While the ABS has been producing multifactor productivity (MFP) estimates for the market sector as a whole for many years, the estimates contained in this paper are the first estimates presented for individual industries within the market sector by the ABS. The estimates represent the results of a significant body of work undertaken within the ABS over the last five years. The paper provides an in-depth analysis of each industry, discussing the validity of the results as well as the data used in construction of the estimates. While the industry analyses indicate plausible results, there are some difficult conceptual and measurement issues at the industry level that are less problematic or do not exist at all at the total market sector level. Because of the difficult measurement issues involved, the ABS qualifies these initial estimates of industry productivity by labelling them as 'experimental'.
The first updated estimates are expected to be released in November 2007, shortly after the release of the annual publication - Australian System of National Accounts (cat. no. 5204.0). Aggregate market sector estimates will still be published in Australian System of National Accounts.
The ABS invites comments on the estimates from readers. After considering comments received on these initial estimates, the ABS will incorporate any necessary adjustments to the estimation of industry level productivity estimates.
This research has identified a number of areas of productivity measurement that require further work. These are summarised below. Although the ABS remains keen to extend and improve the quality of the estimates as quickly as possible, some of the areas flagged for future research are considered longer term projects. Measurement issues such as adjusting for quality change in service industry outputs have been the subject of discussions for many years by national accountants and price statisticians. Any resolutions of these and other outstanding measurement issues will flow through to quality improvements in productivity statistics.
For the most part, the methodology and data underlying the industry level MFP estimates are the same as those used to estimate annual market sector MFP currently published in cat. no. 5204.0, and the labour productivity estimates published in the quarterly national accounts. The main difference is that market sector MFP includes taxes less subsidies on products, whereas the industry level MFP estimates do not. To achieve consistency with aggregated industry estimates, a possible change might be to remove taxes less subsidies on products for the market sector. This would mean that market sector MFP would be presented in basic prices rather than in purchasers' prices (the current treatment). Any such changes would be discussed with users prior to implementation.
More research needs to be conducted on output volume estimates to gauge the possible impact on MFP estimates of changes to industry output composition over time. Likewise, the potential impacts on MFP of quality change that might not be fully captured in industry output volumes could be better assessed. While this applies to all industries, it is particularly relevant to some service industries where the nature of outputs have changed rapidly over time. Adding to this difficulty is the heterogeneous nature of the output for many industries. The measurement of output volumes for services is made particularly difficult by the intangible nature of most services, making quality change particularly difficult to quantify. The Communications industry is a particular example where the product mix produced within the industry has changed significantly between 1985-86 and 2005-06.
Further sensitivity analysis on various aspects of capital services measurement, specifically the assumptions surrounding capital estimation at the industry level is to be undertaken. Although considerable work of this nature has already been undertaken to produce the estimates in this paper, future work would extend this to look at other components of the user cost of capital equation such as the rate of depreciation and the effect of capital gains. Based on the outcome of research to date, further work is to be conducted on the use of a mix of asset price deflators and the CPI as the price deflator in the user cost equation, with the idea being to form a better ex ante measure of producer's asset inflation expectations across all industries. The other assumptions used to produce capital estimates, including depreciation and age-efficiency profiles could also be subject to further research.
One of the main methodological challenges is linked to the problem of negative rental prices. Despite the testing of various options for the rental price equation, no single approach eliminated the occurrence of negative rental prices in all cases. Possible solutions that will be assessed include replacing the current endogenous methodology for the rate of return with an exogenous rate of return e.g. the long term bond rate, or smoothing endogenous rates of return. Another possible area of research relates to volatility of asset prices and how this is best handled in the model.
At a specific industry level there are a number of areas for further research. For instance, in the Mining industry, the services of subsoil assets are not currently included in the capital services measure for Mining. Also, the age-efficiency profile for the mineral exploration asset could be reviewed (currently the assumption is made that mineral exploration assets yield constant capital services throughout their service life). Another area that requires further analysis is the assumption of fixed estimates of productive capital stock of land in the Agriculture, forestry & fishing industry.
The impact of capacity utilisation is another topic for further research. The neo-classical assumptions that underpin MFP estimation assume that capacity utilisation is either constant or at least at full capacity at the peak of the cycle. More investigation is needed into the potential impact of changes in capacity utilisation on MFP for some industries. Of the industries studied, only Accommodation, cafes & restaurants has an indicator which may show a change in capacity utilisation, through room occupancy data, but this measure is at best limited in its usability. Given the difficulties of adjusting for capacity utilisation, further investigations could be undertaken into the impacts of measuring MFP between cycle peaks at the individual industry level, and particularly for those industries that are capital intensive.
There are a number of areas for further investigation in relation to labour inputs. For instance, a Tornqvist index using compensation of employees to weight hours worked could be developed. The current hours worked index is a simple index, which has an implicit assumption that the marginal product of labour is the same across all industries.
Another area for further research is to adjust the industry hours worked estimates for changes in labour composition. When compiling productivity estimates, increases in aggregate labour composition, such as increased experience and higher qualifications, will not be reflected as increases in labour inputs, but will be reflected in MFP. An aggregate market sector adjustment has been developed and results are published in cat. no. 5204.0, but no industry level adjustments are possible at this stage.
Concurrent with measuring industry level MFP, the ABS will continue to estimate aggregate market sector MFP. This aggregate measure can be estimated directly or by summing the individual industry MFP estimates, but different results would be obtained. If the ABS is to publish industry level MFP estimates then the issue of consistency between the current method and any alternative set of results needs to be addressed since compiling the industry level estimates would naturally raise analytical questions as to their contributions to aggregate MFP growth. Future research will be undertaken to evaluate methods for estimating aggregate and industry level MFP for both value added and gross output measures.
Changes to the national accounts
The national accounts will be subject to a number of changes as a result of the future implementation of ANZSIC 2006 and the revision of the 1993 SNA. These new classifications and standards are expected to be first implemented in the annual national accounts for 2008-09 and the September quarter 2009 quarterly national accounts. These changes will also impact on the industry and total market sector MFP estimates at that time. A considerable amount of data development and analysis will be required to incorporate these changes in the MFP estimates.