Statistical methodology research

Commencement date: May 2021
Project status: In progress

The Census of Population and Housing (Census) is one of the ABS’ most critical, legislated functions. The ABS is using administrative data to enhance the efficiency and quality of the 2021 Census. This project guides how Census data is collected, improves the quality of the Census count, and may be used to fill data gaps arising from events such as natural disasters.

Administrative data will be used to: 

• Help maximise Census response by identifying areas where people may need extra support to complete the Census and having strategies in place to deliver this support.
• Help improve the Census count by refining our methods to estimate the number of households unoccupied on Census night, and to adjust where necessary to ensure the Census count is accurate.
• Help us prepare for any unexpected events that could impact the Census response. Identifying other data sources is a key way to be prepared, in case there is a need to fill in any gaps where information has been difficult to collect.

Administrative data will also be used as part of the quality assurance process, once 2021 Census data have been collected. The ABS doesn’t collect, use or share administrative data to make people comply or to charge or fine people.

To promote transparency of the work that is being conducted, the ABS has published information on using administrative data for the 2021 Census.

The Census is critical to informing the planning and delivery of Government and community services, business decisions and academic research. This project will improve the quality of data from the 2021 Census of Population and Housing by supporting strategies to reduce the undercount and overcount of people in Australia on Census night. Better data means better decisions to improve the lives of Australians.

March 2020: 2021 Census Administrative Data Privacy Impact Assessment

Commencement date: May 2021
Project status: In progress

The ABS Survey of Income and Housing (SIH) collects information about the income, housing and living arrangements of Australian households and how they change over time. This project involves the production of high-quality data from the SIH using integrated data to supplement or substitute reported data. The feasibility of using integrated data across other purposes in SIH will also be explored, including opportunities to reduce future provider and processing burden in SIH through supplementation or substitution of survey data with administrative data.

Maintaining high quality SIH information is critical for policy makers and researchers seeking data to inform a variety of social issues such as inequality, poverty, housing affordability and costs of living. The use of integrated data for the SIH may also reduce provider and processing burden.

Commencement date: April 2020
Project status: In progress
 

This exploratory project will use linked data to investigate non-response and collection mode choice for Labour Force Survey, Survey of Income and Housing and General Social Survey data during suspension of face-to-face interviews due to COVID-19. This work is exploratory and will be used to assess the benefit that could arise from using administrative data to help correct for survey non-response by household characteristics.

This project will assist in maintaining the quality of statistics by using administrative data to assess whether the weighted survey sample is representative of household characteristics in the populations of interest. The data collected by the Labour Force Survey, Survey of Income and Housing and General Social Survey during the COVID-19 pandemic will be valuable for current policy decisions related to the labour market and households, and for economic and social research in future.

Commencement date: 11 May 2020
Project status: In progress

The pilot study aims to examine whether electricity data could be used to enhance or bolster ABS economic statistics, or to produce more timely indicators that would be of value in the context of disruptions due to COVID-19.

The advantages of using electricity consumption data are the coverage and ability to show timely changes. Analysis of electricity consumption data can complement other household and economic statistics in supporting evidence-based policy making in the current uncertain environment. The data also presents unique opportunities to expand and enhance ABS’ suite of statistical products.

2/10/2020 - Using electricity data to understand COVID-19 impacts, 2020 (cat no. 4661.0)

Commencement date: February 2020
Project status: In progress

This project uses international student data from the Provider Registration and International Students Management System and Department of Education, Skills and Employment, to assess the quality of existing linkage to migration data on the MADIP person linkage spine (‘the spine’). It will determine the feasibility of using international student data to enhance coverage of the spine, with the possibility of birthing new units onto the spine. The data will be integrated as a once-off linkage.

Spine quality is important for improving linkage rates for all research projects seeking to link into MADIP, in particular, projects seeking to understand migrant populations. Improving spine coverage and quality will lead to better analysis and policy outcomes.

Commencement date: 17/04/2019
Project status: In progress

This project is researching how linked administrative data could improve the Household Income and Expenditure Survey (HIES) and the Census.

