Understanding income in administrative data, the Census and ABS surveys

Comparing income information across administrative data, the Census and ABS surveys

Released
27/04/2023

Administrative income data added to the Census

This article describes new administrative income data now available through Census TableBuilder Pro, and compares it with the current Census income question and the ABS Survey of Income and Housing. Features of this new data are highlighted in an article exploring New Census insights on income in Australia using administrative data.

The Census captures a person's income using a tick-box question with 16 income ranges to choose from. The question intends to cover all forms of income that the person may receive, including salary, government benefits, superannuation, and interest. For more information about this variable, including an example of the Census question, see INCP in the Census dictionary.

While the Census income question is very important, it only provides limited information. To enhance the income information available in the Census, we added new income variables to the Census dataset using linked administrative data.

What is administrative data?

  

Collecting income in future Censuses

The ABS reviews the questions we ask in the lead up to each Census. Public consultation for the 2026 Census started in early 2023. The quality and coverage of administrative income data may make it a potential candidate to replace or supplement the existing Census income question. For more information about this process see the 2026 Census topic review information on the ABS website.

We also welcome your input on using administrative data to enhance the Census in 2026 and beyond. You can contact the Census Admin Data Ventures team directly to provide feedback on the use of administrative data.

Census administrative income variables

There are 3 new income variables which are each available for the person, family, and household levels:

  • Total income (administrative data) (weekly)
  • Main source of income (administrative data)
  • Main type of government benefit payment (administrative data).

In total 11 variables have been created from the linked Census administrative dataset. Information on all variables is available in the Income and work section of the Census dictionary.

The Total income variables indicate the income that a person, family or household has received based on all income recorded in administrative data for the 2020–21 financial year. This is a similar concept to the existing Census income variables, but contains more detailed income ranges.

Comparing person income in the Census and administrative data
 Census income questionAdministrative data
Number of categories16 weekly income categoriesWeekly income in $10 increments (1,000 categories)
Highest income category$3,500 or more per week$10,000 or more per week

 

Main source of income and Main type of government benefit payment aren’t collected directly in the Census. The Main source of income variables indicate the main source of income that a person, family or household has received based on the income recorded in administrative data. Some examples of sources include wages and salary; government benefits and allowances; and investment income.

Similarly, the Main type of government benefit payment variables indicates the main type of government benefit payment that a person, family or household received based on all government benefits, pensions, and allowances recorded in administrative data. Some examples include the Age Pension; Newstart Allowance and Jobseeker; Family Tax Benefits; and Disability Support Pension.

How we calculate administrative income

Administrative income is calculated using linked data from the Multi-Agency Data Integration Project (MADIP) asset. MADIP is a secure data asset combining information on health, education, government payments, income and taxation, employment, and population demographics (including the Census) over time.

We have used administrative income information from the Australian Tax Office (ATO) and the Department of Social Services (DSS) datasets within MADIP, and these are linked to the Census via the Person Linkage Spine (Spine). This means that we have administrative income information available for every Census respondent who was able to be linked to the Spine.

More specifically, the ATO data includes personal Income Tax Returns and Payment Summaries for the 2020–21 financial year. DSS data includes each government payment administered by Centrelink that was made to people throughout the 2020–21 financial year.

For each person aged 15 years and over (who is not an Overseas Visitor), we calculate the total amount of income received in the 2020–21 financial year from each of the following sources:

  • Employee wages and salary (ATO)
  • Government benefits and allowances (DSS and ATO)
  • Own unincorporated business income (ATO)
  • Superannuation income (ATO)
  • Investment income (ATO)
  • Other income (such as child support payments or workers’ compensation) (ATO).

We use the total income from all these sources to determine Total personal income. We define the Main source of personal income as the income source with the largest amount for the financial year. We also calculate the total amount received for each type of government benefit payment. The Main type of government benefit payment is the one with the largest amount for the financial year. It is based on the amount of government benefit payment received, regardless of any incomes recorded in other sources. For example, a person whose Main source of income is an employee wage and salary, but also receives a small amount of Family Tax Benefit, will have Family Tax Benefit as their Main type of government benefit payment.

