On the dependence structure of realized volatilities

Volatility plays an important role when managing risks, composing portfolios, and pricing financial instruments. However it is not directly observable, being usually estimated through parametric models such as those in the GARCH family. A more natural empirical measure of daily returns variability is the so called realized volatility, computed from high-frequency intra day returns, an unbiased and highly efficient estimator of the return volatility. At this time point, with globalization effects driving markets' volatilities all over the world, it becomes of great interest to assess volatilities' co-movements and contagion. To this end we use pair-copulas, a powerful and flexible statistical model which allows for linear and nonlinear, possibly asymmetric forms of dependence without the restrictions posed by existing multivariate models. Given the importance of the Brazilian stock market in the Latin America, in this paper we characterize the dependence structure linking the realized volatilities of seven Brazilian stocks. The realized volatilities are computed using an 8-year sample of 5-minute returns from 2001 through 2009. We include a more comprehensive study involving seven emerging markets, addressing the issue of contagion in a more general scenario.