RDP 2003-10: Productivity and Inflation 7. Discussion

Section 5 found a significant causal flow from inflation to productivity growth that varied by industry. Section 6 established that this finding is reasonably robust. This section discusses our results by expanding on some of the issues up to now implicit in this paper.

A group of arguments suggest the inflation-productivity growth nexus is all about capital accumulation. These effects should be revealed in differing effects of inflation on labour and multifactor productivity growth – we explore this first. We then set out the contention that the relationship is a story of comparative industrial structure. In particular, we focus on one aspect of industrial structure, the distribution of firm sizes. This comes from synthesising two observations. The first is that the relationship's sign appears to correlate with the importance of small relative to large firms in the industry's production. The other is that industry inflation, reflecting the industry environment, matters more than aggregate inflation for industry productivity growth. We proffer this as a partial explanation of our results – other, unobserved, factors also undoubtedly play a role. We conclude this section by searching for alternative means of understanding this relationship, but a lack of data frustrates this inquiry.

7.1 Capital Accumulation

Inflation's output costs are often traced to its effect on capital accumulation. Specifically, it is argued that sub-optimal levels of capital investment arise from the distorting effects of taxation on capital income, with this fiscal distortion accentuated by its interaction with monetary policy.[35] A similar argument is found in references to inflation's confounding effect on the ‘organisation of markets’, particularly markets for financial assets and savings, leading to capital being less efficiently allocated.

The inflation-productivity growth literature evokes these arguments to explain the observed decline of labour productivity growth as price growth accelerates.[36] We can test whether this contention is correct by observing the relationship between inflation and multifactor productivity growth. As multifactor productivity measures output per unit of both labour and capital input, it is ‘net’ of the effects of a slower accumulation of capital per worker.[37] To simplify, if inflation only affects labour productivity growth, then slower capital accumulation is the likely explanation for the inflation-productivity nexus; but if inflation also affects multifactor productivity growth, at least part of the explanation lies elsewhere.

We do not find evidence that supports the capital story. When we ask inflation to explain both multifactor and labour productivity growth, our results are highly similar for multifactor productivity growth – recall Table 4. Higher inflation means slower output growth per unit of labour input and capital input. That there is some shifting of significance between the labour and multifactor productivity growth regressions may indicate that capital possibly plays some role, but that it is far from primary.[38]

7.2 The Role of Industrial Structure

A range of indicators tie the productivity-inflation relationship to industry-specific factors – first and foremost, the results vary by industry and, also, it is industry price deflators that are significant in our model.[39] The model's preference for industry deflators over aggregate deflators tells us that our observation of inflation affecting productivity growth is not only repeated instances of economy-wide trends, but shows firms reacting to price changes in their immediate environment. How firms react to these changes varies by industry, as shown by the previous regressions results. While a range of factors differentiates the industries, we focus on the distribution of firm sizes in this section – in large part because this is most easily quantified.

Let us first conjecture about the process within any given firm. Higher inflation creates a less certain environment for planning, greater cash-flow pressure, an increase in real investment in inventories, delays in account payments and greater bad debts from insolvent debtors. These increased costs of monitoring agents and managing production processes would all be expected to reduce productivity levels and lead to less scope for implementing ongoing improvements to productivity. In larger firms the problems might be expected to be greater. The larger the firm, the more complex becomes the manager's job. Most simply, faster price rises adds an extra layer of uncertainty, so thwarting their task.

The way this general reduction in efficiency translates to industry-wide findings depends on compositional effects. In industries dominated by large firms we would expect there to be minimal compositional effects and, thus, that the within firm effects would also be reflected at the industry level. This is consistent with our finding that higher inflation Granger-causes slower productivity growth in the more concentrated industries. There are few compositional effects obscuring the basic effect and we can be fairly confident that the industry relationships also reflect firm level relationships. In less concentrated industries with more small firms, compositional effects become more important.

Among smaller firms, churning – i.e., entry and exit of firms – is greater.[40] Firm size and failure risk are negatively related, and so we would expect a higher incidence of insolvency in industries that are less concentrated and have smaller firms on average. We also know that the least efficient firms are the most susceptible to insolvency.[41] Bringing these observations together, we conjecture that higher inflation pushes small low-efficiency firms out of business. Inefficient firms are more likely to fail and small firms have less reserves to see them through hard times. The compositional change resulting from this ‘cull of the slowest’ could then lift the average productivity level for those industries where small firms are responsible for a larger proportion of output. This might still imply a negative effect on productivity at the aggregate level to the extent that the affected industry is reduced below its efficient size.

7.3 The Aggregate Effect

The preceding work has used the industry level results to gain a richer understanding of the inflation-productivity relationship. Nonetheless, one may still be interested in the average effect. The aggregate results alluded to in the introduction and presented in Table 4 are one way of looking at this.

At the aggregate level, Granger causality tests suggest that there may be a negative relationship at the 10 per cent level of significance.[42] An alternative is to look at cross-sectional and mean group estimates. We follow Pesaran and Smith (1995) in calculating consistent estimates of the average effect. The mean group estimate, which involves averaging the individual industry estimates, shows a negative but insignificant average relationship. The cross-sectional regression finds a negative and significant average relationship.[43] Thus, there is general agreement across the three methods about the average effect – inflation broadly leads to reduced productivity growth.

Some issues remain open in the work above. We focus on one particular difference between industries. There are others we have not accounted for and that must also influence the prices-productivity growth relationship. More fundamentally, this paper has not specifically addressed the longer run dynamics of the inflation-productivity nexus. It cannot distinguish between a sustained acceleration in productivity growth and a shift in the level of productivity distributed over a period of time. Thus, while we expect that the ‘between-firm’ effect will dissipate in the medium- to long-run, we do not have the data to conclusively prove this.


Feldstein outlines a number of inter-related routes through which inflation interacts with capital accumulation: Inflation distorts the measurement of profits, of interest payments, and of capital gains. The resulting mismeasurement…cause[s] a substantial increase in the effective tax rate on the real income from capital employed in the nonfinancial corporate sector. (Feldstein 1982a, p 153) [35]

Recall from Table 2 that the bulk of the studies only measure productivity in terms of output per unit of labour input, and do not also correlate inflation with multifactor productivity growth. [36]

For exactness, recall that our labour input measures are not employment levels but total hours worked. [37]

In addition, growth of capital services is not statistically significant in our model, although it visually seems to lead the productivity measures by 2–3 years. Finally, neither the real nor the nominal interest rate sits significantly in our model. [38]

At this point, it is useful to review Tables 4, 5 and 6. [39]

On the Australian relationship, see e.g., Bickerdyke, Lattimore and Madge (2000); on the relationship across the OECD economies excluding Australia, see e.g., Scarpetta et al (2002). [40]

Both Bickerdyke et al (2000) and Scarpetta et al (2002) provide empirical evidence that insolvency risk increases as efficiency declines. [41]

While the lags of inflation are jointly significant (p-values of 0.07 and 0.05 for labour and multifactor productivity growth, respectively), the sum of the lags is negative but is not statistically different from zero. For brevity, the entire model and results are not reported here, but are available upon request. [42]

The full details of these regressions are available on request. [43]