Two Bayesian sampling schemes are outlined to estimate a K-state Markov switching model with time-varying transition probabilities. Data augmentation for the multinomial logit model of the transition probabilities is alternatively based on a random utility and a difference in random utility...