Statistical Properties and Financial Applications of Selected Classes of Markov-Switching GARCH-Type Models

We investigate several families of univariate GARCH-type specifications in which both the unconditional dependence structure and the model coefficients evolve according to an unobserved Markov regime process. For these models we establish key probabilistic and inferential properties, and we develop a practical estimation strategy based on a suitably adapted EM-type procedure. A set of numerical experiments illustrates the accuracy and robustness of the resulting nonlinear inference.