PRODUCT OVERVIEW

The Australian Census Longitudinal Dataset (ACLD) is a 5% random sample of persons enumerated in Australia on each Census night linked together using statistical techniques.

THE 2006-11-16 THREE WAVE ACLD

The 2006-11-16 three wave ACLD is a representative sample of almost one million records from the 2006 Census (Wave 1) brought together with corresponding records from the 2011 Census (Wave 2) and the 2016 Census (Wave 3).

Record linking and panel sample

The 2006 Panel sample of records was originally linked to the 2011 Census and released in 2013. In the three wave ACLD, the 2006 Panel has been re-linked to the 2011 Census to take advantage of improved linking methodology since the initial release. At the completion of this linkage process, 756,945 (77%) of the 979,662 records from the 2006 ACLD Panel sample were linked to a 2011 Census record to create the linked 2006-2011 ACLD with an estimated precision of approximately 95%, or a false link rate of approximately 5%.

These record pairs were then linked to the 2016 Census via the 2011 Census record in each pair. Any 2006 Panel record which had not been successfully linked to a 2011 Census record was not given the opportunity to link to the 2016 Census. This achieved 605,618 links (80% of the 2011 records in the 2006 Panel) at an estimated 98.6% precision for a direct 2011-2016 linkage. 62% of links from the original 2006 Panel sample linked to both the 2011 and 2016 Censuses.

There are two main reasons why some records were not linked across Census files:

• Records belonging to the same individual were present at both time points but these records failed to be linked because they contained missing or inconsistent information.
• The person had no record in the later Census. This can be because they were no longer in scope for the Census due to migration from Australia, or if there was a death between 2006 and 2016, or they may simply have been missed in the Census.
  

USING THE THREE WAVE ACLD FOR ABORIGINAL AND TORRES STRAIT ISLANDER POPULATION RESEARCH

There were 21,987 records identified as Aboriginal and/or Torres Strait Islander in the 2006 Panel sample, of which, 9,798 (44.6%) were linked across 2006, 2011 and 2016. Linkage rates were lowest for Aboriginal and Torres Strait Islander people living in the Northern Territory (33.1%) and Very Remote areas (35.3%) in 2006. These lower linkage rates are related to generally higher item non-response in the Census, and more variation across age, date of birth, address and name information for the Aboriginal and Torres Strait Islander population (with variation in name information applying to 2011-2016 linkage only).

In data linkage projects, geographic boundaries function as blocking variables that restrict the search for links to records which agree on the defined geography. They are also used as linking variables, and when combined with other linking fields (such as hashed name [2011-2016 only], age, sex and date of birth), they provide a high level of uniqueness, and reduce the likelihood of linking to an incorrect record. However, not every dwelling enumerated in the Census has a unique address, particularly dwellings in Very Remote areas which may only have a name for the community. This means other blocking variables may be used to link records in the ACLD with consequential impacts on linkage rates for these records.

While the ACLD is weighted to be representative of the Aboriginal and Torres Strait Islander population, the potential impact of missing links should be considered relative to the nature of the analysis being undertaken. Further, as explored in this article, a person’s Indigenous status is not always consistent across time. Consequently, a person who did not identify as Aboriginal and/or Torres Strait Islander in 2006, but did in 2011 or 2016, will have a non-Indigenous or Indigenous status not stated record in the 2006 panel.

Weighting, benchmarking and estimation

Once a record was linked in the three wave ACLD, it was weighted to the population benchmark based on the 2011 and 2016 Estimated Resident Population (ERP), which is adjusted by the estimated probability a person belongs to the longitudinal population. This probability is formed using the Census reported address five year ago variable from the 2011 or 2016 Census.

The 2006-11-16 weights range between 15.9 and 341.3. The mean weight was higher for Aboriginal and Torres Strait Islander people (53.5) than non-Indigenous people (30.0). For both Aboriginal and Torres Strait Islander people and non-Indigenous people, the mean weight was higher for:

• Males
• People living in the Northern Territory
• People living in Very Remote Australia
• Young people aged 20-24 years

Where can I find more detail about linkage rates and weighting in the three wave ACLD?

Information about data linkage methodology used in the three wave ACLD and linkage results can be found in Information Paper: Australian Census Longitudinal Dataset, Methodology and Quality Assessment, 2006-2016 (ABS cat. no. 2080.5). Additional linkage results for the three wave ACLD are in Table 10 in the data cube titled 'ACLD 2006-2011-2016 Identification as an Aboriginal and Torres Strait Islander person in the Census over time' available in the downloads tab.

Information about weighting can be found in the Explanatory Notes section of Information Paper: Australian Census Longitudinal Dataset, Methodology and Quality Assessment, 2006-2016 (ABS cat. no. 2080.5).


SOURCES OF ERROR IN THE ACLD

All reasonable attempts have been taken to ensure the accuracy of the longitudinal dataset. Nevertheless other potential sources of error, in addition to missed links, including sampling, linking and Census quality error should be kept in mind when interpreting the results.

The key sources of error are:

• Sampling Error
• Linking Accuracy
• Census data quality (includes respondent error, processing error, partial response and undercount)

Further information about these sources of error is available in Microdata: Australian Census Longitudinal Dataset, ACLD (ABS cat. no. 2080.0).

Why is it important to consider sources of error when looking at changes in Indigenous status over time?

