Measuring Net Overseas Migration in Australia

This article provides an overview of how Net Overseas Migration (NOM) is measured including, the key data sources, concepts and revision process. In particular, it describes the different ways final and preliminary estimates are produced, their relationship to one another and what this means for understanding changes in NOM. The article also explains why NOM may take some time to fully reflect changes in migration patterns, using examples from the recent COVID-19 pandemic which severely disrupted international travel and migration.

Australia's population growth is composed of two main components, these being natural increase (births minus deaths), and Net Overseas Migration. NOM is the measure of change to Australia’s population resulting from migration into and out of Australia.

While natural increase is comparatively stable over time, the level of NOM is more subject to change. As a result, changes in Australia's population growth rate are largely driven by changes in NOM.

NOM was continuously the largest component of annual population growth from late 2005 until 2020, when the global COVID-19 pandemic resulted in international travel restrictions and changes in people’s travel capabilities and intentions.

The sudden closure of Australia’s international border, the duration of the travel restrictions, and subsequent levels of overseas migration after restrictions were lifted, have combined to give rise to a period with the most volatile changes in overseas migration flows in Australia’s history. 

One flow-on effect from this volatility, has been larger revisions to preliminary overseas migration estimates. 

NOM is the difference between the number of incoming travellers who are added to the population (overseas migrant arrivals) and the number of outgoing travellers who are subtracted from the population (overseas migrant departures). In order to determine which border crossings are overseas migrant arrivals and departures, the ABS uses a threshold of twelve months duration in or out of Australia, as this aligns with the United Nations (UN) definition of international migration. The UN definition is operationalised in Australia through the 12/16 month rule:

• A person who is not already resident in Australia, is added to the estimated resident population (ERP) if they travel to Australia and remain present in Australia for 12 months or more (an overseas migrant arrival) . Likewise, an Australian resident is removed from the population if they leave Australia for 12 months or more (an overseas migrant departure).  
• The 12 months does not have to be continuous and is measured over a 16-month period. This is important, as it means someone who has migrated to Australia can take a short trip outside Australia within the 16 month-window and still count as a migrant arrival. 

In order to produce timely population estimates, a person considered to have migrated, is added to, or removed from, the Australian population from the date they arrived or departed, i.e. from the beginning of the 16-month calculation period.

Being based solely on duration of presence in Australia (or absence from Australia), NOM estimates include all people regardless of their initial reason for travelling, country of citizenship or visa status; however foreign diplomatic personnel and their families are excluded. 

To produce NOM, the ABS uses border crossing data collected by the Department of Home Affairs via electronic immigration systems at the border and from incoming passenger cards. Traveller durations in Australia are calculated using the 12/16 month rule to identify which border crossings qualify as migrations. 

Generally, most border crossings do not result in a migration. For example, during 2018, there were over 42 million border crossings in Australia, however only 1.9% of these crossings qualified as migrations. The overwhelming majority of movements into and out of Australia are for short trips such as holidays and business travel.

From July 2017, due to the removal of the outgoing passenger card (a declaration travellers would fill out when departing Australia), the ABS has also used Medicare enrolment data as a secondary source of state of residence information for a small proportion of Australian residents where this not available in the Home Affairs data.

This border crossing data, which is the input data for NOM, is also published by the ABS in Overseas Arrivals and Departures, Australia.

Some examples of the 12/16 rule in practice are: 

• Example 1: Traveller A, not an Australian resident, enters Australia in July 2018. They are away from Australia in April 2019 but return and continue to stay in Australia until November 2019. In the 16 months following their entry into Australia, they spent 15 months in Australia and 1 month out, meaning they have spent at least 12 months out of 16 in Australia. Traveller A is counted as an overseas migrant arrival and is included in the population from July 2018. Under a model requiring 12 continuous months spent in or out of Australia to identify a migration movement, Traveller A would not have been an overseas migrant arrival in July 2018.
• Example 2: Traveller B, an Australian resident, departs Australia in January 2017 then returns in August 2017. They do not leave Australia again before May 2018. From January 2017 to May 2018, the traveller spends 9 months in Australia and 7 months away. The movement in January 2017 is therefore not regarded as an overseas migrant departure and Traveller B remains included in Australia’s population, even while they are away.

Final overseas migration estimates can only be calculated when Australian border crossing data becomes available for the 16 months following a reference period. To produce more timely NOM estimates, modelled preliminary NOM is produced which allows travellers to be added or subtracted from the population, from their initial date of arrival or departure, without the ABS waiting for the full 16 months to elapse.

Estimates of overseas migration are required less than six months after the reference quarter for the production of quarterly ERP. At that time, complete travel data for the 16 months following a reference quarter are not available meaning that producing NOM estimates required for ERP would require waiting a further 12 months.

To obtain the required estimates in the absence of the full 16 months of travel data, preliminary overseas migration estimates are initially modelled. This is achieved by using the 12/16 month rule results of travellers with similar characteristics from one year earlier. These results are then applied to travellers for the current period. This method is used because the behaviours of earlier travellers tend to be a very good predictor for travellers one year later.

