Research On Financial Systemic Risk Based On Information Theory

The financial system is a typical nonlinear complex system.The research field on financial systems based on the theories of complex systems,applied mathematics,statistical physics,nonlinear science,etc,has gradually become more and more popular,which is called econophysics.However,since the 21 st century,China's financial systems have suffered some violent shocks,such as the subprime mortgage crisis in the USA and the COVID-19 outbreak.Due to the growing number of subsystems and complexity of financial markets,it is difficult to describe systemic risk using economic models.This situation poses many challenges to risk identification and control.Therefore,how to identify and warn financial systemic risk has become the focus.This thesis uses related methods of information theory and the theories of complex systems to analyze information flow and structural evolution characteristics in the financial market under different risk levels,aiming to find out some indicators that can effectively characterize financial systemic risk.The main studies are as follows:Taking the subprime mortgage crisis in 2008 as the background,apply Transfer Entropy(TE)to process one-minute high-frequency return data of stocks of A-shares in Shanghai and construct 726 information flow networks based on a fixed threshold,then analyze these networks which are divided into different crisis periods by the theories of complex systems.The results show that:(1)The information flow measured by TE can effectively reflect the information interaction of the stock market and the change of risk.Specifically,the stock market generates more information flow when intraday stock prices are falling and when the market falls sharply.(2)In the crash period,topological structure parameters of networks are higher such as the network clustering coefficient,network structure entropy,and so on,and there are obvious important nodes in networks.(3)According to the characteristics of the industry networks based on the coarsening of stock networks,the stock market tends to exchange information around some industries.These core industries mainly involve food,clothing,housing,and transportation.And the anti-risk ability of the industry networks is gradually increasing.Based on the background of the COVID-19 sweeping the world in 2020,employ the sliding window method and TE to process daily return data of stocks of A-shares in Shanghai.The correlation between the information flow generated by stocks and the number of new cases of COVID-19 is quantitatively measured by mutual information.Then,build the directed maximum spanning tree networks and analyze systemic risk for each period of the epidemic.The main results are:(1)The impact of COVID-19 on financial markets appears in the early outbreak period with a rapid increase in information flow and the incoming information flow is more sensitive to changes in the epidemic.(2)In terms of the topological structure of the networks,the spread of systemic risk has higher infectious efficiency and faster infectious speed during the severe period of the epidemic and the initial period of changes in the stock market.(3)The clustering of stocks belonging to the same industry is susceptible to policy impact.In the severe epidemic period,the clustering of nodes belonging to the pharmaceutical and biological industry is more obvious than in other periods.(4)In the severe period of the epidemic,the stock market is more resistant to random attacks but less able to resist deliberate attacks.