Economic Analysis and Policy 50 (2016) 4151

Contents lists available at ScienceDirect

Economic Analysis and Policy

journal homepage: www.elsevier.com/locate/eap

Full length article

Trade liberalisation, inward FDI and productivity within

Australias manufacturing sector
Christopher Turnbull a,1 , Sizhong Sun a,2 , Sajid Anwar b,c,
a
College of Business, Law and Governance, James Cook University, Townsville, QLD 4811, Australia
b
School of Business, University of the Sunshine Coast, Maroochydore DC, QLD 4558, Australia
c
School of Commerce, University of South Australia, Adelaide, SA 5001, Australia

article info abstract

Article history: Using a simultaneous equations model, this paper aims to analyse the endogenous re-
Received 16 September 2015 lationship among trade liberalisation, inward foreign direct investment and productivity
Received in revised form 13 February 2016 within Australias manufacturing sector. Using two-digit quarterly time-series data, from
Accepted 15 February 2016
1988 to 2012, we find empirical support for trade liberalisation as a mechanism for im-
Available online 20 February 2016
proving productivity within the domestic manufacturing sector. However, we find that in-
ward foreign direct investment has not had a statistically significant impact on productivity
JEL classification:
F14
within the sector. By drawing on these findings, among others, we make some recommen-
F21 dations for Australias international trade, foreign investment and manufacturing industry
O11 policies.
2016 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights
Keywords: reserved.
Trade liberalisation
FDI
Productivity
Manufacturing sector
Australia

1. Introduction

The manufacturing sector has long been a key pillar of the Australian economy, accounting for as much as 16% of
domestic output in the early 1970s. However, the domestic economy has undergone significant structural change in recent
decades, characterised by an expansion of the services sector, and a relative decline in manufacturing and agricultural output
(Lowe, 2012). According to the Reserve Bank of Australia (2001), structural change in the Australian economy has been
predominantly driven by technological advancement and reduced protection from international competition. This paper
focuses on the latter, paying particular attention to the impact of international competition on productivity within the
domestic manufacturing sector.
The most notable channels through which domestic manufacturers have been exposed to foreign competition are imports
and inward foreign direct investment (FDI). As depicted in Fig. 1, both nominal and effective rates of assistance (NRA
and ERA, respectively) afforded to the aggregated two-digit Australian and New Zealand Standard Industrial Classification

Corresponding author at: School of Business, University of the Sunshine Coast, Maroochydore DC, QLD 4558, Australia. Tel.: +61 7 5430 1222.
E-mail addresses: Chris.Turnbull@my.jcu.edu.au (C. Turnbull), Sizhong.Sun@jcu.edu.au (S. Sun), SAnwar@usc.edu.au (S. Anwar).
1 Tel.: +61 2 6102 9979.
2 Tel.: +61 7 4781 1681.

http://dx.doi.org/10.1016/j.eap.2016.02.004
0313-5926/ 2016 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights reserved.
42 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151

Fig. 1. Assistance afforded to two-digit manufacturing industries. Note: Estimates of ERA and NRA from after the year 2000 are available from the
Productivity Commission upon request.
Source: Productivity Commission (2000).

(ANZSIC) manufacturing sector have declined substantially in recent decades. Moreover, microeconomic reforms and
bilateral agreements have assisted Australias total stock of inward FDI to grow from around 15% of GDP in 1980 to over
35% of GDP in 2012.
Whilst much of the existing theory in international economics suggests that increased liberalisation of trade and FDI
should result in improved host-nation productivity, productivity levels in Australian manufacturing have stagnated since
the early 2000s (Parham, 2004). Consequently, this paper aims to explore the relationship among trade liberalisation,
inward FDI and productivity within the aggregated two-digit ANZSIC manufacturing industries. Specifically, in exploring
this relationship, the paper focuses on addressing the endogeneity issue that has been largely ignored in previous
studies.
As a result of limited research exploring this issue in the Australian context, this paper aims to provide an important
evaluation of the Australian international trade, foreign investment and manufacturing industry policy arrangements.
Additionally, this paper also aims to update the existing literature by analysing current datasets and providing a useful
starting point for further empirical research.
The remainder of this paper is structured as follows. Section 2 offers a brief background on international competition and
the manufacturing industry in Australia. Section 3 provides a review of key literature. Section 4 outlines the methodology
underpinning the empirical analysis. Section 5 identifies data sources and constructs variables. Empirical results are
presented and discussed in Section 6. Section 7 offers policy recommendations and concludes the paper.

2. The empirical setting

In the years following the Second World War, attitudes towards international competition in Australia have changed
dramatically. Under the Menzies-led governments of the 1950s and 60s, protectionism was the cornerstone of industry
policy. Specifically, the Menzies governments offered significant tariff protection to domestic manufacturing industries
while concurrently attempting to attract FDI (Freedman and Stonecash, 1997). Although protectionist policies achieved
their aim of expanding the domestic manufacturing sector and providing an abundance of employment opportunities to
Australians, by the late 1960s calls to improve economic efficiency gained traction.
Over the following two decades the level of trade protection afforded to domestic industries declined substantially,
assisted largely through a universal 25% tariff reduction in 1973. The ERA afforded to domestic manufacturers began to
decline more rapidly through the late 1980s and 1990s following significant microeconomic reforms and, since the late
1990s, the ERA has remained relatively steady at around five percent, reflecting the sectors exposure to increasingly higher
levels of import competition (Productivity Commission, 2000).
Whilst the microeconomic reforms undertaken in Australia throughout the 1980s and 90s helped to significantly reduce
trade barriers, they also continued to attract FDI. According to Crotti et al. (2010), these reforms played an important role in
attracting greater FDI in Australia; rising from 15% of GDP in 1980 to over 35% in 2012. However, it is important to note that
over the five years to 2013, FDI in manufacturing averaged less than 16% of total inward FDI in the Australian economy.
As the Australian economy has increased its openness to international competition, the manufacturing sector has
undergone significant change. In the mid-late 1980s, manufacturing was responsible for close to 17% of total employment
and produced nearly 10% of domestic output. However, as in most developed economies, Australias manufacturing sector
has been the victim of major structural change in recent decades. In 2012, manufacturings share of total employment and
domestic output had fallen to around eight and six percent, respectively.
 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151 43

