Report on the quality of 2021 Census data: Statistical Independent Assurance Panel to the Australian Statistician

The Australian Bureau of Statistics (ABS) conducts a Census of Population and Housing (the Census) every five years. The data collected provides valuable information for policy making and research and is used widely, including setting electoral boundaries and other such applications.

On 18 June 2020, the ABS announced that Census night for Australia’s 18th Census would be 10 August 2021. The ABS planned the 2021 Census to achieve three strategic outcomes:

• Smooth running: The Census process is easy, simple, and secure.
• Strong support: Governments, businesses and the community have confidence in the Census and there is a high level of community participation.
• High quality data: Census data is high quality and widely used to inform on policy and areas of importance to Australia.

As in 2016, the 2021 Census was planned to be multi-modal, with the expectation that approximately 75% of Australian households would complete their Census online. This expectation was exceeded, with nearly 80% of responding households completing their Census online. This was enabled by the development of an entirely new online Census form and digital service, as well as the implementation of a communication campaign that supported a window of time to complete the Census, rather than a single night. The overall dwelling response rate for Australia was 96.1%, an excellent outcome given the target of greater than 95% and the challenges of the pandemic.

People were actively encouraged to complete their Census form online. A letter was delivered to most households with a unique login ID and instructions to complete online. Paper forms were delivered to households where completing online was less likely to be preferred or possible, and paper forms were also available on request to everyone in Australia. The ABS ensured a range of innovative approaches were used so that the enumeration of the entire population was as complete as possible. For the first time, people could complete their Census online without having a Census login code (see Appendix E).

The 2021 Census collected responses on two new topics: Long-term health conditions and Australian Defence Force service. These were the first significant changes to the information collected in the Census since 2006. Changes were also made to the way certain topics were collected. Notably, response categories of 'Aboriginal' and 'Torres Strait Islander' were added to the Ancestry question; respondents had the option of identifying as non-binary sex; and the Language question was updated to refer to language used (rather than language spoken), to enable inclusion of sign languages such as Auslan.

The emergence of the world-wide COVID-19 pandemic in early 2020 introduced a range of challenges across the globe for people, communities, governments, businesses, and countries. Australia experienced the pandemic through several waves during 2020, 2021 and 2022, resulting in the 2021 Census being conducted at a unique and challenging time in Australia’s history. Much of south-eastern Australia was in lockdown on or about Census night, equating to over half of Australia’s population. The ABS’s response was impressive and well managed. Further details on ABS’s management of the Census in the pandemic are available in Section 4 of this report ‘COVID-19 and the 2021 Census’.

The 2021 Census data, collected at such an important and unique time in Australia’s history, provides a timely, contemporary, and detailed view of the circumstances of Australia’s people and communities during the COVID-19 pandemic.

In recognition of the importance of Census data, and the importance of being transparent about data quality, the Australian Statistician established the 2021 Census Statistical Independent Assurance Panel (the Panel). The Panel was established in late 2020 to provide assurance and transparency for the 2021 Census data.

Such a panel had first been established following the 2016 Census. The Report on the Quality of 2016 Census data produced by the 2016 Census Independent Assurance Panel noted that the establishment of an Independent Assurance Panel should be repeated for future censuses. The ABS accepted this suggestion and all other opportunities identified by the 2016 Census Independent Assurance Panel. More information on how the opportunities identified in the 2016 report were addressed is provided in Appendix A.

This report presents the findings of the Panel following its review of the quality of the 2021 Census data and provides information to enable governments, the community, and other stakeholders to make an informed judgment about the fitness-for-purpose and credibility of the 2021 Census data.

The Panel, comprising a group of experts with diverse experience and backgrounds, was tasked by the Australian Statistician to provide an independent assessment of the quality and fitness- for-purpose of the 2021 Census of Population and Housing data.

The Terms of Reference for, and membership of, the Panel are in Appendix D.

The Panel undertook this review as a body independent of the Australian Statistician.

This report provides a high-level analysis of the quality of the 2021 Census data and focuses on key aspects of the Census data. It does not provide detailed quality reports on all data items or geographies.

The Panel was provided access to data from the 2021, 2016 and 2011 Censuses, data from the 2021 Post Enumeration Survey, information on Census processes, and other information as requested. The Panel was also provided Secretariat support from the ABS and engaged throughout the review process with ABS and Census officials.

Under its Terms of Reference, the Panel was required to consider the quality effects of Census design, enumeration, and processing, and to compare the 2021 Census results with the previous two Censuses and various other ABS and non-ABS data sets. The Panel also compared the quality indicators of the Census outputs against selected international censuses. The Panel’s assessment of quality was also guided by the accuracy framework described in Appendix C.

The ABS conducts a Census of Population and Housing every five years as required by the Commonwealth Census and Statistics Act 1905. Regularly taking a census provides a comprehensive snapshot of the nation and enables the updating and maintenance of an accurate time series of Australia’s official population estimates.[1]

The Census counts everyone in Australia on Census night and asks about where they were on the night (referred to as their place of enumeration), their place of usual residence and numerous other questions. In 2021, Census night was designated as 10 August. While Census night is the reference night in respect of which people are asked to complete their Census forms, there is no requirement for people to complete their Census form on Census night. The ABS emphasised this in 2021 with a message to the public that they had a ‘window’ of time (before, on and after Census night) to respond to the Census, to increase response (for more detail see Section 2.3).

While the public were encouraged to complete their Census in this ‘window’, the online Census form was open for approximately 10 weeks. Paper forms were accepted for as long as practicable to allow as many people as possible to respond, with the last paper form included in Census processing having been received in December 2021.

The Census collects data on a broad range of topics including Marital status, Family size, Occupation, Language used at home, Country of birth, Australian citizenship, Income, and Ancestry. In 2021, two new questions, on Long-term health conditions and Australian Defence Force service, were included in the Census. In addition, a third response category was added to the Sex question, to allow people to identify as non-binary sex. Also, the Ancestry question was expanded to add tick boxes for Aboriginal and Torres Strait Islander ancestries.

[1] Australian Bureau of Statistics (2019). Planning the 2021 Census. Available at https://www.abs.gov.au/statistics/research/planning-2021-census#more-information

The purpose of the Census is to measure the number and key characteristics of persons and dwellings in Australia on Census night. This provides a reliable basis to estimate the population for each state and territory and local government area.

The Census also provides the characteristics of the Australian population and its housing for small areas and specific population groups. Census data is used by individuals, organisations, and government to make informed decisions on issues that impact the lives of all Australians.

