posted on 2023-06-16, 05:17authored byGeorge Athanasopouolos, Don Poskitt, Farshid Vahid, Wenying Yao
This article studies error correction vector autoregressive moving average (ECVARMA) models. A complete procedure for identifying and estimating EC-VARMA models is proposed. The cointegrating rank is estimated in the first stage using an extension of the non-parametric method of Poskitt (2000). Then, the structure of the VARMA model for variables in levels is identified using the scalar component model (SCM) methodology developed in Athanasopoulos and Vahid (2008), which leads to a uniquely identifiable VARMA model. In the last stage, the VARMA model is estimated in its error correction form. Monte Carlo simulation is conducted using a 3-dimensional VARMA(1,1) DGP with cointegrating rank 1, in order to evaluate the forecasting performances of the EC-VARMA models. This algorithm is illustrated further using an empirical example of the term structure of U.S. interest rates. The results reveal that the out-of-sample forecasts of the EC-VARMA model are superior to those produced by error correction vector autoregressions (VARs) of finite order, especially in short horizons.