A0476
Title: Copula-based clustering of financial time series via evidence accumulation
Authors: Andrea Mecchina - University of Trieste (Italy) [presenting]
Roberta Pappada - University of Trieste (Italy)
Nicola Torelli - University of Trieste (Italy)
Abstract: Understanding the dependence structure of asset returns is fundamental in risk assessment and is particularly important in a portfolio diversification strategy. When clustering time series of financial returns, it is largely recognized that pairwise association among values in the left tail of their joint distribution should be considered. To this aim, various solutions using copula models have been proposed to define dissimilarity measures based on finite (lower) tail dependence coefficients. Unfortunately, the result depends on the copula model considered and on some choices attaining the clustering procedures. A clustering approach is proposed where evidence accumulated in a multiplicity of classifications is achieved using classical hierarchical procedures and multiple copula-based dissimilarity measures. Specifically, a matrix of co-occurrences of the assets in the same cluster obtained from several partitions is derived. Such a matrix can lead to a more robust partition compared to the result from a single copula model or a specific hierarchical clustering linkage method. As a result, assets in the same cluster are expected to perform similarly during risky scenarios, and risk-averse investors could exploit this information to build a risk-diversified portfolio. An empirical demonstration of such a strategy is presented by using data from the EUROSTOXX50 index.
