Autoregressive Spectrum Estimation Technique Allied to Quarterly Consumer Durables Expenditure Data

Classical spectral techniques can provide sharp insights into the cyclical patterns in a time series of economic data. Various problems in the application of classical spectral techniques, such as the choices of smoothing routine and bandwidth and the appearance of end-effects, inhibit the usefulness of spectral analysis. Alternatively, an autoregressive spectral technique does not share these problems, but does present the difficulty of the choice of the order of the autoregression. This paper applies classical and autoregressive spectral techniques to quarterly consumer durables expenditure data, discusses three approaches to the choice of the order of the autoregression, and compares the results of the different spectral techniques. Autoregressive spectral analysis provides a superior representation for this time series.