• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

3. USE OF BEST PRACTICE METHODS AND SOURCES TO MAXIMISE DATA QUALITY AND MINIMISE RESPONDENT BURDEN

Given the number of organisations involved in the production of agricultural statistics there is likely to be a degree of variability in the application of best practice statistical methods and utilisation of alternative sources of data. The quality of statistical assets will also vary along with the degree of burden placed on respondents. However, the review found that two-thirds of the statistical assets currently being used by stakeholders mostly met their needs. Stakeholder feedback indicated that a large proportion of agricultural statistical assets being used are produced by Commonwealth, State and Territory government agencies (Table 9 Appendix 4). In many cases these government agencies apply best practice in the production of their statistical assets, for example the application of the ABS Quality Framework and the use of the Statistical Clearing House (Box 1 Chapter 3). However, it is difficult to assess how widespread the application of best practice data quality frameworks is across all producers of agricultural statistics.

While there is a clear indication that the majority of end users are satisfied with the currently available agricultural statistical assets there is room for improvement with a number of stakeholders identifying issues with currently available data sets that affect the utility of the data (see Chapter 3 – Data quality).

Improving the use of best practice statistical methods is not a straight forward process and even the application of best practice may not alleviate some issues with data quality. For example improving one dimension of quality may require trade-offs with other dimensions of data quality. NASR consultations indicate that priorities for quality improvement vary between users. For some, improvements in the timeliness and accuracy of ‘headline’ commodity statistics, for example the size of the national livestock herd, are the most important concern. For others, relevance is more important, specifically the availability of more detailed commodity data and/or data at finer levels of spatial disaggregation to support the needs of their industry or policy interest. Consideration should be given to the appropriate balancing of quality dimensions. For example, some users may be willing to trade-off reduced detail in a dataset for improved timeliness or increased accuracy.

While there are a variety of statistical methods and sources in use across the current Australian agricultural statistical system, direct collection from survey respondents is the predominant method in use among both official and non-official sources. This is affecting the sustainability of these statistical programs both in terms of cost (given the rising cost of direct statistical collection), and quality (given the increasing respondent burden and impact on response rates). Two strategies are required to improve the performance of the Australian system in this area: an increased use of alternative data sources, and a stronger focus on increasing respondent engagement and minimising burden.

There is an increasing range of alternative data sources, such as administrative data (Chapter 2) that are potentially under-utilised and which should be explored for their potential to meet demand for agricultural statistics in Australia. The NASR identified a number of such sources, including commodity levy payer records, the National Livestock Identification System (NLIS) and Pigpass. Data held further down the supply chain by, for example, storage and transport bodies such as bulk grain handlers, testing and quality assurance bodies such as the Wool Testing Authority, processors and manufacturers, and retail and wholesale markets could also have a high degree of utility. Access to such data sources may not be straight forward because of confidentiality and privacy issues, and there is a need for a value proposition to encourage private sector bodies to share potentially commercially valuable data. Nevertheless, harnessing of these data sources could contribute to improving both the accuracy and timeliness of agricultural statistics.

Engagement with survey respondents could be improved in line with the best practice principles described in Chapter 4. Improved respondent engagement will have benefits for both the quality and timeliness of statistical collections. Organisations conducting statistical collections should make efforts to better coordinate their survey collections to reduce respondent burden; consider making improvements to the timing and length of their surveys; investigate the use of collection modes that are more convenient for respondents, such as e-forms; and take steps to identify and manage unreasonable survey load. During the NASR consultation a survey calendar listing the dates of surveys undertaken by government and non-government agencies and the likely timing of key production activities was suggested as a simple means of reducing overlap, and consequently, respondent burden; and for informing survey respondents of the range of surveys undertaken each year, and the purpose and timing of each survey. There is also scope to improve the effective use of the Statistical Clearing House as a mechanism for minimising duplication in relation to Australian Government-led agricultural surveys, as it is clear from the NASR consultation that its role may not be fully understood or utilised.

While these strategies would help to manage respondent burden, the NASR consultation identified the need to improve the value proposition for respondents—specifically, finding ways to return data to respondents in a form that is more directly useful for their business. An example might be providing businesses with a benchmark comparison of their data against aggregate data from other businesses in their industry or region. Industry bodies offered to assist with identifying means of doing this through engagement with their constituents.

The burden imposed by statistical collection needs to be weighed against the benefits. Some level of survey burden is necessary to the extent that some core statistics may only be feasibly produced from direct collection. Other means of reducing burden, such as improving the timing and mode of collection, exploring alternative data sources, or removing survey duplication through better coordination, should be explored in preference to reducing any high priority content.