The ABS identifies four main uses for the Census.[1]

1. Allocation of government funds and support for elections
   The official population estimates, which are based on Census population counts, are used in the process of allocation of Commonwealth funds to state, territory and local governments, including Goods and service tax (GST) entitlements, and to determine the number of seats each state and territory has in the House of Representatives. The GST entitlements of the states and territories will amount to more than $80 billion in 2022-23.[2] The population estimates are also used as part of the process for determining electoral boundaries for both Australia and most states and territories.
2. Planning and administration
   The Census provides the characteristics of the population and its housing to support the planning, administration and policy development activities of governments, business and other users. For example, these characteristics have been used to inform the planning of new hospitals and schools.
   While information from other sources provides useful input, only the Census can provide accurate information, nation-wide, and for small geographic areas and population groups.
3. Use in other ABS statistics
   Census data form the basis of many of the ABS’s most widely used products and services. One is the official population estimates, which are updated each quarter. Population estimates are used in compiling the monthly employment and unemployment statistics and the National Accounts. The Census counts of people and dwellings for each geographic area are also used in the framework for selecting the samples for ABS household surveys.
   In recent years, the ABS has sought to further maximise the use of the Census. Along with other Commonwealth partner agencies, it has linked Census data with a wide range of administrative data to create the Multi-Agency Data Integration Project (MADIP), a secure dataset designed to improve and expand the range of statistical and research insights about Australia. MADIP combines data from the Census with information on health, education, government payments, employment, income and taxation, and employment.
4. Community uses
   The ABS releases Census results in analytical articles, QuickStats, Community Profiles, TableBuilder and DataPacks.[3] Governments, members of the public, businesses, researchers and community groups use these products to find comprehensive information for small geographic areas and population groups as well as for states, territories and the country as a whole. In addition to direct engagement with ABS products, many people access ABS statistics indirectly, such as through media articles and reporting of aspects of ABS findings.

[1] Australian Bureau of Statistics (2011). How Australia Takes a Census, 2011: Uses of Census Data [Government Paper]. Retrieved from http://www.abs.gov.au/ausstats/abs@.nsf/lookup/2903.0Main%20Features132011

[2] Australian Government (2022). Part 3: General revenue assistance’, Budget 2022-23. Available at https://budget.gov.au/2022-23/content/bp3/download/bp3_13_part_3.pdf

[3] These are standard Census products available from the ABS website.

One of the most important and well-known uses of the Census is its contribution to Australia’s official population estimates, the Estimated Resident Population. The Estimated Resident Population is compiled through a combination of Census population counts, Post Enumeration Survey estimates, and other administrative sources.[1]

The Post Enumeration Survey[2] is a survey of about 50,000 dwellings, conducted soon after the completion of the Census period. It provides an independent benchmark of the completeness of Census counts. Key changes introduced for the 2021 Post Enumeration Survey were the use of a telephone-first interview approach, and the use of the Address Register (a comprehensive list of all known physical addresses within Australia) for the survey frame.

The Post Enumeration Survey interview process determines whether a person in the sample should have been counted in the Census, and the category in which they should have been counted (such as age, sex, or state of usual residence and whether they are of Aboriginal and/or Torres Strait Islander origin). A matching process determines if they were counted, how many times they were counted, and the characteristics they were counted with. This is done by directly linking Post Enumeration Survey persons and dwellings to their matching Census forms (where they exist). This enables estimates of net overcount or net undercount, which are then applied as part of the five-yearly rebasing of the Estimated Resident Population.

The Estimated Resident Population is a population estimate by geographic area, age and sex and is used to inform many planning decisions by governments and investment decisions by businesses. It is used to plan and forecast needs for housing, schools, hospitals, shopping centres, child and aged care services and transport infrastructure. It is used in the National Agreement on Closing the Gap and more broadly in a range of forecasting and modelling. The Estimated Resident Population is also used to guide the distribution of government funds to states and territories, which flow on to smaller areas such as local governments and is a major factor in the determination of electoral boundaries. The calculation and frequent updating of the Estimated Resident Population is part of the statutory role of the ABS. The Estimated Resident Population forms the basis of the population projections made from time to time by the ABS.

After every census, the Estimated Resident Population is recalibrated, or rebased. Rebasing is done by adding the net overcount or net undercount from the Post Enumeration Survey to the new Census population counts. Further demographic and time adjustments are then made before the Estimated Resident Population data is finalised. Further information on this process is provided at Appendix B.

The Post Enumeration Survey is important as it provides:

• an independent measure of the completeness of the Census counts;
• a critical component for rebasing of the Estimated Resident Population to the latest Census; and
• assistance in identifying improvements for future censuses.

[1] For more information on the Estimated Resident Population, see National, State and Territory Population. This can be accessed at: https://www.abs.gov.au/statistics/people/population/national-state-and-territory-population/latest-release [2] Information regarding the structure and methodology of the Post Enumeration Survey can also be found at: Australian Bureau of Statistics (2022). 2021 Census overcount and undercount. Available at  https://www.abs.gov.au/statistics/people/population/2021-census-overcount-and-undercount/2021

As in 2016, the 2021 Census was ‘digital first’. Most households received Census instructions in the mail to complete online, while paper forms were readily accessible via an automated phone service or through online self-help. Some areas where there were, for example, higher numbers of older people or areas where previous Census response had been lower, received a paper form upfront. People living in remote areas and most people experiencing homelessness had help from Census field staff.

The enumeration model for 2021 built on the success of the 2016 approach, which was based on mail out areas (areas which meet criteria for residential address mailability) and drop off areas (areas with poor address information or difficult to access by mail). The proportion of mail out areas increased from 80% in 2016 to 85% in 2021.

There were three phases of enumeration for the 2021 Census - Approach, Reminder and Visit:

• Approach phase, when letters and, in some areas, paper forms were mailed to addresses (85% of dwellings), with deliveries by early August. The letters requested online participation on Census night but with the option of requesting a paper form. The remaining dwellings (15% of dwellings) were visited by a field officer who hand delivered a paper form. People in these areas still had the option of completing their forms online with online access codes printed on the paper forms. The aim of the approach phase was to ensure that all dwellings were ‘enabled’ by Census night, having received either a letter or a paper form by that date.
• Reminder phase began after Census night with the start of reminder activities to prompt response. The standard reminder was a letter with online access codes mailed to non-responding dwellings in mail out areas.
• Visit phase, when field officer visits to non-responding dwellings commenced. Visits to non-responding dwellings in drop off areas commenced earlier than in mail out areas because reminder letters could not be sent to these addresses.

