Creating a child mobility indicator using the Life Course Dataset

While no single environment defines a child’s outcomes, the places where children live can shape access to schools, services, green space, and community connections.[1] Understanding child mobility in Australia, 2006-2021 is a new experimental release using administrative data in the Life Course Dataset to provide insights into patterns of mobility among children (zero to 14 years) in Australia from calendar years 2006 to 2021. To explore this, a child mobility indicator was developed as part of the pilot Life Course Data Initiative (LCDI). The LCDI aims to build an evidence base for long-term policy responses that improve wellbeing for children and their families.

The Life Course Dataset is a longitudinal administrative dataset that enables linkage of individuals to anonymised[2] address history information from 2006 onward. This makes it possible to observe changes in place of residence (i.e. mobility) over time. By capturing mobility across years, the Life Course Dataset can provide a broader understanding of children’s mobility patterns than is available from the Census of Population and Housing alone, supporting analysis of how mobility intersects with experiences of disadvantage and relates to later wellbeing and life outcomes. 

In contrast, the Census measures mobility at a single point in time by comparing a person’s place of residence on Census night with where they lived one or 5 years earlier. This enables classification of individuals as movers or non-movers, providing valuable information on whether a change in residence has occurred at a point in time. However, because the Census is anchored to one reference point and collected every 5 years, it cannot capture individuals' full mobility trajectories, such as how often they relocate or the distances they move over their life course. Longitudinal administrative data offers an important complement, enabling analysis of sustained and repeated mobility over time. 

In this information paper, we outline the methodological approach used to understand child mobility in Australia using the Life Course Dataset. The paper includes:

• the definition of child mobility used in the Life Course Dataset
• the methodology for deriving the child mobility indicator in the Life Course Dataset
• a comparison with Census measures
• considerations for interpreting child mobility patterns using the Life Course Dataset.

Child mobility refers to changes in children’s place of residence as they grow up. For the purposes of creating the Life Course Dataset child mobility indicator, mobility is defined by the geographic scale of moves within Australia. We distinguish between local or short distance moves occurring within the same Statistical Area Level 4 (SA4) (residential mobility) or longer distance moves across SA4s (internal migration). SA4s are part of the Australian Statistical Geography Standard and are designed to represent distinct labour markets and regional economies. This classification is adapted from Wu (2025).[3]

The child mobility indicator is created using anonymised address histories from administrative data, which record where a child has lived over time based on interactions with government services. These address histories are organised into a series of location episodes, each representing a period of residence and including a start and end date, as well as geocoded location information that indicate the place of residence. Geocoded location information comprises geographic references such as a hashed Address Register Identifier (ARID) that represents each address in the ABS Address Register, and Australian Statistical Geography Standard (ASGS) codes. Using address histories, a child is considered to have moved when the geocoded location information across successive location episodes indicates changes in place of residence.

When counting moves, we apply the following rules: 

• Exclude moves (and children) where location could not be determined: Moves are excluded when the location of either the origin or destination residence is unknown because of insufficient information in the administrative data. Less than 1%–4% of children per year made only unknown moves.
• Treat moves with no change in location as non-moves: Moves with no change in geocoded location information between the origin and destination residences were treated as non-moves.[4] This may occur for a range of reasons. It may reflect real moves where children move out of and back into the same residence or it may be due to administrative processes.

The population for the Life Course Dataset child mobility indicator includes children aged zero to 14 years in calendar years 2006 to 2021 who are part of the Life Course Dataset population. This is a scoped population created from the Person Level Integrated Data Asset (PLIDA). Inclusion in the analysis requires children to have address history information recorded for any year(s) in which they are in scope. 

Between 2006 and 2021, 98.4%–99.2% of children in the Life Course Dataset population had address history information and were included in the analytic population. The analytic population changed over the study period as, for example, children were born, died, immigrated, emigrated or aged out of the cohort.

The child mobility indicator is created using the Life Course Dataset, which draws on integrated data from PLIDA, a secure data asset combining information on health, education, government payments, income and taxation, employment, and population demographics. This linked longitudinal data based on interactions with government services provides a comprehensive picture of Australia’s population. PLIDA can only be used for statistics and research.

To create the Life Course Dataset mobility indicator, there are four key steps:

• Step 1: Creating an ever-resident administrative population.
• Step 2: Scoping the child population at each snapshot date.
• Step 3: Linking anonymised address histories to the child population.
• Step 4: Deriving the child mobility indicator in the Life Course Dataset.

This population is created from the PLIDA Person Linkage Spine, which combines records from the Medicare Consumer Directory (MCD), Data Over Multiple Individual Occurrences (DOMINO) Centrelink Administrative data and Personal Income Tax data from the Australian Taxation Office (ATO). 

Anyone who has appeared in at least one of these datasets since 2006 is included in the ever-resident population. The aim is to cover all people who were living in Australia at any time from 2006 to 2021. 

To create the child mobility indicator in the Life Course Dataset, the second step is to determine which people in the ever-resident population were living in Australia at the snapshot date of 30 June for each year from 2006 to 2021. 

