RDP 2025-01: Are Investment Tax Breaks Effective? Australian Evidence 4. Data

We merge survey data from the ABS with business taxation administrative data from the Australian Taxation Office (ATO) to construct our sample. This provides us with two different sources of investment data. The first is the CAPEX survey. This is a survey of around 10,000 firms each quarter. Consistently large investors as determined by the ABS (large mining companies, banks and a few other large firms) are surveyed every quarter, while the remainder are stratified by industry, location and employment size, drawing from the administrative business register. The advantage of the CAPEX data is that they provide investment by asset type: machinery & equipment (assets that are generally eligible for the investment incentives) and buildings & structures (assets that are not generally eligible for investment incentives). However, the sample is too small to use for our RDD. Once combined with other taxation data (see below) we have a panel with around 90,000 businesses and around 667,000 observations spanning the period 2000/01 to 2018/19. See Appendix B for summary statistics of the data that we use to analyse each policy.

Our second source of investment data is capital expenditure reported in firms' business activity statements (BAS). These are available quarterly covering the period 2001–2021. As these data are reported by almost all firms, the sample is very large.^[6] However, eligible and ineligible investment are not separately reported in these data.

For the analysis, we consider both the intensive and extensive margins. The intensive margin is simply the (natural) log of investment. For the extensive margin, our dependent variable is the log odds ratio for the share of firms investing, calculated at the industry division–turnover category level, as in Zwick and Mahon (2017).^[7] The extensive margin is particularly relevant for the analysis of small business policies, as only around 25 per cent of small firms (< 20 staff) invest each quarter based on tax data, compared to around 80 per cent of large firms (200+ staff). In our RDD design we focus only on the intensive margin.

These investment data are combined with other tax data available on an annual basis. These tax data contain a rich set of information on business type, turnover, assets, liabilities and ownership structure. The tax data also include details of which businesses are connected to others through a business group identifier. This aggregation is useful since the legislated turnover tests are on an ‘aggregated turnover’ basis, which includes the annual turnover of not just the primary business but also of any other controlled or affiliated businesses in related business groups. Lastly, the tax data allow us to differentiate between different legal structures: companies, partnerships, trusts and individual entities. These distinctions are important in the Australian context given companies are subject to a dividend imputation regime, which may limit the value of ITB to the firm.

For the analysis, we follow Rodgers and Hambur (2018) and exclude the mining sector to abstract from any effects from the mining boom. The public sector is removed (based on the 2008 ABS Standard Economic Sector Classifications of Australia). The finance sector is also excluded from the RDD, due to conceptual difficulties measuring investment in the tax data for the sector.

Footnotes

From 2017/18, small firms below $10 million in turnover no longer needed to report on their capital expenditure in the BAS. However, all policies from this point on were aimed at firms above this threshold, so this does not affect our analysis. [6]

More formally, it is defined as ln $\frac{P_{n,s,t}\left(Investmen{t}_{i,t}>0\right)}{1-P_{n,s,t}\left(Investmen{t}_{i,t}>0\right)}$ where n indicates industry, s size and t time [7]