管理科学学报
管理科學學報
관이과학학보
JOURNAL OF MANAGEMENT SCIENCES IN CHINA
2010年
2期
86-96
,共11页
马宇超%陈暮紫%陈浩%王博%陈敏%杨晓光
馬宇超%陳暮紫%陳浩%王博%陳敏%楊曉光
마우초%진모자%진호%왕박%진민%양효광
不良贷款组合%巨额同质资产组合%回收率估计%结构模型
不良貸款組閤%巨額同質資產組閤%迴收率估計%結構模型
불량대관조합%거액동질자산조합%회수솔고계%결구모형
non-performing loan%large homogenous portfolio%recovery rate estimation%structural model
处置不良资产是我国金融业改革和发展的重要问题,大规模批量处置不良贷款是处置不良资产的首选方法之一.不良贷款组合的回收率分布,是不良贷款组合定价的基础.针对不良贷款回收率的双峰分布特性,将不良贷款分为低峰回收贷款和高峰回收贷款,证明了不良贷款组合的回收率收敛于低峰回收率和高峰回收率的条件期望之和.基于信用风险结构模型,进一步证明了高峰回收率条件期望的正态逆变换与低峰回收比率的正态逆变换之间存在线性关系.由于对低峰回收贷款易于判断,很容易估计低峰回收比率,因而可以通过线性回归估计出高峰回收率条件期望,这样就给出了不良贷款组合的整体回收率估计模型.基于这一模型,还给出了估计不良贷款回收率的分位数的计算方法,该方法实际上是VaR方法的推广.
處置不良資產是我國金融業改革和髮展的重要問題,大規模批量處置不良貸款是處置不良資產的首選方法之一.不良貸款組閤的迴收率分佈,是不良貸款組閤定價的基礎.針對不良貸款迴收率的雙峰分佈特性,將不良貸款分為低峰迴收貸款和高峰迴收貸款,證明瞭不良貸款組閤的迴收率收斂于低峰迴收率和高峰迴收率的條件期望之和.基于信用風險結構模型,進一步證明瞭高峰迴收率條件期望的正態逆變換與低峰迴收比率的正態逆變換之間存在線性關繫.由于對低峰迴收貸款易于判斷,很容易估計低峰迴收比率,因而可以通過線性迴歸估計齣高峰迴收率條件期望,這樣就給齣瞭不良貸款組閤的整體迴收率估計模型.基于這一模型,還給齣瞭估計不良貸款迴收率的分位數的計算方法,該方法實際上是VaR方法的推廣.
처치불량자산시아국금융업개혁화발전적중요문제,대규모비량처치불량대관시처치불량자산적수선방법지일.불량대관조합적회수솔분포,시불량대관조합정개적기출.침대불량대관회수솔적쌍봉분포특성,장불량대관분위저봉회수대관화고봉회수대관,증명료불량대관조합적회수솔수렴우저봉회수솔화고봉회수솔적조건기망지화.기우신용풍험결구모형,진일보증명료고봉회수솔조건기망적정태역변환여저봉회수비솔적정태역변환지간존재선성관계.유우대저봉회수대관역우판단,흔용역고계저봉회수비솔,인이가이통과선성회귀고계출고봉회수솔조건기망,저양취급출료불량대관조합적정체회수솔고계모형.기우저일모형,환급출료고계불량대관회수솔적분위수적계산방법,해방법실제상시VaR방법적추엄.
Recovering non-performing loan is a key issue in the innovation and reform of China financial industry. Batch-package method is one of the first alternatives to do with non-performing loan recovering. The core issue in the recovering procedure is to estimate Loss Given Default (LGD) of non-performing loan portfolios. Based on the bi-peak feature of recovery rate, we classify non-performing loans into low-recovery loans and high-recovery loans, and prove that the recovery rate of non-performing loan converges to the sum of low-recovery rate and conditional expectation of high-recovery rate. Using classical structural model, we further derive that there is linear relation between the inverse-normal transformation of conditional expectation of high-recovery rate and the inverse normal transformations of low-recovery ratio. Since it is easy to discriminate the low-recovery loan, and hence easy to estimate low recovery ratio, we can estimate the whole non-performing portfolio recovery rate by linear regression. Then we give an approach to estimate the Loss Given Default of large homogenous non-performing loan portfolio. Based on the model, we give a method to calculate the quantile of non-performing loan portfolio recovery rate, which is an extension of VaR.