3219.0.55.001 - Information Paper: Population Estimates under Australia's New Statistical Geography, August 2011  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 26/08/2011  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All



Total Population Estimates

Age/sex Population Estimates


SA1-Based Population Estimates

LGA Population Estimates


Previously, the SLA was the base spatial unit used to prepare and disseminate sub-state population estimates as at 30 June each year. However, from 2011 onwards, the base spatial unit will be the SA2, as defined in the ASGS.

For the 2011 estimates, the ERP as at census date for each SA2 will be calculated based on usual residence census counts, excluding overseas visitors in Australia, with an allowance for net census undercount and the number of residents temporarily overseas (RTOs) at the census date. The estimates of net undercount will be apportioned to SA2s based on age, sex, Indigenous status, state/territory and GCCSA. The number of RTOs on census night will be estimated based on coding addresses of residence to SA2 from a sample of incoming passenger cards. For some areas, demographic adjustments will be made. As the census is not held on 30 June (the 2011 Census was held on 9 August), further adjustments taking into account births, deaths and migration for the intervening period will be made to obtain the ERP at 30 June.


For estimates as at 30 June 2012 and beyond, ERP for most SA2s will be calculated using a mathematical regression model, where relationships will be established between changes in population and changes in indicators between the 2006 and 2011 Censuses for groups of SA2s. More up-to-date indicator data will then be applied to the regression equation to estimate changes in the population of each area from the 2011 Census.

The regression models use an area's share of indicator data to estimate the share of the state or territory's population for that area. More specifically, the change in share of the state/territory of the indicator data is used to estimate the change in share of state/territory population since the base year.

The choice of indicators varies across the states and territories, depending on availability and indicative ability, and may include dwelling approvals, Medicare enrolments and counts of people on the Australian Electoral Roll.
  • Dwelling approvals. Dwelling approvals data are collected on an ongoing basis by the ABS, with summaries released frequently on the ABS website in Building Approvals, Australia (cat. no. 8731.0). Dwelling counts from the latest census are used as the base number of dwellings by SA2. Updated estimates of dwellings are prepared by adding approvals to the census base. Under the assumption that it takes several months for the dwelling to be constructed and people to move in after it has been approved, lags are incorporated into the approvals data. Dwellings approved six to twelve months before the estimation reference period are incorporated in the regression models, with provision made for some longer lags, in particular for large residential buildings (e.g. large apartment blocks).
  • Medicare enrolments. Changes to the number of Medicare enrolments provide an indication of total population change, which are incorporated into the regression models used to estimate population change. Under the assumption that it takes a few months for a person to change their address on the Medicare system, a lag is incorporated into the Medicare data used in the regression model. Previously Medicare enrolments have been provided to the ABS by Medicare Australia by postcode. The postcode level data were converted to SLA using a postcode to SLA correspondence, which meant the quality of the Medicare data was dependent on the quality of this correspondence. However, from 2011, Medicare Australia will supply enrolments to the ABS at the SA2 level, which is the level required for modelling purposes. As a result, the Medicare data used as an input to the regression model will align directly to the output regions required and should result in better quality ERPs.
  • Electoral enrolments. In the past, counts of people on the Commonwealth electoral roll were provided to the ABS by the Australian Electoral Commission (AEC). Under the assumption that it could take a few months for a person to change their address on the electoral roll, a lag was incorporated into the AEC data used in the regression model. These enrolments were provided by CD, which were then aggregated to SLA.

In areas where indicator data are unreliable and migration can be assumed to be insignificant, population change since the previous census will be estimated by the natural increase (births minus deaths) since the previous census. In some very small areas, population change since the previous census may be assumed to be zero in the absence of reliable indicator data for these areas. All estimates will be scrutinised and validated by ABS population analysts. Local knowledge, including that advised by local governments and state/territory agencies, may be used to adjust the regression model-based estimates for particular SA2s. Estimates at the SA2 level will be constrained to state/territory population estimates.


Once the total population estimates for each SA2 for post censal years are prepared, they are broken down into age and sex components. The estimates by age and sex will be updated from the previous year's estimates using registered births and deaths data, and synthetic estimates of migration based on the previous census. For areas where these data are deemed to be of insufficient quality, adjustments will be made.

Estimates at the SA2 level by single year of age and sex are confidentialised and finally constrained to state/territory population estimates.

SA2 estimates will continue to be released on the ABS website by five year age groups (0-4, 5-9, ..... 80-84, 85 years and over), however calculations will be made at the single year of age level (up to 99 years, then 100 and over). The geographic level at which these estimates will be released, or made available, will be determined once the quality of estimates has been assessed.


In census years, both preliminary estimates (derived from updating the ERP from the previous census) and rebased estimates (based on the current census) are prepared. Differences between these two sets of estimates are known as intercensal errors. In the past this was done at the SLA level, where the rebased, or final, estimates of SLA populations for the previous intercensal years were based on estimates derived by apportioning the intercensal error evenly across the five years. SLA level estimates were constrained to state/territory estimates. For example, rebased 2002 to 2005 estimates were derived by adding one-fifth of the 2006 intercensal error to the previous estimates of the 2002 population, two-fifths to the previous estimate of the 2003 population, and so on.

As the 2011 Census will be based on ASGS, this process of rebasing will be done at the SA2 level.

In order to rebase 2007 to 2010 estimates, preliminary estimates (derived from updating the ERP from the previous census) based on SLAs will be converted to SA2s. The difference between the 2011 preliminary estimate, and the 2011 Census rebased estimate will be applied to the previous intercensal years in a similar method to that previously described. That is, rebased 2007 to 2010 estimates will be derived by adding one-fifth of the 2011 SA2 intercensal error to the previous estimates of the 2007 population, two-fifths to the previous estimate of the 2008 population, and so on.


Census year estimates for SA1s will be prepared by apportioning the population estimate for each SA2 across the SA1s contained within that SA2, using census usual resident counts. In subsequent years, the 30 June population estimates for SA2s will be apportioned across SA1s by taking into account population change implied by indicator data for each SA1 since the census year. The SA1 populations within each SA2 will then be adjusted (on a pro-rata basis) to add to the SA2 population. A confidentialisation method will also be applied to SA1 population estimates.

These estimates will be aggregated to form population estimates for Non-ABS Structures such as Remoteness Areas, Postal Areas, State Suburbs and electoral divisions.

The quality of SA1 estimates will be better than the corresponding CDs of the old geography. CDs were designed primarily for collecting information and not for the dissemination of aggregated statistics. However, SA1s have been designed specifically for output and should more accurately reflect the regions they are designed to represent. Also, while the CD-level population estimates were apportioned within SLAs using just one indicator data series, there will be several data sources available to apportion the SA2-level estimates down to SA1 level.


LGA ERPs will be approximated by SA1 populations, including split SA1s. For LGAs which cut across SA1 boundaries, estimates will be prepared using an annually updated and accurate SA1 to LGA correspondence. Mesh Block-based person data will be used to calculate these splits. This method will potentially be more accurate than the current method of preparing LGA estimates as it will not involve converting population indicator data from postcode to LGA as in the past.