A key feature of the 2021 Census was a change in the messaging about when people should complete their Census. In 2021, Census Instruction Letters were delivered to dwellings across Australia with the banner ‘It’s time to complete your Census – online form open now'. This messaging, reinforced by the Census media campaign, subtly encouraged households to complete their Census as soon as they received their materials and therefore before Census night on 10 August. The letters also emphasised 12 August as a date after which the household may receive contact from the ABS. This messaging was designed to provide a ‘response window’, whereas in previous censuses only Census night was highlighted. Encouraging households to complete their Census early gave them greater flexibility, potentially increasing overall Census response. It was also designed to smooth the response over Census night and thereby assist the ABS in its management of operations. Census results show that the ‘response window’ influenced the timing of response in the 2021 Census, with more than 3 million private dwellings responding before Census night, significantly more than responded early in previous censuses (see Figure 2.3.1). While a significant number still responded on Census night (over 2.5 million), the window was effective in prompting early response and smoothing the submission of Census forms.

Note: Data excludes remote area operations.

Strategies were put in place to maximise the coverage and response of specific population groups, geographic areas and dwelling types. Prioritised population groups included Aboriginal and Torres Strait Islander peoples in urban, regional and remote areas, people from culturally and linguistically diverse backgrounds, people experiencing homelessness, people in aged care, incarcerated people and people with disabilities. As in 2016, various strategies were put in place but engagement with stakeholders was a key strategy across all population groups, with the level of engagement scaled up for 2021. Information hubs were a new aspect of face-to-face assistance in 2021, with Census engagement staff setting up information booths in the community to answer questions, issue paper forms and raise awareness.

A Remote Area Strategy was developed to enhance the enumeration of remote Aboriginal and Torres Strait Islander communities across Australia, covering 51% of the country and including more towns than in 2016. A key change for 2021 was the move to an Aboriginal and Torres Strait Islander led engagement model with a strong emphasis on collaboration with communities and local staff employed where possible. In these areas, the Census was conducted, as in previous years, over an extended period to ensure that Census field staff could cover vast areas and help people complete the Census through face-to-face interviews where required. As in previous censuses recruitment difficulties were experienced in some parts of the country, in particular remote areas, and these were exacerbated by COVID-19 restrictions which limited the movement of field staff within lockdown areas and across state and territory borders. An innovation in the management of remote field operations was the use of a dedicated mobile app, called ‘MyWork’, which greatly improved monitoring of operations and the timely response to issues.

Following the events of Census night in the 2016 Census, when the ABS took down the online form as a precautionary measure after experiencing a series of Distributed Denial of Service attacks, and the recommendations that emerged from several subsequent reviews, the ABS prioritised security and performance, as well as data quality and accessibility in the redevelopment of its ‘digital service’ for the 2021 Census. The digital service included a new public Census website, the online form and a range of self-service options to make it easier for people to complete their Census. In preparation for the 2021 Census, the ABS worked closely with the Australian Cyber Security Centre who provided substantial insight into the development of robust census systems. In addition, the ABS welcomed a review by the Australian National Audit Office, taking all the recommendations from that review into account.

Names and addresses continue to be important data collected on the Census form. As a result of the 2021 Census Privacy Impact Assessment, which was conducted by external experts, names of respondents will be retained for 18 months and form addresses retained for 36 months.

Running a census is a huge undertaking in normal circumstances but conducting one in a pandemic added complexity. While most of the population received their letters or paper forms and participated as planned, COVID-19 lockdowns progressively disrupted some aspects of Census enumeration, reducing contact with households and non-private dwellings such as hospitals and nursing homes.

In response to the emergence of the pandemic, the ABS had developed a Census COVIDSafe step-down plan and framework to manage field staff and community safety during operations. This plan was put through its paces during the major large-scale Census test in 2020, ultimately providing confidence to the ABS, stakeholders and field staff that the 2021 Census could go ahead. However, with more than half Australia’s population in lockdown at some stage during the Census enumeration period, aspects of field operations were significantly disrupted, particularly in New South Wales and Victoria. The ABS’s management of the Census in the pandemic is explained in more detail in Section 4 ‘COVID-19 and the 2021 Census’.

Restricted field operations presented new and unique challenges to determine whether dwellings were occupied due to the combination of a ‘no contact’ approach and ways in which expected occupancy has changed due to COVID-19. However, the ABS had developed a probability model for the 2021 Census to improve the accuracy of determining whether a dwelling was occupied or not on Census night. This was used during Census processing when there was insufficient information from the field to make a determination. The model was also used to validate the final occupancy results for the Census.

A key data quality focus for the Panel has been the comparison of 2021 Census data against data from the 2011 and 2016 Censuses. These comparisons represent an important data quality check. The data chosen by the Panel for assessment was either based on direct relevance to quality such as response rates or their importance to users of the Census, such as counts of persons.

In addition to this analysis, the Panel developed an accuracy framework to explore what is meant by a high-quality census, and examined the accuracy achieved by the 2021 Census and the 2021 Post Enumeration Survey against that framework. This analysis is presented in Appendix C.

There are three key measures of statistical quality used internationally that the Panel employed to assess the quality of the 2021 Census data:

• undercount and overcount (see Section 3.2 and 3.5);
• response rates, including item non-response rates (see Section 3.3); and
• consistency with other data sources (see Sections 3.4, 3.5, 3.6 and 3.7).

The undercount and overcount estimates produced by the Post Enumeration Survey were used to obtain information on the number of people missed, or counted more than once, by the Census.

Dwelling and person response rates and item non-response rates for the Census were reviewed to see if there were any unusual observations in response patterns.

The data from the 2021 Census was compared against the 2016 and 2011 Censuses to check for unexpected differences.

Using the Post Enumeration Survey results, it is possible to estimate:

• the number of people who either were counted more than once or in error in the Census (gross overcount);
• the number of people who should have been counted in the Census but were missed (gross undercount); and
• the net error for Census imputation into dwellings determined incorrectly to be occupied or unoccupied on Census night (net overcount for persons imputed).

The net difference is called the net undercount (or overcount).

In 2021, the ABS has also produced a new coverage measure called gross coverage, which sums the coverage errors. Conceptually, gross coverage aims to identify the population that the Census counted (or received a form for) as a proportion of the population that the Census should have counted. It is the Census count, with the overcount and imputed persons removed as a proportion of the Post Enumeration Survey population estimate, expressed as a percentage.

