Prediction of Financial System Risk Based on Complex Network

Financial systemic risk can be characterized by coexceedances in asset prices in financial markets. This paper proposes a method to determine the lead-lag effect of stock exceedances under a complex network framework and uses the leading stocks to predict the systemic risk occurring in the lagging stocks, which is identified as the exceedance of a composite index in a financial market. As the high-frequency trading data suggests, coexceedances are unlikely to occur simultaneously. This method applies a nonlinear measure of event synchronization to determine the coexceedances between pairwise stocks and explores whether there exist significant lead-lag effects of stock pairs, which are used to construct an adjacency matrix as the basis for building a complex network with stocks as nodes. Combined with the properties of the complex network, constructing an index to predict the possibility of systemic risk of lagging stocks based on the exceedances of leading stocks becomes achievable. The data of constituent stocks in the Shanghai Stock Exchange in China’s stock market is used to conduct empirical research. The results show that in the prediction interval for exceedances provided by the leading stocks, the method proposed can generally predict more than 70% of the systemic coexceedances. It is also proved to be robust and can be well applied to capture major risks in the market