Abstract
The run-up to the Greek default featured marked increases in the cost of insuring sovereign debt from almost all European countries, as evidenced by their credit default swap (CDS) rates. One explanation for the perceived higher default risk in non-periphery countries is that market participants believed a default in the periphery might increase the risk of a future default in the core. To test for such dynamic contagion between credit related events in differing countries, we develop a procedure for tractably estimating high-dimensional Hawkes models using CDS prices. We escape the curse of dimensionality thanks to modelling the market portfolio of risk across countries, which serves as a sort of common factor. We further reduce dimensionality by taking a maximum-likelihood approach to estimation, avoiding having event intensities in the parameter set. Our approximation to the likelihood converges to the true likelihood as the sampling frequency goes to infinity; this ensures consistency even given non-stationarity, and permits tight identification. We find little evidence of shocks to one European country having an instantaneous effect on others, but risk in one country does gradually pull up risk in others.