residuals from this regression in the second stage. As in
Omer et al. 2006, the coefficient on the residuals is insignificant; thus the Durbin-Wu-
Hausman test suggests that no endogeneity problem exists.