Decomposition of Data into Cyclical and Growth Components

Abstract

Time series analysis often assumes zero-mean processes. To accommodate this assumption, non mean-zero components, especially secular growth of time series must be removed before we can begin. More often, economic time series are decomposed into three components; seasonal components, secular growth components and cyclical components or fluctuations about the secular growth paths. Structural information is then extracted from the remaining cyclical components or fluctuations about the growth paths to help predict future fluctuations or to discern patterns of cyclical co-movements of elements making up the time series, or better to characterize business cycles, and so forth.