RDP 2014-08: The Effect of the Mining Boom on the Australian Economy Appendix B: Comparisons to the Monash Model

The responses in the model depend on estimated parameters and specification choices, all of which are subject to uncertainty. So it is useful to compare the AUS-M results to results from other models and previous studies. In this appendix, we compare AUS-M results with those from the Monash Multi-Regional Forecasting (MMRF) model, which other researchers have used to examine related issues. The MMRF model is described in Centre of Policy Studies (2008). Downes and Hanslow (2009) present and discuss a comparison of MMRF and AUS-M for long-run responses to a change in wages.

The Monash model has a simpler treatment of short-term dynamics and macroeconomic relationships than AUS-M. However, these differences do not necessarily affect the long-run results, which we report below. Whereas MMRF is largely calibrated, in AUS-M the elasticities of substitution are estimated directly from the time series data, either as part of demand systems or the joint estimation of production functions. With respect to specification of industry responses, both models use constant elasticity of substitution (CES) production technology for industry value added and are based on ABS input-output data. The AUS-M production functions model constant price industry value added whereas MMRF models industry gross output using a nested CES structure at a much higher level of detail. (Some inputs are combined in fixed proportions at the top level of the nested production functions for some industries but substitution is allowed elsewhere.) The treatment of trade is broadly similar. Reflecting these differences, the two models are useful for addressing slightly different questions.

B.1 Comparison of Terms of Trade Simulations

To facilitate comparisons across models, we focus on a simple experiment of an exogenous change in the terms of trade. To be precise, in both models we shock mining export prices so as to increase the terms of trade by 10 per cent. The shock to AUS-M is around a forecast baseline run out to 2030 while that on MMRF uses the standard long-run closure. We focus on the long-run results – on the assumption that the capital stocks in the model will fully adjust to the shock by 2030.^[6]

In contrast to the AUS-M scenario discussed in Section 3.2, we do not adjust world growth or investment residuals. In contrast to the further AUS-M adjustments discussed in Appendix C, we do not adjust the import composition of construction equipment or net migration flows. These features seemed to complicate the model comparison and several of them are difficult to implement in MMRF in a consistent manner. This simpler experiment design also facilitates comparisons with some previous studies of terms of trade shocks, which we discuss below.

We make two adjustments to MMRF to make the shocks more similar. First, we increase capital productivity in line with adjustments made in AUS-M, discussed in Appendix C.1. The rationale is that the mining boom seems to have led to substantial reductions in capital productivity, as increasingly marginal reserves are brought into production and as the composition of mining has been increasingly capital intensive. The adjustments are implemented by changing the rental value of capital stocks.

Second, we broaden the definition of mining exports in MMRF to include metals such as zinc, lead and aluminium. These are ordinarily classified as manufacturing exports in MMRF but as mining exports in AUS-M. This involves shifting the export demand curves for steel, alumina, aluminium and other metals inwards, so as to give similar changes as other minerals prices.

Table B1 shows the long-run effect of a 10 per cent increase in the terms of trade on other variables in both models.

Given the different approaches to specification and estimation, the similarity of the broad results may be surprising. Both models suggest a moderate increase (of about 2 per cent) in real GDP with consumption and investment rising somewhat more, offset by large increases in import volumes. Both models have appreciations in the real exchange rate that are similar to, but slightly smaller than, the increase in the terms of trade. With respect to industry output, mining production rises substantially, accompanied by smaller increases in services. These increases are partially offset by decreases in agriculture and manufacturing.

There are also differences in the results. For example, export volumes increase slightly in AUS-M, but decline substantially in MMRF. The reduction in MMRF includes a large response by manufacturing and services exports to the higher exchange rate, which more than offsets a rise in mining export volumes.

One of the more important differences is that MMRF provides more finely disaggregated responses. The model provides estimates of output disaggregated by state and detailed industrial categories. The full output of the simulation runs to hundreds of thousands of estimates. In Table B1 we provide a very brief illustration, with estimates of output by state and a few select industries.

B.2 Comparison with Previous MMRF Studies

Three previous studies have examined the effect of terms of trade changes using MMRF. McKissack et al (2008) examines the impact of a 20 per cent increase in the terms of trade generated by a shift in world demand for iron ore and coal. The study is mainly concerned with the distribution of employment effects across states and industries. It uses a short-run closure for the model simulation where capital stocks are fixed. Total employment is also fixed (although labour freely flows between industries and regions). Impacts on macroeconomic aggregates like consumption, the exchange rate and national income are not reported. Presumably national income is around 5 per cent higher, and much of this would be redistributed to the household sector via a higher exchange rate. (Real wages are 2 per cent higher.) GDP rises by 0.3 per cent reflecting the movement of employment from low-productivity industries to the mining industry. Employment in mining and construction are higher while manufacturing employment falls by 7 per cent. Other industries are relatively unaffected.

The fall in manufacturing employment in this short-run simulation is likely to be a result of the impact of the exchange rate movement. There is no short-run boost to manufacturing from higher investment – the authors note this as a limitation of the analysis.

The Productivity Commission (2009) looks at the impact of increased labour mobility on the economy's response to an increase in the terms of trade. The study, using a modified version of MMRF, finds that a uniform 10 per cent increase in mining export prices (which would represent roughly a 6 per cent improvement in the terms of trade) leads to a 2.4 per cent increase in GDP and a 4 to 7 per cent increase in the real wages of blue-collar workers. The results are from the standard long-run closure of the model where there is full adjustment of capital stocks. Hence the much larger impacts on GDP than the Treasury study using the short-run closure where output only responds as a result of the reallocation of labour. (The simulation with reduced labour mobility reduces the GDP impact by 0.3 of a percentage point.)

Thompson et al (2012) extend the Productivity Commission study and look at the impact of the increase in the terms of trade and the potential impact of labour mobility in more detail, using both long-run and short-run closures. The short-run results are similar to the Treasury results, the main differences arising from the different pattern of export price shocks imposed. In the long run the study finds a 30 per cent improvement in the terms of trade leads to a 3 per cent increase in GDP, and an 8 per cent increase in household consumption. Manufacturing employment falls by 11 per cent while employment in mining and services is up by 13 per cent and 4 per cent respectively. Limiting labour mobility reduces the long-run GDP increase by 1 percentage point.

Footnote

As there is still some cyclical movement in the data, the average deviations over the period 2025:Q3 to 2030:Q2 are reported in Table B1. [6]