Modelagem da volatilidade dos índices financeiros IBOVESPA, Dow Jones e Standard & Poors utilizando modelos da classe ARCH

Abstract

We examine lhe volatility process of three financial indexes Ibovespa, Dow Jones and Standard & Poors (500), using the FIGARCH (Fractionally Integrated GARCH) model and traditional ARCH (Autoregressive Conditional Heteroskedasticity) models. Four distributions was taken in to account for the returns; Nor mal, Student-t, skewed Student-t and the Generalized error dislribution(GED). The selected models were compared regarding out-of-sample forecasting accuracy and goodness-of-fit statistics. The empirical results show that ali series show signs of the levcrage cffcct for the variance and the best dcnsitics ovcrall were Student-t and the skewed Student-t. Regarding the several criterions used, the FIGARCH model with skewed Student-t outperform traditional ARCH models for the Ibo vespa, and with a Student-t, for the Standard & Poors (500) indexes. In the case of Dow Jones, the traditional GARCH model with l-Student for the returns was the best model.