Applications of copula-based models in portfolio optimization
Date of Award
Doctor of Philosophy (Ph.D.)
First Committee Member
Keisuke Hirano, Committee Chair
A copula is a special multivariate distribution function with marginal distributions being the standard uniform. It fully captures the dependence structure between two or more random variables. It has been shown that the assumption of multivariate normality may not be sufficient to describe the multivariate dependence between financial returns. The three essays in this dissertation show how we can use copula-based models to describe multivariate dependence between financial returns, then apply them to the portfolio optimization problems. We find that even though copulas matter from a statistical perspective, they may not be very important economically for portfolio optimization.The first essay shows how to flexibly model the contemporaneous dependence structure between financial variables, where we use stochastic volatility models to capture the marginal processes. The second essay applies copula-based models to describe the co-movements between asset returns and investigates the impact of asymmetric dependence on optimal portfolio weights. The third essay studies the economic importance of the asymmetric dependence for portfolio optimization, and pays special attention to the issue of parameter and model uncertainty.
Xu, Yue, "Applications of copula-based models in portfolio optimization" (2005). Dissertations from ProQuest. 2228.