The trend of a time series is its long-run, slowly changing movement once short-term fluctuations are removed. Statisticians usually split a series into a trend, a seasonal component and an irregular component. Because the long-run trend and the business cycle are hard to separate, many offices estimate a combined trend-cycle: the smooth path left after seasonal adjustment and the removal of irregular noise.
Published trend estimates, such as those long produced by the Australian Bureau of Statistics, apply weighted moving averages, such as Henderson filters, to seasonally adjusted data. They show the underlying direction more clearly than raw figures, but the latest points are revised as new data arrive, because the filter has fewer later observations to work with at the end of the series, so turning points show up with a lag. Economists also estimate trends with statistical filters such as Hodrick–Prescott, and the trend in real output is closely related to potential output.
Ref 1100001849