臺大管理論叢 NTU Management Review VOL.30 NO.2

The Benefits of Disclosing Internal Control Weaknesses: Evidence from Taiwanese Banks 194 4.3 Effects of ICW Disclosures on Bank Value This study first estimates Model (1) to address the potential endogeneity problem involved in reporting ICW. Table 5 shows the empirical results. The pseudo- R 2 is 39%, indicating that ICW can be largely explained by these determinants. The coefficients of BDPLEDGE , BLOCKHD , and GOVHD are all be significantly positive ( p < 0.1). These results indicate that banks with more shares pledged by board directors, banks with more shares held by blockholders, and banks with government ownership tend to disclose non- zero ICWs. In contrast, the coefficients of EXGROWTH and BDMGER are significantly negative ( p < 0.1), indicating that banks with extreme growth and banks with managers on the board are less likely to disclose non-zero ICWs. 10,11 Table 6 shows the results of the second stage. The R 2 is 60%, indicating that MV can be fairly explained by the independent variables in Model (2). The coefficient of ICW is positively significant ( p < 0.05). 12 Regarding the book value ( BV ) and net income ( NI ), the associated coefficients are positively significant (both p < 0.1), as generally documented in the value-relevance literature. 10 Two variables may also be related to the ICW disclosures of banks. HOLDING defines whether the bank belongs to a financial holding company. BIGN indicates whether the bank is audited by a large accounting firm. In Taiwan, large accounting firms are affiliated with Arthur Andersen, Deloitte Touche Tohmatsu, Ernst & Young, KPMG, and PricewaterhouseCoopers. However, empirically, the two variables do not add incremental explanation and are automatically dropped by the statistical software. Thus, HOLDING and BIGN are omitted in Model (1). 11 The VIF of the variables in our first stage are all smaller than 10, which indicates there is no multicollinearity problem. However, to be conservative, we eliminate LEGALHD because it has high correlation coefficients (approximately 0.5) with three other variables. The main results do not change. 12 Although we add government ownership ( GOVHD ) to our model (1), it is still worth noting the effect of government ownership on our main tests. Hence, we perform the empirical analyses by separating our observations into two groups: government-owned and non-government-owned banks. We find that for non-government-owned banks, the coefficients of ICW disclosure measures are positive and significant, similar to our main findings. For government-owned banks, the coefficients of ICW disclosure measures become less significant. However, the sample size of government-owned banks is 31, which is relatively small, and caution should therefore be taken when interpreting this result.