RDP 2002-02: Australian Use of Information Technology and its Contribution to Growth 6. Two Interesting Questions

Having covered the basic results from the growth-accounting exercise we turn to two more interesting questions.

6.1 What is the Counterfactual?

In assessing the benefit from computer use we have not explicitly considered what the alternative to computer investment may have been. Thus, discussion about the net benefit from information technology is difficult. This section constructs estimates of the net benefit from computers by making a number of assumptions. As with all counterfactual exercises these assumptions are not the only ones that could be made, but seem reasonable to us. As such the results they generate should be considered ‘ballpark’ figures rather than precise estimates.

We focus on what would have happened if the computer industry had not been characterised by the rapid productivity increases that are at the heart of the industry's uniqueness. We look at what would have happened to output if the computer price deflator had remained unchanged rather than falling rapidly as, in fact, occurred. The falls in the deflator reflect both improvements in the power of computers and falls in their price but the exact division between these two is unimportant. Specifically, we assume the same nominal spending on computers as occurred but compute a new series for real spending by assuming that the computer price deflator was flat from 1989/90. If the industry production functions are Cobb-Douglas this calculation will, in fact, reflect what fully optimising firms would have done – income shares of inputs remain constant regardless of relative price fluctuations. Alternatively, if the relevant industry production functions are not unit elastic then this will be an approximation.

Another way of considering this exercise is that we decompose the contribution of computers to output growth into the parts attributable to price falls and those attributable to increased expenditure. We take 1989/90 as the base year for these calculations as it marks the beginning of our sample. Using this interpretation it is possible to view the counterfactual as the situation that would have resulted if the producers of information technology had retained all of the productivity gains and not passed any on to users in the form of lower prices. This is, perhaps, the more relevant view of this exercise for Australia. It allows us to identify the technological and output gains Australia has realised solely from being a user of information technology, i.e., by focusing on those improvements in computer manufacturing that have been passed on to users in the form of lower prices.

Combining the nominal investment series with the new price deflator gives us a series for real investment. With the new series for real investment we calculate a series for the counterfactual capital stock. We approximate the depreciation function used by the ABS by computing the average depreciation on the capital stock for each industry in each year. Predictably, these steps result in a series for the productive capital stock that grows much more slowly than the actual. As we assume that firms are optimising each period the income shares calculated with the actual series should reflect the technological parameters of the production function, regardless of whether the underlying function is Cobb-Douglas or not. Thus, we use these same weights in calculating the contribution of computers to output growth in our counterfactual exercise.

6.1.1 Results

Table 3 presents the relevant results from this exercise:^[15]

As the results make clear a substantial proportion of the gains from information technology have come in the form of lower prices. Because of our assumption in the counterfactual exercise that prices remain at 1990 levels the difference also grows with time. Prices for hardware fell by an average of 16 per cent per year over the 1990s to be six times cheaper in 2000 than in 1990. Thus, had productivity improvements not been passed on in the form of lower prices, output in Australia could have been lower by an average of (0.25+0.70)/2 ≅ 0.5 per cent per annum. Or, compounding the individual annual effects, output in 1999/2000 would have been 4.8 per cent lower than it actually was.

Viewed another way, this shows that Australia has benefited from intense competition in a sector where productivity advances are quickly passed on to users through lower prices. This finding does not suggest that there are no gains to producers and innovators in information technology, merely that users have received some fairly large benefits along the way – users have by no means been left behind in the IT revolution.

Another implication is also clear. Australia has increased its nominal spending on computers significantly. This increase in expenditure alone accounts for approximately 0.6 per cent per annum of output growth. This is the direct reflection of the increase in computer investment from just under 2 per cent of GDP in 1989/90 to over 3 per cent in 1999/2000 and the corresponding increase in computers' share of income (in the market sector) from 3.3 per cent in 1989/90 to 5.1 per cent in 1999/2000. This increase represents a sustained shift in production methods to more intensive use of computers. This raises the possibility that, in contrast to our assumptions, firms have been unable to fully adjust to the new technology optimally each year. We address this potential problem in the next section.