Response rates are standard measures used internationally to assess data quality for population and housing censuses. The Panel considered three types of response rates: dwelling response, person response, and item non-response. These measures are defined below and are used in this report to assess overall data quality.

The dwelling response rate is used to measure how many occupied private dwellings in Australia completed the Census. Dwelling response occurs when a form is returned for a private dwelling identified as occupied on Census night. The dwelling response rate is calculated as a percentage by dividing the number of responding private dwellings by the number of private dwellings identified as occupied on Census night (including those where no form was returned). The dwelling response rate excludes non-private dwellings. Most people were counted in a private dwelling on Census night.

The person response rate is used to measure how many people in Australia completed the Census. Person response occurs when a person is included on a form returned from either a private dwelling identified as occupied on Census night, or from a non-private dwelling. The person response rate is also calculated as a percentage by dividing the number of responding people by the total number of people in the Census counts (including those imputed into dwellings where no form was returned).

Unlike the dwelling response rate, the person response rate includes people in non-private dwellings. This takes into consideration responses on Census night from people who were in dwellings such as hotels, hospitals, nursing homes, boarding houses, mining camps and other staff quarters, among other non-private dwelling types.

It is important to note that the quality of these response measures depends on how well the ABS has determined the occupancy of private dwellings on Census night. If unoccupied dwellings are incorrectly classified as non-responding occupied dwellings, this will make the response rates appear lower than they were. Conversely, if non-responding occupied dwellings were mistakenly categorised as unoccupied this will make the response rates appear higher.

Item non-response rates are important for understanding the quality of individual data items. Item non-response is calculated as a percentage by dividing the number of households or people who provided a response to a particular question (item) by the number of households or people (including imputed people) for whom the question (item) would have been applicable.

To assess the quality of key indicators and particular data items, 2021 Census data was compared against their equivalents from the 2016 and 2011 Censuses. While differences were anticipated (and found) given 10 years of population growth and societal change, and particularly given the 2021 Census was conducted during the COVID-19 pandemic, analysis was focussed on identifying unexpected and inexplicable differences that might point to data quality issues.

Comparisons were also made against the Estimated Resident Population as at June 2021. The Estimated Resident Population used for comparison was based on the 2016 Census updated using birth, death and migration data for the intercensal period.

The Panel also used trends in other relatable data sources to identify any trends in the Census data that might require further investigation. For the two new Census variables (Long-term health conditions and Australian Defence Force service), external data was used to assess their validity.

Net undercount for any category of person is the difference between the Post Enumeration Survey estimate of the number of people who should have been counted in that category and the actual Census count including persons imputed into non-responding dwellings during Census processing.

Net undercount is often presented as a rate. The rate is the net undercount (or overcount) as a percentage of the Post Enumeration Survey estimate for a given population (i.e. as a percentage of the number of people who should have been counted in the Census).[1] The disaggregations of net undercount (together with Census and other administrative data) are used in the compilation of the official population estimates for Australia, the Estimated Resident Population.

The total net undercount from the Post Enumeration Survey can be examined by the components of persons counted on Census forms, and persons imputed (see Section 3.2.2). This enables examination of the number of persons missed in the Census, the number of persons counted more than once or in error in the Census, and overcount due to over-imputation. Each of these components is important for understanding the quality of the 2021 Census data.

For 2021, the ABS produced a new coverage measure called gross coverage. Conceptually, gross coverage is a measure of the population that the Census counted (or received a form for) as a proportion of the population that the Census should have counted. It is the Census count, with the overcount and imputed persons removed as a proportion of the Post Enumeration Survey population estimate, expressed as a percentage. Gross coverage is useful for understanding how well the population was captured in the Census, and the proportion of the population for which the full suite of data was obtained. It is an important lens for understanding the quality of Census data.

For comparability with the Post Enumeration Survey, all Census counts in Section 3.2 are presented on a usual residence basis and exclude Other Territories.

[1] For more information on the Post Enumeration Survey or net undercount, see Australian Bureau of Statistics (2022). 2021 Census overcount and undercount. Available at  https://www.abs.gov.au/statistics/people/population/2021-census-overcount-and-undercount/2021.

The 2021 Census counted 25,417,999 usual residents of Australia (excluding Other Territories), who were in the country on Census night (including 898,484 persons who were imputed into non-responding dwellings assumed to be occupied on Census night). This was 190,044 persons fewer than the Post Enumeration Survey estimate of the usual resident population who were present in Australia on Census night. This equates to a total net undercount rate of 0.7%, down from 1.0% in 2016.

The net undercount rate for persons on Census forms decreased from 3.7% to 2.8% between 2016 and 2021, and the net overcount rate for persons imputed also decreased from 2.7% in 2016 to 2.1% in 2021 (see Table 3.2.1). Both decreases are consistent with an improvement in data quality.

As these estimates are based on a sample, they are subject to sampling errors. The sampling errors are shown for the core estimates of the net undercount in Table 3.2.1. More detailed estimates are shown in Appendix C. The net undercount rates are also subject to other types of errors such as non-response bias, recall error, matching error, and invalidity of the assumptions in the model to adjust for non-response. These are also discussed in Appendix C.

The overall net undercount can be disaggregated into a number of components of undercount and overcount, as shown in Table 3.2.2. These components are described as follows.

This comprises persons missing from completed Census forms and persons in dwellings missed in the Census (i.e. dwellings not known to Census). Gross undercount in the 2021 Census was estimated to be 1,032,660 persons, or 4.0% of the population. In 2016, this component represented 4.9% of the population (1,150,588 persons).

This comprises persons included on Census forms multiple times or included in error (e.g. they were not in Australia on Census night but were included on a Census form). Gross overcount in the 2021 Census was estimated to be 314,788 persons, or 1.2% of the population, the same proportion as in 2016 (274,673 persons). This indicates that the introduction of the No-Census Number option in 2021 which was used to encourage response (see Appendix E), and potentially may have also introduced a risk of greater duplication in responses from dwellings, but did not in the event result in increased overcoverage as most duplicate dwellings were removed in processing.

This measure is the difference between gross undercount and gross overcount, for persons on Census forms, reflecting the extent to which undercount has been offset by overcount. Net undercount for persons on Census forms was estimated to be 717,872 persons, or 2.8% of the population in 2021. The corresponding figure was 3.7% in 2016 (875,915 persons).

