• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

Feature Article - Trend estimates and unusual real events in ABS Statistics


INTRODUCTION

Time series of economic statistics are often affected by the occurrence of unusual events. In recent years, examples of major unusual events have included the Sydney Olympics (October 2000) and the introduction of the New Tax System (June 2000). This article (see endnote 1) illustrates how unusual real events can impact on Australian Bureau Statistics (ABS) time series, in particular, the trend estimates. It also provides information on how unusual events are treated in the seasonal adjustment process.


SEASONAL ADJUSTMENT PROCESS

Seasonal adjustment is an analytical tool that estimates and removes the effects of normal seasonal influences from time series. Seasonal influences (including all systematic calendar effects) tend to occur regularly once or more per year. They include the effect of periodic factors such as weather, holidays, social conventions, administrative practices, and the composition of the months or quarters in terms of trading days. Typically, the ABS decomposes an original time series into trend, seasonal and irregular components. Seasonally adjusted estimates are produced by estimating the seasonal component and removing this component from the original series. The seasonally adjusted estimates reflect the combined effect of the irregular, i.e. short term variations, and the trend (defined as the remains from removing the irregulars from the seasonally adjusted estimates), i.e. the underlying direction of the series. The trend represents the fundamental direction of the series and captures the longer-term direction of the series as well as the inherent business cycles. Trend estimates are calculated by applying a smoothing procedure to the seasonally adjusted estimates. The details of ABS trend estimates method can be found in ‘A Guide to Interpreting Time Series - Monitoring Trend’ (ABS, 2003, cat. 1349.0).


WHAT ARE UNUSUAL EVENTS ?

Unusual events can be grouped into two categories according to their causes:-methodological changes and real world effects.


Methodological changes

The introduction of changes to the methodology of a collection will often cause an impact on the statistical estimates. These changes may include improvements to the scope, coverage, definitions, survey forms, sample design and estimation techniques, or any other change in the way the ABS collects the data. Examples of methodological changes affecting ABS time series include the introduction of an adjustment for new business provisions in the Retail Trade series, data reclassifications in business surveys, use of data available from the new taxation system in some economic surveys and the introduction of a redesigned questionnaire in the Labour Force Survey.

Statistical impacts arising from methodological changes could bias the original movement estimates and, if not addressed, could add noise in the measurement of real world changes. The ABS aims to maintain the continuity of important time series over periods when methodological changes are introduced to the collection. When they occur these statistical impacts will be brought to the attention of the users and, where possible, removed from the original series through an adjustment process. The methodology for the adjustment and the impact of the adjustment on the estimates is described in the relevant ABS publications.

In cases where the ABS does not seek to maintain the continuity of a time series, the ABS will usually provide information on the magnitude of the break in the series caused by the methodological change. This is normally done by using estimation techniques, although in some rare cases a parallel run may be used.


Real world effects

Real world effects are unusual events which have a significant impact on ABS statistics. Examples of real world effects include the 1989 airline pilots’ ; strike, which impacted on the Tasmanian passenger arrivals by air series; major construction projects such as the Sydney Olympic stadium in 1996, which affected the Non-residential building approvals series; the gold sell-off by the Reserve Bank of Australia in 1997, which affected the Exports of other goods series; and changes to taxation legislation, such as the introduction of the Goods and Services Tax in July 2000, which affected the Retail Turnover series.

This article will focus mainly on the impact of real world effects. Real world effects can be subdivided into those where the cause can be identified (as in the examples above) and others, which are apparent in the data, but which cannot be explained entirely satisfactorily by known events.


HOW UNUSUAL EVENTS MAY AFFECT THE ABS TIME SERIES

The seasonal adjustment procedure used by the ABS is based on the X11 method developed by the US Bureau of Census (Shiskin et. al, 1967). In this method, any large temporary or permanent change in the original time series will have a statistical impact on the trend estimates. The seasonal factors and seasonally adjusted estimates will also be affected. All of these estimates are affected to different degrees in the seasonal adjustment process:

• For the estimation of seasonal factors, the current seasonal adjustment process is relatively robust as it will discount ‘extreme’ irregular values so that the seasonal factors are not distorted in the short term. For example, for series with low volatility any extreme irregular value larger than 1.5 times the standard deviation (a quantity to measure the spread of the volatility over time) will be neglected, and the seasonal adjustment process will treat the occurrence of the extreme irregular as an outlier, replacing it with a smaller value for the purpose of estimating the seasonal factors.

• Seasonally adjusted estimates are derived by dividing the original series by the appropriate seasonal factor. When unusual events occur the seasonally adjusted estimates will include the impact from the event. However, because of the robustness of the seasonal factors, the accuracy of the seasonally adjusted series will not be compromised by the occurrence of the irregular (provided that it is understood that the seasonally adjusted number includes the effect of the real world event).

• Trend estimates are obtained by smoothing the seasonally adjusted estimates. Assuming the irregular is ‘normal’ noise in the data, the ABS smoothing method (ABS, 2003, cat. 1349.0) removes ‘normal’ irregulars and produces trend estimates. The impact of extreme one-off unusual events will produce large ‘abnormal’ irregulars in the seasonally adjusted series. As a consequence the trend estimates from the smoothing method are distorted from the real underlying movement of the series if no correction for the impact is introduced.

FIGURE 1: TASMANIAN PASSENGER ARRIVAL NUMBERS BY AIR, 1980 TO 1989.

FIGURE 2: TASMANIAN PASSENGER ARRIVAL NUMBERS BY AIR, 1980 TO 1999.

FIGURE 3: VALUE OF AUSTRALIAN NON-RESIDENTIAL BUILDING APPROVALS, 1994 TO 1999.

FIGURE 4: VALUE OF AUSTRALIAN NON-RESIDENTIAL BUILDING APPROVALS, 1994 TO 1999,
WITHOUT AN EXTREME CORRECTION AT JULY 1997 OR APRIL 1995.

FIGURE 5: RETAIL TRADE - TREND ESTIMATE BEFORE AND AFTER CORRECTION.

FIGURE 6: RETAIL TRADE - THE PRE-GST AND POST-GST TREND.

FIGURE 7: P.A.Y.E. REFUNDS, 1985 TO 1995, WITH CORRECTION FOR SEASONAL BREAKS

FIGURE 8: P.A.Y.E. REFUNDS, 1985 TO 1995, WITHOUT CORRECTION FOR SEASONAL BREAKS.