Title: Dynamic mixed frequency pooled copula
Authors: Audrone Virbickaite - Colegio Universitario Estudios Financieros (Spain) [presenting]
Hedibert Lopes - INSPER (Brazil)
Abstract: Modeling dependence between financial returns through copula by exploiting the information available from daily and intraday data is proposed. The two alternative copula models, obtained from low and high frequency data, are combined via density pooling approach and show superior performance in terms of sequential predictive log Bayes factors. The intraday copula is estimated via two alternative specifications: hierarchical and mixture model. While hierarchical model is the best in class of models that are based only on high frequency data, incorporating low frequency information provides additional gains in predictive model performance.Proposed pooled model is applied to 3-variate and 10-variate daily log returns of financial assets, traded at the NYSE.