Markov-Switching Models with state-dependent time-varying transition probabilities

Markov-switching models with covariate-dependent transition functions that are subject to exogenous discrete stochastic changes are considered. These changes are associated with simultaneous stochastic changes in the covariance structure of the observable variables. Simulation experiments are carried out to assess the quality of large-sample approximations to the distributions of the maximum-likelihood estimator and of related statistics in such a model, and to examine the implications of misspecification due to unaccounted breaks in the transition mechanism. The practical use of the model is illustrated by analyzing the relationship between Argentinian sovereign bond spreads and output growth