As discussed in this article and others^1,2,3, Indigenous status is a self-reported measure collected through the Standard Indigenous Question (SIQ) – part of the ABS Indigenous Status Standard. This means a person’s Indigenous status can change for any number of reasons and it is not unexpected to observe changes in a person’s Indigenous status in a longitudinal dataset such as the ACLD.

In the Census, the SIQ can be responded to by each person on the Census form or by any responsible adult who fills in the Census form on behalf of all household members (for example a parent may respond to the Census on behalf of their child). A person’s Indigenous status in the Census can therefore reflect how they choose to identify themselves, or how a member of their household has identified them. Also, the mode of collection differs in Aboriginal and Torres Strait Islander communities and some remote areas, where tailored procedures are used to conduct the Census. In these locations, an interviewer household form is used and respondents complete the Census with a trained interviewer.

It is also possible that some of the changes in a person’s Indigenous status in the three wave ACLD are a result of error (sampling error, linking accuracy and Census data quality). While the three wave ACLD has high linkage precision rates and two-thirds (66%) of people who have ever identified as Aboriginal and Torres Strait Islander in the three wave ACLD are consistently identified, it is important to consider the possibility of error when interpreting the data.


COMPARISONS TO THE 2006-2011 AND 2011-2016 ACLD AND THE CENSUS

The three wave ACLD contains data for records that were able to be linked across 2006, 2011 and 2016. This means data extracted from the three wave ACLD will be different to data extracted from the 2006-2011 and 2011-2016 ACLDs.

The three wave ACLD uses an enhanced linkage methodology that was also used in the 2011-2016 ACLD⁴. This enhanced methodology, combined with the cumulative effect of linkage rates, sources of error and variations in the Census overcount and undercount can contribute to differences when comparing data extracted from the 2006-2011, 2011-2016 and the 2006-2011-2016 ACLD. Users should therefore exercise caution in drawing comparisons between all available ACLD waves.

The cumulative effect of linkage rates, sources of error and variations in the Census overcount and undercount and the longitudinal weighting process of the ACLD also means that the ACLD is not strictly comparable with the Census. The weighting process for the three wave ACLD factors in the probability of a record being selected in the 2011 Panel sample, the probability of a record being linked, and estimates of Census overcount and undercount, whilst the Census does not. This will result in differences in estimates and proportions extracted from the ACLD to Census counts. Users should therefore exercise caution in drawing comparisons between the ACLD and the Census.

For further information on the above, see Information Paper: Australian Census Longitudinal Dataset, Methodology and Quality Assessment, 2006-2016 (ABS cat. no. 2080.5).

Where can I find further information about the impact of changes in Indigenous status on statistical datasets?

Endnote 1 suggests some resources that discuss the impact of changing Indigenous status on statistical datasets and the flow on effect to interpretation of data for Aboriginal and Torres Strait Islander people where Indigenous status is a used as a data variable. Most of these resources relate to administrative datasets where changing Indigenous status can affect the quality and reliability of analysis.


REFERENCES

¹ ABS 2014, Information Paper: Review of the Indigenous Status Standard, 2014, cat. no. 4733.0, ABS, Canberra.
² ABS 2013, Information Paper: Perspectives on Aboriginal and Torres Strait Islander Identification in Selected Data Collection Contexts, 2012, cat. no. 4726.0, ABS, Canberra.
³ Australian Institute of Health and Welfare & ABS 2012, National best practice guidelines for data linkage activities relating to Aboriginal and Torres Strait Islander people, AIHW cat. no. IHW 74. Canberra: AIHW.
⁴ Chipperfield, J, Brown J & Watson N 2017, The Australian Census Longitudinal Dataset: using record linkage to create a longitudinal sample from a series of cross-sections, Australian & New Zealand Journal of Statistics, 59(1), pp. 1-16.


ENDNOTE 1

• Biddle, N & Crawford, H 2015, The changing Aboriginal and Torres Strait Islander population: Evidence from the 2006–11 Australian Census Longitudinal Dataset, Centre for Aboriginal Economic Policy Research, Australian National University, Canberra.
• Biddle, N & Markham, F 2018, Indigenous identification change between 2011 and 2016: evidence from the Australian Census Longitudinal Dataset, Centre for Aboriginal Economic Policy Research, Australian National University, Canberra.
• Christensen, D, Davis, G, Draper, G, Mitrou, F, McKeown, S, Lawrence, D, McAully, D, Pearson, G, Rikkers, W, & Zubrick, S R 2016, Evidence for the use of an algorithm in resolving inconsistent and missing Indigenous status in administrative data collections, Australian Journal of Social Issues, 49:14.
• Lawrence D, Christensen D, Mitrou F, Draper G, Davis G, McKeown S, McAully D, Pearson G & Zubrick S R 2012, Adjusting for under-identification of Aboriginal and/or Torres Strait Islander births in time series produced from birth records: Using record linkage of survey data and administrative data sources, BMC Medical Research Methodology, 12:90.
• Lovett, R 2013, The Lowitja Institute Data Linkage Roundtable: Background Document, The Lowitja Institute, available from: https://www.lowitja.org.au/page/research/research-roundtable/data-linkage
• Thompson, S C, Woods, J. A & Katzenellenbogen , J M 2012, The quality of Indigenous identification in administrative health data in Australia: insights from studies using data linkage, BMC Medical Informatics and Decision Making, 12:133.