Markov Switching Smooth Transition GARCH Model

Abstract

A Markov switching asymmetric GARCH model which imposes more leverage effect of the negative shocks is considered. The asymptotic behavior of the second moment is investigated and an upper bound for it is calculated. A bayesian strategy through Gibbs and griddy Gibbs sampling is used to estimate the parameters. Finally we study the performance of the model by two real data sets. We show that this model has the best in-sample fit via DIC and provides a better forecast when the negative skewness is large enough.