Information systems and internal control: evidence from China

The internal control evaluation reports are typically audited by the same auditor for financial statements. However, if the financial statement auditor has been consulted for the construction of the internal control, firms have to hire a different firm to audit their internal control evaluation reports to maintain auditor independence.

The Guidance on Enterprises’ Internal Control Evaluation [9] which governs firms’ internal evaluation reports specifies the content and steps of internal control evaluation. It classifies weaknesses in internal control into three categories: material weaknesses, significant weaknesses and general deficiencies (in declining order of severity). Although firms are only required to describe material weaknesses, many voluntarily describe other weaknesses too. We include all three categories in our analysis as long as firms provide enough information to determine the nature of the weaknesses. It requires firms to disclose in their internal control evaluation reports at least the following materials: (1) the board of directors’ certification of the truthfulness of internal control reports; (2) the overall evaluation of internal control; (3) the basis for internal control evaluation; (4) the scope of internal control evaluation; (5) the steps and methods of internal control evaluation; (6) weaknesses in internal control; (7) remediation of internal control weaknesses and planned remediation of material internal control weaknesses; and (8) conclusion on the effectiveness of internal control. Furthermore, the Accounting Division of the Ministry of Finance published an article on China Accounting News (the official newspaper by the Ministry of Finance) which provided a sample internal control evaluation report (July 30, 2010).

These firms may simultaneously disclose NISW.

To test whether our regressions suffer from multicollinearity, we check the variance inflation factors (VIF) in our regressions presented in Tables 4 and 5. None of the VIF is above 2 (a VIF above 5 indicates a multicollinearity problem). Hence multicollinearity is unlikely a concern.

From Eq. (2), we can see that ADJROA = (β₁ + β₆ Seg) × ISW + control variables, where Seg is median-adjusted. In other words, the impact of ISW on ADJROA is not a constant, but a linear function of Seg, β₁ + β₆ (Seg-1), where 1 is the median value of Seg. We can interpret β₁ as the effects of ISW on ADJROA when Seg = 1. For firms with more segments, β₁ + β₆ (Seg-1) becomes negative. When Seg = 2, β₁ + β₆ (Seg-1) = 0.008 − 0.010 × (2 − 1) = −0.002, and when Seg = 3, β₁ + β₆ (Seg-1) = 0.008 − 0.010 × (3 − 1) = −0.012.

Among the 552 observations used in our regressions, there are 515 firm-years with one segment, 4 with two segments, and 33 with three or more segments, respectively.