RDP 2023-09: Does Monetary Policy Affect Non-mining Business Investment in Australia? Evidence from BLADE 2. Data

2.1 Tax data on investment

Our analysis uses quarterly data on gross investment for the near universe of Australian firms for the period between September 2001 and June 2017 from ABS BLADE.^[1] The particular data we use are reported in firms' quarterly Business Activity Statement (BAS), in which they report the value of capital purchases they made in the quarter. Our analysis is restricted to the non-primary non-financial private sector. The primary sector (agriculture and mining) is excluded as we expect that commodity prices, rather than monetary policy, are the key driver of investment in that sector. The finance sector is removed due to conceptual difficulties in measuring investment and output in this sector from tax data. And the public sector (health, education, public administration) is removed as fiscal policy is likely to play a more important role in investment in this sector. We also remove firms with no employees as is standard in the literature, though this has very little effect on the results.

We consider both the intensive (how much firms invest) and extensive (whether they invest) margin of investment. We model the extensive margin as an indicator ID_i,t, that takes on the value 1 if firm i invests in time t and 0 otherwise. We model the intensive margin as the log of investment Inv_i,t.

Table 1 provides a summary of the data. The full sample contains around 34 million observations, of which around 29.6 million are from small firms (below 20 employees), 3.9 million are from medium-sized firms (20–199 employees) and the rest are from large firms (200 or more employees). There are a little over 2 million unique firms in each quarter.

Of the 34 million observations, around 9.6 million have positive investment, while the rest have zero investment. This implies that in any quarter, around 28 per cent of firms are investing. The share is lower amongst small firms, where only around 25 per cent invest in a given quarter, and much higher amongst large firms, where around 78 per cent invest in a given quarter (Table A1). The share of firms investing varies moderately over the business cycle, having been strong in the mid-2000s alongside strong economic activity during the mining boom, but lower over the 2010s as overall investment and economy activity also softened (Figure C5). This provides some initial evidence that the extensive margin may be an important margin of adjustment.

Even conditional on investing, the amount of investment is highly heterogenous. While the average investment, conditional on investing is $170,682, the standard deviation is almost $11 million. This also indicates that investment activity is highly positively skewed, with a small number of very large investment observations.

2.2 Survey data on expected investment

Data on firms' expected investment is sourced from the ABS survey of New Capital Expenditure (CAPEX). This is a quarterly survey of around 10,000 firms. It is a census of the largest investors, with the rest of the sample being made up of a random stratified sample. For our analysis we do not apply sampling weights. As such, for regressions over the entire sample we will overweight large firms (at least in terms of the number of large firms).

Our CAPEX sample is made up of around 519,000 observations covering from September 2001 to June 2021. Consistent with the relatively larger number of large firms in this sample, in any given quarter around 48 per cent of firms have positive investment (Table 2).

As well as asking firms about their current investment, it asks about their short-term and long-term investment expectations. The horizon of these expectations differs based on the calendar quarter, but the short-term expectations tend to cover investment over the next one to two quarters, while long-term expectations cover investment over a period of two to four quarters, starting in one to two quarters in the future. For example, in December 2022 firms will be asked about their expected investment over the next two quarters (short-term expectation), and their expected investment from September 2023 to June 2024 (the 2023/24 financial year; long-term expectations).^[2] As discussed in Cassidy, Doherty and Gill (2012) and Berkelmans and Spence (2013), these forecasts are noisy estimates of actual investment, particularly at longer horizons, with firms tending to systematically underestimate their investment.

Footnotes

After 2017 the data collection method changed. For data consistency we end the sample in 2017. [1]

In March 2023 the short-term expectation will be for the next quarter, and long-term is again for the 2023/24 financial year. In June 2023, the short-term expectation is for the next two quarters (September and December 2023), and the long-term expectation is for the subsequent two quarters (March and June 2024). And finally in September 2023, the short-term expectation is for the next quarter (December 2023) and the long-term expectation is for the subsequent two quarters (March and June 2024). [2]