Stable Mixture GARCH Models

A new and quite general model class for modeling asset returns and forecasting Value at Risk is proposed. It combines a dynamic multi-component GARCH structure with the stable Paretian distributional assumption. The new model nests several successful models for modeling asset returns, including the stable Paretian, mixtures of normals, normal--GARCH, stable--GARCH, and normal mixture GARCH. An extensive out-of-sample forecasting exercise confirms the superiority of the model compared to its special cases and other competitors, providing evidence that the stable Paretian assumption, in a multi-component setting coupled with a rich GARCH-type structure, is a tenable assumption for the distribution of asset returns.