RDP 2002-02: Australian Use of Information Technology and its Contribution to Growth 5. Results

With the appropriate data assembled it is a simple matter to calculate income shares, growth rates and the residual, multifactor productivity. In line with previous practice the income-share weights used in the growth-accounting exercise are averages of the income shares in the two years the growth is measured over.

Thus, computer growth from 1990/91 to 1991/92 is weighted by the average income share of computers in these two years. This is, in part, an allowance for the fact that calculations are made at discrete intervals whereas the derivation is properly applied to continuous data.

To aid the exposition we begin by discussing the aggregate economy-wide results before moving on to the more detailed industry-level estimates. The aggregate results are generated by weighting the industry-level results together by their respective output shares. Specifically, assuming that there are two industries indexed by 1 and 2, total output is given by Y = Y1+Y2 and all other terms are the same as in Equation (9), the overall contribution of capital to output growth can be calculated as:

5.1 Aggregate Results

It is common practice to average results over a run of years due to the substantial variations in productivity that can occur in any given year. These variations are generally related to cyclical forces so the preferred period covers an entire productivity cycle; this way the cyclical forces should ‘average out’. Most US studies have presented results averaged over the first and second halves of the 90s. As Parham et al (2001) argue, this choice can substantially distort the findings, as the periods involved are not complete cycles. Thus, we present results in this section for the period 1993/94–1999/2000, the latest productivity cycle as defined by the ABS. We do, however, present results using the ‘traditional’ split as well. Despite problems, this split still provides a useful picture of how IT investment has changed over the 1990s.

Table 1 presents the results for output growth and the various contributions to growth. It also presents memo items of the growth rate of the computer stock and other inputs. The time periods average the growth contributions calculated for each (financial) year in the sample. Thus, the first column presents the average of growth for the 1993/94 financial years through to the 1999/2000 financial years – as such it is based on levels data from 1992/93 through 1999/2000.

Table 1: Contributions to Growth
  1993/94–1999/2000 1990/91–1994/95 1995/96–2000/01
Output growth 4.51 1.78 3.86
Contributions from:
IT capital 1.13 0.89 1.26
Hardware 0.70 0.42 0.84
Software 0.43 0.47 0.42
Other capital 0.62 0.15 0.60
Labour hours 1.09 −0.37 0.57
MFP 1.67 1.11 1.43
Income shares:
Hardware 2.3 1.8 2.4
Software 2.5 2.4 2.5
Other capital 34.2 34.8 34.0
Labour 60.9 61.1 61.1
Growth rate of inputs:
Hardware 36.7 22.0 36.9
Software 15.5 18.8 17.1
Other capital 1.1 0.2 1.0
Labour 1.5 −0.8 0.7

Note: All numbers are expressed as percentages per annum.

Generally speaking, we see that Australia has experienced extremely high levels of multifactor productivity growth in addition to significant gains from the ‘capital-deepening’ effects from computer technology. While MFP growth was the largest single contributor to growth between 1993/94 and 1999/2000, capital deepening attributable to computer use added over 1 per cent per annum to output growth.

Looking at the change between the first and second halves of the decade we see that MFP and labour use have been the primary sources of the acceleration in output growth. This partially reflects the fact that the first half of the sample includes a recession that was associated with labour reductions and lower rates of productivity growth. IT investment has been relatively steady over the decade accounting for only 0.37 per cent of the 2.08 per cent pickup in growth across the decade. That being said, the contribution of computer capital to growth has been at a very high level and contributes so little to the pickup because it was relatively unaffected by the early 1990s recession. What is also clear is the amazing growth in the stock of hardware and software. This has been sustained by both increasing nominal expenditure and large price falls fuelling an increase in real inputs. We will return to the split between these two forces in Section 6.

5.2 Robustness Checks

As mentioned above, the differing aggregation assumptions may lead to different estimates of MFP and the contribution to growth of computers. Table 2 compares the results using Equation (9), as already reported in Table 1, with those from using Equation (12), the ABS method.

Table 2: Comparison of Aggregation Methods
  1993/94–1999/00   1990/91–1994/95   1995/96–2000/01
This paper
(Table 1)
Equation
(12)
This paper
(Table 1)
Equation
(12)
This paper
(Table 1)
Equation
(12)
Output growth 4.51 4.70   1.78 2.07   3.86 4.04
Contributions from:
IT capital 1.13 1.24   0.89 0.80   1.26 1.42
Hardware 0.70 0.77   0.42 0.35   0.84 0.88
Software 0.43 0.47   0.47 0.45   0.42 0.54
Other capital 0.62 0.40   0.15 0.29   0.60 0.35
Labour hours 1.09 1.20   −0.37 −0.07   0.56 0.77
MFP 1.67 1.86   1.11 1.05   1.44 1.50
Income shares:
Hardware 2.3 2.3   1.8 1.8   2.4 2.4
Software 2.5 2.6   2.4 2.4   2.5 2.5
Other capital 34.3 36.7   34.7 37.0   34.0 36.6
Labour 60.9 58.4   61.1 58.8   61.1 58.5
Growth rate of inputs:
Hardware 36.7 32.1   22.0 19.3   36.9 36.2
Software 15.5 18.9   18.8 19.2   17.1 21.7
Other capital 1.1 4.4   0.2 2.7   1.0 4.7
Labour 1.5 1.7   −0.8 −0.1   0.7 1.0

Notes: All numbers are expressed as percentages per annum. Differences in output growth rates are due to the treatment of taxes.

