2013 Cambridge Business & Economics Conference ISBN : 9780974211428
Sector-specific effects of the Australian Mining Boom:
Dutch Disease or Dutch Delight?
Jonathan R. Hambur* and Neville R. Norman**
Paper prepared for the CBEC Conference,
Murray Edwards College,
University of Cambridge, June 2013
*Reserve Bank of Australia, Martin Place, Sydney, NSW 2000, Australia.
** Economics Departments of the Universities of Melbourne (Victoria 3053 Australia) and Cambridge (CB3 9DD, United Kingdom). Corresponding author: ;
[Abstract]
Sector-specific effects of the Australian Mining Boom:
Dutch Disease or Dutch Delight?
July 2-3, 2013
Cambridge, UK
2013 Cambridge Business & Economics Conference ISBN : 9780974211428
Studies of the economic impact of mining booms on Australia tend to be either theory without evidence or empirical work based on broad aggregates and limited time spans. Here we investigate industry-specific impacts based on the latest data and VAR econometric techniques. We find mixed evidence for the notion that the latest Australian mining boom is having an adverse impact on manufacturing overall, but some specific adverse effects emerge when industry subdivisions are studied individually. Contrarily, other subdivisions appear to have been impacted positively by the mining boom. Our findings enable specific consequences of such a significant change-generator as the mining boom to be explored. They also offer the warning that Dutch Disease theory is based on overly-simplified assumptions and overly-aggregated economic sectors. There are many, not just two, speeds in a modern economy, some advanced, some retarded, and many left relatively unaffected by significant shocks such as mining booms. This paper has many issues of concern to the business environment in a contemporary global setting, as embraced by the Cambridge Business and Economics Conference series.
July 2-3, 2013
Cambridge, UK
2013 Cambridge Business & Economics Conference ISBN : 9780974211428
July 2-3, 2013
Cambridge, UK 2
2013 Cambridge Business & Economics Conference ISBN : 9780974211428
I:Introduction
Since the mid-2000s the real prices of many of Australia’s non-rural commodities have increased significantly, leading to large resource-industry profits and investment and contributing substantially to overall economic growth. Many commentators have observed disparate growth between resource and other sectors, creating what is widely called a “two-speed economy”. This crude characterisation has led to advocacy for government action, ranging from resource taxes and sovereign wealth funds, to industry bailouts andeven to the devaluation of the Australian dollar.
Such a divergence in economic performanceis directly predicted by ‘Dutch Disease’ (DD) theory. The theory contends that, through a mix of real exchange rate appreciations and reallocation of inputs, a resource boom will causetradeable sectors of the economy, like manufacturing, education and tourism, to shrink.
Received theory, media descriptions and previous empirical work tend to focus on broad groups such as manufacturing overall, or ‘trabeables’ and ‘non-tradeables’.Conversely, we focus onthemuch narrower sub-divisions of manufacturing in testing empirically which specific economic activities of Australia, if any, have‘contracted’DD. Specifically we use the flexible and embracing Vector-Auto-Regression (VAR) approachto test whether the mining boom has caused the touted ‘de-industrialisation’,while allowing for other causal-factor filters and differing responses in each subdivision of industry. We alsoadvance from the existing literature by using a different lag-selection method that is more suited to investigating long-term structural adjustments, such as those predicted by DD theory.
II: Literature Review
The term ‘Dutch Disease’ (DD) was coined by the Economist magazine in the 1970s to describe the apparent de-industrialisation of the Dutch economy following the discovery of natural gas.[1]As noted by Sachs and Warner (1995), Mikesell (1997) and Iimi (2007), DD is part of the wider resource-curse literature. This literature seeks to explain therelatively poor growth performance of many resource-rich countries.[2] Explanations for the resource-curse can be split into two broad categories (Devlin and Lewin, 2005): (a) those emphasising institutions and rent-seeking behaviour; and (b) those focusing on how resources affect the economy’s structure and its overall economic growth.[3]DD is part of the latter category.
Corden and Neary (1982) (CN) is the seminal work on DD. The model it develops is the basis of most latter published work.[4]The CNmodel containsa booming sector, a traded sector and a non-traded sectorwhich operate in a small open economy, meaning that the prices of booming and traded sector output are set internationally, while those of non-tradable outputaredetermined domestically. Labour and capital are internationally immobile,allbooming sector output is exported, real wages are perfectly flexible (implying full employment), and there are no monetary considerations or complications. Employment, wages, prices and output commence in equilibrium. CN theory then investigates theoretically the effect of a Hicks-neutral technological change on these variables for three input mobility structures.[5] These structures are:
i) sector-specific capital and mobile labour;
ii) capital mobility between the two domestic non-booming sectors and domestically mobile labour; and
iii) full domestic capital and labour mobility.
