5260.0.55.002 - Estimates of Industry Multifactor Productivity, 2015-16  
Previous ISSUE Released at 11:30 AM (CANBERRA TIME) 05/12/2016   
   Page tools: Print Print Page Print all pages in this productPrint All RSS Feed RSS Bookmark and Share Search this Product

This release includes two additional experimental tables (Tables 25 and 26). Table 25 presents year on year contributions to growth by industry. Table 26 presents contributions to growth over industry-specific growth cycles. This article describes the methodology adopted to develop the growth cycles, and discusses some of the results.

BACKGROUND

Multifactor productivity (MFP) is widely used as an indicator of technological change. In the short to medium term, MFP estimates are subject to data limitations and assumptions, such as variations in capacity utilisation, economies of scale and scope, reallocation effects of capital and labour, and measurement error.

Variations in the utilisation of inputs would ideally be measured as changes in inputs when MFP is calculated. However, due to current data limitations, adjusting capital service flows for variation in utilisation is not possible.

Taking into account these factors, MFP estimates are useful when viewed as average growth rates between growth-cycle peaks as this assumes a constant rate of capacity utilisation. Growth cycles are chosen with reference to peak deviations which are determined by comparing MFP estimates with their corresponding long-term trend (Footnote 1). The peak deviation between these two series is the primary indicator of a growth-cycle peak. General economic conditions at the time are also considered. In this way, most of the effects of variations in capacity utilisation and much of the random error is removed. However, average growth rates may still reflect any systematic bias resulting from the methodology and data used.

Currently, the ABS publishes market sector MFP growth cycles, both in the Australian System of National Accounts (cat. no. 5204.0) and in Estimates of Industry Multifactor Productivity (cat. no. 5260.0.55.002).

While the market sector growth cycles are useful when analysing the performance of the aggregate market sector, they are not appropriate when analysing individual industry productivity as the factors that cause cyclical fluctuation vary across industries. Barnes (2011) found significant variation between industry growth cycles and the market sector. To provide a better understanding of productivity performance and underlying drivers at the industry level, individual industry productivity growth cycles are required.

The ABS has extended the estimation of growth cycles to the 16 industries that comprise the market sector (Divisions A to N, plus Divisions R and S) on an experimental basis. In addition, the approach used to identify growth cycle peaks has been strengthened to ensure the robustness of peak identification.


THE METHODOLOGY OF IDENTIFYING INDUSTRY GROWTH CYCLES

Growth cycles at the industry level are determined by comparing the original MFP estimates with their corresponding long-term trend estimates. However, to ensure that peaks are resilient to the anticipated revisions to MFP due to data source revisions, a multiple filter approach has been adopted at the industry level.

The purpose of using a multiple filter approach at the industry level is to strengthen the criteria for data that is inherently more volatile, and mitigates the risk of a growth cycle peak being subsequently revised. Common filters used to extract trends from economic series are adopted to estimate the industry growth cycles: Henderson 11, H (11), Hodrick and Prescott, HP, (1997), and Christiano and Fitzgerald, CF, (2003).

The criteria used to identify industry growth cycles are as follows:

  • A peak is considered robust if it is identified by all three filters, and shows deviations equal to, or greater than one percentage point (Footnote 2).

Where identified robust peaks were found less than four years apart (peaks inclusive), additional rules were required to obtain growth cycles of a reasonable length:
  • Choose the peak with the relatively largest deviation
  • If the difference in deviation is negligible, choose the peak which produces the longer cycle
  • If the two adjacent peaks have a similar deviation size and suggest a similar cycle length, refer to macro-economic conditions
  • Consider the ‘nearly’ robust peaks (i.e., suggested by the three filters but with a deviation of less than one percentage point) next to neighbouring troughs (such as the global financial crisis)

The Augmented Dickey-Fuller (ADF) test is adopted to provide evidence on whether a series has a deterministic trend or unit root and, thus, the order of integration of the series. Results from the ADF tests vary across industries. For consistency, the ABS has determined and produced growth cycles for all 16 market sector industries (Footnote 3).

Growth cycles are also available for industries that have long periods of decline in MFP growth. Peaks in this context still represent a deviation from a (declining) trend, and thus indicate where an industry has halted the decline in productivity for a short period. This phenomenon can be seen in Mining and Electricity, gas, water and waste services post 2000-01, and Rental, hiring and real estate services from 1995-96 to 2009-10.

Figure 1 illustrates how the multiple filter approach is implemented. The figure shows the deviations from the long-term trend for Manufacturing MFP. Some of the peak deviations, and thus the cycle, are easy to identify (e.g. for 1993-94 and 2011-12). Others such as the 2007-08 peak require some judgment by considering the prevailing economic conditions at the time, and whether the set of cycles are of reasonable length.
Figure 1. Difference between original and long-term trend (Manufacturing)

Figure 1. Difference between original and long-term trend (Manufacturing)

Figure 1. Difference between original and long-term trend (Manufacturing)

Notes: H (11): Henderson term 11, HP: Hodrick and Prescott, CF: Christiano and Fitzgerald.


