Detecting contagion in a multivariate time series system: An application to sovereign bond markets in Europe.

Summary This paper proposes an original three-part sequential testing procedure (STP) with which to test for contagion using a multivariate model. First, conditional on breaks in the conditional mean, the procedure identifies distinct structural breaks in the volatility of a given set of countries. A further structural break test applied to the correlation matrix identifies and then dates the potential contagion mechanisms. As a third element, the STP tests for the distinctiveness of the break dates previously found. As a result of using multi-dimensional data, the STP has high testing power and is able to locate the dates of contagion more precisely. The application to European long-term interest rates shows that immediate contagion from Greece does not take place, but the dynamic spillovers are shown to increase after controlling for breaks in the different model parameters. For other countries we find evidence of both contagion and flight-to-quality mechanisms.