Table 7.

Mechanisms linking non-financial misconduct to audit fees

Variables(1)(2)(3)(4)
LAGLAGLAFLAF
Panel A: Audit effort and complexity channel
LNPEN0.000 (0.454)0.010*** (6.028)
LNPEN × FRGN0.006** (2.133)
LNNUM0.009* (1.671)0.102*** (6.008)
LNNUM × FRGN0.069*** (2.652)
FRGN−0.003 (−0.368)−0.003 (−0.354)0.120*** (5.119)0.118*** (5.056)
Intercept4.335*** (64.700)4.338*** (65.550)10.408*** (51.309)10.460*** (57.172)
Control variablesYesYesYesYes
Year and industry FEYesYesYesYes
N10,23110,23110,23310,233
Adjusted R20.3140.3150.7940.795
Panel B: Engagement risk premium channel
Variables(1)(2)(3)(4)
LAFLAFLAFLAF
LNPEN0.008*** (4.390)0.007*** (3.613)
LNPEN × BIG40.018** (2.033)
LNPEN × AF0.001*** (3.471)
LNNUM0.084*** (3.991)0.078*** (3.469)
LNNUM × BIG40.220** (2.275)
LNNUM × AF0.005*** (2.841)
BIG40.243*** (4.766)0.253*** (4.937)0.214*** (4.318)0.216*** (4.329)
AF0.000 (0.185)0.001 (0.475)
Intercept10.367*** (53.565)10.408*** (58.981)10.498*** (47.632)10.557*** (53.993)
Control variablesYesYesYesYes
Year and industry FEYesYesYesYes
N10,23310,2339,9749,974
Adjusted R20.7940.7950.7950.796
Panel C: Specialist demand and auditor switching
(1)(2)(3)(4)
VariablesToSPToSPLAFLAF
LNPEN0.157*** (3.134)0.011*** (7.569)
LNNUM1.628** (2.322)0.117*** (7.310)
LNPEN × ToSP0.025* (1.739)
LNNUM × ToSP0.292 (1.579)
ToSP−0.099 (−0.744)0.012 (0.109)
Intercept−0.457 (−0.156)−0.223 (−0.074)10.340*** (50.213)10.388*** (55.245)
Control variablesYesYesYesYes
Year and industry FEYesYesYesYes
N30430410,23310,233
Adjusted R20.7960.797
Pseudo R20.6110.602
Note(s):

Panel A: *, ** and *** indicate significance (two-tailed) at 0.10, 0.05 and 0.01 levels, respectively. Standard errors are robust to heteroscedasticity and clustered by firm. LNPEN × FRGN is the mean-centered interaction term measured as (LNPEN – mean value of LNPEN) × (FRGN – mean value of FRGN). LNNUM × FRGN is the mean-centered interaction term measured as (LNNUM– mean value of LNNUM) × (TURNOVER – mean value of FRGN). The control variables used in Columns 1–4 are obtained from Model 1. Panel B: *, ** and *** indicate significance (two-tailed) at 0.10, 0.05 and 0.01 levels, respectively. Standard errors are robust to heteroscedasticity and clustered by firm. LNPEN × BIG4 is the mean-centered interaction term measured as (LNPEN – mean value of LNPEN) × (BIG4 – mean value of BIG4). LNNUM × BIG4 is the mean-centered interaction term measured as (LNNUM– mean value of LNNUM) × (BIG4 – mean value of BIG4). LNPEN × AF is the mean-centered interaction term measured as (LNPEN – mean value of LNPEN) × (AF– mean value of AF). LNNUM × AF is the mean-centered interaction term measured as (LNNUM– mean value of LNNUM) × (AF – mean value of AF). The control variables used in Columns 1–4 are obtained from Model 1. Panel C: *, ** and *** indicate significance (two-tailed) at 0.10, 0.05, and 0.01 levels, respectively. Standard errors are robust to heteroscedasticity and clustered by firm. LNPEN × ToSP is the mean-centered interaction term measured as (LNPEN – mean value of LNPEN) × (ToSP – mean value of ToSP). LNNUM × ToSP is the mean-centered interaction term measured as (LNNUM– mean value of LNNUM) × (ToSP – mean value of ToSP). The control variables used in Columns 1–4 are obtained from Model 1. In Columns 1–2, control variables (i.e. LOSS, GC, MW, RESTATE, FRGN, BIG4, SP, LNTENURE, AUD_CH and CEOEXP) were omitted from the probit model because they perfectly predicted the outcome or lacked sufficient variation, which is necessary to ensure convergence and valid estimation

or Create an Account

Close Modal
Close Modal