Review of 2021 Census Topics

The new National Housing and Homelessness Agreement will commence on 1 July 2018. A key output of the agreement is the development of a Housing and Homelessness Data Improvement Plan in 2018–19. The Housing and Homelessness Data Working Group will develop this plan. We expect the plan to include improvements to a range of data sources underpinning the national Agreement including (potentially) Census housing data in the following areas.
1. Landlord type
Government monitoring of the performance of the housing system would benefit from better alignment of concepts between the Census and social housing administrative data. For example, the Census enumerates fewer households in social housing than the administrative public housing, state-owned and managed Indigenous housing, community housing and Indigenous community housing data collections held by the Australian Institute of Health and Welfare (AIHW) (see Appendix B of Housing assistance for Indigenous Australians (2014) for more details on this issue). These data are provided to the AIHW by state and territory housing authorities for national reporting purposes.
We think there are opportunities to better share these administrative data that the AIHW collects with the ABS.
2. Household income
Currently, household income is derived by the ABS from reported categorical personal income data (Census question 33). This level of precision does not support analysis of housing affordability such as calculating the proportion of income spent on (or remaining after) housing costs. This is problematic as the Census is the only independent, national data source that provides whole of population data that could be used to monitor housing affordability across small geographic areas.
There are alternative methods that could potentially improve the reliability of the household income data.

The new National Housing and Homelessness Agreement will commence on 1 July 2018. A key output of the agreement is the development of a Housing and Homelessness Data Improvement Plan in 2018–19. The Housing and Homelessness Data Working Group will develop this plan. We expect the plan to include improvements to a range of data sources underpinning the national Agreement including (potentially) Census housing data in the following areas.
1. Landlord type
One aspirational, overarching national outcome of the National Housing and Homelessness Agreement is ‘a well-functioning social housing system that operates efficiently, sustainably and is effective in assisting low income households and priority homelessness cohorts to manage their needs’.
The Census enumerates fewer households in social housing than the administrative public housing, state-owned and managed Indigenous housing, community housing and Indigenous community housing data collections held by the AIHW (see Appendix B of Housing assistance for Indigenous Australians (2014) for more details on this issue). These data are provided to the AIHW by state and territory housing authorities for national reporting purposes.
We think there are opportunities to better share these administrative data that the AIHW collects with the ABS (see response to Assessment criteria 3).
Government monitoring of the performance of the housing system would benefit from better alignment of concepts between the Census and social housing administrative data. For example, the landlord type categories captured in the Census (see excerpt below) cannot be accurately mapped to the types of housing funded within community housing programs and regulated by the National Regulatory System for Community Housing (NRSCH). It is unclear, for example:
• which Census landlord category should be reported by tenants in housing provided by local councils and Indigenous councils
• how people would respond to the Census question if they lived in a caravan park dwelling managed by state/territory government or by a Community Housing Organisation.
This results in inability to produce data describing community housing tenants that is consistent with administrative data held by funding departments and the NRSCH. For example, governments need to separately identify housing provided by government entities, as this housing may be funded under state/territory community housing programs (thus is of particular interest to community housing funding departments) but is not regulated under the NRSCH (and thus should be excluded from comparisons with data sourced from the NRSCH).
In 2016, Landlord type was captured via question 57 of the Census:

The Census:
• captures rich data not well suited to administrative data collection/not included in nationally-agreed data set specifications,
• has the number of respondents required to produce reliable statistics for smaller priority cohorts, and
• enables comparison of housing outcomes across housing tenure and landlord type.
This misalignment of concepts constrains governments’ ability to use Census data to:
• understand the profiles and outcomes of households in social housing and compare the housing outcomes across housing tenure and landlord types, and
• manage gaps in the coverage of administrative community housing data collections.
The Commonwealth Grants Commission has advised that, for the 2020 review of GST distribution methodology, it is looking to develop an approach which scales Census data with administrative data held by the AIHW in order to adjust for the Census undercount of households in social housing.

2. Household income
One aspirational, overarching national outcome of the National Housing and Homelessness Agreement is ‘affordable housing options for people on low to moderate incomes’.
Currently, household income is derived by the ABS from reported categorical personal income data (see excerpt below). This level of precision does not support analysis of housing affordability such as the proportion of income spent on (or remaining after) housing costs. This is problematic as the Census is the only independent, national data source that provides whole of population data that could be used to monitor housing affordability across small geographic areas e.g. local government areas. Low-level geographic analysis of housing affordability is desired by governments as housing affordability varies by location.
In 2016, household income was captured via question 33 of the Census:

There are alternative methods that could potentially improve the reliability of the household income data (see response to Assessment criteria 3).

