Productivity and Regional Economic
Performance in Australia
Edited by Christine Williams, Mirko Draca and Christine Smith
Office of Economic and Statistical Research
Queensland Treasury
This collection of papers has been published by the Office of Economic and Statistical Research,
Queensland Treasury, with the intention of generating and promoting informed debate on
productivity and economic growth issues, particularly in relation to policies to further improve
State productivity and economic performance.
The publication examines productivity and growth from a number of viewpoints. The opinions
expressed in this collection of papers are those of the individual authors and should not be
considered in any way to necessarily reflect the views and opinions of Queensland Treasury, the
University of Queensland, Griffith University, or the Queensland Government.
The information herein has been provided by the authors and derived from sources believed to
be reliable. However, Queensland Treasury does not guarantee or make any representations as
to its accuracy or completeness. Therefore, any information, statement, opinion or advice
expressed or implied in this publication is made on the basis that the State of Queensland, its
agencies and employees are not liable for any damage or loss whatsoever that may occur in
relation to its use.
Office of Economic and Statistical Research
Queensland Treasury
Level 16, 61 Mary Street, Brisbane Qld 4000
PO Box 37, Brisbane Albert Street Qld 4002
Telephone: (07) 3224 5326
Facsimile: (07) 3227 7437
Email: [email protected]
www.oesr.qld.gov.au
www.treasury.qld.gov.au
© Queensland Government 2003
Copyright protects this publication. Except for purposes permitted under the Copyright Act 1968,
reproduction by whatever means is prohibited without the prior written permission of the Under Treasurer,
Queensland Treasury.
ISBN: 0-9751137-1-2
Foreword
This publication provides a major contribution to our understanding of the performance of
the Queensland economy, and will assist in both further research and public policy. A better
understanding of the drivers of our economic growth and performance enables the
Government to develop more focused policy to pursue our primary objective of increasing
prosperity and living standards for all Queenslanders. This will allow the Government to
more successfully address its key priorities of employment creation and positioning
Queensland as the Smart State.
Productivity and Regional Economic Performance in Australia contains seven research
articles that explore the productivity performance of Queensland and other Australian States.
Queensland has experienced a long period of economic growth well above the Australian
average. It has long been realised that much of this has been due to a strong net inflow of
migrants. It is also well known that Queenslanders are comparatively young and active – our
migrants tend to be of working age with families, and on average more of our working age
population seek to work. But Queenslanders also work smart and hard – as papers in this
book indicate, much of our growth has also been due to strong productivity growth.
Productivity growth is the key to the Smart State. It creates jobs and higher incomes. The
research in this book confirms the importance of the policy approaches of the Government –
innovation, education and skills, technology, and research and development. It will therefore
be useful and interesting to all who have some responsibility for, or interest in, the economic
progress of Queensland, whether in the private sector, financial markets or in the public
sector.
I acknowledge and thank all who were involved in the production of this research and of this
book. In particular, I am pleased by the involvement of a number of eminent Queensland
academic economists in the research and by the collaborative nature of their work with
Treasury economists. Treasury has a tradition of both internal research and engagement with
researchers in the universities and the private sector, and this book is another strong
contribution to our body of research knowledge about Queensland.
The Honourable Terry Mackenroth MP
Deputy Premier, Treasurer and Minister for Sport
– iii –
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .vi
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .vii
Notes on the Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .viii
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1
Jimmy Louca
2 New Growth Theories and their Implications for
State Government Policy Makers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
John Foster
3 Variations in Economic and Labour Productivity Growth among the
States of Australia: 1984-85 to 1998-99 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
Duc-Tho (Tom) Nguyen, Christine Smith and Gudrun Meyer-Boehm
4 Economic Growth Performance of the Australian States and Territories:
An Extended Shift-Share Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .67
Christine Smith and Duc-Tho (Tom) Nguyen
5 Multifactor Productivity and Innovation in Australia and its States . . . . . . . . . . . . . .99
Jimmy Louca
6 Recent Convergence Behaviour of the Australian States:
A Time Series Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .139
Philip Bodman, Mirko Draca and Phillip Wild
7 Human Capital Investment and Economic Growth
in the Australian Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .155
Mirko Draca, John Foster and Colin Green
Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .175
–v–
List of Figures
1.1
1.2
1.3
1.4
Regional Economic Performance, 1984-85 to 2000-01 ......................................3
Per Capita Income Performance, 1984-85 to 1998-99 ........................................6
Multifactor Productivity and Innovation ............................................................10
Human Capital in Australia and its States ..........................................................14
2.1
2.2
Microeconomic Reform ......................................................................................22
The Neo-Schumpeterian Logistic Growth Path..................................................28
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
3.10
3.11
3.12
Per Capita Income ..............................................................................................46
GSP Per Hour......................................................................................................48
Labour Productivity in Agriculture, Forestry and Fishing ................................50
Labour Productivity in Mining ..........................................................................51
Labour Productivity in Manufacturing ..............................................................52
Labour Productivity in Electricity, Gas and Water ............................................53
Labour Productivity in Wholesale Trade............................................................54
Labour Productivity in Finance and Insurance ..................................................55
Labour Productivity in Property and Business Services ....................................56
Labour Productivity in Government Administration and Defence ....................57
Labour Productivity in Personal and Other Services ........................................58
Labour Productivity, All Industries less Mining ................................................63
4.1
Classification of Regions according to Total Shift
(Comparative-Static and Dynamic Approaches) ................................................74
Classification of Regions according to Total Shift ............................................84
4.2
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
Multifactor Productivity in Australia ..............................................................103
Real Incomes in Australian States ....................................................................111
Labour Productivity in Australian States ..........................................................112
Multifactor Productivity in Australian States ..................................................114
Trends in Business Expenditure on R&D ........................................................119
Trends in Patent Grants ....................................................................................120
Innovation Output and Input Indicators............................................................122
Returns to R&D in Australia and its States ......................................................129
The Contribution of R&D to MFP across Australian States,
1985-86 to 1999-2000 ......................................................................................131
6.1
6.2
Gross State Product ..........................................................................................145
Dispersion of Gross State Product....................................................................147
7.1
Educational Completion by State, 2000 ..........................................................165
– vi –
List of Tables
3.1
3.2
3.3
3.4
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
Interstate Comparisons of Growth in Output, Population, Employment,
and Labour Productivity, 1984-85 to 1998-99....................................................44
Labour Productivity: Convergence Behaviour within Individual
Industries, 1985-86 to 1998-99 ..........................................................................59
Labour Productivity Growth and Share of Total Labour by Industry,
1985-86 to 1998-99 ............................................................................................61
Interstate Comparison of Growth in All-industry Real GSP and Labour
Productivity, with and without Mining ..............................................................62
Aggregate Results from Application of Comparative-Static
Shift-Share Analysis............................................................................................72
Aggregate Results from Application of Dynamic Shift-Share Analysis ............73
Disaggregated Results from Application of Comparative-Static
Shift-Share Analysis............................................................................................76
Disaggregated Results from Application of Dynamic Shift-Share Analysis......78
Aggregate Results from Application of Simple Productivity-Extended
Shift-Share Analysis............................................................................................84
Disaggregated Absolute Results from Application of Simple
Productivity-Extended Shift-Share Analysis ......................................................86
Disaggregated Relative Results from Application of Simple
Productivity-Extended Shift-Share Analysis ......................................................87
Aggregate Results from Application of the Multifactor
Productivity-Extended Shift-Share Analysis ......................................................92
5.1
5.2
5.3
5.4
5.5
A Decomposition of Economic Growth in Australian States ..........................104
Labour Productivity, Capital Deepening and MFP ..........................................106
Real Incomes in Australian States, 1985-86 to 2000-01 ..................................109
Economy-wide Returns to R&D in Australia ..................................................125
Econometric Results: Multifactor Productivity, 1984-85 to 1999-2000 ..........128
6.1
6.2
Gross State Product ..........................................................................................146
Summary of Pairwise Convergence Results ....................................................149
7.1
7.2
7.3
7.4
7.5
7.6
Distribution of Educational Attainment in the OECD ....................................161
Decomposition of Gemmell (1996) Human Capital Index for Australia ........162
Labour-Income Index of Human Capital for Australia ....................................163
Educational Completion Rates by State and Territory, 2000 ..........................165
Education Attainment by State and Territory, 2000 ........................................166
Human Capital and Economic Performance by State, 2000 ............................167
– vii –
Notes on the Contributors
Philip Bodman is a Senior Lecturer in the School of Economics at the University of
Queensland. He specialises in macroeconomics, applied econometrics, and labour
economics. His research interests include aspects of economic growth, particularly the
relationships between trade, education, migration and economic growth; econometric
modelling and dating of business cycles and international influences on the Australian
economy; analysis of the sustainability of deficits; the economics of education and
unemployment; empirical aspects of fiscal federalism; and the economics of crime. He has
published in journals that include the Canadian Journal of Economics, Applied Economics,
and the Economic Record.
Mirko Draca is a Senior Research Officer at the Centre for Economic Policy Modelling at
the University of Queensland. His current research interests include education economics,
applied econometrics and Austrian economics.
John Foster is Professor and Head of the School of Economics at the University of
Queensland. He is a Fellow of the Academy of Social Science in Australia, Life Member of
Clare Hall College, Cambridge, and Distinguished Associate of the ESRC Centre for
Research on Innovation and Competition at the University of Manchester. His current
research interests include the macroeconomy as a complex adaptive system; the application
of self-organisation theory to statistical/econometric modelling in the presence of structural
transition; the theory of competition and competition policy; legal and regulatory
interactions with the process of economic evolution; and the role of innovation and
education in determining economic growth. He has extensive consultancy experience with
a range of organisations, including the European Commission, the National Australia Bank
(UK), the International Labour Organisation, Queensland Treasury and the Victorian
Department of Treasury and Finance.
Colin Green is a Senior Research Officer at the Centre for Economic Policy Modelling at
the University of Queensland. He specialises in labour economics and applied
microeconometrics. Recent research includes studies of worker retrenchment,
neighbourhood effects in the Australian youth labour market, and the role of casual
employment in internal labour markets.
Jimmy Louca is a Senior Economist in the Economic Policy Branch, Queensland Treasury.
He has been involved in developing an economic strategy framework for Queensland
Treasury, has contributed to several State Government employment policy initiatives and
has worked on the labour market sector of the Queensland macroeconometric model. His
research interests include understanding the causes of unemployment and the economic and
demographic drivers of productivity growth.
Gudrun Meyer-Boehm is a Research Assistant and PhD student in the School of Economics
at Griffith University, and an Economist in the Intergenerational Modelling Team, Office of
Economic and Statistical Research, Queensland Treasury.
– viii –
Notes on the Contributors
Duc-Tho (Tom) Nguyen is Professor of Economics at Griffith University, where he has
held senior posts, including Dean of Commerce and Administration and Head of the School
of Economics. Previously, he worked at the Australian National University, University of
Illinois, University of Adelaide, and Australian Public Service. He has authored or coauthored numerous professional papers and co-edited several books on economics and
finance. He has also acted as a consultant to various government and international bodies,
including the Australian Agency for International Development, the International Labour
Organisation, and the United Nations Development Programme.
Christine Smith is an Associate Professor in Economics and Dean of the Faculty of
Commerce and Management at Griffith University. Her main research interests are in the
areas of multiregional economic and demographic modelling, project evaluation (including
cost benefit analysis and economic impact analysis) and conflict management.
Phillip Wild is a Research Fellow at the Centre for Economic Policy Modelling at the
University of Queensland. He specialises in macroeconometric modelling and time series
econometrics. He was Principal Research Officer (1993-1996) on a large ARC project
investigating nonlinear models of monetary aggregates (University of Queensland) and
Simon Research Fellow (University of Manchester, UK) (1996-1999). He has published in
major international journals including Macroeconomic Dynamics, The Journal of
Evolutionary Economics and The Cambridge Journal of Economics. Current research
focuses on the development of the ARDIMS input-output econometric model.
Christine Williams is Director of Economic Policy Branch, Queensland Treasury. Her
current research interests are focused on the performance of the Queensland economy and
vary widely, from the development of a macroeconomic model of the Queensland economy
to an assessment of the key drivers of economic growth in Queensland.
– ix –
1 Introduction
Jimmy Louca
Queensland has experienced a golden era of economic growth over the past decade and a
half, recording stronger rates of growth in output, real wages and employment than that in
the rest of Australia. This impressive economic performance prompted the Office of
Economic and Statistical Research within Queensland Treasury to develop the Drivers of
Economic Growth project, a collaborative exercise involving the Office, the University of
Queensland and Griffith University. The principal aim of the project is to identify the
fundamental factors that have caused Queensland to record historically higher economic
growth than that in the rest of Australia and to isolate those factors that will play an
important role in shaping future economic growth, in order to assist the formulation of state
economic policy. The papers in this volume, Productivity and Regional Economic
Performance in Australia, contain the findings and policy implications from the first stage
of this research project.
Productivity as a source of economic growth is a central focus of this volume. An economy
can grow by either accumulation of its inputs, namely labour and capital, or improvements
in productivity, that is, the rate at which inputs are transformed into output. Productivity
growth is the main source of increases in living standards and sustainable growth in
employment and economic output. Growth in productivity creates more output from given
inputs, generating a greater amount of income to be shared among residents of an economy,
raising real per capita incomes – the main economic indicator of material living standards.
In the labour market, any increase in labour productivity allows employers to raise real
wages by a commensurate amount without increasing labour costs per unit of output,
helping to sustain employment growth. More generally, productivity growth enables
producers to raise supply without raising costs, allowing aggregate demand to grow at a
faster rate without the need to pass cost increases on to consumer prices, generating noninflationary sustainable economic growth.
Productivity growth is thus central to the attainment of the key economic policy priorities
of the Queensland government. However, it also plays an important role in the delivery of
the state government’s social policy priorities, including safer and more supportive
communities, a healthy environment, community engagement and a better quality of life.
For instance, the rise in real incomes generated by productivity growth also raises tax
revenue without the need to raise tax rates, allowing governments to more easily increase
spending on education, health and aged care, environmental protection, crime and poverty
prevention and cultural activities (Baumol et al., 1988). As people’s incomes grow, they also
tend to have more concern for the environment and other aspects associated with a better
quality of life.
–1–
Productivity and Regional Economic Performance in Australia
Understanding the Drivers of Economic Growth
Several chapters in this volume concentrate on the determinants of productivity growth.
Productivity growth is driven by efficiency improvements (making better use of existing
technology) and technological progress itself. Capital deepening (increases in the capital to
labour ratio) also affects labour measures of productivity. While the tariff reductions, labour
market deregulation and microeconomic reforms that dominated the policy agenda in
Australia in the 1980s were based largely on efficiency considerations, attention turned to
innovation and human capital as determinants of technological advance in the 1990s.
Several developments led to this shift in emphasis. For instance, despite its success, there
has been a slowdown in the process of microeconomic reform following debate over the
extent of benefits delivered, mixed results across industries, unintended redistributions of
income and a growing realisation that such reforms create one-off improvements in
efficiency rather than sustained productivity growth (Industry Commission, 1995; Quiggin,
1998).1 In contrast, the information and communication technologies (ICT) boom that
drove an acceleration in productivity growth and the ‘New Economy’ era in the United
States (Gordon, 2000), along with related productivity spillovers in ICT-importing countries
such as Australia (Parham et al., 2001), have highlighted the importance of innovation and
technology diffusion to economic growth. To quote the OECD (2001, p. 51):
The ability to harness the potential of new scientific and technical knowledge
and to diffuse such knowledge widely has become a major source of
competitive advantage, wealth creation and improvements in the quality of
life. In order to reap the benefits from these changes, governments will have
to put the right policies in place.
A seminal contribution of this volume is the several chapters that look at how interstate
differences in productivity determinants have influenced economic growth across
Australian states over the past decade and a half (see Figure 1.1a). While much attention has
been given to the factors that have driven an acceleration in productivity growth in Australia
as a whole since the 1980s (Productivity Commission, 1999), research into how
productivity gains have been distributed across the states is sparse. This is surprising, given
significant interstate differences exist in educational attainment, rates of research and
development (R&D) expenditure, industrial structure and the impact and focus of
microeconomic reforms. Further, many of the policy tools able to address these factors are
available to state governments, providing considerable scope to influence productivity
growth and thus economic growth. The states of New South Wales, Victoria and Western
Australia continue to have significantly higher levels of per capita income than the states of
Queensland, South Australia and Tasmania (see Figure 1.1b). This provides clear evidence
of interstate differences in productivity determinants and highlights the need for state-based
policies that promote productivity growth.
1
In Chapter 2, Foster outlines several factors that have led to a slowdown in the process of microeconomic reform, marking the
introduction of the GST as an end point in this phase.
–2–
Introduction
Figure 1.1: Regional Economic Performance, 1984-85 to 2000-01
(a) Economic growth
Average annual growth,
1985-86 to 2000-01 (%)
5
4
3
2
1
0
NSW
Vic
Qld
SA
WA
Tas
WA
Tas
(b) Real output per capita
Real gross state product
($'000s) / Resident population
40
1984-85
2000-01
Qld
SA
35
30
25
20
15
10
5
0
NSW
Vic
Significant interstate differences in per capita incomes prompted authors of several chapters
in this volume to study the issue of convergence. The convergence hypothesis argues that
economies with lower per capita incomes should subsequently record faster growth, thereby
catching up to higher income economies over time. This process occurs through two
channels in particular. Given that developing economies face a higher return to capital, they
should record faster rates of capital deepening until their capital to labour ratio and the
return to capital is equalised with that of higher income economies. Lower income
economies also face convergence opportunities through absorbing the latest technologies
available in higher income economies. States such as Queensland, South Australia and
Tasmania are therefore expected to have recorded stronger growth relative to their
counterparts over the past decade and a half, as they converge on the per capita income
levels enjoyed in New South Wales and Victoria. Failure to catch up in this way suggests
structural impediments to convergence, and a number of studies in this volume assess state
performance in relation to these issues.
–3–
Productivity and Regional Economic Performance in Australia
The research gathered in this volume was not of course done in isolation. While this book
is a product of a small group of economists from the University of Queensland, Griffith
University and Queensland Treasury, it reflects the collective wisdom arising from a vast
literature on economic growth. There is a wealth of research available worldwide that helps
to develop an understanding of economic growth and appropriate policies, and the work in
Queensland has been informed by this literature and by contact with other economists,
policy makers and institutions. Several chapters in this volume draw on and synthesise
theoretical developments and empirical evidence.
Growth Theory: Its Implications for the Role of Government
Foster has been a key contributor to the recent literature on economic growth. In Chapter
2, he presents a clear and cogent exposition of a new set of theoretical perspectives on
economic growth that can assist policy makers. After briefly reviewing the microeconomic
reform process, he provides appraisals of the two theories that have emerged over the past
two decades as the most popular explanations of the process of economic growth. The first,
‘endogenous growth theory’, evolved from neoclassical growth theory proposed by Robert
Solow in the 1960s, while the second, ‘neo-Schumpeterian growth theory’, has been built
upon the insights of Joseph Schumpeter over half a century ago. Foster argues that neoSchumpeterian growth theory is more helpful in understanding economic growth and of
greater assistance to policy makers than endogenous growth theory. He provides examples
of how both theories can aid policy makers, and concludes by using education policy as an
example of how such theories can alter thinking about policy priorities.
Foster provides an intuitive and insightful critique of endogenous growth theory, illustrating
how market failures associated with human capital and innovation as sources of productivity
growth provide a revitalised government role in subsidising education and providing
appropriate patents law. However, he argues that, apart from such general prescriptions, the
theory provides little in the way of precise details about policy implementation, mainly
because it deals with aggregate generalisations and often uses unrealistic assumptions to
mathematically maintain tractability and consistency with the neoclassical framework. He
cites the assumption of a ‘one good’ final goods sector as perhaps the most serious
deficiency, given that in practice product variety abounds and consumers provide the most
decisive selection force. As a result, the theory ignores the marketing and entrepreneurial
efforts and uncertainty innovators face as they attempt to commercialise their products in
the market place – a large part of the innovation process and crucial area for policy.
In contrast, neo-Schumpeterian growth theory recognises that actions take place in a world
of incomplete knowledge, variety and uncertainty. It posits that variety in the stock of
knowledge generates diversity in the products and processes available, and incorporates
uncertainty, whereby only the most productive innovations survive the competitive test. This
selection process leads to ‘creative destruction’ in terms of the rise and fall of particular
firms and industries, and produces the necessary change in the economy to produce ongoing
economic growth. Foster argues that, whereas endogenous growth theory stresses market
failure as a rationale for government policy, the neo-Schumpeterian perspective stresses a
wider role for government in relation to the uncertainties people face in making economic
decisions. While uncertainty results in variety and selection, there is no guarantee that this
–4–
Introduction
will result in rational choices with economically valuable outcomes. The neoSchumpeterian policy theme is thus to anticipate and deal effectively with emerging
uncertainty by creating conditions in which quantifiable risk makes rational economic
choices possible.
This theme underpins many of the economic policy principles that are canvassed in the
chapter. One principle that is espoused is that governments should pay at least as much
attention to the ‘process of destruction’ as the ‘process of creation’. Foster argues that
government must be prepared to deal with the uncertainty that emerges among redundant
employees in the case of corporate failures, which he cites was lacking in the recent failures
of Ansett and OneTel. Rather than subsidising failing firms to continue in production,
retrenched employees should be compensated for the loss of specific human capital and be
supported to re-skill, in order to avoid losses of human capital and to provide new
entrepreneurial opportunities. Similarly, he argues that the high failure rate of small firms
that is necessary for the development of dynamic industries should involve a greater
government commitment, beyond bankruptcy law protection, to creating new opportunities
for failed entrepreneurs.
Another policy principle advocated is a focus on ‘regulatory innovation’ rather than
deregulation. Neo-Schumpeterian growth theory suggests government should introduce,
adapt and remove regulations, depending on the stage of economic evolution of various
firms and industries, in contrast to the idea underlying much of the microeconomic reform
agenda that a set of general theoretical principles could guide policy makers in all situations.
For instance, Foster argues that encouraging competition may or may not be appropriate,
depending upon an industry’s stage of development. Further, he argues that policy makers
must recognise that competitive selection will eventually result in monopolistic or
oligopolistic conditions. As a result, protective or facilitating regulations may be required to
reduce uncertainty in emerging industries that would be inappropriate in mature and
powerful industries, where certainty, inertia and the exercise of power must be challenged
by policies that remove barriers to entry, for instance.
The chapter concludes by using education policy as an example of how growth theory can
alter thinking about policy priorities. Foster suggests that individuals perceive participation
in higher education as a human capital investment decision, given little evidence of an effect
on demand for higher education following the introduction of the Higher Education
Contribution Scheme (HECS). In the case of highly vocational degrees, significant private
fee payments should thus be involved from a neo-Schumpeterian perspective, since
individuals are making choices that suggest they perceive quantifiable risk rather than
uncertainty. Foster notes that while human capital theory would favour policies that shift
resources away from non-vocational education, where benefits are more difficult to
quantify, neo-Schumpeterian theory provides a role for government in supporting high
standards in this area, stressing basic literacy and analytical skills as sources of variety in
knowledge, invention and growth. He provides a neo-Schumpeterian taxonomy of how
different human capital investments support ‘inventive’, ‘innovative’, ‘maintenance’ and
‘strategic’ behaviours, and suggests how the composition of investment must be altered
according to the changing structure of the economy in order to foster ongoing growth.
–5–
Productivity and Regional Economic Performance in Australia
Regional Productivity Growth in Australia
In Chapter 3, Nguyen, Smith and Meyer-Boehm turn to an empirical analysis of whether
labour productivity levels across the states have tended to converge over the past decade
and a half, in view of the importance of this process to convergence in per capita incomes.
They find clear evidence of divergence in per capita output across the states over the period
1984-85 to 1998-99, confirming the findings of previous Australian studies relating to
the 1970s and 1980s.2 This is shown in Figure 1.2a, which plots for each state the initial
level of per capita income in 1984-85 against the trend growth in per capita income over
1984-85 to 1998-99.
Convergence would suggest that the trend line through this scatter diagram would be
negatively sloped, indicating that states with initially lower levels of per capita output
subsequently record higher rates of growth in per capita output. However, in this case, the
line is positively sloped, suggesting evidence of divergence. Nguyen, Smith and MeyerBoehm find that the discrepancy between a ‘low income’ group comprising Queensland,
South Australia and Tasmania and a ‘high income’ group of New South Wales, Victoria and
Western Australia actually became more pronounced during the 15 years in question. This
is an important stylised fact that will be addressed by other chapters in this volume. Western
Australia recorded the fastest growth in per capita income over this period, moving to the
top of the high income group by 1998-99, while Tasmania recorded the slowest growth,
worsening its relative position within the low income group (see Figure 1.2b).
Figure 1.2: Per Capita Income Performance, 1984-85 to 1998-99
(a) Divergence in per capita income
Annual trend growth rate (%)
3.2
WA
2.6
Qld
2.0
NSW
Vic
SA
1.4
Tas
0.8
1.8
1.9
2.0
2.1
2.2
2.3
2.4
2.5
2.6
Predicted initial value (in logs)
2
Nguyen, Smith and Meyer-Boehm (Chapter 3) and Bodman, Draca and Wild (Chapter 6) in this volume both provide summaries
of recent Australian convergence studies.
–6–
Introduction
(b) Level of per capita income
NSW
Vic
Qld
SA
WA
Tas
6.0
Per capita income (in logs)
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
Once allowance is made for the effects of growth in population and labour force, however, it
turns out that the divergence pattern was much less pronounced. Indeed, labour productivity
levels neither diverged nor converged during the 1990s, with almost all of the divergence in
labour productivity during the period as a whole having occurred in the second half of the
1980s. Nguyen, Smith and Meyer-Boehm (2000) were probably the first to report this fact.
In attempting to explain the overall divergence trend in labour productivity for the period as
a whole, the authors carry out an industry by industry analysis and find that divergence at
the aggregate state economy level was caused mainly by interstate differences in industrial
structure, rather than by similar industries across states recording dissimilar growth. In
particular, during the second half of the 1980s, Western Australia (the highest labour
productivity growth state) benefited from having a proportionately larger mining sector,
which is itself capital intensive and has in turn recorded strong labour productivity growth
across all states over the period. When mining is excluded from the analysis, the authors
find no evidence of divergence or convergence and are left with a remarkably similar set of
labour productivity growth rates across the states.
In short, the mining boom in Western Australia has distorted the true underlying picture,
which is essentially one of parallel growth paths. In any case, the labour productivity growth
paths experienced by the various states did not conform to the usual convergence process
whereby lower productivity states record stronger capital deepening and catch up to higher
productivity states.
The authors draw two main policy conclusions from their study. First, they note that states
recorded similar labour productivity growth despite any interstate differences in
government policy that may have existed over the period. The authors argue that this
contrasts with the concern of previous studies that states have greatly differed in their ability
to adapt to changing domestic and external conditions, with possible consequences for
relative productivity performances. Second, though, they note that while there is no longer
evidence of divergence when mining is excluded from the analysis, similar productivity
growth across states has by the same token resulted in no convergence in per capita output
–7–
Productivity and Regional Economic Performance in Australia
either. The authors argue that this provides scope for lower income states to implement
policies that address any impediments that may have offset this convergence tendency, in
order to raise their relative per capita output levels.
Nguyen and Smith delve further into regional and sectoral explanations of interstate
differences in economic performance in Chapter 4. They use shift-share analysis in order to
explain economic growth across the states in terms of national growth effects, the impact of
industry mix, and regional effects. With this technique, states with industrial structures
conducive to above average growth are those that have a larger share of output attributable
to industries with national growth rates above the overall rate of economic growth
nationally. Similarly, states that exhibit regional advantage are those that record growth in
their industries above the national rates of growth in the same industries. Together, the
industry mix and regional effect determine whether a state recorded above or below average
economic growth over the period. The authors also extend their shift-share analysis in order
to determine whether state economic growth has been driven primarily by contributions
from employment growth or productivity change, the first such application of this type in
the Australian context.
The authors find that industrial mix and regional advantage have played a different role in
Western Australia and Queensland – the two states that grew at above average rates over the
past decade and a half. Western Australia was the only state to have both an industrial mix
conducive to above average growth and to experience even higher growth than expected on
this account alone (regional advantage). In comparison, Queensland had an industrial
structure conducive to below average economic growth, given it was less reliant relative to
other states on industries recording the fastest growth nationally. However, this was more
than offset by regional advantage, with most industries in the state recording growth rates
above their counterparts nationally. In contrast, Tasmania, Victoria and South Australia were
in the undesirable position of being characterised by both an industry mix conducive to
below average growth and regional disadvantage.
The productivity extensions in the chapter also provide some interesting findings, and are at
the cutting edge of the application of this technique internationally in the sense that they
relate to economic output rather than employment. The authors find that Queensland’s
regional advantage was driven by an above average contribution from employment change
to growth, with the state experiencing a below average contribution from changes in labour
productivity. This does not imply below average productivity growth. As Chapter 3 shows,
Queensland recorded the second highest labour productivity growth rate behind Western
Australia over the past decade and a half. However, Queensland also recorded the fastest jobs
growth of any state, meaning the share of its growth left attributable to labour productivity
fell below the national average. It was also found that the above average contribution of
productivity to growth in Western Australia stems more from non-labour factors, consistent
with Chapter 3 in which excluding the capital-intensive mining sector significantly lowered
labour productivity growth in this state. However, the authors note that state capital stocks
are required for a more detailed analysis of state level multifactor productivity.
Louca accepts this challenge in Chapter 5, by constructing state level capital stocks in order
to study multifactor productivity (MFP) rather than labour productivity. Labour
productivity, defined as output per hour worked, is only a partial productivity measure, as
–8–
Introduction
it can be raised by capital deepening. MFP is a more comprehensive indicator, measuring
output per joint unit of labour and capital. MFP growth occurs when output grows without
any rise in labour or capital and differs from labour productivity growth by excluding the
impact of capital deepening. Chapter 5 examines MFP across Australian states as a source
of interstate differences in economic growth and improving living standards over the past
decade and a half and then studies innovation as one explanation for these interstate trends
in MFP and economic performance. The chapter fills an important gap in the literature, with
little research previously conducted on MFP at the state level. As a result, it presents a
number of interesting findings and issues for policy that complement and build upon the
results of the other chapters in this volume that deal mainly with labour productivity.
The chapter highlights three notable stylised facts on interstate MFP. First, states recording
the highest economic growth over 1985-86 to 2000-01, namely Queensland and Western
Australia, also recorded the highest MFP growth. Queensland’s rapid population growth
caused it to have the highest growth contribution from labour accumulation, while Western
Australia’s mining boom saw it enjoy the highest contribution from capital – both results
consistent with findings of Chapters 3 and 4. Second, the contribution of MFP to rising real
per capita incomes ranged from 30% to 80% across the states, with the influence of terms
of trade and demographics on per capita incomes varying between states. Finally, and most
importantly, while Louca also presents evidence of diverging labour productivity levels
over the past decade and a half, he finds that interstate differences in capital deepening have
masked an underlying process of MFP convergence among five of the six states. This is
shown in Figure 1.3a, where the trend line is negatively sloped through this scatter diagram
plotting initial levels of MFP against MFP growth, suggesting states with initially lower
MFP levels subsequently recorded higher rates of MFP growth. The figure also shows how
Queensland and Western Australia were found to record MFP growth above rates expected
based on convergence channels alone (with these states lying above the trend line).
This last finding complements the results of Nguyen, Smith and Meyer-Boehm in this
volume. In Chapter 3, they find that Western Australia’s greater reliance on the capitalintensive mining sector largely explained their finding of divergence in labour productivity,
causing this high income state to record stronger capital deepening than lower income
states, contrary to what convergence dynamics would predict. In Chapter 5, Louca finds that
stronger jobs growth relative to any other state has caused Queensland to record
significantly less capital deepening relative to high income states, contrary to traditional
convergence theory. Thus, abstracting from capital deepening shows that MFP has been
converging across the states. As the authors of both chapters note, such a result should not
be surprising, given technology should transfer more easily across state economies that are
geographically close and face similar efficiency incentives under a federal microeconomic
reform program.
Louca finds that innovation activity sheds considerable light on interstate MFP trends. An
inspection of innovation indicators shows that the states that recorded the highest MFP
growth over the past decade and a half also recorded the highest growth in business R&D
and patent grants, while an econometric analysis reveals that business R&D growth explains
up to 75% of the variation in MFP growth across the states (see Figure 1.3b). The analysis
also finds evidence of interstate R&D spillovers and some equalisation in the returns to
domestic R&D across the five major states, both consistent with a process of convergence
–9–
Productivity and Regional Economic Performance in Australia
in MFP. In particular, the returns to business R&D seem to have been highest, but have
fallen most, in Queensland and Western Australia, which began the period with relatively
low commitments to R&D. This is consistent with the idea that such states initially faced
greater opportunities to profit from R&D and thus invested most heavily, causing their MFP
levels to converge on the MFP levels enjoyed in New South Wales and Victoria relatively
faster than other states.
Figure 1.3: Multifactor Productivity (MFP) and Innovation
(a) Convergence in MFP
1.8
Qld
Average annual MFP growth,
1985-86 to 2000-01 (%)
1.6
1.4
WA
NSW
1.2
Vic
1.0
SA
0.8
0.6
0.4
Tas
0.2
0.0
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
MFP level, 1984-85
(b) Contribution of business R&D to MFP
Contribution to average annual
MFP growth, 1985-86 to 1999-2000 (% point)
Domestic R&D
Interstate R&D spillovers
other factors
2.0
1.5
1.0
0.5
0.0
NSW
Vic
Qld
– 10 –
SA
WA
Tas, NT
& ACT
Introduction
Louca discusses several policy issues from his findings, using Queensland as a case study.
He argues that stronger jobs growth in Queensland requires higher rates of investment
relative to other states if Queensland is to record similar rates of capital deepening,
highlighting a vital challenge for investment policy in this state, given the importance of
capital deepening to convergence in per capita incomes. Similarly, he notes that, while fast
growth in business R&D has allowed Queensland to record MFP growth at a rate above that
expected from convergence dynamics, the state’s level of MFP, along with its R&D
intensity, still remains below that in the larger states of New South Wales and Victoria.
Louca argues that past convergence itself has partly been the result of the initially higher
returns to R&D facing this state, suggesting policies in Queensland will need to continue to
adapt to, and capitalise upon, threats and opportunities inherent in a changing technological
environment and remove structural impediments to this process, in order to raise this state’s
R&D intensity, level of MFP and per capita income level toward that in higher income
states.
Bodman, Draca and Wild take a different approach to convergence in Chapter 6. The
previous chapters in this volume studied convergence in terms of a gradual equalisation in
the level of per capita output across economies. In contrast, Bodman, Draca and Wild define
‘long-run convergence’ to exist between any two economies when the tendency for the
difference in their per capita output levels to narrow has been completed and a stable per
capita output gap between the two economies has been reached. In a sense, this type of
convergence deals with equalisation in the long-run growth of per capita output between
economies, and permits differences in per capita output levels to exist.
In order to evaluate the long-run convergence hypothesis, the authors are able to apply time
series methods that test for a stationary income gap between any pair of state economies
over time, given long-run convergence implies that any changes in the size of the income
gap between two economies should be transitory rather than permanent in nature. The
authors argue that the time series approach has several advantages over traditional
approaches to convergence, including the ability to test for convergence between different
pairs of economies and thus ‘convergence clubs’, rather than test only for a general pattern
of convergence across a larger group of economies.
Bodman, Draca and Wild find strong evidence for their convergence hypothesis across
Australian states. In particular they found that out of a possible fifteen pair-wise tests across
the states, seven for per capita output and ten for labour productivity indicated statistically
significant evidence of long-run convergence. The authors make a number of general
observations based on this evidence. First, while other studies in this volume testing for
convergence in labour productivity or per capita output in traditional terms found evidence
of divergence over the past decade and a half, this appears to represent a transitory rather
than permanent departure from a long-run stable income gap that exists between many pairs
of state economies. Further, while economic restructuring in terms of microeconomic
reform, trade liberalisation and labour market deregulation has had wide-ranging effects on
economic activity over the past decade, it has not caused any permanent changes in the
income gap between most pairs of state economies.
The authors also provide some interesting insights into individual state economies. They
argue that while Queensland has closed part of the shortfall between its per capita output
– 11 –
Productivity and Regional Economic Performance in Australia
level and that of New South Wales and Victoria, a stable income gap (long-run convergence)
had now been reached between it and the southern states. The authors interpret this as
limiting the extent to which Queensland can raise its relative per capita output through
traditional convergence processes such as capital deepening. Rather, they argue that
investments in human capital that raise the rate of technological progress are required for
Queensland to improve its relative level of per capita output. In contrast, the authors find
that the lowest income state, Tasmania, had the poorest results from the pair-wise
convergence tests, which they argue is inconsistent with traditional convergence dynamics.
However, this finding is consistent with the results in Chapter 5, where convergence in MFP
appeared to be operating among the Australian states, excluding Tasmania.
In advocating greater accumulation in human capital in Queensland, Bodman, Draca and
Wild thus give a specific example of the type of policy that Nguyen, Smith and MeyerBoehm in Chapter 3 stress would be crucial to removing impediments to convergence.
However, it should be noted that the interpretation placed on capital deepening by Bodman,
Draca and Wild in this case differs slightly to that in Chapter 5. There, Louca indicates that
capital deepening will play a crucial role for Queensland in the future, given stronger jobs
growth in this state will require higher investment rates relative to other states if Queensland
is to record rates of capital deepening comparable with the rest of Australia. Clearly, the
contribution of capital deepening to convergence in per capita incomes across the states is
an important area for future research, with implications for how State infrastructure policies
can complement education policies in raising productivity.
Human Capital and Innovation
In an enlightening study, Draca, Foster and Green examine in greater detail, in the final
chapter of this volume, how investment in human capital contributes to state economic
growth. A central theme of their chapter is that the focus on financing of education that has
dominated debate since HECS – itself an example of sophisticated policy design – must be
complemented with a discussion of the appropriate composition of human capital
investment if future education policies are to be effective at state level. The chapter thus
builds on Foster’s discussion on the composition of investment in relation to education
policy in Chapter 2 of this volume. The authors canvass previous research into the impacts
of education on labour market outcomes, study the causes of the rise in the Australian
human capital stock, and examine the contribution of human capital to interstate differences
in per capita incomes in order to provide a number of policy conclusions in the area of
education.
In reviewing previous work into the effect of educational attainment on labour market
outcomes, the authors question the validity of the ‘skill upgrading’ theory that has
dominated policy discussion. This theory suggests that improvements in the education level
of the labour force generate uniform improvements in terms of more jobs in high skill
professions at better rates of pay. However, they cite research to the contrary, including
evidence that a rise in demand for skilled workers has reduced the number of low skill
workers but caused a greater fall in their relative wages, increasing the dispersion in
earnings. They conclude that the skill upgrading theory needs to be replaced with a more
rigorous analysis of both labour market dynamics and the composition of educational
– 12 –
Introduction
investment in order to provide strategic policies that address issues such as conditions in the
low wage sector and earnings inequality.
The authors provide interesting insights into the composition of growth in Australian human
capital. They note that a move to mass higher education has raised tertiary attainment levels
in Australia above the OECD average, but that OECD countries have expanded
qualifications more evenly, with Australia possessing below average secondary level
outcomes despite improvements in school completion rates and vocational training. The
authors construct human capital stocks to show large rises in the stocks of secondary
qualifications and tertiary qualifications over the period 1970 to 1995. They find that labour
force growth accounted for most of the rise in secondary qualifications, while a rise in
enrolment rates drove tertiary qualification growth since the 1980s, consistent with a move
to mass higher education (see Figure 1.4a). The authors argue that demographic and other
influences will slow both the rate of increase in enrolments and labour force growth in the
future. While the former effect will be a general OECD trend, they argue that Australia’s
greater reliance on labour force growth means that convergence with educational attainment
levels of leading OECD economies is not assured.
Most importantly, Draca, Foster and Green highlight the importance of human capital to
interstate differences in per capita output. They illustrate that the three states possessing the
highest per capita incomes, namely New South Wales, Victoria and Western Australia, also
had the highest human capital stocks defined in terms of the distribution of educational
qualifications (see Figure 1.4b). In a growth accounting exercise, the authors show that
differences in human capital can explain as much as 87.0% of the difference in per capita
output between New South Wales and Queensland and 45.1% between New South Wales
and South Australia. Crucially, differences in secondary level qualifications were most
important in explaining differences in output per capita. They find that Queensland would
gain most by raising its human capital stock, with an estimated rise of $5,652 per capita and
an additional $14.7 billion in gross state product if it equalised its human stocks with New
South Wales based on 1996 data, reinforcing the need for Queensland to intensify efforts in
human capital accumulation.
The authors conclude with two main policy proposals. First, they argue that human capital
can account for large differences in interstate per capita output, raising the need for statebased human capital policies. They question the current emphasis of state development
strategies that focus on attracting physical or financial capital to particular regions, arguing
their value in raising productivity is limited. Second, the authors advocate comprehensive
programs in the area of early childhood education, following growing evidence that early
human capital investment promotes later investment and that by alleviating deficits created
in the early years in life, policy makers are able to avoid large costs incurred in later years.
This focus on early education is consistent with the authors’ growth accounting results that
illustrate the importance of below tertiary level education, but also highlights a concern over
Australia’s below average attainment in secondary and lower level education. The authors
argue that for ‘lifelong learning’ policies to be meaningful, they must begin with early
intervention programs that maximise the potential for ongoing skill upgrading through later
stages in life.
– 13 –
Productivity and Regional Economic Performance in Australia
Figure 1.4: Human Capital in Australia and its States
(a) Drivers of human capital growth
Labour force growth
Increase in enrolment rate
Increase in human capital index (%)
70
60
50
40
30
20
10
0
1970-80
1980-90
1990-95
1970-80
Secondary qualifications
1980-90
1990-95
Tertiary qualifications
(b) Education completion rates, 2000
Degree
Year 12
Share of working age population,
May 2000 (%)
70
60
50
40
30
20
10
0
NSW
Vic
Qld
– 14 –
SA
WA
Tas
Introduction
Summary of Findings and the Way Forward
The various determinants of productivity growth are clearly interrelated and this can be
illustrated by comparing the empirical results of Chapters 5 and 7 in particular. In Chapter
5, Louca finds that variation in business R&D activity can explain up to 75% of the
disparity in MFP growth across the states, while in Chapter 7, Draca, Foster and Green find
that a similar amount of the variation in per capita output across the states can be explained
by differences in human capital. These results can be reconciled when noting that human
capital and R&D spending are closely related. It is the analytical and creative skills
embodied in people that determine the rate at which new products can be developed and
new technologies absorbed, while the resulting R&D activity itself adds to the existing
stock of knowledge. It is for this reason that Foster in Chapter 2 comments that the ‘sharp
distinction’ made in endogenous growth theory between human capital and innovation is
‘somewhat artificial’ and stresses the neo-Schumpeterian perspective whereby variety in the
social stock of knowledge generates innovation and economic growth.
The relationship between productivity determinants is crucial for correctly interpreting
empirical work on economic growth and policy formulation. A casual inspection of the
results of a recent Productivity Commission paper on the impact of increasing skills on
Australia’s productivity surge finds only a limited effect. However, the authors stress that
only the direct impact of skills on labour productivity is estimated, rather than its influence
through innovation, which could be considerably greater (Barnes and Kennard, 2002).
Misinterpreting such results may lead policy makers to underestimate the importance of
human capital to economic growth in Australia. Similarly, Australian studies estimating the
benefits of microeconomic reform often ignore its influence on innovation, with growing
international evidence that greater competition prompts firms to innovate in order to obtain
a competitive advantage over their rivals (Aghion et al., 2002). Policies to hasten innovation
in particular industries are likely to fail if incumbent firms are not exposed to adequate
competitive pressures, while policies aimed at raising R&D spending will be ineffective if
education policies do not promote appropriate skill attainment.
The link between innovation, human capital and other drivers of economic growth helps
summarise the findings of this volume. The principal aim of the Drivers of Economic Growth
project is to identify the factors that have both caused Queensland to generate higher
economic growth historically and will be important to future state growth. The chapters
herein suggest that Queensland’s faster rate of economic growth over the past decade and a
half has been underpinned by faster productivity growth, driven by higher rates of business
R&D growth (Chapter 5), and stronger labour accumulation, underpinned by faster
population growth (Chapters 4 and 5). However, per capita output in Queensland still
remains well below that in New South Wales and Victoria, largely due to a relatively lower
innovative capacity reflected in a smaller human capital stock (Chapters 6 and 7) and
negligible capital deepening in Queensland over the period (Chapter 5). This contrasts with
the Western Australian experience, the other high growth state, where capital deepening has
allowed this state to surpass the per capita output levels enjoyed in New South Wales and
Victoria (Chapters 3 and 4). These findings have drawn out a number of issues for policy
concerning government strategies in relation to innovation and entrepreneurship (Chapter 2),
capital deepening (Chapter 5) and the composition of investment in education (Chapter 7).
– 15 –
Productivity and Regional Economic Performance in Australia
The policy issues raised in this volume support a number of initiatives already in place
under the Queensland Government’s economic strategy, which focuses on promoting
innovation, human capital investment and improving economic fundamentals as sources of
productivity growth and sustainable economic growth. For instance, the Queensland State
Education 2010 Strategy aims to significantly raise year 12 completion rates and introduce
a preparatory year of schooling, consistent with the evidence that secondary education
explains a large part of per capita output differences and the importance of early childhood
education to lifelong learning. State innovation strategies such as the Smart State Research
Facilities Fund, which provides funding for the construction of science and technology
R&D infrastructure, and the Queensland government’s proactive approach to forging
partnerships between the private and public sectors are also consistent with growing
evidence that emerging areas of technological opportunity are increasingly dependent on
public sector research, but also require partnerships that allow public inventions and
discoveries to be transformed into commercially viable products that create wealth in the
wider economy.
It is envisaged that future research under the Drivers of Economic Growth project will delve
further into issues with more detailed policy implications. The research under the first stage
of the project reflected more of a fact-finding exercise, given previously little Australian
work conducted into state economic growth. However, the results so far draw attention to
several issues worthy of future consideration. These include the complex interaction
between human capital accumulation, labour market outcomes and income distribution at
the state level and the determinants of interstate variation in innovation activity. The work
under the Drivers of Economic Growth project reflects the state government’s continued
commitment to an economic strategy that fosters high rates of productivity growth,
economic growth and improvements in living standards for current and future generations
of Queenslanders.
