RDP 2018-07: The GFC Investment Tax Break Appendix A: Data

The specific data items we use from BLADE are reported in businesses' business activity statements (BAS). The measure of investment we use is capital purchases, which includes plant and equipment, as well as land and buildings. It does not include intangible assets (ATO 2017a). We use total revenue as the measure of revenue. We exclude the mining division to abstract from the mining boom, and the finance division due to data issues.

The Capex Survey seeks responses from a sample of around 9,000 businesses each quarter. The largest businesses in Australia, roughly 3,000, are included in every quarter of the survey. This is necessary to accurately estimate aggregate capital expenditure, given the skewed distribution of capex. The capex of smaller businesses is estimated based on a random sample that is stratified by industry and size. For example, in a particular quarter, there were around 1,200 manufacturing businesses with between 50 and 99 employees in Australia. The survey sought responses from 10 of these businesses, and a weight of 120 was applied to each response when estimating total capex for this subset of businesses. Responses from the larger businesses that are always in the sample are assigned a weight of 1.

The design of the Capex Survey changed in the September quarter of 2009, in a way that is material for the analysis in this paper. Non-employing businesses were included in the survey population from this quarter. This generated an increase in aggregate estimates of capex, as estimates now approximated the capex of a larger set of businesses. Our primary approach to dealing with this change is to construct capex estimates for the population of employing businesses only. As with the BLADE data, we exclude businesses in the mining division to abstract from the direct effects of the mining boom (particularly in the graphical analysis).