Evolution and Implications of Finance-Prudential Network under Extreme Conditions: Empirical Studies on Chinese Credit Guarantee Network

極端條件下金融審慎網絡的演化與蘊涵: 中國信用擔保網絡的實證研究

Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date31 Jul 2018


The guarantee relationship between two enterprises refers to the responsibility of an enterprise for another enterprise’s financial obligation if that enterprise fails to meet its loans. A guarantee network is originated from the social cooperation in real life. However, a guarantee network has many unique characteristics, such as low density, high local connection, prudence, and power-law property, which are different from social networks studied in the literature before. Especially, guarantee contracts are conducted under law because firms are connected by a large amount of money and strong trust relationships. Thus, we call such a network as a prudential financial network. In this research, we use a unique data set from China Banking Regulation Commission which contains all the loans of middle-size and above firms in the all national-wide banks to construct the guarantee networks, to study the evolution of the networks under extreme conditions.

In addition, our dataset ranges from January 2007 to March 2012 that covers three important extreme events in China, namely, the 2008 financial crisis, the Chinese 4 trillion stimulus plan, and the re-adjustments aftermath. We focus our attention on the evolution of topologies, regional sub-networks, sub-patterns, and structural stability of guarantee networks in the three periods by utilizing the advantages of the special dataset. An in-depth data-driven quantitative study on guarantee networks should be conducted. In this research, we also analyze the risk of default or contagion of guarantee networks along with the guarantee network evolution. Specifically, the detailed chapters are illustrated as follows.

In Chapter 3, we provide a detailed global topological analysis on the Chinese guarantee network and highlight the effect of financial crisis and monetary policies on the evolution of network topologies. On the one hand, the static and dynamic analyses on network indicators confirm a number of stylized facts that are verified for other real complex systems. Guarantee networks are highly sparse, incomplete, and exhibit a small world property and a power-law degree distribution. On the other hand, we present data-driven insights on the association of the topological structure of guarantee networks with economic shock (the 2008 financial crisis) and monetary policies (i.e., Chinese 4 trillion stimulus plan and latter adjustments). Specifically, the empirical and exponential random graph model results provide that (a) the guarantee network becomes small due to the huge number of bankruptcies of small and medium firms during financial crisis, and (b) the loose monetary policy along with the stimulus program increase the mutual guarantee behavior among firms that resulting in a highly reciprocal and interconnected network. The following adjustments of monetary policies reduce the interconnection of the network.

After investigating the nationwide guarantee network, we perform a detailed evaluation on regional sub-networks that have significantly different networks between each other. In Chapter 4, we demonstrate the topological differences among provinces and explain the causes of such topological diversities. First, we construct a detailed statistic on the topological properties of 31 regional sub-networks based on the loan guarantee data of 31 Chinese provinces. Furthermore, we delve deep in investigating the correlation of the topologies of regional sub-network and macro- and micro-economic factors with canonical correlation analysis. The important observations from canonical correlation analysis are that regional macro-factors and firm-level financial characteristics affect the scale (node number, edge number, size of giant component, and shortest path length) and local connectivity (reciprocity, clustering coefficient, and degree) of the guarantee network. Our analysis on the main factors of network topologies may serve as a guide on controlling network topologies from the perspective of macro- and micro-economic factors.

Network motifs are microstructures (sub-patterns) that recur within a network more frequently than expected at random, which constantly correspond to several functional sub-patterns in different networks. In Chapter 5, we focus on the analysis on motif detection of heterogeneous and dynamic Chinese guarantee networks. We obtain statistical important sub-patterns (motifs) of 2- and 3-node that reveal firms have special preferences to form mutual, triangles, and 2-out-star guarantee relationships. Specifically, we consider the heterogeneity of firms and determine that firms with relatively large assets tend to form the sub-pattern of 2-out-star and that small firms tend to form mutual guarantee relationships. In addition, we develop an easy and fast criterion to locate the high default sub-patterns or firms in the guarantee network. In the last part, we investigate the functionally important sub-patterns from the perspectives of default risk and contagion capability. Our analyses show cyclic triads and completely connected subgraphs of 3-node demonstrate low default risks with high contagion capability and sub-patterns of edge, and 2-out-star are perfect options to reduce or control the contagion probability in guarantee networks. The conclusions in our study could provide a good analysis on firms’ favorite guarantee patterns and serve as a guide on structural design principles to reduce or control default and contagion in guarantee networks.

We have discussed the evolution of guarantee networks from the angles of network topology, regional sub-networks, and sub-patterns in previous studies. However, the structural stability is still unclear. To solve such problem, we propose a simulation model for the stability assessment of guarantee networks. Generally, we employ a simulation model to perform different types of attacks (random and targeted attacks) on the guarantee network. Our purpose is to investigate the evolution of network resilience and to determine the linkages among network stability, network topological properties, and microstructures. The simulation results corroborate that the network becomes stable during financial crisis and network stability decreases with the economic stimulus program. During the period of stimulus plan, the turning points of network stability are matched with the adjustments on monetary policies. In addition, the analysis reveals that network stability is associated with the changes on network topologies and sub-patterns. The network topologies and sub-patterns become slightly interconnected and network stability increases due to financial crisis. Conversely, the increase on degree, reciprocity, and sub-patterns of mutual and triangles with the stimulus plan enhances network fragility. Finally, we conclude that the micro-structures of edge and 2-out-star are the stable sub-patterns in the contagion process in terms of sub-patterns of guarantee network, which are consistent with the observation in Chapter 5.

    Research areas

  • complex network, guarantee network, sub-pattern, motif, contagion, stability