The implications of the technological change are separated into two distinct effects, the ‘resource-movement effect’(mobile labour is drawn from other sectors) and the ‘spending effect’ (real appreciation leads to increased domestic purchasing power and higher real imports).[6]
Many extensions have been made to this basic model. Several papers allow for more mobile labour and capital, both domestically (Corden and Neary, 1982; Corden, 1984) and internationally (Bruno and Sachs,1982; Corden, 1984; Kuralbayeva and Vines, 2008).Corden (1984) also considers a lagging tradeable sector made up of several component industries. In this case, while the whole sector may contract, some industries may expand.[7]
Several papers also incorporate market imperfections. Corden (1984) considers the effects of real and nominal wage rigidities on unemployment. Van Wijnbergen (1984), incorporates sticky real wages and real exchange rates (RER), noting that such rigidities may lead to increased unemployment. Benjamin et al. (1989) considers the case where the domestic tradable sector’s goods are not perfectly substitutable for goods on the world market and finds that ‘tradeable’ subdivisions may grow as a resultof the boom, if they have sufficiently low Armingtonelasticities (elasticity of substitution between products from different countries). Similarly, Norman (1977) notes that linkages to the booming sector - coupled implicitly with imperfect substitutability - may lead tradable industries to grow following the boom, rather than shrinking. It further notes that the realappreciation associated with the boom may make imported inputs cheaper, lowering costs for the tradeable sector and thus allowing it to expand output.
The empirical DD literature has had mixed results in identifying DD effects. Much of the early empirical work on DD, such as Hutchison (1990, 1994) and Bjornland (1998), focused on the effects of the discovery of North-Sea oil on the economies of Norway, the UK and the Netherlands. More recent papers focus on other oil booms, such as in Russia(Algieri, 2011;Dobrynskaya and Turkisch, 2010), Kazakhstan (Egert and Leonard, 2008)and OPEC countries (Fardmanesh, 1991). Other natural resource booms have also been considered, such as coffee booms in Colombia (Kamas, 1986; Raju and Melo,2003).[8]While some papers have usedeconometric techniques to identify DD in other industrialised countries, such as Canada (Beine et al., 2012), few papers have used econometric techniques toconsider whether Australia has ‘contracted’ DD.[9]
The empirical literature can be split between cross-country analysis and case studies. Cross-country papers, such as Harding (2010) and Ismail (2010),tend to use panel-data techniques. Ismail (2010)uses panel techniques with subdivisional data to examine DD’seffects on differentsubdivisions. The paper finds permanent increases in oil prices hurt the manufacturing sector, andthis damage is greatest in those subdivisions where capital intensity is highest and those countries with more open capital markets.
Case study papers tend to take one of two approaches. The first is to consider counterfactuals. For example Larsen (2005, 2006) use tests for structural breaks to compare Norway’s economy with its neighbours’ pre and post the discovery of oil. The other approach is to estimate reduced-form equations or VAR/VECM.[10] Of particular note is Hutchison (1994), which uses a VECM to analyse whether the North-Sea oil boom created DD-type effects in the UK, Norway and the Netherlands. Itfinds that, while the increased oil production had little adverse effect on manufacturing, increased oil prices did have an adverse impact. Bjornland (1998) carried out a similar exercise using a Structural VAR,but focused only on Norway and the UK. Itconcludedthat the North-Sea oil discovery adversely affected UK manufacturing, but positively affected Norwegian manufacturing.
III:The 21st Century Australian Mining Boom
The resource boom occurring in Australia since the mid-2000s has been largely driven by demand from China. This increased demand has led to an increase in non-rural commodity prices, have more than doubled and tripled in $AUD and $USD terms respectively since the start of 2004 (Fig. 1).As a result, since the start of 2004 mining company profits before tax have risen by over 400% in nominal terms and 300% in real terms (ABS, 2012a, 2012c). Despite this, mining production has not increased markedly, increasing by only around 30% over the decade to June 2012, compared growth of around 40% in the preceding decade (ABS 2012g). Explanations for the inelasticity of mining production include infrastructure bottlenecks (Bloxham, 2011), skill shortages, and the ‘Global Financial Crisis’, which may have delayed much needed investment.