This process, known as scoping, applies a series of rules to align the ever-resident population with the concept of the Estimated Resident Population (ERP). The ERP is the official population measure and includes all people who usually live in Australia, including those temporarily overseas. It excludes short-term visitors.

For details on how this process is carried out, refer to Scoping the population in the Life Course Dataset. 

Once the whole population is scoped, we further scope it to only include children aged zero to 14 years at the snapshot date in each year, based on month and year of birth recorded in the administrative data. This enables us to determine the resident child population at the snapshot date for the years 2006 to 2021.

To create the child mobility indicator from the Life Course Dataset, the third step is to link anonymised address history information spanning 1 January 2006 to 30 June 2025 to in‑scope children. While the analytic period focuses on mobility occurring between 2006 and 2021 for children aged zero to 14 years, address records from later years are included to accurately capture the type of moves that are initiated within this period, but where the destination address was recorded after 2021 or when the child turned 15 years. 

Address history information is taken from the Core Locations module, a PLIDA dataset that consolidates address information from multiple datasets to determine a single location for each person from 2006. The source datasets for the Core Locations module are: 

• Medicare Consumer Directory (MCD)
• Australian Taxation Office (ATO)
• Centrelink Administrative Data (DOMINO).

Instead of storing full address text, the Core Locations module uses a secure hashed ARID that represents each address in the ABS Address Register. Hashing transforms the ARID into a unique and anonymised numerical code that does not contain the original address information. To create the child mobility indicator, we used a version of the Core Locations module that was developed for forming household structures using the Life Course Dataset. This version uses imputation to fill the gaps when hashed ARIDs are missing. 

The Core Locations module contains address history information for persons in PLIDA. Address history information is stored as a series of location episodes, which include the hashed ARID, other geocoded location information, and a start and end date. Some children may have at least one address history that starts prior to the year in which they were born. These records potentially reflect that the location episode was inherited from their parent(s). Given these records may reflect valid address histories for the child, they were retained in subsequent analyses.

To create the child mobility indicator, we use each child’s address history and determine when a change in residence occurs. The process works as follows:

Step 4.1: Examine location episodes in address histories to identify potential moves.

Each child’s address history consists of one or more location episodes. A location episode represents a period when the child lived at a particular residence and includes a start and end date, an ARID and geocoded location information based on the 2021 ASGS. A potential move is observed when a location episode ends. 

Step 4.2: Compare successive location episodes to determine whether a move has occurred and classify its type.

Where a child has successive location episodes, we compare the ARID, Statistical Area 1 (SA1), and Statistical Area 4 (SA4) of the origin (previous episode) and destination (next episode) to determine whether a change in residence has occurred. These are based on 2021 ASGS codes. Moves are classified following Wu (2025) [3]:

• Local move: The destination episode has a new address (different ARID) but the same SA1. This is akin to moving within a local area.
• Short distance move: The destination episode has a different SA1, but the same SA4. Given that SA1s represent neighbourhood-scale geography (around 400 people), changes between SA1s within an SA4 likely capture short distance moves.
• Long distance move: The destination episode has a different SA4. SA4s represent large regional areas (above 100,000 people), often corresponding to labour markets or distinct regions, making them useful for identifying long distance moves. 

The type of move made by a child is classified according to the geographic scale of the change rather than the exact physical distance travelled. For example, a move within a large labour market region may cover a long physical distance, while a move between labour market regions may be relatively short. Classifying moves by type provides a consistent and interpretable measure of mobility.

Step 4.3: Assign the move year. 

Each move is assigned to the year in which the previous location episode ended. This is referred to as the move year.

The Life Course Dataset child mobility indicator is an experimental product. To understand its quality, patterns of child mobility from the Life Course Dataset were compared to the 2021 Census using the Usual Address One Year Ago (UAI1P)  and Five Years Ago (UAI5P) variables. While the patterns of child mobility between the Life Course Dataset and 2021 Census show close alignment, there are some differences, which can be attributed to conceptual and technical differences between the datasets.

These include:

• Differences in population scope: The Life Course Dataset aims to reflect the population scope of the ERP, which includes usual residents of Australia even if temporarily overseas. In contrast, the Census includes only persons present in Australia on Census night. To better align population scope across datasets, we excluded persons from the Census who were overseas in 2020 or 2016, overseas visitors in 2021, or for whom UAI1P or UAI5P were not applicable.
• Address reporting errors: Administrative data may contain inaccuracies due to delayed or incomplete address updates, individuals approximating their location, or misreporting when they moved. Census data may also include inaccuracies due to recall errors. 
• COVID-19 impact: Australia’s response to the COVID-19 pandemic had an unprecedented impact on administrative data and the lags associated with it. For more information see Net interstate migration review. 
• Differences in how data is captured: The Census and Life Course Dataset capture mobility differently. For example, the Census provides a snapshot of where people lived at the usual address one year ago and 5 years ago, whereas administrative data records addresses when they are reported during interactions with government services.
• Unknown moves: In the Census UAI1P or UAI5P variables, a ‘not stated’ response means there was insufficient information to determine whether a person had changed address one or 5 years ago. In contrast, in the Life Course Dataset, ‘unknown moves’ occur when a location episode ends, but it is not possible to confirm whether the subsequent address is different from the previous one.

