Managing Census quality

The ABS is committed to helping users understand all aspects of data quality so they can assess the usefulness of data for their needs. To achieve high quality data from the Census, extensive effort is put into addressing sources of error through quality control measures across Census processes and products, including how the ABS informs users about Census data quality.

For details on how the Census addresses each of the elements of the ABS’s quality framework, see the 2021 Census Quality Declaration. 

There are four principle sources of error in Census data: respondent error, processing error, non-response and undercount. Quality management of the Census program aims to reduce error as much as possible and to provide a measure of the remaining error to data users to allow them to use the data in an informed way.

The Australian Census is self-enumerated. This means that people are required to complete the Census form themselves, rather than having the help of a trained interviewer. Self-enumeration carries the risk that wrong answers could be given, either intentionally or unintentionally. Error can be introduced if the respondent does not understand the question or does not know the correct information about other household members. The ABS has a number of ways to minimise respondent error.

Choosing suitable content

Census topics have simple questions and the total number of questions is managed to ensure reliable responses. Some topics, such as sex, age, registered marital status, indigenous status, usual residence and internal migration, are included in the Census to meet legislative requirements to provide information for electoral purposes and the distribution of government funds. Other topics are selected for inclusion in the Census following extensive community consultation. These topics are selected for their national importance, need for data at a small population or area level, if the data is not available elsewhere or there is an ongoing need for data on this topic.

Question and form design

The Census form is designed so that questions are easily understood and simple for respondents to answer. Questions are tested through focus groups and cognitive interviews to ensure they are clear, well worded and can be answered on behalf of others. Field tests with these questions are conducted in various cities and rural locations and specific tests are also conducted on the usability and functionality of the online form.

Raising public awareness

To achieve high quality Census data, it is essential that people understand the importance of being counted and of giving the right answers on the Census form. Raising public awareness through advertising and community briefings contributes to high levels of participation in the Census. It also draws attention to the assistance that is available for people who may have problems filling out their Census form.

For more information on planning and running a Census see Planning the 2021 Census.

Most Census data is recorded through the online Census form. Data on paper Census forms is recorded using automatic processes, such as scanning, Intelligent Character Recognition and other automatic processes. Quality assurance procedures are used during Census processing to ensure processing errors are kept at an acceptable level. Sample checking is undertaken during coding operations, and corrections are made where necessary.

Repair for paper forms

Once paper forms are received they are checked for damage and errors, such as tears, multi-mark responses and illegible handwriting. Where required, these problems are fixed manually to assist the automatic coding processes.

Coding errors

Most responses are coded automatically using official classifications with legal value checks built into the system. Errors are more likely to arise during automatic coding of written text answers. Clerical staff resolve problems that arise if text responses cannot be automatically matched to the index of possible responses. Their work is subject to a quality management process where a 10% random sample of coded records are re-coded by a different coder and differences are evaluated by specialist coders. This ensures that systematic errors are detected and remediated.

Validation

The completed data is put through a series of automated checks to ensure internal consistency. The data is also scrutinised for changes over time by comparison with previous Census data and other data sources, and across categories, where expected trends can be identified and unexpected trends investigated.

In preparing Census data for output, various derivations and re-codes are applied to the data to produce the variables listed in the 2021 Census dictionary. Data is processed further to create the range of Census data tools and products. A series of checks occur at each stage of the output process to ensure data consistency and accuracy.

The Census does not receive Census forms from a small proportion of households. The Census aims for a response rate from private dwellings greater than 95% to minimise the risk of non-response bias in Census data.

Additionally, when completing Census forms some people do not answer all the questions which apply to them. While questions of a sensitive nature are generally excluded from the Census, all topics have a level of non-response. This level can be measured and is generally low.

In instances where a household fails to answer a question, a 'not stated' code is allocated during processing, with the exception of non-response to age, sex, marital status, place of usual residence and place of work. These variables are imputed for non-responding dwellings based on a similar responding dwelling. See Imputation for more information.

The goal of the Census is to obtain a complete measure of the number and characteristics of people in Australia on Census Night and the dwellings in which they live. However it is inevitable that a small number of people will be missed and some will be counted more than once. In Australia more people are missed from the Census than are counted more than once. The net effect is an undercount.

Field procedures

Every effort is made to ensure that all households receive a Census form and that these are completed and returned. Information is mailed to households in most areas of Australia on how to complete the Census online, with paper forms mailed or delivered to other areas. Returned Census forms are matched back to the Census frame (list of units, e.g. persons, households, businesses) to register that the household has completed the Census. All forms are registered to the dwelling they were delivered to so that data processing staff can account for forms received as well as those still to be returned by mail or online.

Those households who do not respond receive reminder letters and visits by Census field officers. Ensuring all dwellings are contacted and all persons have provided a response is a critical measure of the completeness of the Census.

Some groups of people in the population are at greater risk of being undercounted in the Census than others. Targeted enumeration strategies have been developed to ensure a more complete count of these groups, dwellings and areas. Strategies are also in place to ensure accessibility to Census forms through the most appropriate means for people with disabilities.

Post Enumeration Survey

The Census Post Enumeration Survey (PES) is run shortly after the Census to independently assess the completeness of the Census count. The survey is collected from a sample of households and asks basic demographic details as well as location on Census Night. This information is used to determine if and where a person should have been counted in the Census. It also supports the matching of PES person records to Census records to establish the number of times they were actually counted. Results from the PES can help evaluate the effectiveness of Census collection procedures and data processing so improvements can be made for future censuses.

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 PES 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. Census PES results are discussed in more detail in 2021 Census overcount and undercount.