臺灣博碩士論文加值系統

為因應金融環境的快速變遷，行政院金管會鼓勵本國銀行業加強風險管理能力，然而多數銀行對信用風險模型尚在初步瞭解階段，故本國銀行應借鑒國際上先進的信用風險管理經驗以發展適合自己銀行的信用風險模型。
本文以國內某商業銀行之中小企業授信戶為研究對象，試圖利用Logit迴歸模型(Logostic Regression Model)得到各個授信戶的違約機率之後，主觀將信用等級分成五個等級，估計各等級之平均違約機率與違約機率的波動性，同時參考樣本銀行回收率的經驗值，依據瑞士信貸第一波士頓銀行(Credit Suisse First Boston, CSFB)所發展的信用風險加成模型(CreditRisk+ Model)，建立預期違約損失分配，計算整個放款投資組合的預期及未預期損失，並使用風險貢獻來管理放款組合的信用風險。實證結果如下：
(1)本文利用Logit迴歸模型估計各信用等級之違約機率平均值與標準差，其中信用等級1，其違約機率平均值為0.31%，標準差為0.0034；信用等級2，其違約機率平均值為4.33%，標準差為0.0258；信用等級3，其違約機率平均值為19.20%，標準差為0.0528；信用等級4，其違約機率平均值為42.64%，標準差為0.0929；信用等級5，其違約機率平均值為89.66%，標準差為0.1396。呈現信用等級越差的公司，違約機率越高且違約機率的波動性越大。
(2)依據CreditRisk+模型，建立投資組合之預期違約損失分配，求得各種百分位損失水準的臨界值，以建立損失控制機制：為了因應投資組合未來可能發生的預期損失，需要提撥2,040,388千元(佔放款核貸金額7.08%)的年度信用準備(annual credit provision, ACP)來吸收；為了因應未預期損失的衝擊，需要提列4,450,526千元(佔放款核貸金額15.44%)的額外資本作為緩衝，此即為經濟資本(economic capital, EC)。
(3)將風險貢獻最大的債務人從投資組合中移除後，預期損失減少了4.15%；99%損失水準的臨界值減少了16.7%；所需要的經濟資本也顯著減少了18.3%。較低的經濟資本要求反映出新投資組合的風險減少且多樣化程度變好。
(4)最後，針對CreditRisk+ 模型的四項輸入參數來進行敏感度分析，以信用暴險額的敏感度最高，違約機率的波動性次之。因此，可以透過暴險評等限額的方式，有效規避潛在非必要的暴險，達到分散投資組合風險的目的。

In order to adapt to the fast change of the financial environment, the Financial Supervisory Commission, Executive Yuan urges local banks to improve their risk management ability. However, most local banks are still in the initial stage of getting familiarized with credit risk models. Therefore, local banks are advised to take in the experience from the advanced credit risk models already developed by internationally acclaimed financial institutions in order to develop credit risk models suitable to their operations.
The dissertation examines small and medium-sized enterprises obligors in a certain local bank. It is attempted to calculate the default rate of every obligor by Logistic regression model and subjectively group the obligors into five ratings. Then, the mean default rate and default rate volatility of each rating are estimated before referring to the historical value of the recovery rate of the sample bank. Furthermore, the expected and unexpected losses of the loan portfolio are calculated after establishing the loss distribution according to the CreditRisk+ model developed by Credit Suisse First Boston. Finally, the credit risk of the loan portfolio is managed by using the risk contributions. The empirical results are as follows:
(1)The article used Logit model to estimate the means and standard deviations of the default rate of each credit rating. When the credit rating is 1, the mean of the default rate is 0.31% and its standard deviation is 0.0034; when the credit rating is 2, the mean of the default rate is 4.33% and its standard deviation is 0.0258; when the credit rating is 3, the mean of default rate is 19.20% and its standard deviation is 0.0528; when the credit rating is 4, the mean of default rate is 42.64 and its standard deviation is 0.0929; when the credit rating is 5, the mean of default rate is 89.66% and its standard deviation is 0.1396. As a result, it can be concluded that the worse the credit rating of a company, the bigger the default risk as well as the volatility.
(2)The portfolio loss distribution is established according to the CreditRisk+ model and the loss percentiles are procured to set up a loss control mechanism. In order to handle the expected loss of the portfolio which might happen in the future, an annual credit provision for 2,040,388 thousand dollars (7.08% of the loan amount approved) is required; in order to deal with the impact of the unexpected loss we need to raise an extra capital of 4,450,526 thousand dollars (15.44% of the loan amount approved). This is so-called “economic capital.”
(3)When eliminating the obligor with the biggest risk contribution from the portfolio, the expected loss decreased by 4.15%; the 99th percentile level diminished by 16.7%; the economic capital needed is significantly reduced by 18.3%. Lower economic capital requirements reflect lower risk of the new portfolio and wider diversification.
(4)Finally, a sensitivity analysis is conducted by inserting four different input parameters according to the CreditRisk+ model. The outcome shows that the sensitivity of credit exposure is the highest, followed by the default rate volatility. Therefore, potential unnecessary exposure can be avoided and the risk of the loan portfolio can be diversified by introducing rating exposure limits.