Figure 1With regard to the resource-movement effect, since 2004 manufacturing workers’ ‘real-product wages’ (Nominal Wage/PPI) - which represents wages as a cost (Corden, 2012; Lowe 2011) -have fallen by around 5% (ABS, 2012d, 2012h).This apparent fall in the cost of labour is counter to DD’s prediction that the cost of labour in the tradeable sector should rise as a result of the rising marginal product and wages in the booming sector. This indicates that any resource-movement effect with respect to labour has been small. Explanations for this include: the capital intensive nature of the resource sector, spare capacity in the economy at the start of the boom, a sizeable compensating differential required for workers to be willing to work in the mining sector, the resource and manufacturing sectors targeting different types of workers.
Despite the apparent lack of resource-movement effect, increased income flowing into the economy
- part of the income effect - is evident in higher average-weekly and consumer wages
(ABS 2012a, 2012d; Gregory and Sheehan, 2012). Much of this income has entered the economy through mining companies via their increased spending on inputs and investment, higher tax and dividend payments, and through wealth effects created by their rising stock prices.
DD predicts that this higher income will lead to a real appreciation. This has been the case in Australia, with the ‘Trade-weighted Real Exchange Rate’ or ‘Real Effective Exchange Rate’ (REER) - often used to proxy for the RER - increasing significantly in the latter part of the decade (RBA, 2012b). Similarly, both the price ratios of tradable to non-tradable goods, and goods to services, have dropped appreciably (ABS, 2012b), while Australia’s Terms-of-Trade has risen to record levels (ABS, 2012f). These relative price movements are further indication of a loss of international competitiveness in the tradable sector, as predicted by DD.
As for the ‘de-industrialisation’ predicted by DD, since 2004chain-volumemanufacturing sales have fallen by around 5%, while Chain-Volume GDP has grown by around30% (ABS, 2012e, 2012f) .Over the same period, the growth performances of manufacturing subdivisions were mixed. While some subdivisions,such as FBT,grew moderately,some - such as textiles, transport and primary metals -shrank, and others, such as chemicals and machinery,continued to grow at similar rates
(Fig. 2 and 3).[11]Overall, while many would contend manufacturing in Australia has been shrinking for some time, the process does appear to have accelerated in some subdivisions since the beginning of the mining boom.
4: Empirical Methodology
While DD theory gives insight into which variables may be salient in testing for DD, it gives little insight into the exact structure and dynamics of the adjustments. For example, as noted by Hutchison (1990, 1994), how quickly the deindustrialisation occurs will vary between economies based on underlying structural parameters. Given this lack of guidance, a VAR approach was considered appropriate. Rather than imposing an a-priori structure, based on economic theory, it allows the data to reveal an appropriate structure. The VAR also provides several methods for assessing whether the tested relationship actually exists. These include Granger-causality tests (Granger, 1969), Impulse Response Functions (IRFs), and Variance Decompositions (VDCs).[12]
The VAR can be expressed as:
whereZt is a vector of k endogenous variables, πi is a matrix of k autoregressive coefficients at lag i, Xt is a vector of q exogenous variables and is a matrix of q coefficients on the exogenous variables. The error term, et, is assumed to contain no serial correlation and have a covariance matrix:
which is estimated using the average sums-of-squares of the LS residuals.
For the base specification, Zt consisted of mining income, the tradable subdivision’s sales and the REER. A mining income variable was chosen over a production variable because the boom appears to bemore evident in mining profits than mining production. Regarding the sales variable, to allow for identification of differing subdivisional responses to the resource boom theVAR was run separately using either aggregate manufacturing’s chain-volume sales, orthe sales of a subdivision. Both sales and profits were measured in log forms. Several dummies were also included. Seasonal dummies were used, along with two dummies to account for different stages of the GFC. Further, a boom dummy was included to capture the exogenous mining profit boom. Finally, a dummy for Q1 2009was included to capture a large negative shock to the REER in this quarter. A further specificationwas also considered, which included a measure of subdivisional input prices, deflated using the subdivision’s Producer-Price-Index (PPI).[13]
Augmented-Dickey-Fuller and unit-root tests were used to test for stationarity. Those variables found to be non-stationary were then modelled in differences. However, where co-integration was presentwe used aVECMalong the lines of Engle and Granger (1987).