基于多层复杂网络结构分析的关联贷款风险识别模型  

Risk Identification Model of Related Loans Based on Analysis of Multi⁃layer Complex Network Structure

作　　者：张恩勇 刘超 李永立[2] 夏丽娟 Zhang Enyong;Liu Chao;Li Yongli;Xia Lijuan(School of Economics and Finance,Xi’an Jiaotong University,Xi’an 710061;School of Economics and Management,Harbin Institute of Technology,Harbin 150001)

机构地区：[1]西安交通大学经济与金融学院,西安710061 [2]哈尔滨工业大学经济与管理学院,哈尔滨150001

出　　处：《管理评论》2024年第5期3-11,共9页Management Review

基　　金：国家自然科学基金项目(72171059,71771041);黑龙江省哲学社会科学研究规划项目(21XWB124)。

摘　　要：关联贷款是指因某种(或若干种)关系而联系在一起的多笔贷款,这些贷款容易“牵一发而动全身”,引起集体性违约。由于关联贷款中存在复杂的多头贷款及隐蔽的影响关系,使得银行难以识别关联贷款并采取针对性措施。本文基于多家商业银行的贷款担保等数据构建了多层复杂网络,提出了一种识别关联贷款结构的算法,并讨论了执行效率;统计了不同关联贷款结构的实际违约率,并与4种风险指标进行对比;最后检验了不同关联贷款结构违约风险的显著性。结果表明:(1)本文构建的多层网络关联贷款模型及提出的识别算法,能大幅提高关联贷款识别效率及精准度,克服单层网络无法解决的关联贷款双重隐蔽性问题;(2)介数、聚类系数指标与关联贷款真实违约率表现较为一致,而基于清偿能力与风险距离的指标对关联贷款真实违约率的预测存在失灵现象;(3)当关联贷款网络中具有圈联结构与汇结构时,贷款的违约风险会显著增加。本文构建的模型及算法为识别多银行、多关系的关联贷款提供了理论基础,对指导银行识别风险网络结构、管控关联贷款风险具有实践意义。Related loans refer to loans that are linked together due to a certain(or several)relationship(s)and these loans as a whole are likely to cause group default.Because of the complicated cross⁃bank loan and concealed influence relationship in related loans,it is difficult for banks to identify related loans and take effective measures.Based on the loan guarantee data of several commercial banks,this paper constructs a multi⁃layer complex network,then designs an algorithm to identify the related loan structure and analyzes the ef⁃fectiveness;the default rate of different related loans structures is calculated and compared with 4 risk indicators;finally,the signifi⁃cance of default risk of different related loans structures is tested.The result shows:(1)the multi⁃layer network related loan model and the recognition algorithm constructed in this paper greatly improve the efficiency and accuracy of related loan identification,and overcome the dual concealed problem of related loans that cannot be solved by the single layer network;(2)the betweenness and clustering coeffi⁃cient indicators are more consistent with the true default rate of related loans,while the indicators based on the clearing payment capacity and risk distance fail to predict the true default rate;(3)when there are circle⁃linked structure and sink structure in related loans net⁃work,the default risk of loans increases significantly.The model and method constructed in this paper provide a theoretical basis for identifying multi⁃bank and multi⁃relationship related loans,and have practical significance for banks to detect risk network structures and to control related loan risks.

关 键 词：多层网络 关联贷款 结构识别 风险识别 

分 类 号：F832.4[经济管理—金融学]