– 16 –
Introduction
References
Aghion, P., Bloom, N., Blundell, R. & Howitt, P. (2002), Competition and Innovation: An
Inverted U Relationship, NBER Working Paper No. 9929.
Barnes, P. & Kennard, S. (2002), Skill and Australia’s Productivity Surge, Productivity
Commission Staff Research Paper, AusInfo, Canberra.
Baumol, W.J., Blinder, A.S., Gunther, A.W. & Hicks, J.R.L. (1988), Economics Principles
and Policy, Australian Edition, Harcourt Brace Jovanovich, Marrackville, Sydney.
Gordon, R.J. (2000), ‘Does the ‘New Economy’ measure up to the great inventions of the
past?’ Journal of Economic Perspectives, 14 (4), 49-74.
Industry Commission (1995), The Growth and Revenue Implications of Hilmer and
Related Reforms, AGPS, Canberra.
Nguyen, D.T., Smith, C. and Meyer-Boehm, G. (2000), ‘Variations in economic and
labour productivity growth among the states of Australia, 1984/85-1998/99’,
Proceedings of the Australian Conference of Economists, 3-6 July, Economic Society of
Australia.
OECD (2001), Science, Technology and Industry Outlook: Drivers of Growth:
Information Technology, Innovation and Entrepreneurship, OECD Publications, Paris.
Parham, D., Roberts, P. and Sun, H. (2001), Information Technology and Australia’s
Productivity Surge, Productivity Commission Staff Research Paper, AusInfo, Canberra.
Productivity Commission (1999), Microeconomic Reform and Australian Productivity:
Exploring the Links, Productivity Commission Research Paper, AusInfo, Canberra.
Quiggin, J. (1998), ‘A growth theory perspective on the effects of microeconomic
reform’, in Microeconomic Reform and Productivity Growth, 26-27 February,
Productivity Commission and Australian National University Workshop Proceedings,
pp. 79-99.
– 17 –
2 New Growth Theories and their
Implications for State Government
Policy Makers
John Foster
Introduction
The objective of this paper is to discuss a new set of theoretical perspectives on the process
of economic growth that can assist policy makers, following on from a decade where
microeconomic reform dominated much of the policy agenda. All economic policy is
implicitly or explicitly based upon some economic theory or economic model. Even though
these may not be discussed at all in documents concerning policy recommendations, they
nonetheless play an important role in setting the policy agenda, in the interpretation of
economic information, in the design of policy instruments and in the goals that policy seeks
to address. For example, the microeconomic reform program in Australia was guided by
conceptions of perfect competition, in parallel with optimally efficient price mechanisms,
as an ideal state to aim for.
Here the goal is not to evaluate the success or failure of the microeconomic reform process
but to move on to consider how economic policy thinking concerning the attainment of high
and sustainable rates of economic growth can be guided by the ‘new’ growth theories that
have emerged over the past decade or so. There is no doubt that the static neoclassical
economic theory that guided thinking concerning the microeconomic reform process is
lacking as a means of understanding economic growth or for suggesting policies to promote
sustained economic growth. Neoclassical economists themselves have largely accepted this
and the first strand of new growth theory that I shall deal with – endogenous growth theory
– is an extension of the neoclassical economic framework to deal specifically with the
dynamic context of economic growth. As we shall see, this provides important new insights
and a revitalised role for government in fostering research and development and the
accumulation of human capital in the economy. However, although policy makers who look
carefully at endogenous growth theory are usually excited by its general thrust, they are
often somewhat disappointed when it comes to the precise recommendations concerning the
details of policy implementation. This is primarily because endogenous growth theory grew
out of macroeconomics and it remains a body of theory about aggregate generalisations.
The second strand of new growth theory that I shall deal with covers many of the issues and
mechanisms discussed in endogenous growth theory but from an entirely different
theoretical perspective. Neo-Schumpeterian growth theory is not built from neoclassical
economic foundations but from the insights of Joseph Schumpeter written down in the first
half of the twentieth century. Those who have studied some history of economic thought
– 19 –
Productivity and Regional Economic Performance in Australia
will recall that Schumpeter wrote a monumental two volume treatise on business cycles and
other important books dealing with economic growth, development, innovation,
entrepreneurship and the nature of the firm. However, Schumpeter’s powerful system of
ideas concerning economic growth lost popularity in the postwar era, mainly because it
lacked a deductive structure that could be easily expressed in formal mathematics and it did
not connect either with neoclassical microeconomics or Keynesian macroeconomics. Over
the past fifteen years this has changed: much greater precision has been used in expressing
and extending Schumpeter’s analytical system, Keynesian macroeconomics has become
unfashionable and neoclassical microeconomics has evolved into a more adaptable game
theoretical body of logic. The accompanying shift in policy focus away from short-term,
demand side considerations to a long-term, supply side perspective on the economy has
provided favourable conditions for a revival in interest in Schumpeterian ideas. It will be
argued in this paper that, in many respects, the result has been a body of theory that is much
more helpful in understanding economic growth, and of greater use to policy makers, than
endogenous growth theory.
The organisation of the paper is as follows: in the first section, I shall begin by explaining
why microeconomic reform, and the economics that lies behind it, is inadequate to deal with
economic growth. In the second section, endogenous growth theory will be introduced in a
highly intuitive way. The third section introduces neo-Schumpeterian growth theory. In the
fourth section, the usefulness of these two strands of new growth theory for policy making
is discussed. To illustrate how new growth theory can change thinking about policy
priorities, the important role of education policy in promoting economic growth is discussed
in the fifth section. The last section contains some concluding remarks.
Microeconomic Reform
Over the past decade or so, Australia has been engaged in a process labelled broadly as
microeconomic reform. This has involved the privatisation of public enterprises,
deregulation, the removal of various government subsidies, alteration in the system of
taxation, stronger competition policy, industrial relations reform and the reduction of
external protection. This reform process was predicated upon the presumption that the
economy was in a sub-optimal state and that, by liberating market forces and exposing
enterprises to more competition, there would be significant efficiency and productivity
gains. The removal of market imperfections and market incompleteness was seen, either
explicitly or implicitly, as moving the economy towards a perfectly competitive ideal state.
Even though it was well understood that reality could never approach such a state, being
closer to it was seen as welfare enhancing.
The era of microeconomic reform followed on from perceived failures of governments in
attempting to intervene in their economies, either to plan supply or to engage in Keynesian
operations to control unemployment and inflation through the management of aggregate
demand. Paradoxically, the introduction of microeconomic reforms was an indirect outcome
of attempts by economists to retain a commitment to Keynesian policies in the 1980s, in
the face of the non-interventionist challenge of ‘new classical’ economic thinking.
Macroeconomics had switched from a demand to a supply side perspective, but few
economists were satisfied with the new classical depiction of the supply side as a perfectly
– 20 –
New Growth Theories and their Implications for State Government Policy Makers
competitive system, held in equilibrium by a highly efficient price mechanism, eschewing
any attempts by governments to intervene to stabilise their economies.
Keynesians had always argued that the economic system was imperfectly competitive,
inducing price and wage rigidities that allowed stabilisation policies to be effective, but this
often required ad hoc arguments outside conventional economic theory or overly complex
theories. The ‘new Keynesian’ view was to accept the simple new classical model as a valid
depiction of the long run but to argue that the short run is characterised by market
imperfections, market incompleteness and information asymmetries. Although this was
intended to provide a depiction of the supply side that could justify a continued commitment
to stabilisation policies, it also provided support to those who argued that extensive
microeconomic reforms were necessary to improve the allocation of resources and
economic welfare. Generally, proponents of this view were neoclassical microeconomists,
not new classical macroeconomists.
Thus, we entered the 1990s with a policy commitment to macroeconomic stabilisation in
the short, and medium term, mainly through the operation of monetary policy, and a
commitment to microeconomic reform to meet long-term efficiency and productivity
objectives on the supply side. The latter was also viewed as providing a long-term solution
to unemployment by lowering the natural rate of unemployment. Fiscal policy, which had
been the main engine of Keynesian stabilisation policy, diminished in its role as a policy
instrument except perhaps in each year prior to an election. Budgetary balance, or surplus,
became an ongoing fiscal target, as microeconomic reform became the primary policy
instrument. Thus, the focus of economic policy moved from macroeconomics to
microeconomics and the goal of ultimately creating an economy that was efficient, flexible,
competitive and open enough to render discretionary fiscal policy unnecessary, with finely
tuned monetary policy being directed mainly at inflationary pressures.
In many respects, microeconomic reform was a useful and effective policy, since it led to the
removal of a number of arrangements that had the sole purpose of extracting economic rents
from consumers and it removed from government control the production of goods and
services that could be supplied more efficiently by the private sector. However, although the
early estimates of the benefits of microeconomic reform were large (Industry Commission,
1995), these have been subject to challenge by some economists (Quiggin, 1998), which in
turn has spawned several vigorous ripostes. There is little doubt that, in some instances,
outcomes did not accord with what neoclassical economic theory would have suggested –
sometimes prices rose when they were expected to fall or they fell because of very intense
competition only to be followed by corporate collapses and price reversals. Furthermore, even
if efficiency increased, the public did not always feel that a fairer outcome had been achieved
– in the context of neoclassical economics, what is implied is that the Paretian principle, that
no one should be rendered worse off by a policy change, was viewed as having been breached.
The resultant disaffection led to some minor political gains for the far left and far right of
politics and an unholy alliance against economic rationalism and globalisation.
Mixed results and political considerations resulted in a marked slowdown of the process of
microeconomic reform by the end of the 1990s. In this regard, the introduction of the Goods
and Services Tax (GST) in Australia can be viewed as an end point in a phase where
microeconomic reform was the central instrument of economic policy, because it
– 21 –
Productivity and Regional Economic Performance in Australia
highlighted many of the practical, political and administrative difficulties involved. Like
many other reforms, it was justified in terms of longer term efficiency and productivity
gains, based upon the logic of neoclassical economic theory, while involving clearly
identifiable redistributions from lower to higher income groups, at least in the short term.
Prior to the introduction of the GST, more and more economists had begun to argue that the
net marginal benefits of further reforms were likely to be low. Importantly, it was also
argued that, in any event, economic growth is not primarily dependent upon how well
market outcomes measure up with regard to the ideal notion of perfect competition, even
though the efficiency of markets and their prevalence are clearly important issues.
Microeconomic reform can cause productivity to grow for a while, as inefficiencies of
various types are eliminated and incentive structures are improved, but it cannot lead to
sustained economic growth. Microeconomic reform is generally regarded as an attempt to
eliminate X-inefficiency (Liebenstein, 1966), i.e. a tendency for productivity to be inside
the best practice production frontier, or to improve the functioning of the market mechanism
and associated incentive structures to obtain a price outcome that allocates resources more
effectively. In Figure 2.1, given a best practice production function OZ, the former involves
a movement from A to B and the latter from C to B (the slope of the production function is
the marginal product of the input, which is equated with the real price of the input if profit
maximisation exists – market liberalisation lowers this real price and increases input supply
leading to increased use of X and increased output of Y at point B). Economic growth can
be enjoyed but only temporarily. For sustained per capita economic growth to occur, the
aggregate production function must rise upwards continually.
Figure 2.1: Microeconomic Reform
Output
(Y)
Z
B
reduction of
X-inefficiency
C
market
liberalisation
A
0
Input (X)
With an upwardly mobile production function, microeconomic reform takes on an entirely
new dimension. Downsizing, rationalisation, wage cutting, etc. are replaced by a need for
policies that introduce new institutions, regulations and infrastructure to ensure that the
– 22 –
New Growth Theories and their Implications for State Government Policy Makers
economy can continually reach its potential. Much of the recent discussion of the ‘new’ or
‘knowledge’ economy is concerned with these issues and it has been recognised that it is
very difficult to employ the traditional analytical tools of neoclassical economics in the
presence of such complex economic dynamics and ongoing structural change in the
economy. This poses a clear difficulty for policy makers, both in thinking about the drivers
of economic growth and in offering coherent policies to promote it. It is this void that new
growth theorists, working in two distinct traditions, have sought to fill.
Endogenous growth theory, which will be dealt with first, can be traced back to the seminal
contributions of Paul Romer in the Journal of Political Economy in 1986. NeoSchumpeterian growth theory had its beginnings in the book entitled An Evolutionary
Theory of Economic Change by Richard Nelson and Sidney Winter, published in 1982. Up
until very recently, there was little connection between these two new theoretical schools,
despite the fact that there are strong similarities between them in several respects. However,
the former is built upon neoclassical economic principles, whereas the latter draws upon the
much more open-ended evolutionary economic tradition of Joseph Schumpeter. To put it in
a nutshell, the lack of connection arises from a fundamental methodological difference:
endogenous growth theories are built upon analytical foundations that involve equilibrium
outcomes that are ahistorical in nature, whereas neo-Schumpeterian growth theory is
constructed on considerably less well-known foundations that involve non-equilibrium
processes that are historical, not in a descriptive, but in an analytical sense.
Endogenous Growth Theory
Endogenous growth theory became considerably more prominent than neo-Schumpeterian
growth theory because of its well-understood neoclassical economic roots and the fact that
it revisited an important empirical mystery that had been given much attention in the 1960s,
namely the existence of the ‘Solow residual’ in economic growth models. Nobel laureate
Robert Solow discovered that, once the effects of increases in the aggregate stocks of capital
and labour are accounted for, there remains a large chunk of economic growth that is
unexplained even if the price mechanism is presumed to be operating to the best of its
ability. In terms of Figure 2.1, the aggregate production function did indeed appear to be
continually rising upwards. Furthermore, it was realised that without this residual, the
economy seemed to have no way of sustaining its growth in output per capita terms.
Without technical progress, per capita growth in the Solow model always tends towards a
stationary state. In other words, economic growth seems to originate beyond the economy,
if we choose to depict it in terms of an aggregate neoclassical economic model.
Solow (1956) employed a special Cobb–Douglas production function to relate the input of
labour (L), the input of capital (K) and the productivity of labour (A) with output (Y):
_
Y = K (AL)1-_
(1)
Taking logs and differentiating with respect to time, we get the following relationship
between the rate of growth of output, capital, labour and technological change (gA):
gY = _ gK + (1-_) gL + (1-_) gA
(2)
– 23 –
Productivity and Regional Economic Performance in Australia
This particular form of the production function ensures constant returns to scale and renders
technical progress (gA) exogenous to the model. Market forces are presumed to ensure that
growth will stay in equilibrium and not have a ‘knife-edge’ character, as suggested by Roy
Harrod half a century ago. Because of the presence of ongoing technical progress,
equilibrium is a steady state rate of growth of output per capita, rather than a stationary
level. Different rates of technical progress and different transitional dynamics, as stocks of
labour and capital change, can then be used to explain inter-country differences in levels of
output and rates of economic growth.
Convergence in output per capita can be explained by technological transfer mechanisms
that equalise rates of technical progress. Also, if labour and capital are paid their marginal
products, the neoclassical model predicts that poor economies will catch up and converge
with the output and income per capita levels observed in relatively rich economies. Thus,
economies can converge to the extent that their aggregate production functions eventually
demonstrate similar microeconomic characteristics in terms of technological opportunities.
The extent to which this is the case is usually examined through growth accounting. This is
based on the simple intuition that economic expansion is the result of two sources of
growth. First, there are increases in the available inputs of labour and capital. Second, there
are increases because of more productive use of these inputs attributable to, for example,
technical change and improvements in the organisation and management of production.
Total factor productivity (TFP) growth manifests itself as the residual left over once the
contributions of inputs to output are calculated, i.e. it is presumed to be (1-_)gA. Residuals
are often the product of measurement error, so one of the main goals of early research in
this field was to seek additional variables that affect production to minimise this residual.
As Quah (2001, p. 5) puts it: ‘the idea being that the smaller TFP became, the more
successfully one had explained economic growth’. Given that Solow (1957) found that the
TFP residual accounted for 80% of growth in per capita output in the United States, this was
understandable. Some researchers, such as Jorgenson (1988), found that it was possible to
remove the Solow residual entirely in the US case through more careful measurement of
factors of production and the inclusion of additional factors. However, in inter-country
comparisons, consistent and accurate data are scarce, so there has been a tendency to restrict
attention to capital and labour. It was the considerable variation in the size of the TFP
growth residual across countries when applying the Solow model that was one of the
primary motivations for the development of endogenous growth theory.
In discussing the origins and principles of endogenous growth theory, there will be no
attempt to provide technical details. These can be found in the very clear Introduction to
Economic Growth by Jones (1997) and in the more advanced Endogenous Growth Theory
by Aghion and Howitt (1998). In the mid 1990s, Romer (1994, p. 3) explained that
endogenous growth theory ‘distinguished itself from neoclassical growth by emphasizing
that economic growth is an endogenous outcome of an economic system, not the result of
forces that impinge from outside’. However, developments in endogenous growth theory do
not stop with this broad distinction. In leading the renaissance of growth theory from the
mid 1980s onwards, Romer (1986) and others went on to explore the interconnections
between knowledge, education and technical change in much greater detail than had
previously been the case.
– 24 –
New Growth Theories and their Implications for State Government Policy Makers
The core of Romer’s (1986) model is a formalisation of the process of invention and
innovation that is absent in the Solow model. Technological and organisational
improvements emanate from ideas. Ideas are goods that have the unique characteristic that
they are non-rivalrous, i.e. if one person uses the idea of another, the latter does not lose that
idea. This provides the basis for increasing returns from the generation of ideas. However,
in order to persuade people to invest in research to produce useful ideas, there must be a
degree of excludability, i.e. a right to own a patent; otherwise it will not be worthwhile to
invest in research. Two roles for government are immediately suggested: first, since nonrivalrous goods are essentially public goods, there is a clear role for government in assisting
in their generation; second, there is a key role for government in enforcing patents law in a
manner that allows enterprises to enjoy sufficient returns without overprotecting to the
extent that the ideas cannot spread.
Romer (1986) argues that ideas are generated by those engaged in research and
development (R&D) and rendered effective by patents. Thus, the numbers employed in
R&D and the numbers of patents granted are used as determinants of TFP growth. Because
ideas spill over, a given population of R&D researchers can generate an ever-increasing
stock of ideas, captured by increasing numbers of patents. Thus, there can be continuing
steady state economic growth without the neoclassical tendency to stationarity.
Technological transfer can spread ideas across the world and therefore can induce
convergence between economies with comparable absorptive capacities for ideas. Such a
capacity depends crucially upon the nature of education systems and, perhaps more
fundamentally, the kind of culture that a country has bequeathed from its history.
There is, however, a key difficulty in this line of argument. The spillovers involved in the
creation of ideas generate increasing returns to knowledge and therefore increasing returns
to production. This provides an explanation of TFP growth, but it also suggests that we
should observe quite a bit of imperfect competition in growing economies, not because of
the exercise of political power but because of technological progress. Since Alfred Marshall
onwards, imperfect competition has been the recognised outcome of increasing returns to
scale. This poses a fundamental problem for the perfectly competitive, constant returns to
scale, Solow model upon which Romer builds his model.
Romer (1990) circumvents this problem by envisaging the economy comprises three
sectors. The first is a final goods sector that is perfectly competitive, producing a
homogenous good, but employing heterogeneous capital goods in combination with
different amounts of labour. In turn, these capital goods are produced by an intermediate
goods sector that consists of monopolists that produce each of the heterogeneous capital
goods. The firms in the intermediate goods sector buy patents from the research sector. The
researchers extract a price for the patent that compensates them for their effort, but the
eventual benefits that arise to the economy as a whole, because of the impact on future
research, can be considerably greater.
The fact that a ‘futures market’ is missing in the case of research occurs for the obvious
reason that we are dealing with uncertainty. It is argued that this represents a market failure
that will result in too little research being undertaken, particularly of a fundamental type,
introducing an obvious role for government. Furthermore, the fact that the consumer surplus
associated with research is greater than the profits accruing to researchers also implies that
– 25 –
Productivity and Regional Economic Performance in Australia
too little research will be undertaken. However, a negative externality is also entertained as
a possibility in cases where there is too much duplication in research effort, lowering
research productivity. Again, there is a role for government here.
A final step in completing the basic endogenous growth theory framework is, following
Mankiw, Romer and Weil (1992), to argue that it is human capital, rather than simply labour,
that enters into the aggregate production function. This is a straightforward extension of the
Solow model, but it provides an explicit role for education and training – higher levels of
human capital will generate higher levels of aggregate output, either directly in production
or through an increased capacity to absorb transferred technologies. Because general
education has a significant public good dimension, i.e. it produces core skills that are too
general to associate with specific returns in particular occupations and thus constitute
spillovers, the government is attributed a key role in its provision. The evidence, using
comparative data on years of schooling across countries, provides support for a link
between education and output per capita.
Since Romer’s (1986) article, there have been many contributions that elaborate upon the
operation of these endogenous mechanisms. These have included models that have focused
on the effects of human capital (Lucas, 1988), international trade (Grossman and Helpman,
1991) and knowledge itself (Quah, 2001). There is also considerable empirical literature
that attempts to investigate the extent to which there is support for the various dimensions
of endogenous growth theory. Surprisingly, given that growth is something that takes place
in history, the bulk of new growth econometrics is not conducted on time series but instead
on country cross-section data, using mainly the Summers and Heston (1991) database.
There are difficulties with this because implicitly each country is in equilibrium at whatever
level of development it happens to be at. Yet it is also argued that their convergence occurs
because some countries are experiencing transitions, suggesting that they are not in
equilibrium. This seems unsatisfactory, given a common Cobb–Douglas production
function is imposed across countries as disparate as the United States and the Sudan.
This is recognised to some extent by grouping countries according to key institutional,
infrastructural and cultural differences that are likely to affect TFP growth in systematic ways
(see Jones, 1998 and Durlauf and Johnson, 1995). However, it is not clear that what is
discovered about differential development in this way is pertinent to the actual growth of
countries over selected time periods (Durlauf, 2001). As Ahn and Hemmings (2000) point
out ‘... the breadth and range of models provides empirical researchers with a carte blanche
when it comes to choosing variables in regression analysis. Furthermore, even when research
focuses on a specific model, the precise variables that should be used to test it in regressions
is not clear’. Despite the important insights provided by endogenous growth theory, the
empirical literature still seems to offer fairly provisional findings concerning some of the
more esoteric aspects of the theory. This stems mainly from the fact that endogenous growth
theory takes a rather circuitous route in incorporating novel factors viewed as fundamental
drivers of growth. The combination of neoclassical theoretical foundations and radical
theoretical additions results in interpretative difficulties in empirical models.
The strength of endogenous growth theory lies in its analytical sophistication in comparison
with the Solow model. Importantly, it introduced a systematic way of analysing the
interactions between education, research and innovation in the process of economic growth.
– 26 –
New Growth Theories and their Implications for State Government Policy Makers
However, the theory is limited in the extent of its applicability and relevance by the fact that
it deals with the determination of ahistorical equilibrium outcomes, not processes, even
though it is clear that economic growth is a process that unfolds in history. In the late 1990s,
this limitation led to the emergence of endogenous growth theories that attempted to be
more process oriented. These theories are even labelled Schumpeterian, e.g. those of Aghion
and Howitt (1998) and Weitzman (1998), to distinguish them from earlier endogenous
growth theories. However, Aghion and Howitt (1998), in 690 pages, do not refer to the rich
literature that exists on neo-Schumpeterian theories of economic growth or to the seminal
contribution by Nelson and Winter (1982). It is to this separate strand of new growth school
that we shall now turn.
Neo-Schumpeterian Growth Theory
It is also the case that a detailed analysis of neo-Schumpeterian growth theory will not be
provided here. However, a good place to start in accessing this literature is in Foster and
Metcalfe (2001), Nelson (1995) and, of course, Schumpeter (1912 and 1942).
As has been stressed, the main problem with endogenous growth theory lies in its analytical
foundations. It relies upon an optimising framework in which there are strong presumptions
concerning knowledge; yet the extensions to the Solow model involve questions that
presume the absence of complete knowledge. In reality, knowledge can never be complete
– investments in human capital accumulate knowledge, and the spillover effects involved in
new knowledge forever disturb the stock of effective knowledge. Inventors and
innovators/entrepreneurs cannot optimise simply because they operate in uncertain
contexts. This is why Schumpeter stressed the importance of the psychological disposition
of entrepreneurs so much. He made a fundamental distinction between the cognitive and
emotional dimensions of human behaviour, with the latter being crucial in the states of
uncertainty faced by entrepreneurs. Recent evidence in cognitive psychology (Damasio,
1995) confirms that Schumpeter’s insights were valid.
As has been noted, endogenous growth theory builds up from the notion that a production
function can summarise the supply side, in conjunction with a demand side mediated by
prices that maintain the economy in general equilibrium. Growth occurs because the
equilibrium production frontier is presumed to move over time. In contrast, the neoSchumpeterian approach is to concentrate not on sets of equilibrium outcomes but on
ongoing dynamic processes. If growth is indeed endogenous then this means that, over time,
it occurs without being fully determined by outside forces. Schumpeter conceived cyclical
upswings as developmental phases where the diffusion of innovations, through the
organisational efforts of entrepreneurs combining inputs of material, energy, capital and
labour, lead to increasing amounts of output. This output is heterogenous, as are the
organisational structures and the production processes that are employed. Consistent with
this, the neo-Schumpeterian approach is not to pretend that economic agents, products or
firms are identical but to accept, head on, the fact that there is considerable variety of all
kinds in the economy.
Indeed, it is this variety that is seen as the fundamental source of economic growth. Variety
in the stock of ideas is essential for the emergence of innovations. Primary
– 27 –
Productivity and Regional Economic Performance in Australia
‘macroinventions’ (Mokyr, 1990) lead to cascades of innovations that diffuse into new
processes and products employing less radical inventions. Thus, the innovation diffusion
process involves the selection of applicable ideas. This is not a process without end because
structural irreversibility and ‘lock-in’ ensure that the adoption of some innovations excludes
the adoption of others. Although profit seeking is evident, profit maximisation cannot be
said to be strongly in evidence. Innovation diffusion processes are explorations of
uncertainty and they rely upon individual imagination and a capacity to share novel ideas
and techniques with others, drawing heavily upon emotions as well as cognition.
Innovation diffusion gives rise to a large variety of processes and products. The second
stage of a growth process involves competitive selection whereby the most productive
processes and the most desirable products survive. Crucially, this does not lead towards a
state of perfect competition but rather monopolistic or oligopolistic conditions. This is
known as the process of replicator dynamics whereby the most productive processes and the
most desirable products come to dominate without the necessity of universal optimisation.
When processes of innovation and competition are in operation, they trace out growth paths
that follow sigmoid curves such as the logistic curve or the Gompertz curve. This provides
a formal representation of a growth process over time (see Figure 2.2). However, although
any mathematical function must be deterministic, it does not imply underlying structural
stability because it summarises a historical process of continual structural change. There is
no equilibrium, or disequilibrium, involved. Being historical, the function traces a nonequilibrium path. Thus, when growth eventually drops to zero, the system often enters a
structurally unstable state.
Figure 2.2: The Neo-Schumpeterian Logistic Growth Path
Output
(Y)
new growth path
K
slow decline
fast
decline
0
structural
instability
structural
instability
structural
stability
Time
However, although neo-Schumpeterian growth theory is concerned with non-equilibrium
processes, at any point on a diffusional growth trajectory there will be tendencies that will
bring growth back on track when there are exogenous shocks. This is a normal homoeostatic
– 28 –
New Growth Theories and their Implications for State Government Policy Makers
feature of all dissipative systems. Crucially, however, the ‘basin of attraction’ involved will
vary in shape at different points on the growth trajectory and it is this that is the key to the
extent of structural stability (Foster and Wild, 1999). When a structural transition occurs, it
can result in the demise of the system, a shift to a new logistic growth path or the emergence
of a core role in the facilitation of the growth of other systems. We can specify the form of
the logistic curve mathematically. For example, the Mansfield version, used widely in
studies of innovation, is as follows:
Yt = Yt-1 - Yt-1b [1 - (Yt-1 /K)]
(3)
where b is the diffusion rate, K is the capacity limit and t denotes a time interval.
It is possible to take account of exogenous effects, suggested by conventional economic
theory, in the neo-Schumpeterian framework. These can impact upon the rate of diffusion
or the limit to which growth tends:
(Yt - Yt-1) / Yt-1 = b(...)[1 - {Xt-1 /K(...) - a(...)}]
(4)
where both the diffusion rate and the capacity limit can now be influenced by other
variables and there is an a(...) function included to allow for competitive influences.
Because the growth process is presumed to be primarily endogenous, these neoclassical
effects, which can be captured in the augmented logistic diffusion model in equation (4),
have a secondary impact, operating in a similar way to that suggested by Alfred Marshall a
century ago (Foster, 1993). In Foster and Wild (2000), it is shown that treating neoclassical
effects in this way can result in them being more clearly defined than in a conventional
neoclassical model, essentially because they have less analytical work to do.
In addition to embracing traditional neoclassical concerns about the role of prices, neoSchumpeterian theory can also incorporate the factors highlighted in endogenous growth
theory. More R&D will yield more ideas and more inventions and innovations. Education
will also lead to more growth because it will provide more core competencies, as well as
essential variety in the collective stock of imaginative ideas. The development of better
market institutions and credit markets will extend the limits to which growth processes tend
and will facilitate more adaptive behaviour. An effective competition policy will ensure
both exit and entry in mature industries.
A crucial difference between endogenous growth theory and neo-Schumpeterian growth
theory is that the latter embraces uncertainty from the outset. Optimisation is an
unsatisfactory method of depicting processes of invention, innovation and entrepreneurship
despite the fact that it enables us to write down tractable mathematical models. In Romer’s
(1990) three-sector model, researchers optimise but, in reality, only a small fraction of ideas
will ever have any commercial value. The essence of research is that a significant amount
of effort comes to nothing in order that a few ‘gems’ are uncovered. The same is true of
innovation but to a lesser extent – only a few innovations survive the competitive test.
In Romer’s model, the perfectly competitive final goods sector produces one good with
different capital goods and varying amounts of labour. This is perhaps the most serious
deficiency of the model. In reality, the final goods sector is full of product variety and it is
the choices of consumers that are the most decisive selection force. To omit this, and along
– 29 –
Productivity and Regional Economic Performance in Australia
with it the massive marketing efforts involved in promoting products in circumstances of
uncertainty, leaves the model fundamentally deficient in attempting to understand the
processes involved in economic growth in any detailed sense.
In recognising the existence of uncertainty and the fact that economic decisions are still
made when it is present, neo-Schumpeterian growth theory is cast in reality, not in terms of
a timeless abstraction. The notion that underlies much of the microeconomic reform agenda,
that all policy makers need is a familiarity with a set of simple general theoretical principles
that are applicable in all situations, is rejected. Each firm, industry, region and economy is
unique, with a particular interface with political, cultural, institutional and environmental
conditions. To understand why growth occurs, and why it occurs at a particular pace and is
subject to limits and bifurcations, requires a very special knowledge of the process that is
under investigation. In this regard, a particular strength of the neo-Schumpeterian approach
is that it interfaces closely with a large empirical literature that exists on the economics of
innovation (see Bryant, 2001). This literature deals with particular cases in great detail
while, at the same time, the logistic curve that is central to neo-Schumpeterian growth
theory plays a central role.
Economic Policy Principles
Rational economising is relevant to conditions where there is quantifiable risk. Uncertainty,
in the Knightian sense that there is a set of known outcomes that is incomplete and the
probability assigned to each outcome is inaccurate, leads to behaviour that is driven by
emotions (Foster, 2000). This is important because it results in variety and selection, but
there is no guarantee that there will be outcomes that are valuable in an economic sense.
Government, as the representative body of society, must strive to ensure that conditions are
such that economic decisions can be made in the presence of the opportunities and threats
that characterise uncertain situations. Uncertainty must be translated into quantifiable risk.
For example, patent laws have to be designed and enforced before inventing can be
transformed from an enjoyable hobby into an activity where economic returns can be
entertained. Market rules and contract laws must be devised and policed before
entrepreneurs can imagine that their opportunistic schemes can turn into wealth.
Dealing with uncertainty is as much an art as a science because it involves the acquisition
of a detailed understanding of emerging possibilities and a capacity to act effectively to
assist the creative efforts of economic agents. In mainstream economics, it is often argued
that government should act when there is market failure. However, this is a very limited way
to look at government intervention – in uncertain situations, the market mechanism does not
fail but is simply absent because the required conditions for its existence are not yet fully
present. Market emergence would be a better term, with market failure being left for
situations where, for example, the onset of monopolistic power in a mature industry results
in the weakening of market forces.
For government, there are two difficult issues to deal with. First, because invention and
innovation must necessarily involve both unproductive activity and unmeasurable
spillovers, conventional measures of inefficiency and waste cannot apply. Thus, it becomes
very difficult to distinguish genuine failure from rent-seeking when R&D subsidies are used
– 30 –
New Growth Theories and their Implications for State Government Policy Makers
by governments. Also, the presence of uncertainty means that conventional benefit–cost
methods, designed for situations of quantifiable risk, can be misleading. However, this does
not mean that such methods should not be used, since they help to clarify the budgetary
implications of policy initiatives and can provide baseline estimates upon which discussion
and debate can be based. Therefore, such studies must be viewed as playing an exploratory
rather than a definitive role in policy making.
The second issue that must be addressed is that the encouragement of competition may be
appropriate or inappropriate, depending upon the stage of development that an industry
happens to be in. Furthermore, when it is appropriate, policy makers must recognise and
deal with the fact that monopolistic conditions will be the eventual result, despite the
productivity gains that are likely to be enjoyed as competition operates. Considerable
challenges are posed for policy makers in the face of such dynamic complexity.
A key role for government is in the introduction, adaptation and removal of regulations and
laws, as the needs of the economic system change. This requires the development of an
endogenous capability to anticipate the future direction of economic evolution in particular
cases from current and past tendencies. To do this requires an understanding of where in
their particular growth trajectories firms and industries lie. Thus, protective and facilitating
regulations may be required in emerging industries that would be highly inappropriate in
mature industries. The neo-Schumpeterian perspective suggests that the need to nurture
emergent creativity must be counterbalanced by a willingness to allow mature firms and
industries to experience crises, failures and strong pressure from new entrants. Uncertainty
must be reduced for the emergent and certainty must be challenged in the case of the mature
and powerful – this is the essence of promoting ‘creative destruction’.
In the case of corporate failures, government must be prepared to deal with the severe
uncertainty that emerges among redundant employees. This is something that has been clearly
lacking in the case of recent failures of companies rich in human capital, such as Ansett and
OneTel. The neo-Schumpeterian perspective is that the disruption that these people experience
is crucial to the process of economic growth, and therefore they must be strongly supported
by government in a range of ways when failures occur. The neo-Schumpeterian general rule
of policy making is always the same: try to anticipate and deal effectively with emerging
uncertainty by creating conditions in which quantifiable risk makes rational economic choices
possible. Failure to do this results in damaging inertia, the exercise of monopoly power,
significant losses of new entrepreneurial opportunities and serious losses of human capital,
both personally and socially, when corporate failures occur. An intimate understanding of
firms and industries and where they are on their evolutionary trajectories really does make
history matter when enacting policies to promote economic growth.
Of course, there are ‘moral hazards’ involved when government commits itself to assist
some of those rendered unemployed by the process of creative destruction. With every day
that passes, small businesses fail and personal responsibility must be assumed for
commercial misjudgments. However, large corporate failures can involve severe
dislocations of people who possess very specialised forms of human capital that can only
be traded on very thin and incomplete labour markets. The provision of support and
assistance to these people, in their attempts to regain employment or to retrain, can often
yield significant social as well as personal returns. Policy rules must be developed to
– 31 –
Productivity and Regional Economic Performance in Australia
identify conditions where government intervention is justifiable. In this regard, the
experience of Sweden provides a useful starting point.
Having summarised some general principles concerning economic policy making in
relation to promoting economic growth, we can offer a more detailed articulation of these
principles. Here, as has been pointed out, although endogenous growth theory stresses the
importance of the number of people engaged in R&D and the number of effective ideas, i.e.
patents, it is necessary to turn to neo-Schumpeterian growth theory for specific guidance. In
this regard, it should be noted that many of the neo-Schumpeterian perspectives on growth
policy discussed are not new because they have been enunciated in the context of innovation
policy (see Bryant, 2001). The following seem relevant:
• Knowledge and physical infrastructure investments must be made in a timely and
flexible manner to facilitate current and future entrepreneurship and innovation while
infrastructure projects with definable risks, being commercial ventures, should be
undertaken independently or in partnership with the private sector.
• Institutions such as new market and contractual arrangements emerging from the private
sector facilitate technological and organisational innovation, and should be translated
into legal forms in close consultation with the innovators themselves.
• Policies that foster cooperation must be used to increase networking, collaborative
R&D, joint market research and marketing among small firms in emergent industries.
• Large maturing organisations experiencing technological or organisational ‘lock-in’,
which can result in inertia and the exercise of power, should be monitored to identify the
emergence of monopolistic practices and, if necessary, be provided with advice
concerning adaptive strategies (see Richard Foster, 1986).
• Effective competition policies must exist to prevent mature firms from colluding to
extract economic rents from the exercise of power and the erection of barriers to entry.
• The high failure rate of small enterprises (Hannan and Freeman, 1989) that is necessary
for the development of successful and adaptive industries should involve increased
commitment by government, beyond the protection offered by bankruptcy laws, to the
creation of new opportunities for failed entrepreneurs.
• Failing firms should not be subsidised by government to continue in production.
Retrenched employees should instead be compensated for the loss of specific human
capital and be supported to re-skill and explore new capabilities where this is possible.
• Subsidies to emergent firms, who have limited access to credit because of high levels of
uncertainty, should be confined to ‘capability building’ through improved access to
technologies, organisational advice and R&D assistance. Support of this kind often
involves only small amounts of funding.
• Since the bulk of specific human capital training is acquired on the job, vocational
training should primarily be in parallel with employment. Because vocational training is
related to well-defined returns, it should to a significant degree be charged for at market
rates, borne by employers and/or employees.
– 32 –
New Growth Theories and their Implications for State Government Policy Makers
• General education yields non-specific human capital, and therefore private returns to it
are uncertain. However, the spillovers are very significant, with resultant core skills and
variety in ideas generating high social returns. Thus, government should subsidise
general education heavily but demand very high standards in literacy, numeracy and
analytical thinking at every level.
Running through these principles is a key Schumpeterian message: when there is creative
destruction, governments should pay at least as much attention to the process of destruction
as to the process of creation. The destruction of firms and industries need not lead to chaos
but rather to the necessary conditions for further creativity. The systems that governments
confront in the economy are all dissipative structures, that is, they import energy and export
entropy, or their money equivalents, to survive. In this regard, neo-Schumpeterian growth
theory is connected with systemic principles enunciated in non-equilibrium
theormodynamics. Thus, ‘entropy sinks’ are just as important as new energy sources, or
their equivalent, if systems are to grow and prosper (Burley and Foster, 1992). This must be
borne in mind in all practical policy contexts, particularly with regard to the release of
labour with obsolete skills from organisations.
Something that becomes apparent in thinking about these principles is that the sharp
distinction made in endogenous growth theory between human capital formation and the
process of invention/innovation is somewhat artificial. Indeed, education policy and
industrial policy are intimately related. It is for this reason that the distinction is not made so
sharply in neo-Schumpeterian growth theory. However, both strands of new growth theory
view the stock of human capital and related education policies as very important in
understanding economic growth. Therefore, it is useful to use education policy as a case
study to demonstrate how new growth theory can lead to a re-orientation of economic policy.
The Case of Education Policy
In recent years, economic policy makers in most OECD countries have become increasingly
interested in education and training policies because of the evidence, contained in a
burgeoning literature concerning the drivers of economic growth, that ‘human capital’ is a
major determinant of the relative economic performance of countries over time. Human
capital theory has tended to be used to analyse private decisions to invest in education and
training, rather than the public provision of education. Human capital theorists have seen
the latter in two contrasting lights. First, ultra-economic rationalists have viewed public
sector education and training as inefficient and wasteful because it is not sufficiently
exposed to the rigor of the market and competition – privatisation is recommended. Second,
human capital theorists, more comfortable with the notion of the mixed economy, have
argued that, for a range of reasons, the demand and supply of human capital by the private
sector is subject to market incompleteness, market imperfections, market failures and
externalities that strongly justify public involvement. However, irrespective of how efficient
public sector delivery is, it has been clear since the very early studies, such as the pioneering
study undertaken by E. Dennison in the 1950s of the determinants of economic
performance, that publicly funded education and training has had a very large role to play
in the process of economic growth.
– 33 –
Productivity and Regional Economic Performance in Australia
A good example of the application of human capital theory in education policy can be found
in the Higher Education Contribution Scheme (HECS) system designed by Bruce Chapman,
a prominent Australian labour economist, in the late 1980s. In levying a private charge for
tuition in higher education, 20% of the return to graduates was deemed to be private while
the other 80% was viewed as a spillover return to society as a whole. This judgment was
based on a well-established literature that existed at that time on the private and social
returns to education. (It is also the case that roughly 80% of higher education costs in the
United States are borne by government.) Under the Coalition, the private HECS
contribution was raised to an average of 30%, primarily with budgetary considerations in
mind. The upshot was that Australian government spending on higher education as a
proportion of GDP became relatively low in comparison with other OECD countries.
However, there has been little indication of a consequent fall in demand for higher
education, either immediately after the introduction of HECS or following the later increase
in charges, indicating that there appears to be some foundation for the notion that
individuals perceive, to a significant extent, participation in higher education as a human
capital investment decision. This indicates that, particularly in highly vocational degree
programs, such as medicine, individuals are making choices that suggest they perceive
quantifiable risk rather than uncertainty. From a neo-Schumpeterian perspective, it is thus
appropriate that the provision of highly vocational degrees should involve significant
private fee payments. The difficulty that arises with education policy lies in non-vocational
education where returns are uncertain but potential spillovers are high.
The human capital perspective has not had much of an impact upon the design of Australian
education policy below the tertiary level, i.e. at the state level. Traditional goals to ‘educate
people for life’ have dominated the agenda in public education. Included in these goals there
has, of course, been a desire to equip people with the analytical skills, communication skills
and knowledge of the economic system essential for entry into employment at all levels.
However, there has been little in the way of systematic attempts to weigh up the economic
costs and benefits of different forms of education in order to assign priorities. Neither has
there been much willingness to compare educational performance either between or within
schools in a consistent way. In the United States, this has also led to perceived difficulties,
as Heckman (1999, p. 20) states:
Public school systems in the US are local monopolies with few competitors.
The American high school system is a creation of the 20th Century and it is a
world unto itself. Within it, an artificial adolescent culture is left to flourish
which often discourages academic achievement and the pursuit of knowledge
even in the best schools in the best neighborhoods ... The incentives of many
principals and teachers to produce useful knowledge, or any knowledge at all,
are weak although there are surely many dedicated professionals. They are
often unresponsive to the changing demand for skills or to the market realities
that will confront their students when they leave school. They are not
accountable to anyone because it is not easy to monitor them. One valuable
source of information – parental and student perception of the qualities of
teachers and schools – is rarely used to punish poor teaching. The educational
technocrats dismiss them as ‘subjective’ and unreliable.
– 34 –
New Growth Theories and their Implications for State Government Policy Makers
However, there are considerable problems in measuring the economic benefits of education in
detail because of the spillovers involved. Understandably, educationalists fear that, given the
diffuse and difficult to quantify benefits of non-vocationally oriented education, the economic
perspective leads too easily to policy prescriptions that involve shifting resources into
vocationally oriented education and training. This is a serious issue because neo-Schumpeterian
growth theory tells us that what seems to be non-vocational ‘education for life’, which enriches
the artistic and creative knowledge and analytical skills of individuals, can have a large impact
upon the economy since it can be the source of the diverse entrepreneurial and innovative
behaviour that lies at the very foundation of the economic system.
When some economists think of non-vocational education, they see it as involving
investments in ‘social and cultural capital’. However, this is a relatively new perspective in
the mainstream of economics that has not taken much hold, largely because of the
quantification problems alluded to. Thus, there is some basis for the fears of educationalists
concerning the application of economic analysis in education policy contexts. However,
these do not constitute a justification for not examining the reasons why there is, for
example, an increasing drain of students from public to private schools. Concerns with
declines in teaching quality, educational standards and educational attainment (particularly
among males) seem to be important causes of this shift and there is a clear case for
economic evaluations of the costs and benefits involved, at the very least to provide some
benchmarks to aid debate and discussion.
Problems of any kind in the educational process are, of course, of great concern to education
policy makers. What has catapulted education into the wider domain of economic policy in
the last decade has been the emergence of endogenous growth theory.
As has been noted, its proponents take a wider view of the role of human capital than most labour
economists, who have tended to focus upon the private vocational possibilities of education and
training (i.e. that which an individual could reasonably associate with a profession and an
estimated income flow). Both strands of new growth theory emphasise two dimensions of
knowledge accumulation and associated skill formation: invention and innovation. Traditional
human capital thinking is relevant to the latter – we need to train engineers who are able to take
new ideas and turn them into viable production systems and new products – but the former
requires creativity, which is difficult to teach in a direct and systematic manner. Inventions come
from diversity – this is the ‘variety’ that evolutionary economists, such as Metcalfe (1998), have
always stressed is the foundation of all economic growth.
However, very few creative people succeed in producing ideas of significant economic
value (i.e. from a static economic viewpoint, it seems very wasteful to have a large number
of people engaging unsuccessfully in creative activities but the opposite can be true from a
dynamic perspective). Even when they do succeed, most capture only a small amount of the
benefit. Attempting to be highly creative does not make much sense in traditional human
capital investment terms, so social support of such endeavours at some agreed level is
essential. Critical in this is the provision of an education system where non-vocational
learning can generate key literacy and analytical skills that can, in turn, provide the basis
for the production of essential variety in the social stock of knowledge. As has been noted,
endogenous growth theory does not deal with this in a very satisfactory manner, so it is
necessary to turn to neo-Schumpeterian growth theory for an analytical framework.
– 35 –
Productivity and Regional Economic Performance in Australia
Once we accept that different types of human capital play different roles in the ‘knowledge
economy’, it is necessary to rethink how they fit together. Foster (1999) offers a neoSchumpeterian taxonomy concerning all investment behaviour that can be adapted to deal
with human capital investment.