INTERPRETING INDUSTRY PRODUCTIVITY GROWTH CYCLES

This section discusses results by industry, and highlights some of the factors that should be taken into account when interpreting industry growth cycles.

While year-on-year Manufacturing MFP is volatile, comparing productivity between the peak years provides an indication of the longer term changes in productivity growth over the past 25 years. Figure 1 shows that MFP for Manufacturing displays an upward trend until 2003-04 before falling away. Figure 2 shows that average growth rates across the cycles indicate that MFP in the sector declined from 2003-04 to 2007-08, and was positive between 2007-08 and 2011-12.

Figure 2 shows that, over the four cycles, output growth has declined, turning negative in the last complete growth cycle. In the 2003-04 to 2007-08 cycle, negative MFP growth contributed to weakened output growth. However, in the 2007-08 to 2011-12 cycle, output growth declined 1.1% and hours worked growth contributed -1.7%.

Manufacturing is one of only two industries (the other industry is Information, media and telecommunications) where growth cycles coincide with the market sector cycles, 1998-99 to 2003-04 and 2003-04 to 2007-08.

Figure 2: Contribution to Growth - Manufacturing

Figure 2: Contribution to Growth - Manufacturing


Figure 3: Contribution to Growth - Agriculture, Forestry and Fishing

Figure 3: Contribution to Growth - Agriculture, Forestry and Fishing



Agriculture MFP is largely driven by major weather events. Years that have optimal farming conditions more often than not see high productivity growth and thus determine the length of productivity growth cycles. Average growth rates between peaks will depend on weather conditions, which results in the shorter, more frequent growth cycles shown in Figure 3.

Mining is an industry that has seen significant Gross fixed capital formation (from 2003-04). Figure 4 shows that a build up in capital stock has contributed to negative MFP recorded in the latest two complete growth cycles. Recent data indicate that this period of negative MFP growth has ended.

Conversely, Information, media and telecommunications shows declining contributions from capital services over each of its four cycles, and varying contributions from hours worked and MFP. Output growth, however, has been declining since 1994-95.
Figure 4: Contribution to Growth – Mining

Figure 4: Contribution to Growth – Mining


Figure 5: Contribution to Growth - Information, Media & Telecommunications

Figure 5: Contribution to Growth -  Information, Media & Telecommunications



Currently, the most common number of productivity cycles in an industry is four, however there are some divergences. Accommodation and food services currently has five productivity growth cycles, reflecting the demand driven nature of this industry. The contribution of capital services and hours worked are positive for each of the productivity cycles, resulting in a fall in MFP in years with low output growth.

Figure 6: Contribution to Growth - Accommodation and Food Services

Figure 6: Contribution to Growth -  Accommodation and Food Services


Figure 7: Contribution to Growth - Professional, Scientific & Technical Services

Figure 7: Contribution to Growth -  Professional, Scientific & Technical Services



Conversely, there have been only three complete cycles in the Professional, scientific and technical services industry. The length of the first two cycles reflects other industry’s cycle length of 3 to 4 years. The last complete cycle has a longer length of eleven years, reflecting the slow recovery of this industry’s productivity growth following the low point during the global financial crisis.

Overall, the industry growth cycle results show that different industries can have very different productivity growth cycles. Care needs to be taken when interpreting year to year industry productivity growth, or applying a standardised averaging method across all industries.

REFERENCES

Barnes, P. (2011) “Multifactor Productivity Growth Cycles at the Industry Level”, Productivity Commission Staff Working Paper, Canberra.
Christiano, L. J., and T. J. Fitzgerald. (2003) “The band pass filter”, International Economic Review 44: 435-465.
Dickey, D. and W. Fuller (1979), “Distributions of the estimators for autoregressive time series with a unit root”, Journal of the American Statistical Association 74, 427-431.
Henderson, R. (1916), “Note on Graduation by Adjusted Average”. Transactions of the American Society of Actuaries, 17, 43-48
Hodrick, R.J. and Prescott, E.C. (1997) “Postwar U.S. Business Cycles: An Empirical Investigation”, Journal of Money, Credit and Banking, 29, pp. 1–16.

INQUIRIES

For further information about these and related statistics, please contact the Director of Productivity Statistics on 02 6252 5860, email national.accounts@abs.gov.au, or phone the National Information and Referral Service on 1300 135 070.

FOOTNOTES

1 The trend is estimated using a Henderson 11-term moving average filter. Back
2 Peaks suggested by the three filters but have deviations slightly less than 1 percentage point are considered if they were found more than four years apart. Back
3 When including the requirement of an underlying trend to their criteria, Barnes (2011) did not calculate growth cycles for Arts and recreation services. Using the Augmented Dickey-Fuller test, ABS confirms Barnes (2011) finding of no evidence of trend in Arts and recreation services industry. Back