What we see is that there are few substantial changes in the picture. While individual results vary (up to 0.3 per cent in the case of labour hours in the first half of the 1990s) the broad sweep of the results is unaffected. On this basis we are reasonably confident about the robustness of our results to the particular aggregation method chosen.

5.3 Industry Results

Rather than look at all industries, we will focus on some pertinent results from certain industries to highlight the main findings. A table of the results for all industries can be found in Appendix A. The contribution of computers to growth by sector is presented in Figure 4 below.

Figure 4: Computer Contribution to Growth
Figure 4: Computer Contribution to Growth

Note: Contribution to growth measures the growth in output attributable to the growth of computer hardware and software capital as identified in Equation (8).

Looking at the broad patterns we see that traditional primary and secondary industries have not been significant beneficiaries from the IT revolution. That is, they have not incorporated high levels of IT capital directly into their production techniques. These results are more or less similar to the pattern of investment shown in Figure 3. The principal beneficiaries have been the service sectors, in particular communications and finance. This bias towards the service sectors is interesting. In the US most of the gains in productivity, and by implication the gains from computers, have been concentrated in manufacturing. This reflects the high level of computer and computer-related production in the US. Further evidence referred to by the Productivity Commission (Parham et al 2001) suggests that the gains from use have been concentrated in wholesale and retail trade, and finance, insurance and real estate. This suggests some overlap between the Australian and US experiences. At the same time, the results for electricity, gas and water (EGW) and communications highlight some results that are peculiar to Australia. These sectors have been deregulated as a part of the microeconomic reform undertaken over the 1980s and 1990s in Australia. The high levels of investment, particularly in computers, in these sectors would seem to be a product of this regulatory change.

Nonetheless, this pattern masks some important points. Examination of input-output tables for Australia allows us to identify the contribution of an input, say high-tech goods, to final output and the intensity of use of that input relative to others in the production process. This enables us to obtain a better indication of the effect on various industries from particular new economy goods and services. For instance, communication and business services incorporate much of the new economy. We can focus on the extent to which communication and business services are used as an input to production by industries to gain a better idea of the influence of high-tech capital on these industries. We do this by calculating the total requirement for inputs, including any intermediate use of goods and services, by industries as a percentage of final output for each industry. These ‘total requirement coefficients’ represent the dollar amount of an input required to produce $100 of final output for each industry.

Figure 5 shows that ‘old economy’ industries, while low direct users of IT, are high users of ‘new economy’ inputs. While it is difficult to separate IT services from other services, they are most likely to be concentrated in the communications and business services sectors. For example, web design firms are generally classified in the business services industry and Internet providers in the communications industry. Based on data for 1996/97, primary industries[12] used just under $3.00 of communications services and around $13.00 of business services for each $100 of output produced.[13] This usage is similar to that in manufacturing, EGW and construction. On average, service sectors used just under $4.00 of communications services and around $16.00 of business services to produce $100 of output. Thus, the gap in usage between primary and secondary sectors, and service sectors may not be as large as the results in Figure 4 suggest. Many primary sectors would seem to contract out their IT requirements.

Figure 5: Total Requirement Coefficients
Dollar amount of an input required to produce $100 of final output
Figure 5: Total Requirement Coefficients

This highlights a fact that is obvious but deserves mention. The heaviest users of computer technology in Australia provide services to other industries. Thus, while the growth-accounting procedure we use highlights high MFP growth and large gains from capital deepening through computer usage in the communications sector, other sectors benefit through falls in telecommunications costs.[14] Similarly, large gains in productivity in retailing through the use of bar coding and scanning have translated into lower prices at supermarkets and faster checkout times.

Turning to specific industry results, the electricity, gas and water (EGW) sector results are particularly interesting. This industry experienced a decline in multifactor productivity yet a marked increase in IT investment. Looking more closely reveals that high rates of MFP growth in the first half of the 1990s were a product of significant job cuts in the industry. Hours worked declined by around ⅓ over the decade with most reductions coming in the first half. Similarly, currently low levels of calculated MFP growth are a reflection of very high levels of investment in IT capital. If these investments have not been fully integrated in the production process, it may be a few years before output responds. Nonetheless, the predominant force in this industry over the decade has been the dramatic structural reorganisation culminating in the privatisation of much of the sector. Given this change it is difficult to draw any strong conclusions about the performance of the sector.

Communications was similarly deregulated. This deregulation has spurred investment by many new participants as well as by the incumbent. The output growth effects of this high investment, seen in Figure 4, have been sustained for the past decade. In contrast to EGW there has also been significant MFP growth in communications throughout the decade (Appendix A). We address the extent to which these MFP patterns can be attributed to IT investment in the next section.

Footnotes

Agriculture; forestry, fishing and hunting; and mining. [12]

See Appendix B for a table of total requirement coefficients for each industry. We use data for 1996/97, as they are the latest available. [13]

The deregulation of national phone companies is obviously a major factor in phone call costs. Nonetheless, the development of computerised switching devices, fibre optic cables and the internet have all combined to make communications much easier and cheaper than they were previously. [14]