A0761
Title: A flexible predictive density combination model for large financial data in regular and crisis periods
Authors: Francesco Ravazzolo - Free University of Bozen-Bolzano (Italy) [presenting]
Roberto Casarin - University Ca' Foscari of Venice (Italy)
Stefano Grassi - University of Rome 'Tor Vergata' (Italy)
Herman van Dijk - Erasmus University Rotterdam (Netherlands)
Abstract: A flexible predictive density combination model is introduced for large financial data sets which allow for dynamic weight learning and model set incompleteness. A dimension reduction allocates the large sets of predictive densities and combination weights to relatively small subsets. Given the representation of the probability model in nonlinear state-space form, efficient simulation-based Bayesian inference is proposed using parallel sequential clustering and filtering, implemented on graphics processing units. The approach is applied to combine predictive densities based on the individual stock returns of daily observations of the S\& P500 over a large period which includes the last two (financial) crises. Substantial predictive and economic gains are obtained, in particular, in the tails using Value-at-Risk. Evidence on model set incompleteness and dynamic cluster patterns provide valuable signals for improved modelling and more effective financial policies.
