|Page tools: Print Page Print All|
FEATURE ARTICLE 3: THE 2006-2011 INTERCENSAL PERIOD AND REVISIONS TO HISTORICAL ERP
WHY REVISIONS ARE NECESSARY
In compiling the preliminary population estimates based on the 2011 Census, ABS introduced a key innovation which improved the quality of ERP. The introduction of Automated Data Linking (ADL) into processing of the 2011 Census Post Enumeration Survey (PES) resulted in an improved, and lower, measure of net undercount than would have been derived had the same methodology used in processing the 2006 PES been used. In other words, the 2011 PES found that more people were actually counted by the Census than would have been suggested by the previous PES methodology.
As a result of this methodological change, the undercount adjustment used in rebasing the 2011 population estimates was also considerably lower than the undercount adjustment that was applied in the rebasing of the 2006 population estimates. ABS was able to estimate from a statistical impact study that if the ADL methodology and supporting technology had been available during the processing of the 2006 PES, the 30 June 2006 population estimates would have been lower by more than 200,000 people.
Given the different methodologies that were used in the processing of the 2006 PES and 2011 PES, the intercensal error for the 2006-2011 period is 294,000, which is around three times higher than the intercensal errors that have been measured historically.
The conventional demographic treatment for intercensal error is to spread the total amount evenly through the series for the previous 5 years, as the error is usually assumed to have accumulated over 5 years and cannot be attributed to a particular source. This conventional treatment resulted in a downward revision of population growth over the 5 year period 2006-2011, from 1.8% (average annual growth) that had been indicated in the population components of births, deaths and migration since 2006 to 1.5%.
This large reduction to growth created a challenge to users' understanding of historical population growth. The reduction implied growth rates which are artificially too low, because:
1. they are the result of an ABS change in method (ADL); and
2. they are in stark contrast to the growth implied by an assessment of components of growth.
These challenges were recognised by the ABS, and general advice on how to best use the preliminary estimates was provided in a feature article titled 'Advice on the use of 2011 Preliminary Rebased ERP' in Australian Demographic Statistics (cat. no. 3101.0) on 27 September 2012.
For these reasons, in the 2011 final rebased ERP release on 20 June 2013, a different treatment is required to the conventional demographic treatment for intercensal error. It is necessary to adjust historical ERP to reflect the different undercount adjustments that would have been used had the 2011 PES methodology been available in earlier years.
CONSULTATION PROCESSES WHICH HAVE INFORMED THE ABS TREATMENT
Since the 20 June 2012 release of Australian Demographic Statistics (cat. no. 3101.0), the ABS has extensively engaged with users of ERP to inform the process of reviewing the treatment of intercensal error ahead of the release of final rebased population estimates on 20 June 2013.
This process culminated in an intensive phase of consultation from the 27 September release through to 27 November 2012 following the publication of an article articulating the ABS intention to revise historical ERP data.
ABS sought and received a number of key perspectives from members of a number of fora, including the ABS Population Estimates Technical Workshop, the Australian Statistics Advisory Council, and the ABS State Statistical Forum. Invitations were also sent to stakeholders to submit written submissions. In addition, ABS held a wide range of meetings with users of ERP.
During this consultation process, the ABS heard consistently from users that a high quality ERP which provides coherent information on both population levels and growth is essential for informed decision making, policy development and evaluation and other planning processes.
Most feedback received through this process supported a 20 year treatment of intercensal error for final rebasing to ensure that population growth in the period up until 30 June 2011 most closely reflected the best estimate of historical growth; that reflected in the population components of births, deaths and migrations statistics.
There was wide support for principles (see below) to be developed to guide the consideration of a methodological response, in particular that any revisions should be based upon the best data available to ABS.
Some organisations indicated support for a 10 year treatment and voiced concerns about the strength of the evidence base back to 1991. There was also some concern about perceived differences in coverage between the Net Overseas Migration (NOM) component of population growth compared with the growth reflected in the Census-based population estimates, particularly for the overseas-born population. The ABS has confirmed through recent analysis that this is not a major contributor to the large intercensal error, and will publish further information on this issue in early 2013.
One organisation also cautioned the ABS against making any revisions and departing from the conventional treatment of intercensal error.
ABS gave very careful consideration to all of these perspectives before making its decision to proceed with revising historical ERP over a 20 year period.
PRINCIPLES UNDERPINNING THE TREATMENT
The ABS developed six guiding principles that were used to develop a robust treatment for revising historical ERP.