Although international competition, including imports and inward FDI, is thought to have played a major role in the
relative decline of Australias manufacturing sector, a number of other factors have contributed to structural change. Most
notable among these are rising costs of production (including labour costs), increased consumer demand for services, and
the lasting strength of the Australian currency. Notwithstanding the relative decline of manufacturing in Australia, the sector
continues to produce around $100 billion of domestic output each year.3

3. International Trade, FDI and Productivity Literature4

3.1. Trade liberalisation

A number of studies have considered the impact of trade liberalisation on various aspects of the Australia economy
(for example employment, economic growth, and productivity). While many studies have explored the impact of trade
liberalisation on the macro economy, a considerable proportion of the existing literature has focussed on productivity
spillovers experienced by the domestic manufacturing sector (see for example Bloch and McDonald, 2001, Chand, 1999,
Mahadevan, 2002, Palangkaraya and Yong, 2011, Paul and Marks, 2009 and Sanidas and Jayanthakumaran, 2007). Given
the traditionally important role of manufacturing in the Australian economy, it is not surprising that much of the research
pertaining to trade liberalisation has focussed on this sector.
An early benchmark for research into the impact of trade liberalisation on Australian manufacturing productivity was
provided by Chand (1999), who uncovered a negative and significant link between the NRA and multifactor productivity
(MFP) growth among eight two-digit ANZSIC manufacturing industries. This negative impact, which provided strong
empirical support for trade liberalisation, was identified by estimating an augmented CobbDouglas production function
populated with annual time-series data from 196869 to 199495.
Notably, the study found that a 10% reduction in the NRA caused up to a five percent increase in the growth rate of MFP.
Additionally, Chand (1999) highlighted the existence of inter-sectoral heterogeneity in the response to trade liberalisation;
a finding that is explained by the broad variety of industries that are considered part of the aggregated manufacturing sector.
Although the significance of this problem was highlighted in the study, no attempt was made to address the functional form
problems of the model, with the problem industries instead dropped from the estimation.
Chands (1999) empirical analysis was based on the CobbDouglas production function, augmented with variables from
new growth theory. The CobbDouglas production function, which measures technological progress using Solow (1957)
residuals, has been commonly employed throughout the literature as a method of measuring productivity (see for example
Driffield and Love, 2005, Fadinger and Fleiss, 2011, Hanel, 2000, Mahadevan, 2002 and Singh, 2011), though the framework
is not without criticism.
Some studies, such as Fox and Kohli (1998) and Paul and Marks (2009), argue that the CobbDouglas framework is an
inappropriate method of measuring productivity because it utilises an outdated functional form, aggregates outputs, and
cannot easily incorporate the effects of an open economy. Whilst these arguments are widely acknowledged, more flexible
functional form frameworks, such as Translog models, present their own problems, particularly in the significance of second
order conditions (Mahadevan, 2002; Murillo-Zamorano and Vega-Cervera, 2001).
Mahadevan (2002), who also explored the aggregated two-digit ANZSIC manufacturing sector between 196869 and
199495, helped to somewhat alleviate the inadequacies of MFP by employing a Stochastic Frontier framework. Although
the framework incorporated a model based on the CobbDouglas production function, it allowed MFP to be divided into
sub-components of technological progress and technical efficiency. Borrowing the dataset utilised by Chand (1999), the
model estimation was performed using Aitkens generalised least squares regression.
Significantly, Mahadevan (2002) found that trade liberalisation, measured by both nominal and effective rates of
assistance, positively impacted technological progress among domestic manufacturers, but had no significant impact on
technical efficiency. The distinction between technological progress and technical efficiency is an important contribution
to the literature, although it also presents some additional problems. Most notably, measures of technical efficiency remain
subject to the functional form problems outlined by Fox and Kohli (1998).
Although firm and establishment-level studies are scarce in the Australian literature, due mainly to data limitations,
a small number of studies have also explored the impact of trade liberalisation at the sub-industry level. Most notably,
Bloch and McDonald (2001) and Palangkaraya and Yong (2011) provide detailed analyses of firm and establishment-level
responses to trade liberalisation, respectively. Importantly, these studies suggest that trade liberalisation yields the greatest
productivity gains in industries with a high level of domestic participation and in industries which are exposed to the greatest
reductions in assistance, respectively.

3 A detailed account of structural change in the Australian economy is provided in the Australian Department of Industrys Australian Industry Report
2014.
4 The review presented in this section focuses predominantly on Australia. Of course, the literature on the determinants of FDI is vast. A more thorough
review of related literature involving other countries can be found in Anwar and Nguyen (2014), Anwar and Cooray (2015), Kahouli and Maktouf (2015),
along with a brief discussion in Section 3.2 of this paper.
44 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151

