Variations in the Utilisation of Productivity Inputs

Multifactor productivity (MFP) is measured as a ratio of output per unit of labour and capital factor inputs to production. Businesses use labour and capital inputs to varying degrees of intensity depending on prevailing economic conditions. The intensity of use of inputs to production, measured against a possible maximum capacity, is defined as utilisation.  

One of the key assumptions underlying productivity estimation is that the utilisation rate of the factor inputs in the production process are constant, and do not take into account technological change. This assumption is applied because changes in the utilisation of factor inputs are typically unobservable. To deal with this assumption, the ABS recommends analysing MFP averages over growth cycles. By reducing the impact of changes in utilisation, MFP growth cycle averages align more closely with their conceptual definition of technology change. 

In the short run, utilisation rates of labour and capital can vary due to a range of factors, including changes in the economic cycle or economic shocks. For example, in cyclical downturns, businesses may retain skilled labour even though there is reduced economic activity. This helps businesses to reduce future costs around recruitment and training when economic activity recovers. Capital inputs are often fixed in the short term, and reducing capital utilisation may be the only option for many businesses in response to temporary shocks, such as lower demand for goods and services. 

This article examines the extent to which the assumption of constant utilisation rates may have distorted productivity measures over the last two decades, by stress testing a selection of productivity assumptions, especially in relation to capital.

There are four potential methods for measuring utilisation adjusted capital services. [3] In general these all relate to using observable utilisation estimates in labour markets and the total economy as indicators for unobservable capital utilisation: 

• Method 1: The ratio of actual to potential market sector real output, where potential real output growth is assumed to be the average annual growth between growth cycle peaks.[4]  However, potential output needs to be projected from the most recent growth cycle peak in 2017-18 to the latest year. 
• Method 2: Economy wide employment rate (derived as 1 - unemployment rate), sourced from the LFS. To the extent that labour is augmented by capital, changes in employment rates may provide timely insights into changes in capital utilisation. [5] 
• Method 3: The economy wide under-employment rate, sourced from the LFS. This approach captures the extent that workers are willing to work longer hours. It assumes that as the availability of work declines relative to willingness to work, that capital utilisation rates may decline accordingly. Alternatively, businesses may be reducing their capital input rather than their utilisation. However, it reflects under-employment conditions outside the market sector. While the assumption in this method, that is, that skills and qualifications in labour supply match labour demand, is unlikely to hold, this method may provide a useful upper bound test for utilisation adjusted capital services. 
• Method 4: Market sector under-employment rate, sourced from LFS. [6] This approach is similar to the economy wide under-employment approach except that the under-employment rate is sourced separately for 16 market sector industries and then aggregated to derive an implied employment rate that reflects willingness to work. This approach has the advantage that industries outside the market sector are excluded and the main industry sources of under-employment rate are identifiable.[7]  

For each method, the resultant capital utilisation indicator is multiplied by the official market sector capital services index to obtain the utilisation adjusted capital services index.

From 2000-01 to 2014-15, growth in published capital services averaged 4.6% per year. However, capital services growth slowed considerably, particularly over the last 5 years, averaging 1.6% per year (Chart 1).

The published market sector capital services index sat midway between the four alternative methods over most of the time span. There is no evidence of systemic drift, with the four methods showing only temporary divergence from published market sector capital services. The strongest divergences occurred for the output gap method (Method 1), particularly near the 2011-12 growth cycle peak, and then again in 2019-20. Variations in utilisation tended to be less for the labour market-based methods (Methods 2, 3 and 4).  

In 2019-20, which captures the impact of natural disasters and part of the COVID-19 pandemic, the output gap method implies that capital services growth fell significantly in 2019-20. Labour market-based methods were mixed, with Method 3 falling slightly and Methods 2 and 4 increasing. JobKeeper and JobSeeker programmes were the most likely reason for the variation, as they had a large impact on the labour market, but no direct impact on the output gap. Despite the variation, the alternative methods still reflect restrictions to business ability to use capital at full capacity in the last 14 weeks of 2019-20.

Since MFP is derived residually, there is an inverse relationship to the changes to capital services.  The impacts on MFP as a consequence of the utilisation adjustments to capital were more prominent because MFP growth is small relative to capital services growth (Chart 2). For 2019-20 Methods 1 and 3 saw MFP change from published -0.6% to positive.  Method 1 (output gap) resulted in the largest change, increasing to 2.6%. 

For most of the time series, the overall MFP the economic narrative described by the four alternative methods has not changed compared to published market sector MFP. In years where differences are notable, measures generally converge in the following years. This is likely similar for 2019-20 results despite the fact that COVID-19 containment measures have made variation much wider and more difficult to interpret.

The four methods showed that on balance, it is unlikely that unmeasured utilisation changes distorted the official market sector productivity measures in the long run. Any distortions were temporary as the methods often converged towards the official index. A more significant change in utilisation rate has likely occurred in 2019-20, which reflects the economic impacts of the COVID-19 containment measures, as well as natural disasters. The longer term impact of COVID-19 on productivity will become clearer as future growth cycles develop.

Productivity measures therefore provide a robust framework for analysing trends in technology change and contributions to output growth, especially over growth cycles, or the long run, that are fairly resilient to utilisation distortions from economic shocks to the economy.  Therefore, it is recommended that analysis over productivity growth cycles is useful as a means to reduce such distortions.

[1] Moreover, estimates of the change in labour composition, due to education and experience can be separately identified.  Labour composition is an approximation for quality change in labour services. 
[2] For a more detailed discussion, see Chapter 19 of Australian System of National Accounts, Concepts Sources and Methods (Cat. 5216.0).
[3] A possible fifth method is to use KLEMS intermediate inputs energy index as a proxy for utilisation rate adjustment. However it would be one year shorter than the other scenarios. 
[4] A similar approach was used in early Treasury modelling such as NIF-10, where capacity utilisation is approximated by actual/potential output and was used to estimate gross investment in other buildings and structures. See The NIF-10 Model of the Australian Economy, page 34, Treasury (1981).
[5] It has been shown that the inverse of the trend unemployment rate and the National Australia Bank business utilisation rate move in a similar direction. See for example, the March 2019 edition of Business Insider.
[6] Both of the under-employment scenarios imply an adjusted employment rate of 1 minus the under-employment rate, where the adjusted employment rate becomes the indicator. 
[7] Both of the under-employment scenarios assume workers are willing to work longer hours in their main job as distinct from working outside their industry of main job.