Blackwell Reference Online

Extract

Seasonality is an important feature of many economic time series. Indeed, examining the quarterly or monthly growth rates in real macroeconomic series over the postwar period, a number of studies have found that 50 percent or more of the variation can typically be “explained” by the quarter or month in which the data are observed: see, for example, Miron (1996 , ch. 3) or Osborn (1990) . Such measures should not be interpreted as necessarily implying that seasonality is fixed over time, but they do emphasize that patterns tend to repeat over successive years. Given the empirical importance which can be attached to seasonality, it is self-evident that its treatment will play a major role when forecasting quarterly or monthly economic time series. Over the last 20 years, empirical workers in economics have paid a great deal of attention to the long-run properties of time series and, in particular, to whether such series are difference stationary or trend stationary. The vast bulk of this work has been conducted in a nonseasonal context, or at least the seasonality issue has been side-stepped by the use of seasonally-adjusted data. The seminal study of Nelson and Plosser (1982) adopted the statistical unit-root test of Dickey and Fuller (1979) and concluded that U.S. time series are generally integrated of order one, denoted I (1), implying that first differences should be taken ... log in or subscribe to read full text