The project links data from the 2015-16 HIES to the Multi-Agency Data Integration Project (MADIP) integrated asset (specifically Census, personal income tax, and social security data). The linked data will not form part of the enduring MADIP asset.

The research is investigating the potential of these data to:

• Improve the collection of data to reduce the cost and burden on the taxpayer and respondents
• Supplement the range of income statistics the ABS provides to the Australian community
• Improve the quality of income data collected in the HIES and the Census

Information on personal and household income is critical to informing the planning and delivery of Government and community services, business decisions and academic research. Better data means better decisions to improve the lives of Australians.

The project will provide public value through:

• Potential improvement of the quality of ABS income statistics
• Potential reduced costs for the ABS’ Household Survey Program
• Understanding of potential new statistical solutions based on linked administrative data, contributing to better policy outcomes and more targeted public spending

Commencement date: 29/03/2019
Project status: Complete

Multi-location businesses tend to report their business activity from their head office. As a result, business data is heavily biased to the location of the head office, when in reality much of the business activity may be taking place elsewhere around Australia. This leads to an unknown number (and location) of business locations associated with these large, complex businesses, making it difficult to accurately estimate the count and location of business premises across Australia.

This project aimed to take the first steps towards resolving this problem.

This project evaluated the use of the Multi-Agency Data Integration Project (MADIP) (specifically Census and personal income tax data) and the ABS Business Register to determine whether Census business location information (noted down by people on their Census form) can provide some insights into the multiple locations of large businesses.

The project was for internal research and development purposes only.

More accurate business location information may provide governments and researchers with a better understanding of the drivers of economic activity across Australia. Policy makers and researchers will be able to undertake analysis on the location information and composition of businesses and will also be able to track the growth and decline of businesses and their locations from Census to Census, enabling a better understanding of the direct regional effects of structural changes in the economy.

The project has successfully demonstrated the utility of Census business location information to help inform the multi-location business problem. In particular, analysis of the linked file has enabled the identification of approximately 600% more business locations (at the SA2 level) for multi-location businesses than what appear on the ABS Business Register.

The output from the report will be passed to the Business Register Unit (BRU) to help assist them with their FY 2019/20 project to spatially enable and improve location information on the ABS Business Register.

Commencement date: 25/03/2019
Project status: In progress

Aboriginal and Torres Strait Islander population estimates are critical to the formation of government policy aimed at improving outcomes for this important population. Because of the challenges in sourcing data of sufficient quality, these estimates are currently updated only once every 5 years based on incoming Census data.

This project is a feasibility study to assess the potential of the Medicare Voluntary Indigenous Identifier (MVII) data to support Aboriginal and Torres Strait Islander population estimates. The project links the 2016 Medicare Enrolments Database (MEDB) population to the 2016 Census of Population and Housing to compare MVII with Indigenous status data from the 2016 Census. This project is an important step in building confidence in the MVII data for population estimation.

This project uses anonymised IDs to protect the privacy of individuals.

Improved and more frequent estimates should result in better targeting of government funds to relevant regions and communities around Australia.

This feasibility work continues ABS efforts to uncover potential for improved and more frequent updates of these estimates, which represent a long-standing information gap.

Commencement date: 09/11/2018
Project status: In progress

This project will investigate the feasibility of integrating survey weights from the Business Characteristics Survey (BCS) to an extract of the Business Longitudinal Analysis Data Environment (BLADE). The aim of this is to allow the production of BCS estimates by some BLADE data items. For example, BCS data is nominally output by Industry Division and Employment size, however this project may allow outputs to be produced using non-BCS BLADE data items such as derived Turnover from BAS data.

Note that this project is a feasibility study and thus findings may not ultimately allow for the integration of BCS weights to the BLADE asset. If it does, the integrated dataset will only be used to produce aggregated outputs for approved classification variables for internal analytical projects.

This project supports research and analysis that feeds into innovation policy development and evaluation in Australia.

Commencement date: 23/10/2018
Project status: Complete
Completion Date: June 2019

This project sought to understand labour force reporting in the 2016 Census and the August 2016 Labour Force Survey. The ABS will use the results of this project for internal improvement processes.

