更早期預警模型：臺灣銀行道德指標的建立及影響

Earlier Warning Model: Establishment and Influence of Moral Hazard Ratio in Taiwan Banking Industry

本文討論影響道德風險如何直接及間接影響銀行績效。本文依一股一權的觀念，計算了二個道德風險指標，第一個以La Porta., Lopez and Shelifer (1990) 與Classens, Djankov and Lang (1999)的觀念及方法建立，其中一股是現金流量請求權，一權是股東投票權，二者差距大表示道德風險嚴重，視為國際型道德風險指標(MH I)，第二個道德風險的一股也是現金流量請求權，但是一權卻是參酌台灣情況的董事投票權，可視為本土型道德風險指標(MH II)。本文發現這二種道德風險能直接及間接影響銀行績效，當道德風險越大，都會造成銀行的逾放比上升及資產報酬率下降，尤其以台灣董事為基礎的本土型道德風險指標更是顯著。

Corporate Governance is a more common and important issue nowadays. In previous studies, Moral Hazard, proxied by shareholding structure, may be a core reason of bad firm performance; however, it's not enough to show the concept of one-share one-vote. This paper investigates how Moral Hazard affects the banking performance and the relationship between the banking performance and its determinants. Moral Hazard is based on the concept of one-vote one-share, but a different view with shareholding of directors and board structure. Two measures of one-share one-vote are adopted to present Moral Hazard in this paper. First, La Porta., Lopez and Shelifer (1999) and Claessens, Djankov and Lang (1999)suggested to track the controlling shareholder in each firm, then used his cash flow rights and voting rights to construct one-share and one-vote, respectively. This measure is referred to as the international Moral Hazard (MH) in this paper. Ⅰ The next one is local Moral Hazard (MHⅡ), according to specific Taiwan case, where one-vote is proxied by the seats controlled by controlling shareholder in board of directors, whereas the one-share is still based on his cash-flow right. The more deviation from one-share one-vote, the more Moral Hazard is and the worse the firm performance may be. This paper employs 33 banks’ data from 1998~2000 in Taiwan. Basic statistics, liner regressions, switching regressions, and sensitive analysis show that Moral Hazard affects the banking performance. Hypothesis I, which suggests that a lower Moral Hazard index is beneficial to the bank performance (evalued by ROA and NPL), gains support by using MH I. When MH II is used, the relevant t-statistics is even larger, suggesting an even stronger support. Hypothesis II, which describes an interaction effect between financial variables and Moral Hazard Index, is supported by most regression. For example, in a good moral hazard regime, lending could increase the return but in a bad moral hazard regime, lending decreases the return. Our results are fruitful. First, the relationship between moral hazard and bank performance is nonlinear, i.e., distressed banks are found to belong to the group with the highest moral hazard index, whereas banks with the highest moral hazard index are not necessarily suffered the distress. Next, MH Ⅱ is found to outperform MH I with a margin First, variations of MH II is larger than MH I, suggesting that agency problem may be underestimated when international moral hazard is used. It is worth developing a local moral hazard index for each country. Next, distressed bank are related to the moral hazard indices. We find that distressed banks tend to display high moral hazard indices but a bank with high moral hazard indices are not necessarily in distress. Third, our moral hazard indices provide a necessary condition to predict the banks’ CAMEL. Employing CAMEL, we find that a bank with lower moral hazard indices does not necessarily outperform a bank with higher moral hazard indices. By contrast, a bank with highest moral hazard indices typically performs the worst. Last, both of them are useful in predicting returns (ROA) and risk (NPL) but MH II performs slightly better than MH I. Specifically, MH I is useful in predicting the return of banks but it has no information for the risk. By contrast, MH II are informative for both returns and risks.