吳榮杰臺灣大學：農業經濟學研究所顏晃平Yen, Huang-PingHuang-PingYen2007-11-282018-06-292007-11-282018-06-292006http://ntur.lib.ntu.edu.tw//handle/246246/58991　　本研究主要利用存活分析概念，瞭解農會信用部存活時間與失敗風險程度的關係，並探討影響農會信用部存活時間長短之主要因素，以及預測農會信用部最終失敗率情形。 　　研究觀察期間設定為民國84年至93年間，於觀察期間內符合失敗事件定義(逾放比率超過15%以上)之信用部，計有184家。而於觀察期間結束後(民國93年之後)，仍未有逾放比率超過15%事情發生之截斷(censored)樣本，共計有90家，合計全體樣本共274家。其中失敗樣本所占比率為67%，而未失敗樣本(或截斷樣本)占33%。 　　主要使用分析工具為非參數及參數法存活分析模型，首先以非參數存活分析模型，探討農會信用部存活時間和失敗風險大小的關係。接著加入自變數的考量，主要以參數法存活模型作為分析工具，並放寬傳統參數法之若干假設，進一步使用區隔母體與隨時間變動之自變數存活模型進行分析，最後，並進行最終失敗率之預測。 　　根據實證結果，可得出如下之結論： (1)逾放比率超過15%以上之農會信用部占整體樣本的比率相當高。 (2)農會信用部面臨事件發生危機率每年並非固定不變，其中農會信用部高風險情況持續期間相當長，直到近幾年來風險才逐漸降低。 (3)成本效益管理、淨值大小、資金流動性、區域及是否加入存款保險皆是影響信用部存活時間長短的原因。 (4)在觀察期間內未發生失敗之信用部，未來仍會有失敗風險產生。The purpose of my work was to address the following issues:It was found that duration(survival time) and risk on credit department of farmers’ associations. Is the timing of failure depend on different factor , and what is probability of eventual failure on credit department. A credit department was defined as failed if it loan ratio above 15% during the 1995-2004 period. Over 2004 year, the sample was defined as censored data. The complete sample data were 184 units, and censored data were 90 units. Total samples have 274 units. In order to address these issues , two methods were used for that purpose: nonparametric and parametric survival model. First ,we was used nonparametric survival model to analysis relationship between risk and time-to-failure on credit department of farmers’ associations. Next, we consider independent variables, and parameter estimates from the parametric survival model , such as standard log-logistic model, split population log-logistic model, and log-logistic model with time-varying covariate. Finally, we predit probability of eventual failure. In conclusion, we have obtained the following from empirical results: (1) It was taken in high percent that credit department of farmers’ associations overdue loan ratio above 15%. (2) Notice that estimated hazard is not constant during research period, especially, hazard that increases for sufficiently long durations, and declines near years. (3) We find that basic indicators of credit department of farmers’ associations condition ,such as cost-benefit management, net income,liquidity,location, and whether to joint deposit insurance are related significantly to the timing of credit department failure. (4) Over research period, it still was risking for not failing credit department.第一章 緒 論 1 第一節　研究動機與目的 1 第二節 研究方法與步驟 3 第三節 研究範圍與資料來源 6 第四節 本文架構 6 第二章 農會信用部背景與現況 8 第一節 金融改革下之農業金融法概況 8 第二節 問題農會信用部之認定與處理機制探討 14 第三節 農會信用部經營現況 25 第三章　理論基礎與文獻回顧 35 第一節　存活分析之理論基礎 35 第二節　存活分析之相關文獻回顧 41 第四章 實證模型之建構 52 第一節 失敗定義與觀察期間設定 52 第二節　研究變數之設定 55 第三節　分配模型之選擇 60 第四節　實證模型之設定 67 第五章 實證估計與結果 73 第一節　非參數存活分析結果 73 第二節 參數法存活分析結果 82 第三節 參數法存活分析之預測 91 第六章　結論與建議 93 第一節 結　　論 94 第二節 建　　議 96 附錄一 97 參考文獻 108714456 bytesapplication/pdfen-US農會信用部存活分析區隔母體存活分析危機率Survival ModelSplit Population Survival ModelHazardSurvival Probabilitythesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/58991/1/ntu-95-D91627001-1.pdf