We calculate Total household income for each occupied private dwelling by adding up the Total personal income for each person aged 15 years or older in that dwelling on Census Night.

We calculate Total family income by adding up the Total personal income for each family member aged 15 years or older who was present in the household on Census Night.

Conceptual differences to Census income measures

Total personal income (weekly) as collected in the Census (INCP) aims to capture the total of all income a person usually receives. It is collected through a single question on the Census form which measures income in pre-defined ranges. Administrative income provides a measure of all income a person usually receives based on income recorded over the 2020–21 financial year in administrative data sources, which is divided by 52 to give an estimate of weekly income.

The following table outlines key conceptual differences between income information collected by Census and administrative income.

 Census income questionAdministrative income data
Purpose

Census data is collected by the ABS to provide a statistical snapshot of Australian people and dwellings on Census Night.

As well as providing counts, it collects a range of person and household characteristics including personal, family, and household income.

Administrative income data is collected by ATO and DSS as part of their government administrative functions.

For this project, ATO and DSS income data is sourced via MADIP, a secure data asset combining information on health, education, government payments, income and taxation, employment, and population demographics (including the Census) over time. 

FrequencyCollected once every five years.

ATO: Income Tax Return and Payment Summary data available annually by financial year.

DSS: Government benefits, pensions, and allowances data includes transaction-level information. The data are supplied to MADIP on a monthly basis.

Reference periodUsual income, as reported on Census Night. If the person’s income was affected by COVID lockdown, guidance was given to report the pre-COVID income.All income recorded in the 2020–21 financial year.
Collection methodology

Collected using a self-completed online or paper form. The form may be completed by one household member on behalf of others, or each person may complete their own questions.

Collected from ATO Income Tax Returns and Payment Summaries, and from DSS data on government benefits, pensions, and allowances.

These datasets are linked to the Census using the Spine

Content

Total income collected as a categorical variable using broad income ranges.

Income for each person is determined by a single question asking the total of all income a person usually receives. 

Household and family income is derived using a median-based approximation from ranged responses to the personal income question.

Actual dollar income (in $10 ranges) calculated by adding a number of different ATO and DSS income items. 

Income for each person is determined by adding amounts a person receives as recorded on an Income Tax Return or Payment Summary, and/or from government benefits, pensions, and allowances from DSS. 

Household and family income is derived from personal income dollar amounts.

Coverage / non-response error

Income is ‘not stated’ for people who:

  • did not complete the Census
  • completed the Census but did not answer the income question.

Income is not available at all for Census respondents who were unable to be linked to the administrative data. This may be due to not having enough information to link the Census record to a person in administrative data, or the person does not exist in the administrative data.

Where a person was able to be linked to administrative data but no income information was recorded, their income was deemed to be zero.

Measurement errorRequires the household reference person to correctly recall the total income a person usually receives. 

Dollar amounts of income are used as recorded in administrative data and should generally reflect all income earned.

Some types of income may not be recorded in administrative data. These include:

  • cash in hand income not reported to the ATO
  • superannuation income which is not currently captured in administrative data (unless included on an Income Tax Return)
  • other income not reported to the ATO.

 

Availability of income information in administrative data

Availability refers to the percentage of persons aged 15 years and over (who were not Overseas Visitors) who we were able to link to the Spine. Not everyone is able to be linked; not all Census responses contain enough information to enable linking, and some Census respondents may not have a corresponding record in administrative data. For further information on linkage of administrative data see Data Integration. Respondents who we cannot link to administrative data are given the value of ‘admin not available’.

Where a Census respondent was able to be linked to the Spine, but no administrative income information was recorded, their income was deemed available, and was calculated as Nil income. For example, people who did not work or receive government benefit payments in the 2020–21 financial year would not have any administrative income recorded and hence did not receive any income.

Overall, we were able to derive income from administrative data for 89.6% of persons in the Census aged 15 years and over (who were not Overseas Visitors). This was 3.2 percentage points lower than the proportion of people who responded to the income question on the Census (92.8%).

Looking at availability in States and Territories, the Australian Capital Territory had the highest rate of income availability in administrative data (92.2%) and Northern Territory the lowest (75.9%). These territories also had the highest and lowest rate of response to the income question on the Census. NT had the biggest gap between the administrative data rate and the Census response rate (12.3%).