6.2 Is Computer Use Associated with Higher MFP Growth?

We now turn to the second question – are there spillovers from computer use to MFP growth? One reason that we might worry about spillovers is because of mismeasurement of the real capital stock. It is very difficult to measure the real quantity of computer capital. Comparing a 2GHz Pentium 4 with a 40GB hard drive and 128MB of RAM with a 1GHz Pentium III with a 20GB hard drive and 64MB of RAM is hard enough let alone trying to compare a Pentium 4 with a Macintosh G4. While the Bureau of Economic Analysis (BEA) in the US addresses some issues with its hedonic index, it is clearly imperfect. The rapid pace of improvement in information technology compounds these problems meaning that, even if a good estimate of the real capital stock at one point of time can be computed, it is likely to be inaccurate within a short period of time. Given the likelihood that computer capital is mismeasured some people have suggested that there may be a spillover of computer productivity into MFP estimates.

Alternatively, and potentially additionally, there may be disembodied technological change associated with computer use. Thus, computer use may be associated with new ways of organising business that are inherently more productive. In this case, productivity improvements would not be directly tied to the quantity of computer capital used but result merely from the fact that firms had reorganised their operations to use computers. In this case there may also be some correlation between MFP growth and computer use.

Thirdly, adjustment costs may mean that firms have less capital at a given point in time than the profit-maximising level. If this is the case, the marginal product of computers will be higher than measured through the growth accounting-technique. This, in turn, means that computers' contribution to growth will be understated and MFP growth overstated.

Finally, firms may have overinvested in computer capital. That is, firms were swayed by the hype surrounding the ‘new economy’ and undertook excessive investment in computers. In this case, the output from computers would be lower than the price paid for them. If this were the case one might expect to see a negative correlation between IT use and MFP growth.

To examine if any of these problems may be present we look at the industry-level data for Australia to see if there is any correlation between computer use and MFP growth across industries. We present two figures to examine this question. The first shows MFP growth and IT income shares for each industry (except agriculture) for the periods 1989/90–1994/95 and 1995/96–2000/01.^[16] It suggests that higher computer use may be associated with higher MFP growth.

This figure shows some suggestion of a positive correlation. Nonetheless, the correlation is dominated by a few industries that experienced strong growth in MFP and had high computer use, most notably communications. There are many reasons one might question this approach. Different industries may grow at different rates for reasons other than computer use. Alternatively, there may be distortions to these results due to deregulation. Some of the best performers are service industries. Telecommunications, for example, was deregulated in 1991 and might have been expected to experience higher productivity as a result. To control for this we present a figure showing the change in MFP growth between 1990/91–1994/95 and 1995/96–2000/01 against the change in computer usage. This allows for industry fixed effects, in other words, that some sectors may have a higher rate of growth independently of their computer usage.

The results from this transformation of the data suggest that there is no obvious MFP spillover. With the exception of the one outlier (EGW), there is no obvious correlation in Figure 7. This suggests that the relation seen in Figure 6 reflects factors other than IT usage. However, the lack of correlation also provides no evidence that there has been over-investment in computers.

The fact that EGW is such an outlier suggests that its results have been significantly affected by industry deregulation. Thus, it would be wise to wait for more data before drawing any general conclusions about the effect of IT on the EGW sector.

Notwithstanding the results in Figure 7, it is still possible that there is a correlation between productivity and computer usage. The ABS, in constructing its estimates, does not assume that rates of return are equalised across industries. Instead it calculates an internal rate of return for each industry such that all its other estimates are consistent. That is, the internal rate of return for each industry becomes the residual and absorbs all the errors made in the process. If the internal rate of return is higher in industries that use computers more intensively this may be a sign of spillovers from computer use to broader productivity growth. We have, unfortunately, been unable to ascertain if this correlation exists.

Footnotes

This table uses the most recent numbers available to us from the ABS, which, in this case, end in the 1999/2000 financial year. [15]

Agriculture is excluded from the figure due to the large fluctuation in MFP associated with weather patterns. It is possible to use other measures on the vertical axis, such as IT's contribution to growth; changing measures has little effect on the results. [16]