【作者】 王东浩；

【导师】 张秋生；

【作者基本信息】 北京交通大学 ， 企业管理， 2012， 博士

【摘要】 金融体系中,不良贷款的产生一直对金融机构构成着严重威胁,甚至是一个国家的经济。设立资产管理公司处置不良贷款是世界各国一项通行且行之有效的做法。源于我国银行金融机构经营模式,如何防范金融不良贷款的发生及发生后的损失管理问题研究成为是我国金融体系最核心的研究内容之一。特别是在我国开展实施巴塞尔Ⅱ新资本协议的今天,深入研究不良贷款的处置和损失问题对我国银行信用风险管理水平的提升及开发银行内部评级系统中的核心参数—违约损失率LGD都具有重大的现实意义。由于我国特殊的金融发展历程,我国银行业的违约不良贷款在2006年前基本划拨或出售给了四大资产管理公司。在新的金融环境下,根据国务院及财政部对资产管理公司新的发展定位,我国四大资产管理公司将不会结业清算,回归母体行,而是仍将以不良贷款处置为主业发展金融控股集团。因此,在未来可预见的时间内,资产管理公司仍将是我国金融不良贷款处置管理的主要力量。一方面,由于历史轨迹的改变,我国金融资产管理公司亟需通过对历史数据的采集、整理和挖掘,探索一条提升不良贷款处置管理水平的新途径；另一方面,由于我国金融不良贷款对银行金融机构的重大影响,研究不良贷款的特征及回收率估计模式,对我国新资本协议的实施不可或缺。违约损失数据库的建设在我国一直是个空白,利用大规模历史数据及其挖掘成果进行不良贷款处置管理的国内外研究和实证也很少。同时,与银行业、学术界和各信用评级机构所关注违约率的研究相比,由于数据缺乏、影响因素众多和形成原因复杂多样等问题,对违约损失率(LGD)国外学者也是从90年代中后期才开始关注,而国内的研究起步更晚,大部分还处于定性的描述,成熟的定量研究及模型开发几乎还是空白。由于信用环境和经营模式的差异,符合中国国情违约损失数据库的开发、数据的挖掘利用、违约损失率模型的开发研究都是我国金融业和学界亟需填补和完善的研究空白。由于违约损失率和回收率的互补关系,本文将交互使用这两个概念。本文尝试从资产管理公司的视角出发,结合新资本协议对金融风险管理的要求,通过建立我国大型的金融不良贷款损失数据库,在海量的不良贷款回收数据的基础上,总结归纳了我国不良贷款的特征和适合我国国情的违约损失率模型开发方法,并形成系统软件进行了实证。开拓性地研究了一套利用数据、挖掘成果、IT技术来整体提升资产管理公司不良贷款处置和损失管理的方法。具体结构如下：第二章综述和总结了不良贷款发生及处置清收的国内外模式及处置方法；第三章对不良贷款处置核心定价问题开展了理论研究。从金融资产定价理论出发,探讨了信用风险定价问题和围绕信用资产违约损失率计量的各种方法。追根溯源实现了不良贷款违约损失定价问题从理论到实践的全面梳理和研究；第四章在全面设计我国金融不良贷款数据库和数据准备的基础上,本文分析了影响我国不良贷款回收率水平的主要因素：宏观因素、行业地区因素、债务人因素、债项因素等；第五章总结了适合我国国情的不良贷款定价模型框架和回收率估计模型并对主要模型变量的贡献度进行了分析；第六章以资产管理公司为例,给出了在实际研究成果基础上所构建的我国第一个基于计算机技术的、从数据管理到过程风险管理到处置定价管理的实际管理框架。其中包括数据采集和数据处理,不良贷款处置方式的选择、不良贷款的定价及风险监测。应用实践表明,本研究的成果,不仅大大提高了资产管理公司的工作效率,而且也提高了不良贷款的回收率。第七章总结了全文的研究成果,并结合我国国情给出相应的政策建议。本文的创新点在于：(1)通过中国的海量不良贷款违约损失数据,实证总结了影响我国不良贷款回收率的基本因素：宏观经济因素、行业地区因素、债务人因素和债项因素。影响因素的得出为中国不良贷款的定价模型建设奠定了理论基础；(2)总结了适合我国国情的不良贷款回收率的建模方法应根据我国不良贷款的违约损失率U型分布的特征,通过先判别后回归的方式来进行回收率的计量工作,同时为提高模型实用性,应采取单户模型和打包模型结合的方式。同时模型建设中应该使用分布模型来更好的认识我国不良贷款的分布特征和影响因素；(3)研究总结了我国金融不良贷款违约损失数据库基本设计要素,包括：业务变量的内容,业务逻辑结构,数据采集清洗的有效机制。提出了我国金融环境下建立违约损失数据共享机制的必要性和重要意义,为今后我国金融不良贷款定价、管理及风险研究工作奠定了良好的研究基础；(4)通过实证研究,提出只有通过基础违约损失数据库的建设,在进行大量的数据特征和损失率影响因素挖掘的前提下,构建适合中国国情的模型才能有效的实现金融不良贷款的处置管理。同时为我国金融不良贷款处置管理及风险控制在整体决策层面上提供了实际案例和参考模式。更多还原