This component largely represents over-imputation of persons during Census processing. Over-imputation occurs when too many people are imputed into non-responding dwellings, either because the non-responding dwelling was incorrectly deemed to be occupied on Census night, or too many people were imputed into a dwelling that was correctly deemed to be occupied. The net contribution of the latter was relatively small. People were also incorrectly imputed into non-private dwellings on Census night, either due to an overestimate of the Census night occupancy of non-private dwellings, or because people were counted a second time on a form from their private dwelling residence. This process provides the balance of the over-imputation.

In the 2021 Census, net overcount for imputed persons was estimated to be 527,828, or 2.1% of the population. The corresponding figure for 2016 was 649,509 persons (2.7% of the population).

The ABS has reported that the process of determining whether a home is occupied on Census night has become more difficult with the increase in the number of people living in high-density, secure buildings and the decreasing likelihood of making doorstep contact. In 2021, COVID-19 lockdowns in some areas of Australia meant that doorstep contact was reduced, however, the introduction of self-service options on the Census website made it easier for people to report if they would not be at home on Census night. In addition, an occupancy determination model was developed for 2021 which helped to more accurately determine whether a dwelling was occupied or unoccupied on Census night in the absence of reliable field information. These two innovations have helped ameliorate the impact of restricted field activities due to the pandemic. More details of the impact of imputation are shown in Appendix C.

This is the net result of the components of undercount and overcount. Total net undercount for 2021 was 190,044 persons, or 0.7% of the population. This was a decrease of 16.1% from the corresponding figure for 2016 (226,407 persons or 1.0% of the population).

This section refers to total net undercount when distinguishing from the components of overcount and undercount for persons on Census forms or persons imputed. For the remainder of this report, total net undercount is abbreviated to net undercount unless an explicit distinction is required.

It is important to note that the Post Enumeration Survey is designed to provide the best measure of Census coverage at a single point in time, rather than as a time series across Censuses. Therefore, the Post Enumeration Survey estimates of undercount and overcount are not strictly comparable over time due to changes in both Census and Post Enumeration Survey methodologies. For example:

• a new estimation process was introduced in the 2006 Post Enumeration Survey, along with the inclusion of remote areas and discrete communities[1] in the Post Enumeration Survey sample;
• Automated Data Linking between Post Enumeration Survey and Census person records was introduced in 2011;
• the 2016 Census underwent a number of changes to its collection methodology, with some changes to Post Enumeration Survey enumeration procedures as a result; and
• the 2021 Post Enumeration Survey introduced further changes to the collection methodology, namely the use of the Address Register (rather than an area-based frame) as the survey frame and the widespread use of telephone interviewing (using an approach letter to elicit response). There were also improvements made to enumeration procedures as well as changes made due to COVID-19 lockdowns.

[1] See Glossary

While net undercount provides a picture of the net coverage error in the Census, as a single measure it can conceal the elements that help us understand the quality of Census coverage; for example, where high gross measures of overcount and undercount net out to a small net undercount. Gross coverage enables us to look at how well the population was captured without the added layers of people counted more than once or in error (overcount), and imputed people (where no Census form was received for a dwelling deemed occupied on Census night).

Figure 3.2.1 shows net undercount and gross coverage error. Gross coverage error is 100 minus gross coverage; for example, if gross coverage is 90%, the gross coverage error is 10%.

In 2016, the net undercount rate was 1.0% and the gross coverage error rate was 7.1%. In 2021, the net undercount rate dropped to 0.7% and the gross coverage error rate dropped to 6.1%, similar to 2011, indicating that better coverage was achieved compared to 2016.

Figure 3.2.2 shows the total net undercount rate for all states and territories for the past three censuses.

In 2021, as in previous censuses, the Northern Territory recorded the highest total net undercount rate of all states and territories (6.0%) up from 5.0% in 2016 but lower than 6.9% in 2011. The Australian Capital Territory recorded the lowest total net undercount rate (-0.6%, i.e. a net overcount). This is consistent with previous Post Enumeration Surveys, although this is the second Census in a row when the Australian Capital Territory has recorded a net overcount.

While the two territories continue to reflect the minimum and maximum total net undercount rates, trends across the states have shifted, with New South Wales, Victoria and Queensland experiencing decreases in net undercount in 2021 compared to 2016 and 2011. In contrast, South Australia, Western Australia and Tasmania reversed the trend experiencing increases in net undercount in 2021 compared to 2016. Western Australia’s net undercount in 2021 was higher than the previous two Censuses.

The components of overcount and undercount for the states and territories are shown separately in Table 3.2.3. The net undercount rates for persons on Census forms are similar for the six states but differ markedly for the Northern Territory (10.4%) and the Australian Capital Territory (1.6%). The largest improvements between 2016 and 2021 have been in New South Wales and Victoria, with net undercount rates (for persons on Census forms) both decreasing by 1.5 percentage points. The largest decline has been in the Northern Territory and Western Australia, with net undercount increasing by 1 percentage point and 0.8 percentage points respectively. In contrast, the net overcount for persons imputed is lower in all states and territories in 2021 when compared with 2016, except for the Northern Territory, which has remained stable. The Northern Territory has the highest net overcount rate for persons imputed, which is not surprising given the Census non-response rate was also the highest in the Northern Territory.

Comparing net undercount rates with gross coverage error for states and territories (Figure 3.2.3) shows that both net undercount and gross coverage error have decreased in New South Wales, Victoria, Queensland, and the Australian Capital Territory. While net undercount has increased in South Australia, the gross coverage error has remained stable, indicating good Census coverage with less over-imputation. In Western Australia and Tasmania both net undercount and gross coverage error have increased, while in Northern Territory the net undercount has increased but the gross coverage error has decreased indicating better coverage by the Census despite an increase in the net undercount.

Previous censuses (both in Australia and overseas) have observed that young adults are more likely to be missed, whereas older adults are more likely to be counted or overcounted in a census. Newborn children are often mistakenly omitted from census forms, and younger males are also traditionally more likely to be missed than females as they tend to be more mobile and more likely to be working away from home. As can be seen from Figure 3.2.4, these trends hold true for the 2021 Census. Relatively high total net undercount rates were observed for the 20-24, 25-29 and 30-34 year age groups (2.1%, 1.7% and 1.7%, respectively) and the highest total net undercount rate was observed for 10-14 year olds (2.3%). In contrast, older age groups had lower total net undercounts (usually overcounts), with the largest net overcount of 2.0% observed for people aged 85 and over, and 1.5% observed for people aged 75-79 years.

