RDP 2023-03: Doing Less, with Less: Capital Misallocation, Investment and the Productivity Slowdown in Australia 5. Understanding the Slowdown

There is strong evidence that productivity-enhancing reallocation of capital within industries has slowed. A key question from a policymaker's standpoint is why? This can inform any potential policy response and can also indicate whether the slowdown is likely to be temporary or ongoing.

To help think about potential drivers, it is useful to return to the model outlined in Decker et al (2020) that the regression framework is based on. The model points to three factors that could account for a weakening in the relationship between firm capital growth and productivity:^[10]

• Increases in ‘Correlated wedges’ that distort decisions for more productive firms, causing them to grow (or appear to grow) more slowly.
• Increases in ‘Frictions’ that make it more costly for firms to adjust their labour or capital inputs.
• Declines in ‘Revenue returns to scale’, either due to decreases in returns to scale in production or demand becoming more inelastic, that make the gain from increasing inputs smaller.

We explore three potential explanations that map to these changes, the first of which is more benign, and the latter two which are more concerning.

• The increasing importance of intangible and digital capital, particularly for more productive firms, which could act like a correlated wedge, lowering measured investment for more productive firms.
• Financing constraints becoming more binding, which represents an increase in frictions.
• Declining competition, which equates to a more inelastic demand curve and so a decline in revenue returns to scale.

We choose these three explanations due to their prominence and importance in other papers, though note that there are many other potential avenues that could be explored in future work.^[11]

5.1 Intangible and digital capital

One potential explanation for the apparent weakening in the relationship between capital growth and productivity could be the increasing use of intangible and digital capital. Numerous papers have highlighted the increasing importance of digital and other forms of intangible capital (e.g. advertising and goodwill) in firms' production processes over recent decades (e.g. Haskell and Westlake 2017). However, these types of capital are often not well-measured in firm-level data. For example, while the BLADE tax data on assets do capture certain types of intangible capital (e.g. the value of patents), other types, such as accumulated brand value gained through advertising, are not generally captured. Similarly, some digital capital such as software tends to be recorded as an expense, rather than as capital.

As such, one potential explanation for the results could be that highly productive firms are increasingly investing in intangible and digital capital, rather than physical capital. As this is not captured in our capital growth measure this could make it look like the relationship is weakening, even if it is not.^[12] That said, the bias could also flow the other way if low-productivity firms are ‘catching-up’ and increasingly investing in intangibles as their importance becomes more widespread throughout the economy (rather than just being used by the most productive firms).

Testing for this explanation is inherently difficult, as intangible capital investment is not well measured. As a simple approach, we examine whether the slowdown has been more evident in sectors that tend to use digital or intangible capital more intensively. If we think that more productive firms have increasingly turned to intangible and digital capital, and that this is driving the results, we may expect to see the slowdown in productivity-enhancing capital reallocation be larger in sectors which can and do tend to make more use of such capital. Put another way, we might expect the bias to be worse in sectors where there is a lot of incentive and scope for the high productivity firms to use intangibles.

We use the measures of digital intensity from Calvino et al (2018) based on cross-country data, and of intangible intensity from Demmou, Stefanescu and Arquie (2019) based on US data. We split industries into the most intensive quartile, and other industries, and re-run the model over these sub-samples.

Consistent with the idea that intangible or digital capital use can lead to measurement issues, we see no relationship between productivity and capital growth in the most intensive sectors (Table 3). This is similar to Decker et al (2020), who find no significant relationship between productivity and capital growth in high-tech sectors in the 2000s. More importantly though, the weakening of the relationship is driven by other sectors. This provides some evidence that the results are not driven by increased use of mis-measured intangible and digital capital, though we cannot rule out that this plays some role.^[13]

5.2 Financial constraints and credit supply

Firms have to fund investment, either through internally generated cash flows or via external debt or equity finance. Another potential explanation for the slower flow of capital to more productive firms could be that it has become more difficult for firms to obtain external funding, increasing frictions in investing. This could, for example, reflect declining credit supply, or increasing financial frictions due to firms having less collateral or greater information asymmetries. For example, changing views around risks and regulatory changes following the GFC could, in theory, have tightened firm financing constraints (though Australia was less directly affected by the crisis).

There is a large literature examining the effects of financing frictions and credit availability on firm investment and growth (e.g. Rajan and Zingales 1998). This literature measures financial frictions and credit supply in numerous ways, and there is no one agreed upon ‘best’ measure.

We take two relatively simple approaches from the literature to try to identify groups of firms who may be more exposed to changes in financial frictions and credit supply, to try to tease out the role of financing availability. First, we construct some simple firm-level measures of financial fragility. Second we use industry-level measures of external finance dependence.

