Learning Network Structure of Financial Institutions from CDS Data

Default contagion is a concern especially after witnessing the financial crisis. How the financial institutions are connected is crucial to how likely they will default together. However, the information on financial network structure is not available. This paper builds a probabilistic graphical model relating the network structure to observable prices. The financial network structure can be learned using CDS data. Conditional default probability and expected conditional loss can be calculated accordingly. As the information is extracted from the CDS price, all quantities are under the risk-neutral measure, and thus account for the market fear that can trigger bank run. This knowledge can be used to gauge systemic risk and inform policy decisions