|Page tools: Print Page Print All|
FEATURE ARTICLE: FINAL REBASING OF AUSTRALIA'S SUB-STATE POPULATION ESTIMATES, JUNE 2007 TO 2011
Following the process illustrated in the diagram above, the ABS used the 2011 Census results to produce 'preliminary rebased' population estimates for 30 June 2011, and for all previous intercensal years (June 2007 to June 2010). Total estimates were released in Regional Population Growth, Australia (cat. no. 3218.0) and age/sex estimates were released in Population by Age and Sex, Regions of Australia (cat. no. 3235.0) in July - August 2012.
In this release, the preliminary rebased population estimates have been updated to final rebased estimates for 30 June 2007 to 30 June 2011.
To construct the final rebased ERP from the Census count, four adjustments have been made since preliminary rebasing.
1. A refined methodology of distributing net undercount to sub-state regions was incorporated. In preliminary rebasing, there were three equal factors used to distribute undercount to SA2s within each part of state (Greater Capital City Statistical Area and Rest of State/Territory): Age; Sex; and Indigenous status. In final rebasing, the undercount was allocated to SA2s within each part of state in a two-tier process. Age and sex were the primary factors used to distribute undercount, with Indigenous status then used as a secondary factor because of the relatively small proportion of the Australian population who are Aboriginal and Torres Strait Islander. The same total Aboriginal and Torres Strait Islander and Non-indigenous undercount for each part of state was still assumed, but the method used to distribute across SA2s was refined. This guaranteed the most demographically plausible sub-state estimates. It is important to note that this methodological refinement had no impact on distribution of ERP at the part of state level.
2. There was the standard revision to the number of Australian residents who were temporarily overseas (RTOs) on Census night for each SA2 using data on international traveller movements obtained from the Department of Immigration and Citizenship. At the national level the final estimate was 484,700, compared with the preliminary estimate of 475,300 (an additional 9,400 people).
3. Final demographic adjustments have also been applied in constructing a final rebased ERP for Census night. Demographic adjustments were applied to resolve any statistical anomalies in the age sex composition of the derived population estimates. To achieve plausible age-sex distributions for final rebasing, 11,700 people were added to the national population. The application of this adjustment varied across SA2s.
4. A further step in arriving at the final base ERP figure for 30 June 2011 was to revise the components of population change (births, deaths and migration) used to backdate the resulting Census night estimate (9 August 2011) to 30 June 2011. The national backdating adjustment in the final rebasing process (38,800 people) was similar to that of preliminary rebasing (35,500 people). Again, the application of the backdating adjustment varied across SA2s, according to fertility, mortality and migration dynamics of the different regions.
Table 1 shows the final components used to arrive at ERP by Greater Capital City Statistical Areas (GCCSAs), including adjustments due to net undercount, RTOs and backdating.
The conclusion of the final rebasing process involved constructing an ERP series for the intercensal period which took account of the difference between the newly established 30 June 2011 ERP based on the 2011 Census, and the original 30 June 2011 estimates modelled from the 30 June 2006 Census base. This difference was distributed evenly over the four intervening years of the intercensal period as 'intercensal error' (discussed further below).
On this occasion the final rebasing process also included an additional one-off 'recasting' of ERP back to June 1991 to accommodate the impact of a methodological improvement in the estimation of Census undercount. For more information on the recasting process, refer to Feature Article: Recasting 20 years of Australia's sub-state population estimates in Population by Age and Sex, Regions of Australia, 2012 (cat. no. 3235.0).
UPDATING THE INTERCENSAL ERP FIGURES DURING THE FINAL REBASING CYCLE
As well as constructing a final ERP base figure for each region at 30 June 2011, estimates from June 2007 to June 2010 were also recalculated to produce final rebased population estimates.
Inevitably there is a difference between the newly established 30 June 2011 ERP based on the 2011 Census, and the original 30 June 2011 series estimated from the 2006 Census using a modelling methodology. This difference is referred to by demographers as the 'intercensal error' and is generally distributed evenly over the intercensal period. Rebased 2007 to 2010 estimates were derived by adding one-fifth of the 2011 intercensal error to the previous estimates of the 2007 population, two-fifths to the previous estimate of the 2008 population, and so on. A summary of the intercensal errors for the 2006-2011 period is provided in table 2 below.
The main reasons for updating these intercensal ERP figures are to provide a more accurate population estimate for each year, and also to ensure that the estimates from the 2006-2011 intercensal period will be comparable with all future estimates, thus creating a consistent time series of ERP.
These adjustments were the concluding stage of the final rebasing process and, following this, it is not expected that any subsequent revisions will be made to these figures.
The intercensal error refers to the difference between the final rebased (2011 Census-based) ERP figures for 30 June 2011 and the unrebased figures updated from the 2006 Census.
For Australia, the preliminary (unrebased) June 2011 ERP over-estimated the final rebased June 2011 ERP by 1.2% (278,300 people). For the states and territories, the 2011 intercensal errors ranged from -0.6% (Australian Capital Territory) to +2.3% (Queensland).
Summary statistics of the absolute values of these errors can be used to assess the accuracy of a number of population estimates. To give an indication of the quality of SA2-based estimates, a set of unrebased 2011 estimates was prepared, updated from 2006 Census-based estimates and adjusted to aggregate to final rebased state/territory ERP. These were compared with the final rebased 2011 SA2 estimates. The average absolute value of the intercensal errors for this series of 2011 SA2 estimates (excluding areas with less than 1000 people) was 3.5%.
Average absolute intercensal errors for these 2011 SA2 estimates decreased with increasing population size; that is, SA2s with large populations recorded the smallest percentage errors while small SA2s had the largest percentage errors.
NEW TERMINOLOGY FOR 2016 REBASING
The ABS is aware that the term 'intercensal error' is often misinterpreted. While there is considerable information available to explain what this demographic term refers to, the word 'error' is too commonly considered to be a synonym for 'mistake'. As a result, the ABS will use the terms 'preliminary intercensal difference' and 'final intercensal difference' in the 2016 rebasing cycle.
FUTURE DIRECTIONS AND FURTHER INFORMATION
Following this issue which contains final population estimates for sub-state regions based on the 2011 Census, it is expected that no subsequent revisions to the 2006-2011 intercensal period will be made.
For more information, contact Tricia Dyson on (08) 8237 7662, or email email@example.com.
These documents will be presented in a new window.