1504.0 - Methodological News, Mar 2008  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 31/03/2008   
   Page tools: Print Print Page Print all pages in this productPrint All

Changing the Methodology for Measuring the Impact of Introducing a New Industry Classification for Sub-annual Surveys

An update in the Australian and New Zealand Standard Industry Classification (ANZSIC) from the 1993 version to the 2006 revision is likely to cause a statistically-induced change to the level and seasonal pattern of estimates for a number of ABS sub-annual business and labour employer surveys. These impacts would occur for sub-annual surveys in the September quarter of 2009, when samples designed on an ANZSIC06 basis are introduced. In order to be able to produce relevant statistics and a consistent series, it is important to be able to explicitly measure the impact of the change, then account for (backcast) this impact in a revised series such that users' interpretation of the statistics is unaffected.

Changes in circumstances have caused the strategy to measure the impact of the ANZSIC update to be revised. The original generic strategy to measure these impacts for sub-annual business and labour employer surveys involved parallel estimation. This meant producing estimates on both ANZSIC93 and ANZSIC06 bases over a five-quarter period from June quarter 2008 to June quarter 2009 inclusive, using existing samples designed for ANZSIC93 outputs but topped up where necessary to produce more reliable estimates based on ANZSIC06. A parallel estimation strategy was chosen as it is a relatively cost-effective way of measuring impacts on both the levels of a series and the seasonality of estimates.

A number of other changes planned for 2009 have necessitated rethinking the impact measurement methodology. These changes include:

  • changes resulting from revision 1 of the System of National Accounts 1993 (SNA93), particularly changes in Standard Economic Sector Classifications of Australia (SESCA) and Standard Institutional Sector Classifications of Australia (SISCA);
  • an update of frame size benchmarks used for stratification purposes;
  • a realignment of Government Finance Statistics units on the Business Register;
  • changes in scope to some collections, e.g. the introduction of non-employing businesses in the Quarterly Business Indicator Survey; and
  • redesigns of samples to take advantages of efficiencies possible from all of the above changes.

Much of the information from these changes will not be available on survey frames until late 2008 or early 2009. As a result, sub-annual collections will now generally be measuring the statistical impact of all these changes using a parallel run methodology, i.e. for (at least) one quarter, generally the June quarter of 2009, a parallel sample will be selected from a frame stratified on the new basis. Data will be required from more businesses under the parallel run methodology than under the parallel estimation methodology, albeit for fewer reference periods, as data will need to be collected from both the old sample based on ANZSIC93 and the new sample based on ANZSIC06, updated benchmarks, SISCA etc. Two sets of estimates will be produced, one on each basis, with the impact of the changes being the difference.

The new parallel run strategy will be more effective than parallel estimation at being able to capture impacts on the level of estimates due to all the changes, including effects from a substantial sample rotation which will occur as a result of redesigned samples. A weakness of the new strategy is the reduced parallel period, which means any changes to seasonality of estimates will be very difficult to determine. We are exploring an alternative 'modelling' strategy using different data sources such as estimates from the completely enumerated sector of samples and taxation data from Business Activity Statements, and will establish an appropriate methodology to assess changes in seasonality.

Survey areas affected by the change are now in the process of updating their plans to take account of the new strategy, with particular focus on ensuring systems and processes are ready to cope with the parallel run, as well as developing plans to quality assure the parallel processes and outputs.

For more information please contact Justin Farrow on (02) 6252 5795 or Paul Schubert on (02) 6252 6591.