Help and Information
The Department of Health (through the Indigenous Health Division (IHD)) contracted Prometheus Information to develop a set of population estimates by Indigenous status for 2011, and projections to 2016 at the Statistical Areas Level 2 (SA2) across Australia, adapting the model and approach they had previously developed for the former Office for Aboriginal and Torres Strait Islander Health (OATSIH).
The Indigenous estimates for 2011 were to sum to ABS Estimated Resident Populations (ERPs) for the Indigenous population at the SA2, Indigenous Region and State/Territory level whenever such estimates were available, and to extent that the ABS figures were consistent across these different geographic levels.The estimates for 2012, 2013, 2014, 2015 and 2016 were to sum to:
- the published ABS Estimated Resident Population (2012 and 2013) for the total population at the SA2 level; and
- the published ABS projections (2014, 2015, 2016) for the Indigenous population at the Indigenous Region level.
Note: These estimates have not been adjusted to reflect the 2016 ERP (available from ABSat SA2 level, but not by age). Nor have they been updated beyond 2016.
Overview of project scope and results
Data acquisition and preliminary estimation
Many data files were obtained or created with the main source being the ABS. In some cases ABS files did not cover the whole of Australia and needed to be extended so they had data for every SA2 in Australia. Moreover, it was not always the case that files were consistent (e.g., estimates from different series may not agree in the aggregate and may be irreconcilable at a small area level). In such cases the data were adjusted to provide consistency - the main impact was on overseas territories.
Running the estimation process
The overall strategy for estimation involved a combination of Bayesian estimates and iterative proportional fitting, as used in the previous work for OATSIH. However, because of the new geographical structure (requiring concordances from 2006 Collection Districts (CDs), 2006 Indigenous Areas and 2006 Statistical Local Areas to 2011 geographical areas) and data elements, a reasonably extensive modification to the previously developed estimation software was required.
The estimation process was repeated numerous times as data problems were identified and fixed. The software detected inconsistencies and mismatches and these were dealt with manually - the process involved correcting data or seeking out missing data. Once all data issues were identified and addressed the iterative processes converged in all regions, SA2s and States/Territories.
The matching against the ABS reference datasets is summarised in the following Table.
Summary of matching of estimates to ABS reference dataset
|Matching by categories||2011||2012||2013||2014||2015|
|State matching - total population (maximum* of estimates minus ABS State population ERP or projections by age/sex)||0.00||0.00||0.00||0.00||0.00|
|Indigenous Region matching - Indigenous population (maximum of estimates minus ABS Indigenous region population projection by age/sex)||0.00||0.00||0.00||0.00||0.00|
|SA2 matching - total population (maximum of estimates minus SA2 ERP or projected population by age/sex)||0.00||0.03||0.23||No ABS SA2 estimates available||No ABS SA2 estimates available|
*In absolute value terms
Review and revision
Once estimates and projections had been created, they were reviewed and examined for anomalies that might be caused by a combination of the nature of the starting data and the nature of the projection process. The results for 2014 and 2015 had to be made without having ABS estimates of population by age and sex at the SA2 level. We created short term estimates by projecting the age groups forward using the trend from 2012 to 2013 and scaling them to match State and Territory totals. In 2015 there were 2 Indigenous Regions (702 and 707) in which ABS projected Indigenous population exceeded our total population for some age-sex groups. We allowed the fitting to adjust the SA2 populations to account for this phenomenon. The final estimates were rounded to 4 decimal places.
These estimates were made to assist the IHD in the calculation of rates and in understanding likely changes in populations at the sub-state level. Although the estimates have been produced at the SA2 level, they are not intended to be used at such a fine geographic level, but rather to provide a base from which they can be aggregated to larger geographic areas.
The IHD agreed to make these estimates available for general use on the condition that Prometheus Information Pty Ltd is acknowledged as the developer of the data, under a contract with the Department of Health. PHIDU, a unit funded by the Department of Health, is distributing the estimates on their behalf.