3.2. Foreign investment

While numerous studies have focussed on the linkages between trade liberalisation and various aspects of the Australian
economy, relatively few have considered the role of foreign investment. Some noteworthy exceptions include studies by Iyer
et al. (2009), Crotti et al. (2010), and Kirchner (2012).
Iyer et al. (2009) examined the implications of outward orientation on Australias economic growth. Using quarterly
time-series data from 1988 to 2003, the study explored the impact of five different measures of outward orientation on GDP
growth. Namely, the study considered the role of imports, exports, FDI, foreign portfolio investment (FPI), and other foreign
investment (OFI), independently. Using a co-integrated vector autoregressive (VAR) model and Granger causality tests, Iyer
et al. (2009) found both imports and FDI to have positive and significant effects on economic growth. Furthermore, the study
provided evidence to suggest that FDI was the only form of foreign investment to Granger-cause economic growth in the
long-run.
Although the empirical analysis performed by Iyer et al. (2009) focussed primarily on economic growth, its simultaneous
consideration of both trade and investment provides a helpful starting point for studies exploring many other impacts of
outward orientation. The main justification for this type of consideration is that trade and foreign investment channels are
highly interrelated. Therefore, the exclusion of one of these variables could lead to under or overestimation of the other.
Furthermore, measures of foreign investment should control for the heterogeneous impacts of FDI, FPI, and OFI (Iyer et al.,
2009). The distinction of foreign investment type is further highlighted by Kirchner (2012) in uncovering the determinants
of inward FDI flows to Australia.
Drawing on quarterly time-series data from 1989 to 2004, Kirchner (2012) estimated a model for inward FDI in Australia.
By employing a VAR model and Granger causality tests, the study found FDI to be positively related to economic growth and
productivity growth but negatively related to foreign portfolio investment (Kirchner, 2012). Additionally, FDI was found to
Granger-cause productivity growth, although this finding was only marginally significant at the 10% level. This conclusion,
however, was based on a rather crude measurement of trade openness; computed as a ratio of total trade (both imports and
exports) to GDP. Such a practice is criticised by many, including Iyer et al. (2009), who argue that imports and exports are
not homogeneous in their impacts on growth and should therefore be considered separately.
Another important finding of Kirchner (2012) is the significance of distinguishing foreign investment into its sub-
components, as indicated by Iyer et al. (2009). Moreover, the empirical evidence presented in the study suggests that
international trade and FDI may be substitutes, or in other words, tariff-jumping behaviour may occur in the Australian
context.
A somewhat contrasting perspective of trade and investment is offered by Crotti et al. (2010), who explored the
relationship between inward FDI and bilateral trade and investment agreements. Using panel data from 1993 and 2003, the
study analysed the impact of trade and investment accords on FDI flows to Australia from its six largest source countries: the
US, the UK, Japan, the Netherlands, New Zealand, and Germany. Empirical results were estimated through both ordinary and
generalised least squares regression of common constant and fixed effects models. Significantly, the study found Australias
bilateral trade and investment agreements to have a positive impact on inward FDI flows. However, the study specifically
highlighted the need for industry-level analyses to provide greater insight into the impacts of increased FDI, particularly on
manufacturing and mining industries.
Although relatively few studies explore the productivity spillovers of FDI on Australias manufacturing sector, the
broader literature related to productivity spillovers of FDI is vast. Notable examples of studies which have investigated
the productivity spillovers of FDI in other countries include Javorcik (2004), Haskel et al. (2007), Suyanto and Salim (2010),
and Suyanto and Salim (2013).
It is commonly believed that FDI from large multinational firms, which are typically exemplars of best-practice
knowledge and technology, will result in positive productivity spillovers for domestic firms. This is in addition to the
productivity-enhancing benefits that FDI is hypothesised to deliver through heightened competition. However, the empirical
evidence to support this hypothesis is mixed.5
Strong support for positive productivity spillovers hypothesis is offered by Haskel et al. (2007), who analysed UK
manufacturing firms at the plant-level from 1973 to 1992. Specifically, the study concluded that a robust and positive
correlation existed between the productivity of a domestic plant and foreign activity within the same industry. This
conclusion is corroborated by Javorcik (2004) who found evidence of positive productivity spillovers from FDI at the firm-
level in Lithuania. However, the study found positive spillovers were only achieved by firms with a mixture of domestic and
foreign ownership.
On the other hand, Suyanto and Salim (2013), who investigated FDI spillovers on productivity within the Indonesian
pharmaceutical sector, found evidence that FDI had negative spillovers on the technical efficiency of domestic competitors,
but positive spillovers for domestic suppliers. Heterogeneous impacts of FDI on domestic productivity are also found by
Suyanto and Salim (2010), who found that spillover effects can manifest through different sources and vary according to
firm-specific characteristics.

5 Suyanto and Salim (2010) offer a more thorough explanation of how FDI can lead to positive productivity spillovers.
 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151 45

3.3. Gaps in the existing literature

Based on the above preceding literature review, it is evident that the majority of existing studies offer support for trade
liberalisation as a mechanism for improving productivity in the domestic manufacturing sector (Bloch and McDonald, 2001;
Chand, 1999; Mahadevan, 2002; Palangkaraya and Yong, 2011; Paul and Marks, 2009). In addition, recent studies have
provided some insight into the determinants of inward FDI, as well as evidence to suggest that FDI has had a positive impact
on Australias economic growth in recent decades (Crotti et al., 2010; Iyer et al., 2009; Kirchner, 2012). Studies investigating
the productivity spillover effects of FDI in other countries also help to provide context for this paper.
Notwithstanding the importance of these contributions, the lack of studies simultaneously exploring trade liberalisation,
foreign investment and manufacturing productivity is a substantial gap in the existing literature. The significance of
such an analysis is best explained by Iyer et al. (2009), who highlight the endogeneity of international trade and
FDI.
Resultantly, this study aims to contribute to the existing literature by simultaneously considering trade liberalisation,
inward FDI and domestic manufacturing productivity. Such an analysis is necessary given the significant degree to which
trade and investment have been liberalised in Australia in recent decades and the concurrent lack of growth in MFP
in the domestic manufacturing sector. Accordingly, the following analysis will provide valuable empirical evidence for
policymakers to evaluate the suitability of Australias existing trade, investment, and industry policies. Additionally, the
study will serve to update the existing literature and provide a platform for future research into industrial productivity in
open economies.

