This document was added or updated on 27/07/2020.

USING MICRODATA RESPONSIBLY

This section provides information about using ABS microdata appropriately, including:
Using microdata for a statistical purpose
When can microdata be shared?
Sharing statistical output from microdata
Collaborating on microdata research
Using more than one microdata file together
Aggregate data and microdata

Using microdata for a statistical or research purpose

ABS legislation requires that microdata is used for statistical or research purposes. You must not use microdata for compliance or regulatory purposes. Examples of statistical or research purposes include:

• estimation of population characteristics to provide indicators of financial literacy of Australian households
• statistical modelling of predictors and correlates of employment and work performance among those with carer responsibilities compared to persons without these responsibilities
• use of data as input to mathematical models to enable analysis of impacts of potential policy changes in the area of health insurance
• study the relationship between disability and labour market outcomes to target future employment programs
• provision of statistical training to staff or students

The ABS reserves the right to seek additional information from applicants regarding their statistical purpose for the use of microdata.

When can microdata be shared

Access to microdata can only be granted by the ABS. Microdata must not be shared with other individuals as this is a breach of the Undertaking signed by your organisation. Any person who needs access to microdata or uncleared DataLab output must apply and be approved by the ABS before accessing the data. This includes members of your research project team who will see, discuss or use microdata or uncleared DataLab output. Sanctions may be brought against organisations and individuals found to be sharing access to microdata with unapproved individuals or organisations.

Where researchers from different organisations are collaborating on a project, the ABS may provide permission to share specific microdata and uncleared DataLab output. See Collaborating on microdata research below.

When can microdata output be shared

Output that can be shared:

• Cleared DataLab output: All DataLab output that has been cleared by the ABS may be shared or published.
• Basic microdata output: Tables or other aggregated output (for example, averages and model parameters) may be disclosed, published or disseminated by the user.

Examples of basic microdata aggregate data that do not reveal microdata and are not required to remain in a secure environment include:

• tabulations containing cells with 3 or more unweighted contributing sample units
• details of fitted models, such as regression parameters
• summary and test statistics from estimates
• linear and non-linear regression coefficients

When disseminating aggregated output or reports, you must ensure that you correctly attribute ABS microdata as the source. See Microdata user obligations for Licence provisions, copyright and attribution.

Output that cannot be shared:

• unit record data (microdata) must not be disclosed or disseminated by the user, and must be securely stored
• all data and output from the DataLab that has not been cleared by the ABS
• unit record data from basic microdata

Collaborating on microdata research

Microdata collaboration can occur when consultants conduct work on behalf of an organisation, or when multiple organisations are working together on a research product.

Collaborating researchers need to be approved to access microdata:

• if you are working directly with microdata you must have approval from the ABS to access the specific microdata product
• if you are involved in discussing uncleared DataLab data or output you must be approved to access the microdata in that project

Contact microdata.access@abs.gov.au to ensure the correct approval has been granted for all individuals in your project

Collaborating researchers can work on a project without seeking further approval if you will only be viewing or discussion DataLab output that has been cleared by an ABS officer. For example, if a consultant provides a report with conclusions drawn from analysis of the microdata to another organisation, then that organisation is not required to have approved access to the microdata.

DataLab project workspaces:

• Researchers may work together on a project within the DataLab. In this case, the ABS may provide a shared workspace only accessible by the project researchers.
• People approved to participate in the project may share their uncleared DataLab research and results within the DataLab workspace, whether it is microdata or aggregate data. This is because all the researchers have been approved to access that microdata and access is within the secure DataLab environment.

Using more than one microdata file together

You may merge microdata files where:

• the data has been released as a multi-level or linked file and
• identifiers have been provided for this purpose and
• you have been approved to do so as part of your project if you are using the microdata in the DataLab

You must not match microdata files where identifiers have not been provided for that purpose. Different microdata releases may include information about the same record. For example a person may be selected in one survey and also be included in an administrative microdata file about people. In this case, identifiers will not be provided to facilitate matching that record across the two microdata files. You must not attempt to match that record based on their characteristics on the two files. This includes matching two ABS microdata files or matching ABS microdata with microdata from any other source.

An example of a multi-level microdata file is person data in one file and household data relating to those persons in another file. In this case, each record on the person file will include a person and household identifier. This enables you to group persons into households, and to undertake research about those people as a household group. Merging microdata files is only permitted where identifiers have been provided as part of the release.

ABS releases microdata files from many sources including surveys, censuses and administrative data provided by other organisations. In some cases the ABS may undertake a project to match or link records based on their characteristics. You must not attempt to match or link microdata records yourself. When the ABS links records to create a linked microdata file, it is assessed to ensure that the confidentiality of the records is protected before being released. Where the linked data is provided in multiple files, they may only be used together if record identifiers have been provided for that purpose and you have been approved to do so as part of your project.

Researchers may want to concatenate or pool two microdata files. This is where the records from two microdata files are stacked so that all of the records are in a single microdata file. While it is generally safe to do this, you must seek approval as part of your DataLab project. Different versions of the same file, such as basic microdata and detailed microdata from the same survey and reference period, must never be concatenated or used together.

If you have been approved for more than one DataLab project, you must not use microdata files approved for one project in a different project.

Researchers may also want to add aggregate data characteristics to a microdata file. This may be appropriate, provided it is not undertaken in a way that assists identification of the record. For example, if median salary by sex and age was added to a person microdata file, this would not represent the actual income for each record but may benefit the research.

Aggregate data and microdata

Aggregate data is produced by grouping information into categories and combining values within these categories, for example, a count of the number of people of a particular age (obtained from the question ‘In what year were you born?’). Aggregate data examples include tables, means, medians, ranges and regressions.

Microdata is a dataset of unit records where each record contains information about a person, organisation or other record. This information can include individual responses to questions on surveys or administrative forms, which are characteristics of that record.