Bayesian Analysis of Stochastic and Deterministic Processes in The Error Correction Model

In this article a method for joint estimation of the number of stochastic trends and the deterministic processes in a multivariate error correction model is presented. This approach takes advantage of the Laplace method of approximating integrals and, the second important contribution of the paper, careful elicitation of the prior for the cointegrating vectors from a prior on the cointegrating space. The approach follows the classical approaches of James (1969), Anderson (1951) and Johansen (1988 and 1991) and performs well when used to estimate the number of stochastic trends compared with information criteria in finite samples in Monte Carlo experiments.