When comparing a child’s previous location in the Life Course Dataset with the Census at an aggregate level, results align well. In 2021, 14.7% of children aged one to 14 years in the Life Course Dataset are recorded to have moved one year ago, compared with 14.2% in the Census. A move one year ago in the Life Course Dataset is determined when a child’s ARID differed on the Census month (August) in 2021 as compared to a year prior (2020), whereas in Census it is people with UAI1P = 2 (Elsewhere in Australia). 

The percentage of children aged 5 to 14 years recorded in the Life Course Dataset as having moved 5 years ago (48.1%) broadly aligns with the Census (41.5%). A move 5 years ago in the Life Course Dataset is determined when a child’s ARID differed on the Census month in 2021 as compared to 5 years prior (2016), whereas in Census it is people with UAI5P = 2 (Elsewhere in Australia).

The distribution of mobility one year ago by age between the Life Course Dataset and the Census aligns well. The Life Course Dataset indicates 16.1% of children aged one to 4 years in 2021 were living in a different location one year ago (17.9% in the Census). This prevalence reduced with age. The Life Course Dataset indicates 14.4% of children aged 5 to 9 years and 13.9% of children aged 10 to 14 years were living in a different location one year ago. Similarly, the Census indicates 14.0% of children aged 5 to 9 years and 11.7% of children aged 10 to 14 years lived in a different location a year ago.

There is also alignment in the percentage of children aged 5 to 14 years in 2021 who moved 5 years ago, with less than 6 percentage points difference between the Life Course Dataset and Census for children aged 5 to 9 years, and less than 8 percentage points difference between data sources for children aged 10 to 14 years. Both data sources show the percentage of children with a different current and previous location 5 years ago reduces with age. 

The Life Course Dataset child mobility indicator is an experimental product. At the national level, the Life Course Dataset child population with address history information closely aligns with ERP for each year from 2006 to 2021. Over this time, the Life Course Dataset child population was between 0.5%-1.3% lower than ERP. While the overall population counts in this dataset broadly align with ERP, the assignment of children to specific geographic locations may differ. This reflects factors such as address availability, quality, and update frequency in administrative data. Other factors, such as those listed below, should also be considered when using and interpreting this product.

• The child mobility indicator is based on address histories recorded through interactions with government services. Children who did not link to the Core Locations module were not included (0.8%–1.6% per year). These children may have no or insufficient address information recorded with government services. This means that any moves made by those children are not reflected.
• Incomplete address histories contribute to moves being classified as ‘unknown’. While incomplete geocoded location information can occur in different regions, it is more common in remote areas where geocoding accuracy is typically poorer. Among children with address histories between 2006 and 2021, 4.3%–11.2% were missing an ARID, 3.7%–8.4% were missing an SA1, and 0.9%–1.7% were missing an SA4. Where geocoded location information was missing for either the previous or current location episode, a change in residence could not be confirmed and the resulting move was excluded from the analysis. Children were excluded from the analysis if all of their moves were classified as ‘unknown’. 
• The Life Course Dataset population scoping approach overestimates the number of infants aged zero years across all years from 2006 to 2021. This overestimate is due to challenges in applying age-based activity scoping rules for infants in administrative data, and may affect mobility estimates at age zero years. For more information see Scoping the population in the Life Course Dataset.

• Address history information in administrative data is based on interactions with government services. Families may not always update their address straight away or at all following a move, or may misreport when or to where they moved. This means that some children in administrative data may be incorrectly counted as having moved or not moved in a particular year. 
• Undercounting mobility is likely for populations that interact irregularly or not at all with government services, such as families who are highly mobile or experiencing vulnerability. Short‑term stays with relatives, refugees, or transitional housing may also not appear in administrative data, potentially leading to underestimating mobility.
• Individuals are assigned to no more than one address per location episode. This means more complex living arrangements, such as children living in shared care arrangements, were not captured, potentially making children appear more residentially stable.

This release has been produced via a collaboration with the Centre for Community Child Health of Murdoch Children’s Research Institute.

1. Goldfeld S, Villanueva K, Tanton R, Katz I, Brinkman S, Giles-Corti B, Woolcock G (2021), 'Findings from the Kids in Communities Study (KiCS): A mixed methods study examining community-level influences on early childhood development', PLOS ONE, vol. 16, no. 9, e0256431, accessed 30 April 2026.
2. An anonymised address refers to a secure hashed Address Register Identifier (ARID) that represents each address in the ABS Address Register. Hashing transforms the ARID into a unique and anonymised numerical code that does not contain the original address information. 
3. Wu G (2025), ‘Repeat internal migration: a life-course trajectory approach’, PhD Thesis, School of the Environment, The University of Queensland, accessed 30 April 2026.
4. Robertson O, Nathan K, Howden-Chapman P, Baker MG, Carr PA, Nevil P (n.d.), ‘Residential mobility for a national cohort of New Zealand-born children by area socioeconomic deprivation level and ethnic group’, BMJ Open 2021;11:e039706, doi: 10.1136/bmjopen-2020-039706, accessed 23 July 2026.