Type de document/Document Type	Thèse/Dissertation
Auteur/Author	Valdesogo Robles, Alfonso
URN	BelnUcetd-09232009-110603
Langue/Language	Anglais/English
Titre/Title	Multivariate volatility models using copulas
Intitulé du diplôme/Degree	ECON 3 - Doctorat en sciences économiques
Département/Department	ESPO/ECON - Département des sciences économiques
Jury/Advisors	Nom Titre Czado, Claudia Membre du jury/Committee Member Hafner, Christian Membre du jury/Committee Member Laurent, Sébastien Membre du jury/Committee Member Davila, Julio Président du jury/Committee Chair Bauwens, Luc Promoteur/Director Heinen, Andréas Promoteur/Director
Mots-clés/Keywords	Multivariate volatility Canonical vine copula Asymmetric dependence
Date de défense/Defense Date	2009-09-30
Résumé/Abstract In this thesis we use the notion of copulas in order to create flexible multivariate volatility models that can capture some stylized facts presented in the financial data, such as leptokurtosis, skewness and asymmetric dependence. In the first chapter we investigate multivariate regime-switching models of copulas. We provide further evidence on asymmetric dependence in international financial returns. We find that canonical vine models with asymmetric copulas perform better than models that impose symmetric dependence. These findings have important for financial implications in risk management and portfolio selection. In the second chapter we propose a new method for the construction of flexible large-dimensional copulas. This method uses the structure of canonical vines until a certain level and a multivariate copula. We show the use of this method with factors in a financial application. In the third chapter we propose a new dynamic model for volatility and dependence in high dimensions where the dependence structure is modelled with a dynamic canonical vine copula (CAVA). We show that once the stock returns are conditioned on the market and the sector returns, most of the dependence has been captured adequately. We find that many of the restrictions imposed by the Dynamic Conditional Correlation (DCC) model are not fulfilled. Moreover the CAVA model performs better than the DCC in terms of Value-at-Risk. Finally, in the fourth chapter we introduce a dynamic model of dependence based on a D-vine copula and we analyze if the dependence structure is constant over time and if it is asymmetric. We use two different data set, six exchange rates and five Asian equity indexes. We find that in both data set the dependence structure is time varying. Moreover, the dependence structure is symmetric for the exchange rates, whereas for the Asian equity indexes it is asymmetric.