第壹章 緒論……………………………………………………1
第一節 研究動機………………………………………………1
第二節 研究目的………………………………………………3
第三節 研究架構………………………………………………4
第貳章 文獻探討………………………………………………6
第一節 信用評等違約機率量化之技術………………………6
第二節 主要信用風險模型……………………………………10
第三節 文獻評析………………………………………………13
第參章 研究方法………………………………………………14
第一節 Logit迴歸模型流程設計 ……………………………20
第二節 信用風險加成模型流程設計…………………………21
第三節 信用風險加成模型的應用……………………………30
第四節 研究限制………………………………………………32
第肆章 實證分析………………………………………………33
第一節 樣本說明………………………………………………33
第二節 Logit迴歸模型實證結果 ……………………………36
第三節 信用風險加成模型實證結果…………………………40
第四節 銀行放款投資組合管理分析…………………………43
第伍章 結論與建議……………………………………………51
第一節 研究結論………………………………………………51
第二節 研究建議………………………………………………53
第三節 研究特色………………………………………………55

參考文獻…………………………………………………………57

中文部分
1.江玉娟(2005)。「商業銀行如何建置符合新巴塞爾資本協定的信用評分制度」國立中央大學財務金融研究所。
2.何太山(1976)。「運用區別分析建立商業放款信用評分制度」國立政治大學企業管理研究所。
3.李智芳(2005)。「中小企業信保案件之違約機率、回收率與信用風險值的實證研究」。國立中央大學財務金融研究所。
4.周詩添(1978)。「企業授信風險評估模式之研究」淡江大學管理科學研究所。
5.林竹君(2004)。「商業銀行如何衡量企業之償還率」。國立中央大學財務金融研究所。
6.林建州(2001) 。「銀行授信客戶違約機率之衡量」國立中山大學財務管理學系研究所。
7.張永吉(2005)。「新版巴塞爾資本協定本國銀行業信用風險之探討-CreditRisk+信用風險模型研究」。東吳大學商學院企業管理學系碩士在職專班碩士論文。
8.陳侑宣(2004)。「商業銀行如何利用信用風險值檢視授信政策」。國立中央大學財務金融研究所。
9.陳淑玲、陳漢沖(2003)。「信用評等風險量化之研究」。貨幣觀測與信用評等，42，100-109。
10.陳瑞娟(2006)。「中小企業信用風險加成模型之實證研究－以台灣之H銀行為例」。國立交通大學財務金融研究所。
11.陳錦村、陳木在(2001)。商業銀行風險管理。台北市：新陸書局股份有限公司。
12.黃小玉(1989)。「銀行放款信用評估模式之研究--最佳模式之選擇」淡江大學管理科學研究所。
13.敬永康(2001)。「新版巴塞爾協定－內部評等制度(一)」。貨幣觀測與信用評等，32，63-72。
14.敬永康(2001)。「新版巴塞爾協定－內部評等制度(二)」。貨幣觀測與信用評等，33，129-137。
15.敬永康(2001)。「新版巴塞爾協定－內部評等制度(三)」。貨幣觀測與信用評等，34，116-122。
16.黃仁德、陳淑郁(2004)。「信用風險衡量－信用風險加成模型」。台灣金融財務季刊，5:3，77-111。
17.劉貞君(2005)。「信用風險值的探討」。輔仁大學金融研究所。
18.潘雅慧(2002)。「新巴塞爾資本協定及信用風險模型化之研析」。中央銀行季刊，24:2，35-53。
19.盧陽正、陳達新、郭哲男、駱茂榮(2001)。「新一代信用風險模型的介紹與應用—Credit Portfolio View和CreditRisk+」。台灣期貨市場，3:2，3-33。
20.賴耀群(1977)。「銀行放款信用評估模式之研究」淡江大學管理科學研究所。
英文部分
1.Altman, E. (1968). “Financial ratios, discriminant analysis and the prediction of corporate bankruptcy.” The Journal of Finance, 23(4), pp.589-609.
2.Carey, M. (1998). “Credit risk in private debt portfolios.” The Journal of Finance, 53(4), pp.1363-1387.
3.Carey, M. and M. Hrycay (2001). “Parameterizing credit risk models with rating data.” Journal of Banking and Finance, 25, pp.197-270.
4.Carty, L.V. and D. Lieberman (1997). Historical Default Rates of Corporate Bond Issuers, 1920-1996. New York: Moody’s Investors Service, Global Credit Research.
5.Credit Suisse Financial Products (1997). CreditRisk+: A Credit Risk Management Framework, are available directly from CSFB’s web site at: http://www.csfb.com
6.Crouhy, M., D. Galai, and R. Mark (2000). “A comparative analysis of current credit risk model.” Journal of Banking and Finance, 24:1-2, pp.59-117.
7.Jacobson, T. et al. (2006). “Internal ratings systems, implied credit risk and the consistency of banks' risk classification policies.” Journal of Banking and Finance, 30(7), pp.1899-1926.
8.Maddala, G.S. (2001). Introduction to Econometrics, 3rd Edition. New York: John Wiley & Sons.
9.Ohlson, J.T. (1980). “Financial ratios and the probabilistic prediction of bankruptcy.” Journal of Accounting Research, 18, pp.109-131.
10.Platt, H.D. and M.B. Platt (1991). “A note on the use of industry-relative rations in bankruptcy prediction.” Journal of Banking and Finance, 15, pp.1183-1194.
11.Rohatgi, V.K. (1976). An Introduction to Probability Theory and Mathematical Statistics. New York: John Wiley & Sons.
12.The CreditRisk+ spreadsheets (cr1 and cr2) along with a technical document, are available directly from CSFB’s web site at: www.csfb.com/institutional/research/credit_risk.shtml
13.Tracy W.F. and M. Carey (2000). “Credit risk rating systems at large US banks.” Journal of Banking and Finance, 24, pp.167-201.
14.Wilson, T. (1997). “Measuring and managing credit portfolio risk (I).” Journal of lending and credit risk management, pp.61-72.
15.______. (1997). Ratings Performance 1996 Stability & Transition. Standard & Poor’s.