Abstract: Based on the gravity model, this paper tries to recognize the spatial network relationship of provincial local government bond risks. It employs the social network method to analyze the spatial network structure characteristics and the dynamic evolution trend of the bond risks during the debt swap period and empirically examines the spatial convergence of the bond risks. The results show that: (1) the local government bond risks present a rather typical“scale-free network”structure, and the network contagion mechanism has a“small world phenomenon”; (2) local government bond risks have a stronger spatial network connectivity, correlation and accessibility, and they are dynamically strengthened year by year; (3) local government bond risks present obvious characteristics of cohesive subgroup structure, and the connected effect varies among different subgroups; (4) the bond risks of different provinces have a large difference in the“power”status of infection in the spatial network, and the ability of risk transmission among them is relatively stable; (5) the local government bond risks have significant convergence characteristics, and the convergence speed is different under different circumstances, indicating that the effect of the debt swap policy is more obvious. Therefore, it is necessary to prevent bond risks from such aspects as strengthening the coordinated management of bond risks across provinces, preventing bond risk contagion among regions, enhancing early warning of bond risk contagion in different regions, promoting classified bond risk management in different provinces, further promoting bond risk mitigation, and so on.

Key words: local government bond risk, spatial correlation network, the method of social network analysis, risk contagion