4. Methodology

4.1. Model specification

A Simultaneous Equations Model (SEM) is employed to investigate the relationship among trade liberalisation, FDI and
productivity within the Australian manufacturing sector. Following previous studies see for example Anwar and Sun (2011)
and Sun (2011) the model is specified as follows:
uvat = 0 + 1 erat + 2 fdit + 3 klt + 4 gt + 5 salest + 6 mt + t (1)
fdit = 0 + 1 uvat + 2 erat + 3 gnit + 4 wt + t (2)
erat = 0 + 1 uvat + 2 fdit + 3 fpit + 4 gnit + t (3)
where uva is unassisted value added, constructed following Chand (1999); era is the effective rate of assistance; fdi is inward
foreign direct investment; kl is the capital-to-labour ratio; g is government expenditure on fixed capital; m is imports of
intermediate inputs; gni is gross national income; w is the weekly wage of employees; fpi is foreign portfolio investment;
and , , and , represent the error terms in the three equations, respectively.6
In the above SEM, Eq. (1) is formulated loosely following the growth accounting framework first introduced by Solow
(1957), but is extended to include variables inspired by endogenous growth theory, which is explained in detail in Rivera-
Batiz and Romer (1991). Instead of exploring economic growth, the framework is utilised to model unassisted value added
(UVA) by manufacturing firms operating in Australia, as performed by both Chand (1999), and Mahadevan (2002). While
the traditional capital and labour inputs have been incorporated through the capital-to-labour ratio (kl), Eq. (1) also includes
trade liberalisation (era), foreign direct investment (fdi), public infrastructure (g ), the size of the domestic manufacturing
sector (sales) and imports of intermediate inputs (m).
In Eq. (2), inward FDI is modelled as a function of endogenous productivity (uva) and trade liberalisation (era) variables,
and exogenous variables representing the size of the domestic economy (gni) and the domestic cost of production, namely
the wage rate (w).
Finally, trade liberalisation is modelled in Eq. (3) through the ERA afforded to domestic manufacturers (era). The equation
includes endogenous variables for productivity (uva) and inward FDI (fdi), and exogenous variables representing private
investment in domestically producing firms (fpi) and the size of the domestic economy (gni).

4.2. Estimation and inference procedure

The empirical estimation is performed using the three-stage least squares (3SLS) regression technique for evaluating
SEMs, as developed by Zellner and Theil (1962). After the estimation, autocorrelation plots are generated for each of the
predicted residuals. Upon observing the autocorrelation plots, the presence of serial correlation is identified for each of the
predicted residuals. With the existence of serial correlation, even though point estimates of coefficients remain consistent,
they are not efficient. To account for this problem, we employ the block bootstrap technique to compute standard errors for
our point estimates.

6 Variables written in lower-case represent the logarithmic transformation of original variables.

46 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151

Table 1
Summary statistics of selected data series.
Data series (units) Observations Mean Std. Dev. Min Max

Unassisted value added ($m) 95 4294.298 1911.784 1193.855 6802

Labour productivity ($000/labour hour/week) 95 108.1608 50.61971 27.3079 184.4366
FDI ($m) 95 250833.5 140651.9 77007.35 555373.5
ERA (%) 95 7.544211 4.387912 4 17.9
NRA (%) 95 4.74 2.895646 2.1 10.5
Capital-to-labour ratio ($000/labour hour/week) 95 62.99813 15.19095 37.9475 99.99666
Public infrastructure ($m) 95 6287.621 3428.052 2732 14 911
Sales ($m) 95 64137.68 20972.43 33 516 102 925
Imported intermediate inputs ($m) 95 13799.97 6533.417 3717 27 714
GNI ($m) 95 186220.8 79624.77 85 830 354 269
Wages ($/week) 95 629.8286 62.82187 528.2392 716.6065
FPI ($m) 95 508552.5 350119.1 103093.9 1 205 017

Unlike the standard bootstrap, the block bootstrap captures the dependence among time series by stratifying the data
into sequential blocks. It is important to note, however, that the resulting estimates are sensitive to the selected block length.
In our case we follow an approach outlined by Inoue and Shintani (2006) in selecting this block length. Namely, we have
chosen the block length () to be approximately = T 1/3 , where T is the number of observations in each series. Since each
series contains 95 observations, we compute bootstrap standard errors using a block length of 5 observations.

5. Data sources and variable construction

The empirical analysis is performed using quarterly time-series data, from 1988Q3 to 2012Q1, drawn from Australian
Bureau of Statistics (ABS), Productivity Commission (PC) and OECD databases. Where necessary, series are seasonally
adjusted and measured in constant prices. Additionally, all series are converted to their natural log form for scaling and
ease of interpretation.
Data collected from the ABS include time-series for gross value added in manufacturing; the total flow of inward FDI
to Australia; total expenditure on fixed capital in two-digit manufacturing industries; total government expenditure on
fixed capital; income from sales of goods and services produced in the two-digit manufacturing industries; total imports
of intermediate goods; total hours worked in the two-digit manufacturing industries; and gross national income (GNI).7
Additionally, data pertaining to the NRA and ERA for domestic manufacturing industries are obtained from the PC by
request. Data for average weekly earnings in manufacturing are obtained from the OECD databases. A full description of the
data, including sources, is provided in the Appendix. Summary statistics of the sample used in this paper are presented in
Table 1.