Males are more likely to be missed in the census compared to females. In 2021, the total net undercount rate for males was 1.3% while for females it was 0.2% and this is consistent with previous censuses. The total net undercount rates for males and females were 1.5% and 0.4% for 2016, respectively, and 2.2% and 1.2% for 2011, respectively. For males, those aged 30-34 years had the highest total net undercount rate of all age groups (3.6%) followed by 20-24 year olds (3.0%). For females, those aged between 10-14 years had the highest total net undercount at 2.2%. This was a similar pattern to the 2011 Census. Total net undercount for children aged 0-4 years more than halved in 2021 compared to 2016 (1.7% compared to 5.1%, respectively).

Figure 3.2.5 shows that net overcount for imputed persons is more evenly spread across nearly all age groups in 2021 compared with 2016, and this appears to be driven by changes in Census imputation methodology with younger adults more likely to be imputed in 2021 than in previous censuses. As in previous censuses, people aged 0-19 years have much lower rates of over-imputation, while in 2021 people aged 85 years and over had the highest. In 2016, there were consistently higher rates of over-imputation for people aged 55 years and over and this has reduced in 2021.

Figure 3.2.6 shows the net undercount for persons on Census forms by age. While total net undercount remains highest in the young adult age groups, particularly for 20-24 and 25-29 year olds, overall it is generally lower and more evenly spread across all age groups in 2021 than was the case in 2016.

Figure 3.2.7 illustrates the overall outcome: the total Census count (responses plus imputation) is lower than the Post Enumeration Survey population estimate for younger age groups and very close to the estimates for people aged 40 to 64 years. For people aged 65 years and over, the Census counts (including imputation) are slightly larger than the Post Enumeration Survey population estimates. There has been an improvement in the imputation approach used in 2021 with the counts including imputed persons being closer to the Post Enumeration Survey estimates than was the case in 2016, especially for people under the age of 40 years.

Note: Excludes Other Territories and overseas visitors.

The net undercount rates for country of birth in Figure 3.2.8 are displayed for the 10 highest ranked countries (in terms of population residing in Australia), according to the 2021 Census.

While there has always been some non-response to the Country of birth question, with increasing uptake of the online Census, non-response has decreased. For those who do not respond to the question, since Census does not impute a country of birth for these people, they are not included in Census counts for these categories. However, they contribute to net undercount estimates based on the category in which they should have been counted, as reported in the Post Enumeration Survey. There were 1,372,273 people (5.4%) whose country of birth was not stated in the 2021 Census or for whom the response was unable to be coded to a country, a decrease from 1,622,118 people (6.9%) in 2016.

Reflecting the overall decrease in net undercount seen across the population, there has been a decrease in net undercount compared to 2016 across nearly all of the top 10 countries of birth listed. The exception is New Zealand, with a net undercount rate of 8.5% in 2021, up from 5.5% in 2016. Persons born in South Africa have the second highest rate in 2021 (6.3%), followed by Australia (6.2%). Persons born in Italy had the lowest net undercount of the top 10 countries in 2021 at 1.8% and this was a significant decrease from 5.5% in 2016.

People who have come to Australia from other countries and whose first language is not English may find completing a Census form more difficult than other Australians. For several censuses, special strategies have been employed to promote an understanding of the Census among migrants and to provide assistance in a range of languages.

The Post Enumeration Survey shows that the Census population data, adjusted by the Post Enumeration Survey, is fit-for-purpose for its important use in rebasing the Estimated Resident Population at the national and state and/or territory level.

The total net undercount is small at the national level (0.7%) and for the states and territories with the exception of the Northern Territory and Western Australia. The Northern Territory continues to have the highest total net undercount.

Aboriginal and Torres Strait Islander peoples continue to be under-represented in the Census. The Post Enumeration Survey estimates the net undercount rate to be about 17% for each of the last three Censuses (for more see Section 3.5 Aboriginal and Torres Strait Islander peoples) despite increased investment in the enumeration of this population. COVID-19 lockdowns and border closures had an impact on the availability of field staff within states and interstate, and it is likely that these factors had an impact on the enumeration of Aboriginal and Torres Strait Islander peoples in some areas. The enumeration of Aboriginal and Torres Strait Islander peoples continues to be a priority area for attention by ABS. Despite this high net undercount, the Post Enumeration Survey enables the Estimated Resident Population estimates for this population to be rebased.

The net undercount rate for persons on Census forms (2.8%) has decreased compared with 2016 (3.7%), reflecting in particular a decrease in persons missing from Census forms (4% in 2021 compared to 4.9% in 2016), while persons counted more than once or in error has remained stable at 1.2%. The net overcount rate for persons imputed has decreased from 2.7% to 2.1%. This reduction in over-imputation is the result of improvements in the correct identification of occupancy in the Census, even though this was more difficult because of pandemic-based contact restrictions at the time of the Census. The Post Enumeration Survey also shows that while age skewing in the Census data has improved in 2021 compared to 2016, there are not enough persons aged 39 years and under in the Census counts.

The overall response rates for the 2021 Census are higher than for the 2016 Census, but lower than those of the 2011 Census. They are similar to response rates achieved in other censuses in Canada and the United Kingdom (both conducted during the pandemic in 2021), and higher than the response rate in New Zealand, when their Census was conducted in 2018 (pre-pandemic). This was an excellent outcome for Australia given the challenges of conducting a Census during a pandemic.

In the 2016 Census, too many non-responding dwellings were deemed as occupied on the Census night that were unoccupied. This was due to dwellings being deemed occupied where there wasn’t sufficient information in the field to determine occupancy. This resulted in some dampening of estimated response rates. Improvements in response rates in 2021 are largely due to an increase in the number of responding dwellings as a proportion of all dwellings which may in part be due to the pandemic causing more people to be counted at home. Improvements are also related to improved occupancy determination. Irrespective, it is an important quality improvement in the 2021 Census.

Person response rates can be viewed in two ways:

• where people were on Census night (their state of enumeration); and
• where people usually live (their state of usual residence).

In reviewing both indicators across each state and territory, the response rates achieved in 2021 were higher than those of 2016. They were also generally lower than those of 2011 with the exception of the Australian Capital Territory and Western Australia for place of enumeration, but they are comparable.

The person response rate (state of enumeration) for the 2021 Census was 95.8%, up from 94.8% in 2016 but lower than the rate in 2011 of 96.3% (see Figure 3.3.1). The highest person response rates were observed in the Australian Capital Territory (97%) and in South Australia (96.9%), while the lowest was for the Northern Territory (91.3%).

