基于复杂网络理论的银行系统性风险传染研究

发布时间：2018-01-05 01:18

本文关键词：基于复杂网络理论的银行系统性风险传染研究　出处：《西南财经大学》2014年硕士论文　论文类型：学位论文

【摘要】：2008年美国爆发的金融危机给世界各国的金融系统造成了重大的经济损失。在危机过后,人们开始痛定思痛,思考金融危机在世界上造成大面积损失的原因。研究发现,这主要是世界金融系统的安全性方面存在着隐患。因此,对于风险的研究就要求人们从过去的微观层面上升到宏观层面,各个国家加强自身金融系统的宏观审慎管理成为防止下一次金融危机大面积蔓延的重中之重。从宏观角度来看,就要把参与经济活动的各金融机构组织,金融市场作为一个整体来考虑,以整体的角度来研究和防范危机在系统中蔓延。以整体的思想来研究系统中风险传染的方法很多,不过近年来使用复杂网络的方法变得备受推崇。欧洲央行在一篇金融稳定评论中就曾经指出,现在社会金融系统中的各要素关系错综复杂,要研究这复杂的关系,就需要使用复杂网络的研究方法。复杂网络是由数学上的图论发展演变而来的二种新兴学科,它涵盖内容广泛,主要包括了系统性科学,管理学,计算机科学,数学,统计学等多方面的理论和知识。正因为如此,复杂网络才被学者们认可,是研究复杂系统最有效的一种工具,就可以将复杂系统抽象成为一种复杂网络。在金融系统中,存在着大量的金融机构,金融组织,各机构组织在系统中起的不同的作用,而且之间经常发生一些业务往来。然而,这种业务往来关系并不是一成不变的,而是随着时间的推移发生着变化。从这个角度来看,金融系统是一个复杂的系统,我们就可以用复杂网络的理论来进行分析研究。金融系统中的金融机构的联系关系主要是通过资金流动形成的,而资金的流动又主要是通过支付系统来完成的,因此我们可以说对应于金融系统的支付系统也会表现出复杂的特性。本文根据大额支付系统建立网络拓扑模型,估计支付系统中银行间交易金额,然后,使用复杂网络的理论方法构建银行系统网络模型,对该网络的拓扑结构性质及网络结构进行分析,最后根据网络结构特性分析系统性风险在网络中的传播规律。本文的具体内容如下：第一部分为文献综述部分,首先,整理了一下国内外学者对于银行网络结构的分析情况,然后对银行间系统性风险传染情况做了整理,主要是从四个方面介绍的,有系统性风险的涵义,国内外学者根据收集到的数据对系统性风险的具体测量,对于系统性风险的一些研究方法,其中主要评述了使用复杂网络理论的方法,综述的最后部分是对于系统性风险爆发的可能性做了整理。第二部分主要介绍本文的数据来源,主要来源于三个部分,支付体系发展报告,各银行的年度报表,以及前支付结算司司长欧阳卫民的一篇文献。这部分最后对于支付系统中数据暴露不充分的现状,本文尝试使用回归分析,变量替代的方法,估算出各银行交易金额在整个支付系统中的比重,然后进一步估算出整个支付系统中任意两家银行之间的交易额。第三部分是先简单介绍了复杂网络的相关理论,然后分析了银行业系统的复杂性,得出可以用复杂网络进行分析的结论,最后根据前面估算出的交易矩阵,对银行业交易金额网络的网络结构进行了分析,主要表现在网络的拓扑参数以及网络结构。网络拓扑参数主要是平均路径长度和度分布,网络结构主要是聚集系数以及网络匹配模式,研究结果表明该网络具有小世界的特性,即该网络具有小的平均路径长度以及较大的聚集系数,为后面对网络进行针对性攻击,分析鲁棒性提供了前提。第四部分介绍了本文在研究银行间危机传染时的具体方法,主要就是对没有破产的银行对上一轮破产的银行进行轧差,然后再和银行资产额比较进行判断,通过不断的迭代过程,来观察危机产生的影响。最后了通过Matlab仿真模拟运行的结果,通过假设一家银行,两家银行,多家银行破产分别做模拟,来对比观察系统中风险传染规律。通过仿真模拟的结果我们可以发现,我国的银行业系统是非常安全的,没有外部因素的干扰是不会出现系统性风险的。第五部分是结论与展望,主要介绍了本文所得出的结论,另外也指出了本文中的不足及局限性,并且为后面的进一步研究做了展望。本文的创新点主要有两点,第一点是数据的估算,考虑需要用到各个银行交易额在整个系统中的比重,但是这种数据不容易收集到,因此本文试着寻找一个相关变量,这个变量的数据能够收集到,并且和待估计的变量之间具有正相关关系,然后用该变量替换需要估计出银行交易额的比重,另外也部分地运用了最大熵估计的思想,把所以已知能够收集到的数据运用在内。第二点是分析危机传染的过程。在分析危机传染的过程中先对系统建立网络模型,然后根据网络模型抽象出对应的矩阵,危机传染过程就对应在矩阵上元素的变化。具体就是对没有破产的银行对上一轮破产的银行进行双边或者多边轧差,然后再和银行资产额比较进行判断,通过不断的迭代过程,来观察危机产生的影响。观察为了反映这种变化规律,通过Matlab软件仿真模拟实验,并通过不断赋予不同的初值,对比观察网络中风险传染情况。
[Abstract]:The financial crisis in the United States has caused great economic losses to the world ' s financial system in 2008 . After the crisis , people began to feel pain and think about the causes of large - area losses caused by the financial crisis in the world . Complex network is a kind of newly developed subject developed by mathematical graph theory . It covers a wide range of theories and knowledge , such as systematic science , management science , computer science , mathematics , statistics and so on . The relationship between the financial institutions in the financial system is mainly formed through financial flows , and the flow of the funds is mainly completed through the payment system , so we can say that the payment system corresponding to the financial system can also show the complicated characteristics . According to the large amount