Common cycles: A frequency domain approach

In this paper we decompose the Serial Correlation Common Feature (SCCF) of Engle and Kozicki (1993) in the frequency domain. A collection of time series is said to share a common cycle if there exists a linear combination of the predicted series with a zero spectral density at some frequency. Estimation and inference can be performed using an Instrumental Variables (IV) approach or a Canonical Correlation Analysis (CCA). The asymptotic and finite sample properties are studied and an analysis of the comovement between Germany, Austria and the United Kingdom is presented.