A Markov-Switching Model with Mixture Distribution Regimes

Abstract

© 2018, Springer International Publishing AG, part of Springer Nature. This study proposes the mixture Markov-switching autoregressive model, which allows variation in error distribution across different regimes. This model is generalized from the ordinary MS-AR model owing to two considerations, but related to each other. First, we have concern about the mixture of distributions or populations, which often prevails in economic time series. Second, when using the MS models to analyse economic fluctuation, we doubt if each regime in the model can have distinct distribution. All of these concerns are addressed by an empirical study.