Markov-switching models with endogenous explanatory variables

摘要: The maximum likelihood estimation of a Markov-switching regression model based on the Hamilton filter is not valid in presence endogenous explanatory variables. However, we show that there exists an appropriate transformation allows us to directly employ filter. transformed explicitly for vector bias correction terms as additional regressors, and new disturbance term uncorrelated with all regressors model. Within this framework, quasi procedure presented. A test endogeneity Wald statistic or ratio also