This paper proposes the notion of polynomial serial correlation common feat
ures as a measure of non-contemporaneous cyclical co-movements in multiple
time series. Statistical inference within this modeling is easily performed
by reduced rank regression. We show the implications of the PSCCF in terms
of the Beveridge-Nelson decomposition and we illustrate their relevance fo
r empirical analyses. (C) 2001 Elsevier Science B.V. All rights reserved.