RDP 2002-02: Australian Use of Information Technology and its Contribution to Growth 2. Previous Research
January 2002
- Download the Paper 187KB
Although there have been a number of active contributors to the literature in the US there have been comparatively few studies based on experiences in other countries. For the most part this reflects data availability; the US is one of only a few countries with sufficient data to perform the growth accounting exercises required by this type of project.[2]
Within the US, Oliner and Sichel (1994, 2000) have been prominent contributors to the literature. They find that the contribution of the use of information technology, i.e., computer hardware, software and communication equipment, to productivity growth grew rapidly in the second half of the 1990s. They also find that technological advances in the production of computer-related goods have contributed to the higher productivity growth rates witnessed in the second half of the decade. More specifically, they estimate that the use of information technology related goods (contributing 0.5 per cent) and the advances in the production of computers (contributing 0.25 per cent) accounted for around three-quarters of the 1 per cent increase in labour productivity growth between the first and second halves of the 1990s in the US.
Gordon (1999) reaches a very different conclusion. He argues that the production of computer hardware accounts for the entire increase in trend labour productivity between the first and second halves of the 1990s. This implies that the use of computers contributed nothing to the growth in trend productivity. Of the 1 per cent increase in measured labour productivity throughout the decade he attributes 0.7 per cent to cyclical factors and the remaining 0.3 per cent to increased productivity of computer producers. This implies trend productivity growth in the US has remained roughly constant throughout the 1990s outside the computer-related sectors.
In October 1999 the US national accounts were revised and this led Gordon to modify his findings. Nonetheless, Gordon (2000) reaches similar conclusions to his earlier paper. In addition he argues that information technology has not had as large an effect on output growth as did the wave of great innovations introduced around the turn of the century such as electricity and the internal combustion engine. He argues that computer demand has primarily risen as a consequence of lower prices, that much of the development of the Internet represents a substitution away from, or a duplication of, pre-existing activities, and that much of the investment in new technology simply represents a defence of market share by incumbents rather than a conscious effort on behalf of firms to increase investment more generally. Furthermore, comparing the periods 1972–1995 and 1995–1999, he argues that much of the rise in productivity seen in the second half of the 1990s reflects an unsustainable cyclical effect and is not a direct result of higher investment in computers. Specifically, he estimates that of the 1.4 per cent increase in productivity growth since 1972–1995, 0.6 represents a cyclical effect. He attributes the entire remainder to faster multifactor productivity growth in computer-related sectors and suggests that there has been no revival in productivity growth rates in other sectors.
Jorgenson and Stiroh (2000) do not use a growth-accounting framework in their analysis; nevertheless they reach a similar conclusion to Oliner and Sichel (1994, 2000). They find that computer, and more generally information technology, investment can account for a large portion of the productivity growth that spurred the accelerated output growth in the late 1990s. They argue that rapid progress in semiconductor technology made it possible for downstream industries to reduce prices, enabling households and firms to invest in high-tech assets, which in turn drove strong output growth. However, they maintain that, while the production of high-tech products is a driving force behind recent productivity growth within the high-tech sectors, it generally does not spillover into other industries. Finally, they highlight the uncertainty surrounding the sustainability of rapid progress in high-tech industries and recognise the effect that lower productivity growth in tech-producing sectors and slower capital accumulation by high-tech-using sectors would have on growth.
Other studies for the US have generated more divergent estimates of the contribution of computers to growth. Whelan (2000) estimates that the use of computer hardware contributed 0.8 per cent to output growth between 1996 and 1998. This is higher than others' estimates.[3] This is primarily due to a difference in measurement rather than concept. In particular, Whelan's measure of the capital stock is around one-third larger than the one used by Oliner and Sichel. This increases his estimate of the income share and in turn the contribution to growth of computer capital. Whelan's capital stock is larger because he does not allow for any loss of efficiency in the productive capital stock before retirement whereas Oliner and Sichel make this adjustment.
Kiley (1999) on the other hand concludes that computer hardware has consistently detracted from growth since the mid 1970s. This stems from his assumption that there is an installation cost to new investment that detracts from growth. He argues that investment in computers involves high installation costs and, consequently, a negative contribution of computers to growth. He estimates that in the steady state the contribution from computers will be around 0.5 per cent per year. Although there may be some significant costs involved when installing new types of software and hardware, it is arguable whether they are as large as Kiley suggests.
Generally, the literature from the US suggests that there are gains to be had from both the production and use of high-tech goods. Roughly speaking, of the 1 per cent increase in productivity growth between the first and second half of the 1990s, around half of the acceleration can be assigned to the use of IT-related goods and around a quarter to advances in the production of computers. That said, productivity developments tend to be restricted to the computer-related sectors. Evidence suggests that there has been little, or no, revival in the productivity growth rates of other sectors attributable to the accelerated growth in high-tech industries in the US economy.
Although the US has been the predominant source of research there has also been some work conducted on this issue in the UK. As an alternative to the traditional historical growth accounting framework, Bakhshi and Larsen (2001) develop a dynamic general equilibrium (DGE) model to distinguish between investment-specific and sector-neutral sources of labour productivity growth for the UK economy. This approach emphasises the importance of substitution effects. That is, that rapid technological progress in the production of a given good, say high-tech products, leads to falling prices and increased investment, i.e., substitution towards computers. Using their DGE model, they find that technological progress in high-tech industries may account for around 25 per cent of labour productivity growth in the long run. Although they identify the effect that progress in the production of high-tech goods may have on productivity growth, they make no reference to the contribution that the use of high-tech goods may have on productivity and consequently output growth.
In Australia there have been only a few investigations of the effect of computers on productivity growth. The Productivity Commission (Parham, Roberts and Sun 2001) investigates the role of information technology in the output and productivity growth of the Australian economy and compares it with the US experience. In Australia, the contribution of information technology to labour productivity growth began to accelerate around 1996. They also find that a large proportion of new investment has been in the form of IT capital. Thus, firms have generally been replacing older non-IT capital with IT capital. They provide evidence in support of the view that there are productivity gains to be had from the use of high-tech goods and not just from their production. The Productivity Commission's report also investigates the MFP gains from information technology use. They find that finance and insurance, and wholesale trade display a positive relationship between IT use and MFP growth. They cannot detect a consistent relationship for other industries.
Wilson (2000) tells the story that extensive policy reform over the last 15 years has led to what he calls the ‘first wave’ of productivity improvements for Australia and that the effect of new economy developments could allow us to benefit from a ‘second wave’ of productivity gains in years to come. He estimates that this ‘second wave’ could add anywhere between 0.5 and 0.8 per cent to Australia's annual productivity growth rate over the next ten years. This is attributed to increased investment in computer equipment and the growing use of e-commerce by business.
Toohey (2000) analyses the contribution that investment in information technology has made to Australia's productivity performance and benchmarks his findings to the US experience. His results suggest that although both the US and Australia experienced stronger output growth over the second half of the 1990s, information technology investment contributed 0.25 per cent per annum more to growth in the US than in Australia. This is due to the rapid build-up of computer hardware in the US relative to Australia. He finds that the US and Australia are remarkably similar in terms of IT's contributions to growth, and that, although labour accumulation was a stronger contributor to US economic growth than to Australia's, MFP growth was stronger in Australia in the second half of the 1990s than in the US.
Wilson (2000) and Toohey (2000) both present estimates based upon the ABS's published capital stock numbers. For reasons that will be explained below these data create a number of conceptual problems and so lead to problems interpreting or comparing the numbers with, for example, the US studies. This paper represents the first attempt to use the appropriate capital stock measures for Australian data.