Feature Article - Estimates of productivity in the Australian National Accounts
Measures of productivity growth are important in understanding long term improvements in Australia's living standards and changes in Australia's international competitiveness. This article discusses the recent release of the annual estimates of productivity in the Australian System of National Accounts, 2004-05 (cat. no. 5204.0) and announces a work program to extend and improve the ABS's productivity estimates.
The latest annual estimates of market sector productivity show declines across the three productivity measures for 2004-05. Labour productivity fell by 1.3%, capital productivity fell by 2.2% and multifactor productivity (MFP) fell by 1.7%(Footnote: The terms 'labour productivity' and 'capital productivity' are used throughout the article to refer to measures of output per unit of labour input and output per unit of capital input respectively. The terms are used for convenience, although it should be noted that the measures reflect the contribution of other factors of production to changes in output as well as that contributed by the specific factor of production.). The falls reflect relatively weaker growth in output, 2.2% for 2004-05, and strong growth in hours worked, 3.5%, and capital services 4.5%.
The latest productivity estimates from the national accounts confirm that 2003-04 is the latest productivity growth cycle peak. The latest complete productivity growth cycle is now from 1998-99 to 2003-04. Average MFP growth for the latest cycle was 1.0% per year, which is lower than the previous cycle average rate of 2.1% per year, and slightly lower than the long term average (1964-65 to 2003-04) of 1.2% per year.
The section 'Productivity measurement' provides some background to the concepts of productivity measurement. 'Annual productivity estimates' analyses the latest annual estimates and, 'Productivity cycles' examines productivity growth in the context of the productivity growth cycles. Finally, the last section provides an outline of the planned work program for productivity measurement.
The primary aim of productivity analysis is to understand the drivers of growth in output. At a basic level, productivity growth occurs when the volume of output rises faster than the volume of inputs but beyond this simple notion a range of conceptual and measurement issues arise. In the first instance, productivity growth can be defined in relation to a single input (e.g. labour) or to a combination of inputs (e.g. labour and capital). Also, output growth might be defined in relation to total sales or it might be defined as growth in value-added, that is, output less intermediate costs. This section introduces these issues.
The ABS currently publishes a number of productivity estimates for the market sector of the economy:
The market sector of the economy consists of the following industries: Agriculture, forestry & fishing, Mining, Manufacturing, Electricity, gas & water, Construction, Wholesale trade, Retail trade, Accommodation, cafes & restaurants, Transport & storage, Communication, Finance & insurance, and Cultural & recreational services.
- labour productivity (LP), estimated by dividing an index of the volume of value added (VA) by an index of labour input (L), i.e., LP = VA / L
- capital productivity (KP), estimated by dividing an index of the volume of value added by an index of capital inputs (K), i.e., KP = VA / K
- multifactor productivity (MFP), estimated by dividing an index of the volume of value added by a combined index of labour and capital inputs, i.e., MFP = VA / (L+K).
All three productivity measures for the market sector are available from 1964-65 to 2004-05. Labour and capital productivity measures are partial productivity measures and provide a measure of growth in output to growth in one of the factor inputs, either capital or labour. The most obvious limitation of labour and capital productivity measures is that they attribute to one factor of production - labour or capital - changes in efficiency attributable to all factors of production. For example, the effects of technical progress, improvements in management practices and economies of scale, could affect measures of both capital and labour productivity. Further, a labour productivity measure takes no account of the amount of capital available to labour, or how this amount changes over time. Changes in the amount of capital available to labour is related to the concept of capital deepening, which is discussed further in the section 'Annual productivity estimates'.
The limitation of partial productivity measures gives rise to the development of MFP, which is a more comprehensive productivity measure. MFP measures the ratio of growth in output to growth in two or more factor inputs and represents that part of the change in output that cannot be explained by changes in the inputs. The ABS approach to measuring MFP is based on neoclassical economic theory using a translog production function in conjunction with two assumptions, constant returns to scale and the marginal products of capital and labour being equal to their respective real market prices. The estimates are derived by dividing the volume measure of gross value added by a combined labour and capital volume index. Value added is defined as gross output less intermediate inputs (materials, energy, business services etc. used up in the production process).
The labour input measure is hours worked. The hours worked estimates are derived as the product of employment and average hours worked per employed person. A shortcoming of the standard hours worked measure is that it does not take into account changes in the quality of labour. When compiling productivity estimates, increases in aggregate labour quality, such as increased experience and higher qualifications, will not be reflected as increases in labour inputs, but will be reflected in MFP. The ABS has developed and published an experimental annual quality-adjusted series for hours worked, and related productivity measures, going back to 1983-84. The method used to compile the index is described in more detail in Quality-adjusted Labour Inputs (Reilly, Milne and Zhao 2005, cat. no. 1351.0.55.010).