From this perspective, we can think of four types of human capital investment expenditure:
1. Inventive investment: those investment expenditures in ideas that can result in new
inventions, new organisational ideas and new forms of human capital. General education
in abstract logic, language and communication skills, the sciences and the arts will be
important components of this type of investment.
2. Innovative investment: investment that develops a capacity to transform ideas into
economically valuable processes and products. The acquisition of specialised
knowledge and skills, for example in engineering or architecture, permit technological,
organisational and aesthetic improvements that can generate rising productivity and
economic value.
3. Maintenance investment: expenditure on the knowledge and skills necessary to keep
economic systems and sub-systems going. In addition to production tasks themselves,
many of the other routine tasks in firms in, for example, purchasing, production control,
inventory control, maintenance scheduling and equipment repair fall into this category.
Despite the routine nature of much of this activity, there are ‘learning curve’ efficiency
gains available over time.
4. Strategic investment: expenditure on knowledge and skills that help to secure and defend
economic power. These skills are political in character and enable, for example, strategic
collusion between firms, an understanding of how to construct entry barriers, knowledge
of how to eliminate competitors and an understanding of marketing strategies that
extract economic rents from consumers.
At any level of aggregation, there will be flows of investment in human capital in all four
categories. The economic impact of human capital investment will depend on the relative
importance of each and how capital in each combines and interacts.
Category (1) investments have very uncertain outcomes from a microeconomic perspective,
but it is the effects on overall economic performance that are fundamental. They underpin the
effectiveness of other forms of investment. Category (2) investments involve the acquisition
of skills that allow organisational, process and product innovations to occur, demanding
high-level engineering, design, and scientific knowledge and skills. The productivity that
emanates from the application of such human capital tends to rise along logistic diffusion
curves. Category (3) involves essential investments that allow the fruits of the innovation
process to yield products and services in efficient ways – operational and managerial skills
that maintain and improve efficiency in a myriad of ways are always required. The efficiency
gains that flow from the application of the resultant human capital can be observed and
measured in various ways by economists. Category (4) investments provide a stock of human
capital that fuels strategic behaviour, both cooperative and non-cooperative, in the economic
system. The result of this may not be productivity enhancing in the short term and may lead
to the power-related redistribution of income and wealth. However, in the long term, this kind
– 36 –
New Growth Theories and their Implications for State Government Policy Makers
of behaviour can provide the basis for the operation of selection mechanisms that open up
new opportunities for those who have invested in (1) and (2).
The balance of these human capital investments determines long-term economic
performance. If all the emphasis is on (1) without much innovation potential, an economy
will not prosper. However, new growth theory stresses that it is possible for a particular
economy to grow without much (1), provided that there is knowledge and skill transmission
into (2) and (3) from outside. An excess of (3) can result in a routinised economy that overemphasises static efficiency, which ultimately declines because of a shortage of (1) and (2)
and a tendency for investments in (4) to rise. If category (4) is the dominant form of human
capital investment in an economy, economic performance is likely to be poor and/or
punctuated by crisis.
There has been a tendency by economists to see economic performance and growth in terms
of the size of the physical capital stock and its changing technological capabilities.
However, the ultimate source of economic growth must be human capital and the pace of
economic growth will be influenced by the composition of human capital across our four
categories. In particular, just how education and training is spread across (1), (2) and (3) will
be important. Each economy will have different needs at different points in time, depending
on economic structure and how it is changing. Policies on education and training must also
take account of the impact of the ‘learning by doing’ that takes place in the course of
employment, since this will differ in different industries as they rise and decline.
In an economy that produces large quantities of heterogeneous services, this source of
productivity will be very significant. In this regard, the emphasis on general rather than
specific education in the United States, right up to undergraduate level, would seem to be
appropriate to the service dominated economy that has evolved. However, both Romer (2000)
and Heckman (1999) argue that there is a compositional problem in the US economy, whereby
the general nature of education has led to too few top specialists in science, engineering and
the professions. Since this is accompanied by the production of too many ‘low quality’
generalists, they argue that the result is the observed widening of the remuneration gap.
Although these economists are concerned with the balance of different human capital
investments, they do not approach it from a neo-Schumpeterian perspective, which might
suggest that the problem lies not with too few scientists and engineers but rather with the
relatively low quality of general education at the sub-tertiary level. Answers to such questions
remain unclear and will remain so until much more research is undertaken in this field.
Concluding Remarks
In this chapter, it has been argued that the formulation of economic policy to target the rate
of economic growth requires new theoretical perspectives to guide government intervention
in the economy. The two strands of new growth theory offer such perspectives. Both stress
that economic growth is to a large extent an endogenous process but one in which
government has a crucial role to play. Endogenous growth theorists talk in terms of
government compensating for ‘market failures’ while neo-Schumpeterians see the role of
government more widely in relation to the uncertainties that people face in making rational
economic decisions.
– 37 –
Productivity and Regional Economic Performance in Australia
The ‘big spending on big projects’ mindset that characterised the Keynesian era in policy
making is being superseded by a view that does not see the economy, or the people who
participate in it, in homogenous and quantitative terms. Qualitative considerations
concerning the degree of variety that exists and the extent to which it can lead to innovation
and productivity growth are increasingly emphasised. Thus, areas traditionally separated
from economic policy, such as education policy, have become strongly linked. In turn,
education policy and industrial policy are becoming viewed as complementary. Government
priorities will increasingly become targeted not on deregulation but on regulatory
innovation that assists creative individuals and groups to produce new products and operate
in new markets. Government spending decisions will be geared increasingly to the emergent
needs of an evolving economic system and to the enhancement of quality, for example in
education, rather than the simple attainment of quantitative targets, such as total numbers of
year 12 completions or tertiary students.
The move to longer term policy priorities, which are uncertainty-reducing and therefore do
not show quick or clearly defined returns, will be difficult for all governments because of
the financial viability requirements that now have to be demonstrated in relation to all
spending proposals. Also, the bureaucratic structure of the public service, originally set up
for fairly routinised service delivery activities, is a difficult context for creative and openended policy making initiatives. In addition, the pressure brought to bear by the short-term
political goals of incumbent governments provides an added constraint on the pursuit of
longer term economic objectives. In many respects, it is likely that a radical policy agenda
will not be forthcoming until politicians of all colours can form some sort of a consensus on
policy priorities.
Unfortunately, the recent era of microeconomic reform has culminated in a distinct lack of
consensus on economic policy priorities within both of the main political parties, so it will
not be easy for a new consensus to emerge, no matter how novel the economic theory
involved might be. However, neo-Schumpeterian thinking can be extended to the policy
formation process itself. Such a process also involves a certain kind of entrepreneurship,
innovation diffusion and competitive selection processes. Provided that democracy and
other political processes remain healthy and open, there is no reason why creative
destruction cannot also give us new policies to deal with the problems and challenges we
face in achieving a stronger, fairer and more sustainable economy.
– 38 –
New Growth Theories and their Implications for State Government Policy Makers
References
Aghion, P. and Howitt, P. (1998), Endogenous Growth Theory, MIT Press, Cambridge, MA.
Ahn, S. and Hemmings, P. H. (2000), Policy Influences on Economic Growth in OECD
Countries: An Evaluation of the Evidence, OECD Economics Department Working Paper
No. 246.
Bryant, K. (2001), ‘Promoting innovation: An overview of the application of evolutionary
economics and systems approaches to policy issues’, in J. Foster and S. Metcalfe (2001),
Frontiers of Evolutionary Economics: Competition, Self-Organisation and Innovation
Policy, E. Elgar, Cheltenham, United Kingdom, 361-384.
Burley, P. and Foster, J. (1992), Economics and Thermodynamics: New Perspectives on
Economic Analysis, Kluwer, Boston.
Damasio, A. (1995), Descartes Error: Emotion, Reason and the Human Brain, Putnam, New
York.
Durlauf, S.N. (2001), ‘A manifesto for a growth econometrics’, Journal of Econometrics,
100, 65-69.
Durlauf, S.N. and Johnson, P.A. (1995), ‘Multiple regimes and cross-country growth
behaviour’, Journal of Applied Econometrics, 4 (10), 365-384.
Foster, J. (1993), ‘Economics and the self-organisation approach: Alfred Marshall
revisited?’, The Economic Journal, 103, 975-991.
Foster, J. (1999), ‘The rise of unemployment and working poverty: An evolutionary
macroeconomic perspective’, in M. Setterfield (ed.), Rapid Growth and Relative Decline:
Modelling Macroeconomic Dynamics with Hysteresis, St Martin’s Press, New York.
Foster, J. (2000), ‘Is there a role for transaction cost economics if we view firms as
complex adaptive systems?’, Contemporary Economic Policy, 18, 369-387.
Foster, J. and Metcalfe, S. (2001), Frontiers of Evolutionary Economics: Competition,
Self-Organisation and Innovation Policy, E. Elgar, Cheltenham, United Kingdom.
Foster, J. and Wild, P. (1999), ‘Econometric modelling in the presence of evolutionary
change’, Cambridge Journal of Economics, 23, 749-770.
Foster, J. and Wild, P. (2000), ‘Detecting self-organisational change in economic
processes exhibiting logistic growth’, Journal of Evolutionary Economics, 9 (1999)
109-133. Reprinted in U. Cantner, H. Hanusch and S. Klepper (eds) (2000), Economic
Evolution and Complexity, Physica-Verlag, New York, 159-184.
Foster, R. (1986), Innovation, Summit Books, New York.
Grossman, G. and Helpman, E. (1991), Innovation and Growth in the Global Economy,
MIT Press, Cambridge, MA.
Hannan, M.T. and Freeman, J. (1989), Organisational Ecology, Harvard University Press,
Cambridge, MA.
Heckman, J.J. (1999), Policies to Foster Human Capital, NBER, Working Paper No. 7288.
– 39 –
Productivity and Regional Economic Performance in Australia
Industry Commission (1995), The Growth and Development Implications of Hilmer and
Related Reforms AGPS, Canberra.
Jones, C.I. (1998), Introduction to Economic Growth, Norton, New York.
Jorgenson, D.W. (1988), ‘Productivity and postwar U.S. economic growth’, Journal of
Economic Perspectives, 2, 23-41.
Leibenstein, H. (1966), ‘Allocative vs X-efficiency’, American Economic Review, 56, 392-415.
Lucas, R.E. Jr (1988), ‘On the mechanics of economic development’, The Journal of
Monetary Economics, 22, 3-42.
Mankiw, G.N., Romer, D. and Weil, D.N. (1992), ‘A contribution to the empirics of
economic growth’, Quarterly Journal of Economics, 107, 407-437.
Metcalfe, J.S. (1998), Evolutionary Economics and Creative Destruction, Routledge, London.
Mokyr, J. (1990), The Lever of Riches, Oxford University Press, New York.
Nelson, R. (1995), ‘Recent evolutionary theorising about economic change’, Journal of
Economic Literature, 33, 48-90.
Nelson R. and Winter S. (1982), An Evolutionary Theory of Economic Change, Belknap
Press, Cambridge.
Quah, D. (2001), Technology Dissemination and Economic Growth: Some Lessons for the
New Economy, Mimeo, Economics Department, LSE, September.
Quiggin, J. (1998), ‘Micro gains from micro reform’, Economic Analysis and Policy, 28, 1-16.
Romer, P. (1986), ‘Increasing returns and long-run growth’, Journal of Political Economy,
94 (5), 1005-1037.
Romer, P. (1990), ‘Endogenous technical change’, Journal of Political Economy, 98, 71102.
Romer, P. (1994), ‘The origins of endogenous growth’, Journal of Economic Perspectives,
8, 3-22.
Romer, P. (2000), Should the Government Subsidise Supply or Demand in the Market for
Scientists and Engineers?, NBER Working Paper No. 7723.
Schumpeter, J. (1912), The Theory of Economic Development, Cambridge University
Press, Cambridge, United Kingdom.
Schumpeter, J. (1942), Capitalism, Socialism and Democracy, Harper, New York.
Solow, R.M. (1956), ‘A contribution to the theory of economic growth’, Quarterly Journal
of Economics, 70 (February), 65-94.
Solow, R.M. (1957), ‘Technical change and the aggregate production function’, Review of
Economics and Statistics, 39 (August), 312-320.
Summers, R. and Heston, A. (1991), ‘The PENN World Table (mark 5): An expanded set of
international comparisons 1950-1988’, Quarterly Journal of Economics, 106, 327-368.
Weitzman, M.L. (1998),‘Recombinant growth’, Quarterly Journal of Economics, 113, 331-360.
– 40 –
3 Variations in Economic and Labour
Productivity Growth among the States
of Australia: 1984-85 to 1998-99 1
Duc-Tho (Tom) Nguyen, Christine Smith and Gudrun Meyer-Boehm
Introduction
How successful have individual Australian states been in terms of their economic and
labour productivity (LP) growth relative to the other states? Can any given state lift its game
above the others by means of either judicious government policy or superior private sector
performance? Is there much scope for the poorer performers to improve their relative
positions? Questions such as these are potentially of great interest to policy makers and
advisers in state and federal governments, business leaders, academic researchers, and the
public at large. Indeed, it appears that in recent years there has been an upsurge in Australia
of public interest in the topic of income distribution generally, and more particularly in the
finding that income disparities may have widened considerably; see, for example, the
editorial published by the Weekend Australian on 17-18 June 2000.
In this chapter we examine differences in the economic and LP growth performances of the
six states of Australia over the period 1984-85 to 1998-99. Our analysis builds on and
extends previous studies in this area in several ways. First, we examine LP as well as per
capita income. In so doing, we find that while considerable cross-state variations exist in the
growth rates of gross state product (GSP) per capita, the rates of LP growth have been far
more similar to one another. The discrepancy between these two pictures has been due
mainly to demographic changes.
Second, we investigate the industrial structure of each state, as well as the differences across
states in terms of LP growth within each industry. Our aim is to assess whether structural
differences or differential growth rates have been the major cause of overall interstate
variations in LP growth. Our results suggest that the former factor has been dominant during
the period studied.
Third, we examine recently available data and find that, contrary to international and
historical experiences, the levels of GSP per capita and LP in the various states of Australia
have tended to diverge over the past 15 years. When mining is excluded, however, the
pattern that emerges is one of neither convergence nor divergence – instead, we are left with
a set of remarkably similar growth paths. This leads to a number of interesting policy
implications, some of which may not have been immediately obvious before.
1
This chapter is based largely on a paper with the same title which was presented to the 2000 Conference of Economists, held by the
Economic Society of Australia at the Gold Coast on 3-6 July 2000. The authors wish to thank other members of the Drivers of Economic
Growth project for helpful discussions. Special thanks are due to Jim Hurley and Christine Williams for their assistance with data
sources and interpretations. Responsibility for the views expressed and for any remaining errors rests with the authors only.
– 41 –
Productivity and Regional Economic Performance in Australia
The chapter is organised as follows. The first section presents a brief review of previous
results and some background information. In the second section, we examine aggregate (allindustry) data relating to variations in economic and LP growth across states, while the
third section extends this type of interstate analysis to the level of individual industries. The
fourth section takes a different approach to the problem: here we examine differences in the
national (all-state) rates of growth across industries. The final section contains a summary
of the main results and draws out some policy implications.
Previous Results and Background
Harris and Harris (1992) examined differences in the rates of economic growth and levels
of GSP per head of population in the states of Australia during the period 1953-54 to 1990-91.
They found that real GSP per head for New South Wales and Victoria stayed above the
national average throughout this period, while for Queensland, South Australia and
Tasmania it stayed below the national average. For Western Australia real GSP per capita
went from below average to above. There was a general pattern of convergence, with
deviations from the national mean of GSP per head in most states tending to become
smaller, except for Tasmania. At the same time, states in the below-average group tended to
experience higher rates of growth.
This type of cross-sectional convergence pattern has been the subject of much research and
discussion in the comparative economic growth literature. For a very small sample of this
literature, see Baumol (1986), Abramovitz (1986), Dowrick and Nguyen (1989), Barro and
Sala-i-Martin (1992), Quah (1993), Sala-i-Martin (1996) and Bernard and Jones (1996). In
contrast with the inter-country studies, where both convergence and divergence have been
observed, interregional investigations have thus far tended to yield findings of convergence.
As Sala-i-Martin (1996, p. 1325) concluded:
The empirical evidence on regional growth and convergence across the United
States, Japan, and five European nations ... confirm[s] that the estimated
speeds of convergence are surprisingly similar across data sets: regions tend
to converge at a speed of approximately two percent per year. We also show
that the interregional distribution of income in all countries has shrunk over
time.
It may be that sub-national regions are more likely to share a common technological
framework, a common culture, and common legal, social and other characteristics, which
would allow them to experience convergence in productivity more readily than is the case
with different nations or groups of nations.
In this literature a distinction is often made between two key aspects of the convergence
phenomenon. Using the terminology first proposed by Sala-i-Martin (1990), beta
(`)-convergence is said to exist when, among a cross-section of economies, there is a
negative relationship between the initial level of income (or LP) and the rate of growth in
income during the ensuing period. If `-convergence holds, poorer economies will tend to
grow more rapidly than richer economies. By contrast, sigma (m)-convergence is said to
hold when the cross-sectional dispersion of income (or LP) declines over time. If
m-convergence holds, income disparities across the relevant economies will tend to
– 42 –
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
diminish. While these two concepts of convergence are related, they are not identical. In
particular, `-convergence is a necessary, but not sufficient, condition for m-convergence (Salai-Martin 1996, pp. 1328-9). In terms of this terminology, Harris and Harris’s (1992) results
are consistent with both `- and m-convergence.
Cashin (1995) examined a much longer period than Harris and Harris, namely from 1861 to
1991, and subjected the data to formal statistical testing. (He also included New Zealand as
a former Australasian colony.) He found evidence of both `- and m-convergence in the per
capita income levels of the seven economies. However, most of the decline in crosssectional dispersion had occurred in the nineteenth century. During the subsequent 90-year
period (1901-1991), dispersion did not display a clear downward trend.
Neri’s (1998) study was essentially a re-examination of Cashin’s analysis, but with an
alternative data set, and with the differences across sub-periods being examined more
closely. Neri also included the Northern Territory and the Australian Capital Territory as
separate units of the cross-section, and dropped New Zealand. While confirming Cashin’s
observations regarding both types of convergence in the sub-periods prior to 1976, Neri
emphasised that from the mid 1970s to the early 1990s, there was neither convergence nor
divergence in the ` sense, while there was clearly a rise in cross-sectional dispersion of
income per capita (evidence of m-divergence). He suggested that this widening of the
income gaps between states has been due mainly to the ability of the more successful states
to adapt to national and global changes through changes to their sectoral compositions.
The above studies were all based on analyses of data for GSP per head of population. Yet
movements in population need not be identical to movements in labour force or actual
labour usage. If, for example, a state’s population is being pushed up by a large inflow of
retirees or other people of non-working age, its per capita income level will tend to decline
relative to other states, even if its LP is keeping pace with the other states. Similarly, if
employment is rising less rapidly in a given state while its LP is growing at the same rate
as the national average, its income per capita will tend to fall relative to the other states.
Finally, if a given state’s labour force participation rate is rising faster than the national
average, its per capita income will also tend to rise faster.
In what follows we will consider both per capita income and LP levels. Further, we will
examine the differences in industrial structure across the states to determine whether and to
what extent these differences affect their growth performance. Our data relate to the period
1984-85 to 1998-99, and come mainly from the Australian Bureau of Statistics (2000) via
the dX EconData database.
There are two sets of real GSP data: one based on the System of National Accounts 1968
(SNA68) and covering the years 1984-85 to 1996-97, the other based on the System of
National Accounts 1993 (SNA93) and covering the years 1990-91 to 1998-99. We adopt the
latter data set as the main source, and apply simple splices to extend it back to 1984-85
using the earlier data set. As for employment, we consider both the number of persons
employed and the total number of hours worked. While our discussion below is based
mostly on the latter measure, movements in the two were very similar and the main results
are unaffected if the former measure is used instead. For the present purposes, the Northern
Territory and the Australian Capital Territory are excluded, mainly because of the small
sample sizes involved in the compilation of data relating to their individual industries.
– 43 –
Productivity and Regional Economic Performance in Australia
Interstate Comparison of Economic and Labour Productivity Growth
Real GSP
As a starting point, Table 3.1 presents an interstate comparison of rates of growth in real
GSP. These are measured as trend growth rates, and are obtained as the slope coefficients
from regressions of logs of real GSP on a simple trend and a constant. It is evident that
Western Australia and Queensland outperformed the other states on this basis, registering
growth rates of 4.7% and 4.5% per year respectively, while New South Wales, Victoria and
South Australia were in the 2.2% to 2.9% range, and Tasmania recorded only 1.5% per year.
Table 3.1: Interstate Comparisons of Growtha in Output, Population, Employment,
and Labour Productivity, 1984-85 to 1998-99, per cent per year
State
GSPb
NSW
2.9
1.1
Vic
2.5
Qld
Population Employed
persons
Hours
worked
GSP
per
capita
GSP per
employee
GSP per
hour
worked
1.5
1.5
1.8
1.4
1.4
0.9
1.1
1.1
1.6
1.4
1.4
4.5
2.3
3.1
3.0
2.2
1.4
1.5
SA
2.2
0.6
0.7
0.7
1.6
1.5
1.6
WA
4.7
1.8
2.5
2.4
2.9
2.3
2.3
Tas
1.5
0.5
0.6
0.4
1.0
0.9
1.1
All states
3.1
1.3
1.7
1.7
1.9
1.4
1.5
a Growth rates are obtained by fitting an exponential trend to the relevant series with a constant term.
b Rates based on chain volume measure of gross state product, at 1997-98 prices.
Source: ABS data via DX database.
Real GSP per capita
Of course, much of these differences could be explained by variations in population growth.
Data for this variable are also summarised in the table. Queensland and Western Australia
again dominated, with 2.3% and 1.8% per year respectively, and again Tasmania recorded
the lowest growth rate, 0.5%. Combining these movements, we find that even in per capita
income terms, Western Australia and Queensland still registered the highest growth rates,
and Tasmania the lowest, although the differentials across the states were now much
smaller: 2.9% for Western Australia versus 1.0% for Tasmania.
Figure 3.1a displays movements in real GSP per capita during this period, with the data
being presented in logs to better illustrate growth trends. The growth paths can clearly be
divided into two broad groups: the top group comprised New South Wales, Victoria and
Western Australia, and the lower group consisted of Queensland, South Australia and
Tasmania. It can also be seen that Western Australia moved from the lower range of the top
group to the upper range, having surpassed New South Wales and Victoria in the early
1990s. Over the same period, Tasmania failed to keep up and by 1998-99 could be thought
of as being in a third group of its own.
– 44 –
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Our results thus far are consistent with the findings of previous studies, in terms of both (a) the
relative rankings of the states, and (b) the absence of evidence in support of convergence during
this recent period. Indeed, from Figure 3.1a and especially from the diverging paths of Western
Australia and Tasmania, one would be inclined to suspect that there has been some divergence.
Figure 3.1b displays the trend growth rates experienced by the states against their initial
levels of income. To reduce the risk of errors in measuring the latter variable, especially
those due to short-term fluctuations, the actual values are replaced by the predicted values
generated from the corresponding trend growth regressions. We have also experimented
with an alternative method for specifying the initial levels, namely taking the average of the
first few (say, three) years. The results from the two methods are quite similar. It can be seen
from Figure 3.1b that the data did not support `-convergence, nor `-divergence. Regressing
trend growth rate on predicted initial income level (and a constant term) confirms that there
was neither a (statistically significant) negative nor a positive relationship between them,
although reservations about the low degrees of freedom must be kept in mind. The slope
coefficient is 0.72, with t-statistic of 0.50 and R2 of 0.06.
Figure 3.1c shows movements over time of the cross-sectional standard deviation of the
logs of per capita incomes. This is effectively the unweighted coefficient of variation (CV)
that is frequently used as a measure of cross-sectional dispersion. The figure indicates a
rising trend. A regression of the CV series against a linear time trend and a constant
confirms that the positive trend is highly significant, with t-statistic of 10.88 and R2 of 0.91.
Such divergence is in contrast with international and Australia’s own historical experiences,
where interregional incomes have tended to converge, especially over long periods. It is,
however, in keeping with Neri’s (1998) results for the 1976-1991 sub-period, and can be
seen as both confirmation and extension of those results to a subsequent decade.
Labour productivity
We now turn to an analysis of employment growth. As Table 3.1 shows, the relative
rankings of the states with respect to growth in the number of persons employed, or in the
total number of hours worked, are largely unchanged from those obtained for growth in the
total number of residents. However, the growth rate differentials between states with high
population growth, such as Queensland and Western Australia, and those with low rates of
population growth, such as Tasmania and South Australia, tend to be larger for the persons
employed and hours worked measures than for the total population measure. This suggests
that the higher rates of population growth in Queensland and Western Australia were
associated with inflows of migrants who were more than proportionately of the working age
and able to gain employment. Rather than having an adverse effect on per capita income
growth, therefore, population growth in these cases tended to lift income growth rates by
raising the proportion of the overall population who are of working age.
After adjusting GSP growth for employment growth rather than population growth, we find
an interstate pattern of remarkably even performances in terms of LP growth. As the last two
columns of Table 3.1 indicate, LP growth rates recorded by the various states were similar
to one another. Even though Tasmania’s growth rates were still the lowest, the gaps between
it and the other states were small, especially when one considers statistical variations and
errors. The only true exception was Western Australia, which continued to record a
substantially higher growth rate than the others.
– 45 –
Productivity and Regional Economic Performance in Australia
Figure 3.1: Per Capita Income
(a) Level of per capita income
NSW
Vic
Qld
SA
WA
Tas
6.0
Per capita income (in logs)
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
1986-87
1988-89
1990-91
1992-93
1994-95
1998-99
1996-97
(b) Beta-convergence of per capita income
Annual trend growth rate (%)
3.2
WA
2.6
Qld
2.0
NSW
SA
Vic
1.4
Tas
0.8
1.8
1.9
2.0
2.1
2.2
2.3
2.4
2.5
2.6
Predicted initial value (in logs)
(c) Sigma-convergence of per capita income
CV
Trend
Coefficient of variation
0.15
0.14
0.13
0.12
0.11
0.10
0.09
1986-87
1988-89
1990-91
– 46 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.2a illustrates the time paths of LP levels in the six states during the period of
analysis. The relative rankings are consistent with historical trends and patterns. Once
again, Western Australia joined the top group, which had previously consisted of New South
Wales and Victoria, and again Queensland, South Australia and Tasmania remained in the
lower group. Compared with the pattern presented in Figure 3.1a for per capita income,
however, the divergence tendency here appears far less pronounced. As Figure 3.2b
indicates, and a corresponding regression confirms (t-statistic = -0.04, R2 = 0.00), there was
no evidence of either `-convergence or `-divergence. Figure 3.2c shows that there was still
a tendency toward m-divergence; a regression confirms that this rising trend in the CV is
significant at 5% (t-statistic = 2.70, R2 = 0.38).
Interstate Comparison of Labour Productivity Growth by Industry
In this section we replicate the above analysis for a number of representative industries,
including agriculture, forestry and fishing; mining; manufacturing; electricity, gas and
water; wholesale trade; finance and insurance; property and business services; general
government; and personal and other services.
Agriculture, forestry and fishing: Figures 3.3a, 3.3b and 3.3c present a summary view of
LP growth in agriculture in the various states. It can be seen from Figure 3.3b that most
states displayed a tendency toward `-convergence. However, Western Australia, which
started out with a high LP level, continued to record a relatively high growth rate, thus
tending to pull away from the other states. Partly as a result of this, the CV of state LP levels
registered an upward trend that is significant at the 10% level. In short, the data indicate no
`-convergence (or `-divergence) but significant m-divergence.
Mining: As Figure 3.4a illustrates, Victoria’s LP level in mining was consistently far above
the levels in other states. There is no strong evidence to support either `-convergence or
`-divergence. As Figure 3.4b shows, Western Australia, South Australia and Queensland all
started with very similar levels of initial LP, yet recorded considerably different growth
rates. Nor was there any sign of a significant secular trend in the dispersion of LP levels.
Manufacturing: This is the largest of all the industries listed. As Figure 3.5a shows,
Queensland’s LP level tended to remain substantially below the levels in other states. From
Figure 3.5b we can see that New South Wales, South Australia, Western Australia and
Queensland all started from similar initial LP levels but then experienced divergent growth
rates. While there is no firm indication of either `-convergence or `-divergence, significant
m-divergence is found at the 10% level.
Electricity, gas and water: Figure 3.6a illustrates the strong growth of Tasmania’s LP in
this industry. After rising from the bottom position, the state caught up with, and then
overtook, the leading states. This can probably be accounted for by increasing reliance on
hydro-power in Tasmania, following the decommissioning of the last remaining oil-fired
power station during this period. There are other examples of ‘cross-overs’. These are
consistent with the finding of significant `-convergence, as shown in Figure 3.6b, and at the
same time m-divergence, as portrayed in Figure 3.6c.
– 47 –
Productivity and Regional Economic Performance in Australia
Figure 3.2: GSP Per Hour
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
5.2
5.0
4.8
4.6
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
2.8
WA
2.2
Qld
SA
Vic
1.6
NSW
Tas
1.0
0.10
0.11
0.12
0.13
0.14
0.15
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.100
0.095
0.090
0.085
0.080
0.075
1986-87
1988-89
1990-91
– 48 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Wholesale trade: As is evident from Figures 3.7a, 3.7b and 3.7c, wholesale trade
experienced both `- and m-convergence. Note that some states recorded negative LP growth
over several years. While this may accurately reflect some difficult times that were
experienced by the industry, it also points to the inherent problems of measuring output in
a service-oriented industry.
Finance and insurance: Figures 3.8a, 3.8b and 3.8c present perhaps the clearest example
of convergence in both senses of the term. Note, however, that most of the m-convergence
took place during the relatively short period of the latter half of the 1980s, and that during
the entire following decade the CV of LP levels was fairly steady. This pattern of
developments may have been due to a combination of both the introduction of modern
technologies in this sector and the rationalisation of work practices in the more competitive
environment following financial deregulation.
Property and business services: Figure 3.9a displays a general pattern of falling LP levels.
Apart from New South Wales, all states experienced considerable declines in LP. This
divergent pattern resulted in an increase in cross-state dispersion (see Figure 3.9c). The
rising trend is found to be significant at the 5% level.
Government administration and defence: Figure 3.10a shows that Queensland and (to a
lesser extent) Victoria outperformed the other states with respect to LP growth in this
industry. LP levels in these two states were among the lowest nationally at the beginning of
the period, but then rose rapidly, enabling them to catch up and then surpass the others. The
improvement in Queensland’s position may have been a result of the considerable public
sector reforms that took place during the 1990s. In contrast to Queensland and Victoria,
Tasmania generally remained below the other states, until the late 1990s when it achieved
a sharp rise in LP. Combining these movements, the overall picture is one where there is
evidence of `-convergence, but m-divergence.
Personal and other services: Figures 3.11a, 3.11b and 3.11c are of interest for at least two
reasons. First, they relate to an industry that experienced negative LP growth, in common
with several other service industries. Second, because of considerable swings in state LP
levels, cross-sectional dispersion was not reduced significantly even though the growth
rates did conform reasonably well to a `-convergence pattern.
– 49 –
Productivity and Regional Economic Performance in Australia
Figure 3.3: Labour Productivity in Agriculture, Forestry and Fishing
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.0
5.8
5.6
5.4
5.2
5.0
4.8
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
5
Vic
4
SA
WA
3
Tas
Qld
2
NSW
1
0.14
0.15
0.16
0.17
0.18
0.19
0.20
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.25
0.20
0.15
0.10
0.05
0.00
1986-87
1988-89
1990-91
– 50 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.4: Labour Productivity in Mining
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
9
8
7
6
5
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
9
Tas
WA
8
7
NSW
Vic
6
SA
Qld
5
0.0
0.5
1.0
1.5
2.0
2.5
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
1.0
Coefficient of variation
0.9
0.8
0.7
0.6
0.5
0.4
1986-87
1988-89
1990-91
– 51 –
1992-93
1994-95
1996-97
1998-99
Productivity and Regional Economic Performance in Australia
Figure 3.5: Labour Productivity in Manufacturing
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.2
6.1
6.0
5.9
5.8
5.7
5.6
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
2.4
NSW
Annual trend growth rate (%)
SA
1.8
Vic
Tas
WA
1.2
0.6
0.0
0.30
Qld
0.31
0.32
0.33
0.34
0.35
0.36
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
0.12
Coefficient of variation
0.10
0.08
0.06
0.04
0.02
0.00
1986-87
1988-89
1990-91
– 52 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.6: Labour Productivity in Electricity, Gas and Water
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
8.0
7.6
7.2
6.8
6.4
6.0
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
12
Tas
SA
10
8
NSW
Vic
WA
6
Qld
4
0.3
0.4
0.5
0.6
0.7
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
0.30
Coefficient of variation
0.25
0.20
0.15
0.10
0.05
0.00
1986-87
1988-89
1990-91
– 53 –
1992-93
1994-95
1996-97
1998-99
Productivity and Regional Economic Performance in Australia
Figure 3.7: Labour Productivity in Wholesale Trade
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.2
6.0
5.8
5.6
5.4
5.2
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
2.5
Annual trend growth rate (%)
SA
2.0
WA
Tas
Qld
NSW
1.5
1.0
0.5
0.0
0.20
Vic
0.22
0.24
0.26
0.28
0.30
0.32
0.34
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
0.18
Coefficient of variation
0.16
0.14
0.12
0.10
0.08
0.06
1986-87
1988-89
1990-91
– 54 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.8: Labour Productivity in Finance and Insurance
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
7.0
6.5
6.0
5.5
5.0
4.5
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
14
Tas
12
10
Qld
8
Vic
SA
WA
6
4
0.10
NSW
0.15
0.20
0.25
0.30
0.35
0.40
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
0.40
Coefficient of variation
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1986-87
1988-89
1990-91
– 55 –
1992-93
1994-95
1996-97
1998-99
Productivity and Regional Economic Performance in Australia
Figure 3.9: Labour Productivity in Property and Business Services
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.3
6.1
5.9
5.7
5.5
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
1
NSW
0
-1
SA
WA
Vic
Qld
-2
-3
Tas
-4
-5
0.36
0.38
0.40
0.42
0.44
0.46
Predicted initial value (in logs)
0.48
0.50
0.52
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.22
0.20
0.18
0.16
0.14
0.12
0.10
0.08
1986-87
1988-89
1990-91
– 56 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.10: Labour Productivity in Government Administration and Defence
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.3
6.1
5.9
5.7
5.5
1986-87
1988-89
1990-91
1992-93
1994-95
1998-99
1996-97
(b) Beta-convergence of labour productivity
4
Annual trend growth rate (%)
Qld
3
2
Vic
NSW
1
Tas
SA
0
WA
-1
-2
0.27
0.29
0.31
0.33
0.35
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.20
0.18
0.16
0.14
0.12
0.10
0.08
0.06
1986-87
1988-89
1990-91
– 57 –
1992-93
1994-95
1996-97
1998-99
Productivity and Regional Economic Performance in Australia
Figure 3.11: Labour Productivity in Personal and Other Services
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
5.8
5.7
5.6
5.5
5.4
5.3
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
1.2
Annual trend growth rate (%)
Qld
0.8
0.4
Vic
0.0
WA
-0.4
Tas
-0.8
SA
NSW
-1.2
0.20
0.22
0.24
0.26
0.28
0.30
0.32
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
1986-87
1988-89
1990-91
– 58 –
1992-93
1994-95
1996-97
1998-99
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Summary: Table 3.2 presents an overview of the evidence concerning cross-state
convergence (or divergence) within each individual industry. It can be seen that LP levels
within many industries displayed patterns that suggest `-convergence, as indicated by the
negative sign, with varying degrees of confidence, of most estimates of the slope coefficient
in the regression of trend growth rate on initial income level. Yet the data for all industries
combined do not indicate clearly either convergence or divergence. Why is this so? One
possible reason is that the industries, primarily in the service sector, that show clear signs
of convergence are not the largest industries. Another is that, as we have seen above, for
different industries different sets of states were responsible for the converging behaviour, so
that overall they did not converge.
Table 3.2: Labour Productivity:
Convergence Behaviour within Individual Industries, 1985-86 to 1998-99
Regression of trend growth rate
on predicted initial value
Industry
Slope
t-value
coefficient DF = 4
Agriculture,
forestry and fishing
-9.81
-0.36
Mining
-0.53
5.31
Manufacturing
R2
Regression of CV (standard
deviation of logs) on time trend
Slope
t-value
coefficient DF = 12
(x102)
2.11 *
R2
0.03
0.56
-0.74
0.12
-0.49
0.25
0.02
0.24
1.97 *
0.24
0.63
2.48 **
0.34
1.61
0.18
-0.55
0.27
0.02
Electricity, gas and water
-17.39
-2.90 **
0.68
Construction
-16.78
-0.49
0.06
0.24
Wholesale trade
-17.54
-4.94 **
0.86
-0.44
-3.74 **
0.54
Retail trade
-61.68
-2.67 *
0.64
-0.13
-1.05
0.08
Accommodation, cafes
and restaurants
-23.47
-0.96
0.19
-0.13
-0.71
0.04
Transport and storage
-19.76
-1.06
0.22
0.26
0.97
0.07
Communication services
-24.91
-2.89 **
0.68
-0.34
-1.01
0.08
Finance and insurance
-36.51
-4.60 **
0.84
-1.70
-3.50 **
0.51
0.11
0.00
0.78
4.47 **
0.63
Government administration
and defence
-39.65
-2.17 *
0.54
0.54
2.88 **
0.41
Education
Property and
business services
1.59
-13.42
-1.10
0.23
-0.13
-0.62
0.03
Health and
community services
-2.64
-0.23
0.01
0.12
0.99
0.08
Cultural and
recreational services
-9.31
-0.82
0.15
0.15
0.58
0.03
-3.72 **
0.78
-0.28
-1.34
0.13
Personal and other services -21.40
Total (all industries listed)
Total less Mining
**
*
2.67
0.27
0.02
0.17
-0.00
-0.65
0.10
-0.00
Significant at 5 %
Significant at 10 %
– 59 –
5.27 **
-0.86
0.70
0.06
Productivity and Regional Economic Performance in Australia
There are a number of industries (such as electricity, gas and water; and manufacturing) that
exhibited either `-convergence or no clear behaviour with respect to `-convergence, yet
clearly were subject to m-divergence. One possible explanation for this apparent
discrepancy is that, for some industries, exogenous (e.g. technology-driven) shocks may
have significantly affected the system, and these effects may have temporarily dominated
the convergence tendency, which would eventually reassert itself once the system settles
down again, in the absence of major new shocks. Another possibility is the case of crossovers where some states would come up from below the national average, catch up with the
others (thus fulfilling the conditions of `-convergence), but then would just keep rising
further, thus contributing to m-divergence.
Of the 17 industries studied, seven exhibited some significant tendency with respect to
cross-sectional dispersion. Of these, five showed m-divergence while two displayed
m-convergence. Yet the overall picture, when all the sectors are combined, is unambiguously
one of m-divergence, with the finding enjoying a higher degree of confidence than any of
the industry specific results (t-statistic = 5.27). One might attribute this to the fact that more
industries exhibited divergence than convergence, and the divergence displaying industries
(such as manufacturing) tended to be more influential in terms of shares of both output and
labour. But an alternative explanation is possible: it may be that, as Neri (1998) suggested,
overall divergence has been caused mainly by differences in the states’ industrial structure
and by structural changes. We now turn to an examination of this issue.
Inter-industry Comparison of Labour Productivity Growth
In Table 3.3 we present summary data relating to the rates of LP growth in each industry for
all states combined, as well as the industrial composition of employment in each state,
measured as the average shares of that state’s employment being devoted to the various
industries. It is evident that the industries with the highest trend growth rates were mining;
electricity, gas and water; communication services; and finance and insurance. This may
have been due to rapid technological progress as well as changes in the labour–capital mix
and industrial practices. The industries that recorded the slowest rates of LP growth (in
some cases even negative rates of growth) were property and business services;
accommodation, cafes and restaurants; personal and others services; and other servicerelated industries such as cultural and recreational services, and education. As pointed out
above, this may well reflect fundamental problems with the measurement of output in a
service industry.
The data presented in Table 3.3 can be used to answer the following question: Were some
states disadvantaged by the fact that their industrial structures were less conducive to high
LP growth than others? Consider, for example, the hypothetical case of a state that is heavily
oriented toward the service industries. Given that these industries tend to record low rates
of growth, the state is likely to be disadvantaged in any interstate comparisons of recorded
LP growth.
– 60 –
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Table 3.3 Labour Productivity Growth and Share of Total Labour by Industry,
1985-86 to 1998-99, per cent
Industry
Agriculture, forestry
and fishing
Average trend
growth rate
of labour
productivity
Average share of total labour in each state
NSW
Vic
Qld
SA
WA
Tas
CV
2.79
5.52
6.05
8.27
8.75
7.66
10.03
22.72
Mining
6.69
1.07
0.33
1.75
0.90
4.59
1.56
86.71
Manufacturing
1.63
15.88
19.32
12.45
17.35
11.88
14.49
18.91
Electricity,
gas and water
7.55
1.41
1.30
1.06
1.37
1.26
1.80
17.29
Construction
0.74
7.57
6.86
8.78
6.34
8.65
6.74
13.59
Wholesale trade
1.36
7.37
6.87
6.49
6.28
6.39
5.47
9.98
Retail trade
1.32
12.66
13.01
14.11
13.19
13.28
14.10
4.40
Accommodation, cafes
and restaurants
-0.54
4.16
3.15
4.59
3.42
3.75
4.15
13.96
Transport and storage
2.02
5.62
5.07
6.09
4.58
5.15
4.65
10.96
Communication
services
7.38
1.99
2.00
1.69
1.68
1.55
1.64
10.60
Finance and insurance
7.10
5.01
4.46
3.17
3.48
3.55
3.14
19.09
-1.07
9.11
8.78
8.27
7.36
9.01
5.59
18.73
Government administration
and defence
1.45
3.66
3.95
3.91
3.58
3.72
5.88
18.63
Education
0.24
6.02
6.53
6.57
7.05
6.49
6.51
5.01
Health and
community services
0.93
7.68
7.30
7.63
9.26
7.64
8.77
9.40
Cultural and
recreational services
0.12
2.01
1.70
1.94
1.73
1.89
1.95
6.86
-0.36
3.27
3.33
3.24
3.66
3.54
3.53
4.99
Property and
business services
Personal and
other services
Total
(all industries listed)
1.76
100.00 100.00 100.00 100.00 100.00 100.00
– 61 –
Productivity and Regional Economic Performance in Australia
It turns out, however, that with one exception (to be discussed below) there was a fairly high
degree of similarity across states with respect to the industry shares of total labour used. In
a majority of cases the coefficient of variation is less than 15%. It is true that variations in
the share of manufacturing, in particular, could be quite influential in view of the relatively
large size of the industry. The labour share of this industry in the most manufacturing
intensive state (Victoria) was 19.3% compared with only 11.9% in the least manufacturing
intensive state (Western Australia). Nevertheless, variations in labour shares across states
tended to interact in an offsetting fashion with differentials in the LP growth rates achieved
by the different states in each industry, so that they all ended up with similar aggregate LP
growth rates.
As mentioned above, there was a very notable exception to this. Because the mining
industry’s LP growth rate was so much higher than the average for other industries, and
because the industry accounted for such a large share of total employed labour in Western
Australia, this state’s LP growth performance was substantially affected by the industry. To
illustrate this point, we have replicated the calculations with mining excluded from
measures of both output and employed labour.
As shown in Table 3.4, without mining, Western Australia’s overall LP growth performance
would have been slightly below average (1.5% per year, compared with the national average
of 1.6%). Moreover, as portrayed in Figures 3.12a and 3.12c, there would have been no
strong indication of m-divergence among the overall LP levels of the Australian states
during the period of study. A regression confirms that there would have been neither
m-convergence nor m-divergence (t-statistic = -0.86, R2 = 0.06).
Table 3.4: Interstate Comparison of Growth in All-industry Real GSP
and Labour Productivity, with and without Mining, per cent per year
Trend growth rates
State
Real GSP
Real GSP/hr worked
with
Mining
without
Mining
with
Mining
without
Mining
NSW
3.2
3.2
1.8
1.7
Vic
2.6
2.6
1.5
1.5
Qld
4.7
4.7
1.8
1.7
SA
2.3
2.4
1.8
1.8
WA
4.9
3.8
2.6
1.5
Tas
1.8
1.8
1.5
1.4
All states
3.4
3.2
1.8
1.6
– 62 –
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
Figure 3.12: Labour Productivity, All Industries less Mining
(a) Level of labour productivity
NSW
Vic
Qld
SA
WA
Tas
Labour productivity (in logs)
6.0
5.9
5.8
5.7
5.6
5.5
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
(b) Beta-convergence of labour productivity
Annual trend growth rate (%)
1.9
SA
1.8
Qld
NSW
1.7
1.6
Vic
WA
1.5
Tas
1.4
1.3
0.24
0.25
0.26
0.27
0.28
0.29
0.30
0.31
Predicted initial value (in logs)
(c) Sigma-convergence of labour productivity
CV
Trend
Coefficient of variation
0.074
0.072
0.070
0.068
0.066
0.064
0.062
0.060
1986-87
1988-89
1990-91
– 63 –
1992-93
1994-95
1996-97
1998-99
Productivity and Regional Economic Performance in Australia
Summary and Policy Implications
In this chapter, we have examined differences across the states with regard to their rates of
growth in real GSP per capita and LP during the period 1984-85 to 1998-99. We have found
that some of the interstate variations in per capita income growth rates could be attributed
to variations in the rate of population growth, as states whose populations grew more
rapidly also tended to receive larger inflows of working age migrants, who were then able
to gain employment and raise the ratio of employed persons to total population.
Once allowance has been made for the effects of population growth, there remained very
small differences in LP growth performance across the states, except for Western Australia
and, to a lesser extent, Tasmania. Even for Western Australia, which enjoyed a markedly
higher LP growth rate, the difference had largely disappeared by the early 1990s.
This is not to say that there were no interstate differences in industry specific LP growth rates
or in the industrial composition of the employed labour force. Indeed, as seen above there has
been a wide range of patterns of interstate and inter-industry variations, and in some cases the
variations were quite sizeable. It is rather remarkable, however, that such differences have
tended to largely offset one another, leaving fairly small differences in the aggregate (allindustry) LP growth rates. Thus, while a given state may enjoy an advantage from having an
industrial structure more conducive to high growth, it may at the same time be disadvantaged
by lower LP growth rates compared with the other states in a range of industries.
