WHAT SEASONAL ADJUSTMENT TOOLS DOES THE ABS USE?
The main tool used in the ABS is called SEASABS (SEASonal analysis, ABS standards). SEASABS is a software package that applies successive moving averages using the X11 processing method to produce seasonally adjusted estimates. SEASABS is available free to most organisations on signing an agreement.
WHAT IS A MOVING AVERAGE?
Moving averages successively average a shifting time span of data in order to produce a smoothed estimate of a time series. This smoothed series can be considered to have been derived by running an input series through a process which filters out certain cycles.
WHAT IS THE END POINT PROBLEM?
It is not possible to use a symmetric filter at either the beginning or the end of a time series as there are not enough time points. This is one problem that leads to the revision of time series estimates.
WHAT MODELS CAN BE USED TO DECOMPOSE A TIME SERIES?
Time series of original estimates are decomposed into seasonal, trend and irregular components through the application of decomposition models. These models are usually additive or multiplicative, but can also take other forms such as pseudo-additive. We usually choose the one that yields the most stable seasonal component, and leaves the residual noise component in an appropriate form.
WHAT ARE STOCK AND FLOW TIME SERIES?
Time series can be classified into two different types: stock and flow. Stock series are a measure of certain attributes at a point in time and can be thought of as "stocktakes". Flow series measure a volume of activity over a given period. For example, the total profits taken in a month is a flow measure.
WHAT ARE MOVING HOLIDAY EFFECTS?
Moving holiday effects are calendar related effects caused by holidays which occur each year, but whose exact timing shifts in the Gregorian calendar. Examples are Easter and Chinese New Year.
WHAT ARE TRADING DAY EFFECTS?
Trading day effects are calendar related effects related to the number of occurrences of each of the days of the week in a given month. For example having 4 weekends in March in 2000, but 5 weekends in March 2002 would cause higher turnover at a public swimming pool for March 2002.
WHEN IS SEASONAL ADJUSTMENT INAPPROPRIATE?
Any series that contains seasonal patterns should be seasonally adjusted. However, if your data is very noisy or has weak seasonal patterns it may be difficult to seasonally adjust. In this case the seasonal patterns will be harder to identify and seasonal adjustment will be less reliable.
WHAT IS DIRECT AND INDIRECT SEASONAL ADJUSTMENT?
Often time series are related in an aggregative way. For example, estimates in individual states should sum to the national total. When seasonally adjusting aggregate time series we can either adjust the aggregate (e.g. national) series directly or sum the seasonally adjusted estimates derived from the component (e.g. state) series to get an indirectly seasonally adjusted total.
WHAT IS CONCURRENT SEASONAL ADJUSTMENT?
Concurrent adjustment is a seasonal adjustment process in which updated estimates of the seasonal pattern are prepared each month using all of the available data. Concurrent seasonally adjusted estimates are much faster at identifying any changes to the seasonal pattern and will usually lead to more appropriate seasonally adjusted and trend estimates than a forward factor adjustment in which the seasonal pattern is only estimated once per year.
WHAT DO I NEED TO KNOW TO DO SEASONAL ADJUSTMENT MYSELF?
For each series, you need to choose between a few different decomposition models, and choose between seasonal filters and between trend filters. You need to recognise large extremes, trend breaks, and seasonal breaks, and you may need an understanding of other effects like Trading Day and Moving Holiday corrections. You can study the ABS publications listed below or attend a nearby offering of "Understanding Time Series", where you can learn more and discuss your data with an ABS Time Series expert.
WHERE CAN I FIND OUT MORE ABOUT SEASONAL ADJUSTMENT?
The ABS produces a number of publications providing in-depth reference material, such as
1. Information Paper: A Guide to Interpreting Time Series – Monitoring Trends, 2003 (cat. no. 1349.0)
2. Information Paper: An Introductory Course on Time Series Analysis – Electronic Delivery, Jan 2005 (cat. no. 1346.0.55.001)
3. Time Series Analysis Frequently Asked Questions, 2003 (cat. no. 1346.0.55.002)
All these are available free from www.abs.gov.au.
For further questions, or to enquire about data analysis or training opportunities such as "Understanding Time Series", contact firstname.lastname@example.org or visit www.abs.gov.au.