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

The Benefits of Disclosing Internal Control Weaknesses: Evidence from Taiwanese Banks 200 4.4.4 Classification of ICWs Theoretically, not all ICW disclosures imply the same severity of weakness. However, regulators in Taiwan do not require banks to perform such classifications; therefore, in practice, banks do not classify their ICW disclosures. This study classifies the ICWs by their types and materiality and re-perform the analysis. However, we delete some ICW content due to data availability. By identifying the ICW content which can be collected, seven, among 28, ICWs are classified as the severe ICWs ( SEVICW ). Based on the 28 observations that show non-zero ICWs, this study explains that the market value of equity using SEVICW , book values, earnings, other control variables, and inverse Mills ratio obtained in the first stage of Heckman procedure. The coefficient of SEVICW is found to be insignificant ( p = 0.852), indicating that disclosing less and more severe ICWs are not valued differently by investors. Such result is consistent with the argument that disclosing non-zero ICWs reveals that the firm is diligent in identifying ICWs, and ex ante it is not expected that disclosing less and more severe ICWs is related to different levels of diligence. Though, due to the small size, i.e., 28 observations, this study suggests that caution is necessary when interpreting such result. It would be more appropriate to re-examine whether the severity of disclosed ICWs is related to varying valuation in the future when the sample size that discloses non-zero ICWs substantially increases. 4.4.5 Other Methods to Mitigate the Endogeneity Problem This study follows some relatively recent studies (e.g., Lennox et al., 2012) and applies propensity score matching to address the endogeneity problem of reporting ICWs. First, the propensity score on the basis of Model (1) is estimated, and then a bank reporting non-zero ICWs is matched with a bank reporting zero ICWs without replacement. After 1:1 matching, the differences of variable means between the treated group and control group become insignificant, reflecting the quality of matching. A significant difference of 0.078 is found in the means for ICW ( p = 0.018). Thus, once the bank characteristics are balanced between two groups, the treatment effect of ICW correlates positively with the bank value. In other words, the main findings are unaffected by adopting the propensity score matching method. In order to alleviate the common endogeneity problems, the approach constructed by Larcker et al. (2007), and adopted by Laksmana (2008) and Johnstone, Li, and Rupley (2011) is also applied. The procedures are summarized as follows. First, the ICW on several determinants (model (1)) is regressed, and the residuals are obtained. The residuals from the regression indicate whether a firm’s ICW is above or below a benchmark value.