Producing agriculture financial statistics using linked data

The Agriculture Statistics NSC, with support from Analytical Services Branch and Economic Conditions section, is currently undertaking a feasibility study to investigate the possibility of linking data collected as part of the Agricultural Census (commodity production data) with Business Income Tax (BIT) and Business Activity Statement (BAS) data from the Australian Taxation Office (ATO). This project has been made possible through user funding provided by the Department of Agriculture, Fisheries and Forestry (DAFF), who saw potential and were keen to assist the ABS add value by linking these two rich data sources. If successful, the linked dataset will provide a rich source of information useful to discussions on the economic performance of businesses operating within the agriculture industry.

The first phase of the feasibility study was to link, at the unit record level, the responding in-scope units from the 2005-06 Agricultural Census (approx 138,000) with the BIT and BAS data held in the ABS. Analytical Services Branch provided some advice on the linking methodology. A form of deterministic linking was implemented using Australian Business Number (ABN) along with other business structure information. Some specific issues encountered when linking with ABN included:

• Tax records from multiple ABNs being consolidated under one head ABN;
• A business (with one ABN) operating over multiple locations (multiple Agricultural Census records); and
• The statistical reporting ABN may not be the same as the tax reporting ABN.

Business structure information was used to deal with these issues and ensure the maximum number of meaningful links were made.
Initial linking results are very encouraging and indicate that approximately 90 per cent of 2005-06 Agricultural Census records can be linked to 2005-06 BIT and BAS records. The other important point is that unlinked Agricultural Census records have roughly the same proportional distribution by state and commodity as the linked records providing confidence that bias in our results will be minimised. Imputation is currently being considered as the best option for dealing with the unlinked records.

The next phase of the feasibility study, which is currently underway, is to focus on a conceptual comparison of Tax Data and ABS Financial Data, to ensure that the linked data set can be used to measure the appropriate financial concepts. Generally speaking, the concepts align fairly well, however there are some slight differences in the definition of some individual data items. At present, the funders and other potential users of the estimates (e.g. ABARE, National Accounts Branch) are being consulted regarding the financial concepts that are being measured.

ABS staff have provided a presentation to DAFF outlining the project achievements to date which has generated considerable interest from staff within the funding body. The project also provides ongoing opportunities for the ABS. The current and ongoing consultation will ensure that any estimates derived from the linked data set will inform and support the information needs relating to agricultural statistics.

For more information on this project, please contact Bella Liao on (02) 6252 7311 or Jeffrey Wright on (02) 6252 5294.