FEATURE ARTICLE 4: ADVICE ON THE TRANSITION TO REBASED POPULATION ESTIMATES FOR NATIONAL AGREEMENT REPORTING
UNDERSTANDING THE FIVE-YEARLY ERP REBASING
The ABS produces official population estimates for Australia and each of the states and territories on a quarterly basis. These estimates are known as estimated resident population, or ERP. They are published in Australian Demographic Statistics (cat. no. 3101.0), together with a range of other important population information - most notably the population change components of births, deaths and migration.
ERP is calculated using the 'demographic balancing equation', which derives a population at the end of a period by starting with the population at the start of the period, adding births that occurred, subtracting deaths that occurred and factoring in the net migration outcome for the period.
Every five years the ABS uses information from the most recent Census, which is the best source of information about Australia's population, to 'rebase' (or 'recalibrate') ERP. To do this, the ABS uses data from the Census and makes an adjustment for net undercount or overcount (i.e. people missing from Census counts or counted multiple times) as determined by the Census Post Enumeration Survey - a short survey run in the period immediately following the Census. An adjustment is also made for Australians who were temporarily overseas on Census night. From this, the ABS produces the official estimate of the number of people who usually reside in Australia - in respect of 30 June of the Census year.
The difference between the estimates published before rebasing and the rebased estimates at 30 June of the Census year is referred to as the 'intercensal error'- that is, the accumulated difference between the change reflected in the population components over the intercensal period, and that reflected in the previous and new bases.
The intercensal error for the preliminary rebased population estimate for the 2006-2011 period was 294,000, around 247,000 (or 84%) of which can be directly attributed to a change in the ABS methodology used to calculate the undercount adjustment.
In determining how to calculate the final 2011 Census rebased ERP estimates, the ABS consulted key users of population estimates on whether the larger than normal intercensal error caused by the methodological change warranted a different treatment to the conventional one used in producing the preliminary estimates. As a result, on 18 December 2012 the ABS announced in Australian Demographic Statistics (cat. no. 3101.0) that it intends to adjust historical ERP series to reflect the different undercount adjustments that would have been used had the 2011 undercount methodology been available in earlier years. For further information, please refer to The 2006-2011 Intercensal Period and Revisions to Historical ERP in Australian Demographic Statistics (cat. no. 3101.0).
Historical population data from 1991 to 2011 has now been adjusted, to ensure population growth in the rebased series in recent years reflects that observed in the population components of births, deaths and migration. The revised data appear in this release of Australian Demographic Statistics (cat. no. 3101.0).
REVISIONS TO OTHER DEMOGRAPHY DATA AFTER EVERY REBASING
Just as the rebasing process occurs on a five-yearly cycle, so does the process of revising other demography data which use ERP as a denominator. Using the final rebased estimates, all of these estimates will be re-derived, using the new denominator ERP values. For example, all fertility measures and mortality measures (with the exception of life expectancy, which is never re-estimated) will be re-derived, onto the new ERP base.
While revisions to demographic data would normally only involve revisions over a period back to the previous Census base (in this case 30 June 2006), the current rebasing process has delivered recast ERP back to 30 September 1991. Therefore fertility and mortality measures will also be revised back to 1991 and released per the schedule outlined in Table 1.
DEMAND FOR ABS ADVICE
While ERP and other demography data have been extensively used in long-standing reporting arrangements, such as the Report on Government Services (ROGS), this is the first rebasing cycle since the national agreement reporting arrangements began. This advice therefore provides a broad outline of what parties involved in reporting should consider, with the release of rebased population data in 2013 and 2014.
As with all advice the ABS provides, the advice given in this article should be considered in relation to the specific requirements of data users and the appropriateness of the advice for their needs.
ADVICE #1 - THE USE OF ERP AS A DENOMINATOR
Where ERP has been used as a denominator, from 20 June 2013 it will be possible to use 2011-based estimates (from 30 August 2013 for sub-state ERP). The only exception to this is Aboriginal and Torres Strait Islander and Non-indigenous population estimates, for which 2011-based time series estimates will not be available until April 2014. 2006-based estimates and/or projections for these populations should continue to be used as denominators during the interim period.
