On the Correspondence between Data Revision and Trend-Cycle Decomposition

Abstract

This paper places the data revision model of Jacobs and van Norden (2011) within a class of trend-cycle decompositions relating directly to the Beveridge-Nelson decomposition. In both these approaches identifying restrictions on the covariance matrix under simple and realistic conditions may produce a smoothed estimate of the underlying series which is more volatile than the observed series.JEL classification: C22, C53, C82

Multivariate decompositions and seasonal gender employment

“…An important strand of literature employs so-called unobserved components (UC) models to study trends and cycles as separate phenomena of economic interest. While conventional UC models, including those underlying seasonal adjustment, assume that the individual components are uncorrelated, a substantial literature questions this in the context of analyzing trend and cyclical movements in seasonally adjusted data; important contributions include Clark (1989), Morley, Nelson and Zivot (2003, henceforth MNZ), Morley (2007), Sinclair (2009), Dungey et al (2013Dungey et al ( , 2015. Very recently, Hindrayanto, Jacobs, Osborn and Tian (2019, henceforth HJOT) extend the analysis to include a seasonal component, showing (in the univariate context) that the assumption of uncorrelated trend, cycle and seasonal components can be questionable.…”