An alternative solution to the Autoregressivity Paradox in time series analysis

This note concerns with the marginal models associated with a given vector autoregressive model. In particular, it is shown that a reduction in the orders of the univariate ARMA marginal models can be determined by the presence of variables integrated with different orders. The concepts and methods of the paper are illustrated via an empirical investigation of the low-frequency properties of hours worked in the US. (C) 2011 Elsevier B.V. All rights reserved.