【Abstract】 Non-performing loan (NPL) has long been a great threat to financial institutions, and even to the whole nation’s economy. It’s a prevailing and effective way to dispose of non-performing loans by setting up asset management companies(ACMs) which are specialized in NPL disposalling. Due to the business model of banks in our country, it’s one of the very core interests of research in our financial system to avoid NPL and to avoid the problems in NPL disposal.Moreover, since our country is nowadays implementing Basel Ⅱ, a thorough research into NPL disposalling and recovering is magnificently meaningful to both the enhancing of banks’ability in credit risk management and the development of Loss Given Default(LGD), which is a key parameter in bank’s internal rating-based approach.According to the unique path of development of financial industry in our country, most of the NPLs in bank industry had been conveyed to or sold to the4state owned ACMs by2006. Under the new financial environment, the4ACMs are not going to liquidate thus eventually coming back to the parental company, but to grow to holding companies whose major business is NPL disposalling. So, in the foreseeable future, ACMs are still the dominant force in NPL disposalling. On one hand, in order to keep up with time, the ACMs need to develop a new way to improve their pricing capability in NPL by means of data collecting, sorting and mining; on the other hand, studying on NPL characteristics and on the estimate model of recovery rate (RR=1-LGD) is vital to the implementation of Basel Ⅱ.Our country was and is still blank with LGD data base, and research on NPL disposalling and management with historical data and data mining has been rare both inside the country and worldwide. Meanwhile, compared with the research on Probability of Default(PD), which has been given a lot of attention by banks, academics and credit rating institutions, research on LGD started only from late90s by some foreign scholars, even later did we start ours. Still, most of the researches are of qualitative description, and quantitative researches and modeling are rare. Dud to the discrepancies in credit atmosphere and business model, the development of LGD data base, data mining and utilizing and LGD modeling which can comply well with China’s condition are badly needed both in financial industry and in academia. This dissertation is composed from an ACM’s perspective, and by constructing a massive NPL data base with China’s data conforming to Basel II’s regulations on financial risk management, it develops a data mining model together with a system software that is based on various theoretical presumptions. It creatively finds a way to enhancing ACMs’ability to disposal and manage NPLs using data, mining and IT technology. The framework is as follows:Chapter2gives a thorough analysis and summary of how NPL is happened and how it is settled both in foreign countries and in China;Chapter3focuses on pricing of NPL. It starts with pricing theories and then discusses credit risk pricing and various means of measurement of credit asset LGD. It gives an overall description and study on NPL pricing from theory to practice;Chapter4analyzes variable design and utilizing model in both foreign and domestic LGD data base construction. Then it summarizes and points out the fundamental path for LGD data base construction that complies with our country’s basics. It also offers suggestions on data sharing model in our country;Chapter5analyzes major influencing factors for NPL recovery rate:macro economy,industry and locality, debtor, debt. At the same time, it summarizes NPL pricing model and recovery rate estimate model that are compatible with China;Chapter6takes AMC as an example and illustrates the real managerial operational management framework. This framework is the first in our country that is based on computer technology covering from data management to process risk management and to disposal and pricing management. It involves data collecting and processing, selection on NPL disposalling methods and NPL pricing and risk monitoring. Facts show that the fruits of the research not only improve ACM’s efficiency but also increase the recovery rate for NPL.Chapter7summaries the whole article and offers policy suggestions based our country’s conditions.The innovation of this paper are:1. come up with a complete and detailed methodology to fully construct NPL LGD data base in our country, offering both theory and practice to fundament data processing for financial risk management;2. First to come up with a systematic analysis on affecting factors for NPL recovery rate by analyzing China’s massive default data;3. Summarizes the method for NPL recovery rate modeling in China and actual model development;4. Probe into a compatible way for our financial institutions to disposal and manage NPL risk management model with the help of data base construction and data mining, together with measurement model.更多还原