Asymptotic Properties of the Maximum Likelihood Estimator for Markov-switching Observation-driven Models

Abstract: A Markov-switching observation-driven model is a stochastic process $((S_t,Y_t)){t \in \mathbb{Z}}$ where (i) $(S_t){t \in \mathbb{Z}}$ is an unobserved Markov process taking values in a finite set and (ii) $(Y_t)_{t \in \mathbb{Z}}$ is an observed process such that the conditional distribution of $Y_t$ given all past $Y$'s and the current and all past $S$'s depends only on all past $Y$'s and $S_t$. In this paper, we prove the consistency and asymptotic normality of the maximum likelihood estimator for such model. As a special case hereof, we give conditions under which the maximum likelihood estimator for the widely applied Markov-switching generalised autoregressive conditional heteroscedasticity model introduced by Haas et al. (2004b) is consistent and asymptotic normal.