中文摘要：

网络模型已经成为研究银行系统性风险的重要方法。然而现有研究忽视了银行系统性风险的小概率特点，同时也缺少度量银行系统性风险的统一标准。为此，本文提出了基于网络模型的银行系统性风险度量方法：银行系统性风险VaR和银行系统性风险ES。首先，本文采用蒙特卡洛模拟方法，模拟银行外部冲击造成银行间网络损失的大样本。在银行间网络损失大样本中，估计银行系统性风险VaR和银行系统性风险ES。这两个测度能够捕捉到银行间网络损失的尾部特征，解决了对比随机冲击结果无法反映银行系统性风险的问题。其次，在模拟实验中，本文利用真实银行间网络结构参数，对模拟的三种银行间网络进行校准，保证了研究结论真实性和可靠性。最后，在模拟实验中发现：（1）外部冲击会引发违约传染的连锁反应，并导致银行间网络损失分布从近似正态分布转变成尖峰厚尾分布，最后变成双峰分布。（2）网络集中度越高发生违约传染连锁反应的概率越小，但是传染的破坏力会更大。（3）银行间网络的潜在传染作用会极大的放大银行系统的风险，而且违约传染效应是呈指数增长的。

英文摘要：

The interbank network is convenient for liquidity adjustment of the interbank market. But, the network configuration also increases risk contagion among the interbank market. However, the relationship between the network configuration and the systemic risk is a disputable issue in the research area. On one hand, the systemic risk is a typical small probability event which only happens in extreme circumstance. On the other hand, simulated interbank market network in the research area is different from real interbank market configuration. So building a realistic network model and researching the behavior of real network model in extreme circumstance are the key issues. Based on reality interbank network model, two evalua- tion parameters of the systemic risk： VaR and ES are presented in this paper. Firstly, Monte Carlo method is utilized to simulate the external impact of interbank system. Then, the systemic risk VaR and ES which can reflect the small probability characters of the systemic risk are estimated and the tail properties of in- terbank system loss are captured. Secondly, real bank parameters are utilized to calibrate three kinds of in- terbank network in simulation. Such a method ensures the reality and reliability of simulation results. Finally, three valuable conclusions are drawn： （1） External impact will trigger contagion. The in- terbank system loss will change from norm distribution to heavy tail distribution and then to bimodal dis- tribution. （2） The contagion probability of high density interbank network is smaller than that of low den- sity network, but the destruction is much higher. （3） The potential contagion will enlarge the systemic risk and default contagion effect will increase exponentially. A model which can evaluate the extent of the destruction of the systemic risk in extreme condition is presented. Furthermore, the simulation results com- prehensively reveal the relationship between the network configuration and the systemic risk.