5.2.1 Firm-level dependence

We consider two measures that could make firms more exposed to a decline in the availability of external funding:

• Gearing, measured as non-current liabilities as a share of assets. Highly geared firms may have more trouble raising additional external finance to fund future investment.
• Negative cash flows, measured as having negative earnings before interest and depreciation. These firms will not be generating internal cash flows, and so will need to fund investment via external funding.

Table 4, columns 1 and 2 show the results for gearing, where we define highly geared as gearing above the median in the industry in the year. The weakening in the relationship is evident mainly for the less-geared firms. Results are similar if we define high gearing using the top quartile or decile.^[14]

The results are similar if we look at cash flow (columns 3 and 4). Cash flow negative firms have a much weaker relationship between productivity and capital growth, suggesting they are constrained. But the weakening in the relationship has mainly occurred for the cash flow positive firms.

This provides some evidence that increasing financing frictions and decreasing credit supply do not account for the slowdown in capital reallocation. However, there are a number of concerns with these firm-based metrics. First, firms may appear highly geared, or have negative cash flow, as they are in a growth phase and this could complicate the interpretation (though the inclusion of the age variable may assist here). Second, just because the other firms are not financially fragile based on these metrics does not mean that they have not been affected by increases in financing frictions. For example, a cash flow positive firm with large financing needs would still be affected by a decline in credit supply. And finally, it might be that we are actually capturing firms that are distressed and unable to invest whether or not financial conditions worsen (e.g. La Cava 2005). This is consistent with the lack of relationship between productivity and investment for cash flow negative firms. As such, we turn to industry-based metrics.

5.2.2 Industry-level dependence

To abstract from the above concerns, and incorporate a broader sense of firms' potential funding needs, we consider some industry-level measures of external finance dependence. If credit and financing frictions have contributed to slower reallocation, we might expect to see a more substantial slowdown in industries that make more use of external finance.

To consider this we use industry-level measures of external finance constructed by Demmou et al (2019). These measures are constructed using the approach pioneered in Rajan and Zingales (1998), and measure the external financing need as the gap between average investment and internally generated cash flows for the industry. Demmou et al construct these measures using US COMPUSTAT data.

Using US-based metrics has some advantages and some disadvantages. The obvious disadvantage is that industries may finance themselves differently in Australia and the United States. Still, there are a number of advantages: using an external dataset limits potential endogeneity concerns; the US data are based on a much longer sample than we have available; the US-based metrics are well-documented and have been used in numerous different studies, making them a robust and comparable set of indicators.

Demmou et al (2019) produce these measures on an ISIC Revision 4 basis, and we map them to ANZSIC 4-digit industries. We allocate industries as being amongst the top quartile of most exposed firms, or being in other quartiles.

We see that the slowdown has been somewhat larger and more significant in the most financially dependent sectors (Table 5, columns 1 and 2). We see similar results using measures of financial dependence that also account for investment in intangibles and R&D (Table B8), providing further evidence that the results are not simply driven by sectors that may be increasingly investing in (unmeasured) intangibles. The results are also robust to removing mining and agriculture (Table B2) or large firms.^[15] In this case, though, the results are less robust to using the capital productivity measure.

Interestingly, the same pattern is not observed if we consider external equity dependence, with the slowdown being evident in the less equity dependent sectors (columns 3 and 4). The relationship between investment and productivity is much weaker in sectors with high equity dependence, potentially indicating lesser availability of equity finance in the first place and greater financing frictions. However, the slowdown is more visible in the less equity finance dependent sectors. Taken together, this suggests that the slowdown relates to a tightening in access to debt finance.

Notably, if we consider employment growth, rather than capital growth, reallocation slowed similarly in sectors with both high and lower external finance dependence (Table 6). In this case both the initial relationship between employment growth and productivity, and the change in the relationship, are broadly similar. This finding is robust to using labour productivity, rather than MFP. Given financing frictions are more likely to impinge directly on investment this is not surprising (e.g. Spaliara 2009), and highlights the value in exploring capital reallocation to better understand the factors weighing on business dynamism and productivity growth.^[16]

Overall, the results indicate that those sectors that tend to rely most heavily on debt finance experienced the greatest slowdown in capital reallocation, with the gap in outcomes between high-and low-productivity firms falling the most. This provides some evidence that financing constraints may have become more binding over this period, preventing high-productivity firms from investing.

