|Page tools: Print Page Print All|
FEATURE ARTICLE 1: PATTERNS IN SOUTH AUSTRALIAN RETAIL TURNOVER
From another perspective, the compound average monthly growth rate for South Australian retail trade turnover for this twelve month period was -0.02%, while the corresponding growth for Australia was +0.17%.
For the period of negative monthly growth for South Australia's retail turnover (July 2010 to January 2011), the main retail groups contributing to this decline were Clothing, footwear and personal accessory retailing, Department stores, Food retailing and Other retailing. In addition, Department stores showed negative monthly growth for each of the twelve months of this study, while Other retailing had negative or zero growth for eleven of the twelve months.
TIME SERIES ANALYSIS - INTRODUCTORY REMARKS
In the previous section the performance of South Australia's retail sector was examined for recent months using trend series data exclusively. The results indicated that retail turnover in the state had undergone a decline during the twelve months of the study. A number of questions remain; for example, why did the downturn happen? Is this an unusual event? The 'why' remains to be seen, but the uniqueness of the event can be tested by looking at historical data over a longer time span, that is, by looking at a long time-series.
As defined in the ABS information paper An Introductory Course on Time Series Analysis (cat. no. 1346.0.55.001), a time series is a collection of observations of well defined data items observed through time. In the data analysis that follows a ten year time series will be used.
The actual, or real world, representation of retail turnover comprises the original series of data; that is, the original series tells the reader what actually happened. It provides information about current levels of activity but it does not clarify whether changes from one period to the next are due to price changes, seasonal variations, or extra-ordinary events. More complete and meaningful impressions are obtained when the original data are subjected to time series analysis. Time series analysis removes from these original data seasonal influences and irregular events.
Movements in an original series of estimates can be attributed to a combination of three factors:
The trend, original and seasonally adjusted estimates contain complementary information and a complete perspective will be achieved by studying all three series. For measuring the fundamental behaviour of a series, for clarifying short-term movements, and for comparing data at different points in time (e.g. month to month, or year apart), the trend series will generally be the best analytical tool.
ANALYSING SOUTH AUSTRALIAN RETAIL TURNOVER OVER TIME
RETAIL TURNOVER - ORIGINAL SERIES
When examined for the ten year period January 2001 to January 2011, the original series of monthly South Australian retail turnover data demonstrates regular fluctuations with peaks of turnover in December falling away to low points in February each year. This pattern appears to be the main repeating sequence, that is, it represents a regular, or systematic calendar related variation - a seasonal variation. Not surprisingly, the primary seasonal influence is the Christmas season, while the February results are affected by the fewer number of trading days in that short month. A smaller, somewhat regular peak appears post-February, that is, in the March-April period, suggesting a contribution from the increased number of trading days in these two months and some influence of the movable holiday period of Easter.
For the period under study, there was an average increase of 24.7% in the retail turnovers between November and December and an average 24.0% decline in the retail turnovers between December and January.
Of particular note is the November to December 2008 increase of 29.4%, which was nearly one fifth greater than the ten year average of 24.7%. The December 2008 estimate amounted to a $421.1 million increase in turnover, easily the largest seen for the ten years of the series. The increased levels of spending estimated may have resulted from the Australian Government's stimulus packages but could also reflect other irregular impacts like changes in interest rates and other influences associated with global economic conditions. The Retail Trade series does not measure how the stimulus packages were used - they may have been spent on retail activities (possibly bringing forward some expenditure or creating additional expenditure), non-retail activities, used to reduce debt or contributed to savings.
RETAIL TURNOVER - SEASONALLY ADJUSTED SERIES
When the seasonally adjusted series is generated by distributing the seasonal patterns of the original data across the whole year, considerable modification of the original series occurs. There is still some residual/irregular activity present, but this activity appears quite minor and indications are that seasonality is the major factor influencing monthly retail turnover in South Australia. In other words, seasonality is much stronger than volatility.
The process of seasonal adjustment aims to smooth out the effects of seasonal variations arising from, say, Christmas. This adjustment to the original data series results in South Australia's peak retail turnovers for December and the troughs between December and January being ironed out.
For example, in seasonally adjusted terms, the ten year average for the change in retail turnover between November and December was +0.2% - much lower than the corresponding increase of 24.7% in the original series. Similarly, the ten year average change for the December to January period was an 0.8% increase in seasonally adjusted series, compared with an average 24.0% decrease in original series terms.
However, even after removing the seasonal effects, the retail turnovers for December 2008 and January 2009 still show a peak and a trough respectively, where changes in turnover from the previous month are a 4.0% increase to December 2008 (average is 0.2%) and a 1.3% decrease to January 2009 (average is 0.8% increase). The November to December 2008 increase in turnover ($55 million) was by far the largest for the time series.
Thus, it appears that something other than seasonality was responsible for recent (average) changes seen with the original series; for example, the Australian Government's stimulus payments to households may have had some affect on spending/turnover.
RETAIL TURNOVER - TREND SERIES
It is apparent from inspection of the original and seasonally adjusted series that both these measures of monthly retail turnover are tending to increase over time, that is, there appears to be an upward trend in the data. To arrive at the underlying behaviour of the original series, and to isolate the trend line, further smoothing operations are carried out on the seasonally adjusted series to remove residual/irregular influences. The graph below depicts the seasonally adjusted series and the derived trend series, where "bumps" in the seasonally adjusted lines can be seen to have been smoothed out in the trend lines.
After adjusting the original series of retail turnover data for seasonal effects, and for irregular factors, the underlying trend in retail turnover across time is discovered. In this format it is revealed that retail turnover in South Australia increased for most of the series. Then between December 2008 and June 2010 the underlying trend in retail turnover flattened and began to decline from July 2010.
The effects of unexpected, extraordinary events have been smoothed out in the trend series.
In recent times, the main findings in trend series retail turnover are as follows:
While the ABS provides three different perspectives of data - the original (unadjusted) series, the seasonally adjusted series and the trend series - this article highlights the benefits of using the trend series to reveal the underlying direction of the data.
From a time series perspective, the trend series levels out the irregularities in the seasonally adjusted data. It reveals that (trend) retail trade turnover in South Australia increased for most of the reference period (January 2001 - January 2011) but began to decline from July 2010.
For a more comprehensive study of time series interpretation and analysis the reader is directed to the following information papers:
ABS 1998, Information Paper: Seasonal Influences on Retail Trade (cat. no. 8508.0)
ABS 2005, Information Paper: An Introductory Course on Time Series Analysis (cat. no. 1346.0.55.001)
To clarify the activity underlying the movements in the raw (original) data a smoothing mechanism is applied. Information about the methodologies applied in smoothing raw data may be found in the following publications:
ABS 2003, Information Paper: A Guide to Interpreting Time Series - Monitoring Trends (cat. no. 1349.0)
ABS 2003, Time Series Analysis Frequently Asked Questions (cat. no. 1346.0.55.002)
ABS 2006, Australian and New Zealand Standard Industrial Classification (ANZSIC), 2006 (cat. no. 1292.0)
ABS 2011, Retail Trade, Australia, Jan 2011 (cat. no. 8501.0)
These documents will be presented in a new window.