Does the ABS Henderson-Trending Process Harm Forecasting Accuracy? An application Using a Selection of Australian Macroeconomic Variables


Speaker:Liam Lenten, La Trobe University, Australia
Date: Friday 3 November 2006
Time: 16:15
Location: Lecture Room D, Streatham Court

Further details

Using a structural time-series model, the forecasting accuracy of a wide range of 18 Australian macroeconomic variables is investigated. Specifically of importance is whether the Henderson Moving Average procedure used by the Australian Bureau of Statistics (ABS) to generate ‘trended’ series, distorts the underlying time-series properties of the data for forecasting purposes. Since the ABS regularly publishes both seasonally adjusted and trended data along with the original series for many variables, this is an issue of utmost importance. However, given the weight of attention in the literature to the seasonal adjustment processes used by various statistical agencies, it appears that ‘trending’ procedures have received somewhat less attention, which this study hopes to address.

An unobserved components model is utilised to generate out-of-sample forecasts for each of the 18 series, using both the trended and seasonally adjusted (since the ABS applies the procedure to the seasonally adjusted series, not the original) series. The two sets of forecasts are then made comparable by ‘detrending’ the trended forecasts, and comparing both series to the realised seasonally adjusted values. Forecasting accuracy is measured by a suite of common methods, and a test of significance of difference is applied to the respective Root Mean Squared Errors (RMSEs). It is found that, the Henderson procedure applied by the ABS does not lead to deterioration in forecasting accuracy in Australian macroeconomic variables on most occasions, though the conclusions are very different between the one-step ahead and multi-step ahead forecasts. Overall therefore, one need not necessarily hesitate to use ABS trended data for forecasting purposes.