By and large, the advantages and disadvantages across states have tended to cancel out each
other. In particular, despite any differences that may have existed in the sets of policies
adopted by the various state governments, the states have ended up with rather similar rates
of LP growth.
This result is in contrast to the proposition (or concern) that states have differed
substantially in their ability to adapt themselves to suit external conditions, and that this has
resulted in substantial variations in LP growth rates. We have found that the only state with
a clearly superior LP growth performance was Western Australia, but even there the
difference, attributable mainly to a mining boom that was perhaps due as much to good
fortune as wise management, has largely disappeared.
Neither the above finding, nor the fact that the interstate dispersion of LP in Australia has
for decades been very low (of the order of 10%) by international standards, can provide any
grounds for complacency over the issue of income disparities across the states. First, policy
makers in slow population growth states (such as Tasmania and South Australia) still need
to retain a focus on demographic changes, and work to avoid the situation where the state
may end up being unable to attract and retain its fair share of dynamic, employable persons
who would contribute to a rise in living standards for all residents.
Second, the historically and internationally prevalent tendency toward interregional
convergence of LP appears to have been absent during the past 15 years, and, from Neri’s
(1998) results, during the preceding decade as well. For states that have remained in the
lower income group (such as Queensland, South Australia and Tasmania), the convergence
tendency should have provided an advantage over the top group states. That this advantage
has been offset by other factors only means that the lower group states have some scope for
– 64 –
Variations in Economic and Labour Productivity Growth among the States of Australia: 1984-85 to 1998-99
identifying these disadvantages and working toward improving their LP growth
performance.
Finally, all Australian states still need to remain vigilant about monitoring best practices
among comparable economies overseas, and act to facilitate the adoption of these practices
by organisations operating within their boundaries. After all, the international convergence
tendency should present all of them with the opportunity to grow more rapidly than, and
thereby catch up with, the most advanced economies in the world.
– 65 –
Productivity and Regional Economic Performance in Australia
References
Abramovitz, M. (1986), ‘Catching up, forging ahead, and falling behind’, Journal of
Economic History, 46 (2), 385-406.
Australian Bureau of Statistics (2000), various data, Canberra, in dX EconData, database,
EconData Pty Ltd, Melbourne.
Barro, R. and Sala-i-Martin, X. (1992), ‘Convergence’, Journal of Political Economy,
100 (2), 223-251.
Baumol, W.J. (1986), ‘Productivity growth, convergence and welfare: What the long-run
data show’, American Economic Review, 76 (5), 1072-1085.
Bernard, A.B. and Jones, C.I. (1996), ‘Productivity and convergence across the U.S. states
and industries’, Empirical Economics, 21, 113-115.
Cashin, P. (1995), ‘Economic growth and convergence across the seven colonies of
Australasia: 1861-1991’, The Economic Record, 71 (213), 132-144.
Dowrick, S. and Nguyen, D.T. (1989), ‘OECD comparative economic growth 1950-85:
Catch-up and convergence’, American Economic Review, 79 (5), 1010-1030.
Harris, P. and Harris, D. (1992), ‘Interstate differences in economic growth rates in
Australia, 1953-54 to 1990-91’, Economic Analysis & Policy, 22 (2), 129-148.
Neri, F. (1998), ‘The economic growth performance of the states and territories of
Australia: 1861-1992’, The Economic Record, 74 (225), 105-120.
Quah, D. (1993), ‘Empirical cross-section dynamics in economic growth’, European
Economic Review, 37 (2/3), 426-434.
Sala-i-Martin, X. (1990), ‘On growth and states’, Ph.D. Dissertation, Harvard University,
Cambridge, MA.
Sala-i-Martin, X. (1996), ‘Regional cohesion: Evidence and theories of regional growth
and convergence’, European Economic Review, 40 (6), 1325-1352.
– 66 –
4 Economic Growth Performance of the
Australian States and Territories:
An Extended Shift-Share Analysis
Christine Smith and Duc-Tho (Tom) Nguyen
Introduction
There are comparatively few studies of the spatial and sectoral aspects of Australian gross
domestic product (GSP) dynamics.
Donovan (1981) was concerned with disaggregating the official Australian Bureau of
Statistics (ABS) national accounts to generate unofficial state accounts, including estimates
of GSP at factor cost for the period 1953-54 to 1977-78. His main conclusion was that ‘state
growth rates have been persistently different to a possibly significant extent’ (p. 227).
Harris and Harris (1992) examined differences in the rates of economic growth and levels
of aggregate gross state product (GSP) and GSP per capita over the period 1953-54 to 1990-91.
Aggregate GSP grew at above the national rate in Western Australia, Queensland and South
Australia/Northern Territory combined over this period as a whole, while the remaining
regions (New South Wales/Australian Capital Territory combined, Victoria and Tasmania)
recorded growth rates below the national average. The major focus of the paper was with
identifying differences in GSP per capita. The main conclusion drawn in relation to
aggregate GSP was that ‘the Australian economy over the period ... exhibited significant
interregional variations in growth rates, so that any explanation as to the causes or sources
of ... growth must take account of ... regional factors’ (p. 142).
Harris (1998) was primarily concerned with analysis of interstate disparities in GSP per
capita over the period 1977-78 to 1994-95 using coefficients of variation as the measure of
dispersion. He did, however, present data relating to interstate differences in aggregate GSP
growth and concluded that ‘... no state had an above-average or below-average GSP growth
rate every year. Instead each state had a mixture of years when GSP growth was above
average and years when it was below average’ (p. 206). The only factor used in an attempt
to explain differentials in aggregate GSP growth was population growth.
Cashin (1995) examined a longer time period, namely 1861 to 1991 (and included New
Zealand as a former Australasian colony). His concern was with more formal statistical
testing for convergence in real per capita GSP rather than a comparison of growth rates for
aggregate GSP.1 An attempt was made to examine the extent to which agricultural shocks
impacted on the rate of convergence.
1
Maxwell and Hite (1992) and Cashin and Strappazzon (1998) focused on convergence in regional per capita incomes and so
are not directly relevant to this paper.
– 67 –
Productivity and Regional Economic Performance in Australia
Neri’s (1998) study was essentially a re-examination of Cashin’s analysis with an
alternative data set (excluding New Zealand) and with differences across sub-periods
examined more closely. An attempt was made to explore the role of sectoral composition
differences in explaining the rate of convergence. The sectoral composition variable
employed was a fairly crude measure of the extent to which a state’s production is
concentrated in the faster growing sectors of the economy. Nevertheless, this variable
proved to be a highly significant variable in explaining interstate convergence.
Nguyen, Smith and Meyer-Boehm (2000) examined the same time period as this present
chapter and utilised the same data set. Their primary concern was not with testing for
convergence in real GSP per capita but rather convergence in real GSP per employed worker
(i.e. labour productivity). Chapter 3 comprises a summary of their main findings. Like
Cashin (1995) and Neri (1998) they also point to the potential importance of industrial
structure differences in explaining variations in interstate economic growth performance.
Although our analysis builds on and extends the results of previous studies in this area,2 at
the same time it differs from them in three important respects. First, previous studies deal
with the time period up to 1991-92, whereas our focus is on a more recent time period,
namely 1984-85 to 1998-99. Second, we concentrate our attention on growth or the change
in aggregate GSP, while most previous studies are primarily concerned with GSP
(or income) per capita. Third, our primary concern is not with documenting and describing
interregional differences in economic growth performance over the period 1984-85 to
1998-99, but perhaps more importantly with attempting to identify those factors responsible
for producing these differences.
In particular we apply and extend further the modern variations of shift-share analysis
developed by Barf and Knight (1988), Rigby and Anderson (1993) and Haynes and Dinc
(1997). This enables us not only to decompose GSP growth into components attributable to
national growth, industry mix and regional location advantage (or disadvantage), but also to
assess whether changes in GSP are primarily attributable to employment change or to
productivity change.
The chapter is organised as follows. The first section provides an overview of shift-share
methodology with an emphasis on the particular variants of it adapted for use in this study.
The second section presents the results of the application of dynamic versus conventional
shift-share analysis, while the third and fourth sections present the results of productivityextended shift-share analysis. The final section contains a summary of the main conclusions
and draws out some policy implications.
As in Chapter 3, our data relate to the period 1984-85 to 1998-99 and come mainly from the
Australian Bureau of Statistics (2000) via the dX EconData database. There were two sets of
GSP data available: one based on the System of National Accounts 1968 (SNA68) and
covering the years 1984-85 to 1996-97, and the other based on the System of National
Accounts 1993 (SNA93) and covering the years 1990-91 to 1998-99. We adopted the latter
data set as the main data source and applied simple splices to extend it back to 1984-85 using
2
These studies include, for example, Donovan (1981), Harris and Harris (1991), Harris and Harris (1992), Maxwell and Hite
(1992), Cashin (1995), Cashin and Strappazzon (1998), Harris (1998), Neri (1998) and Nguyen et al. (2000).
– 68 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
the earlier data set.3 These GSP data were available in total and for 17 different industries.
They were converted to real 1997-98 dollars using national industry-specific GSP price
deflators.4 In terms of employment, we considered using both the number of persons
employed and the total number of hours worked. Because the latter incorporates the effect of
interstate, inter-sectoral and inter-temporal differences in part-time versus full-time work, it
is generally regarded as the more appropriate employment indicator and was the one finally
adopted.5 Factor income and wage, salary and supplement income were also derived for each
state and territory and the nation from the dX database.6
Shift-Share Analysis
Shift-share analysis is a relatively simple technique for analysing changes in economic
activity in a region (or set of regions) over a specific time period. The indicator of economic
activity could be output levels, employment levels, the number of establishments, household
income levels or a variety of other variables for which data are available. Being highly
adaptable to the type of data available, implementable with only a limited time series of such
data and comparatively easy to use with modern spreadsheet software, the technique has been
widely used since it was first developed in 1960.7 At the same time, however, its simplicity
has led it to be vulnerable to a wide range of criticisms.8 In an attempt to redress a number of
these criticisms, various authors have sought to extend and reformulate the technique.9 We
apply a number of these modern extensions and reformulations to the Australian context in
this paper. Before reporting our results it may be useful to provide an overview of the basic
technique and the main variants employed in this study.
3
For a discussion of the differences between these two systems of national accounts, see Australian Bureau of Statistics (1997).
4
As a result inter-regional differences in rates of inflation are ignored, while differences in relative price changes for different
sectoral outputs are incorporated. We experimented with a number of alternative approaches to converting nominal GSP
figures to their real counterparts: for example, use of the national aggregate GDP deflator series across all sectors and regions
(which ignores both inter-regional and inter-industry differences in price changes) and use of the ratio of capital city consumer
price indexes (CPI) for each state/territory to the corresponding national CPI to generate a state-specific GDP deflator series
from the national GDP deflator series (which ignores differences between the CPI and GDP deflator concepts, as well as
inter-industry differences in price changes). The choice of deflator impacted quite significantly on the final results for a number
of sectors and regions, so this represented a non-trivial research question. Results from alternative resolutions of this question
are available from the authors upon request.
5
See, for example, Rigby and Anderson (1993) and Haynes and Dinc (1997) for arguments in favour of use of hours worked
rather than number of employees. Once again the choice of the measure of employment impacted quite significantly on the
final results for a number of sectors and regions, so this represented a non-trivial research question. Results from alternative
resolutions of this question are available from the authors upon request.
6
Once again, differences between the SNA93 and SNA68 systems necessitated a splicing of the data to get estimates of the ratio
of labour income to total factor income over the entire study period (see Australian Bureau of Statistics, 1997).
7
See, for example, Dunn (1960), Perloff et al. (1960), Fuchs (1962), Dawson (1982), Ledebur and Moomaw (1983), Markusen
et al. (1991), Denning (1996), Noponen et al. (1996) and Noponen et al. (1998). There are, however, comparatively few
published applications of the technique in the Australian context. See, for example, Smith (1979) and Donovan (1981).
8
For overviews of these criticisms see, for example, Hewings (1977), Richardson (1978), Fothergill and Gudgin (1979), Stevens
and Moore (1980), Dawson (1982) and Dinc et al. (1998).
9
See, for example, Beeson (1987), Haynes and Machunda (1987), Barff and Knight (1988), Knudsen and Barff (1991),
Markusen et al. (1991), Rigby (1992), Rigby and Anderson (1993), Noponen et al. (1996), Noponen et al. (1997), Dinc and
Haynes (1998a), Dinc and Haynes (1998b), Graham and Spence (1998) and Noponen et al. (1998).
– 69 –
Productivity and Regional Economic Performance in Australia
Traditional comparative static shift-share analysis
The traditional shift-share model decomposes change (i.e. growth or decline) in an economic
variable (e.g. output or employment) into three elements:
• that due to the national rate of change in the variable of interest (labelled the ‘national
growth’ effect);
• that due to the industrial structure of the region (labelled the ‘industrial mix’or ‘proportionality’
shift); and
• a residual element (labelled the ‘competitive’ or ‘differential’ shift) traditionally interpreted
as indicating the locational advantages (or disadvantages) of the regional economy.
The sum of these three elements equals the ‘actual change’ in the variable of interest within
the region over the given time period, while the sum of the last two elements is often
referred to as the ‘total shift’. Appendix 4 provides technical details of this variant of the
shift-share model.
Dynamic shift-share analysis
A common early set of criticisms of the traditional shift-share model related to its treatment
of the time dimension. In particular, the traditional model considers conditions only at the
beginning and the end of an often quite lengthy study period, with the industrial mix in the
initial year used to define the proportionality (and hence the differential) effect. This ignores,
of course, any intervening changes in the industry mix at the national level. A number of
potential solutions have been proposed to this choice-of-weights problem. Stillwell (1969)
recommended the use of final year weights rather than the traditional initial year weights for
industry mix. Fuchs (1962) argued for the use of an average of initial and final year weights.
Thirlwall (1967) suggested that the study period be divided into a number of sub-periods to
reduce the impact of this problem. Barf and Knight (1988) addressed this issue and adapted
the traditional model for use with annual data over an extended time period. Their dynamic
shift-share model adjusts annually for change in industry mix at the national level as well as
continuously updating the size of the region’s total GSP. It is argued that by using annual
growth rates their dynamic approach provides for a more accurate allocation of regional GSP
change among the three components defined above.10 Appendix 4 provides technical details
of this variant of the shift-share model.
Simple productivity-extended shift-share analysis
Population growth rates and labour force participation rates differ between regions at any
given point in time (and even over extended periods of time). A region experiencing lower
(higher) population growth and/or lower (higher) labour force participation growth will,
other things being equal, display slower (faster) than expected growth in employment. Interregional differences in employment growth rates in turn impact on the corresponding GSP
growth rates. Similarly, inter-regional differences in labour productivity change (both in
10 For further discussion of the justification for this claim, see Barf and Knight (1988) and Dinc et al. (1998).
– 70 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
aggregate and for any given sector) also influence the corresponding GSP growth rates.
Unless the impacts of these two factors (employment growth and productivity change) can
be identified separately, then the insights into regional GSP growth performance provided
by shift-share analysis are somewhat limited.
Rigby and Anderson (1993) developed a productivity-extended version of dynamic shiftshare analysis that can be adapted to permit the various shift-share effects described above
to be further disaggregated into two parts:
• a productivity constant component, where GSP per hour worked is in effect held
constant but employment levels (hours worked) are allowed to change; and
• an employment constant component, where employment levels (hours worked) are in
effect held constant but GSP per hour worked is allowed to change.11
Appendix 4 provides technical details of our adaptation of this variant of the shift-share
model.
Multifactor productivity-extended shift-share analysis
Haynes and Dinc (1997) argue that the simple productivity-extended shift-share method
developed by Rigby and Anderson (1993) ignores the contribution to inter-regional growth
differences made by factors of production other than labour. As the total factor productivity
literature suggests, the results derived from a labour-only study of productivity could be
misleading and should be interpreted with caution. Appendix 4 provides technical details of
our adaptation of the multifactor productivity extended variant of the shift-share model.12
Results of Application of Comparative-Static and Dynamic
Shift-Share Analysis
Aggregate level results
The aggregate results from the application of the comparative-static and dynamic versions
of shift-share analysis are given in Tables 4.1 and 4.2 respectively. Although the numerical
values of the various shift-share components differ between these two tables, the signs of
the total, proportionality and differential shifts remain unaffected by the choice of method.
11 The Rigby and Anderson (1993) method has to date been applied only in the context of analysing employment change and not
GDP change. As a result their method involves decomposing employment change effects (total shift, proportionality shift,
differential shift, etc) into a productivity constant component (where output levels per employee are held constant but output
levels are allowed to change) and an output constant component (where output levels are held constant but output per
employee levels are allowed to change).
12 The Haynes and Dinc (1997) method has to date been applied only in the context of analysing employment change and not GDP
change, hence the need for their method to be adapted for use in this chapter.
– 71 –
Productivity and Regional Economic Performance in Australia
Table 4.1: Aggregate Results from Application of
Comparative-Static Shift-Share Analysis
(a) GSP/GDP at factor cost ($m), 1985-86 to 1998-99
Actual
change
National
growth
Total
shift
Proportionality
shift
Differential
shift
Boudeville
classification
NSW
59,003.304
60,390.340
-1,387.036
1,700.671
-3,087.707
Type 4
Vic
38,111.126
47,288.503
-9,177.378
-908.342
-8,269.035
Type 8
Qld
34,353.496
24,592.481
9,761.015
-126.976
9,887.990
Type 6
SA
8,548.359
13,887.162
-5,338.803
-551.366
-4,787.438
Type 8
WA
24,433.610
16,294.784
8,138.826
601.725
7,537.102
Type 2
Tas
2,052.916
4,025.440
-1,972.523
-517.791
-1,454.732
Type 8
NT
2,087.267
2,018.937
68.330
187.249
-118.919
Type 3
ACT
3,497.017
3,589.448
-92.431
-385.169
292.739
Type 5
Aust
172,087.096
172,087.096
17,968.171
2,489.644
17,717.831
(b) GSP/GDP at factor cost (%), 1985-86 to 1998-99
Actual
change
National
growth
Total
shift
NSW
34.29
35.09
-7.72
68.31
-17.43
Type 4
Vic
22.15
27.48
-51.08
-36.48
-46.67
Type 8
Qld
19.96
14.29
54.32
-5.10
55.81
Type 6
SA
4.97
8.07
-29.71
-22.15
-27.02
Type 8
WA
14.20
9.47
45.30
24.17
42.54
Type 2
Tas
1.19
2.34
-10.98
-20.80
-8.21
Type 8
NT
1.21
1.17
0.38
7.52
-0.67
Type 3
ACT
2.03
2.09
-0.51
-15.47
1.65
Type 5
Aust
100.00
100.00
100.00
100.00
100.00
– 72 –
Proportionality
shift
Differential Boudeville
shift
classification
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
Table 4.2: Aggregate Results from Application of Dynamic Shift-Share Analysis
(a) GSP/GDP at factor cost ($m), 1985-86 to 1998-99
Actual
change
National
growth
Total
shift
Proportionality
shift
Differential Boudeville
shift
classification
NSW
59,003.304
59,489.362
-486.058
1,532.396
-2,018.453
Type 4
Vic
38,111.126
45,442.506
-7,331.380
-892.706
-6,438.674
Type 8
Qld
34,353.496
26,829.506
7,523.990
-261.720
7,785.710
Type 6
SA
8,548.359
12,767.928
-4,219.569
-708.891
-3,510.678
Type 8
WA
24,433.610
18,206.665
6,226.945
828.249
5,398.696
Type 2
Tas
2,052.916
3,699.954
-1,647.038
-347.744
-1,299.294
Type 8
NT
2,087.267
2,000.663
86.604
158.045
-71.441
Type 3
ACT
3,497.017
3,650.512
-153.495
-307.628
154.133
Type 5
Aust
172,087.096
172,087.096
13,837.539
5,827.491
13,597.441
(b) GSP/GDP at factor cost (%), 1985-86 to 1998-99
Actual
change
National
growth
Total
shift
Proportionality
shift
NSW
34.29
34.57
-3.51
26.30
-14.84
Type 4
Vic
22.15
26.41
-52.98
-15.32
-47.35
Type 8
Qld
19.96
15.59
54.37
-4.49
57.26
Type 6
SA
4.97
7.42
-30.49
-12.16
-25.82
Type 8
WA
14.20
10.58
45.00
14.21
39.70
Type 2
Tas
1.19
2.15
-11.90
-5.97
-9.56
Type 8
NT
1.21
1.16
0.63
2.71
-0.53
Type 3
ACT
2.03
2.12
-1.11
-5.28
1.13
Type 5
Aust
100.00
100.00
100.00
100.00
100.00
– 73 –
Differential Boudeville
shift
classification
Productivity and Regional Economic Performance in Australia
The net proportionality shift (column 4 of Tables 4.1 and 4.2) highlights the effects of
industrial mix or structure on a region’s growth performance. Only three regions, New
South Wales, Western Australia and the Northern Territory, experienced positive net
proportionality shifts. They were the only regions with industrial mixes conducive to
growth in GSP higher than the national average.
The net differential shift (column 5 of Tables 4.1 and 4.2) highlights those regions that
experienced GSP growth performances of a different sign and/or magnitude to that expected
on the basis of their industrial structure. Three regions, Queensland, Western Australia and
the Australian Capital Territory, experienced positive differential shifts, each capturing a
larger share of the nation’s GSP growth than their industrial structures would have
suggested, with Queensland and Western Australia being the most significant in this respect.
All other regions experienced negative differential shifts, with Victoria and South Australia
figuring most prominently in this regard.
The net total shift (column 3 of Tables 4.1 and 4.2) combines the corresponding net
proportionality and differential shifts. These results demonstrate clearly that the agricultural
and mining based regions of Queensland, Western Australia and the Northern Territory are
the only ones to record positive net total shifts. Queensland and Western Australia, for
example, accounted for 54.3% and 45.3% of the nation’s positive net total shift respectively
in Table 4.1.13 All other regions experienced negative net total shifts, with Victoria and
South Australia accounting for 51.1% and 29.7% of the nation’s negative net total shift
respectively in Table 4.1.14
Column 6 of Tables 4.1 and 4.2 classifies regions according to the relative size of the
proportional and differential shifts along the lines of the Boudeville (1966) system given in
Figure 4.1.15
Figure 4.1: Classification of Regions according to Total Shift
(Comparative-Static and Dynamic Approaches)
Proportionality Shift (PS)
Positive
Differential
Shift (DS)
Negative
Positive
Negative
PS>DS
Type 1
Type 5
DS>PS
Type 2
Type 6
PS>DS
Type 3
Type 7
DS>PS
Type 4
Type 8
Source: Adapted from Boudeville (1966).
13 The corresponding values in Table 4.2 are 54.4% and 45.0% respectively.
14 The corresponding values in Table 4.2 are 53.0% and 30.5% respectively.
15 In Figure 4.1 the left-hand column (PS>0) contains regions with industrial structures conducive to higher than average GSP
growth, while the right-hand side column (PS<0) contains regions with industrial structures conducive to lower than average
GSP growth. The upper two rows (DS>0) contain regions with GSP growth higher than their industrial structure would suggest,
while the lower two rows (DS<0) contain regions with GSP growth lower than expected given their industrial structure.
– 74 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
Western Australia is a type 2 region. It is the only state that, over the study period as a
whole, displayed an industrial structure conducive to real GSP growing at a rate higher than
the national average (PS>0), and concomitantly experienced even higher real GSP growth
rates than would have been expected on this account alone (DS>0).
Like Western Australia, the regions of New South Wales and the Northern Territory also
displayed industrial structures conducive to GSP growing at a rate faster than the national
average (PS>0). However, unlike Western Australia, they experienced GSP growth lower
than would be expected on the basis of this initial advantage (DS<0). In the case of New
South Wales, the positive industrial mix (PS) effect was more than offset by the negative
regional attractiveness (DS) effect, to produce a negative total shift and a type 4 regional
typology. By contrast, for the Northern Territory the negative regional attractiveness effect
was insufficient to offset the positive effects of its industrial structure and it recorded a
positive total shift and a type 3 regional typology.
Queensland and the Australian Capital Territory have industrial structures that were not
conducive to GSP growing at a rate higher than the national average (PS<0). Nevertheless,
they experienced GSP growth higher than would be expected on the basis of this initial
disadvantage (DS>0). In the case of Queensland, this positive regional attractiveness (DS)
effect was sufficiently large to offset the negative effects of its industrial mix (PS) with the
result that it recorded a positive total shift and a type 6 typology. By contrast, the Australian
Capital Territory produced a negative total shift and is classified as a type 5 region.
All other regions, Tasmania, Victoria and South Australia, were in the unenviable position
of not only having a disproportionate proportion of slower growth industries relative to the
national average, but in addition experiencing slower growth than would have been
expected on this account alone. They are classified as type 8 regions.
Disaggregated industry level results
Further insights can be obtained from a more detailed analysis of results focusing on
particular industries and regions. The results disaggregated by industry are given for the
various shifts in Tables 4.3 and 4.4 for the comparative-static and dynamic versions of the
method respectively.
From these tables, especially the third part of the tables (headed ‘proportionality shift’), it
can be seen that the ‘high’ growth sectors, which contribute positively to each region’s
proportionality shift, are mining; accommodation, cafes and restaurants; communication
services; finance and insurance; and property and business services. All other sectors are
‘slow’ growth sectors, which contribute negatively to proportionality shifts in each region.16
For brevity in what follows, we discuss in the text results for the dynamic shift-share
analysis (i.e. Table 4.4 results) and point out via footnotes differences that arise from the use
of comparative-static shift-share analysis (i.e. Table 4.3 results).
16 High (slow) growth sectors are those that have sectoral growth rates at the national level that exceed (fall below) the overall (all
industry) national growth rate.
– 75 –
Productivity and Regional Economic Performance in Australia
Table 4.3: Disaggregated Results from Application
of Comparative-Static Shift-Share Analysis
(a) Actual GSP/GDP change ($m), 1985-86 to 1998-99
Sector
SA
WA
1,325.995 1,425.451 1,028.733
889.718
384.453
Mining
1,591.941
Manufacturing
Agriculture, forestry and fishing
Electricity, gas and water
Construction
NSW
NT
ACT
Aust
244.690
63.225
-2.408
5,359.857
-63.958 7,893.555
130.331
484.656
-4.534 12,514.888
4,616.799 3,673.110 3,203.638 1,421.094 1,736.799
134.469
97.125
-7.752 14,875.283
817.535
Vic
Qld
425.235 2,057.662
251.859
747.342
4,837.774 1,682.784 2,118.468
574.791
Tas
729.517
162.076
43.947
196.308
220.901 2,237.452
-10.085
38.997
222.591 11,348.882
3,523.375
Wholesale trade
2,970.381 2,116.675 2,281.118
170.762
873.035
23.287
50.401
54.244
8,539.904
Retail trade
2,758.244 1,456.722 2,474.423
268.246
942.599
143.293
33.457
144.033
8,221.019
310.517
Accommodation, cafes and restaurants 1,946.162
320.450
100.100
87.287
100.690
5,162.120
Transport and storage
2,732.114 2,240.120 2,067.542
590.863 1,048.112
51.577
139.766
103.428
8,973.523
Communication services
4,212.049 3,334.390 2,102.629
635.018 1,248.029
185.600
167.496
255.707 12,140.918
8,929.695 6,968.003 3,169.871 1,106.802 1,591.495
374.296
68.577
225.356 22,434.094
12,684.359 7,145.158 3,604.212 1,061.052 2,576.169
-5.766
284.340
607.308 27,956.831
Finance and insurance
(incl. nominal industry)
Property and business services
693.959 1,602.954
Government administration
and defence
1,515.131
71.002 1,620.751
38.679
251.708
176.798
116.492
993.580
4,784.141
Education
2,579.147 1,647.766 1,770.950
304.093
591.081
65.548
148.375
115.040
7,221.999
Health and community services
3,103.064 2,916.311 2,650.760
684.588 1,236.952
277.172
177.570
256.779 11,303.197
Cultural and recreational services
1,122.300 1,278.451
Personal and other services
1,260.614
All industries (net)
720.595
91.081
279.039
-39.384
87.886
45.316
3,585.285
784.130 1,131.846
244.111
493.165
38.914
-2.329
191.331
4,141.781
59,003.304 38,111.126 34,353.496 8,548.359 24,433.610 2,052.916 2,087.267 3,497.017 172,087.096
(b) National shift ($m), 1985-86 to 1998-99
Sector
Agriculture, forestry and fishing
Mining
Manufacturing
NSW
Vic
Qld
SA
WA
NT
ACT
Aust
243.806
106.925
11.402
7,637.977
563.234 1,464.018
49.732
228.778
3.934
6,378.246
12,205.423 10,954.126 3,778.255 2,822.616 2,308.884
806.573
96.579
2,149.263 1,601.005 1,571.016
774.007 1,180.553
1,037.359 1,485.020 1,546.171
871.040
432.412
Tas
Electricity, gas and water
2,056.495 1,660.156
503.289
228.220
34.267
Construction
4,363.046 3,311.478 2,283.259 1,026.612 1,275.186
320.328
211.829
115.213 33,087.668
26.607
5,812.486
255.476 13,047.215
Wholesale trade
4,538.313 3,578.372 1,631.183
860.221 1,037.953
221.335
80.001
103.457 12,050.835
Retail trade
4,330.490 3,330.456 2,019.930 1,094.277 1,188.036
335.850
185.092
207.632 12,691.763
Accommodation, cafes and restaurants 1,855.990 1,011.502
395.671
105.581
69.909
Transport and storage
4,120.565 2,485.646 1,808.776
753.926
739.360 1,078.713
327.980
230.295
113.285
Communication services
74.606
35.686
1,333.862 1,015.879
542.433
260.317
295.729
Finance and insurance
(incl. nominal industry)
3,006.538 1,628.728
491.069
382.415
388.668
54.237
61.409
Property and business services
6,401.315 4,842.529 2,214.946 1,167.019 1,542.596
284.386
159.373
92.241
4,612.798
145.662 10,722.303
53.816
3,612.328
76.047
6,089.111
378.191 16,990.356
Government administration
and defence
2,692.284 2,114.476 1,098.881
658.116
695.130
209.871
220.253 1,311.113
9,000.124
Education
3,161.578 2,858.999 1,336.914
898.569
884.390
285.973
121.473
329.201
9,877.096
Health and community services
4,074.940 3,193.630 1,580.425 1,151.380 1,231.643
339.337
124.649
202.148 11,898.153
Cultural and recreational services
1,443.812
958.077
419.680
297.736
361.579
113.288
58.874
160.810
3,813.856
Personal and other services
1,619.067 1,258.422
644.579
430.890
462.746
122.024
110.554
116.498
4,764.780
All industries (net)
60,390.340 47,288.503 24,592.481 13,887.162 16,294.784 4,025.440 2,018.937 3,589.448 172,087.096
– 76 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
(c) Proportionality shift ($m), 1985-86 to 1998-99
Sector
NSW
Agriculture, forestry and fishing
WA
Tas
NT
ACT
-352.114
-72.718
-31.892
-3.401
-2,278.120
541.899 1,408.562
47.848
220.112
3.785
6,136.642
-6,718.209 -6,029.460 -2,079.658 -1,553.647 -1,270.875
-443.960
-53.160
-641.044
Mining
Vic
-477.519
Qld
-468.575
SA
-230.857
998.064 1,428.769 1,487.603
Manufacturing
Aust
-63.417 -18,212.386
Electricity, gas and water
-809.902
-653.813
-343.038
-170.295
-198.208
-89.879
-13.495
-10.479
-2,289.110
Construction
-567.930
-431.049
-297.208
-133.632
-165.989
-41.696
-27.573
-33.255
-1,698.332
Wholesale trade
-1,322.208 -1,042.535
-475.234
-250.620
-302.401
-64.484
-23.308
-30.141
-3,510.931
Retail trade
-1,525.439 -1,173.172
-711.532
-385.465
-418.492
-118.305
-65.200
-73.140
-4,470.745
Accommodation, cafes and restaurants
221.023
120.456
89.782
39.058
47.119
12.573
8.325
10.985
549.321
Transport and storage
-672.053
-405.402
-295.007
-120.588
-175.935
-37.561
-18.477
-23.757
-1,748.780
3,149.206 2,398.459 1,280.667
614.600
698.207
176.142
84.253
127.057
8,528.591
Communication services
Finance and insurance
(incl. nominal industry)
8,070.441 4,371.989 1,318.175 1,026.516 1,043.301
145.588
164.839
204.133 16,344.983
Property and business services
4,131.748 3,125.624 1,429.644
753.256
995.673
183.558
102.868
244.105 10,966.475
Government administration
and defence
-1,261.163
-990.497
-514.756
-308.285
-325.624
-98.311
-103.175
-614.173
-4,215.984
Education
-849.875
-768.538
-359.380
-241.547
-237.736
-76.873
-32.654
-88.494
-2,655.097
Health and community services
-203.763
-159.694
-79.028
-57.574
-61.587
-16.968
-6.233
-10.108
-594.955
Cultural and recreational services
-86.530
-57.419
-25.152
-17.844
-21.670
-6.790
-3.528
-9.638
-228.572
-211.694
-164.540
-84.279
-56.339
-60.504
-15.955
-14.455
-15.232
-622.999
1,700.671
-908.342
-126.976
-551.366
601.725
-517.791
187.249
-385.169
0.000
Personal and other services
All industries (net)
(d) Differential shift ($m), 1985-86 to 1998-99
Sector
NSW
Vic
SA
WA
Tas
NT
346.568
-443.986
73.602
-11.808
-10.409
-976.112 -1,169.091 5,020.976
32.751
35.765
-12.252
0.000
-228.144
53.706
-59.549
0.000
Agriculture, forestry and fishing
-182.224
Mining
-443.482 -2,488.554
Manufacturing
-870.415 -1,251.556 1,505.041
Electricity, gas and water
Construction
Wholesale trade
Retail trade
-429.057
301.965
Qld
-73.708
152.126
698.790
ACT
Aust
0.000
-754.484
219.341
312.674
424.436
23.735
23.175
180.179
0.000
1,042.657 -1,197.645
132.417
-672.079 1,128.255
-288.716
-145.259
0.370
0.000
-245.725
-419.162 1,125.170
-438.839
137.483
-133.563
-6.293
-19.071
0.000
-46.806
-700.562 1,166.025
-440.566
173.055
-74.252
-86.435
9.541
0.000
Accommodation, cafes and restaurants
-130.850
-437.999
759.246
-56.521
-122.340
-18.054
9.053
-2.536
0.000
Transport and storage
-716.398
159.877
553.773
-27.909
145.334
-141.158
44.957
-18.477
0.000
Communication services
-271.019
-79.948
279.529
-239.898
254.093
-65.148
47.557
74.835
0.000
Finance and insurance
(incl. nominal industry)
-2,147.284
967.285 1,360.627
-302.129
159.526
174.470
-157.671
-54.825
0.000
Property and business services
2,151.296
-822.995
-40.378
-859.224
37.901
-473.710
22.098
-14.988
0.000
84.010 -1,052.977 1,036.626
0.000
Government administration
and defence
-311.152
-117.798
65.239
-0.587
296.639
Education
267.444
-442.695
793.417
-352.929
-55.573
-143.551
59.556
-125.668
0.000
Health and community services
-768.113
-117.625 1,149.363
-409.219
66.896
-45.197
59.154
64.740
0.000
Cultural and recreational services
-234.982
377.793
326.067
-188.811
-60.869
-145.882
32.540
-105.856
0.000
Personal and other services
-146.758
-309.753
571.546
-130.440
90.923
-67.155
-98.429
90.065
0.000
-3,087.707 -8,269.035 9,887.990 -4,787.438 7,537.102 -1,454.732
-118.919
292.739
0.000
All industries (net)
– 77 –
Productivity and Regional Economic Performance in Australia
(e) Total shift ($m), 1985-86 to 1998-99
Sector
NSW
Agriculture, forestry and fishing
Mining
-823.268
Vic
-175.554
554.582 -1,059.785
Qld
SA
WA
Tas
NT
-542.283
115.711
-796.100
0.883
-43.700
511.491
-627.192 6,429.538
80.599
255.877
-672.104
0.547
-122.965 -18,212.386
-572.085
Aust
-2,278.120
-8.468
6,136.642
Manufacturing
-7,588.624 -7,281.016
Electricity, gas and water
-1,238.959 -1,408.298
-123.697
142.379
226.228
-66.144
9.680
169.701
-2,289.110
474.727 -1,628.694
-164.791
-805.711
962.266
-330.412
-172.833
-32.885
-1,698.332
Construction
-574.617 -1,401.521
ACT
-13.810
Wholesale trade
-1,567.933 -1,461.697
649.935
-689.458
-164.918
-198.047
-29.600
-49.213
-3,510.931
Retail trade
-1,572.246 -1,873.734
454.493
-826.031
-245.437
-192.557
-151.634
-63.599
-4,470.745
Accommodation, cafes and restaurants
90.173
-317.543
849.029
-17.463
-75.221
-5.481
17.378
8.449
549.321
Transport and storage
-1,388.451
-245.525
258.767
-148.497
-30.601
-178.718
26.481
-42.234
-1,748.780
Communication services
2,878.187 2,318.510 1,560.196
374.701
952.300
110.994
131.810
201.892
8,528.591
Finance and insurance
(incl. nominal industry)
5,923.157 5,339.275 2,678.802
724.387 1,202.826
320.059
7.168
149.309 16,344.983
Property and business services
6,283.044 2,302.629 1,389.265
-105.968 1,033.574
-290.152
124.966
229.116 10,966.475
Government administration
and defence
-1,177.153 -2,043.474
521.870
-619.437
-443.422
-33.072
-103.762
-317.533
-4,215.984
Education
-582.431 -1,211.233
434.036
-594.476
-293.309
-220.425
26.902
-214.161
-2,655.097
Health and community services
-971.876
-277.319 1,070.336
-466.792
5.309
-62.165
52.921
54.632
-594.955
Cultural and recreational services
-321.512
320.374
300.915
-206.655
-82.539
-152.672
29.012
-115.493
-228.572
Personal and other services
-358.453
-474.293
487.267
-186.780
30.419
-83.110
-112.884
74.833
-622.999
-1,387.036 -9,177.378 9,761.015 -5,338.803 8,138.826 -1,972.523
68.330
-92.431
0.000
All industries (net)
Table 4.4: Disaggregated Results from Application of Dynamic Shift-Share Analysis
(a) Actual GSP/GDP change ($m), 1985-86 to 1998-99
Sector
SA
WA
1,325.995 1,425.451 1,028.733
889.718
384.453
Mining
1,591.941
Manufacturing
Agriculture, forestry and fishing
Electricity, gas and water
Construction
NSW
NT
ACT
Aust
244.690
63.225
-2.408
5,359.857
-63.958 7,893.555
130.331
484.656
-4.534 12,514.888
4,616.799 3,673.110 3,203.638 1,421.094 1,736.799
134.469
97.125
-7.752 14,875.283
817.535
Vic
Qld
425.235 2,057.662
251.859
747.342
4,837.774 1,682.784 2,118.468
574.791
Tas
729.517
162.076
43.947
196.308
220.901 2,237.452
-10.085
38.997
222.591 11,348.882
3,523.375
Wholesale trade
2,970.381 2,116.675 2,281.118
170.762
873.035
23.287
50.401
54.244
8,539.904
Retail trade
2,758.244 1,456.722 2,474.423
268.246
942.599
143.293
33.457
144.033
8,221.019
310.517
Accommodation, cafes and restaurants 1,946.162
320.450
100.100
87.287
100.690
5,162.120
Transport and storage
2,732.114 2,240.120 2,067.542
590.863 1,048.112
51.577
139.766
103.428
8,973.523
Communication services
4,212.049 3,334.390 2,102.629
635.018 1,248.029
185.600
167.496
255.707 12,140.918
8,929.695 6,968.003 3,169.871 1,106.802 1,591.495
374.296
68.577
225.356 22,434.094
12,684.359 7,145.158 3,604.212 1,061.052 2,576.169
-5.766
284.340
607.308 27,956.831
Finance and insurance
(incl. nominal industry)
Property and business services
693.959 1,602.954
Government administration
and defence
1,515.131
71.002 1,620.751
38.679
251.708
176.798
116.492
993.580
4,784.141
Education
2,579.147 1,647.766 1,770.950
304.093
591.081
65.548
148.375
115.040
7,221.999
Health and community services
3,103.064 2,916.311 2,650.760
684.588 1,236.952
277.172
177.570
256.779 11,303.197
Cultural and recreational services
1,122.300 1,278.451
Personal and other services
1,260.614
All industries (net)
720.595
91.081
279.039
-39.384
87.886
45.316
3,585.285
784.130 1,131.846
244.111
493.165
38.914
-2.329
191.331
4,141.781
59,003.304 38,111.126 34,353.496 8,548.359 24,433.610 2,052.916 2,087.267 3,497.017 172,087.096
– 78 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
(b) National shift ($m), 1985-86 to 1998-99
Sector
NSW
Agriculture, forestry and fishing
Mining
Manufacturing
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
6.363
6,874.316
2.594
8,325.672
704.226 1,017.140
1,229.037 1,427.632 1,675.389
403.744 3,115.322
98.161
373.793
10,739.481 9,748.478 3,754.341 2,621.662 2,140.089
686.604
108.163
Electricity, gas and water
1,898.576 1,524.989
Construction
4,033.322 2,652.528 2,212.150
892.237
398.693
228.296
79.671
1,820.746 1,584.167 1,433.708
553.521
201.218
42.655
811.615 1,441.048
264.268
156.251
101.805 29,900.624
63.443
5,575.331
244.564 11,815.746
Wholesale trade
4,001.189 2,993.393 1,566.224
671.578 1,003.634
172.566
70.567
92.846 10,571.997
Retail trade
3,821.532 2,840.657 2,076.393
882.537 1,111.520
297.099
136.702
197.454 11,363.894
Accommodation, cafes and restaurants 1,860.117
375.377
111.005
75.564
Transport and storage
3,796.746 2,418.187 1,923.587
896.362
941.840
773.570 1,016.998
306.234
192.847
111.306
Communication services
1,800.138 1,453.260
86.001
52.939
90.575
4,657.074
131.485 10,364.725
790.907
324.037
458.161
Finance and insurance
(incl. nominal industry)
90.674
5,056.117
4,419.208 2,737.340 1,209.192
624.028
691.844
148.688
58.315
124.600 10,013.216
Property and business services
7,546.820 5,062.445 2,512.296 1,119.540 1,754.671
234.488
157.323
457.260 18,844.842
Government administration
and defence
2,467.187 1,928.112 1,274.805
503.571
633.616
215.455
191.897 1,280.889
8,495.533
Education
3,140.501 2,718.952 1,536.134
810.685
869.307
242.722
120.283
289.638
9,728.223
Health and community services
3,972.127 3,333.191 1,811.298 1,125.395 1,207.262
336.220
131.945
205.423 12,122.861
Cultural and recreational services
1,384.897
969.649
513.069
245.371
321.782
81.732
60.445
132.969
3,709.913
Personal and other services
1,557.738 1,153.165
705.937
441.442
495.372
102.583
72.842
137.931
4,667.010
All industries (net)
59,489.362 45,442.506 26,829.506 12,767.928 18,206.665 3,699.954 2,000.663 3,650.512 172,087.096
(c) Proportionality shift ($m), 1985-86 to 1998-99
Sector
Agriculture, forestry and fishing
Mining
Manufacturing
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
-95.879
-54.048
-17.673
-1.454
1,514.459
260.712 1,126.527
67.470
223.570
2.591
4,189.215
-5,334.157 -4,881.262 -1,890.230 -1,350.745 -1,112.881
-349.358
-56.202
-50.505 -15,025.341
-28.123 -2,051.956
-484.953
-429.786
-328.836
-101.829
702.737
920.633
884.974
Electricity, gas and water
-699.905
-529.282
-345.233
-149.649
-211.533
-74.079
-14.153
Construction
-136.163
-289.069
17.806
-84.108
36.129
1.624
-18.936
Wholesale trade
-692.620
-673.942
-289.396
-139.483
-162.587
-40.704
-13.190
-20.171 -2,032.093
Retail trade
5.855
-466.864
-1,059.714
-828.421
-516.789
-262.817
-298.007
-82.393
-42.061
-52.674
Accommodation, cafes and restaurants
192.855
102.281
105.980
34.607
37.694
11.869
7.754
12.006
505.045
Transport and storage
-536.900
-336.416
-236.149
-89.999
-134.067
-26.489
-15.399
-15.783
-1,391.202
2,532.664 2,034.754 1,099.685
455.700
643.519
121.329
70.509
126.641
7,084.801
Communication services
-3,142.876
Finance and insurance
(incl. nominal industry)
5,763.092 3,458.965 1,274.780
716.393
805.926
153.271
86.517
161.934 12,420.878
Property and business services
3,662.603 2,432.727 1,222.149
537.971
843.566
111.891
80.894
220.188
9,111.989
Government administration
and defence
-1,104.708
-792.282
-583.074
-214.274
-273.339
-93.663
-85.846
-564.206
-3,711.393
Education
-807.482
-686.418
-433.692
-200.125
-219.285
-59.315
-31.532
-68.375
-2,506.224
Health and community services
-246.664
-227.224
-156.616
-63.204
-84.608
-18.897
-7.930
-14.520
-819.664
Cultural and recreational services
-45.678
-36.267
-8.763
-7.937
-11.341
-6.188
-2.075
-6.378
-124.629
-172.609
-131.695
-78.317
-50.103
-61.585
-10.064
-6.201
-14.654
-525.229
1,532.396
-892.706
-261.720
-708.891
828.249
-347.744
158.045
-307.628
0.000
Personal and other services
All industries (net)
– 79 –
Productivity and Regional Economic Performance in Australia
(d) Differential shift ($m), 1985-86 to 1998-99
Sector
Agriculture, forestry and fishing
NSW
-9.797
Vic
Qld
SA
WA
Tas
NT
ACT
287.321
-536.808
-728.414 3,651.706
-202.777
Aust
70.441
1.228
-7.317
-35.301
-112.708
-9.719
0.000
45.164
-59.052
0.000
271.070
-76.139
Mining
-339.833 -1,923.030
-502.701
Manufacturing
-788.525 -1,194.106 1,339.527
Electricity, gas and water
-381.136
-743.848
200.338
325.747
387.529
34.938
15.444
160.988
0.000
Construction
940.614
-680.674
-111.488
-506.606
760.276
-275.976
-98.319
-27.828
0.000
Wholesale trade
-338.188
150.177
709.591
0.000
-202.775 1,004.290
-361.332
31.988
-108.575
-6.976
-18.431
0.000
-3.573
-555.514
-351.474
129.086
-71.412
-61.184
-0.747
0.000
Accommodation, cafes and restaurants
-106.810
-304.685
555.134
-30.323
-92.621
-22.774
3.969
-1.891
0.000
Transport and storage
-527.731
158.350
380.104
-92.708
165.181
-114.781
43.859
-12.274
0.000
Communication services
-120.753
-153.625
212.037
-144.719
146.349
-21.730
44.048
38.392
0.000
Retail trade
914.819
Finance and insurance
(incl. nominal industry)
-1,252.605
771.698
685.899
-233.619
93.724
72.336
-76.255
-61.178
0.000
Property and business services
1,474.937
-350.014
-130.234
-596.458
-22.068
-352.145
46.123
-70.140
0.000
Government administration
and defence
152.651 -1,064.827
929.020
-250.619
-108.569
55.006
10.441
276.897
0.000
Education
246.128
-384.768
668.508
-306.467
-58.942
-117.859
59.624
-106.224
0.000
Health and community services
-622.399
-189.656
996.079
-377.602
114.297
-40.151
53.555
65.877
0.000
Cultural and recreational services
-216.919
345.070
216.290
-146.353
-31.401
-114.928
29.516
-81.274
0.000
Personal and other services
-124.514
-237.340
504.226
-147.229
59.378
-53.606
-68.970
68.055
0.000
-2,018.453 -6,438.674 7,785.710 -3,510.678 5,398.696 -1,299.294
-71.441
154.133
0.000
All industries (net)
(e) Total shift ($m), 1985-86 to 1998-99
Sector
Agriculture, forestry and fishing
Mining
NSW
-494.750
Vic
-158.716
362.904 -1,002.397
Qld
SA
WA
Tas
NT
ACT
-404.974
185.492
-632.687
16.394
-16.446
-8.771
-1,514.459
382.273
-467.702 4,778.234
32.170
110.863
-7.128
4,189.215
-552.135
-11.038
Manufacturing
-6,122.682 -6,075.368
-550.703 -1,200.568
Electricity, gas and water
-1,081.041 -1,273.130
-403.290
Aust
-109.557 -15,025.341
-144.894
176.098
175.996
-39.142
1.291
132.865
804.451
-969.743
-93.682
-590.714
796.404
-274.352
-117.255
-21.973
-466.864
Wholesale trade
-1,030.809
-876.717
714.894
-500.815
-130.599
-149.279
-20.166
-38.602
-2,032.093
Retail trade
-1,063.288 -1,383.935
-3,142.876
Construction
-2,051.956
398.030
-614.292
-168.921
-153.806
-103.245
-53.421
86.045
-202.403
661.114
4.284
-54.927
-10.905
11.723
10.114
505.045
Transport and storage
-1,064.631
-178.066
143.955
-182.707
31.114
-141.270
28.459
-28.057
-1,391.202
Communication services
789.868
99.599
114.557
165.033
7,084.801
Accommodation, cafes and restaurants
2,411.911 1,881.130 1,311.722
310.981
Finance and insurance
(incl. nominal industry)
4,510.487 4,230.662 1,960.680
482.774
899.650
225.607
10.262
Property and business services
5,137.540 2,082.713 1,091.915
-58.488
821.498
-240.254
127.017
150.048
9,111.989
100.756 12,420.878
Government administration
and defence
-952.057 -1,857.110
345.946
-464.892
-381.908
-38.657
-75.405
-287.309
-3,711.393
Education
-561.354 -1,071.186
234.816
-506.592
-278.226
-177.174
28.092
-174.598
-2,506.224
Health and community services
-869.063
-416.880
839.463
-440.807
29.689
-59.047
45.625
51.356
-819.664
Cultural and recreational services
-262.598
308.802
207.527
-154.290
-42.743
-121.116
27.441
-87.653
-124.629
Personal and other services
-297.124
-369.035
425.909
-197.332
-2.207
-63.669
-75.172
53.401
-525.229
All industries (net)
-486.058 -7,331.380 7,523.990 -4,219.569 6,226.945 -1,647.038
86.604
-153.495
0.000
– 80 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
New South Wales
All ‘high’ growth sectors expanded less rapidly in New South Wales than in the nation as a
whole (DS<0), with the exception of the property and business services sector. In all cases,
this negative DS was too small to outweigh the positive PS so that a positive total shift was
recorded for these sectors.