The other input used in measuring MFP is capital services. Capital services are a flow measure based on the productive capacity of capital. The capital services produced by an asset over its life are directly proportional to the productive capital value of the asset. The productive capital value, or stock, estimates are derived on the basis of an asset's pattern of decline in efficiency due to age. There are many possibilities for defining the 'age-efficiency profile', but a lack of data makes it difficult to determine the precise profile of each asset and hence, general assumptions are used. The ABS uses hyperbolic functions to describe the age-efficiency profiles. A hyperbolic function means that the efficiency of an asset declines by a small amount at first and the rate of decline increases as the asset ages. Age-efficiency profiles are distinct from, but related to, rates of financial depreciation of assets. That is, the economic efficiency of an asset will not necessarily decline at the same rate as that of an asset's financial depreciation.
The productive capital stock is estimated by applying a perpetual inventory method (PIM) to volume estimates of investment (gross fixed capital formation) at a detailed level in conjunction with age-efficiency profiles. Briefly, for each asset type the PIM adds the current year's investment to previous year's investments, which are multiplied by suitable scalars(Footnote: The scalars take into account the age-efficiency profile and retirement distribution patterns. The scalar in the year the investment is made is equal to one. To represent the decline in the capital service that occurs over time the scalar in subsequent years is less than one. The scalars are different for each asset and change each year of the asset's life.), to give a productive capital stock for each asset. An index of capital services is defined by aggregating the changes in the productive capital stock for each asset using as weights, the rental price of each asset type.
MFP measurement requires combining the labour and capital inputs into a single index. This combined input index takes the separate labour and capital input measures just described and weights them together using the labour and capital shares of total market sector income.
Total market sector income is the sum, for market sector industries, of the gross operating surplus (GOS) of corporations and government, gross mixed income (GMI), compensation of employees (COE), and taxes less subsidies on production and imports. Gross mixed income and taxes include both capital and labour components, and these are split in order to obtain capital and labour income shares. Thus capital's income share is given by the sum of GOS, proprietor's capital income (capital's share of GMI), and net taxes on production levied on capital divided by total income. Labour's income share is then the residual, but more precisely it is the sum of COE, proprietor's labour income (labour's share of GMI), and net taxes on production levied on labour divided by total income.
For more detail about these conceptual and measurement issues readers should refer to the Australian National Accounts: Concepts, Sources and Methods (cat. no. 5216.0) available on the ABS web site.
ANNUAL PRODUCTIVITY ESTIMATES
In 2004-05, labour productivity for the market sector recorded its first fall since 1984-85, falling by 1.3%, reflecting a 3.5% increase in the hours worked measure of labour input, against a 2.2% increase in market sector output (GDP). The quality adjusted labour productivity fell by 1.5%, suggesting an improvement in the composition of labour in the labour force, as well as an increase in the number of hours worked. Figure 1 shows labour productivity on both an hours worked basis and a quality adjusted hours worked basis.
Since 1983-84, the volume of labour input has grown by 1.3% per year on an hours worked basis, and by 1.7% per year on a quality adjusted hours worked basis (See Figure 2). This reflects a 0.4% per year improvement in the quality of labour over the period, measured in terms of educational attainment and work experience. The faster growth in the quality adjusted input measure is reflected in the lower growth in quality adjusted productivity measures, relative to unadjusted estimates.
Figure 1. Labour productivity, (2003-04 = 100.0)
Multifactor productivity recorded a fall in 2004-05 of 1.7%. In addition to the strong growth in labour inputs relative to output, there was a strong increase capital services in 2004-05 of 4.5%. Total combined inputs (capital services and labour) grew by 4.0% compared to output growth of 2.2%. Figure 3 shows the changes in multifactor productivity.
Figure 2. Labour inputs, (2003-04 = 100.0)
MFP and labour productivity indexes exhibit similar patterns (Figures 1 and 3). This is mainly due to capital services exhibiting smoother growth compared to labour inputs. This generally smoother pattern of growth in capital services is reflected in the PIM which models how an owner receives capital services from an asset. It is not the PIM that smooths capital services it is the fact that in any one period the new additions to the capital stock are only a small proportion of the total capital stock available to the economy.
Figure 3. Multifactor productivity, (2003-04 = 100.0)
A common method of examining changes in productivity over an extended period involves identifying and dividing the data into productivity 'growth cycles'. The reason for this relates to the natural variations in productivity growth present within the business cycle and how these relate to measurement issues, such as the ability to capture capacity utilisation within the input statistics. This means that year to year changes in estimates may not be truly indicative of a change in productivity. By analysing average productivity statistics between growth cycle peaks, the effects of some of these influences can be minimised, and therefore allow better analysis of the drivers of growth in different periods. In addition to the annual estimates published in Australian System of National Accounts (cat. no. 5204.0), estimates for identified productivity growth cycles are also presented.
