1504.0 - Methodological News, Mar 2012  
Previous ISSUE Released at 11:30 AM (CANBERRA TIME) 22/03/2012   
   Page tools: Print Print Page Print all pages in this productPrint All RSS Feed RSS Bookmark and Share Search this Product

Improving Economic Activity Survey by using BAS Unit Record Estimates (BURE)

BAS Unit Record Estimates (BURE) are available for the key financial data items reported by businesses on the Australian Taxation Office's Business Activity Statement (BAS). BURE are produced on both a quarterly and financial year basis by applying a number of transformations to the raw BAS data. BURE standardise the reporting period of the raw BAS, remove the GST if it is included in the reported values and impute an estimate for missing returns and correct basic errors. BURE are available for each business for the BAS data items of Total Sales, Total Salary, wages & other payments, Non-Capital Purchases and Capital Purchases.

The extensive use of BURE data in the annual Economic Activity Survey (EAS) has led to a substantial improvement in the accuracy of produced outputs. BURE are used directly in the creation of survey estimates for approximately 700,000 of the smallest non-employing businesses. This reduces the size of the business population in scope of direct collection by 35%. This reduces the reporting burden for small businesses and improves the efficiency of the sample for the remaining population.

The most significant gains for the EAS come with BURE Total Sales and Wages & Salaries as benchmark variables in the estimation process, which is based on using the Generalised Regression estimator (GREG). Using GREG, as opposed to the Horvitz-Thompson estimator, has led to a reduction in the EAS sample from 30,000 to 20,000 businesses, without affecting the quality of the key financial estimates that have good correlation with benchmark variables. This represents a significant reduction in costs for the EAS as well as the burden placed on business providers.

Further savings and improvements in the quality of survey outputs are achieved through the improved efficiency of the editing processes for the EAS, with BURE data being used to evaluate the significance of survey reporting errors. In addition, BURE data is used in the imputation of non-responding businesses.

While the use of GREG estimation delivers savings at a broad output level, more intensive statistical modelling techniques that combine survey data and BURE data, can allow for finer level outputs to be created, which would not otherwise be possible without a larger sample. "Experimental' estimates for the Manufacturing industry have been created for a small set of data items at the state and fine industry level. These estimates are based on a sample of approximately 2,700 businesses, but have levels of quality similar to those, which were achieved previously by the Manufacturing collections with a larger sample of approximately 9,000 businesses.

Like any data source, BURE have a number of deficiencies. The main problem is the difficulty in using BURE for large business structures created using the ABS Economic Unit Model and which have complex relationships between the ABN(s) and businesses. With BURE only being available at the ABN level, the BURE records for complex businesses are produced using a number of approximations which unavoidably lead to a lower quality of data. There are plans for improving the quality of these estimates by using additional sizing information, which is currently collected during the profiling of Large Businesses.

The second problem is related to the fact that there may be a genuine reason for difference of reporting Total sales on the BAS statement and on the Economic Activity Survey form. Some providers complete their EAS form using their Profit and Loss statement (P&L). However, BAS data is reported with an emphasis on correctly reporting GST liabilities/credits. The Annual Integrated Collection (AIC) BSC statistical support team is planning to conduct investigations into units with atypical large BURE Sales and Non-Capital Purchases, to see if modified BURE for these units would be more appropriate for stratification, editing, estimation and data substitution purposes.

Apart from continuing work on developing a greater understanding of BAS data, its quality and its relationship with EAS data, there are intentions to expand the use of BURE data in the Economic Activity survey by 1. investigating options for improving the accuracy of AIC Gross Fixed Capital Formations estimates by evaluating the use of BURE Capitalised Purchases as auxiliary information during stratification and estimation, and 2.increasing the scope of the modelled estimates produced at finer output levels to a larger number of industries and data items.

Further Information
For more information, please contact Irina Pribil (03 9615 7566, irina.pribil@abs.gov.au)