Abstract
Systemic importance of a financial institution is usually measured by the impact on the banking system conditional on the insolvency of this bank and solvency of other banks. However, in reality banks encounter diverse kinds of shocks simultaneously. Therefore, the conditional results give biased estimates of financial institution systemic importance when interactions of various risks are ignored. A number of researchers propose the Shapley value method to deal with risk connections, but it suffers heavy computational cost. In our paper, we propose an analysis of variance-like decomposition method to measure systemic importance of banks in more complicated and realistic environments. This method takes both interactions and individual effects of multiple shocks into reflection and provides a more precise estimation of systemic importance. We find the proposed methodology fits well in the network models. In the numerical example, we provide a discussion of our methodology and the Shapley value method, showing the advantage of the method: Shapley value method needs running of 2(n) subsystems, while our method needs only n + 1 model runs. At last, the proposed methodology is applied to the Chinese banking system with 16 listed banks. Our empirical findings confirm that interactions of different shocks play a significant role in systemic importance of a bank, and thus, the total impact considering interactions should be adopted.