Using Archimedean copulas, we investigate the dependence structure existing between several series of financial assets log-returns that come from different markets. These series are considered as components of a portfolio and they are investigated on a long period including high shocks. To perform such a study, we model the tail of their joint distribution function using a dependence measure (Kendall's tau) and its relationship with the class of Archimedean copulas. Then, we define two different diagnostics to decide which copula best fits the tail of the empirical joint distribution. This approach permits us to understand the evolution of the interdependence of more than two markets in the tails, that is when extremal events corresponding to shocks induce some turmoil in the evolution of these markets.