Bayesian Inference in Cointegrated Systems

Gianni Amisano
Department of Economics, University of Brescia and University of Warwick
Amisano@master.cci.unibs.it

Abstract

This paper develops a Bayesian procedure to conduct inference on the cointegrating rank of a system of I(1) variables, and to verify the validity of over-identifying restrictions on the cointegration parameters. The model is specified in terms of a parameterization which seems to suit best this inferential problem. Exact finite sample distributions for the parameters and the statistics of interest are obtained by means of Monte Carlo integration of the corresponding conditional posterior distributions. A simulation analysis, an application on Danish and Finnish money demand data, and an application on the PPP and UIP UK data are presented.

Keywords: Cointegration, identification, bayesian analysis, highest posterior density confidence intervals, Monte Carlo integration, HAC estimation, Gibbs sampling.