5.1. Endogenous variables

(i) Productivity
Productivity levels for the aggregated two-digit manufacturing sector are measured following the unassisted value-
added (UVA) approach outlined by Chand (1999). Specifically, gross value-added for the aggregated manufacturing sector
is deflated by the ERA afforded to the sector. Such a measure of productivity is justified as it captures value-added at border
prices.
An alternate measure of productivity is available in the form of labour productivity. This measure is computed as
manufacturing gross value added deflated by average weekly hours worked in the sector.
(ii) Foreign direct investment
Due to data constraints, total inward FDI in Australia is used as a proxy for FDI in domestic manufacturing industries.
Specifically, FDI is measured as the quarterly closing position of Australias direct investment liabilities.
(iii) Trade liberalisation
Trade liberalisation is measured through the ERA afforded to the aggregated two-digit manufacturing industry. The ERA is
the preferred proxy for trade liberalisation as it captures net assistance provided to industry. Most notably, the ERA considers
non-border measures such as subsidies, special taxation arrangements, and other special provisions afforded to domestic
industries, in addition to taxes and tariffs incurred on imported manufactures at the border. Additionally, the ERA captures
any benefits achieved by providing assistance to domestic industries.
An alternate measure of trade liberalisation is available in the NRA afforded to the aggregated two-digit manufacturing
industry. The NRA includes only the costs associated with assistance. Namely, these include the degree to which consumers
pay higher taxes and prices to support domestic industries. Both the ERA and NRA are regularly estimated by the PC, with
estimates available in publications like the Trade and Assistance Review 19992000, or upon request.8

7 An appropriate proxy for R&D is not available, and is therefore excluded from the analysis.
8 For the empirical analysis, the ERA and NRA are assumed constant across all four quarters of each financial year, as quarterly estimates are not available.
 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151 47

5.2. Exogenous variables

(i) CapitalLabour ratio Drawing on a basic value-added production function, productivity is hypothesised to be impacted
by both physical capital and labour inputs. Accordingly, these inputs have been incorporated through the capital-to-labour
ratio, as performed by Mahadevan (2002). In computing this measure, physical capital is measured as quarterly expenditure
on fixed capital in the aggregated two-digit manufacturing industries. The labour input is measured as weekly actual hours
worked in the manufacturing sector.
(ii) Public infrastructure
Manufacturing productivity is also hypothesised to be impacted by the level of public infrastructure at the sectors
disposal. Therefore, gross fixed capital formation attributable to general government is used to proxy public infrastructure.
(iii) Size of the domestic industry
The size of the domestic manufacturing industry has been included in the model to account for the impact of economies
of scale. The size of the domestic sector is measured as total income from sales of goods and services attributable to the
aggregated two-digit industries.
(iv) Imported intermediate inputs
As explained by Chand (1999), imported intermediate inputs can impact productivity by offering domestic producers
insight into foreign best practice. To capture this spillover effect, imported intermediate inputs are included in the model as
Australias total trade debits for intermediate and other merchandise goods.
(v) Size of the domestic economy
The size of the domestic economy is hypothesised to affect both trade liberalisation and inward FDI. For example,
governments may be inclined to provide a significant level of support to a developing industry, before reducing support once
the industry becomes fully established. Additionally, a larger economy is expected to attract a higher level of inward FDI,
ceteris paribus. For these reasons, gross national income (GNI) is included as a proxy for the size of the domestic economy.9
(vi) Domestic cost of production
The flow of inward FDI is likely to depend, at least to some degree, on the domestic cost of production. Thus, the domestic
wage rate is included in the model as an exogenous variable. Specifically, the wage rate is measured through average weekly
earnings for all persons employed in the manufacturing sector.
(vii) Foreign portfolio investment (FPI)
FPI is included in the model as a determinant of foreign capital investment in domestic manufacturing industries. Large
FPI is correlated with a strong domestic financial market (Kinda, 2011). Therefore, when FPI is large, domestic firms are less
likely to be disadvantaged in terms of access to finance, compared with FDI invested firms. Hence, governments will be less
inclined to provide assistance and, thus, FPI is included as the quarterly closing position of Australias portfolio investment
liabilities.

6. Empirical findings and robustness

6.1. Empirical findings

The results of the initial 3SLS estimation, which accounts for endogeneity of some right hand side variables, are reported
in Table 2. The 3SLS method is selected due to its well-established suitability for estimating SEMs.
Developed by Zellner and Theil (1962), the 3SLS method estimates all coefficients within the SEM simultaneously by
drawing on the estimated moment matrix of the structural disturbancesa product of the earlier two-stage least squares
estimation technique. The benefit of the 3SLS technique is that it can provide more efficient estimates through the inclusion
of instrumental variables within the SEM. The method also accounts for correlation among the residuals of each equation
by incorporating Generalised Least Squares estimation.
In examining the results of the initial estimation, we find that a one percent reduction in the ERA contributes to about
a one percent increase in manufacturing UVA, which is significant at the one percent level. This result is expected and
corroborates the theory that trade liberalisation leads to productivity spillovers in manufacturing.
Additionally, a 10% increase in total sales attributable to manufacturing is linked with just over a four percent increase
in manufacturing UVA, which is significant at the five percent level. Again, this result is expected and could be explained by
economies of scale.
In terms of government expenditure on public infrastructure, we unexpectedly find that a 10% increase in the variable
results in about a one percent reduction in manufacturing UVA. This result, however, is only marginally significant at the
10% level. Coefficients for FDI, capital intensity and imports of intermediate inputs are statistically insignificant.
As expected, these findings support the claim that reductions in industrial assistance have led to improvements in
productivity within the aggregated two-digit manufacturing industries. Furthermore, our results suggest that productivity
improvements for these industries are partially explained by economies of large scale production, which has been facilitated

9 GNI is preferred to GDP, as it more accurately captures income effects influenced by the terms of trade (Kirchner, 2012).
48 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151

Table 2
3SLS estimates of the initial SEM.
Equation Variables Estimated coefficient Bootstrap standard error

