Predicting the risk of global portfolios considering the non-linear dependence structures

In this paper we estimated pair copula constructions (PCC) for three sets of markets: developed, Latin emerging and Asia-Pacific emerging. To that, we used daily prices from January 2003 to November 2011, totaling 1872 observations. The last 200 observations were separated for posterior validation of the estimated PCC. After, we constructed portfolios for each set of markets and we predicted their daily Value at Risk (VaR) for distinct significance levels, considering the dependence structure previously estimated, in the 200 days of the out-sample period. The results allow concluding that there were differences in the dependence structure of each set of markets. Further, the PCC were validated through backtesting of the predicted VaRs.