While the ABS recommends revising historical series after each rebasing cycle is complete, it acknowledges that this may not be possible in all instances. Where revising indicators every five years presents a major challenge to stakeholders (whether due to resourcing or other constraints), and is not considered to be feasible, the ABS advises that users should be aware that this will result in a break in series at five-yearly intervals. This will directly affect the comparability of the data over time. A 'smoothing' option may be a feasible alternative for some indicators, but it should be noted that this is generally considered to be a sub-optimal solution with its own associated challenges.
ADVICE #2 - THE USE OF OTHER ABS DEMOGRAPHY DATA
From October 2013 and November 2013 (for fertility and mortality measures respectively) it will be possible to use the 2011-based measures. The timing of the release of these measures is outlined in Table 1.
Aboriginal and Torres Strait Islander and Non-indigenous measures will continue to be released using 2006-based denominators until April 2014. These measures will be revised in 2014.
ADVICE #3 - STANDARD POPULATION
Prior to the 2011 ERP rebasing cycle, it was generally accepted that the ABS produces a new 'standard population' every ten years, with the last standard population being 30 June 2001 and the next population expected to be 30 June 2011. However, the ABS and AIHW have recently reviewed whether changing the standard population every 10 years is necessary, given the resourcing required. After concluding that there is very little difference in age-standardised comparisons over longer periods of time, both organisations are proposing that the frequency be extended to every 25 years (i.e. 2001, 2026, 2051 etc.). The ABS and AIHW will undertake consultation on this change ahead of publishing an information paper by the end of September 2013.
ADVICE #4 - THE USE OF INDICATORS FROM ABS POPULATION AND SOCIAL SURVEYS
ABS population and social surveys are not revised as a result of rebasing, with the exception of the Labour Force Survey (LFS). Whereas the LFS is rebenchmarked to be consistent with revised ERP, the supplementary surveys such as the Survey of Education and Work are not revised. While previous estimates of the population are used in the benchmarks and weighting of ABS population and social surveys, these estimates are used in both the numerator and denominator. The compositional information available from these collections should therefore continue to be used.
The ABS recommends that no revisions to any indicators derived from ABS surveys be undertaken, except in instances where LFS data are specifically referenced in a time series capacity. LFS estimates, revised from 1991 onwards to be consistent with recast ERP, are expected to be released in February 2014 (see Labour Force, Australia, December 2012 (cat. no. 6202.0)).
Table 1 - Timetable of release of rebased demography data (and key related releases)
|20 June 2013
|Australian Demographic Statistics (cat. no. 3101.0)
|National and state/territory ERP (September 1991 and onwards revised)
|30 August 2013
|Regional Population Growth (cat. no. 3218.0) and Population by Age and Sex, Regions of Australia (cat. no. 3235.0)
|Annual sub-state ERP (1992 and onwards revised)
|30 August 2013
|Population Estimates, Aboriginal and Torres Strait Islander Australians, 2011 (cat. no. 3238.0.55.001)
|30 June 2011 estimates of Aboriginal and Torres Strait Islander and Non-indigenous populations
|24 October 2013
|Births, Australia (cat. no. 3301.0)
|Revised 2002-2012 fertility measures (e.g. TFR)
|7 November 2013
|Deaths, Australia (cat. no. 3302.0)
|Revised 2002-2012 mortality measures (except for Indigenous and Non-indigenous, which will occur in late 2014)
|7 November 2013
|Life Tables, States, Territories and Australia, 2010-2012 (cat. no. 3302.0.55.001)
|Life tables and life expectancy measures for 2010-2012. Historical life expectancy measures are not revised
|15 November 2013
|Life Tables for Aboriginal and Torres Strait Islander Australians, 2010-2012 (cat. no. 3302.0.55.003)
|Life tables and life expectancy measures for Aboriginal and Torres Strait Islander and Non-indigenous populations, 2010-2012. Historical life expectancy measures are not revised
|18 December 2013
|Migration, Australia (cat.no. 3412.0)
|Revised ERP by Country of birth
|30 April 2014
|Population Estimates and Projections, Aboriginal and Torres Strait Islander Australians, 2001-2026 (cat. no. 3238.0)
|Backcast and projected estimates for the Aboriginal and Torres Strait Islander population - 2001-2026
|By June 2014
|Australian Historical Population Statistics (cat. no. 3105.0.65.001)
|Revised 1992-2001 fertility measures. Revised 1992-2001 mortality measures
OTHER RELATED ADVICE - REMOTENESS CLASSIFICATIONS AND SEIFA INDEXES
While the Remoteness classifications and SEIFA indexes are primarily designed to measure the characteristics of the population at the time of the Census, they may also be used to compare indicators over time. A remoteness category used in one Census (e.g. Inner Regional Australia) would not refer to exactly the same areas as the corresponding category from the previous Census (although there is a high degree of overlap). At each Census, however, Inner Regional Australia would include areas categorised according to the same conceptual basis (i.e. accessibility to services).