All ‘slow’ growth sectors also expanded less rapidly in New South Wales than in the nation
as a whole (DS<0), with the exception of the construction, government administration and
defence, and education sectors. Of the exceptions, only the construction sector had its
positive DS large enough to more than offset the negative PS to produce a positive total shift
for this sector in New South Wales.
Victoria
All ‘high’ growth sectors expanded less rapidly in Victoria than in the nation as a whole
(DS<0), with the exception of the finance and insurance sector. In two sectors (mining, and
accommodation, cafes and restaurants), this negative DS was large enough to outweigh the
positive PS so that a negative total shift was recorded for these sectors.
All ‘slow’ growth sectors also expanded less rapidly in Victoria than in the nation as a whole
(DS<0), with the exception of the agriculture, forestry and fishing; transport and storage;
and cultural and recreational services sectors. Of the exceptions, only the cultural and
recreational services sector had its positive DS large enough to more than offset the negative
PS to produce a positive total shift into this sector for Victoria.
Queensland
All ‘high’ growth sectors expanded more rapidly in Queensland than in the nation as a
whole (DS>0), with the exception of the mining, and property and business services sectors.
For the exceptions, their negative DS was too small to outweigh the positive PS so that a
positive total shift was recorded for these sectors in Queensland.
All ‘slow’ growth sectors also expanded more rapidly in Queensland than in the nation as a
whole (DS>0), with the exception of the agriculture, forestry and fishing; and construction
sectors.17 With the exception of the manufacturing, and electricity, gas and water sectors,18
the remaining ‘slow’ growth sectors each recorded a positive DS large enough to more than
offset their negative PS to produce a positive total shift in Queensland.
South Australia
All ‘high’ growth sectors expanded less rapidly in South Australia than in the nation as a
whole (DS<0). In two sectors (mining, and property and business services),19 this negative
DS was large enough to outweigh the positive PS so that a negative total shift was recorded
for these sectors.
17 In Table 4.3, the only exception here is the agriculture, forestry and fishing sector.
18 In Table 4.3, the construction sector is added to the list of exceptions here.
19 In Table 4.3, the accommodation, cafes and restaurants sector is added to this list of sectors recording a negative total shift in
South Australia.
– 81 –
Productivity and Regional Economic Performance in Australia
All ‘slow’ growth sectors also expanded less rapidly in South Australia than in the nation as
a whole (DS<0), with the exception of the agriculture, forestry and fishing; manufacturing;
and electricity, gas and water sectors. Of the exceptions, only the agriculture, forestry and
fishing; and electricity, gas and water sectors had their positive DS large enough to more than
offset the negative PS to produce a positive total shift for these sectors in South Australia.
Western Australia
All ‘high’ growth sectors expanded more rapidly in Western Australia than in the nation as
a whole (DS>0), with the exception of the accommodation, cafes and restaurants; and
property and business services sectors.20 For the former sector, the negative DS was large
enough to outweigh the positive PS so that a negative total shift was recorded for this sector
in Western Australia.
All ‘slow’ growth sectors also expanded more rapidly in Western Australia than in the nation
as a whole (DS>0), with the exception of the agriculture, forestry and fishing; government
administration and defence; education; and cultural and recreational services sectors. Indeed
the electricity, gas and water; construction; transport and storage; and health and community
services sectors21 each recorded a positive DS large enough to more than offset their
negative PS to produce a positive total shift for these sectors in Western Australia.
Tasmania
All ‘high’ growth sectors expanded less rapidly in Tasmania than in the nation as a whole
(DS<0), with the exception of the finance and insurance sector.22 In two sectors,
accommodation, cafes and restaurants; and property and business services, this negative DS
was large enough to outweigh the positive PS so that a negative total shift was recorded for
these sectors in Tasmania.
All ‘slow’ growth sectors also expanded less rapidly in Tasmania than in the nation as a
whole (DS<0), with the exception of the agriculture, forestry and fishing; electricity, gas
and water; and government administration and defence sectors. Of the exceptions, only the
agriculture, forestry and fishing sector had its positive DS large enough to more than offset
the negative PS to produce a positive total shift for this sector in Tasmania.
Northern Territory
All ‘high’ growth sectors expanded more rapidly in the Northern Territory than in the nation
as a whole (DS>0), with the exception of the mining, and finance and insurance sectors.23
Even for the exceptions, the negative DS was not large enough to outweigh the positive PS,
so that positive total shifts were recorded for all ‘high’ growth sectors in the Northern
Territory.
20 In Table 4.3, the only exception here is the accommodation, cafes and restaurants sector.
21 In Table 4.3, the personal and other services sector is added and the transport and storage sector is deleted from this list.
22 In Table 4.3, the mining sector is added to this list.
23 In Table 4.3, the only exception is the finance and insurance sector.
– 82 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
All ‘slow’ growth sectors also expanded more rapidly in the Northern Territory than in the
nation as a whole (DS>0), with the exception of the construction, wholesale trade, retail
trade, and personal and other services sectors.24 Indeed, the electricity, gas and water;
transport and storage; education; health and community services; and cultural and
recreational services sectors each recorded a positive DS large enough to more than offset
their negative PS to produce a positive total shift for these sectors in the Northern Territory.
Australian Capital Territory
All ‘high’ growth sectors expanded less rapidly in the Australian Capital Territory than in
the nation as a whole (DS<0), with the exception of the communication services sector. In
the mining sector, this negative DS was large enough to outweigh the positive PS, so that a
negative total shift was recorded for this sector in the Australian Capital Territory.
All ‘slow’ growth sectors also expanded less rapidly in the Australian Capital Territory than
in the nation as a whole (DS<0), with the exception of the electricity, gas and water;
government administration and defence; health and community services; and personal and
other services sectors.25 Of the exceptions, only the government administration and defence
sector had its positive DS insufficiently large enough to more than offset the negative PS and
thus fail to produce a positive total shift for this sector in the Australian Capital Territory.
Results of Application of Simple Productivity-Extended
Shift-Share Analysis
Aggregate level results
The aggregate results from the application of our adaptation of the simple productivityextended shift-share analysis are given in Table 4.5. The first point worthy of note with
these results is that they are totally consistent with the results derived from application of
the dynamic shift-share results reported in the second section of this chapter. The total
national growth, total shift, total proportionality shift and total differential shift values are
numerically identical for any given region in Tables 4.2 and 4.5.
As discussed in the first section of this chapter, the contribution made by the productivityextended shift-share approach is the disaggregation of these various ‘total’ shift-share
components into productivity-constant and employment-constant related sub-components.
With productivity held constant, the results show the output effects of variations in
employment levels, while with employment held constant the results show the impact on
output levels of productivity changes.
In discussing these disaggregated results, we make use of the system developed by Rigby and
Anderson (1993) for classifying regions according to the relative contribution to total shift of
its two sub-components. A schematic representation of this classification system is given in
Figure 4.2.
24 In Table 4.3, the agriculture, forestry and fishing; and government administration sectors are added, while the construction
sector is deleted from this list.
25 In Table 4.3, the construction sector is added to this list.
– 83 –
Productivity and Regional Economic Performance in Australia
Table 4.5: Aggregate Results from Application
of Simple Productivity-Extended Shift-Share Analysis
Gross product at factor cost ($m, 1997-98 prices),1985-86 to 1998-99
Sector
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Absolute total output change
Productivity constant
28,358.125 16,647.938 21,735.872
2,757.796
12,660.314
546.971
Employment constant
30,645.179 21,463.188 12,617.624
5,790.563
11,773.296
1,505.945
2,309.616 2,559.889
Total
59,003.304 38,111.126 34,353.496
8,548.359
24,433.610
2,052.916
Productivity constant
28,989.715 22,366.547 12,985.918
6,280.451
8,747.975
1,819.607
957.966 1,781.609
Employment constant
30,499.647 23,075.959 13,843.588
6,487.477
9,458.691
1,880.348
1,042.697 1,868.903
Total
59,489.362 45,442.506 26,829.506 12,767.928
18,206.665
3,699.954
2,000.663 3,650.512
-222.349
937.128
2,087.267 3,497.017
Absolute national growth effect
Absolute proportionality shift (PS)
Productivity constant
589.330
-1,666.119
-1,213.789
-634.945
-1,714.452
-406.633
-121.687
311.631
Employment constant
943.065
773.413
952.069
-73.946
2,542.701
58.889
279.731
-619.259
1,532.396
-892.706
-261.720
-708.891
828.249
-347.744
158.045
-307.628
-1,220.920
-4,052.489
9,963.743
-2,887.710
5,626.791
-866.002
1,473.337
466.649
-797.533
-2,386.185
-2,178.033
-622.968
-228.095
-433.292
-1,544.778
-312.516
-2,018.453
-6,438.674
7,785.710
-3,510.678
5,398.696
-1,299.294
-71.441
154.133
Productivity constant
-631.590
-5,718.609
8,749.954
-3,522.655
3,912.339
-1,272.636
1,351.650
778.280
Employment constant
145.532
-1,612.771
-1,225.964
-696.914
2,314.606
-374.402
-1,265.047
-931.775
Total
-486.058
-7,331.380
7,523.990
-4,219.569
6,226.945
-1,647.038
86.604
-153.495
Type 4a
Type 3a
Type 2a
Type 3a
Type 1a
Type 3a
Type 2a
Type 2b
Total
Absolute differential shift (DS)
Productivity constant
Employment constant
Total
Absolute total shift = PS + DS
Rigby & Anderson classification
Figure 4.2: Classification of Regions according to Total Shift
Total Shift TS(b)
(employment-constant)
Positive
Total Shift TS(a)
(productivity constant)
Positive
Negative
1b
2b
1a
2a
3b
4b
Negative
4a
3a
Source: Adapted from Rigby and Anderson (1993).
Western Australia is a type 1 region: it has a positive total shift for both components. That is,
it is experiencing above average GSP growth both due to the impact of above average
employment growth (TSr(a)>0) and to the impact of above average productivity growth
(TSr(b)>0). The former effects dominate the latter, to produce a type 1a classification overall.
Queensland, the Northern Territory and the Australian Capital Territory are type 2 regions:
they have above average employment growth effects (TSr(a)>0) but below average
productivity growth effects (TSr(b)<0). For Queensland and the Northern Territory, the
employment growth effects dominate the productivity growth effects, so overall the total shift
– 84 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
is positive, and accordingly these are classified as type 2a regions. For the Australian Capital
Territory, the negative productivity effects dominate to produce a negative total shift and a
type 2b regional typology.
Victoria, South Australia and Tasmania are type 3a regions: they have negative total shifts
for both components. That is, they are experiencing below average GSP growth both due to
the impact of below average employment growth (TSr(a)<0) and to the impact of below
average productivity growth (TSr(b)<0), with the former exceeding the latter in magnitude.
New South Wales is a type 4 region: it has below average employment growth effects
(TSr(a)<0) but above average productivity growth effects (TSr(b)>0). The former effects
dominate the latter, to produce an overall negative total shift and a type 4a classification
overall.
From the above observations it can be seen that, even at the aggregate level, some additional
insights can be made available through the use of the simple version of productivityextended shift-share analysis. In particular, it is interesting to contrast the experiences of
New South Wales and Victoria. While both have experienced negative employment growth
effects, the former more than compensated for this by achieving a strong positive
productivity growth effect. Similarly, it is noteworthy that while both Queensland and
Western Australia experienced positive shifts induced by employment growth, the latter also
benefited from a positive shift induced by productivity change, while Queensland had a
negative productivity-induced shift (which, as pointed out, was more than offset by the
positive employment growth-induced shift).
Examination of the relative contribution of proportionality versus differential shifts to any
given region’s ‘employment growth’ and ‘productivity growth’ effects gives an indication
of the impact of industry mix on these effects.
The results in Table 4.5 indicate, for example, that on the basis of industry-mix (i.e.
proportionality shift) alone, Queensland, Western Australia and the Northern Territory
should have experienced negative rather than positive employment growth related total
shifts. The latter were produced therefore by the large positive employment growth related
differential shifts.
The results in Table 4.5 also indicate that on the basis of industry-mix (i.e. proportionality
shift) alone, South Australia and the Australian Capital Territory were the only regions that
should have experienced negative productivity growth related total shifts. However, other
regions to produce such negative total shifts were Victoria, Queensland, Tasmania and the
Northern Territory.
Disaggregated industry level results
Further insights can be obtained from a more detailed analysis focusing on particular
industries and regions. In particular, such an analysis could assist us in gaining an
understanding of factors underlying interstate differences in productivity change. The
results disaggregated by industry are given for the various shifts in absolute terms (i.e.
millions of dollars, 1997-98 prices) in Table 4.6, and in relative terms (i.e. as percentages
of the average levels of real GSP calculated over the entire study period) in Table 4.7.
– 85 –
Productivity and Regional Economic Performance in Australia
Table 4.6: Disaggregated Absolute Results from Application of
Simple Productivity-Extended Shift-Share Analysis
(a) Absolute total shift ‘a’
Sector
Agriculture, forestry and fishing
NSW
Vic
Qld
SA
-623.424 -1,742.786
-421.683
-311.970
Mining
-1,736.692 -1,628.235
-200.861
-337.635
Manufacturing
-7,100.386 -5,434.424
593.021 -1,743.180
Electricity, gas and water
-3,693.941 -2,842.292
Construction
1,414.499
WA
NT
ACT
-107.949
218.461
4.232
-3,584.454
483.462
57.512
-42.773
6.749
-3,398.472
-176.488
-372.530
95.405
-599.334
Tas
Aust
-114.110 -14,252.692
-774.776
-671.940
-785.718
-486.376
198.544
29.020
-9,027.480
775.146 1,278.588
-340.889
449.469
-188.838
15.156
-339.021
3,064.110
Wholesale trade
-434.912
-228.651
515.029
-392.888
286.085
-143.044
-21.653
-33.849
-453.884
Retail trade
-719.925
-437.379
585.317
-182.696
176.102
-73.847
83.316
86.101
-483.011
Accommodation, cafes
and restaurants
1,143.517
611.605
789.263
299.732
214.685
33.646
109.692
93.918
3,296.059
Transport and storage
-211.049
-638.915
141.835
-400.497
131.181
-27.064
144.679
134.305
-725.526
Communication services
-111.406
-48.399
-19.408
-5.873
53.308
-34.693
-44.074
141.001
-69.545
-685.509
-247.410
-531.467
-509.025
-161.567
-112.485
-56.637
-154.827
-2,458.927
8,196.731 6,809.337 3,143.203 1,121.225 2,344.697
123.773
278.201
801.868 22,819.036
Finance and insurance
(incl. nominal industry)
Property and business services
Government administration
and defence
Education
Health and community services
Cultural and recreational services
-620.267 -1,777.804
2,834.894
953.201
331.503
-160.510
367.730
-62.630
-86.337
-122.857
932.957
-30.197
320.727
49.214
133.783
48.334
3,579.807
335.379 1,129.619
-39.728
306.263
121.190
156.018
-9.885
2,952.057
-709.906
84.685 1,088.551
-2,131.171
725.581
6.401
150.897
-31.294
16.820
127.007
2,168.649
532.234
177.016
350.840
-17.222
153.049
80.293
2,352.180
-631.590 -5,718.609 8,749.954 -3,522.655 3,912.339 -1,272.636 1,351.650
778.280
3,646.735
Personal and other services
678.396
Total
397.573
(b) Absolute total shift ‘b’
Sector
Agriculture, forestry and fishing
Mining
Manufacturing
NSW
Vic
128.674 1,584.071
2,099.596
SA
WA
497.461
-33.353
124.342
-234.906
-13.003
2,069.995
-130.068 4,294.771
-25.343
153.635
-13.878
7,587.688
16.709
583.134
NT
ACT
Aust
-226.802
-179.605
-106.443
4.553
-772.649
629.882
848.038
961.714
447.234
-197.253
103.845
6,975.524
Construction
-610.047 -1,744.890 -1,372.270
-249.825
346.935
-85.514
-132.411
317.049
-3,530.974
Wholesale trade
-595.896
-648.067
199.866
-107.927
-416.684
-6.235
1.487
-4.753
-1,578.209
Retail trade
-343.362
-946.556
-187.287
-431.595
-345.023
-79.959
-186.560
-139.522
-2,659.865
Accommodation, cafes
and restaurants
-1,057.472
-814.008
-128.149
-295.449
-269.612
-44.551
-97.969
-83.804
-2,791.013
Transport and storage
-853.583
460.848
2.120
217.791
-100.066
-114.205
-116.219
-162.362
-665.676
2,523.318 1,929.528 1,331.131
316.855
736.560
134.291
158.631
24.032
7,154.346
(incl. nominal industry)
5,195.996 4,478.072 2,492.147
991.800 1,061.217
338.093
66.898
255.583 14,879.806
Property and business services
-3,059.191 -4,726.624 -2,051.288 -1,179.713 -1,523.199
-364.028
-151.184
-651.821 -13,707.047
Communication services
-640.944 -1,143.725
Tas
542.613
Electricity, gas and water
977.704
625.839
Qld
2,612.900 1,569.162
Finance and insurance
Government administration
and defence
-331.790
-79.306
14.443
-304.382
-749.638
23.973
10.932
-164.453
Education
-3,396.248
-361.280
-698.141
-476.395
-598.953
-226.389
-105.692
-222.932
-1,580.222
-6,086.031
Health and community services
-1,822.264
-752.259
-290.157
-401.078
-276.574
-180.238
-110.393
61.241
-3,771.721
Cultural and recreational services
-347.282
-779.749
-518.055
-160.691
-193.640
-89.822
10.621
-214.660
-2,293.278
Personal and other services
-975.520
-766.608
-106.324
-374.348
-353.048
-46.448
-228.221
-26.893
-2,877.408
Total
145.532 -1,612.771 -1,225.964
-374.402 -1,265.047
-931.775
-3,646.735
-696.914 2,314.606
– 86 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
(c) Absolute total shift ‘a + b’
Sector
NSW
Agriculture, forestry and fishing
Mining
-494.750
Vic
-158.716
362.904 -1,002.397
Qld
SA
WA
Tas
NT
ACT
-404.974
185.492
-632.687
16.394
-16.446
-8.771
-1,514.459
382.273
-467.702 4,778.234
32.170
110.863
-7.128
4,189.215
-552.135
-11.038
Manufacturing
-6,122.682 -6,075.368
-550.703 -1,200.568
Electricity, gas and water
-1,081.041 -1,273.130
-403.290
Aust
-109.557 -15,025.341
-144.894
176.098
175.996
-39.142
1.291
132.865
804.451
-969.743
-93.682
-590.714
796.404
-274.352
-117.255
-21.973
-466.864
Wholesale trade
-1,030.809
-876.717
714.894
-500.815
-130.599
-149.279
-20.166
-38.602
-2,032.093
Retail trade
-1,063.288 -1,383.935
398.030
-614.292
-168.921
-153.806
-103.245
-53.421
-3,142.876
Construction
-2,051.956
Accommodation, cafes
and restaurants
86.045
-202.403
661.114
4.284
-54.927
-10.905
11.723
10.114
505.045
Transport and storage
-1,064.631
-178.066
143.955
-182.707
31.114
-141.270
28.459
-28.057
-1,391.202
Communication services
2,411.911 1,881.130 1,311.722
310.981
789.868
99.599
114.557
165.033
7,084.801
(incl. nominal industry)
4,510.487 4,230.662 1,960.680
482.774
899.650
225.607
10.262
Property and business services
5,137.540 2,082.713 1,091.915
-58.488
821.498
-240.254
127.017
150.048
9,111.989
Finance and insurance
100.756 12,420.878
Government administration
and defence
-952.057 -1,857.110
345.946
-464.892
-381.908
-38.657
-75.405
-287.309
-3,711.393
Education
-561.354 -1,071.186
234.816
-506.592
-278.226
-177.174
28.092
-174.598
-2,506.224
Health and community services
-869.063
-416.880
839.463
-440.807
29.689
-59.047
45.625
51.356
-819.664
Cultural and recreational services
-262.598
308.802
207.527
-154.290
-42.743
-121.116
27.441
-87.653
-124.629
Personal and other services
-297.124
-369.035
425.909
-197.332
-2.207
-63.669
-75.172
53.401
-525.229
Total
-486.058 -7,331.380 7,523.990 -4,219.569 6,226.945 -1,647.038
86.604
-153.495
0.000
Table 4.7: Disaggregated Relative Results from Application of
Simple Productivity-Extended Shift-Share Analysis
(a) Relative total shift ‘a’
Sector
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
-1.743
Agriculture, forestry and fishing
-1.120
-3.705
-0.976
-1.478
-2.055
-1.565
9.016
2.162
Mining
-4.714
-3.768
-0.401
-2.886
0.515
1.989
-0.369
8.856
-1.357
Manufacturing
-2.251
-1.890
0.533
-2.215
-0.272
-1.826
2.952
-3.851
-1.612
Electricity, gas and water
-6.564
-6.321
-2.928
-5.578
-4.716
-8.203
15.113
1.458
-5.450
Construction
1.158
0.959
1.948
-1.398
1.037
-2.512
0.321
-4.661
0.861
Wholesale trade
-0.366
-0.255
1.081
-1.976
0.960
-2.779
-1.041
-1.220
-0.144
Retail trade
-0.631
-0.522
0.940
-0.702
0.538
-0.846
2.066
1.487
-0.143
Accommodation, cafes and restaurants
2.046
2.311
2.764
3.243
1.938
1.011
4.919
3.509
2.363
Transport and storage
-0.189
-0.896
0.249
-1.774
0.431
-0.475
4.385
3.453
-0.237
Communication services
-0.204
-0.111
-0.081
-0.060
0.382
-1.319
-2.744
5.141
-0.045
(incl. nominal industry)
-0.499
-0.286
-1.411
-2.640
-0.745
-2.325
-3.205
-3.841
-0.786
Property and business services
3.596
4.448
4.160
3.361
4.485
1.812
5.792
5.825
4.020
Government administration and defence -0.864
-3.170
0.890
-1.074
1.989
-0.973
-1.518
-0.326
-0.858
Education
3.043
-0.886
2.071
-0.126
1.253
0.689
3.701
0.566
1.246
Health and community services
0.808
0.338
2.072
-0.119
0.849
1.210
3.998
-0.161
0.817
Finance and insurance
Cultural and recreational services
0.204
3.790
4.764
0.088
1.593
-1.347
0.944
3.243
1.966
Personal and other services
1.433
1.156
2.515
1.357
2.383
-0.552
7.378
1.945
1.680
Total
-0.036
-0.422
1.090
-0.925
0.719
-1.158
2.247
0.717
0.071
– 87 –
Productivity and Regional Economic Performance in Australia
(b) Relative total shift ‘b’
Sector
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
1.006
Agriculture, forestry and fishing
0.231
3.367
0.039
2.357
-0.114
1.803
-9.694
-6.642
Mining
5.699
1.448
1.163
-1.112
4.572
-0.876
1.327
-18.210
3.031
Manufacturing
0.310
-0.223
-1.029
0.690
-0.350
-0.880
-3.293
0.154
-0.087
Electricity, gas and water
4.643
3.490
2.380
7.040
5.772
7.543
-15.015
5.218
4.211
Construction
-0.500
-2.158
-2.091
-1.024
0.801
-1.137
-2.802
4.359
-0.992
Wholesale trade
-0.502
-0.724
0.419
-0.543
-1.398
-0.121
0.072
-0.171
-0.500
Retail trade
-0.301
-1.130
-0.301
-1.658
-1.054
-0.916
-4.627
-2.410
-0.788
Accommodation, cafes and restaurants
-1.892
-3.075
-0.449
-3.196
-2.434
-1.339
-4.394
-3.131
-2.001
Transport and storage
-0.765
0.646
0.004
0.965
-0.329
-2.006
-3.522
-4.174
-0.218
Communication services
4.617
4.413
5.548
3.244
5.281
5.106
9.878
0.876
4.674
Finance and insurance
(incl. nominal industry)
3.785
5.177
6.617
5.144
4.896
6.988
3.785
6.341
4.754
Property and business services
-1.342
-3.088
-2.715
-3.536
-2.914
-5.329
-3.148
-4.735
-2.415
-0.636
Government administration and defence
-0.462
-0.141
0.039
-2.037
-4.055
0.373
0.192
-0.437
Education
-3.646
-0.451
-1.550
-1.988
-2.340
-3.167
-2.923
-2.609
-2.119
Health and community services
-1.544
-0.757
-0.532
-1.200
-0.767
-1.799
-2.829
0.996
-1.044
Cultural and recreational services
-0.835
-2.715
-3.401
-2.207
-2.044
-3.866
0.596
-5.481
-2.079
Personal and other services
-2.060
-2.228
-0.503
-2.869
-2.398
-1.489
-11.001
-0.651
-2.055
Total
0.008
-0.119
-0.153
-0.183
0.425
-0.341
-2.103
-0.858
-0.071
(c) Relative total shift ‘a + b’
Sector
Agriculture, forestry and fishing
NSW
-0.889
Vic
-0.337
Qld
SA
WA
Tas
NT
ACT
Aust
-0.937
0.879
-2.169
0.238
-0.679
-4.480
-0.736
Mining
0.985
-2.320
0.763
-3.998
5.087
1.112
0.957
-9.353
1.673
Manufacturing
-1.941
-2.113
-0.495
-1.526
-0.622
-2.706
-0.342
-3.697
-1.699
Electricity, gas and water
-1.921
-2.831
-0.548
1.462
1.056
-0.660
0.098
6.676
-1.239
Construction
0.659
-1.200
-0.143
-2.422
1.838
-3.649
-2.481
-0.302
-0.131
Wholesale trade
-0.868
-0.979
1.500
-2.518
-0.438
-2.900
-0.970
-1.392
-0.644
Retail trade
-0.932
-1.652
0.639
-2.359
-0.516
-1.762
-2.561
-0.923
-0.931
Accommodation, cafes and restaurants
0.154
-0.765
2.315
0.046
-0.496
-0.328
0.526
0.378
0.362
Transport and storage
-0.954
-0.250
0.253
-0.809
0.102
-2.481
0.863
-0.721
-0.455
Communication services
4.413
4.302
5.467
3.183
5.663
3.787
7.133
6.017
4.629
Finance and insurance
(incl. nominal industry)
3.286
4.891
5.206
2.504
4.150
4.663
0.581
2.500
3.968
Property and business services
2.254
1.361
1.445
-0.175
1.571
-3.517
2.645
1.090
1.605
-1.494
Government administration and defence
-1.326
-3.311
0.929
-3.111
-2.066
-0.601
-1.326
-0.763
Education
-0.603
-1.337
0.521
-2.114
-1.087
-2.479
0.777
-2.043
-0.873
Health and community services
-0.737
-0.420
1.540
-1.319
0.082
-0.589
1.169
0.835
-0.227
Cultural and recreational services
-0.631
1.075
1.363
-2.119
-0.451
-5.213
1.541
-2.238
-0.113
Personal and other services
-0.628
-1.073
2.013
-1.512
-0.015
-2.042
-3.624
1.294
-0.375
Total
-0.027
-0.540
0.938
-1.108
1.144
-1.499
0.144
-0.141
0.000
– 88 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
Agriculture, forestry and fishing
All regions except the Northern Territory and the Australian Capital Territory recorded below
average employment growth effects in this sector. In absolute terms, the largest below
average employment growth effects occurred in Victoria, New South Wales and Western
Australia (see row 1 of Table 4.6a). In relative terms, the largest below average employment
growth effects were in Victoria, Western Australia and Tasmania (see row 1 of Table 4.7a).
All regions except the Northern Territory, the Australian Capital Territory and Western
Australia experienced above average productivity growth effects in this industry. In absolute
terms, the largest above average productivity growth effects were in Victoria, South Australia
and New South Wales (see row 1 of Table 4.6b). In relative terms, the largest above average
productivity shifts were in Victoria, South Australia and Tasmania (see row 1 of Table 4.7b).
Mining
All regions except Western Australia, Tasmania and the Australian Capital Territory
recorded below average employment growth effects in this sector (see row 2 of Table 4.6).
Most regions26 experienced positive shifts in GSP due to productivity growth in this sector
(see row 2 of Table 4.6).
Overall, Western Australia recorded large above average productivity growth effects that
combined with above average employment growth effects to produce a significant positive
total shift of economic activity into the mining sector. While a number of regions also recorded
positive total shifts into this sector, this was a result of productivity growth rather than
employment growth effects in Queensland, New South Wales and the Northern Territory.27
Manufacturing
Most regions recorded below average employment growth effects in the manufacturing
sector, with the most notable exceptions being Queensland and the Northern Territory. In
absolute terms, the largest negative shifts in GSP due to employment growth were in
Victoria, New South Wales and South Australia (see row 3 of Table 4.6a). This seems to
reflect the effects of microeconomic (especially tariff) reforms. In relative terms, the largest
negative shifts in GSP due to employment growth were in the Australian Capital Territory,
New South Wales and South Australia (see row 3 of Table 4.7a).
Most regions experienced below average productivity growth effects in the manufacturing
sector, with one important exception being New South Wales. In absolute terms, the largest
negative shifts in GSP due to productivity growth were in Queensland, Victoria and Western
Australia (see row 3 of Table 4.6b). In relative terms, the largest negative shifts in GSP due
to productivity growth were in the Northern Territory, Queensland and Tasmania (see row
3 of Table 4.7b).
26 The exceptions here are South Australia, Tasmania and the Australian Capital Territory.
27 While Tasmania also recorded a positive total shift into mining, this was due to above average employment growth since its
productivity growth was negative in this sector.
– 89 –
Productivity and Regional Economic Performance in Australia
Queensland and the Northern Territory are worthy of special note since although they were
significantly above the national manufacturing average in terms of employment growth
effects, they were concomitantly significantly below the national manufacturing average in
terms of productivity growth effects. The net effect in both regions was a rate of growth in
GSP that was higher than the national average for this sector.
Electricity, gas and water
Most regions recorded below average employment growth effects in this sector,28 perhaps
reflecting microeconomic reforms and technological changes (see row 4 of Table 4.6). In
relative terms, the largest negative shifts in GSP due to employment growth were in
Tasmania, New South Wales and Victoria (see row 4 of Table 4.7a).
All regions except the Northern Territory experienced positive shifts in GSP due to
productivity growth effects (see row 4 of Table 4.6). In relative terms, the largest positive
shifts in GSP due to productivity growth were in Tasmania, South Australia and Western
Australia (see row 4 of Table 4.7b).
In the cases of New South Wales, Victoria and Queensland, these positive productivity
growth effects were insufficiently large to offset the negative employment growth effects
and so negative total shifts were recorded for this sector. By contrast, the netting out of these
two contrasting effects produced positive total shifts for this sector in Western Australia,
South Australia and Tasmania. For Western Australia and South Australia, this is primarily
due to the gas rather than the electricity subsector, while for Tasmania the dominance of
hydropower in the electricity subsector is the contributing factor.
Communication services
Most regions, with the notable exceptions of Western Australia and the Australian Capital
Territory, recorded below average employment growth effects (see row 10 of Table 4.6),
again reflecting the impact of microeconomic reforms and technological change. In relative
terms, the largest negative shifts in GSP due to employment growth were in Northern
Territory, Tasmania, New South Wales and Victoria (see row 10 of Table 4.7a).
All regions recorded positive shifts in GSP due to productivity growth effects (see row 10
of Table 4.6). In relative terms, the largest positive shifts in GSP due to productivity growth
were in the Northern Territory, Queensland, Western Australia and Tasmania (see row 10 of
Table 4.7b).
In all regions with negative employment growth effects, positive productivity growth
effects dominated, such that all regions recorded positive total shifts of economic activity
into this sector (see row 10 of Table 4.6c).
Finance and insurance
Microeconomic reforms and technological change resulted in all regions recording below
average employment growth effects in the finance and insurance sector (see row 11 of Table
4.6). In relative terms, the largest negative shifts in GSP due to employment growth were in
28 The exceptions here are the Northern Territory and the Australian Capital Territory.
– 90 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
the Australian Capital Territory, the Northern Territory, South Australia and Tasmania (see
row 11 of Table 4.7a). This reflects a growing contraction of employment in this sector into
the ‘headquarters’ states of Victoria and New South Wales.
All regions recorded positive shifts in GSP due to productivity growth effects (see row 11
of Table 4.6). In relative terms, the largest positive shifts in GSP due to productivity growth
were in the ‘branch-plant’ regions of Tasmania, the Australian Capital Territory and
Queensland (see row 11 of Table 4.7b).
Queensland and Tasmania were above average in terms of both the negative employment
growth related shift and the positive productivity growth related shift characteristic of this
sector. The latter shift dominated in both cases, with the net effect being that each of these
states recorded a total shift into the finance and insurance sector significantly higher than
the national average in relative terms (see row 11 of Table 4.7c).
Broader conclusions
Sectors registering significant positive productivity growth related shifts in GSP are mining;
electricity, gas and water; communication services; finance and insurance; and manufacturing
(see Tables 4.6b and 4.7b). Sectors registering significant positive employment growth related
shifts in GSP are property and business services, health and community services, cultural and
recreational services, and education (see Tables 4.6a and 4.7a).
In general, there is a negative correlation between employment growth and productivity
growth related shifts. As technological changes and/or microeconomic reforms have been
introduced over the period covered by this study, there has been concomitant labour
shedding and higher labour productivity in many sectors. A notable exception is the special
case of Western Australia and the mining sector for which the results presented here are
similar to those found by Nguyen, Smith and Meyer-Boehm (2000).
Tables 4.6b and 4.7b show clearly that most service sectors have registered negative
productivity growth related shifts in GSP. Given the difficulties associated with
measurement of output in service sectors, this suggests that results need to be treated with
caution, especially in those regions such as Queensland and the Australian Capital Territory
where service sectors comprise a large proportion of economic activity.
Results of Application of Multifactor Productivity-Extended
Shift-Share Analysis
The aggregate results from the application of the multifactor productivity-extended shiftshare analysis are given in Table 4.8.
As discussed in the first section of this chapter, the main difference between the simple and
multifactor productivity-extended shift-share approaches lies in the allocation of total
output change (identical in Tables 4.5 and 4.8) to the various components. In Table 4.5, total
output change is fully allocated to labour-related national growth, proportionality and
differential shifts. In Table 4.8, however, the output change attributed to the contribution of
labour is significantly reduced since part of this change is more correctly attributed to the
– 91 –
Productivity and Regional Economic Performance in Australia
other non-labour factors of production. Indeed, the contributions of the latter factors are
estimated to account for 54% of the nation’s total output change. For some states this
contribution is much higher (79% for Tasmania, 72% for South Australia, 62% for Western
Australia and 56% for New South Wales), while for other regions it is lower (53% for
Victoria, 49% for the Northern Territory, 45% for Queensland and 35% for the Australia
Capital Territory). For the former group of states, economic growth appears to have been
significantly influenced by improvements in capital stocks, technology, infrastructure and
other non-labour factors. On the other hand, for the latter group of regions non-labour
factors are less significant in driving output growth.
Table 4.8 also records the results from the application of the Rigby–Anderson regional
typology (given in Figure 4.2) to the labour-only total shift components.
Table 4.8: Aggregate Results from Application of the Multifactor
Productivity-Extended Shift-Share Analysis
Gross product at factor cost ($m, 1997-98 prices), 1985-86 to 1998-99
NSW
Vic
Qld
SA
WA
Productivity constant
28,358.125
16,647.938
21,735.872
2,757.796
12,660.314
Employment constant
30,645.179
21,463.188
12,617.624
5,790.563
11,773.296
Total
59,003.304
38,111.126
34,353.496
8,548.359
24,433.610
Productivity constant
18,000.148
13,892.862
8,048.940
3,902.322
Employment constant
15,105.436
11,348.431
7,053.329
3,192.851
Total
33,105.584
25,241.293
15,102.268
7,095.174
Tas
NT
ACT
546.971
2,309.616
2,559.889
1,505.945
-222.349
937.128
2,052.916
2,087.267
3,497.017
5,422.380
1,131.027
595.082
1,105.479
4,837.550
912.170
514.346
933.498
10,259.929
2,043.197
1,109.428
2,038.976
Absolute total output change
National growth (NG) effect (labour)
Proportionality shift (PS) (labour)
Productivity constant
Employment constant
Total
4,142.124
1,857.730
662.867
435.986
-2.782
14.419
70.215
442.280
-10,548.455
-7,365.893
-2,794.195
-1,907.540
-1,800.384
-473.861
-198.844
-339.154
-6,406.330
-5,508.163
-2,131.328
-1,471.554
-1,803.166
-459.442
-128.628
103.126
Differential shift (DS) (labour)
Productivity constant
-1.449
-3,442.030
6,027.741
-1,846.086
2,628.765
-579.088
763.267
536.818
Employment constant
-722.495
1,659.583
-163.996
-1,389.382
-1,857.531
-576.755
-684.660
-412.628
Total
-723.945
-1,782.447
5,863.745
-3,235.468
771.234
-1,155.843
78.606
124.189
Total shift (labour) = PS + DS
Productivity constant
Employment constant
Total
4,140.675
-1,584.301
6,690.609
-1,410.100
2,625.984
-564.669
833.482
979.098
-11,270.950
-5,706.310
-2,958.191
-3,296.922
-3,657.915
-1,050.616
-883.504
-751.783
-7,130.275
-7,290.611
3,732.418
-4,707.022
-1,031.932
-1,615.285
-50.022
227.316
2,084.577
Total output change (labour) = PS+DS+NG
Productivity constant
22,140.823
12,308.561
14,739.549
2,492.222
8,048.363
566.357
1,428.564
Employment constant
3,834.486
5,642.121
4,095.137
-104.071
1,179.634
-138.446
-369.158
181.715
25,975.309
17,950.682
18,834.686
2,388.152
9,227.997
427.911
1,059.406
2,266.292
Type 2b
Type 3b
Type 2a
Type 3b
Type 2b
Type 3b
Type 2b
Type 2a
475.312
Total
Regional typology (labour)
Other factors contribution
Productivity constant
6,217.302
4,339.376
6,996.323
265.574
4,611.951
-19.386
881.052
Employment constant
26,810.694
15,821.067
8,522.487
5,894.634
10,593.662
1,644.391
146.809
755.413
Total
33,027.996
20,160.443
15,518.810
6,160.207
15,205.613
1,625.005
1,027.861
1,230.725
– 92 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
For most regions, compared with Table 4.5, the regional typology does not change at all
(Queensland) or only in a very insignificant way (from type 2a to 2b for the Northern
Territory, from type 2b to 2a for the Australian Capital Territory and from type 3a to 3b for
Victoria, South Australia and Tasmania).29
Western Australia, however, changes from a type 1a to a type 2b region. That is, it changes
from having a positive productivity-related output change in total to having a negative
productivity-related output change attributable to labour alone. This suggests that most of
its quite significant productivity-related output change has been based on the contribution
of non-labour factors. This demonstrates quite clearly the value of the multifactor extension
introduced by Haynes and Dinc (1997) as adapted for use in this chapter.
The remaining state, New South Wales, also changes from having a positive productivityrelated output change in total to having a negative productivity-related output change
attributable to labour alone. Once again this points to the important role being played by nonlabour factors in generating this state’s overall positive productivity-related output change.
In addition, however, New South Wales also records a change in its employment growth
related output change component from negative in total to positive based on the contribution
of labour alone. It appears then that much of the improvements in capital stocks, technology
and other labour factors in this state have in turn facilitated labour shedding.
Conclusion
This chapter demonstrates the usefulness of the shift-share method as a means of analysing
interstate and intertemporal variations in real GSP growth in the Australian context.
Over the period covered by the study, Western Australia, New South Wales and the Northern
Territory displayed industrial structures conducive to GSP growing at a rate faster than the
national average. Only Western Australia, however, grew even faster than its industrial mix
would have suggested. By contrast, Victoria, Queensland, South Australia, Tasmania and the
Australian Capital Territory had industrial structures conducive to GSP growing at a rate
slower than the national average. Victoria, South Australia and Tasmania, however, grew at a
rate even slower than would have been expected on the basis of their industrial mix, while
Queensland was able to overcome its negative proportionality shift and to grow faster than the
overall national rate of growth.
Over the period, Western Australia was shown to be experiencing above average GSP growth
both due to the impact of above average employment growth and due to the impact of above
average productivity growth. On the other hand, above average GSP growth in Queensland and
the Northern Territory was primarily due to employment growth rather than productivity
growth, since the latter effects were negative in these two regions. By contrast, Victoria, South
Australia and Tasmania were shown to be experiencing below average GSP growth both due
to the impact of below average employment growth and to the impact of below average
productivity growth. On the other hand, below average GSP growth in New South Wales was
primarily due to lagging employment growth rather than productivity growth, since the latter
effects were in fact positive for New South Wales.
29 These changes are not deemed significant, since the signs of the sub-components have remained the same. All that has altered
is the relative size of these sub-components.
– 93 –
Productivity and Regional Economic Performance in Australia
Non-labour factors were demonstrated as most significant in driving GSP growth in New
South Wales, South Australia, Western Australia and Tasmania. In Western Australia and New
South Wales this led to above average productivity-related output change in total being
converted into a below average productivity-related output change attributable to labour alone.
To our knowledge, previous applications of the Rigby and Anderson (1993) and Haynes and
Dinc (1997) variants of the shift-share method have been restricted to the analysis of
employment change. This chapter has successfully adapted these methods for use in the
analysis of GSP or output change. We believe these adaptations represent a significant
contribution to the shift-share literature since the change components generated by these
adaptations are inherently easier to interpret than their Rigby and Anderson (1993) and Haynes
and Dinc (1997) counterparts.
Future research needs to focus on total factor productivity change more explicitly. Some
preliminary research that attempts to do this in the context of Australian states and territories
is reported in Chapter 5.
– 94 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
References
Australian Bureau of Statistics (1997), Information Paper: Implementation of Revised
International Standards in the National Accounts, Cat. no. 5251.0, Canberra.
