Model and distribution uncertainty in multivariate GARCH estimation: a Monte Carlo analysis

摘要

Multivariate GARCH models are in principle able to accommodate the featuresof the dynamic conditional correlations processes, although with the drawback, whenthe number of financial returns series considered increases, that the parameterizationsentail too many parameters.In general, the interaction between model parametrizationof the second conditional moment and the conditional density of asset returnsadopted in the estimation determines the fitting of such models to the observed dynamicsof the data. This paper aims to evaluate the interactions between conditionalsecond moment specifications and probability distributions adopted in the likelihoodcomputation, in forecasting volatilities and covolatilities. We measure the relativeperformances of alternative conditional second moment and probability distributionsspecifications by means of Monte Carlo simulations, using both statistical and financialforecasting loss functions.