Abstract: At present, enterprises form complex relationships with network characteristics based on creditor's rights and debts, guarantees and cross-shareholdings, etc. The “credit chain” which is inherent in the correlation relationship has gradually become an important risk infection path among enterprises. The credit chain formed by the association relationship of a single enterprise's credit risk spreads and enlarges in the industry, which is the credit risk contagion. The outstanding feature of the new energy automobile industry is the upper, middle and downstream “deep linkage”. The scope of risk impact of a single enterprise is no longer limited to the enterprise itself. Through the network of mutual credit support path overflow to the upper and lower credit related parties, the industry credit risk correlation has gradually increased. Based on this, we adopt the complex network method to study the credit risk transmission mechanism of the green-car industry, which on the one hand makes a useful supplement to the relevant achievements of the credit risk transmission of green-car industry, and on the other hand has a strong guiding significance to prevent the industry credit risk transmission from evolving into a systemic risk.
Based on the above background, this paper selects 25 representative enterprises in the new energy automobile industry as research objects, and the data source is the financial data of enterprises in each year in the Wind database. In this paper, the default distance in the KMV model is used to measure the credit risk level of each enterprise in different years, and the Pearson coefficient is used to measure the credit risk correlation degree of each enterprise in different years, and on this basis, the industry credit risk threshold network is constructed. Finally, the evolution of network characteristics is studied in stages through the complex network model. In the aspect of network research, first of all, degree centrality, network centralization, average path length and clustering coefficient are used to describe the evolution of the overall network features, and then the minimum spanning tree method is used to analyze the local features of the network, so as to put forward corresponding policy suggestions for related investors.
This paper finds that: (1)The credit risk network has typical characteristics of “scale-free” and “small world”, with uneven distribution of credit risk correlation. (2)The concentration degree of network correlation shows a trend of fluctuation, and higher correlation concentration reflects the tightness of credit risk correlation structure and becomes the catalyst of risk contagion. (3)The prominent feature of industry network different from that of financial network is that its topology shows the community structure of upper, middle and lower link clustering, and linkage effect of each clustering is different. Such inter-link clustering and inter-clustering linkage form a mechanism of endogenous network contagion.
Based on the above findings, this paper puts forward the following suggestions: (1)Make full use of the advantages of the two types of enterprises. Industrial credit risk monitoring should not only pay attention to the risk exposure of a single enterprise, but also deeply understand the contagion linkage phenomenon formed by the characteristics of the network “small world”, and make full use of the advantages of the “two types” of enterprises to solve the problem of industrial risk contagion. (2)To achieve comprehensive collaborative monitoring of industrial credit risks, explicit external government supervision should have a strong demonstration and guidance role in the overall risk prevention and control of the industry, while implicit market regulation should have a prominent advantage in the management of risk contagion among internal enterprises. Therefore, the monitoring design that adopts the coordination of government “supervision” and market “control” has high application value in risk prevention. (3)To implement classified risk infection early warning, credit risk threshold warning lines for different links of upper, middle and downstream should be formulated based on comprehensive consideration of enterprise policy sensitivity and consumers' expected response, and targeted sub-link dynamic monitoring of industry risk infection should be implemented based on different market conditions.

Key words: complex network theory, green-car industry, credit risk contagion, KMV model, minimum spanning tree

摘要： 在“后补贴时代”背景下,新能源汽车行业信用风险传染成为加剧行业系统性风险波动的关键特征。本文利用2016—2020年新能源汽车行业违约距离数据,运用复杂网络中最小生成树(MST)方法刻画行业信用风险传染机制,研究发现:(1)新能源汽车行业信用风险网络具备典型的“无标度”及“小世界”特征,企业信用风险关联分布不均;(2)网络关联集中度呈波动上升态势,较高的关联集中度体现行业信用风险关联结构紧密性特征进而成为风险全局性传染现象的催化剂;(3)行业信用链风险关联网络异于金融风险网络的突出特点是其拓扑呈现出上、中、下游分环节聚类的社团结构特征,且不同外部政策环境下各聚类间的联动效应有所差异,这种环节内聚类及聚类间的联动是信用风险内生性网络传染的重要机制。基于上述发现,本文提出“充分利用两类企业的优势地位、政府‘监'与市场‘控'协同、进行不同市态阶段下分环节动态监控”的建议,为防范行业信用风险传染演变为系统性风险提供决策参考。

关键词: 复杂网络理论, 新能源汽车行业, 信用风险传染, KMV模型, 最小生成树