COMFORT-CCClass: A Common Market Factor Returns Model for Density Prediction and Portfolio Allocation

A new multivariate returns model is motivated and studied. By extending the CCC model in several ways, it allows for all the primary stylized facts of asset returns, including volatility clustering, nonnormality of asset returns (excess kurtosis and asymmetry), and also dynamics in the dependency between assets over time. A fast EM algorithm is developed for estimation. The predictive conditional distribution is a (possibly special case of a) multivariate generalized hyperbolic, so that sums of marginals (as required for portfolios) are tractable. Each marginal is endowed with a common univariate shock, interpretable as a common market factor, and this stochastic process has a predictable component. This leads to the new model being a hybrid of GARCH and stochastic volatility, but without the estimation problems associated with the latter, and being applicable in the multivariate setting for potentially large numbers of assets. In-sample t and out-of-sample conditional density forecasting exercises using daily returns on the 30 DJIA stocks conrm the superiority of the model performance to competing models. 

http://www.econometric-institute.org/seminars
Twitter: @MetricsSeminars