Productivity growth cycle peaks are determined by comparing the original MFP estimates with their corresponding long-term trend estimates(Footnote: The long-term trend estimates are calculated using an 11-term Henderson moving average of the original, annual indexes.). The peak deviations between these two series are the primary indicators of a growth-cycle peak, although the more general economic conditions at the time are also considered.
Figure 4 below shows the deviations from the long-term trend. Growth cycle peaks have been identified in 1964-65,1968-69, 1973-74, 1981-82, 1984-85, 1988-89, 1993-94, 1998-99 and 2003-04. Whilst some of the peak deviations, and thus the cycle, are easy to identify (for example the 1998-99 and 2003-04 peaks), others such as the 1976-77, 1978-79 and 1993-94 deviations require the consideration of other economic information such as information on employment, prices and general economic circumstances in determining the existence or otherwise of a peak.
Figure 4. Difference between original and long-term trend
Figure 5 provides an indication of the longer term changes in productivity growth over the last 40 years. It is worth highlighting the cycle from 1993-94 to 1998-99, as it represents the largest average increase in MFP in the series. This was the result of strong growth in market sector output outweighing the growth in observed inputs (labour and capital). In contrast to this cycle is the 1984-85 to 1988-89 cycle, where similar growth in market sector output was recorded, but was offset by relatively larger increases in inputs, particularly labour, and hence lower MFP growth.
Figure 5. Multifactor productivity cycles, (2003-04 = 100.0)
Contributions to output growth
Table 1 shows the (compound) annual percentage changes in the productivity indexes and the components of the indexes for the last four productivity growth cycles. This table highlights the high rates of productivity growth experienced in the mid to late 1990s, and that the most recent growth cycle has seen labour productivity and MFP growth rates return to long term average levels. Additionally, the strong growth in capital services relative to labour inputs has continued in the most recent cycle. However, growth in both output and input measures were marginally below long-term averages.
Figure 6 shows the contributions of capital services, labour and MFP to growth in market-sector output across the productivity growth cycles identified since 1964-65. The size of the contribution of capital services is approximately consistent throughout each cycle, particularly compared with the contribution of labour (hours worked) and to a lesser extent, MFP.
The graph also highlights that the fall in average output growth of 1.5 percentage points between the last two cycles, from 4.6% per year to 3.1% per year, was largely attributable to the fall in MFP growth, from 2.1% per year to 1.0% per year (table 2). As a consequence, the contribution of MFP to output growth fell from 45% to 32%.
Figure 6. Contributions to output growth productivity growth cycles, Per cent per year (a)
Table 2. Contributions to output growth, Selected periods(a)
1984-85 to 1988-89
1988-89 to 1993-94
1993-94 to 1998-99
1998-99 to 2003-04
% per year
% per year
% per year
% per year
|Output growth |
|Contributions to output growth |
|Capital services (b) |
|Hours worked(b) |
|Multifactor productivity |
|(a) Compound annual average rates of growth. |
|(b) Growth in input index multiplied by its respective income share. |
Contribution of investment in information technology
One method of analysing the contribution of investment in computer hardware and software (information technology (IT)) to labour productivity growth is to use the growth accounting identity that labour productivity growth is the sum of MFP and the contribution of capital deepening. Capital deepening means that, on average, each unit of labour has more capital to work with to produce output. The contribution of capital deepening to labour productivity growth equals the change in the capital/labour ratio (capital intensity) multiplied by capital's share of total payments for inputs. By separately identifying IT capital and its income, an estimate of the contribution of IT capital deepening to labour productivity growth can be made. The capital deepening effect from IT derives from the increases in the IT-intensity of production. For example, relatively cheaper IT would induce businesses to substitute IT capital for other types of capital and labour. The owners of IT assets capture the benefits in income flows (reflected in the IT capital income shares).
Table 3 provides details on the contributions of IT, other capital and MFP to labour productivity growth. The table shows total capital deepening being relatively constant across the productivity cycles with the exception of 1984-85 to 1988-89. Decomposing total capital into IT and other capital shows IT capital's contribution increasing significantly over the period 1964-65 to 2003-04. Comparing the latest two cycles, 1993-94 to 1998-99 and 1998-99 to 2003-04, there is only a moderate increase in IT capital deepening, from 0.6% to 0.7%. The dark shaded area in Figure 7 highlights the increasing importance of IT capital to labour productivity growth. IT capital's contribution has increased from 3% of labour productivity growth over the period 1964-65 to 1968-69 to 31% of labour productivity growth for the latest cycle, 1998-99 to 2003-04. A qualification of this analysis is that it captures only the direct impact of IT investment on labour productivity. Indirect impacts, such as improved business processes as the result of IT investments, are not reflected and remain as part of the MFP residual.
