Research on systemic risk in a triple network

Abstract

At present, most research on systemic risk focuses on a single interbank market network or bank-asset bipartite network, but the research on how multiple networks interact and affect the systemic risk is still rare. Therefore, this paper constructs a triple network to study the systemic risk, which includes the interbank market network, the bank-asset bipartite network, and the asset-asset relationship network. Different from previous studies using the number of bank failures to measure the systemic risk, this paper uses the DebtRank algorithm to construct the feedback effect between networks and further studies the influence of network average degree and network crowding on the systemic risk under the feedback effect of the triple network. The results show that: (i) The feedback effect between the interbank market network and the bank-asset bipartite network is the accelerator of the systemic risk. (ii) Under any three network properties, the systemic risk increases with the increase of network crowding. (iii) Under different network average degrees, the nature of triple networks is different and the magnitude of the systemic risk is different. In the case of the positive asset-asset network, the larger interbank market network, and the smaller bank-asset bipartite network, the smaller the systemic risk; however, when the asset-asset network is a non-positively relationship (negative and random relationship), if the interbank market network is large, and the bank-asset bipartite network is small, then the systemic risk is large. (iv) When the asset-asset network has a positive relationship, regardless of the size of the interbank market network and the bank-asset bipartite network, the systemic risk increases with the increase of the network average degree; when the asset-asset network has a non-positive relationship (negative correlation and random relationship), the systemic risk first decreases with the increase of network average degree, then decreases to the critical value, and then increases with the increase of network average degree. The model and analysis method of this paper provides a framework for the quantitative research of systemic risk and provides a decision-making basis for regulators and policymakers.