Australian Bureau of Statistics (2000), various data, Canberra, in dX EconData, database,
EconData Pty Ltd, Melbourne.
Barff, R.A. and Knight, III, P.L. (1988), ‘Dynamic shift-share analysis’, Growth and
Change, 19 (2), 1-10.
Beeson, P.E. (1987), ‘Total factor productivity growth and agglomeration economies in
manufacturing 1959-73’, Journal of Regional Science, 27, 183-199.
Boudeville, J.R. (1966), Problems of Regional Planning, Edinburgh University Press,
Edinburgh, 77-80.
Cashin, P. (1995), ‘Economic growth and convergence across the seven colonies of
Australasia: 1861-1991’, The Economic Record, 71 (213), 132-144.
Cashin, P. and Strappazzon, L. (1998), ‘Disparities in Australian regional incomes: Are
they widening or narrowing?’, Australian Economic Review, 31 (1), 3-33.
Dawson, J. (1982), ‘Shift-share analysis: A bibliographic review of technique and
applications’, Public Administration Series, P-949, Vance Bibliographies, Monticello,
Illinois.
Denning, W.G. (1996), ‘A decade of economic change and population shifts in US
regions’, Monthly Labour Review, November, 3-14.
Dinc, M. and Haynes, K.E. (1998a), ‘International trade and shift-share analysis: A
specification note’, Economic Development Quarterly, 12 (4), 337-343.
Dinc, M. and Haynes, K.E. (1998b), ‘International trade and shift-share analysis: A
specification note, rejoinder’, Economic Development Quarterly, 12 (4), 351-354.
Dinc, M., Haynes, K.E. and Qiangsheng, L. (1998), ‘A comparative evaluation of shiftshare models and their extensions’, Australasian Journal of Regional Studies, 4 (2),
275-299.
Donovan, J.B. (1981), ‘Social accounting measures of regional growth’, in R.L. Mathews
(ed.) Regional Disparities and Economic Development, Centre for Research on Federal
Financial Relations, Australian National University, Canberra, 221-254.
Dunn, E.S. (1960), ‘A statistical and analytical tool for regional analysis’, Papers of the
Regional Science Association, 6, 97-112.
Fothergill, S. and Gudgin, G. (1979), ‘In defence of shift-share analysis’, Urban Studies,
16 (3), 309-319.
Fuchs, V.R. (1962), ‘Statistical explanations of the relative shift of manufacturing among
the regions of the United States’, Papers, Regional Science Association, 8, 105-126.
– 95 –
Productivity and Regional Economic Performance in Australia
Graham, D.J. and Spence, N. (1998), ‘A productivity growth interpretation of the labour
demand shift-share model’, Regional Studies, 32 (6), 515-526.
Harris, P. (1998), ‘Changes in the extent of interstate disparities in GSP per head in
Australia 1977-78 to 1994-95: A comparison of the findings of two methods of
analysis’, Australasian Journal of Regional Studies, 4 (2), 193-213.
Harris, P. and Harris, D. (1991), ‘Interstate disparities in real gross state product at factor
cost per head in Australia, 1953-54 to 1990-91’, Australian Journal of Regional Studies,
6, 42-58.
Harris, P. and Harris, D. (1992), ‘Interstate differences in economic growth rates in
Australia, 1953-54 to 1990-91’, Economic Analysis and Policy, 22 (2), 129-148.
Haynes, K.E. and Dinc, M. (1997), ‘Productivity change in manufacturing regions: A
multifactor/shift-share approach’, Growth and Change, 28, 201-221.
Haynes, K.E. and Machunda, Z.B. (1987), ‘Considerations in extending shift-share
analysis’, Growth and Change, 18 (2), 69-78.
Hewings, G.F.D. (1977), Regional Industrial Analysis and Development, St Martins Press,
New York.
Knudsen, D.C. and Barff, R. (1991), ‘Shift-share analysis as a linear model’, Environment
and Planning A, 23, 421-431.
Ledebur, L.C. and Moomaw, R.L. (1983), ‘A shift-share analysis of regional labour
productivity in manufacturing’, Growth and Change, 14 (1), 2-9.
Markusen, A. Noponen, H. and Driessen, K. (1991), ‘International trade, productivity and
regional growth’, International Regional Science Review, 14 (1), 15-39.
Maxwell, P. and Hite, J.C. (1992), ‘The recent divergence of regional per-capita incomes:
Some evidence from Australia’, Growth and Change, 23, 37-53.
Neri, F. (1998), ‘The economic growth performance of the states and territories of
Australia: 1861-1992’, The Economic Record, 74 (225), 105-120.
Nguyen, D.T., Smith, C. and Meyer-Boehm, G. (2000), ‘Variations in economic and
labour productivity growth among the states of Australia: 1984/85-1998/99’, paper
presented at the Conference of Economists, Economic Society of Australia, Gold Coast,
July.
Noponen, H., Markusen, A. and Driessen, K. (1997), ‘Trade and American cities: Who
has the comparative advantage’, Economic Development Quarterly, 11 (2), 67-88.
Noponen, H., Markusen, A. and Driessen, K. (1998), ‘International trade and shift-share
analysis: A response to Dinc and Haynes’, Economic Development Quarterly, 12 (4),
344-350.
Noponen, H., Markusen, A. and Shao, Y. (1996), ‘Is there a trade and defence perimeter?
The regional impacts of trade and defence spending in the United States, 1978-1986’,
Growth and Change, 27, 405-433.
– 96 –
Economic Growth Performance of the Australian States and Territories: An Extended Shift-Share Analysis
Perloff, H.S., Dunn, E.S., Lampard, E.E. and Muth, R.F. (1960), Regions, Resources and
Economic Growth, John Hopkins Press, Baltimore.
Richardson, H.W. (1978), ‘The state of regional economics: A survey article’,
International Regional Science Review, 3 (1), 1-48.
Rigby, D.L. (1992), ‘The impact of output and productivity changes on manufacturing
employment’, Growth and Change, 23, 405-427.
Rigby, D.L. and Anderson, W.P. (1993), ‘Employment change, growth and productivity in
Canadian manufacturing: An extension and application of shift-share analysis’,
Canadian Journal of Regional Science, 16 (1), 69-88.
Smith, C. (1979), ‘Estimation and analysis of regional gross domestic product,
Queensland statistical divisions, 1968/69 and 1973/74’, Master of Regional Science
Dissertation, University of Queensland.
Stevens, B.H. and Moore, C.L. (1980), ‘A critical review of the literature on shift-share as
a forecasting tool’, Journal of Regional Science, 20 (4), 419-437.
Stillwell, F.J.B. (1969), ‘Regional growth and structural adaption’, Urban Studies, 6 (2),
162-178.
Thirlwall, A.P. (1967), ‘A measure of the proper distribution of industry’, Oxford
Economic Papers, 19 (March), 46-58.
– 97 –
5 Multifactor Productivity and Innovation
in Australia and its States
Jimmy Louca
Introduction
Productivity growth is the main source of improvements in living standards and sustainable
growth in an economy. While much attention has been paid to the acceleration in multifactor
productivity (MFP) growth in Australia over the past decade or so (Productivity Commission,
1999), less consideration has been given to whether significant interstate differences in MFP
growth have occurred in this period. However, interstate differences exist in many of the
factors thought to influence MFP, including rates of innovation, human capital, competition,
trade openness and labour market flexibility, possibly leading to interstate differences in MFP
growth. Further, the policy tools available at the state level provide state governments
considerable scope to influence such factors and therefore regional economic performance.
Innovation, in particular, has risen in importance as a driver of MFP and thus as a policy
priority. To quote the OECD (2001, p. 51), ‘the ability to harness the potential of new scientific
and technical knowledge and to diffuse such knowledge widely has become a major source of
competitive advantage, wealth creation and improvement in the quality of life’.
This chapter examines MFP in Australian states as a source of the interstate differences in
economic growth and living standards over the period 1985-86 to 2000-01 and then looks at
innovation as one explanation for these interstate trends in MFP and economic performance.
After outlining the importance of productivity and the difference between its labour and
multifactor measures, this study is then divided into two main parts. The first part estimates
MFP at the state level to examine several stylised facts, including the contribution of MFP
growth to interstate differences in economic growth and real per capita incomes, and the
extent of any convergence in MFP levels across the states, given the importance of this
process to convergence in material living standards between the states. The second part then
examines interstate innovation activity as an explanation for these stylised facts, both
through viewing trends in interstate research and development (R&D) and patents data and
through econometric techniques.
The results highlight four notable trends in interstate MFP. First, the states that generated
the highest average annual economic growth over 1985-86 to 2000-01, namely Queensland
(4.5%) and Western Australia (4.4%), also recorded the highest average annual growth in
MFP (at 1.6% and 1.3% respectively). Second, MFP growth accounted for the largest part
(30-85%) of the rise in real per capita incomes in the states, with the likely adverse impact
of future demographic trends, however, raising the importance of policies that promote
higher MFP growth and workforce participation. Third, while labour productivity levels
diverged over the period, differences in capital deepening have masked an underlying
process of convergence in MFP among the five major Australian states, with Queensland
and Western Australia recording MFP growth above those rates expected based on
– 99 –
Productivity and Regional Economic Performance in Australia
convergence channels alone. Finally, in contrast to other states, Queensland recorded
negligible capital deepening over the period 1985-86 to 2000-01, due to employment
growth well above the rest of Australia, particularly over the Prices and Incomes Accord
period. This suggests that Queensland requires higher investment rates relative to other
states if it is to record similar capital deepening, highlighting a vital challenge for
investment policies for this state.
Interstate trends in innovation activity are found to shed considerable light on these MFP trends.
States that generated the highest MFP growth also recorded the strongest growth in business
R&D expenditure and patent grants. Yet, consistent with convergence, this seems to reflect a
catch-up effect, with the actual stock of innovative activity in New South Wales and Victoria
reflecting their historically higher levels of MFP relative to the faster growing states of
Queensland and Western Australia. An econometric analysis of the returns to business R&D
also finds evidence of interstate R&D spillovers and some equalisation in the rates of return to
domestic R&D across the five major Australian states over the period of the study, both
consistent with a process of convergence rather than divergence in MFP. In particular, the returns
to business R&D seem to have been highest, but have fallen most, in Queensland and Western
Australia. This suggests that these states initially faced the greatest opportunities to profit from
R&D and thus invested most heavily in R&D, causing their MFP levels to converge toward
those enjoyed in New South Wales and Victoria relatively faster than did those of other states. The
results also provide some general implications for R&D policy in particular states, in relation to
promoting MFP growth, economic growth and continued convergence in per capita incomes
toward that in higher income states.
Productivity, Living Standards and Sustainable Growth
Productivity growth is the main source of improvements in living standards and sustainable
economic growth in an economy. Growth in productivity creates more output from given
inputs, generating a greater amount of income to be shared among the resident population,
raising real income per capita and thus living standards.
Small differences in the rate of productivity growth compound over time to create large
differences in living standards. Consider two economies with real income per capita levels
of $20,000 in 2000. Assume the first economy records annual productivity growth of 2%,
while the second records 3% growth over the next 25 years. Holding other factors constant,
income per capita in the first economy will grow by over 60% to around $32,800 in 2025,
but will more than double in the second economy to around $41,900 over the same period.
Productivity growth raises real incomes through three channels in particular. Productivity
gains are either passed on to employees in the form of higher real wages, to consumers in
the form of lower prices, or to employers, firm owners and shareholders in the form of
higher profits (Industry Commission, 1997, p. 80). Productivity growth also raises living
standards through indirect channels. The resulting increase in real incomes it creates also
raises tax revenue without the need to raise tax rates, allowing governments to more easily
raise spending on education, health and aged care, environmental protection, crime and
poverty prevention, and cultural activities (Baumol et al., 1988, p. 349).
– 100 –
Multifactor Productivity and Innovation in Australia and its States
There is an important distinction between the impact of input growth and productivity
growth on living standards. Increasing a given input generates higher income, but also
incurs the cost of using that input. For instance, capital accumulation incurs the costs of
investment expenditure, and depreciation, and uses up depletable mineral and fuel resources
needed to drive machinery and equipment (Baumol, 1988, p. 362). Similarly, labour
accumulation occurs at the cost of leisure for employees and payroll costs for employers.
By raising output without raising inputs, productivity growth avoids such production costs.
Productivity growth is the main driver of sustainable economic growth. Generally, inflation
rises if the growth in aggregate demand exceeds that in aggregate supply. Yet, supply is the
product of inputs and their productivity. Increases in productivity enable producers to raise
supply without significantly raising costs, allowing aggregate demand to grow at a faster
rate without the need to pass cost increases on to consumer prices. Productivity is thus
crucial to high rates of non-inflationary economic growth.
Labour Productivity versus Multifactor Productivity
Labour productivity and multifactor productivity are the two most widely used measures of
productivity. Labour productivity can be defined as real output produced per person
employed or hour worked. Growth in labour productivity plays a crucial role in the labour
market. By reducing labour costs, it stimulates employment growth and also lessens the
need for employers to pass on wage increases to consumer prices. However, labour
productivity is only a partial measure of productivity, since in reflecting the productivity of
a single input, it can be simply raised through capital deepening, where more capital is
added to a given amount of labour.
Multifactor productivity (MFP) is a more comprehensive measure of productivity, since it
measures real output produced per joint unit of inputs, namely labour and capital.1 Growth
in MFP occurs when output is increased without any increase in inputs and is underpinned
by improvements in efficiency and technological progress. Labour productivity and MFP
are clearly related. MFP growth differs from growth in labour productivity by excluding the
impact of capital deepening, whereas labour productivity will grow in line with MFP growth
plus the rate of capital deepening.
Labour productivity is the measure more often used in productivity comparisons between
national economies, industries and particularly between regions, mainly because it is
simpler to calculate, with MFP requiring capital stock estimates and assumptions about the
functional form of the production process. However, this study addresses data and other
limitations in calculating preliminary estimates of MFP across Australian states.
1
This chapter follows the Productivity Commission and the Australian National University (1998, p. xviii) in making a distinction
between multifactor productivity and total factor productivity. The former is taken to reflect the productivity of the two main
factors of production, labour and capital, in generating value added. The latter also incorporates intermediate transactions in
materials and services and uses gross output as a measure of output.
– 101 –
Productivity and Regional Economic Performance in Australia
Multifactor Productivity in Australian States
The first part of this chapter is concerned with the stylised facts of MFP across Australian
states. It first looks at the contribution of MFP growth, along with accumulation in labour
and capital, to state economic growth, and then considers how MFP growth has combined
with capital deepening to determine state labour productivity growth over the period
1985-86 to 2000-01. The contribution of MFP and other demographic, labour market and
terms of trade factors to growth in material living standards in each state is also examined.
Finally, and perhaps most importantly, this section explores whether state MFP levels have
tended to converge over the study period, given the importance of this process to
convergence in real per capita incomes across the states.
The estimates of state level MFP presented in this chapter are derived from the Törnqvist
index number methodology. This technique estimates labour and capital contributions by
weighting the growth of each input by its cost share in output, with MFP growth given as
any unexplained residual output growth. This method is based on a strong theoretical
framework. In particular, by assuming markets with competitively priced inputs, the cost
shares of capital and labour reflect their output elasticities in a Cobb–Douglas production
function with constant returns to scale. Appendix 5A discusses the Törnqvist methodology,
its assumptions and limitations, particularly in relation to applying the approach at a state
level. Data sources and descriptions are contained in Appendix 5B.
The contribution of MFP, labour and capital to economic growth
One advantage of the Törnqvist method is that it allows economic growth to be decomposed
into contributions from accumulation in inputs, labour and capital, and MFP growth. This
enables an examination of the relative importance of MFP as a driver of economic growth
in each state. The results indicate that those states with the highest economic growth over
the period 1985-86 to 2000-01, namely Queensland and Western Australia, also recorded
the highest MFP growth, with MFP accounting for around 20-45% of overall economic
growth across the states over the period.
Before examining the state level results, Figure 5.1 shows the resulting Törnqvist estimates of
MFP growth over the period 1964-65 to 1999-2000 for Australia as a whole. This helps place
in historical context the state level results, available only for the past 15 years. Two points
should be noted about Figure 5.1. First, it divides growth in MFP into periods relating to
productivity cycles, with growth rates calculated over intervals between productivity peaks in
order to avoid the spurious effects of business cycles.2 Second, it also shows that the whole
economy estimates in this study are similar to the Australian Bureau of Statistics (ABS) MFP
estimates for the ‘market sector’, which accounts for around two-thirds of the economy. In
deriving the market sector, the ABS excludes industries whose output is estimated primarily
from input data, since this method effectively assumes zero productivity growth (ABS, 2000,
p. 363). However, this study estimates MFP for the entire economy, since the industry detail
required to calculate market sector estimates is not available at the state level.3
2
MFP tends to rise and fall with the business cycle. Firms face costs in hiring and firing workers (screening potential employees,
training and severance pay). Thus, firms may work labour harder, rather than hire labour, when demand rises, and then lower
the utilisation rate of labour, rather than fire labour, when demand falls. Thus, cyclical fluctuations in output are larger than
fluctuations in inputs, creating pro-cyclical MFP growth.
– 102 –
Multifactor Productivity and Innovation in Australia and its States
Figure 5.1: Multifactor Productivity in Australia, average annual growth
Tornqvist (full economy)
ABS (market sector)
2.0
1.7 1.8
1.8
1.6
1.5
Per cent
1.4
1.2
1.2
1.1
1.0
1.2
1.1
1.1
1.0
0.8
0.9
0.8
0.7
0.6
0.4 0.4
0.4
0.2
0.0
1964-65 to
1968-69
1968-69 to
1973-74
1973-74 to
1981-82
1981-82 to
1984-85
1984-85 to
1988-89
1988-89 to
1993-94
1993-94 to
1999-00
Source: ABS estimates from ABS, Australian System of National Accounts, Cat. no. 5204.0.
Figure 5.1 shows that MFP in Australia grew strongly for the two decades following the mid
1960s, before moderating sharply in the 1980s. Explanations for strong growth over the
former period include postwar reconstruction and technological catch-up aided by liberal
US trade policies, leading to a golden era of growth since World War II (Gordon, 2000;
Industry Commission, 1997, p. 337). Explanations for the subsequent slump in MFP growth
in the 1980s have been numerous, including the oil price and wage shocks (Industry
Commission, 1997, p. 38), less technological catch-up opportunities as Australia became
more industrialised (Dowrick, 1995), a rise in low-skilled employment induced by the real
wage restraint under the Accord (Lowe 1995, p. 95), a shift in industrial structure from
manufacturing to lower productivity services (Industry Commission, 1997, p. 37), and a fall
in R&D intensity (Baumol et al., 1988, p. 349).
However, Figure 5.1 also illustrates that the period covered in this state level study, namely
1984-85 to 2000-01, covers an era in which national MFP growth accelerated markedly.
Explanations for this strengthening in MFP growth since the 1980s have ranged from a restored
low inflation environment, greater domestic competition, more flexible labour markets, greater
R&D intensity and an information and communication technologies boom (Industry Commission,
1995; Parham et al., 2001). However, there has been little research conducted into how these
productivity gains have been distributed across the states over the study period.
Table 5.1 illustrates the Törnqvist results of decomposing state economic growth over the
period 1985-86 to 2000-01 and into three sub-periods. The first relates to the late 1980s
boom (1985-86 to 1989-90), the second covers the recovery from the early 1990s recession
3
The market sector includes the following industries: agriculture; mining; manufacturing; electricity, gas and water;
construction; wholesale trade; retail trade; accommodation, cafes and restaurants; transport and storage; communication
services; finance and insurance; and cultural and recreational services. It excludes property and business services, government
administration and defence, education, health and community services, personal and other services, and ownership of
dwellings. While including zero productivity growth industries should bias downward this study’s estimate relative to the ABS,
there is no persistent bias in Figure 5.1. However, other differences exist between this study’s estimation procedure and that of
the ABS, including the latter’s use of more disaggregated capital and income data.
– 103 –
Productivity and Regional Economic Performance in Australia
(1990-91 to 1994-95), while the third spans a period commonly called the ‘New Economy’
era following the noticeable pick up in productivity growth in the United States (and
Australia) over the late 1990s (1995-96 to 2000-01).
Table 5.1: A Decomposition of Economic Growth in Australian States,a b
annual averages
State
Time period
NSW
Vic
Qld
SA
WA
Tas
Australia
Output
Contribution to output growth
growth
Labour
Capital
MFP
1985-86 to 1989-90
3.3
1.8
1.2
0.3
1990-91 to 1994-95
2.2
0.4
0.8
1.0
1995-96 to 2000-01
4.2
1.1
0.9
2.1
1985-86 to 2000-01
3.3
1.1
1.0
1.2
1985-86 to 1989-90
3.1
2.2
1.3
-0.4
1990-91 to 1994-95
1.4
-0.3
0.5
1.1
1995-96 to 2000-01
4.4
1.2
1.2
2.0
1985-86 to 2000-01
3.0
1.0
1.0
1.0
1985-86 to 1989-90
5.1
3.2
1.0
0.8
1990-91 to 1994-95
4.0
1.5
0.8
1.6
1995-96 to 2000-01
4.5
1.2
1.0
2.2
1985-86 to 2000-01
4.5
1.9
0.9
1.6
1985-86 to 1989-90
3.7
1.6
1.0
1.0
1990-91 to 1994-95
0.7
-0.1
0.4
0.3
1995-96 to 2000-01
2.8
0.4
0.5
1.9
1985-86 to 2000-01
2.4
0.6
0.7
1.1
1985-86 to 1989-90
5.8
2.7
1.7
1.3
1990-91 to 1994-95
4.3
1.2
1.0
2.0
1995-96 to 2000-01
3.3
1.2
1.2
0.8
1985-86 to 2000-01
4.4
1.7
1.3
1.3
1985-86 to 1989-90
1.7
1.5
1.5
-1.3
1990-91 to 1994-95
1.4
-0.2
0.6
1.0
1995-96 to 2000-01
1.4
0.1
0.2
1.1
1985-86 to 2000-01
1.5
0.4
0.8
0.3
1985-86 to 1989-90
4.0
2.2
1.3
0.4
1990-91 to 1994-95
2.4
0.5
0.7
1.1
1995-96 to 2000-01
4.0
1.1
1.0
1.8
1985-86 to 2000-01
3.5
1.3
1.0
1.2
a
As with all MFP estimates, the numbers should be taken as indicative of trends rather than precise estimates of
productivity growth, due to the measurement problems involved (Industry Commission, 1997, p. 29), particularly in
relation to state capital stocks (see Appendices 5A and 5B).
b
State MFP estimates are based on ABS data for consistency of measurement. However, Queensland Treasury produces
preferred MFP estimates for Queensland based on Queensland State Accounts data.
– 104 –
Multifactor Productivity and Innovation in Australia and its States
The results indicate that those states with the highest economic growth over the period
1985-86 to 2000-01 also recorded the highest MFP growth. Queensland recorded the
highest average annual growth in real output and MFP, at 4.5% and 1.6% respectively,
followed by Western Australia (4.4% and 1.3%) and New South Wales (3.3% and 1.2%).
South Australia (1.1%), Victoria (1.0%) and Tasmania (0.3%) all recorded MFP growth
below the national average (1.2%) over the period. The acceleration in MFP growth over
the ‘New Economy’ period was strongest in South Australia, rising from an annual average
of 0.3% over the period 1990-91 to 1994-95 to 1.9% over 1995-96 to 2000-01. Table 5.1
illustrates that the share of MFP growth in economic growth was highest in South Australia
(46.1%), followed by New South Wales (35.2%), Queensland (34.8%), Victoria (31.5%)
and Western Australia (30.6%).
The fact that MFP growth comprised a relatively lower share of economic growth in
Queensland and Western Australia can be explained by noting that these two states also
recorded the highest growth contributions from labour and capital. Queensland recorded the
highest growth contribution from labour (1.6 percentage points), consistent with its faster
population and employment growth relative to any other state over the period. Similarly,
Western Australia recorded the highest growth contribution from capital (1.7 percentage
points), consistent with strong investment in the North West Shelf in the 1990s.
Noting the interrelationships between the three components of economic growth in Table
5.1 also helps explain why the states that recorded the highest MFP growth also recorded
the highest growth in inputs. For example, the strong mining investment in Western
Australia may have not only reflected greater capital accumulation, but also embodied new
technology, helping underpin MFP growth. Similarly, stronger productivity growth in
Queensland may have helped lower labour costs, stimulating employment growth in
addition to that induced by strong growth in real output and the resident population.
Labour productivity, capital deepening and MFP
Each state’s labour productivity growth can be decomposed into the contributions from two
components, when recalling that labour productivity will grow in line with MFP growth
plus the rate of capital deepening. This distinction is important since, while labour
productivity is the more often used measure, variations in capital deepening across states
will mask the true rate of change in technology and efficiency underpinning MFP growth.
This decomposition also gives a regional insight into Dowrick’s (1990) national finding that
the 1980s slowdown in labour productivity was due to an easing in capital deepening, in
addition to moderating MFP growth.
Table 5.2 decomposes state labour productivity growth into capital deepening and MFP
growth. While Queensland (1.6%) recorded the highest growth in MFP ahead of Western
Australia (1.3%) and New South Wales (1.2%), Western Australia (1.8%) recorded the
highest growth in labour productivity, followed by New South Wales (1.5%) and Queensland
(1.5%). The difference in state rankings lies in capital deepening, which contributed
0.4 percentage point to labour productivity in Western Australia but detracted 0.1 percentage
point in Queensland. In fact, Queensland was the only state not to record any significant
capital deepening over the period, with capital deepening and MFP growth accounting for
around 20-40% and 60-80% of labour productivity growth respectively in most other states
over the period.
– 105 –
Productivity and Regional Economic Performance in Australia
Table 5.2: Labour Productivity, Capital Deepening and MFP, annual averages
State
NSW
Vic
Qld
SA
WA
Tas
Australia
a
Time period
Labour
productivity
Contribution from
k/l a
MFP
%
percentage point
1985-86 to 1989-90
0.5
0.2
0.3
1990-91 to 1994-95
1.5
0.5
1.0
1995-96 to 2000-01
2.4
0.3
2.1
1985-86 to 2000-01
1.5
0.4
1.2
1985-86 to 1989-90
-0.3
0.1
-0.4
1990-91 to 1994-95
1.8
0.7
1.1
1995-96 to 2000-01
2.5
0.5
2.0
1985-86 to 2000-01
1.4
0.4
1.0
1985-86 to 1989-90
0.0
-0.8
0.8
1990-91 to 1994-95
1.6
0.0
1.6
1995-96 to 2000-01
2.5
0.4
2.2
1985-86 to 2000-01
1.5
-0.1
1.6
1985-86 to 1989-90
1.1
0.2
1.0
1990-91 to 1994-95
0.8
0.5
0.3
1995-96 to 2000-01
2.2
0.3
1.9
1985-86 to 2000-01
1.4
0.3
1.1
1985-86 to 1989-90
1.6
0.3
1.3
1990-91 to 1994-95
2.4
0.4
2.0
1995-96 to 2000-01
1.4
0.5
0.8
1985-86 to 2000-01
1.8
0.4
1.3
1985-86 to 1989-90
-0.7
0.7
-1.3
1990-91 to 1994-95
1.7
0.7
1.0
1995-96 to 2000-01
1.3
0.2
1.1
1985-86 to 2000-01
0.8
0.5
0.3
1985-86 to 1989-90
0.5
0.0
0.4
1990-91 to 1994-95
1.6
0.5
1.1
1995-96 to 2000-01
2.3
0.4
1.8
1985-86 to 2000-01
1.5
0.3
1.2
Capital/labour.
The estimated slight reversal in capital deepening in Queensland has not been the result of
low rates of capital accumulation relative to other states, but because strong capital stock
growth has been more than offset by even stronger growth in hours worked in Queensland,
causing the capital to labour ratio in the state to remain relatively unchanged over the
period. Queensland recorded average annual capital stock growth of 2.7% over the period
1985-86 to 2000-01, in line with 2.8% growth recorded nationally. However, Queensland
also recorded the highest annual growth in hours worked of any state, at 3.0%, with the state
– 106 –
Multifactor Productivity and Innovation in Australia and its States
accounting for over one-quarter of the total jobs created nationally over the period. In
contrast, capital stock growth exceeded growth in labour hours in all other states.
This provides a regional insight into the findings in Dowrick’s (1990) national level study.
Dowrick was responding to the criticism that wage and price rigidities inherent in the
centralised wage fixing of the early versions of the Accord may have slowed the rate of
MFP growth, either through inhibiting the ability of employers to offer ‘the financial
incentives needed to win union acceptance of technical innovation’ (technological advance)
or attracting ‘skilled labour to its most productive use’ (efficiency). He argued, however,
that a slowdown in capital deepening contributed more to the easing in labour productivity
growth in the 1980s, as the real wage restraint initiated by the Accord stimulated strong
employment growth that largely offset investment growth, leaving the capital to labour ratio
relatively unchanged in the 1980s.
At the regional level, this effect seems to have been concentrated in Queensland, particularly
in the late 1980s and early 1990s – corresponding to the period of the Accord. This highlights
a special challenge for policies related to public and private investment in this state. Put
simply, stronger population and employment growth in Queensland will require faster
growth in capital infrastructure relative to other states if Queensland is to record rates of
capital deepening similar to the rest of Australia. Table 5.2 suggests Queensland has managed
to achieve this more recently, with the State recording average annual rates of capital
deepening more in line with other states over the period 1995-96 to 2000-01.
The contribution of MFP to living standards
The above decompositions of output growth and labour productivity only partly indicate the
contribution of MFP to growth in real income per capita, the standard economic measure of
living standards. If strong output growth is mainly underpinned by population growth, real
output per capita will show little increase. Similarly, if an increase in the productivity of
those employed is offset by a falling proportion of the population in employment, output per
capita will again rise by little.
Clearly, several factors influence growth in real income per capita. This section attempts to
decompose average real incomes into these influences in order to estimate the contribution
of MFP to growth in this measure of living standards in each state over the period 1985-86
to 2000-01. The results indicate that MFP growth was the primary contributor to the rise in
average real incomes in Australia, with its contribution varying from 30% to 80% across the
states. The decomposition also indicates that the likely adverse impact on real income per
capita of future demographic trends will raise the importance of policies that promote higher
MFP growth and increased workforce participation.
The factors influencing living standards can be derived from the relation between living
standards, measured as real output per capita, and labour productivity, measured in terms of
real output per hour worked. Given labour productivity per hour, an increase in the number
of hours worked per person employed will raise real output per capita. Given labour
productivity and average hours worked, a higher proportion of the population in
employment will also raise real output per capita. This is illustrated in equation (1), where
output per capita (Y/Pr), is the product of output per hour worked (Y/H) (labour
productivity), hours worked per employed person (H/E) (average hours worked), and
– 107 –
Productivity and Regional Economic Performance in Australia
employed persons as a share of the resident population (E/Pr):
(1)
E
Y
Y . H . —
—
—
Pr
Pr = —
E
H
The proportion employed in the population (E/Pr) in turn depends on three other factors: the
civilian population aged 15 years and over as a share of the resident population, (Pc/Pr) (the
working age share of the resident population); the proportion of the working age population
in the labour force (L/Pc) (participation rate); and the proportion of people employed in the
labour force, (E/L) (employment rate). Substituting these factors into (1) gives the following
identity:
L
E
H
P
Y
Y
— = — . — . — . — . —c
(2)
Pc
L
E
Pr
H
Pr
The identity in (2) can be expanded into (3) when noting that the employment rate is one
minus the unemployment rate (ur) and labour productivity is comprised of the capital to
labour ratio (K/L) and MFP. Taking logarithms and differentiating with respect to time
converts (3) into a growth identity in (4), when noting the differential of log(1-ur) is
approximately equal to the percentage point change in the unemployment rate (6ur):
Y
K
H
L
P
— = — . MFP . — . (1-ur) . — . —c
L
E
Pc
Pr
Pr
(3)
ˆ = k/l
ˆ + mfp
ˆ + h/e
ˆ - ¨u + l/p
ˆ + p ˆ/p
y/p
r
r
c
c r
(4)
Finally, the influence of the terms of trade should also be included in any living standards
measure. The terms of trade measure export prices relative to import prices. A fall in terms of
trade, through a rise in the price of imports relative to exports, reduces the volume of imports
that can be purchased with a given volume of exports, reducing real purchasing power and real
income (ABS, 1993, p. 112; Industry Commission, 1997, pp. 8-9). Thus, real output should be
adjusted for terms of trade to create a real income per capita measure of living standards.
This adjustment gives the relation for living standards in (5), with real income per capita
growth (i/p̂r) comprised of improvements in terms of trade (t ôt), capital deepening, MFP
growth, increases in average hours, decreases in the unemployment rate, and rises in the
participation rate and working age share of the population:
ˆ = tot
ˆ + k/l
ˆ + mfp
ˆ + h/e
ˆ - ¨u + l/p
ˆ + p ˆ/p
i/p
r
r
c
c r
(5)
Table 5.3 illustrates the results of applying the decomposition of real income per capita in
(5) to each state over the period 1985-86 to 2000-01. The table shows that average annual
MFP growth of 1.2% was the main factor raising living standards in Australia as a whole,
comprising over 55% of annual growth of 2.2% in real income per capita over the period.
Capital deepening, and rises in the participation rate and working age share of the
population each comprised 14-15% of the rise in real income per capita, while a slight fall
– 108 –
Multifactor Productivity and Innovation in Australia and its States
in average hours was offset by a fall in the unemployment rate. This result is in line with
previous national studies, which estimated MFP growth to account for 60-65% of the rise
in per capita real incomes over the past 35-40 years (Industry Commission, 1997, p. 26;
Henry, 2001, p. 50).4
Table 5.3: Real Incomes in Australian States, a b average annual growth,
1985-86 to 2000-01
Component
NSW
Vic
Qld
SA
WA
Tas
Aust
Real output
per capita
Real income per capita
Terms of trade
2.2
0.0
1.9
-0.2
1.9
-0.4
1.7
-0.2
2.8
0.4
1.0
0.0
2.2
0.0
Labour
productivity
Capital deepening
Multifactor productivity
0.4
1.2
0.4
1.0
-0.1
1.6
0.3
1.1
0.4
1.3
0.5
0.3
0.3
1.2
Employed
as share of
population
Participation rate
Working age
Average hours
Unemployment rate
0.2
0.3
-0.1
0.2
0.4
0.3
-0.1
0.1
0.5
0.3
-0.2
0.1
0.1
0.3
-0.1
0.1
0.4
0.3
-0.2
0.2
0.1
0.3
-0.2
0.1
0.3
0.3
-0.1
0.1
a
This study follows the ABS (1993, p. 113) method of revaluing exports with the price deflator for imports to provide a
measure of the purchasing power of exports over imports and substituting this value for the actual constant price value
of exports in deriving real gross domestic product.
b
The terms of trade adjustment for the states has been conducted on relative prices of traded goods, since services data
are not available over the entire period. Another limitation in interpreting the terms of trade result is that some states
import many of their overseas goods via the larger states of New South Wales and Victoria.
Of the states, Western Australia (2.8%) recorded the strongest annual growth in real
income per capita, followed by New South Wales (2.2%), Queensland (1.9%) and Victoria
(1.9%). MFP growth was of most importance in Queensland, comprising 84% (1.6 / 1.9)
of the rise in real income per capita, where it offset detractions from terms of trade, a fall
in average hours and some reversal in capital deepening. In contrast, MFP growth
comprised only 48% (1.3 / 2.8) of the rise in real income per capita in Western Australia,
since terms of trade and capital deepening also contributed to growth in living standards in
this state. Overall, Table 5.3 illustrates that labour productivity and MFP growth have
comprised around 30-85% and 60-85% respectively of the rise in real incomes across the
states over the period 1985-86 to 2000-01.
Several points can be taken from Table 5.3. First, MFP growth and policies promoting it will
become even more important to future living standards given likely demographic trends.
Demographic influences have contributed to rising average real incomes over the past, as the
postwar baby boom has shifted a greater share of the population into working age and into
peak participation rate age groups. However, demographic influences are likely to contribute
less in the future, as the baby boomer population moves into lower participation rate age
groups, falling mortality rates raise the proportion of people older than working age and a
growing service sector and incidence of part-time employment lower average hours worked.
Thus, productivity policies will increase in importance in the future, as higher MFP growth
will be required to offset demographic trends and maintain growth in living standards.
4
The slightly higher MFP contributions in these two studies results from their inclusion of the 1970s period, when strong MFP
growth offset a rise in the unemployment rate and a decline in average hours worked.
– 109 –
Productivity and Regional Economic Performance in Australia
The second point to take from the table is that policies promoting greater participation in
the labour market will also be of more importance in future decades. Clearly, labour market
policies reducing the unemployment rate in each state would add to average real incomes,
as would policies encouraging greater workforce participation. Indeed, Henry (2001, p. 51)
states that ‘the positive contribution of population dynamics might have been responsible
for some complacency, in some quarters, in policies affecting workforce participation’. He
goes on to argue that ‘over the next forty years, with population dynamics detracting from
growth, these policy areas will, necessarily be centre stage’.
However, it should be noted that the negative impact of future demographic trends on living
standards might be overstated for several reasons. Education and preference differences may
cause the baby-boomer generation to maintain a stronger labour force attachment than
previous generations as it moves into the historically non-peak participation age groups.
Further, an increasing proportion of the population above working age and declining average
hours worked may actually represent improvements to living standards, if these trends reflect
the longer life expectancy brought about by improvements in medicine and health and the
increasing value society places on leisure. Clearly, real income per capita is only an economic
measure, and other factors such as leisure, environment, life expectancy and income equality
also shape living standards. Having said this, it is MFP growth that generates the real income
needed to be spent on issues such as education, health and aged care, environmental
protection, and crime and poverty prevention. To quote Baumol et al. (1988, p. 363):
To improve the purity of our air and water and to clean up urban
neighbourhoods, many millions of dollars must be made available …
Continued growth would enable the required resources to be provided without
any reduction in the availability of consumer goods. But without such growth,
we may actually be forced to cut back on our programs. Society could thus end
up with less goods and a worse environment.
Convergence in real incomes, labour productivity and MFP
While the previous section illustrated the contribution of growth in MFP to growth in real
income per capita, the importance of MFP growth lies in how it adds to the level of real
income per capita as a measure of living standards. Figure 5.2 illustrates that while
Queensland recorded the highest MFP growth of all states, real income per capita in this
state remains below that of others, with real income per capita across states seeming to
diverge over the period 1984-85 to 2000-01. This section explores whether the levels of
labour productivity and MFP across states have converged, given their importance to
convergence in real income per capita. The results indicate that while labour productivity
levels have diverged, differences in the rate of capital deepening have masked an underlying
process of convergence in MFP among the largest states, a finding similar to that in Dowrick
and Nguyen’s (1989) international OECD study.
– 110 –
Multifactor Productivity and Innovation in Australia and its States
Figure 5.2: Real Incomes in Australian States, per capita
NSW
Vic
Qld
SA
WA
Tas
40,000
Dollars
35,000
30,000
25,000
20,000
15,000
1984-85
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
2000-01
There are two main arguments put forward as to why labour productivity and MFP levels
should converge between regions. In terms of labour productivity, the return on capital is
said to be higher in smaller, developing economies with less infrastructure than
industrialised economies. In this case, capital deepening will occur in the developing
economies until their capital to labour ratio and the return on investment is equalised with
that of industrialised economies, causing convergence in labour productivity, provided
underlying MFP is similar across countries. In terms of MFP, international trade allows less
advanced economies to import new knowledge, ideas and inventions from technological
world leaders, causing convergence in MFP levels, provided the rate of absorption of such
innovations in less advanced economies is faster than the rate of creation in technologically
advanced countries.
Figure 5.3, however, shows that labour productivity levels between Australian states seem
to have diverged rather than converged. Figure 5.3a illustrates that two distinct groups have
emerged: a high labour productivity group comprising New South Wales, Victoria and
Western Australia, and a lower labour productivity group comprising Queensland, South
Australia and Tasmania. Figure 5.3b plots the initial level of labour productivity in each
state in 1984-85 on the horizontal axis against the subsequent average annual growth rate in
labour productivity over the period 1985-86 to 2000-01 on the vertical axis. Convergence
in productivity levels would require a negatively sloped regression line through the scatter
plot, reflecting the fact that states with initially lower levels of labour productivity tend to
record higher rates of labour productivity growth (see Sala-i-Martin, 1996, p. 1327, and
Chapter 3 of this volume). However, the line is positively sloped, suggesting divergence,
although the slope coefficient is statistically insignificant.
– 111 –
Productivity and Regional Economic Performance in Australia
Figure 5.3: Labour Productivity (LP) a in Australian States
(a) Real LP levels, in 1999-2000 dollars
NSW
Vic
Qld
SA
WA
Tas
44
42
Dollars per hour
40
38
36
34
32
30
28
26
24
1984-85
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
2000-01
(b) Convergence v divergence
2.0
LP growth, 1984-85 to 2000-01
1.8
WA
1.6
NSW
Qld
1.4
SA
Vic
1.2
1.0
0.8
Tas
0.6
26
27
28
29
30
31
32
33
34
LP in 1984-85
a
Defined as real output per hour worked.
Source: ABS, Australian National Accounts: State Accounts, Cat. no. 5220.0; ABS, Labour Force, Australia, Cat. no. 6202.0.
– 112 –
Multifactor Productivity and Innovation in Australia and its States
This finding of possible divergence in labour productivity is similar to that found in other
Australian and overseas studies. Nguyen, Smith and Meyer-Boehm in Chapter 3 of this
volume find that real output per hour worked, across the six Australian states, showed no
sign of convergence over the period 1984-85 to 1998-99 and cite a number of other
Australian and international studies that find convergence in per capita output or labour
productivity before the early 1970s, but divergence thereafter.
However, in a seminal contribution, Dowrick and Nguyen (1989) argued that while labour
productivity levels had diverged between countries since the 1970s, inter-country
differences in the rate of capital deepening had masked an underlying process of
convergence in MFP over the period 1950 to 1985. In an inter-country regression using
OECD data, their finding that trend total factor productivity (TFP) growth over 1950 to
1985 was inversely related to the initial level of TFP remained robust after testing for
parameter stability, sample selection bias and many other measurement issues. Dowrick and
Nguyen (1989, pp. 1018, 1021-22) thus concluded that:
Although income levels have not converged for the wider sample, TFP catchup has been operating … the reason that incomes within the wider group of
countries have not converged is the tendency for poorer countries to have low
investment ratios relative to rapidly expanding populations.
A similar dichotomy appears to exist between labour and multifactor productivity across
Australian states. Figures 5.4a and 5.4b illustrate that MFP levels have shown some signs
of convergence, with states on lower initial levels of MFP subsequently recording higher
annual MFP growth. The broken line in Figure 5.4b is the regression line for all states.
However, the negative slope is statistically insignificant. The unbroken line is a regression
line excluding Tasmania, a notable outlier. The slope of this line is –0.29 and statistically
significant at conventional confidence levels. This indicates that on average across states,
annual MFP growth over the period 1985-86 to 2000-01 will be 0.3 percentage point higher
for every dollar MFP is initially lower in 1984-85.
Importantly, the states situated above the unbroken line in Figure 5.4b, namely Queensland,
Western Australia and New South Wales, can be interpreted to have recorded above average
MFP growth with respect to convergence. That is, these states have generated higher MFP
growth than an underlying process of convergence alone would suggest. In comparison,
South Australia and Tasmania have recorded MFP growth below the rates that would be
expected based on this catch-up hypothesis alone.
– 113 –
Productivity and Regional Economic Performance in Australia
Figure 5.4: Multifactor Productivity in Australian States
(a) Real MFP levels, in 1999-2000 dollars
NSW
Vic
Qld
SA
WA
Tas
8.5
8.0
Dollars per hour
7.5
7.0
6.5
6.0
5.5
5.0
4.5
1984-85
1986-87
1988-89
1990-91
1992-93
1994-95
1996-97
1998-99
2000-01
7.5
8.0
(b) Convergence v Divergence
1.8
MFP growth, 1984-85 to 2000-01
1.6
Qld
1.4
WA
1.2
NSW
SA
1.0
Vic
0.8
0.6
0.4
Tas
0.2
0.0
4.0
4.5
5.0
5.5
6.0
6.5
7.0
MFP in1984-85
As in Dowrick and Nguyen’s (1989) international study, the existence of convergence in
MFP despite divergence in labour productivity across Australian states is because smaller
states with lower initial levels of labour productivity have actually recorded strong growth
in their population and labour input relative to their capital input. This has been particularly
the case for Queensland and to a lesser extent for South Australia. As a result, capital
deepening has added less to labour productivity growth in these states relative to larger
states with higher initial levels of labour productivity, causing labour productivity levels
across all states, on average, to diverge.
– 114 –
Multifactor Productivity and Innovation in Australia and its States
However, when removing the effect of capital deepening, convergence in MFP appears to be
operating in five of the six major states. This result is not surprising, given diffusion of new
technologies should occur more easily across regional economies that are geographically
close, similar in terms of industrial structure and culture, and also face similar incentives to
improve efficiency from a federally driven microeconomic reform program.
The second half of this study looks at innovation as one explanation for the stylised facts on
interstate MFP uncovered here, especially in relation to evidence of convergence in MFP
levels across the five major Australian states over the period 1985-86 to 1999-2000 and the
above average performance of Queensland and Western Australia and the below average
performance of South Australia and Tasmania with respect to this hypothesis in particular.
Innovation and Multifactor Productivity
Innovation can be generally described as a process involving the development, application
and diffusion of new knowledge or new products, qualitative improvements to existing
goods or services, or more efficient production processes (Productivity Commission, 1995,
p. 59; OECD, 2001, p. 51). In the private sector, firms innovate in order to gain a
competitive advantage over their rivals. Successful innovation generates greater output and
income from available inputs, either through the creation of new products that expand
market share or the implementation of more efficient production processes, and is thus a
primary source of MFP growth. In the public sector, innovation efforts often centre around
areas of national defence, health, scientific research and environmental protection, again
providing a fundamental source of improvements in living standards over time.
The second half of this chapter examines innovation as an explanation for the revealed
stylised facts on interstate MFP. It overviews the stages of the innovation process in order
to introduce research and development spending and patents as innovation indicators and
then illustrates how trends in these indicators help explain differences in the levels, growth
and convergence behaviour of MFP across Australian states. An econometric analysis of the
MFP gains due to R&D activity across the states is also conducted, after briefly reviewing
the results of similar exercises conducted at the national level. The state level analysis
provides several novel insights that help to explain state MFP trends such as convergence,
including evidence of interstate R&D spillovers and some equalisation in the rates of return
to domestic R&D across Australian states.
