Bayes Analysis of Partially Cointegrated VAR Systems with Markov Regime Switching

This paper introduces Bayesian inference in a Markov switching partial cointegration model. The partial cointegration allows the cointegration relationships to be switched on and off depending on the regime, unlike conventional cointegration analysis that assumes linear adjustment toward equilibrium takes place at all times. The assumption of global linear adjustment is likely to be inappropriate for variety of economics situations (e.g. presence of transaction costs, adjustment costs, or policy interventions) and so that it is more appropriate to consider cointegration models in which cointegration vectors are subject to change with regime. The model can be estimated by using Bayesian techniques with numerical integration such as Gibbs samplers. We use the natural conjugate priors to derive the posteriors in order to computer the Bayes factors, which are used for the determining the cointegration rank and models specifications.