There is negative correlation between administrative income data availability and Remoteness Area. This explains the relatively lower rate in the NT, which has a relatively larger population in more remote areas. Major Cities had the highest rate of income availability in administrative data (91.0%) and Very Remote Australia the lowest (64.8%). The gap between administrative and Census income data availability increased with increasing remoteness. In Very Remote Australia, the availability of administrative income data is 17.3 percentage points lower than Census responses.

The availability of administrative income information is greater than 87% for all age groups between 15 and 84 years. There is slightly lower availability for 20 to 34 years of age than other working ages, and lower availability in administrative data than the Census question. Availability decreases in older age groups; however, the availability of Census income responses decreases at a greater rate. There is more availability of administrative income information than Census income responses in people aged 85 years and older (81.8% in administrative data compared with 80.7% in Census).

Total income, confrontation with Census income

Comparing counts in broad income ranges

Administrative income generally aligns well with Census income at the national level. The national median personal weekly income from administrative income was very similar to Census income ($789 compared with $805). Comparing the number of people in each Census income range, the administrative income has more people with income in the ranges $1-$149, $400-$499, and $2,000-$2,999 than the Census reported income. Income frequencies align best for people receiving nil income, and for those with weekly income of $3,000 or more.

Similar trends were evident for households, although there was a greater proportion of households with no income in administrative data compared to Census. Trends for families were very similar to households.

Comparing administrative and Census reported income at a person and household level

Because administrative and Census reported income are conceptually different and subject to different measurement errors, we expect significant differences at the person and household level. Despite this we see generally good alignment between administrative and Census reported income at a person level. Nearly 2 in 3 people had their administrative income within one category of Census reported income (64.9%). For example, if a person reported that their total income was $500-$649 per week on the Census, then their administrative income was generally between $400 to $799 per week. One in 7 people had an administrative income which was 2 categories different compared to Census personal income (14.4%). One in 5 people had an administrative income which was 3 or more categories different compared to Census personal income (20.6%). Alignment was better for people and households with larger incomes. This may be influenced by the Census income category ranges being smaller for smaller incomes, and larger for larger incomes.

The alignment between administrative household and Census household incomes was similar, but slightly weaker. Three in 5 households had administrative income within one category of Census household income (59.7%), and one in 6 had income which was 2 categories different (16.6%).

Main source of income, comparison with survey measures

Main source of income refers to the largest source of income a person, household or family received. Census does not collect information on main source of income, but it is collected and published at a household level in the publication Household Income and Wealth, Australia, which presents key information from the Survey of Income and Housing 2019–20 (SIH).

There was close alignment in the largest categories between the administrative data and SIH. 62.8% of households had their main source of income from employee wages and salary, compared with 62.0% in SIH. 23.3% of households had their main source of income from government benefit payments in administrative data, compared with 22.4% in SIH.

One gap evident between administrative data and SIH is in the availability of superannuation income data, as many superannuation payments are not required to be reported annually to the ATO. Therefore, the administrative income data understates the proportion of households whose main source of income is superannuation (1.1% in administrative data compared to 6.0% in SIH). This means that we have under-recorded the amount of income received from superannuation in all of the Total income variables. The main source of income variables will also be missing superannuation income data. This known undercount of people with superannuation as their main source of personal income results in a corresponding overcount of the number of people with other income types as their main source of personal income.

Additional information

This article describes new administrative income data now available through Census TableBuilder Pro.  It also compares it with the current Census income question and with income from the ABS Survey of Income and Housing.  Features of this new data will be highlighted in an upcoming article exploring new Census insights on income in Australia using administrative data.

The Census captures a person's income using a tick-box question with fourteen income ranges to choose from.  The question intends to cover all forms of income that the person may receive, including salary, government benefits, superannuation, and interest.  For more information about this variable, including an example of the Census question, see INCP in the Census dictionary.

While the Census income question is very important, it only provides limited information.  To enhance the income information available in the Census, we added new income variables to the Census dataset using linked administrative data.

Back to top of the page