275 /304
275
臺大管理論叢
第
28
卷第
1
期
4.2 Logistic-Regression Results
We estimate Model 1 using the subsamples of restatement companies and non-
restatement companies separately, and the results are presented in Table 7. For the
restatement sample, the coefficient on SOX404 is -0.932, with a p-value < 0.046,
indicating that the odds of concluding that internal controls are effective for restatement
companies subjected to SOX 404 decreased by 60.62%, as compared with their
restatement counterparts subjected to SOX 302. It suggests that the enactment of SOX 404
reduces Type II errors of ICFR disclosures and is consistent with our H1a. For the non-
restatement sample, the coefficient on SOX404 is not significant, suggesting that there is
no evidence that the enactment of SOX 404 results in a higher Type I error rate. Taken
together, we conclude that the enactment of SOX 404 reduces Type II errors of ICFR
disclosures without increasing Type I errors.
For the restatement sample, companies with few restructuring charges, without FT, or
with low aggregate loss are more likely to conclude that their internal controls are
effective. For the non-restatement sample, concluding that internal controls are effective is
positively associated with MB ratios, the appointment of a Big-4 auditor, and market
capitalization, and it is negatively associated with PE ratios, restructuring charges, the
incidence of FT, and aggregate loss.