Across most of the states and territories, the response rates for occupied private dwellings reflected the pattern for Australia, namely higher in 2021 than in 2016 and lower than in 2011 (see Figure 3.3.3). Some of this is due to a higher level of response in 2021 and some may be due to the occupancy determination issue that was more prominent in 2016. However, Western Australia, the Northern Territory and the Australian Capital Territory are exceptions to this finding, with the rate achieved in 2021 higher than those achieved in both 2011 and 2016. Nevertheless, the rate for the Northern Territory continues to be lower than the other states and territories at 92.8% in 2021 and this reflects the difficulty in enumerating this part of Australia due to its remote nature, diversity of language groups, more transient population and a tighter labour market.

Approximately three per cent (2.9%) of the persons in the 2021 Census population counts were counted in non-private dwellings, down from about four per cent in 2016. The reduction reflects the impact of COVID-19 restrictions on movement in certain jurisdictions. Overall response rates were similar to 2016 (see Table 3.3.1). However, response rates in staff and nurses’ quarters show a marked decline, pointing to difficulty in enumerating settings including staff quarters in remote mining camps. The response rate for people in hospitals was also lower. This result is likely due to the difficulties in enumeration of hospitals during COVID-19 lockdowns.

As is the case in other countries, the 2021 Census relied on and strongly encouraged a digital first approach, while at the same time providing the option of a paper form for those unable or unwilling to complete online. Uptake of the online option by respondents has increased significantly over the past three censuses and has resulted in an associated increase in data quality, as online response is aided by: better sequencing through the form (skipping over questions that are irrelevant based on prior responses); the use of error messages to better guide respondents; and targeted supplementary questions to obtain more detailed responses from respondents.

The Australian Capital Territory had the highest online response of all the states and territories (93.2%), followed by Western Australia (82.4%) (see Table 3.3.3). Tasmania and South Australia had the lowest online response (61.3% and 71% respectively). The Northern Territory had the largest increase in the rate of online take up from 49.4% in 2016 to 74.9% in 2021. It should be noted that online uptake across areas is impacted by the mode of delivery and those states in which more paper forms were delivered had a lower online uptake.

As with overall non-response, total item non-response rates have declined since 2016 and are similar to 2011.

The main contributor to total item non-response is people who do not respond to the Census at all. While key variables (Age, Sex, Marital status and Usual residence) for a non-responding person are imputed (a process that attempts to fill in basic information for the people who didn’t respond to the Census), the remainder of the questions (data items) are set to either ‘not stated’ (and treated as ‘item non-response’) or not applicable, dependent on the imputed age of the person.

Figures 3.3.4 and 3.3.5 show item non-response rates for 2011-2021. The larger blue and orange bars represent the non-response rates which include imputed people (that is people who are not on a Census form). The shorter bars in the same graph show non-response rates for people who responded to the Census but did not answer a particular question. For those people who responded to the Census in 2021, item non-response is lower than in previous censuses for almost all items. Overall, the 2021 Census was more successful than previous censuses in getting people to complete more items as shown in Figure 3.3.4 and 3.3.5. Figure 3.3.6 shows item non-response for a range of dwelling variables.

Analysis of changes in person response rates at the Statistical Area Level 4 (SA4) level between 2016 and 2021 in Table 3.3.4 show that, in 39.8% of the SA4s, the improvement has been more than one percentage point (SA4s are the largest sub-state regions in the Australian Statistical Geography Standard). Most of the inner-city areas across the nation fall into this category. For a further 46.6% of SA4s, there has been an improvement of less than one percentage point. In 13.6% of cases, there was a decline in response. This group is dominated by the more remote areas. This is where field staff recruitment difficulties were most prevalent.

There have been increases in the person and dwelling response rates since 2016, nationally and for all states and territories. This is due to an increase in responding dwellings as a proportion of all dwellings and may also in part be due to improved occupancy determination. However, it should be noted that in some areas (mostly in remote areas), response rates have declined, and quality may be affected in these areas. Households completed the 2021 Census online at a much higher rate than previously, with about 79% of responding dwellings doing so and this will have had a positive impact on data quality. Item non-response rates were generally lower than in 2016, partly due to the lower number of imputed persons in 2021, and partly due to the high online form uptake. These will also lead to an improvement in data quality.

This section provides comparisons of the age distributions of the population recorded in the 2021 Census for Australia and for states and territories with the corresponding age distributions estimated in the unrebased Estimated Resident Population at 30 June 2021.[1] It should be noted that the reference periods for the Census and the Estimated Resident Population differ, with the Census reference period some six weeks after that of the population estimates. Census counts are not backdated in this analysis.

Comparing the age distributions from the two sources provides information about the quality of both sources. The more alignment there is between the two and an understanding of the differences between them provide an indication of coherence and quality.

Both the Census and the Estimated Resident Population are potential sources of any difference observed between the two measures. The ABS investigates these differences and using data from the Census and from the Post Enumeration Survey, adds in numbers of Australian usual residents who were overseas on Census night, then re-estimates the Estimated Resident Population. This is referred to as rebased Estimated Resident Population. This rebased 2021 Estimated Resident Population will be released on 28 June 2022 and will form the basis of all population estimates from 2021 until the next Census in 2026.

Error in the Estimated Resident Population can arise from quality and coherence issues associated with the administrative data sources used for quarterly estimates. These can include delays in the reporting and registering of births and deaths, the use of preliminary estimates of net overseas migration and the quality of information collected on Incoming Passenger cards, and lags, quality and undercoverage of people reporting changes to their usual residence within Australia.

While some delay in registrations is observed each year across all jurisdictions, the pandemic has had a noticeable effect in increasing the delay of birth registrations, most particularly in Victoria where, for the June 2021 quarter, more than half of the records were received after the reference quarter. These delayed registrations are not included in the published Estimated Resident Population for 30 June 2021.

The Census is particularly useful in recalibrating or rebasing the Estimated Resident Population for geographic areas within Australia every five years, as the Estimated Resident Population is affected by the accuracy of measurement of people changing their place of usual residence in Australia which is based largely on Medicare data. Evidence shows that the pandemic and subsequent lockdowns have significantly affected intra and interstate movements over the last two years. The success of the national COVID-19 vaccination program has had a positive impact on the quality of administrative data used to estimate Australia’s state and territory population as the rollout of the vaccination program has seen large numbers of Australians updating their address information. However, this happened towards the end of the inter-censal period so would have been of limited benefit to the Estimated Resident Population estimates for 30 June 2021.

