|Page tools: Print Page Print All|
NEW METHODOLOGY USED TO PREPARE SUB-STATE POPULATION ESTIMATES
Pt = the estimated resident population at time t
Pt+1 = the estimated resident population at time t+1
B = births occurring between time t and t+1
D = deaths occurring between time t and t+1
M = net migration occurring between time t and t+1
The component method is used by the ABS to estimate national, state and territory population change. Traditionally, migration has been difficult to estimate below the state level and therefore a regression modelling method was used to estimate sub-state population change. However, recent developments in the estimation of regional migration have enabled the component method to be implemented to prepare sub-state estimates of the resident population (ERP) for 2017 and onward. This change brings consistency with the method used to prepare state and territory ERP.
For the first time, the data released in this product allows users to break down total population change into its core components of natural increase and net migration, enabling a richer understanding of population change for an area. This article describes the method used to prepare the component estimates in more detail. The Feature Article 'Understanding the 'Hows and Whys' of Regional Population Growth: What's driving change in our regions' demonstrates the power and depth of the additional component data, exploring the driving forces behind population change for sub-state areas. This Feature Article can be accessed via the left navigation panel of this publication.
DATA SOURCES AND METHODS
Each of the components of sub-state population change (births, deaths, internal migration and overseas migration) are prepared by breaking down state and territory component estimates, using the data sources outlined below. This ensures consistency between state and sub-state geographies. For more information on the preparation of components at the state/territory level, see the Explanatory Notes of Australian Demographic Statistics (cat. no. 3101.0).
NATURAL INCREASE: BIRTHS AND DEATHS
Natural increase is calculated based on births and deaths data provided to the ABS by the state and territory Registries of Births, Deaths and Marriages. This data is aggregated to the Statistical Area Level 2 (SA2) and Local Government Area (LGA) levels of the Australian Statistical Geography Standard (ASGS).
Sub-state births and deaths estimates are prepared for financial years to correspond with the 30 June reference date for ERP. To enable the production of timely estimates, births and deaths by year of registration (the year that the birth/death was registered) is used as a proxy for year of occurrence (the year that the birth/death actually occurred). This is consistent with published preliminary state and territory estimates of births and deaths.
Preliminary birth and death estimates are subject to fluctuations caused by lags or accumulations in the reporting of birth and death registrations. This can be caused by late notification to the registrars, delays arising from incomplete information being supplied, procedural changes affecting the registrars or issues within the ABS or registry processing systems.
Sub-state births and deaths data is also prepared and released for calendar years in Births, Australia (cat. no. 3301.0) and Deaths, Australia (cat. no. 3302.0). This data is released for calendar years, rather than financial years, to correspond with the calculation of fertility and mortality rates.
REGIONAL INTERNAL MIGRATION ESTIMATES (RIME)
RIME has been developed by the ABS since 2005 and is based on de-identified Medicare change of address data provided by the Department of Human Services. This data is coded to the Mesh Block level of the ASGS, enabling direct aggregation to SA2 and LGA regions.
The Medicare data is lagged by three months to account for the time between a person moving address and updating their details with Medicare. This lag is consistent with the three month lag assumption used to estimate interstate migration. Data from September of the previous year to September of the reference year is used to reflect movements that occurred in the financial year.
There is recognised undercoverage of certain subpopulations in Medicare, such as temporary migrants (most of whom are not eligible for Medicare) and young adults (who do not always update their details with Medicare in a timely manner). This is accounted for in RIME by applying expansion factors to increase the number of moves by age, in line with expansion factors applied at the state/territory level.
Medicare theoretically covers all Australian residents and non-Australian residents granted temporary Medicare registration. However, some Australian residents do not access the Medicare system, primarily due to access to other health services. One such group is defence force personnel. Therefore, Medicare data is supplemented with information on defence force movements provided by the Department of Defence.
Further detail on the method used to prepare RIME including the use of expansion factors and defence force movements, is available in Discussion Paper: Assessment of Methods for Developing Experimental Historical Estimates of Regional Internal Migration (cat. no. 3405.0.55.001).
REGIONAL OVERSEAS MIGRATION ESTIMATES (ROME)
This is the newest of the sub-state components and has traditionally been the most difficult to estimate for sub-state regions, however it is also generally the smallest component of population change at the sub-state level. The Census tells us that of all people who had a different address one year ago, 90% were elsewhere in Australia (internal migrants) and only 10% were overseas. With the longstanding series of sub-state births and deaths, and the production of the new RIME series, the ABS has now developed a model to prepare ROME. The ROME model breaks down state/territory overseas arrivals and departures into sub-state areas, based on information from the Census.
