Are VAR Models Good Enough?

VAR models are used in practice in preference to VARMA models due to the difficult issues involved in the identification and estimation of VARMA models. This paper examines if VAR models are good enough for forecasting macroeconomic variables. To answer this question, we extend the Tiao and Tsay identification procedure for VARMA models and proposes a complete VARMA modelling procedure. We then examine the properties of this identification procedure through simulation and used it to determine VARMA models for many trivariate sets of macroeconomic variables and compare the out-of-sample forecasting performance of these models against VAR models fitted to the same da