Insights from this work will inform the design of the 2021 Census.

Commencement date: 02/10/2018
Project status: In progress

This project will assess the technical and operational feasibility of contacting ABS survey respondents using outgoing telephone or SMS contact, and will evaluate use of telephone number databases, such as the Telstra Sensis and Integrated Public Number databases, for use in field operations of the ABS household survey program.

More efficient and effective data collection will minimise costs, and reduce burden on households selected in ABS household surveys, by reducing the number of visits to respondents' homes.

As this work is intended to improve operational efficiency of data collection, no publications are planned.

Commencement date: 06/08/2018
Project status: Complete

The ABS and the Australian Institute of Health and Welfare (AIHW) together with the Commonwealth Department of Health and Victorian Department of Health and Human Services are working together to explore ways to connect the integrated data assets held by ABS and AIHW to create enhanced research datasets without needing to duplicate the data integration work already undertaken. The mechanism for achieving this is referred to as interoperability. Interoperability would enable, subject to relevant approvals, in-scope data held by the AIHW or in state agencies where the AIHW has created linkage keys to be combined with data held by the ABS.

In the first phase of this project, the AIHW and ABS created an interim interoperable spine using the common elements of the Medicare Enrolments Database held at each agency. In the second phase, these linkage keys have been used to assemble a file of combined Victorian community health and MADIP data to enable research for Victorian policy development. The third and final stage prepared a business case exploring data integration operability options and for maintaining an enduring interoperable linkage asset beyond 30 June 2020.

If interoperability achieves sufficient quality, AIHW and ABS will be better placed to combine data assets for approved research projects, and could do so in a way that enhances privacy, minimises duplication and data transfer.

Commencement date: 22/05/2018
Project status: In progress

This project will assess the value of emerging methods and technologies for combining different data sources in order to maximise the usefulness of public sector data, and produce useful information to help inform key policy and research questions, such as around performance and innovation in Australian exporting businesses.

This project will be internally assessed with learnings to help inform any future directions – any future directions would be separately advertised.

Commencement date: 28/03/2018
Project status: Complete

The ABS worked with academics at the University of Wollongong to develop a program that takes into account the additional uncertainty using linked data creates in producing estimates from regression analysis. When performing regression analysis, there is always some uncertainty in the estimates produced. This uncertainty is greater in linked datasets and may produce consistent errors that over-estimate or under-estimate the measured effect (bias). This project aimed to develop a method of taking potential biases into account when analysing linked data and thus improving estimates based on linked datasets.

By developing a method to improve regression estimates for linked datasets, the ABS will be able to:

• Correct for biases for data linked from two sources.
• Obtain more reliable results for analysing linked data.
• Expand the available tools within the ABS for analysing linked data.

Commencement date: 27/03/2018
Project status: In progress

This proof of concept study investigates the feasibility of linking administrative datasets from across the criminal justice sector. It will test data integration methodologies and techniques by linking experimental datasets from police, criminal courts, corrective services, and juvenile justice (for selected states and territories only). There are no outputs to be published during this exploratory phase; the intention is to inform the development of the full project for discussion with our key clients, the Justice Boards of Management.

The project is endorsed by the Council of Australian Governments and delivered in partnership with the National Indigenous Australian Agency and the Australian Institute of Health and Welfare.

The exploratory project has been extended to also investigate the feasibility of linking selected experimental justice datasets to other non-justice datasets.

This project will provide greater insight into the flows of persons through the criminal justice system.

Commencement date: 26/03/2018
Project status: On hold

The development of methods to enable greater use of Big Data for the production of official statistics has only recently started to grow. This project aims to apply and extend these methods to available ABS agricultural datasets as a proof-of-concept to determine whether the methods can be applied in the ABS context, and if they will lead to the production of more efficient official statistics.

This project will potentially open up new sources of data for the ABS to use when producing agricultural statistics, resulting in reduced costs and respondent burden, and more frequent and a broader range of statistics.

