Adapting the Litterman Prior for Cointegrated VARs

Abstract

The paper presents a novel prior for Bayesian VAR models, characterized by explicit modelling of cointegration that avoids certain unattractive restrictive properties of the priors used previously. The potential of the prior for easy elicitation from the well-established Litterman beliefs is demonstrated. An efficient procedure for sampling from posterior distribution is provided. The favourable outcome of the forecast comparison exercise gives further support for the use of the methods proposed.