payment system , a network topology model is established , the inter - bank transaction amount in the payment system is estimated , and then the network model of the bank system is constructed by using the theory method of the complex network , and the propagation rule of the systemic risk in the network is analyzed according to the network structure characteristics . The first part is divided into the literature review part . First , the author finishes the analysis of the bank network structure at home and abroad , then arranges the systematic risk infection among the banks , which is mainly introduced from four aspects , and has the meaning of systemic risk . The second part mainly introduces the data source in this paper , mainly comes from three parts , payment system development report , annual report of each bank , and a document from the director of the former payment and settlement division , Eryang Weimin . In the end , this part attempts to use regression analysis and variable substitution method to estimate the specific gravity of each bank transaction amount in the whole payment system , and then further estimates the transaction amount between any two banks in the whole payment system . In the third part , the theory of complex network is simply introduced , then the complexity of banking system is analyzed . Finally , the network structure of the network is analyzed based on the previously estimated transaction matrix . The network structure is mainly the aggregation coefficient and the network matching mode . The results show that the network has the characteristics of small world , that is , the network has the characteristics of small average path length and large aggregate coefficient , which provides the precondition for the target attack and analysis robustness of the network . In the fourth part , the paper introduces the concrete way to study the inter - bank crisis infection , which is mainly to judge the bank which is not bankrupt , then judge the bank ' s assets . Finally , through the simulation simulation , we can find that the banking system in our country is very safe , and there is no systematic risk without the external factors . The fifth part is the conclusion and prospect , mainly introduces the conclusion drawn from this paper , and also points out the shortcomings and limitations in this paper , and prospects for further research in the future . The paper tries to find a relative variable , the data of this variable can be collected , and the corresponding matrix is abstracted according to the network model . The second point is to analyze the effect of crisis infection . In order to reflect this change rule , the simulation experiment is simulated through Matlab software , and the risk infection in the network is observed by constantly giving different initial values .

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F830.3;O157.5

【参考文献】