1. The credibility of population estimates, both level and growth, should be maintained for all spatial levels (i.e. national, state, and sub-state).
2. The use of ADL in the PES has been a major improvement in how we measure Census coverage, and the 2011 net undercount should be used to inform our historical understanding of Census coverage.
3. Population growth for the 2006-2011 period should, as closely as possible, reflect the growth in the population components (i.e. births, deaths and migration) for all spatial levels (i.e. national, state, and sub-state).
4. Any assumptions should be based upon the best available data.
5. Any revision to the historical ERP series should maintain the demographically plausible relationships between the fundamental building blocks of population series (e.g. age-sex profiles).
6. Where revised data exist for population components data, they should be used regardless of whether they were available at the time of previous rebasing processes (e.g. recently identified late registration births for NSW for 2005-2010 should be included).
10 OR 20 YEAR REVISION PERIOD?
A revision period of 10 years, where the PES undercount adjustment for 2006 is the only data revised, cannot provide a coherent or demographically plausible series. This is because the resulting quarter on quarter population change would be well outside what are considered to be acceptable margins of error on the components of growth (births, deaths and migration).
While it might be considered appropriate to revise back to the start of the ERP series in 1971, at which point the undercount would also have been lower had the ADL methodology been available, this was not considered optimal. Revising back further than to 1991 would result in little gain for considerably greater effort.
THE REVISION METHOD
The ABS treatment will involve revising the PES undercount adjustment component of the ERP rebasing methodology for the 1996, 2001 and 2006 bases. The revised PES undercount adjustments will be published in the 20 June 2013 issue of Australian Demographic Statistics (cat. no. 3101.0) and will be determined through reference to:
1. Components of growth, which will be a strong indicator of change in historical ERP levels;
2. Historical intercensal error;
3. Studies into the impact of changes in PES undercount methodology, in particular the introduction of ADL in 2011; and
4. Other data, where available.
The revisions will have the largest impact on 2006 ERP estimates and the smallest impact on 1996 ERP estimates. This will ensure the series is plausible between 1991 and 1996, given the growth reflected in the series prior to 1991, which will remain unchanged.
THE IMPACT OF REVISIONS
The 20 June 2013 issue of Australian Demographic Statistics (cat. no. 3101.0) will include the final rebased ERP estimates, based on the 2011 Census, and will also include all revised series at the national and state and territory levels back to 1991. An additional feature article will be included in this issue comparing the revised series with the previously published series.
The ABS intends to release the revised sub-state series with the release of final rebased sub-state estimates in the August 2013 release of Regional Population Growth, Australia (cat. no. 3218.0).
The intended revision to the 2006 PES undercount adjustment for Australia will be similar to the level obtained through the taking of components of growth back in time from 2011. The intended revision to the 2001 PES undercount adjustment will be smaller in magnitude than the revision applied to the 2006 PES undercount adjustment, and less again for the 1996 PES undercount adjustment.
The graph below provides an illustrative representation of the impact of the 20 year revision at the national level on ERP, with the most noticeable revision being for 2006. The indicative revisions in the graph are 240,000 fewer people at 2006, 130,000 fewer at 2001 and 70,000 fewer at 1996.
Until a revised undercount adjustment is derived for each state and territory, the size and direction of their current 2006-11 intercensal error should be considered an illustrative guide. This information should be considered to be indicative since other data will be used to inform the size of revisions for each state and territory.
Revisions at the sub-state level will be guided by revised undercount adjustment estimates at the capital city and rest of state level, and revisions for a state or territory should not be assumed to apply to all levels of geography within.
IMPLICATIONS FOR OTHER STATISTICAL SERIES
The revisions to historical ERP will have implications for other ABS demographic data, including fertility and mortality rates. The ABS intends to gauge the need for revisions to be applied to other ABS demographic data on an individual collection basis. A summary of the series for which revisions will be required will be included in the 20 June 2013 issue of Australian Demographic Statistics (cat. no. 3101.0).
The revisions to historical ERP data will also need to be applied to other ABS data for which historical ERP is a key input. Two of the most prominent examples of these data include Labour Force and National Accounts. The table below provides an indication of the timing of when these data will be revised and the publication within which they will be released.
ERP is also used within a broad range of non-ABS data series. The ABS will work with a range of stakeholders to support them through the transition period and after revised data are released, and welcomes any requests for advice and support.
FOR MORE INFORMATION
For more information, please contact ABS Demography at email@example.com.
OTHER RELEVANT ABS REFERENCES
These documents will be presented in a new window.