The innovation process
The innovation process contains a number of interrelated stages, including those of research,
experimental development, commercialisation and diffusion. In this context, R&D expenditure
has become one of the most widely used indicators of early stage innovation inputs, while
patents have formed one of the more common indicators of later stage innovation outputs.
Innovation is a process involving the generation, transfer and use of knowledge, with
research forming a fundamental source of knowledge creation. The research stage involves
experimental and theoretical work undertaken either to advance general knowledge (basic
research) or to develop knowledge with a specific application in mind (applied research)
– 115 –
Productivity and Regional Economic Performance in Australia
(ABS, 1998). Basic research is often viewed as science, while applied research attempts to
discover uses for the findings in science or solutions to other problems encountered in
stages further along the innovation process (Industry Commission, 1995, p. 60).
The experimental development stage builds on knowledge gained from research or practical
experience in order to produce new products or processes, or improve those already in place
(ABS, 1998). Statistical information on research and experimental development spending is
usually grouped together and simply denoted as research and development, or R&D,
forming one of the most widely used indicators of innovation.
However, R&D only forms an input into the innovation process. Commercialisation is
crucial in determining whether R&D efforts lead to successful innovation. That is, any new
product or process an organisation creates in its R&D stages must be effectively marketed
and distributed if the new product is to generate and expand market share, or if the new
process is to be implemented and provide a cost advantage over rivals. Entrepreneurship is
fundamental to this stage, with commercialisation requiring the creative and risk-taking
ability to link new products or processes to market opportunities and vice versa.
If innovators are successful in commercialising their inventions, these products or processes
are likely to be absorbed, imitated and built upon by others. This is the process of diffusion,
whereby the knowledge contained in inventions (embodied knowledge) or embodied in the
people who created them (disembodied knowledge) spreads from the innovating
organisation through to potential users, purchasers and the general community (OECD,
2001, p. 53). Clearly, the innovation process is not a linear one from research to diffusion,
but a complex process involving interaction between its components. For instance, once an
innovation is diffused, market feedback may identify problems requiring the initial producer
to conduct further R&D to find solutions that can be incorporated into an improved product
for re-commercialisation.
However, the diffusion process generates externalities or ‘spillovers’, whereby innovation
undertaken in one organisation creates rewards for other organisations that are not fully
reaped by the original innovating organisation (Industry Commission, 1995, p. 64).
Spillovers arise since the knowledge created by R&D exhibits ‘public good’ characteristics,
in that it may be made available to others at little marginal cost (non-rivalrous) or it is
difficult to prevent others accessing it (non-excludable) (Geroski, 1994).
Spillovers provide the classic example of a market failure and thus potential rationale for
government involvement in the innovation process. The market failure reflects the fact that
the private returns or benefits accruing to innovators for any particular innovation will be
lower than the total or social returns enjoyed, possibly resulting in the production of too few
ideas or innovations. Thus, there is a role for government in raising private incentives to
innovate when spillovers exist.
One popular form of government involvement is a system of legally enforceable property
rights, such as a patent system. A patent is a right granted to a product or process that meets
criteria of being new, inventive and capable of creating commercial gains, and that prevents
others from using the invention for a fixed period or requires the payment of royalties for
use. Thus, patents represent one of the most often used indicators of innovation outputs.
– 116 –
Multifactor Productivity and Innovation in Australia and its States
Innovation indicators across Australian states
Interstate trends in R&D expenditure and patent grants seem to shed considerable light on
the stylised facts of interstate MFP growth over the period 1985-86 to 1999-2000, especially
in relation to each state’s performance relative to the convergence hypothesis. In particular,
the states that recorded the highest growth in MFP, namely Queensland and Western
Australia, also recorded growth in business expenditure on R&D well above that in any
other state over the period. However, this seems to reflect a catch-up effect, with the actual
stock of innovative activity in New South Wales and Victoria reflecting their historically
higher levels of MFP. Taken together though, these trends help to explain evidence of
convergence in the levels of MFP in Queensland and Western Australia to those in the larger
states at a relatively faster rate than that in South Australia and Tasmania over the period.
Several limitations associated with using business R&D and patents as innovation
indicators should be noted before attempting to explain interstate MFP trends using these
indicators. Some of the more important limitations are briefly discussed below.
As mentioned earlier, R&D only forms an input into the innovation process such that greater
R&D expenditure does not necessarily imply a rise in the number of successful innovations
(Eglander et al., 1998, p. 9). Indeed, the relation between R&D and innovation is likely to
be time varying and may occur with uncertain or variable lags (Crosby, 2000, p. 256). Also,
statistically measured R&D itself forms only one input into the innovation process and
excludes other activities such as engineering, design, and learning by doing (OECD, 2001,
p. 54). Clearly, the importance of R&D as an input into the innovation process varies
between industries and will therefore vary between regions with different industrial
structures.
There are further limitations at the Australian state level. This study looks only at business
expenditure on R&D, since this is the only sector for which annual R&D data are available
in Australia. Other sectors performing R&D include government agencies, tertiary
education institutions and non-profit organisations. While business R&D raised its share of
total R&D over the period 1985-86 to 1999-2000, it remains only one part of each state’s
total R&D effort. Trends in total R&D may differ to business R&D trends across the states.
Finally, the state in which business R&D is recorded may differ to that in which the R&D
was conducted. This may happen, for instance, when a company with subsidiaries in several
states records its R&D in the state in which its head office is located.
The main advantage of patent grants over R&D data is that they are more likely to be related
to innovation outputs. This is because patents, by definition, are granted to a product or
process only if it is novel by world standards and offers a solution to a problem that
technical experts believe is non-obvious. Yet, four limitations of patents data should also be
noted.
First, the granting of a patent may not indicate that successful innovation has actually taken
place, since the invention may turn out not to be commercially viable when taken to the
market. Conversely, not all successful inventions are patented or patentable (Griliches,
1990, p. 1669). In this respect, Intellectual Property Australia (1998, p. 2) also advocate
examining trade mark, copyright and design registrations, which it is argued ‘relate more
– 117 –
Productivity and Regional Economic Performance in Australia
directly to market activity, and thus ‘will more often reflect successful innovation as distinct
from investment innovation’.
Second, not all patents have the same commercial value in terms of the additional income
or productivity they generate. Many patents will reflect minor improvements of little
economic value, while others may involve technological breakthroughs that prove
extremely valuable (Schankerman and Pakes, 1986; Griliches, 1990, p. 1666).
Third, patent statistics also suffer similar limitations to business R&D. The propensity to
patent will vary across industries. For example, patenting in pharmaceuticals may be crucial
due to the threat of imitation, while patenting in computers may be less valuable due to
shorter product lifespans (Productivity Commission, 1999, p. 171). Further, the link
between patents and innovation may vary over time, with the year to year variations in
patent grants, for example, partly responding to changes in staffing levels or other resources
in patent offices (Griliches, 1994, p. 3).
Finally, there are again added interpretation difficulties at the state level. Once more, the
state in which the patent is recorded may not be the same as that in which the related R&D
was undertaken. Further, some inventions may be jointly developed by a number of states
and will be counted as a patent in each participating state. These limitations illustrate that
patent data at the state level should be regarded as approximate and indicative of trends,
rather than reflecting precise estimates of interstate patenting activity.
Bearing these limitations in mind, Figures 5.5 to 5.7 show several trends in business R&D
and patent grants that help shed light on the interstate differences in MFP performance
recorded over the period 1985-86 to 1999-2000.
Figure 5.5a shows that the states that recorded the highest annual growth in MFP and above
average MFP growth with respect to the convergence hypothesis, namely Queensland
(1.6%) and Western Australia (1.3%), also recorded the highest growth in business R&D
over the period (11.4% and 12.2% respectively). Conversely, states that recorded lower
MFP growth, such as South Australia (1.1%), also recorded much lower growth in business
R&D (6.8%) over the period. Note that the region denoted as ‘other’ in Figure 5.5 comprises
Tasmania, the Australian Capital Territory and the Northern Territory, since Australian
business R&D data are only available for these regions in aggregate.
Figure 5.5b illustrates the ratio of business R&D to GDP. Interstate differences in this
measure may reflect differences in the extent to which state economies are focused on
innovative products (subject to the limitations listed earlier). The figure shows that the
business R&D to GDP ratio rose fourfold to 0.4% in Queensland and more than doubled to
0.5% in Western Australia over the period 1984-85 to 1999-2000, while the business R&D
to GDP ratio in the region including Tasmania rose by the smallest amount. Note that the
business R&D to GDP ratio in New South Wales (0.6%) and Victoria (0.9%) remained
significantly higher than in Queensland and Western Australia in 1999-2000, consistent
with the historically higher levels of MFP recorded in New South Wales and Victoria over
the 1984-85 to 1999-2000 period.
– 118 –
Multifactor Productivity and Innovation in Australia and its States
Figure 5.5: Trends in Business Expenditure on R&D
(a) Annual growth, 1985-86 to 1999-2000
14
12.2
12
11.4
Per cent
10
8
7.6
7.4
6.8
6.2
6
4
2
0
NSW
Vic
Qld
SA
WA
Other
(b) Ratio of business R&D to GSP
1.0
0.9
1984-85
0.9
1999-00
0.8
Per cent
0.7
0.6
0.6
0.6
0.5
0.5
0.4
0.5
0.4
0.3
0.4
0.3
0.3
0.2
0.2
0.1
0.2
0.1
0.0
NSW
Vic
Qld
SA
WA
Other
Source: Estimates based on ABS, Research and Experimental Development, Businesses, Australia, Cat. no. 8104.0,
and ABS, Australian National Accounts: State Accounts, Cat. no. 5220.0.
The above trends suggest that the stronger growth in business R&D in Queensland and
Western Australia may reflect a catch-up effect in innovative activity that is consistent with
evidence of convergence in their level of MFP to that in New South Wales and Victoria. In
particular, faster growth in business R&D indicators in Queensland and Western Australia
compared with South Australia and Tasmania helps explain the above and below average
performance of these two pairs of economies respectively in terms of convergence in their
MFP levels toward those in New South Wales and Victoria.
– 119 –
Productivity and Regional Economic Performance in Australia
Figure 5.6: Trends in Patent Grants a
(a) Annual growth, 1990-91 to 1999-2000
6
5.2
5
4
3.7
Per cent
3
2.0
2
1
0.3
0
0.7
-1
-2
-3
-3.3
-4
NSW
Vic
Qld
SA
WA
Tas
(b) Levels, 1989-90 and 1999-2000
500
1989-90
450
1999-00
400
Number
350
300
250
200
150
100
50
0
NSW
a
Vic
Qld
SA
An adequate state disaggregation of patents is only available since 1989-90.
Source: Intellectual Property Australia (IPA)
– 120 –
WA
Tas
Multifactor Productivity and Innovation in Australia and its States
An examination of patent trends produces some similar conclusions. For instance, Figures
5.6a and 5.6b show that Queensland recorded the highest average annual growth in patents
over the period 1990-91 to 1999-2000 in addition to recording the highest annual MFP
growth over the same period, while New South Wales and Victoria have historically
recorded the highest stock of annual patent grants in addition to recording the highest levels
of MFP. The average annual decline in patent grants recorded in South Australia and the
relatively low level of patents are also consistent with the lower MFP growth recorded in
this state and its below average performance with respect to convergence toward New South
Wales and Victoria. However, one notable discrepancy to the trends in business R&D is the
relatively low rate of annual growth in patent grants in Western Australia over the period.
Figure 5.7 illustrates two ways to compare indicators of innovation outputs to innovation
inputs. Figure 5.7a measures ‘R&D potency’, a measure of the return to R&D in terms of
the number of patents it produces. Note that Queensland, Western Australia and South
Australia all have R&D potencies above that in New South Wales and Victoria. This is not
surprising, given business R&D may be a smaller share of inputs into the production
process in Queensland, South Australia and Western Australia, and thus generate a higher
return, all else being equal.
However, a problem with this R&D potency measure is that the patent data cover all sectors,
while the R&D data cover only the business sector. This is most notable when viewing the
high R&D potency in the ‘other’ category, most likely reflecting the high patent outputs of
ACT-based government agencies, such as the CSIRO, that are funded primarily through
government R&D. Also, note that R&D potency seems to have fallen in several states over
the period. Again, this effect may overstate any decline in total R&D potency, since business
R&D has generally grown faster than R&D in other sectors. These issues are revisited in the
state level econometric analysis.
Figure 5.7b measures the coefficient of ‘inventiveness’, calculated as the number of patent
grants per 100,000 of population (Productivity Commission, 1999, p. 172). For most states,
this figure rose over the 1989-90 to 1999-2000 period, as opposed to the R&D potency
measure, mainly because annual growth in the population was much lower than that in
business R&D over the decade. Queensland recorded the highest level of inventiveness of
any state in 1999-2000 (excluding the ‘other’ region), followed by Victoria and New South
Wales. Once more, though, patents give a contrasting view to business R&D in Western
Australia, where inventiveness actually declined over the period.
– 121 –
Productivity and Regional Economic Performance in Australia
Figure 5.7: Innovation Output and Input Indicators
(a) R&D potency
Ratio of patents to $m real business R&D
1.8
1989-90
1.6
1999-00
1.5
1.4
1.2
1.1
1.0
0.9
0.8
0.7
0.7
0.6
0.6
0.4
0.3 0.3
0.4
0.3
0.3
0.3
0.2
0.0
NSW
Vic
Qld
SA
WA
Other
(b) Inventiveness coefficient
Ratio of patents to 100,000 residents
10
9
8.2
8
7.4
7
6
1989-90
8.5
1999-00
7.3
7.3
6.3
6.2
5.7
5.2
4.9
5
4
3.0 3.2
3
2
1
0
NSW
Vic
Qld
SA
WA
Tas
Source: Intellectual Property Australia (IPA); ABS, Research and Experimental Development, Businesses, Australia,
Cat. no. 8104.0; ABS, Australian Demographic Statistics, Cat. no. 3101.0
A logical extension in attempting to explain the stylised facts of interstate differences in MFP
growth in Australia is to econometrically estimate the gains in MFP due to business R&D
directly, given the limitations of patent data at the state level and in particular the results
relating to Western Australia. Directly linking R&D to actual MFP has several advantages,
some of which include the fact that differences in the commercial significance of innovation
outputs should be reflected in differences in MFP, unlike simple patent counts, and that
trends in MFP should also reflect the gains derived from innovations that are not patented.
– 122 –
Multifactor Productivity and Innovation in Australia and its States
Empirical evidence on returns to R&D in Australia
This section discusses a number of econometric studies looking at the MFP gains due to
R&D in Australia in order to place in context the state-based results in the next section and
introduce some conventions in estimating returns to R&D used therein. This brief survey
indicates that Australian studies estimate returns to domestic R&D ranging from 27% to
196% and international R&D spillovers ranging from 3% to 27%, with most estimates
gathering toward the upper end of these ranges.
Econometric work usually employs one of two methods to examine MFP gains due to R&D.
A production function, as in (6a), can be estimated, relating output to labour and capital
inputs, and to the ‘domestic R&D stock’ (RD) and ‘foreign R&D stock’ (RF). The residual
output not explained by labour and capital (logY - _logL + `logK) is then taken as MFP.
Alternatively, a two-step method can be used, where MFP is first estimated by a Törnqvist
index number technique and then regressed on relevant R&D stocks, as in (6b):
logY = _logL + `logK + a1logRD + a2logRF
(6a)
logMFP = a1logRD + a2logRF
(6b)
The R&D stocks are calculated by accumulating annual R&D expenditure while at the same
time allowing for a depreciation rate. These stocks reflect the fact that R&D is a cumulative
activity that adds to a given stock of knowledge, while at the same time allowing some past
R&D to diminish in value over time as new products or processes make previous knowledge
obsolete (Griliches, 1979, pp. 100-1). Further, the foreign R&D stock for a given country is
usually calculated by using some measure of technological proximity (such as import
shares) to weight together the R&D stocks of all other countries.
The MFP gains due to R&D are also reported in two main ways. The elasticity coefficient
a1 or a2 from (6a) or (6b) can simply be reported, giving the percentage change in MFP for
a 1% change in domestic or foreign R&D stocks respectively. Importantly, a positive
coefficient on the foreign R&D stock suggests evidence of international R&D spillovers,
whereby R&D conducted overseas raises income or MFP domestically.
However, a preferable way to report the MFP gains due to R&D is to report the marginal
product or rate of return to R&D – the rise in MFP due to a unit rise in the R&D stock
(Productivity Commission, 1995, p. QB9). This is of greater policy interest since investment
in R&D should be undertaken to the point where the marginal return equals that available
from the best alternative investment. Note that the output elasticity of R&D simply gives
the relative change in output (bY/Y) divided by the relative change in the R&D stock (bR/R),
as in (7a). The rate of return can thus be calculated in (7b) by multiplying this elasticity (a)
by the ratio of output to the relevant R&D stock, giving the change in output for a given
change in the R&D stock:
(7a)
bY/Y
bY R
a = ____ = __ . __
bR/R
bR Y
(7b)
Y
bY R Y
bY
a . __ = ___ . __ . __ = ___
R
bR Y R
bR
– 123 –
Productivity and Regional Economic Performance in Australia
As an example, assume a1 = 0.05, so that a 1% increase in the domestic R&D stock raises MFP
by 0.05%. If output is 30 times larger than the R&D stock, the return on R&D is given as
150% (0.05 x 30 x 100). That is, a $100 increase in the R&D stock raises income by $150
(Coe and Helpman, 1993, p. 874). Importantly, even if the elasticity (a) remains unchanged
over time, the return to R&D will show diminishing (increasing) returns if output grows
slower (faster) than the R&D stock over time (Griliches, 1998, p. 270). This point is important
when viewing the results of the state-based econometric analysis in the next section.
Coe and Helpman (1993) looked at the returns to R&D in 22 countries over the period 1971
to 1990, in a seminal work built upon by many Australian researchers. They estimated the
panel equation in (8), which differs slightly to (6b) in two respects. First, they used a
dummy variable (G7) to distinguish between MFP elasticities of R&D for 15 smaller
countries (aa) and the G7 countries (aa + ab). Second, they multiplied the import-weighted
foreign R&D stock for each country by its share of imports in GDP (m). This was to account
for the fact that for two countries with the same foreign R&D stocks, the country that
imports more relative to its GDP (more exposed to trade) should enjoy a higher MFP
elasticity to foreign R&D (equal to a2 · mit in this case).
logMFPit = _i + aalogRD + ab· G7 · logRD + a2 · mit · logRF
(8)
Coe and Helpman estimated the MFP elasticity of domestic R&D to be 0.234% for the G7
countries and 0.078% for the remaining 15 economies. Importantly, they found that the
foreign R&D elasticity was higher than the domestic R&D elasticity for most of the 15
smaller economies, with the opposite the case for all the G7 countries. This result reflected
the fact that the smaller countries were both more open to trade (higher import shares of
GDP) and had access to larger foreign R&D stocks relative to their G7 counterparts.
However, Coe and Helpman’s findings in relation to Australia have drawn considerable
comment. They found that Australia was one of four exceptions, having the lowest foreign
R&D elasticity (0.055%) among the group of 15 smaller economies and the third lowest
elasticity for the entire country group, ranking only behind the United States (0.033%) and
Japan (0.027%). This result led Rogers (1995, p. 169), for instance, to argue that ‘Australia
may be relatively poor at benefiting from R&D carried out in the rest of the world’.
The reason Coe and Helpman found a low international R&D spillover for Australia is that
they estimate this country to have had the lowest import share of GDP in the smaller country
group. In contrast, Belgium was measured to have the highest import share of GDP and thus
the highest international spillover. In reviewing Coe and Helpman’s study, the Industry
Commission (1995, Appendix QA, pp. 47-48) thus stated that:
This fact seems to further strengthen their conclusions that the more open an
economy is to trade, then the more able it is to benefit from R&D undertaken
elsewhere. Recent trade liberalisation in Australia over the last decade is likely to
increase the importance of foreign R&D to the Australian economy in the future.
Coe and Helpman only converted the elasticity on domestic R&D to a rate of return in the
body of their paper. They estimated returns to own R&D in 1990 of 123% and 85% for the
G7 and smaller economy groups respectively. However, returns to domestic and foreign
R&D for individual countries can be calculated via information contained in the appendix
– 124 –
Multifactor Productivity and Innovation in Australia and its States
to their paper. This exercise implies Australia had the fifth highest return to domestic R&D,
at 196%, but the fourth lowest return to foreign R&D, at 3%, of the smaller economy group
in 1990.
Several studies have replicated this analysis for Australia. Dowrick (1994) used the Coe and
Helpman dataset to estimate a separate MFP relation for Australia. He found a slightly lower
return to domestic R&D of 166% and a slightly higher return to foreign R&D of 19%
compared with the Australian estimates implied by Coe and Helpman’s results. Rogers
(1995) also tested the sensitivity of the Australian results in Coe and Helpman by replacing
their OECD estimate of MFP with the ABS market sector estimate and using only the G7
countries plus Holland to construct foreign R&D stocks. While Rogers did not convert
elasticities into rates of return, his results suggested that a 1% increase in the domestic and
foreign R&D stock would on average raise MFP by 0.089 and 0.042 respectively, similar to
the elasticities of 0.078 and 0.055 found by Coe and Helpman.
The Industry Commission (1995, Appendix QB) conducted its own analysis into the returns
to R&D, given what it called at the time a ‘limited number of Australian studies’ attempting
to ‘econometrically estimate the returns to R&D at the economy-wide as well as sectoral
level’. Table 5.4 illustrates the economy-wide results from the analysis over the period
1975-76 to 1990-91. It illustrates that the returns to the domestic R&D and foreign R&D
stock varied from 25% to 149% and from 8% to 27% respectively, depending on the
assumptions made about non-market sector MFP, the additional explanatory variables
included, and the estimation procedure used.
Table 5.4: Economy-wide Returns to R&D in Australia
Estimation
Assumptions
Additional variables
method
–
58%
16%
education, time trend
25%
8%
–
87%
23%
education, time trend
43%
12%
capital, labour, education, time
trend, interaction term
149%
27%
non-market sector MFP = 0
Two-step
method
non-market = market sector MFP
Production
function
Return on R&D
Domestic Foreign
constant returns to scale
assumption relaxed
Source: Industry Commission (1995), Research and Development, report no. 44, AGPS, Canberra, Appendix B
Intuitively, the returns to the domestic and foreign R&D were higher when non-market
sector MFP was set equal to market sector MFP, since this gave a higher economy-wide
estimate of MFP. In either case, adding other variables, such as education and a time trend
(often taken to reflect growth in disembodied knowledge) lowered the returns to domestic
R&D (to 25-43%) and foreign R&D (to 8-12%). This led the Industry Commission (1995,
Appendix QB, p. 19) to argue that ‘failure to account for other variables explaining
productivity may bias upwards the estimated returns to R&D’ and could be one reason for
the higher returns found in Coe and Helpman (1993), Dowrick (1994) and Rogers (1995).
– 125 –
Productivity and Regional Economic Performance in Australia
However, the Industry Commission found much higher returns to domestic R&D (149%)
and foreign R&D (27%) when using the production function approach. Two points help to
explain this result. First, the estimated labour and capital contributions were higher under
this production approach than under the Törnqvist approach, resulting in a higher residual
MFP estimate. Second, the production function equation contained a positive coefficient on
an interaction term between domestic and foreign R&D. The Industry Commission
interpreted this as indicating that the overall return to domestic R&D was higher than the
direct return, since own R&D also provided a greater understanding of, and therefore
spillover from, the pool of foreign R&D available, a finding consistent with growing
empirical evidence on the complementarity between domestic and foreign knowledge
stocks (see Jaffe, 1986; Cohen and Levinthal, 1989; and OECD, 2001, p. 55).
Returns to R&D in Australian states
A state level econometric analysis of MFP gains due to R&D provides further explanations
for interstate MFP trends. For instance, the empirical analysis finds evidence of interstate
R&D spillovers and some equalisation in the returns to R&D across the states, consistent
with convergence, rather than divergence, in MFP. In particular, the returns to business R&D
seem to have been highest, but have fallen most, in Queensland and Western Australia. This
suggests that these states faced the highest opportunities to profit from R&D at the start of
the period and thus invested most heavily in R&D, allowing them to converge on the levels
of MFP enjoyed in New South Wales and Victoria relatively faster than other states.
A panel regression of the general form of (9) was used to estimate the returns to business
R&D across Australian states, with the log of MFP for each state regressed on the domestic
business R&D stock (RD), the business R&D stock existing in the rest of Australia (RIS), an
import-weighted foreign business R&D stock (RF), state-specific constants (_i) to account
for unobserved factors (fixed effects) that may influence interstate differences in MFP
levels, and other variables (XYZ) that may explain MFP over time in each state:
logMFPit = _i + `1logRDit + `2logRISit + (`3 · mit) · logRFit + `xyz · XYZit + ¡it
(9)
Importantly, the coefficients `2 and `3·mit in (9) measure the extent of any interstate R&D
spillovers or international R&D spillovers respectively. For each state, the foreign business
R&D stock was calculated by weighting together the business R&D stocks of the G7
countries by each country’s share of total imports into that state. As in Coe and Helpman
(1993), this stock was then multiplied by each state’s share of imports in GDP (mi) to test
whether states with greater trade exposure enjoyed a higher return on MFP from foreign
R&D. For each state, the R&D stock in the rest of Australia was calculated as a simple sum
of business R&D stocks in all other states. In this case, a lack of official data on import
flows between states meant a weighted R&D stock for the rest of Australia could not be
calculated.
The inclusion of other variables (XYZ) in (9) reflects the fact that innovation is only one of
a number of factors that may influence MFP. Growth in MFP is driven by rises in efficiency
(making better use of existing technology) and technological progress itself. Efficiency is
influenced by factors such as the extent of competition, openness to trade, labour market
flexibility and macroeconomic stability, while innovation and human capital are seen as the
major drivers of technological progress.
– 126 –
Multifactor Productivity and Innovation in Australia and its States
As noted in the previous section, excluding these other factors from (9) may inappropriately
bias upward the estimated contribution to MFP of the included R&D variables. Thus,
several other variables were experimented with in (9), including tariff rates to account for
greater trade openness and competitive pressures since the early 1980s, rates of industrial
disputation to account for labour market reform over the period 1984-85 to 1999-2000, high
school retention rates to capture the rise in the potential human capital stock over this
period, and a capacity utilisation variable to account for the cyclical swings in MFP (see
Appendix 5B for data sources and construction).
The panel regression in (9) was estimated across six states (i = 6) over 16 years (t = 1,…,
16). The period only relates to 1984-85 to 1999-2000, since R&D data at the time of writing
were not available for 2000-01. Also, the sixth state in the panel was calculated as a region
comprising Tasmania, the Australian Capital Territory and the Northern Territory, since
Australian business R&D data over the period are only consistently available for the sum of
these three regions.
In estimating (9), the domestic R&D stock, rest of Australia R&D stock, tariff rate, rate of
industrial disputation and capacity utilisation were all found to be statistically significant.
However, foreign R&D and the high school retention rate were insignificant. One
explanation for the insignificance of foreign R&D and the retention rate is that these series
were highly correlated with the state and rest of Australia R&D stocks included in the
model. In such cases, the regression is unable to separate out the individual influences of
each variable, causing some to appear statistically insignificant despite their economic
significance (Griliches, 1979, p. 94).
This represents a common problem in empirically modelling MFP, since its determinants
are highly interdependent. For example, the previous section cited growing empirical
evidence of complementarity between domestic and foreign R&D. Complementarity in this
study would suggest that some R&D in each state is conducted in order to absorb,
implement and improve upon innovations emanating from foreign R&D activity, helping
explain the high correlation between each state’s domestic R&D stock and the foreign R&D
stock. Similarly, it is the human capital stock, embodying the analytical and creative skills
of society, that largely determines the rate of innovation, helping to explain the correlation
between R&D and the high school retention rate series used in this model. Thus, the
statistical significance of included R&D stocks in this study may indirectly reflect the
economic importance of many other factors in driving a faster rate of innovation since the
mid 1980s, including greater foreign R&D intensity, rising human capital and greater
domestic and international competition.
Table 5.5 reports the aggregate empirical results. The coefficient on a state’s own R&D
stock indicates that, on average across Australia, a 1% increase in a state’s own domestic
business R&D stock will raise MFP by 0.056% in that state, similar to the elasticity of 0.055
found in Coe and Helpman (1993) and that of 0.044 found in Rogers’ (1995) Australian
study. Note that one reason for the similarity of the elasticity to business R&D in this study
to the elasticity to total R&D in other studies is that any upward bias on the current
coefficient as a result of underestimating total R&D (by omitting R&D conducted by
universities, government and non-profit agencies) may be offset by a downward bias caused
by the fact that these omitted sectors conduct R&D in areas with greater spillovers (social
returns) than business R&D.
– 127 –
Productivity and Regional Economic Performance in Australia
Table 5.5: Econometric Results: Multifactor Productivity (logMFP), a b
1984-85 to 1999-2000
Variable
Coefficient
Standard error
State R&D stock
Rest of Australia R&D stock
Import tariff
Rate of industrial disputation
Capacity utilisation
0.056***
0.039**
-0.555*
-0.009*
0.691***
0.014
0.018
0.098
0.313
0.005
Diagnostics:
R2
Levin and Lin (1992)
Error correction term
0.963
-5.441***
-5.817***
–
–
–
t-statistic
3.912
2.159
-1.767
-1.857
7.059
–
–
–
a
The terms *, ** and *** denote significance at the 90%, 95% and 99% confidence levels respectively.
b
The equation was estimated in levels (rather than in rate of change form) in order to determine the significance of any
long-run elasticity and thus the rate of return on R&D. However, various diagnostic tests indicate that the equation forms
a cointegrating relation. For instance, results from applying the Levin and Lin (1992) panel unit root tests indicate that
the null of non-stationary residuals can be rejected at the 99% level of confidence, while the coefficient on these
residuals lagged once as an error correction term in a short-run differenced equation was also significant at the 99%
confidence level.
Most importantly though, the positive and significant coefficient on the rest of Australia
R&D stock provides some evidence of interstate R&D spillovers. In this case, a 1% rise in
the R&D stock in the rest of Australia will on average raise MFP by 0.039% in each state.
However, this result should be taken in light of the possible limitations in interpreting R&D
at the state level. For instance, this result may arise in situations in which a company
undertakes R&D in a subsidiary in one particular state and then makes any successful
innovation outputs easily available to subsidiaries in other states. This does not represent a
true spillover, since the costs and benefits of the initial R&D are internalised by the same
organisation.
Nevertheless, this result is most likely indicative of interstate R&D spillovers in the truest
sense, since a crucial determinant of the size of any spillover is the ‘technological
proximity’ or usefulness of the external R&D to the organisation or region of interest.
Technological proximity is likely to be greatest among regional economies that are
geographically close, have similar industrial structures and face similar federally driven
reforms. The finding of interstate R&D spillovers is also consistent with the US study by
Jaffe et al. (1993), which found that spillovers were initially geographically localised, even
after accounting for similar industrial structures between close regions. They found that
patent citations were twice as likely to come from the same state (country) as the cited
patent than would be expected based on the pre-existing concentration of technological
activity in that state (country), as spillovers initially spread across regions before spreading
across countries.
– 128 –
Multifactor Productivity and Innovation in Australia and its States
The empirical results also provide support for MFP gains due to the microeconomic reforms
since the mid 1980s. In particular, Table 5.5 illustrates that a 1 percentage point fall in the
tariff rate in each state would on average raise MFP by 0.56% in each state. This result can
be compared to Chand (1999) who found that a 1% fall in the nominal rate of assistance led
to a rise of 0.18-0.51 percentage point in annual MFP growth for a panel of eight Australian
manufacturing industries over the period 1968-69 to 1994-95. Table 5.5 shows that the
effect of a 1% fall in the rate of industrial disputation is relatively smaller, resulting in a
0.01% rise in MFP on average across the states.
The economy-wide coefficients on the R&D stocks in Table 5.5 can be multiplied by the
output to R&D stock ratio in each state in order to compute rates of return to own state and
interstate R&D. Figure 5.8a illustrates the average rate of return to domestic R&D and the
average interstate R&D spillover for Australia as a whole. Figure 5.8b shows the rate of
return to domestic R&D in each state. Both figures compare the returns to R&D in
1999-2000 with that in 1990-91, as this corresponds to the year in which returns to R&D
were calculated in earlier Australian studies. Comparison over a ten-year period also
provides insights into changes in the return to R&D over time.
Figure 5.8a shows that the Australian average rate of return to business R&D in 1990-91
was 173%, similar to returns to total R&D of around 196% implied in Coe and Helpman’s
(1993) results, 166% in Dowrick (1994) and an upper estimate of around 147% calculated
by the Industry Commission (1995). The current estimate indicates that a $100 rise in the
business R&D stock of any state would have on have average raised income by $173 in that
state in 1990-91. The estimate of an average interstate R&D spillover of 23% in 1990-91
implies that a state’s income would increase by $23 for a $100 rise in business R&D stock
in the rest of Australia.
Figure 5.8: Returns to R&D in Australia and its States
(a) Australian average
Domestic return to R&D
Interstate R&D spillover
200
180
173.4
160
Per cent
140
116.4
120
100
80
60
40
23.8
20
16.0
0
1990-91
1999-00
– 129 –
Productivity and Regional Economic Performance in Australia
(b) Interstate returns to domestic R&D
1990-91
1999-00
400
361
350
337
308
Per cent
300
250
232
201
200
214
160
150
133
114
134
113
100
81
50
0
NSW
Vic
Qld
SA
WA
Other
However, the rate of return to domestic R&D fell over the period 1990-91 to 1999-2000,
with the Australian average rate of return falling from 173% to 116%. This reflects the fact
that the business R&D stock has been growing faster than output, causing some diminishing
returns to R&D as it comprises a larger share of total inputs into the production process. As
stated by the Industry Commission (1995, p. 191):
A standard observation about investment more generally is that, at any point
in time, expected returns to investors decline as investment increases. This is
also so for investments in R&D, and is no more than common sense, as the
most productive opportunities are exhausted first.
This fall in the return to business R&D is consistent with the fall in R&D potency (ratio of
patents to business expenditure on R&D shown earlier in Figure 5.7a, but also bears a
similar qualification. While the current estimates suggest the return to business R&D has
fallen, this does not imply that the return to total R&D has also fallen over the period.
Business R&D in Australia has generally grown faster than R&D in other sectors and has
thus grown faster than total R&D over the period. This means that the fall in the output to
business R&D stock ratio measured here would be greater than any fall in the output to total
R&D stock ratio over the period. Thus, the fall in the return to business R&D shown here
is also likely to overestimate the fall, if any, in the return to total R&D over the period,
holding everything else constant.
However, Figures 5.8a and 5.8b illustrate three important points that help explain interstate
trends in MFP experienced over the period 1984-85 to 1999-2000. First, the existence of
interstate R&D spillovers will themselves act against any process of divergence in MFP
across states, as R&D in any one state will benefit MFP in all other Australian states.
Second, the rates of return to domestic R&D across the five major states appear more
similar in 1999-2000 than 1990-91, with the standard deviation between the returns falling
from 261.8 in 1984-85 to 43.7 in 1999-2000. This evidence of equalisation in returns to
R&D is also consistent with the evidence of some convergence in MFP across the five states
– 130 –
Multifactor Productivity and Innovation in Australia and its States
over the period. Finally, and perhaps most importantly, the returns to domestic R&D were
highest, but have fallen most, in Queensland and Western Australia. These states recorded
the highest returns to business R&D in 1990-91, at 361% and 337% respectively, with these
returns falling to 201% and 134% in 1999-2000. This suggests that these states faced greater
opportunities to profit from R&D in the late 1980s and subsequently invested most heavily
in R&D over the period to 1999-2000, allowing their levels of MFP to show some signs of
convergence toward that in the larger states of New South Wales and Victoria.
With Queensland and Western Australia facing higher returns, business R&D in these two
states added around twice as much to MFP growth than in other Australian states. This helps
explain why these two states have recorded above average growth in MFP relative to the
convergence hypothesis, compared with the below average MFP growth in the states of
South Australia and Tasmania. This is illustrated for South Australia in Figure 5.9. It
combines the average economy-wide coefficients in Table 5.5 with the variation in the R&D
stocks, tariffs rates and industrial disputation rates between individual states, in order to
decompose MFP growth in each state into the relative contributions of each of these factors.
The figure shows that domestic R&D accounted for around 0.7-0.8 percentage point of
annual MFP growth in Queensland and Western Australia, nearly twice its contribution in
South Australia.
Figure 5.9: The Contribution of R&D to MFP across Australian States,
average annual growth, 1985-86 to 1999-2000
Domestic R&D
Interstate R&D
Tariffs
Labour market
Capacity/residual
1.8
1.6
1.4
Per cent
1.2
1.0
0.8
0.6
0.4
0.2
-0.0
-0.2
NSW
Vic
Qld
– 131 –
SA
WA
Other
Productivity and Regional Economic Performance in Australia
Implications for R&D Policy and Future Research
The above econometric results help summarise the importance of R&D to interstate trends
in MFP growth, economic growth and material living standards. Growth in business R&D
has comprised around 75% of annual MFP growth on average across the states, when own
R&D and interstate R&D spillovers are taken together. Note that MFP growth was
estimated earlier in the chapter to account for about one-third of economic growth and
around 55% of growth in real per capita income on average across the states. Combining
these two results suggests that growth in business R&D has driven around one-quarter of
economic growth and around two-fifths of growth in per capita real incomes across the
states over the period 1985-86 to 1999-2000.5
While this chapter has primarily been a fact finding exercise, the aggregate level results do
suggest some general implications for a number of states, relating to R&D policy as a
source of promoting productivity growth, economic growth and higher living standards.
The results also point to important avenues for future research, which in turn may yield
more detailed insights into R&D policy at the state level.
Some general policy implications can be given regarding the best and worst performing
states in terms of MFP growth. For instance, Queensland recorded growth in MFP at rates
above that expected from convergence dynamics alone, underpinned by relatively faster
growth in business R&D. However, Queensland’s level of MFP, along with its R&D
intensity, remains below that in the larger states of New South Wales and Victoria. Clearly,
there is no room for complacency, as Queensland has further ground to catch up.
Convergence in the future is not assured, with past convergence itself partly the result of the
initially higher potential returns to R&D in this state. Policies in Queensland will need to
adapt to, and capitalise on, threats and opportunities inherent in a rapidly changing
technological environment and remove any structural impediments to this process, in order
to increase this state’s R&D intensity, level of MFP and level of per capita income toward
that in higher income states.
In fact, a number of policies implemented by the Queensland state government provide a
good example of a proactive approach to addressing threats and opportunities emerging in
the business R&D sector. For example, current issues include the greater reliance of current
areas of technological opportunity on the basic research findings of public sector
organisations and universities (OECD, 2001) and the fact that globalisation may result in
specialisation-induced low growth if not linked in with appropriate education policies in
resource abundant regions (Dowrick, 1997). In this respect, the Queensland government has
implemented several policies promoting public–private partnerships in emerging areas of
technological opportunity, such as biotechnology and information and communications
technology, in order to ensure that discoveries and inventions in the public sector are
commercialised and create improvements in living standards in the broader economy. The
Queensland government has also proposed several reforms to the education system,
5
In one sense, this static analysis may underestimate the contribution of business R&D to economic growth and living standards,
as it ignores the indirect effect of R&D-induced rises in productivity on other factors, such as jobs growth, investment and
capital deepening. Yet, as with any econometric exercise, the contribution of R&D to MFP growth may also be biased upward if
the included business R&D stocks are indirectly incorporating the influence of other relevant, but omitted, variables. However,
generously accounting for such biases would still leave a sizeable contribution from business R&D to MFP growth and thus a
sizeable contribution to economic growth and improvements in living standards across the states over the period.
– 132 –
Multifactor Productivity and Innovation in Australia and its States
including a full-time preparatory year of schooling and policies to significantly raise high
school completion rates, helping Queensland develop its human capital base in order to take
advantage of technological opportunities and avoid the low growth trap. Indeed, several
chapters in this volume discuss in more detail the importance of state education policies for
innovation and growth.
The empirical results also suggest that a state such as Tasmania may have faced several
impediments to growth in business R&D and thus productivity and economic growth over
the period 1985-86 to 1999-2000. Tasmania appears to be the worst performer in terms of
convergence dynamics, recording a rate of MFP growth well below that expected based on
its initial level of MFP. Figure 5.8b shows that the ‘other’ category, which includes
Tasmania, initially faced returns to business R&D similar to that in Queensland and Western
Australia, but did not record the resulting growth in business R&D that these other states
achieved (see Figure 5.5a). This suggests that various structural impediments in Tasmania
have prevented it from capitalising to the same extent as Queensland and Western Australia
on initially high returns to R&D. This conclusion must be qualified by the fact that the
Australian Capital Territory and the Northern Territory are also included in this ‘other’
category. However, the human capital stock may represent one impediment to R&D growth,
with Draca, Foster and Green in Chapter 7 of this volume illustrating that Tasmania ranks
lowest in terms of educational attainment levels, for instance. This state may also suffer
disadvantages related to its ability to import and absorb knowledge from other regions, due
to its geographically more remote position relative to other Australia states.
Clearly, more detailed insights into the implications for R&D policy across individual states
require a further layer of analysis. That is, while this chapter has attempted to empirically
explain interstate MFP trends by variation in business sector R&D across the states, an
important avenue for future research is to, in turn, explain differences in business sector
R&D growth across the states by interstate variation in a variety of factors thought to drive
R&D spending in the private sector. This is particularly important given this study has
shown that variation in potential R&D returns across the states does not alone account for
subsequent differences in interstate R&D growth, with some states appearing to face greater
impediments to capitalising on potential returns relative to others. A better understanding of
other factors driving differences in R&D growth would help guide R&D and productivityrelated policy at the state level.
Such an econometric analysis could attempt to incorporate various possible explanations for
interstate variation in business R&D growth, including differences in human capital, openness
to trade, geographical proximity, domestic competition, labour market flexibility and
industrial structure. As previously discussed, the level of human capital is a primary driver of
the extent of innovation. There is also growing evidence that greater competition provides
added incentive to firms to innovate in order to obtain a competitive advantage over their
rivals. Similarly, greater labour market mobility allows disembodied knowledge to be diffused
throughout the economy more rapidly, with mobility between the public and private sectors
particularly important, given the growing importance of public sector basic research to private
sector innovation. Finally, interstate differences in industrial structure may also impact on
R&D growth, with some industries having higher propensities to invest in R&D relative to
other industries. Such an empirical analysis could also incorporate an evaluation of the impact
on R&D spending of various federal and state-based policies already in place.
– 133 –
Productivity and Regional Economic Performance in Australia
Concluding Remarks
This chapter has highlighted both the importance of MFP as a source of interstate
differences in economic growth and real per capita incomes since the mid 1980s and the
major role innovation activity has played in explaining these interstate trends in MFP and
economic performance. Queensland and Western Australia, which recorded the strongest
annual economic growth over the period 1985-86 to 2000-01, also enjoyed the highest
growth in MFP, with MFP growth accounting for around 20-45% of economic growth
across the states. As a result, Queensland and Western Australia also recorded convergence
in their levels of MFP toward that in leading income states, namely New South Wales and
Victoria, at a relatively faster rate than South Australia and Tasmania. However, while MFP
growth accounted for over half of the growth in real per capita income across the states over
the period 1985-86 to 2000-01, convergence in MFP has not translated into overall
convergence in per capita incomes, due to significant differences in the rate of capital
deepening across the states over the period.
Interstate trends in business R&D activity appear to shed considerable light on differences
in MFP growth at the state level. An econometric analysis indicates that growth in business
R&D comprised up to 75% of annual MFP growth across the states over the period
1985-86 to 1999-2000. The results also provide some evidence of interstate R&D spillovers
and equalisation in the returns to domestic R&D across the Australian states, both consistent
with a pattern of convergence rather than divergence in MFP. In particular, the returns to
business R&D seem to have been highest, but have fallen most, in Queensland and Western
Australia. This result suggests that these states initially faced the greatest opportunities to
profit from R&D and thus invested most heavily in R&D, causing their MFP levels to
converge toward those in New South Wales and Victoria relatively faster than did those for
South Australia and Tasmania.
These findings point to some general implications for R&D and related productivity policy
at the state level. Queensland provides a good case study. This state has generated growth
in MFP at a rate above that expected from convergence dynamics alone, driven by relatively
faster growth in business R&D, allowing it to, in turn, record above average economic
growth. Yet, Queensland’s level of MFP, along with its R&D intensity, remains below that
in the larger states of New South Wales and Victoria. Past convergence itself has partly been
the result of the initially higher returns to R&D in Queensland, suggesting policies in this
state will need to continue to adapt to, and capitalise upon, threats and opportunities
inherent in a changing technological environment and remove structural impediments to
this process, in order to raise this state’s R&D intensity, level of MFP and per capita income
level toward that in higher income states. Further, ensuring rates of capital deepening
comparable with other states will be crucial for convergence in Queensland’s level of per
capita income with higher income states. Put simply, relatively stronger population and
employment growth have caused Queensland to record a rate of capital deepening below
that in other states on average over the period 1985-86 to 2000-01. This highlights a
challenge for investment policy in Queensland, with the state requiring relatively faster
growth in its capital stock if it is to record rates of capital deepening similar to those in the
rest of Australia.
– 134 –
Multifactor Productivity and Innovation in Australia and its States
This chapter has primarily been a fact finding exercise, given the little attention paid to
interstate MFP and its determinants in the past. However, it does suggest several directions
for future research. At the empirical level, the difficulty in gaining precise state level data in
order to calculate MFP and examine its determinants highlights the need for progress in this
area. Further work in constructing annual data for other sectors performing R&D would also
allow an analysis into the returns to total R&D, rather than the returns to business R&D only.
These extensions will also test the robustness of the results obtained in the current study.
However, perhaps the most promising avenue for future research is to attempt to empirically
explain differences in business sector R&D growth across the states by interstate variation
in the factors thought to underpin R&D spending in the private sector, given the large share
of growth in MFP, economic activity and per capita incomes explained by business R&D.