It is unclear whether the results would be stronger or weaker if we were able to include very young firms. On the one hand, they may be stronger as young firms tend to be more sensitive to financial constraints, given they are growing and may have volatile cash flows which can make it hard to finance investment internally. On the other hand, if young firms are always subject to constraints, and the change largely reflects constraints becoming more binding for older firms, including younger firms may actually weaken the results.

5.3 Competition

One final explanation we explore is declining competition. As also discussed in other papers (e.g. De Loecker, Eeckhout and Unger 2020), decreases in competitive pressure can limit selection effects, meaning that unproductive firms may be less likely to shrink and exit, freeing up resources for higher productivity firms. At the same time, more productive firms may have less impetus to continue to grow and improve to stay ahead of their competitors. More generally, in many models where market power is higher, production, and therefore input use, will be less responsive to productivity shocks. As such, the relationship between productivity, and capital growth and reallocation, may be weaker.^[17]

To examine this, we estimate the following extension to the earlier model:

where ${\mu }_{m,t}$ is the industry-level unweighted average mark-up from Hambur (forthcoming). In running these regressions we de-mean industry mark-ups. This controls for the possibility that there may be some third time-invariant factor that affects both the level of industry mark-ups, and the strength of the reallocation effects. It also means we are focusing on the effect of changes in mark-ups (from the industry average), limiting concerns about issues in identifying the levels of mark-ups.^[18]

Table 7 shows the results for capital growth, alongside employment growth similar to Hambur (forthcoming). Where mark-ups and market power increase, the relationship between productivity and capital reallocation weakens, as with employment growth. For example, taking the national average increase in mark-ups of around 5 percentage points and applying it to the coefficients, the relationship weakened by half over the sample. In relative terms, the effect of markups on capital growth appears slightly larger than the effect on employment growth.

Rising mark-ups also appear to weaken the relationship between gross investment and firm productivity (column 2). Given this link, it's also worth exploring whether there is any evidence that rising mark-ups were associated with lower aggregate investment. As a simple approach, we revisit Figure 3, but instead of separating firms based on their productivity quartile, we separate them based on whether their industry had a large increase in mark-ups over the sample – or more precisely, which quartile their industry fell into in the distribution of industry mark-up changes.

Figure 5 shows the results. While they are less compelling than the productivity grouping results, there is evidence that investment declined most in the industries that experienced the largest increase in mark-ups. This provides some evidence that increasing mark-ups and declining competitive pressures may have contributed to weakness in aggregate investment. And more generally, it provides further evidence that the factors that are driving slowing dynamism and reallocation are also contributing to slower aggregate investment.

Finally, it is worth highlighting that the competition and financial constraints explanations are not necessarily independent. For example, while declining competitive pressures might reflect weaker enforcement by regulators, it could also reflect other barriers that prevent firms entering, growing and competing, like financing constraints. We find some evidence that the two causes could be linked, as the increases in mark-ups were on average larger in the most financially dependent industries (Figure B1). But further work is required to better understand the ultimate causes of declining competition and dynamism.

Footnotes

Decker et al (2020) also note that in the presence of non-convex adjustment costs a narrowing in the productivity distribution could lead to a weaker relationship between input growth and productivity. However, Andrews et al (2019) find no evidence of a decline in the dispersion of productivity. [10]

One example would be changes in returns to scale in production. However, our findings of a weakening relationship between input growth and productivity would be consistent with a fall in returns to scale, while most of the literature points to increasing returns to scale. Still, we can't rule out a role for changing returns to scale in production. [11]

Exclusion of intangible capital from the capital stock could also affect our measures of capital, as the measured set of inputs is lower. This would push up measured productivity for firms with large intangible capital stocks. However, as many of these investments are recorded as expenses, intermediate inputs may be overstated. The net effect on measured productivity is ambiguous and will change over time. Similarly, outsourcing is likely to lower measured value added by raising expenses while lowering labour inputs, leading to an ambiguous effect on productivity. [12]

Results are robust to removing the primary sector and are available in the online Supplementary Information. [13]

Available in the online Supplementary Information. [14]

No large firm results available in the online Supplementary Information. [15]

Firms will generally grow along both the capital and labour margin, and so any friction weighing on one could affect the other. However, it may be easier to identify the effects of financing frictions on the factor that they directly impact, given other factors may be subject to some substitution. [16]

Per Decker et al (2020), the role of competition could also be explored based on measures of mark-up dispersion in the context of correlated wedges (i.e. mark-ups rising for more productive firms). But Hambur (forthcoming) finds minimal evidence of increasing dispersion in mark-ups. [17]

Including an interaction between industry and productivity leads to very similar results. [18]