Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/28874
標題: 農業行庫授信風險評估模式之研究
A Study on Credit Risk Rating Models of Agricultural Banks
作者: 龐欽元
Pang, Chin-Yuan
關鍵字: Agricultural Banks;上市公司;Credit Risk;Finance Ratio;Listed Companies;Full Delivery;Credit Crisis;Factor Analysis;Multiple Discriminate Analysis;Probit Regression Analysis;Logit Regression Analysis;因素分析;多變量區別分析;Probit迴歸分析;Logit迴歸分析;農業行庫;全額交割;信用危機;財務比率;授信風險
出版社: 農業經濟學系
摘要: 
自從政府提倡金融自由化、國際化後,新銀行的相繼成立,金融商品不斷的推陳出新,使得銀行授信業務間的競爭日趨白熱化。銀行之營運槓桿大不如前,授信之利息收入直線下滑。然而多數的銀行在積極拓展授信業務的同時,可能因此忽略了授信的品質,當經濟景氣低迷時,面臨借貸者無力償還或要求展期的情形,增加銀行業務經營的風險性與不確定性。農業行庫(中國農民銀行、台灣土地銀行與台灣省合作金庫)為農業之專業銀行,在調劑農業金融及供給農業有關信用上,扮演著極重要的角色。職是之故,本研究以農業行庫授信戶財務比率資料,建立一套較佳之授信風險評估模式,以作為農業行庫授信風險評估之參考依據。
本研究以台灣證券交易所股票公開上市、上櫃,且與農業行庫有債權往來關係之102家公司為研究樣本,其中以股票交易被裁定為全額交割、停止買賣或終止上市(櫃)之公司,作為信用危機公司之界定標準,共選取信用危機樣本34家﹔另以配對方式,選取產業相同與規模相近之信用正常樣本68家﹔並將樣本分割為兩部分,原始樣本包含69家上市、上櫃公司,預測樣本包含33家上市、上櫃公司。收集信用危機前三年之資產負債表、損益表、股東權益變動表與現金流量表來計算財務比率各20項,而後利用因素分析法萃取每年度之財務比率因素,依各年度所抽取之共同因素建立信用危機發生前三年度之多變量區別分析模式、Probit及Logit迴歸分析模式,並比較其區別與預測效果。
研究結果顯示:
(1)各年度之財務比率因素隨時間而改變,危機發生前一年至前三年,
因素分析將20項財務比率歸納為5-6個共同因素,各年度財務比率之
因素解釋能力分別為71.052﹪、84.358﹪與80.564﹪。
(2)本研究所建立之各模式,其解釋能力與區別能力均具顯著性。危機
發生前一年至前三年,多變量區別分析模式之區別及預測準確率依
序分別為65.22﹪、65.22﹪與60.87﹪﹔63.64﹪、66.67﹪與
54.55﹪。Probit分析模式為76.81﹪、73.91﹪與75.36﹪;
75.76﹪、66.67﹪與75.76﹪。Logit分析模式為78.26﹪、73.91﹪
與77.36﹪﹔75.76﹪、66.67﹪與75.76﹪。
(3)Logit迴歸分析模式無論於分類及測試效果上皆優於多變量區別分析
模式與Probit迴歸分析模式。

Because of the internationalization and Liberalization of the nation''s financial system, new banks have been established one by one, and the financial products have been changed very fast. The competition in loan making among banks has become harder. Then the banking cannot use leverage as ever, and both the revenues of interest and the profits are declining. Most banks start increasing the loan business, they may neglect the quality of loan. When the debtors are involved with recession, they may face cash insolvency and cash inadequacy. Therefore the environment of bank operation becomes more uncertain and risky. Agricultural banks (The Farmers Bank of China, Taiwan Land Bank and Taiwan Cooperative Bank) are the professional banks of agriculture. They play an important role in adjusting agricultural finance and providing credit related to agriculture. As result, banks urgently need a concrete and definite credit risk evaluation model as a criterion before loan making.
The data of this study was derived from 102 sample stock companies listed in Taiwan stock exchange corporation, which the agricultural banks made loans to them. We used those companies with full delivery or were previously removed from the list as the criterion of credit crisis and then chose 34 credit crisis samples for data collection. On the other hand, we separately found 68 normal samples with the similar scales in the same industry by means of pair-making method. And all these 102 samples were divided into two groups: the first subsameple of 69 companies was used as the original set; the second subsameple consisted of 33 companies was used as the predictive set. The balance sheet, income statement, schedule of changes in stockholders'' equity and cash flow statement were traced back to 4 years before failure to compute the 20 annual financial ratios. The financial status of failure and non-failure companies were first compared and contrasted. The factor analysis was then applied to extract the most significant ratios in predicting the business failures. Based on the common ratios extracted annually, multiple discriminant analysis model, probit regression model and logistic regression model were developed by using the data of three years for each model prior to the credit failure. We can compare the differences and the predicted performances of these models.
The empirical results led to the following conclusions:
(1) The significant financial ratios were not identical in
three years. 5 to 6 factors were extracted from 20 ratios
through factor analysis, the variance of significant ratios
were found to be 71.052%, 84.358% and 80.564% respectively.
(2) As the study showed, the explanatory ability and
discriminanting ability in this study were significant.
The correct rates of the classification and the prediction
of multiple discriminate analysis model in three years were
separately 65.22%, 65.22% and 60.87%; 63.64%, 60.87% and
63.64%. Probit regression model were 76.81%, 73.91% and
75.36%; 75.76%, 66.67% and 75.76%. Logistic regression
model were 78.26%, 73.91% and 75.36%; 75.76%, 66.67% and
75.76%.
(3) The results of this research revealed that the logistic
regression model performed better on both the correct
classification and the prediction than the other two models.
URI: http://hdl.handle.net/11455/28874
Appears in Collections:應用經濟學系

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.