Refining the Distribution of GARCH Models: Application to Stock Indexes Returns

Abstract

GARCH models have been extensively used in risk modeling under the normal distribution. Although they generate highly significant coefficient estimates, these models are known to have poor forecasting power. It is therefore interesting to develop a different approach of risk modeling to improve forecasting results. By using the generalized t-distribution in modeling the changes in the distribution of stock indexes returns, the results show a significant improvement in the forecasting power. Moreover, Monte Carlo simulations have confirmed that the indexes returns are better explained by ARCH-type models.

Keywords: generalized t, GARCH, forecast, simulation, index return

JEL Classification: G12, G15, C12, C13, C15, C16, C22