CMStatistics 2017: Start Registration
View Submission - CFE
Title: Testing for conditional dependence between domestic indexes using nonparametric copulas Authors:  Jone Ascorbebeitia - University of the Basque Country (Spain) [presenting]
Eva Ferreira - University of the Basque Country (Spain)
Susan Orbe - University of the Basque Country (Spain)
Abstract: It is well known that the comovements between portfolios are time-varying. The interest is to detect in which level the comovements variation between domestic indexes can be explained by some global risk factors. Concretely, we will study the dependence of Euro Stoxx. Moreover, we are interested in measures of dependence beyond linear Pearsons correlation. This coefficient suits for normal variables, but not for variables with more complicated distributions such as financial variables. To overcome this fact, we propose the use of copulas to analyze the relation between domestic European indexes, conditional to the Euro Stoxx. The use of copulas allows us to model the dependence better than with elliptic distributions. We estimate conditional copulas using nonparametric methods to obtain the joint probability distribution function between indexes and examine the asymptotic properties of the nonparametric estimator. Moreover, we focus on the bandwidth choice of the conditional copula estimator and its sensitivity via a simulation study. This conditional copula method allows us to relate daily data with monthly data in a very simple manner, and therefore to relate stock indexes with macroeconomic variables such as inflation or GDP. We will measure the dependence and the conditional dependence using the Kendall's tau. We provide a statistic to test the significance of tau and its empirical distribution using jacknife.