Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We study the relationship between vector autoregressive moving-average (VARMA) and factor representations of a vector stochastic process. We observe that, in general, vector time series and factors cannot both follow finite-order VAR models. Instead, a VAR factor dynamics induces a VARMA process, while a VAR process entails VARMA factors. We propose to combine factor and VARMA modeling by using factor-augmented VARMA (FAVARMA) models. This approach is applied to forecasting key macroeconomic aggregates using large U.S. and Canadian monthly panels. The results show that FAVARMA models yield substantial improvements over standard factor models, including precise representations of the effect and transmission of monetary policy.