Modeling Multivariate Revisions in a Cointegrated Vector Autoregressive Model

• Activity Number: 563
• Type: Topic Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Riccardo Gatto*+ and Gian Luigi Mazzi and Alain Hecq
• Companies: Eurostat and Eurostat and Maastricht University
• Keywords: Data revision ; Co-movements ; ARMA ; EU business cycle indicator

Abstract

After extracting the most important features of different vintages composing the data revision process, it emerges that the nonstationarity and the presence of autocorrelation are the two most dominant characteristics of these time series. We look whether these features are common to several variables, by carrying out a common trend/common cycle analysis with the goal to finding the timing with which an indicator is close enough to the "real thing". We propose two ways of splitting a real-time data set: considering successive vintages of the same phenomenon and looking at real-time releases. We try also to answer the three questions do different vintages (diagonals and verticals in our study) share the same long-run movements? Do different vintages share the same short-run movements? Should we use one or a combination of different vintages? We show that series coming from the same multivariate system and sharing co-movements must have the same parsimonious individual ARMA representation. This shall help us in selecting the best combination of vintages. An application on real EU monthly industrial production index. is presented.