CDS Spreads and Systemic Financial Risk∗

This paper investigates the information content of bond and Credit Default Swap prices of financial institutions for the measurement of systemic risk in the financial sector, defined as the probability that several institutions default. Because CDS contracts involve counterparty risk, this is reflected in their price, the spread paid to a dealer to insure against default. Then, the set of spreads of all CDSs written by each member of the financial network on the other members together with bond prices reflects the risk-neutral probabilities of default of each institution and of each pair of institutions in the network. In the paper, I show how this information can be aggregated to construct bounds on the probability of systemic events. These bounds are proven to be the tightest ones possible given this information set. Applying this method to the recent financial crisis yields allows to decompose movements in CDS spreads and bond yields in a systemic and an idiosyncratic default components. I show that the spikes observed around March and September 2008 are attributable to idiosyncratic risk, not systemic risk. Systemic risk was low until mid 2008 and increased steadily after then. Essential to this result is the use of both bond and CDS spreads: alternative measures that only look at bonds or CDSs fail to capture this decomposition, and report sharp increases in systemic risk already in 2007. ∗Contact: [email protected]. I owe special thanks to Effi Benmelech, John Campbell and Emmanuel Farhi for their many comments and suggestions. Also, I am very grateful to Ruchir Agarwal, Robert Barro, Mikkel Davies, Ian Dew-Becker, Ben Friedman, Leonid Kogan, David Laibson, Eugene Kandel, Anton Korinek, Greg Mankiw, Parag Pathak, Carolin Pfueger, Tiago Severo, Andrei Shleifer, Holger Spamann, Jeremy Stein, and seminar participants at the Harvard Macro Lunch, the Harvard Finance Lunch and the MIT Sloan Finance Lunch for many helpful suggestions.