Tracking future population changes requires a range of estimation methods

Understanding where populations will change in the future through using population projection models is key for governments and providers of housing, infrastructure and services to be able to plan where they will prioritise resources. However, ensuring such models are accurate means recognising their underlying assumptions, data and potential for error.

New AHURI research critically assesses the population projection methods available to Australian decision-makers and planners. It found that because different users need different levels of understanding about future population changes, it is unlikely the existing methods could be unified into one nationally consistent approach that would be accurate enough for everyone’s needs. The research, ‘Improving small area population projects’, was undertaken for AHURI by researchers from the University of Adelaide, the University of South Australia and the University of Sydney.

Population can be projected using top-down or bottom-up methods

Population projections can range from covering the national scale down to quite small areas; for Australia, the geographical hierarchy narrows down from the national to the state/territory level, and then to a range of smaller ‘administration’ areas such as Greater Capital City Statistical Areas; Local Government Areas (LGAs); postcodes and suburbs; and ABS Statistical Area Levels 4 (SA4, or around 100,00 people) to 1 (SA1, or around 400 people).

Population projections are modelled either as top-down models, where larger spatial units are used to control the smaller area projections, or as bottom-up models, where the smaller area projections are added together to form the population projection of the larger spatial units.

For example, with a top-down projection Australia’s total population can be used as a control total for the state and territory projections and, in turn, these are used to control the smaller SA4 projections. In essence, the sum of the smaller parts—in this case the projected populations of the states and territories—can never exceed the total population of Australia. In the bottom-up model, the projected populations of the states and territories would be added together to form the total population projection of Australia. 

‘There are arguments to support both types of models,’ says lead researcher, Professor Emma Baker of the University of Adelaide. ‘The general population data for larger spatial units, such as Australia and the states and territories for example, are more reliable. As geographic areas get smaller, the population data may be less reliable due to small numbers or missing data. Small area projections, however, will have better input of local policy, land availability and development data, which in turn provides a realistic base for population growth. Conversely, it may overestimate growth trends when any land development that drove historic growth is no longer available.’

Different users have different needs for population projections

The research finds that populations projections are used in two main ways. 

The first sees national and state and territory governments use population projections as an official benchmarking tool for the allocation of funding and resources. In general, this means projections are at the larger spatial level, have longer timeframes and are predominantly based on demographic change. 

The second way, as used by private enterprise, research and consultancy, sees population projections as important reflections of what is likely to happen on-the-ground, and are therefore characterised by shorter time frames; larger variety and quantity of input data; smaller spatial areas; and frequent updating.

‘Ultimately, our research shows that it is unlikely that the top-down and bottom-up methods could be unified to one nationally consistent approach that would fulfil different users’ needs with sufficient accuracy,’ says Professor Baker.  ‘Instead, there is a place for both strong top-down projections and more flexible targeted bottom-up ones.’

The research also found that population estimate errors are larger in populations that are more mobile (e.g. late teenagers and 20s) and in the 0—4 cohort. The latter may indicate an issue with the fertility assumptions, while the former is more likely an issue of the way mobility is measured in Australia. This can lead to problems when developing assumptions for intrastate and interstate migration for smaller population spatial units.

There are ways population projection methods can be improved

In order for future policy to be based on solid and reliable population estimates, the research identifies prioritising:

• consistent approaches and shared information sources 
• good quality, reliable and timely data
• a thorough understanding of land and dwelling supply
• better methods (especially for estimating small area populations)    
• a more widespread understanding of error and accuracy
• a solid pipeline of training in demographic skills to ensure there is a qualified and skilled workforce into the future.