Eq. (1): uva

fdi 0.0465 0.1135
era 0.9864*** 0.0760
kl 0.0524 0.0399
g 0.0979* a 0.0592
sales 0.4242** 0.1887
m 0.0303 0.0536
const. 6.1848*** 1.3175
Eq. (2): fdi
uva 0.5578 0.4386
era 0.5221 0.2923
gni 1.1666*** 0.0550
w 0.1300 0.2149
const. 8.1872** 3.9215
Eq. (3): era
fdi 1.5571* 0.7895
uva 0.9702*** 0.1056
fpi 0.1982 0.2681
gni 1.5791* 0.8695
const. 12.3264*** 2.4057
Note:
*
Represents significance at the 10% level.
**
Represents significance at the 5% level.
***
Represents significance at the 1% level.
a
The coefficient for g is marginally significant at the 10% level.

by the growing size of the domestic market. On the other hand, we find no empirical evidence to suggest that either inward
FDI or capital intensity have directly impacted productivity of domestic manufacturing industries. We also find no empirical
evidence to indicate the presence of any significant skill or knowledge spillovers arising from imported intermediate inputs.
Surprisingly, 3SLS estimation results indicate that increased expenditure on public infrastructure has had a negative
impact on manufacturing productivity. Even though this finding is only marginally significant at the 10% level, such a finding
could be indicative of public infrastructure not providing significant benefits to the manufacturing sector, or that increased
government expenditure has crowded out private capital investment.
In the estimated Eq. (2) presented in Table 2, we find only the coefficient for GNI to be statistically significant. Specifically,
our results indicate that a 10% increase in domestic GNI results in about a 12% increase in inward FDI. This result is significant
at the one percent level. Logically, this result indicates that the domestic economy will attract a greater volume of inward
FDI as it becomes larger, ceteris paribus. Notwithstanding this, we find no empirical evidence to suggest that inward FDI is
directly impacted by domestic manufacturing productivity (UVA), the level of industrial assistance (ERA) or the wage rate.
In the case of Eq. (3) where the dependent variable is trade liberalisation, we find that, except for the estimated coefficient
of fpi, all estimated coefficients are statistically significant with the expected sign. The impact of foreign portfolio investment
on trade liberalisation is statistically insignificant.
Namely, we find a 10% increase in inward FDI leads to about a 16% increase in the ERA afforded to two-digit manufacturing
industries. This result is significant at the 10% level. In contrast, we find increases in manufacturing UVA and domestic GNI
to result in statistically significant reductions in the ERA. Specifically, we find a one percent increase in manufacturing UVA
to result in just under a one percent reduction in the ERA; a result which is significant at the one percent level. Similarly, a
10% increase in domestic GNI leads to about a 16% reduction in the ERA, being significant at the 10% level.
As expected, we find the ERA to be positively impacted by inward FDI. This finding is consistent with governments
predisposition to support domestic industries to compete with foreign firms. Additionally, we find evidence to suggest
that governments will reduce assistance to domestic manufacturing industries as they become more productive and as the
domestic economy grows larger. Whilst these findings are expected, we find no significant empirical evidence to support
the hypothesis that trade liberalisation is impacted by FPI.

6.2. Robustness

In order to test the robustness of our SEM-based empirical results, the model was re-estimated with alternate measures
of productivity and trade liberalisation. Namely, we consider labour productivity in place of UVA, and NRA in lieu of ERA,
respectively. Results from the re-estimated SEM, where the NRA is used in place of the ERA, are presented in Table 3.
Upon examination of the re-estimated model, we observe findings largely consistent with the initial estimation. However,
we uncover some notable variation in the significance of the estimated coefficients, particularly in Eq. (2). Specifically, we
now find public infrastructure to be statistically insignificant in its impact on manufacturing UVA. Additionally, the re-
estimated model suggests that both manufacturing UVA and the NRA have statistically significant impacts on inward FDI.
 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151 49

Table 3
3SLS estimates of the alternate SEM.
Equation Variable Estimated coefficient Bootstrap standard error

Eq. (1): uva

fdi 0.3468 0.3033
nra 1.0252*** 0.1674
kl 0.0973 0.1095
g 0.1589 0.1377
sales 0.9879*** 0.3565
m 0.1174 0.1214
const. 5.1083*** 2.7352
Eq. (2): fdi
uva 0.4306** 0.1761
nra 0.4097* 0.2101
gni 1.3347*** 0.0907
w 0.1727 0.3531
const. 6.8054*** 1.9263
Eq. (3): nra
fdi 2.6852*** 0.9784
uva 0.9400*** 0.1482
fpi 0.2739 0.3458
gni 3.1455*** 1.1023
const. 17.5888*** 3.0760
Note:
*
Represents significance at the 10% level.
**
Represents significance at the 5% level.
***
Represents significance at the 1% level.

We also considered the models sensitivity to alternate measures of the selected variables. For example, when labour
productivity is included in place of UVA and when both labour productivity and the NRA are concurrently included in the
model. The estimated results are not reported here to save space, but are available from the authors upon request. The
discussion below relates to the re-estimation of the three simultaneous equations.
(i) Productivity
Following re-estimation of Eq. (1), we find only the coefficients of nra and sales to be statistically significant with
the expected signs. Specifically, we find a one percent reduction in the NRA to lead to about a one percent increase
in manufacturing UVA. This result is significant at the one percent level and corroborates the earlier finding that trade
liberalisation leads to productivity improvements in manufacturing industries. Additionally, we find a one percent increase
in total sales of manufactured goods to contribute to just under a one percent increase in manufacturing UVA. This result
is also significant at the one percent level and reinforces the earlier suggestion that scale efficiencies lead to improvements
in manufacturing productivity. We once again find FDI to be statistically insignificant in its impact on productivity. In
contrast to the initial estimation, however, we now find public infrastructure to be statistically insignificant in its impact on
productivity.
(ii) FDI
Unlike our initial estimation, the re-estimated model suggests that manufacturing productivity and industrial assistance
are both statistically significant in their respective impacts on inward FDI. Notably, we find a 10% increase in manufacturing
UVA leads to just over a four percent increase in inward FDI, which is significant at the five percent level. This result could
reflect the expectation that higher levels of manufacturing productivity will attract a greater volume of inward FDI, ceteris
paribus.
We also now find that a 10% reduction in the NRA leads to about a four percent reduction in inward FDI, which is
significant at the 10% level. This result could be explained as a reduction in tariff jumping behaviour or, in other words,
a substitution from FDI to imports.
Finally, the re-estimated model suggests that a 10% increase in domestic GNI leads to about a 13% increase in inward
FDI. This result is consistent with our initial findings and significant at the one percent level. Again, this corroborates the
expectation that a larger economy will attract a greater volume of inward FDI, ceteris paribus.
(iii) Trade liberalisation
As expected, the re-estimated coefficients of Eq. (3) are consistent with those of the initial estimation. In the re-estimated
model we find a one percent increase in inward FDI to contribute to approximately a two and two-third percent increase
in the NRA, being significant at the one percent level. We also find that a one percent increase in manufacturing UVA leads
to just under a one percent reduction in the NRA, also being significant at the one percent level. This result reinforces the
earlier suggestion that governments are less inclined to support manufacturing industries that are relatively productive and
self-sufficient. Finally, the results of the re-estimated model suggest that a one percent increase in domestic GNI leads to
just over a three percent reduction in the NRA, which is significant at the one percent level. Again, this result suggests that
governments will be less inclined to support manufacturing industries as the domestic economy grows larger.
50 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151

