RDP 2002-02: Australian Use of Information Technology and its Contribution to Growth 4. The Data

While the theory used in this paper is relatively straightforward, the practice is more complicated. Getting adequate measures of the capital stock is not easy. Furthermore, the task is made more complex since, for the purposes of growth accounting, a slightly different measure of the capital stock to that which is normally used is needed. The capital stock traditionally reported in national accounts statistics is the wealth capital stock whereas the capital stock required for growth accounting exercises is the productive capital stock. The distinction is that the wealth capital stock measures what the stock could be sold for at a given point in time whereas the productive capital stock measures its income-producing capacity. A simple example may illustrate the difference more clearly: suppose the capital stock consists of two computers – a two-year-old Mac and a new Pentium. Suppose, furthermore, that computers last for four years and then cease to function with no residual value (except, perhaps, as a paperweight or modern art). Finally, assume that there is no decline in computer efficiency with age. If the two computers have the same output then the current income capital stock would be 2 Pentium equivalent units. The wealth capital stock, however, would be less. The two-year-old Mac only has half its service life left and so, abstracting from discounting, the wealth capital stock is only 1½ Pentium equivalent units.

It is possible to relax the assumption that computers maintain full efficiency throughout their life. Making different assumptions about the decline in efficiency of computers leads to a slight change in the arithmetic and more substantial effects on the calculated capital stocks. Nonetheless, this assumption does not affect the concepts involved.

The data used in this exercise are unpublished ABS estimates of the productive capital stock and rental returns to different types of capital, all broken down by industry within the market sector.[10] The ABS makes an adjustment for declining efficiency over the life of the capital in addition to calculating the appropriate productive capital measure. Conceptually, these estimates are the same as those of the BLS in the US, whose data has been used by most US researchers in this field. Nonetheless, there are still a number of assumptions that differ between the two estimates that make direct comparison problematic. Furthermore, it is important to note that the results are sensitive to the underlying assumptions for which no good benchmarks exist.

A further limitation in the accounts at the industry level is the apportionment of taxes and subsidies on products. Currently the ABS values industry output at basic prices, i.e., excluding taxes and subsidies on industry outputs but including taxes and subsidies on their production. Their rental income estimates, on the other hand, include taxes and subsidies on products. Ideally we would like all the components of the calculation to be valued on a consistent basis. However, while this is a problem, there are reasons to suspect that the effect may be limited. The growth-accounting exercise is primarily concerned with growth rates rather than levels. To the extent that taxes and subsidies are proportional to output this should not affect the industry-level growth estimates. Furthermore, it is a simple matter to check the size of the effect at the aggregate level since the ABS publishes the growth rate of gross value added (GVA) for the market sector valued at both basic and market prices.[11] This comparison indicates that in any given year the difference could be up to 0.5 per cent. However, over the longer term the differences tend to average out. Thus, between 1989/90 and 1999/2000 the average annual growth rate only differs by 0.1 per cent.

The apportioning of taxes and subsidies also has an effect on the estimate of the capital share of income at the industry level. The ABS capital income estimates include taxes and subsidies on products while the published gross value added by industry estimates do not. This leads to an overestimate of the capital share of income. Lacking estimates of GVA at market prices by industry, it is difficult to be certain of the size of the effect. Nonetheless, an idea of the size of the mismeasurement can be gained in aggregate. Taxes and subsidies on products are, on average, 7–8 per cent of GVA valued at basic prices. This suggests, with an average capital share around 40 per cent in Australia, that the estimate of the capital share may be up to 3 percentage points too high. This figure does, unfortunately, vary by industry as products from some industries receive significant subsidies while others incur significant taxes. For the time being this paper proceeds by noting that the final estimates are based on a capital share that may be up to 3 percentage points too high and, consequently, that the contribution of capital may be overstated by an average of 7–8 per cent and that the contribution of labour may be understated by a similar amount. This could have flow-on effects to the residual MFP component but the size of this effect will depend upon the growth rates of capital and labour and so is not readily estimable. Nonetheless, given the other sources of error in these estimates, an 8 per cent variation will not significantly alter any of the conclusions.


We are grateful to the Capital, Production and Deflators Section of the ABS for providing these data. The data obtained covered the period 1964/65–2000/01. [10]

The growth rate of real GVA valued at market prices is published in ABS Cat No 5204.0 (2000–2001), Table 20. The growth rate of real GVA valued at basic prices can be calculated from their published industry estimates contained in Table 10 of the same publication. [11]