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

The Benefits of Disclosing Internal Control Weaknesses: Evidence from Taiwanese Banks 186 MV and BV reflect the market value and net assets of bank i at the end of year t , respectively. 6 NI equals the net income of bank i in year t . MV , BV , and NI are deflated by total assets at the end of year t -1. Morever, based on Baele, De Jonghe, and Vander Vennet (2007) and Guerry and Wallmeier (2017), bank size ( SIZE ), degree of diversification ( DIV_ASSET and DIV_REV ), and Z-score ( ZSCORE ) are added as the control variables. Furthermore, Model (2) is clustered by years. The hypothesis is that reporting non-zero ICWs is positively valued by the market; therefore, a positive value for β 1 7 is predicted. 4. Empirical Results 4.1 Sample Selection By correlating the ICW disclosures and bank value, disclosure information, financial and stock price data, and values for the corporate governance variables are required. The ICW disclosures are manually collected from the annual reports of every bank in the sample. Financial data, stock prices, and corporate governance variables are extracted from the Taiwan Economic Journal ( TEJ ) database. In the sample of banks, there are 384 bank-year observations with financial data in the TEJ for the 2001–2014 period. This study excludes 35 observations with incomplete financial or corporate governance variables. Furthermore, 208 observations are excluded because their shares are not publicly traded in the stock market. Thus, the final sample comprises 141 observations from 16 banks. 8 See Table 1 for the sample selection process. 6 The empirical results remain unchanged when we measure the market value at the date of financial statement disclosures. 7 Lennox, Francis, and Wang (2012) indicate that when applying twostage method for the selection models, exclusion resctriction is important. In the first stage model, if any variable does not appear as a separate regressor in the second stage model, it is likely to satisfy the exclusion restriction. In our model (1), several variables are excluded from the model (2). Then, Lennox et al. (2012) point out that if the research imposes no exclusion restrictions, usually the model is more likely to suffer from multicollinearity problems. The mean VIF for our model (2) is 1.54, and hence the concern over imposing no exclusion restrictions might be mitigated to some degree. 8 Most banks in Taiwan are owned by financial holding companies, and these banks are not publicly traded. Hence, there are few public banks. By comparing the average total assets, our sample banks account for approximately 70% of all banks. Similarly, the number of public banks in other Asian countries is also low (Bikker and Haaf, 2002; Shen and Chih, 2005). For example, the number of banks in Korea investigated by Bikker and Haaf (2002) is 21, and the number of observations from Thailand in Shen and Chih (2005) is 64 over a 9-year period, indicating an average of nine banks per year.