7. Conclusion and policy implications

This paper utilises a simultaneous equations (SEM) framework to explore the endogenous relationship among trade
liberalisation, FDI and productivity in Australias aggregated two-digit ANZSIC manufacturing industries between 1988Q3
and 2012Q1. Through three stage least squares (3SLS) estimation, we conclude that productivity is positively impacted by
trade liberalisation, but find no empirical evidence to suggest that inward FDI has had an impact on productivity levels
in domestic manufacturing industries. Additionally, we conclude that inward FDI is positively impacted by productivity
of domestic manufacturing, but negatively impacted by trade liberalisation, when measured through the Nominal Rate of
Assistance (NRA) afforded to the sector. Lastly, we find that the NRA is positively affected by inward FDI, but negatively
affected by productivity levels in domestic manufacturing.
Most notably, we find a one percent reduction in the NRA afforded to two-digit manufacturing industries to result in
just over a one percent increase in manufacturing Unassisted Value Added (UVA). Additionally, our findings suggest that a
10% increase in manufacturing UVA leads to just over a four percent increase in total inward FDI to Australia, while a 10%
reduction in the NRA afforded to manufacturing leads to around a four percent decline in total inward FDI. Finally, our results
suggest that one percent increases in total inward FDI and manufacturing UVA, lead to approximately a two and two-third
percent increase and just under a one percent decrease in the NRA afforded to manufacturing industries, respectively.
The findings of this study corroborate those of previous empirical analyses, after we control for the endogeneity issue
encountered when modelling trade liberalisation, FDI and productivity in an SEM framework. Nonetheless, it is important
to acknowledge that this study does not aim to address the impact of trade liberalisation and FDI below the aggregated
two-digit ANZSIC manufacturing industries. Hence, we reiterate the need for sub-sectoral analyses in future additions to
the literature.
The empirical findings of this study have significant implications for Australias manufacturing sector, foreign investment
and international trade policy regimes. Firstly, our finding that government expenditure on public infrastructure has had
a negative impact on manufacturing productivity is surprising. However, this finding is sensitive to alternate measures of
productivity and trade liberalisation and is only marginally significant in the initial estimation. Notwithstanding this, such
a finding highlights the importance of adequate and efficient allocation of public infrastructure.
Secondly, our results provide further empirical support for trade liberalisation as a mechanism for improving productivity
levels in domestic manufacturing industries. Nonetheless, we acknowledge the significant degree to which Australia has
already liberalised its trade policies and the difficulties that will be faced in achieving further productivity gains via this
mechanism. Thus, policymakers should now consider multilateral trade reform to be a top priority if they hope to achieve
further domestic productivity gains in manufacturing. Additionally, governments should continue to pursue new free trade
agreements (FTAs), improve the terms of existing FTAs, and reduce levels of non-border assistance to domestic industries
where practical.
Finally, our empirical findings suggest that only trade liberalisation, and not inward FDI, has a positive impact on domestic
manufacturing productivity. From a policy perspective, this finding suggests that a stronger emphasis should be placed on
reducing both border and non-border assistance to industry, rather than on further liberalising FDI, in pursuing productivity
gains in manufacturing.

Acknowledgements

This paper has greatly benefited from useful comments and suggestions received from the reviewers and the editor.
However, the authors are solely responsible for all remaining errors and imperfections.

Appendix. Variable definitions and data sources

Variable Symbol Definition Source Catalogue no.

Capital-to- kl Natural log of manufacturing sector ABS 5625 &
labour expenditure on fixed capital, deflated 6203.B3
ratio by average actual hours worked by all
persons in the sector
Effective rate of era Natural log of ERA afforded to the PCa N/A
assistance aggregated two-digit manufacturing
industries
Foreign direct fdi Natural log of Australias total direct ABS 5302-27F
investment investment liabilities (closing
position)
Foreign fpi Natural log of Australias total ABS 5302-27F
portfolio portfolio investment liabilities
investment (closing position)
(continued on next page)
 C. Turnbull et al. / Economic Analysis and Policy 50 (2016) 4151 51

Variable Symbol Definition Source Catalogue no.