See response to Assessment criteria 1.

We are proposing improvements in the data that is already being collected/generated by the ABS.
Landlord type
Improvements to the alignment of concepts between the Census and social housing administrative data could potentially be achieved though initiatives such as:
1. Using administrative data to create new administrative social housing flags. Dwelling address data are held by the state/territory housing authorities that deliver/fund social housing as well as community housing regulators. This initiative would produce reliable data as it does not rely on self-reported data. The AIHW could support the ABS to work through privacy-related issues to facilitate state/territory provision of address data to the ABS, as we have successfully managed these issues to deliver the National Social Housing Survey.
2. If the first initiative above is not achievable, we recommend:
a. The response options for landlord type are improved to separately identify tenants in public housing, state-owned and managed Indigenous housing, community housing and Indigenous community housing, and avoid response errors resulting from respondent misunderstanding of how to classify their landlord.
b. Providing more examples of how different landlord types should be coded e.g. housing provided by local councils or land councils.
c. Moving the dwelling-related section to earlier in the Census form. This initiative may help to manage errors associated with respondent fatigue.
There are a number of advantages of implementing the first initiative described above. For example:
• Any administrative flag for social housing dwellings developed for the Census could potentially be rolled out in other ABS surveys to improve the identification of households in social housing. Data from a range of ABS surveys contribute to government monitoring and reporting for the housing system e.g. the Survey of Income and Housing, the General Social Survey, the National Aboriginal and Torres Strait Islander Social Survey and the Australian Aboriginal and Torres Strait Islander Health Survey. These surveys, in particular, would benefit from improved identification of households in social housing.
• Supplementary flags could be developed to identify dwellings that have been transferred from the public housing system to the community housing system and the date in which this occurred. This would facilitate reporting of housing outcomes following transfers of housing stock/tenancy management.
While the introduction of administrative social housing flags would improve the reliability of social housing data captured via the Census, there are some data quality issues to consider. Counts of households in social housing would never exactly match across the Census and administrative data sets e.g. due to incomplete Census coverage of households in Australia. Also, currently, state/territory housing authorities do not have complete and valid addresses for all funded social housing dwellings (especially community housing and remote Indigenous housing).
Household income
Currently, household income is derived by the ABS from reported categorical personal income data (Census question 33). This level of precision does not support analysis of housing affordability such as the proportion of income spent on (or remaining after) housing costs. There are alternative methods that could potentially improve the reliability of the household income data. For example, the ABS could potentially use an alternative method, the ABS Personal Income Tax dataset (PIT), to inform small area household estimates, or add an addional question, asking respondents to self-report their household income (such as collected via the Household Income and Labour Dynamics in Australia Survey).
There are some considerations associated with the use of PIT data. In particular:
• Data are provided by the Australian Taxation Office (ATO) before the processing of all income tax returns for any given year may have been completed. Data provided to the ABS by the ATO are from returns processed up to 31 October, 16 months after the end of the financial year. It is estimated that approximately 3% of taxpayers lodge their income tax returns in the twelve months after the initial processing cut off each year and approximately 5% of taxpayers lodge their income tax returns in the three years following the processing cut off. This means that data provided to the ABS slightly under-estimates the ‘final’ number of income earners and the total income earned.
• The different reporting period to other Census data.
• There would be some missing records for people who are not required to lodge tax forms e.g. people receiving pensions, with low incomes, or receiving tax-exempt superannuation income.
• Definition work would need to be undertaken to assign individuals to Census households.
• Linkage would not be possible based on dwelling address for accountant lodgements or tax records containing a PO Box rather than a place of residence.
• Some households may be under-represented due to poor linkage rates e.g. residents of the Northern Territory, persons aged between 20 and 29 years of age, and persons of Aboriginal and/or Torres Strait Islander descent.

We are proposing improvements to data already captured in the Census.

We are proposing a number of options for improving data already captured in the Census. These initiatives are outlined in the response to Assessment criteria 3. Costs will reflect the initiative(s) chosen for implementation.

The National Housing and Homelessness Agreement is ongoing. As a result, there is a need for time series data relating to Landlord type and Household income, and the topic is likely to remain relevant in the future

See response to Assessment criteria 1.

Thank you for considering this submission. We would appreciate being involved in any further data development work in this area. The AIHW can support engagement with the (yet to be convened) Housing and Homelessness Data Working Group to discuss data development work.
Landlord type
We would also appreciate receiving any results from cognitive or pilot tests relating to Landlord type. This information will help us to better understand Census outputs.
The Commonwealth Grants Commission has advised that it supports this proposal.
We understand that the Department of Social Services has also raised the issue of identification of households in social housing in their housing submission.