3

This paper analyzes the nonlinear interdependencies of stock market indices in the eurozone. We introduce an original dynamic model, which is a combination of the bivariate noisy Mackey-Glass model of Kyrtsou and Labys (2006), extended in a multivariate framework, with DCC-GJR-GARCH errors. The empirical application conducted on seven stock market indices in the eurozone between November 28, 2003 and November 25, 2012 highlights several mechanical interdependencies. The relationships detected in the mean equation confirm the leading role of the French and German indices, an organization of Southern European countries around the Italian index, and the isolation of the Greek and Irish indices. The psychological interdependencies revealed by the DCC-GJR-GARCH specification show two particular groups: the first is composed of the ATHEX, CAC, ISEQ, IBEX, and DAX indices; the second concerns the Italian and Portuguese indices. We find strong correlations for the Franco-German couple, while we note a decorrelation in the first group of the CAC, DAX, ISEQ, and IBEX indices with the Greek index following the sovereign debt crisis. The other indices are characterized by unstable psychological interdependencies over the study period.