Regional Overseas Arrivals
The Census question which asks where a person was living one year ago gives us numbers of overseas arrivals by SA2 and LGA in a Census year. These numbers are then constrained to state/territory overseas arrivals for each year up until the next Census. This can be done with reasonable confidence given that the distribution of overseas arrivals within a state/territory has not changed substantially between previous Censuses. Exceptions, such as high growth and inner city areas, are accounted for by incorporating more up-to-date indicator data sources such as counts of 457 visa holders and overseas students.
Regional Overseas Departures
A model to estimate sub-state overseas departures has been developed based on population change and experimental components data (including overseas arrivals) for each region over the past two intercensal periods. The model is based on an adaption of the component method equation:
For example, for the 2011 to 2016 intercensal period, the 2011 and 2016 populations are known from Census, and births (B), deaths (D), internal migration (IM) and overseas arrivals (OA) are estimated using the methods described above. Overseas departures are then modelled based on the difference between the sum of these remaining components (B – D + IM + OA) and the known population change between previous Censuses (Pt – Pt+1). Overseas departures is not deemed as simply the balancing item between the two series, but rather is modelled based on these differences.
A number of models to distribute overseas departures within each state/territory were tested. The models used different assumptions based on other information for each area. For example, total population, overseas-born population and the number of people who arrived from overseas in the last 10 years. The effectiveness of these assumptions was assessed by comparing the component-based estimates with the actual population change for the intercensal period. The best model apportioned overseas departures within a state/territory based on recent overseas arrivals - those that arrived within the last year. This makes sense, as areas with high numbers of recent overseas arrivals are more likely to have strong connections with overseas communities and a greater propensity to move back.
The model was further refined to boost departures in areas if they met 2 criteria:
Overseas departures are boosted for areas which had more than 20% of their population born overseas and were in the top 2 SEIFA IEO deciles at the previous Census, on a sliding scale. That is, SA2s with more than 60% of their population born overseas, and with the highest SEIFA scores in decile 10 are given the biggest boost, and those with between 20% and 60% of their population born overseas and in SEIFA decile 9 are given a lower boost. These thresholds provided the most accurate modelled estimates for the 2011 to 2016 intercensal period of all the models evaluated. This model is then used to estimate overseas departures each year up until the next Census, with the modelled estimates constrained to corresponding state/territory overseas departures.
CALCULATION OF ERP
Each of the components, prepared as described above, are fed into the component method equation to estimate sub-state populations in non-Census years. As each of the components are constrained to state/territory level estimates, the resulting ERP sums to state/territory ERP.
After the estimates are prepared they are each scrutinised and validated by ABS analysts. Additional population indicator data sources such as dwelling approvals, Medicare stock and counts of people on the Australian electoral, as well as local knowledge advised by state governments is considered and used to adjust figures where necessary.
While the component method is a new way of estimating population change out from the 2016 Census base, this is not considered a break in time series. Even though the previous, regression-based estimates shaped the rebased 2011–2016 intercensal population change, the 2011 and 2016 end points reflect what the respective Censuses have shown for each area. Under the old regression method, the ABS prepared new models every five years based on what we learnt from the last two Censuses. Therefore, under a regression- or component-based approach, the process is reset every five years. The validation process outlined above also ensures that traditional modelling indicators (dwellings, Medicare stock and AEC counts) are still taken into account where necessary.
Note that for just over 100 SA2s, component data is considered unreliable and so population has been held constant rather than estimated using the component method. These SA2s generally have no/small population and cover areas such as national parks, reserves, industrial and commercial complexes and airports. This is consistent with the approach used under the regression method.
RELEASE SCHEDULE AND REVISION CYCLE
Preliminary component-based estimates for the reference year will generally be available by April of the following year, and will be based on preliminary state/territory components and ERP (as released in Australian Demographic Statistics). Over the following year, some of the components will be revised at the state/territory level. Any such revisions will be incorporated into sub-state ERP as part of the next scheduled release of Regional Population Growth. For example, the 2017-18 issue of this product will include preliminary 2018 estimates for the first time, as well as revised 2017 estimates that sum to revised 2017 state/territory ERP and components.
Sub-state ERP by age and sex in non-Census years will also be prepared using the component method from 2017 onwards. That is, estimates by age and sex will be updated from the previous year's estimates by adding natural increase and net internal and overseas migration by age and sex. Age and sex variables are available on the births, deaths and internal migration data sources discussed above. Modelled overseas migration estimates can also be prepared by age and sex, utilising information from the Census and other data sources at the age and sex level. Each of the components by age and sex will be constrained to the corresponding state and territory estimates by age and sex released in Australian Demographic Statistics, and the sub-state totals released in this product.
Component-based estimates of sub-state population by age and sex will generally be available five months after the release of totals. ERP by age and sex for 30 June 2017 will be released on 31 August, 2018 in Regional Population by Age and Sex, Australia, 2017 (cat. no. 3235.0).
These documents will be presented in a new window.