Table 3: Contributions to labour productivity growth for the market sector(a)
1964-65 to 1968-69
1968-69 to 1973-74
1973-74 to 1981-82
1981-82 to 1984-85
1984-85 to 1988-89
1988-89 to 1993-94
1993-94 to 1998-99
1998-99 to 2003-04
|Labour productivity |
|Capital intensity |
|Information technology(b) |
|All other capital services(b) |
|Multifactor productivity |
|(a) Compound annual average rates of growth. |
|(b) Growth in capital-labour ratio multiplied by capital's share of income. |Figure 7. Contributions to labour productivity growth for the market sector, Per cent per year (a)
The ABS as part of its ongoing work to improve and enhance productivity measures is embarking on a series of new projects. The projects outlined here aim to both improve the measures shown in table 1, such as labour productivity, multifactor productivity, labour and capital inputs, and to measure them at the industry level. This section outlines six possible projects on productivity for the ABS, including:
Estimating industry MFP estimates for market sector industries is a high priority for the ABS. To derive MFP, estimates of output, labour and capital at the industry level are required. All of these are currently available through the annual supply and use tables, labour force surveys and capital stock system. Many of the issues discussed below apply to both aggregate (market sector) and industry level estimates. The ABS recently published experimental estimates of industry level MFP (Zheng 2005, cat. no. 1351.0.55.004) and this work will form the basis for future development.
- measuring MFP at the industry level
- improving labour input measures
- improving capital service input measures
- improving non-market sector output measures
- measuring MFP for the Property & business services industry
- developing a productivity measurement database.
Improving labour input measures
Currently, the output and labour input measures used for estimating labour productivity and MFP come from different sources. That is, output measures are based on business survey measures and labour input (hours worked) is from the Labour Force Survey (LFS), a household survey measure. The two surveys may record employment against different industries causing distortions in the industry productivity measures. Another measurement issue to be considered is the levels of hours worked. At present the Labour Force Survey records hours worked data based on a reference week. Resolving these issues would lead to greater accuracy in industry and market sector productivity measures.
Improving capital service input measures
There are many theoretical assumptions underlying the capital services measure, and much debate surrounding their construction. There are a number of measurement issues for improving capital services to consider, such as the assumptions made about capital assets, exogenous versus endogenous rates of return, depreciation rates, changes to the asset boundary, and capacity utilisation.
Improving non-market sector output measures
Historically, output for the non-market industries has been measured using deflated inputs, which are not suitable for measuring productivity. Direct estimates of output volumes for Health and Education were introduced into the National Accounts in 2001 but there remain issues in measuring quality change. The ABS has also investigated direct output volume estimates for other non-market service industries relating to the justice sector, social security and tax administration. Further work in this area is likely to focus on improving measures on Health and Education output in the first instance. The ABS has no plans to incorporate Health and Education into its MFP estimates at this stage.
MFP for Property & business services
At present, market sector MFP estimates exclude the Property & business services industry. This industry is large (second in size after Manufacturing), and its inclusion in the market sector would enhance current productivity estimates. Given its size, it is likely to have an influence on aggregate productivity levels and growth rates. Data are available for its inclusion, however they are only available from 1995-96.
Productivity measurement database
The database would act as a warehouse containing the raw information used in constructing MFP measures. This would be useful for researchers in analysing productivity from their particular perspectives. As much of the data are already available or will be developed, it would be a matter of determining how the database should be constructed, what exactly should be included on the database, and how access to the database should be managed.
SUMMARY AND CONCLUSIONS
Productivity measures are essential in understanding Australia's economic performance. This article gives a brief introduction to the measurement concepts and issues and has highlighted how productivity data can be used to better analyse and understand patterns in economic activity. The future development of industry level estimates of MFP will further assist in analysing the changes in economic growth.
For further information regarding the contents of this paper, or the ABS's work on productivity statistics, please contact Paul Roberts in Canberra on (02) 6252 5360 or firstname.lastname@example.org.
Australian Bureau of Statistics (2000), Australian System of National Accounts: Concepts, Sources and Methods, (cat. no. 5216.0), Canberra.
Australian Bureau of Statistics (2005), Australian System of National Accounts, 2004-05, (cat. no. 5204.0), Canberra.
OECD (2001), OECD Productivity Manual: A Guide to the Measurement of Industry-level and Aggregate Productivity Growth, OECD, Paris.
Reilly, R. Milne, W, and Zhao, S (2005), Quality-adjusted Labour Inputs, Research Paper, (cat. no. 1351.0.55.010), ABS, Canberra.
Zheng, S. (2005), Estimating Industry Level Multifactor Productivity: Methods and Experimental Results, Research paper, (cat. no. 1351.055.004), ABS, Canberra.