Shaw, S and Bell, PA (2019), “Regression data integration and constrained imputation – small domain estimation using an administrative source”, presented at September 2019 Methodology Advisory Committee meeting, Sydney, September 2019

Pham, H and L Ang (2018), “Regression Data Integration – Results From a Study to Improve the Efficiency of Survey Estimates Using Imperfect Auxiliary Datasets”, presented at November 2018 Methodology Advisory Committee meeting, Sydney, November 2018

Tam, S-M, J K Kim, L Ang and H Pham (2018), “Mining the New Oil for Official Statistics”, presented at BigSurv18: Exploring new statistical frontiers at the intersection of survey science and big data, Barcelona, October 2018

Commencement date: 22/09/2017
Project status: Completed

This internal ABS feasibility study investigated an appropriate method for data linkage that uses name information which is encoded as a privacy protection. The project also assessed the impact of using encoded name on data linkage quality and the ability to produce statistical insights, particularly about migrant and Aboriginal and Torres Strait Islander communities, given the high policy and research interest in these populations.

This project will enhance the value of ABS data linkage products by supporting the implementation of best practice for name encoding and data linkage methods.

Commencement date: 21/09/2017
Project status: In progress

This study will assess the potential of administrative data to inform more efficient data collection and improved data quality for the 2021 Census.

Key outcomes include:

• Predicting dwelling occupancy at the time of the 2021 Census
• Informing the imputation strategy for non-responding dwellings.

The output will be research and analysis to inform ABS decision making for the 2021 Census.

The benefits of the project include:

• More efficient resource allocation for the 2021 Census
• Improved dwelling imputation for a clearer community profile
• Filled information gap for housing policy (dwelling vacancy/occupancy rates).

Commencement date: 09/03/2016
Project status: In progress

The project will assess the quality, completeness, and coherence of taxation data with the aim of assessing taxation data’s potential to substitute, enhance, and expand the ABS statistical program.

The three broad streams of work for this project are:

• Improving ATO data quality for statistical purposes;
• Improving understanding of ATO data coherence and concepts; and
• Exploring potential use of ATO data in statistical production.

Most benefits arising from these investigations will be in regards to the data quality of taxation datasets.

Commencement date: 01/10/2013
Project status: Ongoing

The ABS Address Register is a current list of all physical address (both residential and non-residential) in Australia. The main input to the register is the Geocoded National Address File (G-NAF), with continuing supplementation from other available address sources and from desktop/field work undertaken by ABS officers.

Note: No personal information about residents in the addresses are collected or stored on the Address Register.

The Address Register has enabled improved and more efficient data collection and creation of statistics, including facilitating the production of new datasets that have the potential to expand the range of official statistics to better inform Australian society in the future.

Commencement date: 30/05/2013
Project status: Ongoing

This project aims to maintain a high quality real-world statistical business register as the basis for coherent and integrated ABS economic data.

The ABS Business Register is a list of organisations which undertake economic activity in Australia. The data on the ABS Business Register is primarily sourced from the Australian Business Register (ABR), the Australian Tax Office (ATO) and via ABS profiling of large, and/or complex businesses, and the structural relationships between related businesses.

The key users of the ABS Business Register are ABS staff compiling economic statistics. Another integrating authority would be unable to undertake this project because of legislative constraints. The unit record information on the ABS Business Register is not available for release outside the ABS.

The ABS Business Register is an ongoing project.

An integrated infrastructure product from which many economic collections draw their frames is vital for coherence and comparability of economic statistics.

Annual release - Counts of Australian Businesses, including Entries and Exits (cat. no. 8165.0)
This release contains counts of actively trading businesses, rates of entry to and exit from the market sector of the Australian economy, and rates of business survival. These data are sourced from the Australian Bureau of Statistics Business Register.

Commencement date: December 2012
Project status: Complete

The Local Government Finance Quality Study used the ABS Australian Business Register and information from State Auditor General departments to review the quality of coverage for the Local Government Frame produced by the Public Finance section at the ABS.

This investigation confirmed the existing Local Government Frame was accurately capturing all financial activities of the local government sector.

Units that were believed to be missing from the frame were not in fact separate corporations or entities. While displaying some characteristics of separate business entities, these units were confirmed to be business activities that were reported within the constituent local councils.