Gross national gni Natural log of Australias gross ABS 5206
income national income
Imported m Natural log of debits for intermediate ABS 5368
intermediate and merchandise good (total)
inputs
Labour lp
Natural log of two-digit ABS 5206 &
productivity manufacturing gross value added, 6203.B3
deflated by total weekly hours worked
(manufacturing)
Nominal rate of nra Natural log of NRA afforded to the PC N/A
assistance aggregated two-digit manufacturing
industries
Public g Natural log of general government ABS 5204
infrastructure expenditure on fixed capital
Sales sales Natural log of total sales of goods and ABS 5676
services produced in aggregated
two-digit manufacturing industries
Unassisted uva Natural log of gross value added in ABS & 5206
value added aggregated two-digit manufacturing PC
industries, deflated by the ERA
Wage w Natural log of average weekly OECDb N/A
earnings (manufacturing)
a
ERA and NRA estimates are available from the Productivity Commission upon request.
b
OECD Wage data is derived from ABS catalogue no. 6345.

References

Anwar, S., Cooray, A., 2015. Financial flows and per capita income in developing countries. Int. Rev. Econ. Finance 35 (1), 304314.
Anwar, S., Nguyen, L.P., 2014. Is foreign direct investment productive? A case study of the regions of Vietnam. J. Bus. Res. 67 (7), 13761387.
Anwar, S., Sun, S., 2011. Financial development, Foreign investment and economic growth in Malaysia. J. Asian Econ. 22, 335342.
Bloch, H., McDonald, J.T., 2001. Import competition and labor productivity. J. Ind. Compet. Trade 1, 301319.
Chand, S., 1999. Trade liberalization and productivity growth: Time-series evidence from Australian manufacturing. Econ. Rec. 75, 2836.
Crotti, S., Cavoli, T., Wilson, J.K., 2010. The impact of trade and investment agreements on Australias inward FDI flows. Aust. Econ. Pap. 49, 259275.
Driffield, N., Love, J.H., 2005. Who gains from whom? Spillovers, competition and technology sourcing in the Foreignowned sector of UK manufacturing.
Scott. J. Polit. Econ. 52, 663686.
Fadinger, H., Fleiss, P., 2011. Trade and sectoral productivity. Econ. J. 121, 958989.
Fox, K.J., Kohli, U., 1998. GDP growth, terms-of-trade effects, and total factor productivity. J. Int. Trade Econ. Dev. 7, 87110.
Freedman, C., Stonecash, R., 1997. A survey of manufacturing industry policy: From the tariff board to the productivity commission. Econ. Rec. 73, 169183.
Hanel, P., 2000. R&D, interindustry and international technology spillovers and the total factor productivity growth of manufacturing industries in Canada,
19741989. Econ. Syst. Res. 12, 345361.
Haskel, J.E., Pereira, S.C., Slaughter, M.J., 2007. Does inward Foreign direct investment boost the productivity of domestic firms? Rev. Econ. Stat. 89, 482496.
Inoue, A., Shintani, M., 2006. Bootstrapping GMM estimators for time series. J. Econometrics 133, 531555.
Iyer, K.G., Rambaldi, A.N., Tang, K.K., 2009. How trade and Foreign investment affect the growth of a small but not so open economy: Australia? Appl. Econ.
41, 15251532.
Javorcik, B.S., 2004. Does Foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages. Amer.
Econ. Rev. 94, 605627.
Kahouli, B., Maktouf, S., 2015. Trade creation and diversion effects in the Mediterranean area: Econometric analysis by gravity model. J. Int. Trade Econom.
Dev. 24 (1), 76104.
Kinda, T., 2011. On the drivers of FDI and portfolio investment: A simultaneous equations approach. Int. Econ. J. 26, 122.
Kirchner, S., 2012. Foreign direct investment in Australia following the AustraliaUS free trade agreement. Aust. Econ. Rev. 45, 410421.
Lowe, P., 2012. The Changing Structure of the Australian Economy and Monetary Policy. Reserve Bank of Australia, Sydney.
Mahadevan, R., 2002. Trade liberalization and productivity growth in Australian manufacturing industries. Atl. Econ. J. 30, 170185.
Murillo-Zamorano, L.R., Vega-Cervera, J.A., 2001. The use of parametric and non-parametric frontier methods to measure the productive efficiency in the
industrial sector: A comparative study. Int. J. Prod. Econ. 69, 265275.
Palangkaraya, A., Yong, J., 2011. Trade liberalisation, exit, and output and employment adjustments of Australian manufacturing establishments. World
Econ. 34, 122.
Parham, D., 2004. Sources of Australias productivity revival. Econ. Rec. 80, 239257.
Paul, S., Marks, A., 2009. Modelling productivity effects of trade openness: A dual approach. Aust. Econ. Pap. 48, 105123.
Productivity Commission, 2000. Trade & Assistance Review 19992000. Canberra.
Reserve Bank of Australia, 2001. The Manufacturing Sector: Adapting to Structural Change.
Rivera-Batiz, L.A., Romer, P.M., 1991. Economic integration and endogenous growth. Quart. J. Econ. 106, 531555.
Sanidas, E., Jayanthakumaran, K., 2007. Trade liberalization and lagged reactions of trade flows, productivity and internal demand: An application to the
Australian PMV industry. J. Asia Pacific Econ. 13, 2649.
Singh, T., 2011. International trade and economic growth nexus in Australia: A robust evidence from time-series estimators. World Econ. 34, 13481394.
Solow, R.M., 1957. Technical change and the aggregate production function. Rev. Econ. Stat. 39, 312320.
Sun, S., 2011. Foreign direct investment and technology spillovers in Chinas manufacturing sector. Chin. Econ. 44, 2542.
Suyanto, , Salim, R.A., 2010. Sources of productivity gains from FDI in Indonesia: Is it efficiency improvement or technological progress? Dev. Econ. 48,
450472.
Suyanto, , Salim, R., 2013. Foreign direct investment spillovers and technical efficiency in the Indonesian pharmaceutical sector: Firm level evidence. Appl.
Econ. 45, 383395.
Zellner, A., Theil, H., 1962. Three-stage least squares: Simultaneous estimation of simultaneous equations. Econometrica 30, 5478.