CHANGES TO THE RETAIL BUSINESS SURVEY OCTOBER 2013
This issue of Retail Trade, Australia (cat. no. 8501.0, October 2013) introduces some changes to the sample design and methodology of the Retail Business Survey. This is the first sample redesign since the September quarter 2009 when the survey was updated to use the 2006 edition of the Australian and New Zealand Industrial Classification (ANZSIC).
CHANGES TO SAMPLE DESIGN AND METHODOLOGY
Business Activity Statement (BAS) information from the taxation system is used both as a sizing variable for stratification purposes and to form auxiliary information (estimation benchmarks) to support the regression estimation methodology used in the Retail Business Survey. The utilisation of BAS information enables the most efficient design for the survey, keeping sample sizes to a minimum while providing accurate results. The sample design of the Retail Business Survey imposes tight sampling error constraints at the state/territory and broad industry levels. For lower level estimates, sampling error can be larger.
Prior to the October 2013 issue the annualised turnover benchmarks allocated to each business for the purposes of stratification have not been updated every quarter. Keeping the stratification benchmarks static generally reduces the volatility of the movement estimates. However, over time it also reduces the accuracy of the level estimates, particularly when businesses are changing rapidly in size. As a result, the stratification benchmarks gradually "age" over time and the sample may become less representative of the target population. It is now four years since the stratification benchmarks were last updated and there has been significant change in the size of businesses on the frame during that time.
From the October 2013 issue onwards, the annualised turnover benchmarks used for stratification purposes have been updated to the most recent BAS information. In the future, the stratification benchmarks will be updated every quarter. This methodological change will improve the accuracy of level estimates derived from the survey as well as addressing the issue of aging stratification benchmarks which must otherwise be periodically updated. The implementation of new methodology for the October 2013 issue has been accompanied by changes to the stratification boundaries and sample allocation. This ensures the sample is spread optimally across the available size strata to reflect the current turnover distribution in the target population.
HOW THE CHANGES IMPACT ON RETAIL TRADE SERIES
The changes to stratification boundaries and sample allocation have been implemented through the selection of a new sample for October 2013. The new sample retains as many existing units as possible which are redistributed on the basis of their refreshed sizing information. However, a significant number of new units have been included which replace others rotating out.
As there has been no change to the scope of the Retail Business Survey, the sample redesign will have no long-term impact on the level of the estimates in this publication (i.e. publication industries and state/territory totals). However, the unusually high level of sample rotation and redistribution in the implementation month will have resulted in the movement estimates for October 2013 being more volatile than usual. In addition, lower level time series may be impacted by a short-term shift in level.
In subsequent issues, the ABS will estimate the size of any level shifts and consider revisions to historical series to facilitate comparisons over time. The impact of changes on levels may be estimated from the first two cycles which use the new methodology (i.e. the October and November issues).
The Retail Trade series will be monitored over the coming months as more data becomes available using the new sample. Should revisions to the historical series be required to smooth level shifts back into the monthly and quarterly estimates, these revisions will be applied in the December issue to be released on 6 February 2014.
IMPACT ON COMPLETELY ENUMERATED SECTOR ESTIMATES
The changes have resulted in a break in time series between September 2013 and October 2013 for completely enumerated sector and sample sector estimates. The completely enumerated sector comprises 'large' businesses that are in the survey each month. This grouping is not fixed and can change over time, though it tends to remain relatively constant from month to month. However, with the introduction of changes to the stratification boundaries and sample allocation for October 2013, the size of the completely enumerated sector has been increased, offset by a reduction in the size of the sample sector. Therefore, care should be taken when interpreting monthly and annual percentage changes for completely enumerated and sample sector estimates which include September 2013 and October 2013 reference periods.
Seasonally adjusted and trend estimates for the total completely enumerated sector are currently included in Table 16 from the Downloads tab. The seasonally adjusted estimate is currently rising 3.3% ($477.5m). The ABS estimates that the impact of methodological changes on this series are approximately $400m. As a result of this impact, a break in the monthly trend series has been inserted between September and October 2013. Therefore, care should be taken in comparing the series over time. For further information, refer to Information Paper: A Guide to Interpreting Time Series - Monitoring Trends, 2003 (cat. no. 1349.0).
IMPACT ON THE CHAIN VOLUME MEASURES
As no revisions have been applied yet to the historical Retail Trade series in current prices, there has been no subsequent impact on the quarterly chain volume measures. Should revisions be applied in future issues, the chain volume measures will also be recalculated on the new basis.