Likewise for SEIFA, a quantile from one Census (e.g. lowest quintile in the Index of Relative Social Disadvantage) would not necessarily refer to the same areas as the corresponding quantile from a previous Census, but both would have been calculated by applying a common conceptual framework of relative socio-economic disadvantage.
In most cases it is possible to make a valid comparison of indicators using the Remoteness Structure derived from different Censuses. When doing this users should keep the following in mind:
- the concept of accessibility/remoteness has remained constant between 2001 and 2011;
- the geographic unit on which remoteness areas are calculated changed from Collection District (CD) according to the Australian Standard Geographical Classification (ASGC) in the 2006 Census to Statistical Area 1 (SA1) according to the Australian Statistical Geography Standard (ASGS) in the 2011 Census.
- between 2006 and 2011 some localities moved from one remoteness category to an adjacent one - some to a more remote category, others to a less remote one; and
- the overall effect of the change from 2006 to 2011 was minimal, both in terms of the relative land area and population affected.
For further information see ASGS Remoteness Structure - Facts Sheet at https://www.abs.gov.au/geography
or at https://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/1270.0.55.005July%202011?OpenDocument
Information can be obtained by comparing indicators for individuals by SEIFA decile (and quintile) over time where the SEIFA measures are derived from different Censuses. When doing this users must keep the following in mind:
COMPARISONS USING SEIFA AND INDIGENOUS STATUS
- the conceptual basis of the four SEIFA indexes has remained constant between 2006 and 2011 although the variables used in each index and the weights given to variables have changed;
- the relative socioeconomic advantage/disadvantage of an individual may be different from that of the area in which he or she lives;
- the relative socioeconomic advantage/disadvantage of some areas will have changed between Censuses;
- the SEIFA indexes may not be as relevant a measure of the relative socioeconomic advantage/disadvantage of the area at the end of the intercensal period as they were at the start;
- the Census years provide the best time points for comparison; and
- when making comparisons over time, quantiles (e.g. deciles, quintiles) should be used rather than ranks or index scores.
Indicators of the Aboriginal and Torres Strait Islander and non-Indigenous population can be meaningfully compared by SEIFA category. In addition to the points noted in the previous section, users should keep the following in mind when undertaking such analysis:
- the SEIFA indexes assess the relative socioeconomic advantage/disadvantage of areas based on the characteristics of the whole population, they do not assess the relative socioeconomic advantage/disadvantage of individuals within subpopulations; and
- the 2011 SEIFA Index of Relative Social Disadvantage (IRSD) (unlike the corresponding 2006 index) does not include the proportion of Indigenous people living in an area as an input to the measure of relative socioeconomic disadvantage, therefore it is a more useful tool for comparing indicators by Indigenous status than it was in 2006.
For further information see the SEIFA entry under Data & analysis at www.abs.gov.au/census
or at https://www.abs.gov.au/ausstats/abs@.nsf/mf/2033.0.55.001