RDP 2002-02: Australian Use of Information Technology and its Contribution to Growth 7. Discussion

Before considering the implications of these results it is worth reiterating the lack of precision in these estimates. Capital stock estimates are notoriously difficult to construct as are the estimates of income shares for computers. This combined with the treatment of tax in the underlying data mean that there is inevitably a range of error around the estimates. That being said, the estimates are the best available based on the best available data.

The estimates obtained in this paper are larger than previous estimates of the benefit from computer use both in Australia and in the US. Table 4 below compares the estimates from these previous studies and recalculates the results from Table 1 over the shorter sample period that has been used in previous studies.[17]

Table 4: Comparison of Results
  Australia(a)   US
This paper   Toohey Oliner and Sichel
1991–95 1996–99 1991–95 1996–99 1991–95 1996–99
Growth rate of output 1.78 4.47   2.47 4.35   2.75 4.82
Contribution from:
IT capital 0.89 1.20   0.54 0.74   0.57 1.10
Hardware 0.42 0.87   0.25 0.36   0.25 0.63
Software 0.47 0.33   0.21 0.24   0.25 0.32
Communications   0.08 0.13   0.07 0.15
Other capital 0.15 0.76   0.86 1.10   0.44 0.75
Labour −0.37 0.35   0.40 0.87   1.26 1.81
MFP 1.11 2.16   0.67 1.64   0.48 1.16
Income shares:
Hardware 1.8 2.4   1.3 1.4   1.4 1.8
Software 2.4 2.6   1.2 1.4   2.0 2.5
Growth rate of inputs:
Hardware 22.0 41.6   18.9 25.7   17.5 35.9
Software 18.8 13.9   17.8 17.4   13.1 13.0

Notes: (a) Dates used are financial year periods, e.g., 1991–95 indicates 1990/91–1994/95.
Results for Toohey were obtained from the author and include some results not originally published.

There are enough differences between the studies that one should be cautious about drawing too much out of the precise figures. Between the Australian results the sectoral coverage is different as are the underlying capital data. The US results are based upon aggregate economy-wide results whereas ours are based upon industry-level data. Additionally, there is a difference in the treatment of labour inputs. Oliner and Sichel's results for the US include an estimate of quality adjusted labour inputs – generated by looking at changes in education and experience of the labour force. The ABS is currently working on an experimental labour quality adjustment but full results are not yet available. The preliminary results suggest that over the period 1994/95 to 2000/01 the contribution to growth of quality improvements in labour was only 0.15 per cent per annum.[18] Due to the preliminary nature of these estimates no adjustment is included in Table 4. Consequently the MFP estimates for Australia include any ‘residual’ effect from improved labour quality. However, using the preliminary estimates as a guide, subtracting 0.15 per cent per annum from MFP growth would not change any substantive features of the comparison with the US.

Notwithstanding these points, this table highlights the difference between the experience of Australia and the US over the 1990s. While GDP growth rates in the market sector were broadly similar, the US achieved this expansion by employing more labour while Australia has seen little change in the contribution from labour. Instead, most of the gains have been made through MFP growth.

There has also been a slightly larger contribution from capital, and in particular computer capital, to growth in Australia. Given that most of the ‘new-economy’ innovations originated in the US, this seems a surprising result. It may be that Australia achieved higher computer-capital contributions to growth over the 1990s by gradually ‘catching-up’ to the US. But it is hard to be sure. There are enough differences in statistical treatment that it is difficult to make confident statements about the relative contributions of computer capital to growth in the two countries.

What is most remarkable in Australia is the rate of MFP growth. Results from Section 6.2 suggest that we cannot clearly attribute this to our higher use of computers. In the absence of other candidates we can only suggest that the microeconomic reform of the 1980s and 1990s has paid handsome productivity dividends in the late 1990s.

Our decomposition in Section 6.1 allows us to consider some implications for the future. Production methods are in the process of being reorganised towards greater use of computer inputs. This change is reflected in their increasing income share. To the extent that this is a transitional feature we would expect the income share of computers to stabilise at some time in the future. When this occurs, the rising income share will no longer make a contribution to growth. Our estimates suggest that this contribution may be up to 0.6 per cent per year.

The other half of our decomposition highlights the benefits from the rapid technological advances that have occurred. These have added about 5 per cent to output over the course of this decade. Should technological progress or price reductions slow then this contribution to growth would also fall. Nonetheless, despite recent weakness in the industry, there are few signs that the pace of technological improvement or price falls passed on to users are declining. This gives us reason to believe that computers will continue to make a significant contribution to output growth for the foreseeable future.


Wilson (2000) looks at the contributions to labour productivity growth so the results are not directly comparable. The results in Table 4 also differ slightly from those originally reported in Gruen (2001). This is a result of minor revisions to the method, error checking, and some changes in the data. None of the differences change the story in any significant way. [17]

See the feature article in the September quarter 2001 National Accounts, ABS Cat No 5206.0, p 15. [18]