Topic: Contributed on Co-movements in macro and finance time series Title: Bayesian cointegration using a singular distribution on the long-run relations matrix Authors:  Basile Marquier - University of Sheffield (United Kingdom) [presenting]
Abstract: The focus is on new methods of cointegration to establish convergence of economies, their performance, co-evolution and long-run relationships. Multivariate cointegration methods have dominated the econometrics literature over the past 20 years. They offer a framework of identifying relationships of financial assets or exchange rates or more generally of financial time series, hence they are exploited in developing long term decision making, trading and portfolio management. Over the last decade Bayesian methods have been developed for the purpose of estimating the parameters of the well established vector error correction (VEC) model and thus discovering cointegrated time series. By using Markov chain Monte Carlo methods, we propose a Gibbs sampling scheme for estimation of the parameters of the VEC model. The long-run relationships matrix is a square matrix of low rank, called the cointegration rank. This rank is the number of independent cointegration relations between the economic time series. A critical step in the analysis of the cointegration rank will be to try to avoid reliance upon Johansen tests. A singular normal prior is set over the long-run relations matrix. The proposed methodology is illustrated with a panel data of economic and financial variables and a study of the cointegrating coefficients eventually provides information about positive and negative comovements.