1504.0 - Methodological News, Jun 2009  
Previous ISSUE Released at 11:30 AM (CANBERRA TIME) 03/07/2009   
   Page tools: Print Print Page Print all pages in this productPrint All RSS Feed RSS Bookmark and Share Search this Product

New synthetic estimates enhance the retail time series

No, it's not what you're thinking - sales of nylon, polyester and spandex have always been included in the Australian Retail time series. What is new is the synthetically estimated State-by-industry subgroup values for July - October 2008.

You may recall the article Retail Survey Redesigned to Reduce Sample; Time Series Implications Considered from the September 2008 edition of Methodological News. It outlined reductions to the sample size of the Retail Business Survey, the impacts on ABS Retail time series and the introduction of a quarterly Retail publication. While the impacts on the broad level monthly time series were small, fine State-by-industry subgroup level series exhibited strong survey design effects and increased volatility from July 2008. As a result, the release of state-by-industry series was restricted to the quarterly publication.

In November 2008, the Statistician re-instated the full sample size, citing increased scrutiny of economic statistics due to global financial developments and key users identifying more robust monthly retail data as their top priority. This meant that focus returned to the detailed monthly Retail estimates and the ceasing of the quarterly Retail publication after December 2008.

To ensure the consistency and continuity of the ABS monthly Retail time series, the Time Series Analysis section (TSA) re-estimated the fine State-by-industry subgroup level series for July to October 2008, which was a challenge because the small sample size and sample design effects made the series out of character for this period. The re-estimation was done by using historical State-by-industry subgroup data from July 1994 to June 2008 in conjunction with the November 2008 estimate. Treating the July - October estimates as missing, their values were interpolated using an application of the Kalman filter. These initial synthetic estimates then had an iterative reconciliation procedure applied to them to restore additivity up to the less affected higher aggregate marginal values. The final synthetic estimates were published in the March 2009 edition of Retail Trade, Australia (8501.0).

For more information, please contact Sam Allingham on (02) 6252 5718 or sam.allingham@abs.gov.au or Kirk Hampel on (02) 6252 5659 or kirk.hampel@abs.gov.au.