The Census counts used in this section eliminate overseas visitors before the comparison with Estimated Resident Population is made. A challenge in this regard is that the Census definition of a usual resident is not precisely the same as the definition of a usual resident in the Estimated Resident Population calculation. In the Census, a person is counted as a usual resident if they usually live in Australia or if they usually live in another country but are resident in Australia for one year or more. A usual resident for the Estimated Resident Population is a person who is in Australia for 12 months over a 16-month period around the reference date. However, significantly reduced levels of overseas migration in the 18-month lead-up to the Census will have reduced the impact of this difference between the two datasets.

The comparisons shown below are based on the 2021 Census before any adjustment is made for the 2021 Post Enumeration Survey. This means, for example, that the people that the Post Enumeration Survey showed to have been incorrectly added (imputed) to non-responding dwellings (see Section 3.2, Table 3.2.2) are not removed from the Census counts before the comparisons with Estimated Resident Population are made. This approach is taken because the published Census data also include these incorrectly attributed people, although in 2021 it is known that there were fewer imputed people, and occupancy determination was more accurate.

Taking account of the difference outlined above, there is generally a good match between the 2021 Census and the 2016-based Estimated Resident Population for 30 June 2021 but the Census results for some of the smaller states and territories indicate there is room to improve the way administrative data is used between censuses to estimate the state or territory of residence and interstate movement for some recently arrived migrants.

[1] The unrebased Estimated Resident Population for 2021 is obtained by updating the 2016 rebased Estimated Resident Population using births, deaths, internal migration and international migration between 2016 and 2021. Unless otherwise specified, all references to Estimated Resident Population in this section refer to first released unrebased Estimated Resident Population.

For Australia (Figure 3.4.1), the peaks and troughs of the unrebased Estimated Resident Population age distribution are mirrored by the 2021 Census data. Note that the comparison of the two sources for Australia is up to 100 years of age, while for the states and territories it is truncated at 85 years because the numbers get much smaller after this age. ABS will make corrections at older ages as part of the recalculation of the Estimated Resident Population.

Numerically, the 2021 Census and the Estimated Resident Population are very close from age 40 onwards. Furthermore, for the under 40 years population, the 2021 Census counts are a closer fit to the Estimated Resident Population than was the case in 2016.

The largest differences between the two measures are in the childhood years (0-13 years of age) and the young adult years (20-35 years). At these ages, the Census results are lower than the Estimated Resident Population. This means that fewer people were counted in the 2021 Census in these age groups than would have been expected using the Estimated Resident Population. In the first age group, the largest differences were observed for children aged five years (4.7% lower in the Census) followed by babies (0 year olds), at 4.5% lower.

The Estimated Resident Population in the early childhood years is likely to be quite accurate because it is based on the number of births that occurred in the few years prior to the Census although, as noted earlier, an increased delay in birth registrations due to the pandemic is likely to have had an effect in Victoria. The omission of children at very young ages is a common feature of censuses not only in Australia, but also in other countries. However, in Australia there appears to be an undercount of children up to the age of 13 and this is more difficult to explain. Part of the explanation may lie in the different reference periods of the Census compared to the Estimated Resident Population, when some children would have had a birthday between the end of June and 10 August, causing an ‘age mismatch’ between the two data sources. An additional reason might be the movement of children between households as part of custodial arrangements where the children may not have been reported in the Census by either parent. The undercount of young adults (especially males) is also commonly observed in censuses because people of these ages are generally harder for censuses and surveys to contact. The Post Enumeration Survey confirms that the net undercount rate is relatively high for children aged 10-14 years, with the highest net undercount observed particularly for young men in their early 20s and early 30s.

The pattern of differences observed for Australia between the Census and the unrebased Estimated Resident Population is similar across the states and territories, as is the closer alignment of the 2021 Census counts to the Estimated Resident Population compared with 2016. There are generally lower counts of babies, young children, and young adults in the Census and a much closer fit for ages 40 and over between the Census and the Estimated Resident Population. However, there are noticeable differences across the states and territories in proportional terms (see Figures 3.4.3 to 3.4.18).

In New South Wales, there were 6.1% fewer babies in the Census count than in the Estimated Resident Population, while for 26-29 year olds the Census counts were between 5% and 6% lower. Conversely, in Victoria, there were 4.2% more babies in the Census count than in the Estimated Resident Population, and this reflects the delayed registration of births in Victoria due to COVID-19, which in turn has had an impact on the Estimated Resident Population for that age. The largest difference for Victoria was for young adults, with the Census counts for 25-27 year olds more than 8% lower.

South Australia’s Census counts were closely aligned with the population estimates, with 4.3% fewer babies. In South Australia, the count of people aged 19 and 20 years also showed differences with the estimates (3% lower than the population estimates). Census counts in the older ages (65-77 years) were somewhat higher than for the Estimated Resident Population.

In Queensland and Western Australia, there were larger difference in the count of babies, which were 8.9% and 8.7% lower than the Estimated Resident Population respectively, but there were relatively small differences in the counts of young adults compared with the estimates. For young adults in Queensland, the largest proportional difference between Census counts and the population estimates were for those aged 26-28 years where the counts were between 4% and 4.5% lower. In Western Australia, as in South Australia, the Census counts of people aged 19 and 20 years showed the largest difference in the young adult cohort with the counts about 3.5% lower.

In Tasmania and the Australian Capital Territory, there was an unusual pattern in that the Census counts for young adults were much higher than that of the Estimated Resident Population. In Tasmania, the Census counts for the 27 to 32 age group were at least 15% higher than that of the Estimated Resident Population, while in the Australian Capital Territory, the difference was even more pronounced with the Census counts for those aged 18, 19, 26, 27 and 28 years, in particular, more than 20% higher than the population estimates.

This unusual pattern could be due to a number of factors such as the potential movement of students back to their parent’s homes during COVID-19 lockdowns, but early analysis suggests it also reflects inaccuracies in capturing state or territory of residence and/or interstate migration of new migrants to these two jurisdictions. This unusual pattern will be investigated during the construction of the rebased Estimated Resident Population to be published on 28 June 2022. At this stage, it seems highly likely that the unrebased Estimated Resident Population has heavily underestimated the young adult population in Tasmania and the Australian Capital Territory. This provides an example of the value of a five-yearly Census in improving population estimates, especially at the sub-national level.

For the Northern Territory, the most striking variations are the relatively large differences for babies (where the Census count was 18% lower than the population estimate), and for children and young people up to the age of 21 years. As indicated by the Post Enumeration Survey, this is likely to be due to Census undercount, especially of Aboriginal and Torres Strait Islander peoples. The Estimated Resident Population for the Northern Territory is also noticeably higher than the Census count for those in their thirties.