The characteristics of travellers used to find similar groups for preliminary modelling are: age, country of citizenship, direction of first and last movement in the reference quarter, initial ERP status, time spent out of Australia, and visa type. To create the preliminary estimates, the behaviour (or likelihood of migrating) of travellers in the corresponding group from one year earlier is applied to the same group in the current reference quarter. Applying this behaviour from earlier travellers then determines whether current travellers will be counted towards preliminary NOM arrivals or departures.

However, the migration behaviour of travellers changed during the COVID-19 pandemic due to the disruption to travel patterns caused by international border closures. At such a time, it was not appropriate to use patterns of behaviours from one year earlier as this would have resulted in very large revisions once all the travel data became available. From 1 January 2022 to 31 March 2023 for example, preliminary estimates were modelled on traveller behaviour from the corresponding quarter of 2018 as this was the most recent year that was completely unaffected by changed travel behaviours resulting from the pandemic (for many people who travelled in 2019, plans to leave or return to Australia in early 2020 ended up being changed).   

The majority of overseas travellers in a given reference quarter take a trip of short duration. The ERP status (i.e. whether a person is included in the population estimates or not) of most short-term travellers is already known when ERP is produced six months after the reference period (with travel data as recent as six weeks earlier feeding into the calculation). This means for many individuals that their final NOM status has already been accurately determined and will not need to be revised in future quarters.

For example, an Australian resident already counted in the ERP at the beginning of a quarter, may commence a holiday during that quarter and then return to Australia two weeks later. Once four months since their return has passed without a further departure, they can be determined to still be in Australia’s population at the end of the reference quarter. This is because it will no longer be possible for them to be outside of Australia for more than 12 months of the 16 months following their original departure. 

This determination of the ERP status of short-term travellers reduces the proportion of travellers which are modelled to around 30% of people who travelled in that quarter. This substantially increases the predictive accuracy of the model. 

As mentioned above, most short-term travellers can have their final NOM outcome determined based on data available at the time of estimation. However, as more time passes after the initial overseas movement, this outcome can be determined for more individuals. 

In response to increased volatility and uncertainty in preliminary NOM estimates due to COVID-19 travel restrictions, the ABS implemented a more frequent revision cycle to take advantage of the most recent data available each quarter. As a result, preliminary NOM is revised through four preliminary stages until it becomes final in its fifth iteration. The revisions progressively reduce reliance on modelling from around 30% of travellers initially, down to around 1% of travellers prior to the final NOM calculation. This progressive increase in the use of real travel outcomes, rather than modelled traveller outcomes, increases the accuracy with each revision. 

In addition to a smaller proportion of travellers to model, more travel data (e.g. initial ERP status, time spent out of Australia) is also then known about the travellers in this smaller pool. This allows even better alignment to past traveller behaviour, making the modelling more accurate again.

By revising the preliminary NOM each quarter rather than waiting one year, any revisions that may be required are able to be brought into the population estimates sooner, reducing the size of future revisions which would otherwise need to be made.

As the preliminary NOM model uses behaviour of travellers from one year earlier, its accuracy can be affected by changes in traveller behaviour over time. If these changes occur relatively slowly or progressively, the migration behaviour of the current travellers will not differ considerably to those from one year earlier and the preliminary estimates will be highly reflective of the final NOM results. 

However, larger than usual revisions to NOM may occur if the migration behaviour of travellers changes rapidly or suddenly. This is because the travel behaviour of the current travellers could be different enough from a year earlier despite them having similar characteristics. This would result in actual migration outcomes which are different from those produced by the preliminary NOM model. The COVID-19 pandemic provided notable examples of this.

For example, in December 2019, many international students left Australia for a short trip at the end of the academic year, planning to return and continue their studies in early 2020. When they couldn't return due to border closures, the preliminary NOM model's reliance on behaviour of students who had departed a year earlier with no difficulty in returning, led to a larger than usual revision of student departures for that quarter. This sudden change in migration behaviour could not be wholly predicted by the model.

Another example relates to the surge in volumes of migration to Australia which occurred after the borders reopened, resulting in rapid increases in the rates of overseas migrant arrivals. With behaviour changing so rapidly, the differences in behaviour between the preliminary NOM reference period compared with one year earlier were large. The largest of these behavioural differences were mitigated by modelling using 2018 behaviours, the most recent full year that reflected travel without COVID-19 restrictions. This mitigation was in place until such time that borders had been open for over a year. Modelling which used behaviours from one year earlier was then able to be resumed in June quarter 2023.

Over the year leading up to the June quarter 2023, migrant arrivals had increased by 24%. This increase was due to higher numbers of travellers as well as changes in traveller behaviour. While any effects of increased travel volumes were immediately seen in the preliminary NOM estimates, the effects of changes in behaviour were brought into subsequent estimates through the revision cycle.