A better understanding of the relative contribution of factors, such as human capital,
openness to trade, exposure to competition, labour market flexibility and industrial
structure, in driving interstate differences in business sector R&D intensity may provide
more detailed insights into how individual states can tailor their innovation policies in order
to improve productivity growth and regional economic performance in the future.
– 135 –
Productivity and Regional Economic Performance in Australia
References
Australian Bureau of Statistics (ABS) (1968), System of National Accounts, see Australian
System of National Accounts, Cat. no. 5204.0, Canberra.
ABS (1993), Technical Note, Australian National Accounts, Cat. no. 5206.0, September
quarter, Canberra, 112-114.
ABS (1998), Australian Standard Research Classification, Cat. no. 1297.0, Canberra.
ABS (2000), Australian System of National Accounts: Concepts, Sources and Methods,
Cat. no. 5204.0, Canberra.
ABS (various issues), Australian Demographic Statistics, Cat. no. 3101.0, Canberra.
ABS (various issues), Australian National Accounts, Cat. no. 5204.0, Canberra.
ABS (various issues), Australian National Accounts: State Accounts, Cat. no. 5220.0,
Canberra.
ABS (various issues), Australian System of National Accounts, Cat. no. 5204.0, Canberra.
ABS (various issues), Industrial Disputes, Australia, Cat. no. 6321.0, Canberra.
ABS (various issues), Labour Force, Australia, Cat. no. 6202.0, Canberra.
ABS (various issues), Research and Experimental Development, Businesses, Australia,
Cat. no. 8104.0, Canberra.
ABS (various issues), Schools, Australia, Cat. no. 4221.0, Canberra.
Baumol, W.J., Blinder, A.S., Gunther, A.W. and Hicks, J.R.L. (1988), Economics
Principles and Policy, Australian Edition, Harcourt Brace Jovanovich, Marrickville,
Sydney.
Chand, S. (1999), ‘Trade liberalisation and productivity growth: Time-series evidence
from Australian manufacturing’, Economic Record, 75 (228), March, 28-36.
Coe, D. and Helpman, E. (1993), ‘International R&D spillovers’, IMF Working Paper No.
84, European Economic Review (1995), 39 (5), 859-887.
Coelli, T. (1998), ‘Discussant on construction and use of measures to guide
microeconomic reform’, Microeconomic Reform and Productivity Growth, 35-40.
Workshop Proceedings, Canberra, 26-27 February, Productivity Commission and
Australian National University, AusInfo, Canberra.
Cohen, W.M. and Levinthal, D.A. (1989), ‘Innovation and learning: The two faces of
R&D’, The Economic Journal, 99 (397), September, 569-597.
Crosby, M. (2000), ‘Patents, innovation and growth’, The Economic Record, 79 (234),
September, 255-262.
– 136 –
Multifactor Productivity and Innovation in Australia and its States
Dowrick, S. (1990), ‘Explaining the labour productivity slowdown of the 1980s’,
Australian Bulletin of Labour, 174-199.
Dowrick, S. (1994), ‘The Role of R&D in Growth: A Survey of the New Growth Theory
and Evidence’, Paper commissioned by the Industry Commission (1995), Research and
Development Report No. 44, AGPS, Canberra.
Dowrick, S. (1995), ‘The determinants of long-run growth’, in P. Anderson, J. Dwyer and
D. Gruen (eds), Productivity and Growth, Proceedings of a Conference, Kirribilli,
10-11 July, Reserve Bank of Australia, Sydney.
Dowrick, S. (1997), ‘Openness and growth’, in P. Lowe and J. Dwyer (eds), International
Integration of the Australia Economy, 9-41, Proceedings of a Conference, Kirribilli,
11-12 July 1994, Reserve Bank of Australia, Sydney.
Dowrick, S. and Nguyen, D. (1989), ‘OECD comparative economic growth 1950-85:
Catch-up and convergence’, American Economic Review, 79 (5), 1010-1030.
Eglander, A.S., Evenson, R. and Hanazaki, M. (1988), ‘R&D, innovation and the total
factor productivity slowdown’, OECD Economic Studies, 11, Autumn, 8-43.
Geroski, P. (1994), ‘Do spillovers undermine the incentive to innovate?’, in Industry
Economics Conference Papers & Proceedings 9-29, 7-8 July, Bureau of Industry
Economics and Australian National University, AGPS, Canberra.
Gordon, R.J. (2000), ‘Does the “New Economy” measure up to the great inventions of the
past?’, Journal of Economic Perspectives, 14 (4), Fall, 49-74.
Griliches, Z. (1979), ‘Issues in assessing the contribution of research and development to
productivity growth’, Bell Journal of Economics, 10 (1), Spring, 92-116.
Griliches, Z. (1980), ‘R&D and the productivity slowdown’, American Economic Review,
70 (2), May, 343-348.
Griliches, Z. (1990), ‘Patent statistics as economic indicators: A survey’, Journal of
Economic Literature, 28, December, 1661-1707.
Griliches, Z. (1994), ‘Productivity, R&D, and the data constraint’, American Economic
Review, 84 (1), March, 1-23.
Griliches, Z. (1998), R&D and Productivity: The Econometric Evidence, University of
Chicago Press, Chicago.
Henry, K. (2001), Australia’s Economic Development, address to the Committee for the
Economic Development of Australia (CEDA), 40th Anniversary Annual General
Meeting Dinner, 19 November, Sydney.
Industry Commission (1995), Research and Development, Report No. 44, AGPS,
Canberra.
– 137 –
Productivity and Regional Economic Performance in Australia
Industry Commission (1997), Assessing Australia’s Productivity Performance, AGPS,
Canberra.
Intellectual Property Australia (1998), Industrial Property: IP Activity in Australia and the
Asia-Pacific Region, Commonwealth of Australia, Canberra.
Jaffe, A. (1986), ‘Technological opportunity and spillovers of R&D: Evidence from firms’
patents, profits, and market value’, American Economic Review, 76 (5), 984–1001.
Jaffe, A., Trajtenberg, M. and Henderson, R. (1993), ‘Geographic localisation of
knowledge spillovers as evidenced by patent citations’, Quarterly Journal of
Economics, August, 577-598.
Levin, A. and Lin, C.F. (1992), ‘Unit Root Tests in Panel Data: Asymptotic and FiniteSample Properties’, Discussion Paper 92-23, University of California San Diego.
Lowe, P. (1995), ‘Labour productivity growth and relative wages: 1978-1994’, in P.
Anderson, J. Dwyer and D. Gruen (eds), Productivity and Growth, Proceedings of a
Conference, Kirribilli, 10–11 July, Reserve Bank of Australia, Sydney.
Morrison, C. (1998), ‘Construction and use of measures to guide microeconomic reform’,
in Microeconomic Reform and Productivity Growth, 13-34, Workshop Proceedings,
Canberra, 26-27 February, Productivity Commission and Australian National
University, AusInfo, Canberra.
OECD (2001), Science, Technology and Industry Outlook: Drivers of Growth:
Information Technology, Innovation and Entrepreneurship, OECD Publications, Paris.
Parham, D., Roberts, P. and Sun, H. (2001), Information Technology and Australia’s
Productivity Surge, Productivity Commission Staff Research Paper, AusInfo, Canberra.
Productivity Commission (1999), Microeconomic Reform and Australian Productivity:
Exploring the Links, Commission Research Paper, AusInfo, Canberra.
Productivity Commission and the Australian National University (1998), Microeconomic
Reform and Productivity Growth, Workshop Proceedings, AusInfo, Canberra.
Rogers, M. (1995), ‘International knowledge spillovers: A cross-country study’, in S.
Dowrick (ed.), Economic Approaches to Innovation, 166-188, Aldershot, England.
Sala-i-Martin, X. (1996), ‘Regional cohesion: Evidence and theories of regional growth
and convergence’, European Economic Review, 40 (6), 1325-1352.
Schankerman, M. and Pakes, A. (1986), ‘Estimates of the value of patent rights in
European countries during the post-1950s period’, The Economic Journal, 96 (384),
December, 1052-1077.
Solow, R. (1956), ‘A contribution to the theory of economic growth’, The Quarterly
Journal of Economics, 70, 65-94.
– 138 –
6 Recent Convergence Behaviour of the
Australian States: A Time Series
Approach
Philip Bodman, Mirko Draca and Phillip Wild
Summary
The concept of convergence is important to the analysis and measurement of economic
growth. A key implication of the neoclassical growth model is that there is a tendency for
incomes per capita across economies to equalise in the long run. This tendency is a result
of increases in the rate of investment in poorer economies (capital deepening) and technical
change.
Empirically, the extent of convergence (or divergence) is important because it gives us an
indication of how economies are growing relative to other economies and what may be
influencing the relative growth rates. For example, if an economy shows little sign of
converging with similar economies over time, this may indicate structural impediments to
convergence in the economy.
Typically, convergence hypotheses have been tested using cross-sectional econometric
methods. These methods focus on the average behaviour of a group of economies over time.
Specifically, they examine how fast economies have grown compared with their initial income
levels. For example, if the principles underlying the neoclassical growth framework hold,
then in a given group of economies the initially poorer members of the group are expected to
grow faster than the wealthier economies, an effect known as ‘catching up’.
In this chapter we use an alternative approach based on time series methods. This approach
has the advantage of being able to discriminate between the growth behaviour of individual
pairs of economies within a group. This time series approach focuses on the long-run
behaviour of differences in per capita income. If these differences in per capita income are
found to be transitory rather than permanent then it can be concluded that growth is
consistent with a process of long-run convergence.
We implement time series tests of convergence for the Australian states for the period 1985 to
2000. In contrast to cross-sectional methods (which show that the dispersion of incomes
between states has been increasing), our results indicate that some evidence of time series
convergence behaviour can be observed for each state. This indicates that while the economic
changes experienced over the period have been very intense, they have not shifted the
underlying convergence behaviour of economic growth at the interstate level.
The policy implications for Queensland relate to the state’s future economic performance.
Our findings indicate that it is possible that Queensland has reached a steady-state position
in terms of its level of gross state product (GSP) per capita relative to other states. While
– 139 –
Productivity and Regional Economic Performance in Australia
the gap between Queensland’s GSP per capita and other states like New South Wales has
not been fully eliminated over time, it would seem that the neoclassical forces that would
allow for further catching up and ultimate total elimination of such gaps have been largely
exhausted. That is, there may be less scope in the future for Queensland to benefit from the
effects of catching up that are associated with capital deepening and incremental technical
changes, and policy makers may need to induce changes to the underlying structure of the
economy if further gains in relative GSP per capita are to occur.
These changes can take a variety of forms including further microeconomic reform and/or
investments in human capital and the state economy’s capacity for innovation. Further
catching up also depends on the behaviour of other states – if they grow faster or introduce
parallel policies to accelerate economic growth it will be harder for Queensland to catch up
with them in terms of its relative levels of GSP per capita and GSP per person employed.
Introduction
The concept of convergence is at the centre of modern debates on the nature of economic
growth. Solow’s (1956) influence here is well recognised.1 He originated a neoclassical
growth model that made extensive predictions about the role of different inputs in economic
growth as well as the long-run relationship between economies’ income per capita levels
and growth rates over time. The convergence in per capita incomes between economies
(regardless of the initial level of income) in the neoclassical model reflects entirely the
operation of technical change and capital deepening across economies with given
specifications of technology and preferences. However, more recent developments in
growth theory initiated by Romer (1986) and Lucas (1988) challenge the notion that there
is an inherent tendency for the equalisation of incomes between growing economies in the
long run. In these models, scale economies and/or endogenous technical change provide the
grounds for differences in income levels and growth rates to persist indefinitely. Since the
late 1980s, endogenous growth theory has experienced a rapid evolution in its detail and
scope. This includes the emergence of novel neo-Schumpeterian theories of technological
change (Aghion and Howitt, 1998) and the development of a sophisticated empirical
literature on convergence and economic growth (Durlauf and Quah, 1999), as discussed in
previous chapters.
Empirical studies of regional economic growth in Australia have created a rich picture of
spatial income inequality at the state and substate level. Cashin (1995) examined the
behaviour of GSP per capita for the six Australian colonies/states and New Zealand over the
period 1861 to 1991. His study found that the dispersion of per capita income had been
shrinking for most of this period and that the low level of dispersion found in Australia was
comparable to that prevailing in countries such as the United States, Japan, France and the
United Kingdom. This contrasts with the higher levels of dispersion observed in countries
such as Spain, Germany, Italy and India. However, this result is qualified by Cashin’s (1995)
additional finding that the dispersion of GSP per capita across the Australian colonies/states
and New Zealand increased during the 1980s, a result consistent with the earlier findings of
Maxwell and Hite (1992).
1
As is that of Swan (1956).
– 140 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
Harris and Harris (1991) compared state GSP per capita between 1953-54 and 1990-91.
They concluded that the majority of interstate income disparities could be attributed to
state-specific economic shocks. More recently, Cashin and Strappazzon (1998) studied the
dispersion of per capita incomes between and within states using census data from 1976 to
1991. Their cross-sectional research confirmed that the dispersion of state per capita income
increased during the 1980s. The dispersion of substate regional incomes remained constant
and did not intensify for any particular state.
Similarly, the work of Nguyen et al. in Chapter 3 indicates that there has been a tendency
for the levels of GSP per capita to diverge over the past 15 years, although the exclusion of
mining results in a pattern of neither convergence nor divergence.
Clearly then, a common feature of these studies has been that they have suggested a
tendency towards regional divergence (non-convergence) since the late 1970s. Furthermore,
they have used cross-sectional methods to analyse convergence behaviour at the regional
level and, with the exception of Nguyen et el. (2003), have not so far considered the
developments in the 1990s. This chapter employs a time series approach to convergence to
analyse the evolution of state GSP per capita and GSP per person employed. The time series
approach has several theoretical advantages to cross-sectional methods and produces
complementary insights into the existing Australian literature on regional economic growth.
Utilising a time series perspective also lets us consider the implications of the latest theories
of economic convergence, particularly the idea of ‘convergence clubs’ and the role of
income distribution in economic growth.
This chapter is organised as follows. The first section reviews the time series and crosssectional approaches of convergence and gives the strict definitions for the hypotheses of
long-run convergence and catching up. The second section outlines how the time series
approach is operationalised in a framework for empirical testing, including tests for
structural change. The third section reports the results of these tests for GSP per capita and
GSP per person employed. The conclusion discusses options for extending these tests and
integrating the time series and cross-sectional approaches to convergence.
Interpreting Convergence
Bernard and Durlauf (1996) comprehensively address the interpretation of the neoclassical
growth model’s convergence hypothesis, focusing on the permanence of contemporaneous
output differences between economies. The neoclassical model predicts that output per
capita differences are transitory such that economies that are initially capital-poor grow
faster than capital-rich economies in a process of catching up. Intuitively, this theory
matches the experience of economies such as Japan and other East Asian countries in the
postwar era.
We cover the following issues in interpreting convergence: the distinction between crosssectional and time series tests, Bernard and Durlauf’s (1996) two definitions of
convergence, and details of the application of the time series approach.
– 141 –
Productivity and Regional Economic Performance in Australia
Cross-sectional versus time series tests
Empirical tests of convergence can be classified as following either a time-series or crosssectional approach. Cross-sectional approaches focus on the correlation between initial per
capita income levels and the subsequent per capita growth rates for a group of economies.
In this framework, as illustrated in Chapter 3, a negative correlation implies that, on
average, economies with lower levels of initial income have been growing faster and
catching up with high-income economies. Time series approaches examine the long-run
behaviour of the differences in per capita income between economies. Specifically, the time
series approach is based on the implication that long-run convergence means that shocks to
output differences will be transitory in nature only and therefore make them testable using
standard time series methods, such as tests for unit roots (stochastic trends).
Both approaches make different assumptions concerning the meaning of convergence and the
statistical properties of the data that are being studied. As a result, empirical research based
on these approaches have delivered contrasting results. Studies that have used cross-sectional
tests have generally rejected the null of no convergence. This has been the case for datasets
for the industrialised countries (Dowrick and Nguyen, 1989), the United States regions, the
Japanese prefectures, the countries of Europe and even for broader cross-country samples. In
contrast, research based on time series tests has generally failed to reject the null of no
convergence for a similar range of datasets (Quah, 1991; Bernard and Durlauf, 1995).
Defining convergence
The definition of convergence plays a major role in the operational differences exhibited in
the cross-sectional and time series approaches to empirical testing. Bernard and Durlauf
(1996) propose two definitions that are designed to capture the main implications of the
neoclassical growth model. These definitions describe the convergence between two
economies denoted as i and j respectively. Convergence in a set of I economies is
analogously defined for situations where every pair of economies in I exhibit convergence.
The first definition of convergence outlined by Bernard and Durlauf (1996) relates to the
difference in output between two economies over a fixed time interval. In this definition,
convergence corresponds with the tendency for the difference between economies to
narrow. Therefore, economies i and j converge between the dates t and t + T if there is an
expectation that the disparity in log GDP per capita will decrease.
E(yi,t+T - yj,t+T | ¼t) < yi,t - yj,t
(1)
where ¼t denotes all information available at t.
This definition represents convergence as catching up, since it is based on changes in the
expected difference in output between economies.
– 142 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
The second definition of convergence relates to the equality of long-term forecasts of output
at a fixed time. That is, two economies converge if the long-term forecasts of log per capita
output are equal at t, a fixed time. This definition is important because if the effects of a
shock to output differences can be expected to persist indefinitely then such shocks will be
impediments to convergence. This definition is represented as:
lim k‰' L (yi,t+T - yj,t+T | ¼t)=0
(2)
These two definitions are of course closely related, with definition (2) implying definition
(1) for some T if convergence is occurring. That is, a decrease in the disparity in log GDP
per capita is a natural concomitant of long-run convergence.
These formal definitions have an impact on how we can interpret the findings of different
cross-sectional and time series tests for convergence. Specifically, they allow us to identify
at least two caveats to the interpretation of cross-sectional tests of convergence that regress
growth rates on initial income for a selected group of economies. First, since the crosssectional test employs weighted averages of growth rates and initial incomes, it can
overstate the convergence dynamics operating within a given group of economies. That is,
cross-sectional tests are unable to distinguish between the pairs of economies that are
converging in a given group and those that are not. Second, cross-sectional tests do not
provide evidence of whether economies converge in line with the second definition of
convergence outlined above. This allows researchers to conclude that a group of economies
are converging when they could instead be making the transition to different steady states,
as defined by their initial conditions.
Time Series Convergence in Australia: 1985-2000
Implementing time series tests
Time series tests of convergence hinge on the statistical properties of differences in per
capita output between economies. Tests are employed to analyse the stationarity of these
differences over time such that (yi,t - yj,t) does contain a unit root and possibly a
deterministic trend. Intuitively, the presence of such components would be an indication
that shocks to the difference in output per capita between economies is persistent. In
contrast, when the bivariate difference (yi,t - yj,t) is stationary it can be concluded that output
differences are transitory and compatible with a hypothesis of economic convergence.
Empirically, this test is formulated as an Augmented Dickey-Fuller (ADF) type test2 with
the null hypothesis of a unit root:
yi,t - yj,t = µ + _(yi,t-1 - yj,t-1) + `t +
n
bk ¨(yi,t-k - yj,t-k ) + ¡t
Y
k=1
(3)
where yj,t is output in economy j in period t, µ is a constant drift term and t is a deterministic trend.
2
For a clear exposition of unit root tests, including the ADF test and the problems associated with unit root testing,
see Enders (1995).
– 143 –
Productivity and Regional Economic Performance in Australia
The format of this test lets us differentiate between two distinct versions of the convergence
hypothesis: catching up and long-run convergence. They are defined as follows:
• Catching up: This convergence hypothesis relates to the tendency for the difference in
per capita output to narrow over a period of time. Practically, this refers to economies
that are out of equilibrium over a fixed interval of time. The tendency for two
economies’ per capita output differences to decline means that the economies are
growing according to a process that is consistent with convergence. Statistically,
catching up implies the absence of a unit root in the (yi,t - yj,t) times series process but
admits the possibility of a significant non-zero time trend in the same process.
• Long-run convergence: This is the strong version of the convergence hypothesis and
relates to economies in long-run equilibrium. It is characterised by the absence of both
a unit root and a time trend in (yi,t - yj,t). This hypothesis implies that catching up has
been completed such that the economies in question have converged to their steady-state
levels of output.
This framework for testing and understanding convergence behaviour can be modified in a
number of ways to account for more complex types of convergence dynamics, particularly
those related to structural changes and discontinuities in output.
Testing convergence
In the following we analyse the time series convergence behaviour of the six Australian
state economies for the period 1985 to 2000. This period was marked by the implementation
of a series of intensive microeconomic reforms by state and federal governments in
Australia. These reforms included a range of measures such as labour market deregulation,
trade liberalisation, privatisation and the introduction of a national competition policy. As a
result, this period was characterised by many significant changes in all areas of the
Australian economy.
We analyse the behaviour of quarterly GSP per capita and GSP per person employed, which
are plotted in Figures 6.1a and 6.1b. On the basis of these figures the state economies can
be divided into two groups. First, there is the group comprised of Queensland, New South
Wales and Western Australia that have relatively smooth fluctuations in their recorded levels
of GSP per capita and per person employed. Second, the plots of GSP per capita and GSP
per person employed for Victoria, Tasmania and South Australia indicate that some
significant structural changes may have occurred. In particular, there is an interesting
relationship between the behaviour of GSP per capita and GSP per person employed in
Victoria. The recession of the early 1990s led to a sharper fall in GSP per capita in Victoria
than that experienced in other states (see Figure 6.1a).
– 144 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
Figure 6.1: Gross State Product
(a) Per capita, logarithms
NSW
Vic
Qld
SA
WA
Tas
4.10
4.05
4.00
3.95
3.90
3.85
3.80
3.75
Sep-85
Sep-88
Sep-91
Sep-94
Sep-97
Sep-00
(b) Per person employed, logarithms
NSW
Vic
Qld
SA
WA
Tas
4.30
4.25
4.20
4.15
4.10
4.05
4.00
Sep-85
Sep-88
Sep-91
Sep-94
Sep-97
Sep-00
Tables 6.1a and 6.1b rank the states according to their levels of GSP per capita and GSP per
person employed in 1985-86 and 1999-2000. The data indicate that New South Wales is the
leading state in terms of GSP per person employed. New South Wales also has the highest level
of GSP per capita in 1985-86 but comes second to Western Australia in 1999-2000.
Interestingly, Queensland improves its ranking in both GSP per capita and GSP per person
employed from sixth in 1985-86 to fourth in 1999-2000. This major shift is also reflected in the
final column, which indicates that Queensland’s cumulative growth in GSP per capita and in
GSP per person employed over the period 1985-86 to 1999-2000 was higher that all other states
except Western Australia.
– 145 –
Productivity and Regional Economic Performance in Australia
Table 6.1 Gross State Product
(a) Per capita
1985-86
1999-2000
GSPPCa
Ranking
$
1985-2000
Change
%
Ranking
State
GSPPCa
$
NSW
32,265
1
43,375
2
34.4
Vic
31,884
2
41,721
3
30.9
Qld
26,332
6
37,104
4
40.9
SA
28,990
4
34,452
5
18.8
WA
31,822
3
45,150
1
41.9
Tas
28,613
5
30,653
6
7.1
a
Gross state product per capita
(b) Per person employed
1985-86
Ranking
1999-2000
GSPPEa
Ranking
$
85-86 to 99-00
Change
%
State
GSPPEa
$
NSW
58,153
1
74,262
1
27.7
Vic
55,675
2
71,099
3
27.7
Qld
47,346
6
62,085
4
31.1
SA
52,378
5
61,631
5
17.7
WA
53,623
3
72,403
2
35.0
Tas
53,293
4
57,321
6
7.6
a
Gross state product per person employed.
Figures 6.2a and 6.2b show how these changes are reflected in the dispersion of GSP per
capita and GSP per person employed among the six states between 1985 and 2000. This is
a traditional indicator of convergence in the cross-sectional framework and is known as
sigma or m-convergence. The time series plot of m given in Figures 6.2a and 6.2b indicate
that the dispersion increased during this period and that the increase appears to be most
pronounced in the first half of the period, that is, the mid 1980s to the early 1990s.3 Overall,
this finding is consistent with the results of previous research using cross-sectional
approaches (Cashin, 1995; Maxwell and Hite, 1992) which found that the dispersion of state
incomes had been increasing since the 1970s. However, the slower rate of change for m later
in the period indicates that recent strong economic growth in Australia may have arrested
the increase in the dispersion of state income to some extent.
3
Cross-sectional beta or `-convergence models (after de la Fuente, 1997) were also estimated for a group of six states between
1985 and 2000. The results for these models were sensitive to the choice of starting dates and influenced by noise in the quarterly
data. The estimates of ` did not generally support the hypothesis of convergence. Full details are available from the authors on
request.
– 146 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
Figure 6.2: Dispersion of Gross State Product
(a) Per capita
Standard deviation GSPPC / GDP
0.18
Recession
1990:2–1991:4
0.16
0.14
0.12
0.10
0.08
0.06
Sep-85
Sep-88
Sep-91
Sep-94
Sep-97
Sep-00
Sep-97
Sep-00
(b) Per person employed
Standard deviation GSPPE / GDP
0.22
Recession
1990:2–1991:4
0.20
0.18
0.16
0.14
0.12
0.10
Sep-85
Sep-88
Sep-91
Sep-94
Source: Recession dates come from Bodman and Crosby (2002).
Empirically, the two hypotheses of catching up and long-run convergence are nested in the
ADF-type specification outlined in equation (3). Generally, the hypothesis of convergence
revolves around the presence of a unit root in the difference of output series (yt - y j). The
presence of a unit root (i.e. _ =1) implies that the output differences between the economies
cannot be considered to be transitory in the time series sense. However, if |_| < 1, the null
of no convergence can be rejected. Provided that the null of a unit root is rejected we can
then distinguish between catching up and long-run convergence by testing the significance
of the trend term, i.e. the hypothesis ` = 0.
– 147 –
Productivity and Regional Economic Performance in Australia
The results can therefore be summarised in terms of three cases related to equation (3):
• Long-run convergence: _ = 0; ` = 0
• Catching up: _ = 0; `  0
• No convergence: _ = 1; ` = 0,1
The third case is deliberately characterised as ‘no convergence’ rather than divergence. As
Bernard and Durlauf (1995, p. 99) point out, even if economies do not converge according
to the definitions given above, ‘they may still respond to the same long-run driving
processes, [that is] they face the same permanent shocks with different long-run weights’.
Practically, what this means is that an explicit process of economic divergence is not an
automatic implication of not rejecting various nulls of no convergence.
To operationalise the analysis, we employ the methodology developed by Dolado et al.
(1990) for the evaluation of deterministic regressors within the standard unit root testing
framework. Dolado et al. make the point that the presence of deterministic regressors (i.e.
trend and drift terms) influences the power of unit root tests. In particular, omitting a trend
where it is part of the data generating process imparts an upward bias to a unit root test. To
avoid these problems, we adopt the procedure of Dolado et al. of ‘testing down’ from a full
ADF model that includes trend and drift terms. Furthermore, this procedure assists in
accurately discriminating between the three convergence hypotheses outlined above.
A summary of the results obtained from implementing these tests are reported in Tables 6.2a
and 6.2b. The second column reports the final conclusion of the tests while the third column
reports the significance levels supporting these conclusions. Detailed results, including
calculated values, are reported in Appendix 6A. These tables report the results for the 14
possible pairwise combinations of GSP per capita and GSP per person employed among the
states. Bootstrapped critical values were calculated when various diagnostic tests recorded
the presence of non-normal residuals in the redundancy test for a deterministic trend.
The results reported in Tables 6.2a and 6.2b provide strong evidence of catching up and
long-run convergence among the Australian states for the period 1985 to 2000. The null of
no convergence is rejected for each pairwise combination of GSP per capita and GSP per
person employed. Interestingly, there is more evidence of long-run convergence in the case
of GSP per person employed than GSP per capita. Significant long-run convergence is
evident in 10 pairwise cases of GSP per person employed compared with seven cases for
GSP per capita. This suggests that levels of labour productivity across the states are
converging more strongly than per capita income.
Queensland stands out as a state whose growth path is consistent with traditional
convergence dynamics. Long-run convergence in GSP per capita and GSP per person
employed was found for every pairwise case except those involving Tasmania. In contrast to
Queensland, Tasmania’s performance does not seem to be consistent with the traditional
convergence dynamics where initially poor economies significantly catch up with other
economies in the longer run.
– 148 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
Table 6.2: Summary of Pairwise Convergence Results
(a) Gross state product per capita
Pairwise case
Conclusion
Significance
New South Wales
Victoria
Long-run convergence
Long-run convergence at 1%
Catching up at 10%
New South Wales
Queensland
Long-run convergence
Long-run convergence at 5%
Catching up at 10%
New South Wales
South Australia
Catching up*
Catching up at 1%
New South Wales
Western Australia
Long-run convergence*
Long-run convergence at 10%
New South Wales
Tasmania
Catching up*
Catching up at 1%
Victoria
Queensland
Long-run convergence
Long-run convergence at 1%
Victoria
Western Australia
Long-run convergence
Long-run convergence at 1%
Victoria
South Australia
Catching up*
Catching up at 1%
Victoria
Tasmania
Catching up*
Catching up at 1%
South Australia
Queensland
Long-run convergence*
Long-run convergence at 1%
South Australia
Western Australia
Long-run convergence
Long-run convergence at 1%
South Australia
Tasmania
Catching up at 5%
Long-run convergence at 1%
Catching up at 5%
Western Australia
Queensland
Long-run convergence
Long-run convergence at 1%
Western Australia
Tasmania
Catching up*
Long-run convergence at 1%
Catching up at 5%
Tasmania
Queensland
Catching up*
Catching up at 1%
* Indicates that bootstrap critical values were calculated. See Appendix 6B for more details.
– 149 –
Productivity and Regional Economic Performance in Australia
Table 6.2: Summary of Pairwise Convergence Results
(b) Gross state product per person employed
Pairwise case
Conclusion
Significance
New South Wales
Victoria
Long-run convergence
Long-run convergence at 1%
New South Wales
Queensland
Long-run convergence
Long-run convergence at 1%
New South Wales
South Australia
Long-run convergence
Long-run convergence at 1%
Catching up at 10%
New South Wales
Western Australia
Long-run convergence
Long-run convergence at 5%
New South Wales
Tasmania
Catching up
Catching up at 1%
Victoria
Queensland
Long-run convergence
Long-run convergence at 5%
Victoria
Western Australia
Long-run convergence
Long-run convergence at 1%
Victoria
South Australia
Long-run convergence
Long-run convergence at 1%
Catching up at 10%
Victoria
Tasmania
Catching up
Catching up at 1%
South Australia
Queensland
Long-run convergence
Long-run convergence at 1%
South Australia
Western Australia
Long-run convergence
Long-run convergence at 1%
South Australia
Tasmania
Long-run convergence
Long-run convergence at 1%
Catching up at 10%
Western Australia
Queensland
Catching up
Catching up at 10%
Western Australia
Tasmania
Long-run convergence
Long-run convergence at 1%
Tasmania
Queensland
Catching up
Catching up at 10%
– 150 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
Conclusion
This chapter has provided a preliminary analysis of the time series convergence behaviour
of the Australian states over the period 1985 to 2000. The most striking finding of this
analysis is that some form of convergence behaviour is apparent for every pairwise case
studied here. This finding contrasts with the evidence of increased dispersion apparent in
the m-convergence indicators reported in Figures 6.2a and 6.2b. Interestingly, this is
consistent with Bernard and Durlauf’s (1996) arguments, namely that the use of weighted
averages in cross-sectional tests can overstate the convergence dynamics operating in a
group of economies. This finding has a number of important implications for policy making
and for future research into the issue of regional convergence. We discuss each topic in turn.
Policy implications
Interpreting the theory of economic convergence and using it to guide policy making is not a
straightforward task. The central challenge here is to translate the abstract conclusions suggested
by neoclassical growth theory into a concrete analysis of comparative economic growth. In the
following, we identify three important issues for the consideration of policy makers.
First, it is necessary to understand what the time series tests of convergence conducted here
convey about state economic development since 1985. Following Bernard and Durlauf (1996),
these tests investigate whether the underlying pattern of growth observed here is consistent with
a rigorously defined convergence process. Therefore, our findings do not represent a statement
about convergence in the levels of GSP per capita and GSP per person employed as much as a
statement about the nature of economic growth. In fact, overall convergence in level terms
appears to have been limited over the period 1985 to 2000, with the overall dispersion of GSP
per capita and GSP per person employed increasing slightly since 1985.
Second, these results let us draw conclusions about the impact of economic restructuring on
state economic growth, particularly those changes that have resulted from the policy of
microeconomic reform pursued by state and federal governments since the 1980s.
Specifically, while the process of microeconomic reform has been very intense and has had
wide-ranging effects on economic activity, it has not shifted the underlying convergence
behaviour between the state economies. In particular, economic restructuring has not
resulted in divergence at the state level. This finding therefore allows us to establish some
boundaries for the analysis of comparative economic growth among the states.
Finally, the steady-state nature of the conclusion of long-run convergence has implications
for the future economic growth of the states. Using Queensland as an example, we can see
that it is catching up to New South Wales and Victoria in the sense that approximately
30-40% of the gap in GSP per capita between Queensland and these two states in 1985 had
been eliminated by 2000.4 The results of the tests for convergence in GSP per capita for
Queensland indicate that long-run convergence has now been achieved in terms of the
neoclassical convergence process. Practically, this means that while incomes per capita have
not been equalised between Queensland and New South Wales or Victoria, Queensland has
reached a steady-state in its relative GSP per capita levels in relation to these two states.
4
Approximately two-fifths of the gap between Queensland and Victoria has been eliminated, together with one-third of the gap
between Queensland and New South Wales.
– 151 –
Productivity and Regional Economic Performance in Australia
Therefore, to the extent that Queensland has reached a steady-state position in its relative economic
growth, there are limits to the scope of the gains that the state can make from ‘natural’ convergence
processes in the future to try and further eliminate the absolute income differences that remain.
Basically, the natural convergence processes are those that are consistent with the neoclassical
growth model’s focus on capital deepening and the inherent tendency for incomes to equalise
across economies in the long run. Therefore, if a state such as Queensland is to make gains in GSP
per capita relative to other ‘richer’ states in the future, it will have to deliberately induce changes in
the underlying structure of its economy. This includes investments in human capital and focusing
on other endogenous factors that enhance the State’s aggregate production frontier.
Simply put, now that the State has caught up in terms of capital deepening it must catch up in
terms of the microeconomic specifications of its aggregate production function.5 It must be
noted that the scope of the policy actions suggested by this analysis is wide. The
microeconomic structure of the economy can be altered through further microeconomic reform
(particularly as these reforms relate to the productivity of the education system) and/or through
government investment. Of course, such policies will only assist catching up if other states do
not grow faster or introduce parallel policies to accelerate the rate of economic growth.
Issues in regional convergence
The conclusions of long-run convergence and catching up reached in this chapter are
qualified by the increased dispersion of GSP per capita and GSP per person employed
apparent in the state economies and the weaker findings for Tasmania. When considered in
conjunction with the results of previous cross-sectional studies, these results illustrate the
complex nature of regional convergence dynamics in Australia. It must be emphasised that
the contrasting results for the cross-sectional and time series approaches studied are
consistent with Bernard and Durlauf’s (1996) commentary on empirical convergence issues.
Technically, the findings illustrate the hazards of employing cross-sectional weighted averages in
studying the convergence behaviour of groups of economies. While data constraints prevent a full
investigation of ‘convergence clubs’ in the Australian context, research focusing on structural
change and the time series properties of m- and `-convergence measures over longer periods
could provide further insights into the convergence dynamics of Australia’s regional economies.
Practically, this study indicates that while increased dispersion is present it has not yet led
to a divergence in growth in GSP per capita and GSP per person employed according to time
series definitions of convergence. This is a surprising finding in so far that the narrow time
period considered here contains a number of transitory shocks to output associated with the
business cycle that could be expected to favour conclusions of divergence in the short term.
5
Note that some of the differences in the level of GSP per capita among the states may be attributable to cost factors. That is, variations
in the cost of living, particularly housing, can affect living standards but not necessarily be captured by traditional measures of GSP.
– 152 –
Recent Convergence Behaviour of the Australian States: A Time Series Approach
References
Aghion, P. and Howitt, P. (1998), Endogenous Growth Theory, MIT Press, Cambridge.
Bernard, A. and Durlauf, S. (1995), ‘Convergence in international output’, The Journal
of Applied Econometrics, 10, 97-108.
Bernard, A. and Durlauf, S. (1996), ‘Interpreting tests of the convergence hypothesis’,
The Journal of Econometrics, 71, 161-173.
Bodman, P. and Crosby, M. (2002), ‘The Australian business cycle: Joe Palooka or dead
cat bounce’, Australian Economic Papers, 41 (2), 191-207.
Cashin, P. (1995), ‘Economic growth and convergence across the seven colonies of
Australasia: 1861-1991’, The Economic Record, 71 (213), 132-144.
Cashin, P. and Strappazzon, L. (1998), ‘Disparities in Australian regional incomes: Are
they widening or narrowing?’, Australian Economic Review, 31 (1), 3-26.
de la Fuente, A. (1997), ‘The empirics of growth and convergence: A selective review’,
The Journal of Economic Dynamics and Control, 21 (1), 23-73.
Doldado, J., Jenkinson, T. and Sosvilla-Rivero, S. (1990), ‘Cointegration and unit roots’,
The Journal of Economic Surveys, 4, 249-273.
Dowrick, S. and Nguyen, D.T. (1989), ‘OECD comparative economic growth 1950-85:
Catch-up and convergence’, American Economic Review, 79 (5), 1010-1030.
Durlauf, S. and Quah, D. (1999), ‘The new empirics of economic growth’, in J. Taylor
and M. Woodford (eds), The Handbook of Macroeconomics, Elsevier Science Press,
North Holland, Amsterdam.
Enders, W. (1995), Applied Econometric Time Series, John Wiley and Sons, New York.
Hamilton, J.D. (1994), Times Series Analysis, Princeton University Press, New York.
Harris, P. and Harris, D. (1991), ‘Interstate disparities in real gross state product at factor
cost per head in Australia 1953-54 to 1990-91’, The Australasian Journal of Regional
Studies, 6 (1), 42-58.
Lucas, R. (1988), ‘On the mechanics of economic development’, The Journal of Monetary
Economics, 22, 3-42.
Maxwell, P. and Hite, J.C. (1992), ‘The recent divergence of regional per capita incomes:
Some evidence from Australia’, Growth and Change, 23 (1), 37-53.
Nguyen, T., Smith, C. and Meyer-Boehm, G. (2003), ‘Variations in economic and labour
productivity growth among the states of Australia: 1984-85 to 1998-99’, in C. Williams,
M. Draca and C. Smith (eds), Productivity and Regional Economic Performance in Australia,
Office of Economic and Statistical Research, Brisbane.
Press, W.H., Teukolsky, S.A., Vetterling, W.T. and Flannery, B.P. (1992), Numerical Recipes in
Fortran: The Art of Scientific Computing, second edition, Cambridge University Press, UK.
– 153 –
Productivity and Regional Economic Performance in Australia
Quantitative Micro Software (2001), EVIEWS 4 Users Guide, Quantitative Micro
Software, Irvine, California, 360-366.
Romer, P. (1986), ‘Increasing returns and long-run growth’, Journal of Political Economy,
94 (5), 1003-1037.
Solow, R. (1956), ‘A contribution to the theory of economic growth’, The Quarterly
Journal of Economics, 70, 65-94.
Swan, T. (1956), ‘Economic growth and capital accumulation’, The Economic Record, 32,
334-361.
– 154 –
7 Human Capital Investment and
Economic Growth in the Australian
Economy
Mirko Draca, John Foster and Colin Green
Introduction
Education has been in vogue as a central topic of policy discussion in advanced economies
since the early 1990s. Arguably, this trend began in earnest with the Clinton administration’s
early declared commitment to improving education and the subsequent adoption of this
focus by other national governments. In Australia, the state of the education system at all
levels is a regular topic of intense public debate. Indeed, the high importance of education
is a given in these debates. It is now, after economist J.K. Galbraith’s famous phrase, a part
of the ‘conventional wisdom’.
However, Australian research into the macroeconomic impact of changes in the educational
composition of the workforce has produced some ambivalent findings. Furthermore, the
debate on education policy in Australia has regularly focused on the level of government
expenditure on education, as opposed to the actual economic features and consequences of
human capital investment. In turn, this has resulted in misguided thinking. In a stimulating
contribution, Gregory (1995, p. 296) noted:
Government expenditure on education is only five per cent of GDP and while
large swings in this ratio (from say, 4.7% to 5.1%) may exert considerable
effects on education institutions and students, it is difficult to believe that the
impact on macro labour market outcomes and economic growth can be large
in the short-to-medium term. These expenditure variations are quite small
relative to variations in investment in machines and buildings and variations
in other policy instruments such as interest rates, government deficits and
wages. We should not expect too much from our education policy.
Following this observation, this chapter discusses the economic potential of education
policy in Australia at the state and national level. We emphasise the need for Australian
governments to articulate explicit policies for human capital investment at the
macroeconomic level. The scope for education policy to improve economic outcomes is
significant if the considerations raised by Gregory and others can be incorporated into future
policy design. However, realising this potential requires a major re-orientation of the current
debate on education policy in Australia. In particular, we argue that the focus on the
financing of education that has dominated debate since the introduction of the Higher
Education Contribution Scheme (HECS) needs to be complemented with a discussion of the
composition of human capital investment.
– 155 –
Productivity and Regional Economic Performance in Australia
We outline this argument in four sections. The first section reviews existing Australian
research on the effects of educational expansion on the labour market. This research
indicates that educational expansion has had complicated effects on labour market
outcomes. In turn, this indicates that the simple ‘skill upgrading’ hypothesis that has
dominated policy discussion needs to be substantially revised. Specifically, the assumption
that skill upgrading results in uniform improvements in conditions at all levels of the labour
market needs to be replaced with a more rigorous analysis of labour market dynamics.
The next two sections contain some growth accounting exercises that fill gaps in the current
research on educational expansion. The second section follows the research by Gregory
(1995) and examines the sources of educational expansion via two decompositions of the
growth in human capital since 1970. In particular, we decompose the overall growth in
human capital by its sources in labour force growth and enrolment rates.
The third section analyses the relationship between human capital and economic
performance at the state level, using the cross-sectional approach outlined by Bhatta and
Lobo (2000). We find that differences in human capital can explain variations in gross state
product (GSP) per capita among the Australian states. Comparative static projections also
indicate that various states could benefit from expansions in educational attainment.
However, we note that there is a need to consider the impact of educational expansion on
wage determination in future research on differences in human capital at the state level.
The concluding section provides a detailed discussion of the policy issues canvassed in this
chapter. We discuss the need to develop human capital investment policies at the state and
national levels in line with a revised approach to the skill upgrading hypothesis. The need
to focus on the composition rather than the financing of human capital investment is also
emphasised. We illustrate this argument with two potential approaches for the promotion of
human capital formation: state-based human capital investment policies and comprehensive
programs in the area of early childhood education.
Overall, this chapter aims to illustrate some of the main themes of this volume. In particular,
we stress the need to connect the analysis of productivity to specific government policies.
Different patterns in research and development, human capital and industrial structure
among the states create a need for state governments to modernise their economic
development policies. Specifically, the traditional focus on developing the physical capital
stock needs to be supplemented with new policies to improve human capital and research
and development. Furthermore, these policies must be elaborated in ways that are consistent
with the goals of dynamic and allocative efficiency.
Educational Expansion and the Labour Market
As discussed, while the high importance of education is an accepted feature of
contemporary policy debate in Australia, the content of this debate is characterised by some
major omissions. In particular, the impact of past human capital investments on Australian
labour market outcomes has not been a focus for discussion. Clearly, the impact of previous
investments determines the scope of current and future education policy.
– 156 –
Human Capital Investment and Economic Growth in the Australian Economy
The policies of educational expansion followed by Australian governments since the 1970s
have been underpinned by what Gregory (1995) terms a ‘skill upgrading’ approach. The
skill upgrading approach to education policy can be summarised in terms of a high-wage,
high-skill argument or hypothesis. That is, improvements in the education level of the
labour force are assumed to generate more jobs in high-skill professions at better rates of
pay. However, the evidence on education and earnings in Australia suggests that the
relationship is more complicated. In the following we comment on the credibility of this
hypothesis and comment on the need to develop more rigorous underpinnings for future
education and labour market policy.
The economics of educational expansion
Gregory (1995) explores the relationship between education expansion and labour market
outcomes at the macroeconomic level. In the following we focus on two aspects of
Gregory’s analysis: the determinants of the increase in the demand for education and the
impact of increased labour quality on economic performance.
First, Gregory argues that the increased demand for education seen since the 1970s is
closely related to changing labour market conditions. Specifically, the increase in education
participation at the secondary and tertiary level was driven by a fall in the effective cost of
an additional year of education. The increase in part-time employment opportunities for
young people has allowed them to earn income while studying and therefore reduced the
opportunity cost of further education. In turn, the reduction in full-time job opportunities
has re-enforced this trend towards combining work and study. Further, increases in
government education subsidies provided incentives for education participation.
Second, Gregory examines the relationship between labour quality, productivity and real
wages. His review of data from the Income Distribution Survey (IDS) is dominated by a
surprising result (Gregory 1995, p. 318):
These data seem to suggest that improvement in the education quality of the
labour force, measured by education qualifications weighted by earnings, has
a limited role to play in the increasing real wages and productivity of
Australian workers. They also suggest that the increased education is not
necessarily a major contributor to economic growth.
Of course, this finding is mitigated by the complexity and scale of the economic changes
that have occurred in Australia since the early 1970s. As Gregory notes, large increases in
real wages were driven by macroeconomic factors, obscuring the productivity impact of
labour quality. The weights assigned to different workers are also limited by the compressed
earnings relativities that prevail at the lower end of the education distribution. Finally, the
ratio of new educated workers to the overall size of the labour force is low by definition.
This means that the effects of education expansion are subject to long and uncertain lags as
younger, more educated groups progress through the age–earnings profile and gradually
raise the average level of attainment.
Despite these caveats, the issues raised by Gregory are critical because they challenge the
assumption that there is an automatic relationship between skill upgrading, higher pay and
improved economic performance. In particular, Gregory identifies the central flaw of the
– 157 –
Productivity and Regional Economic Performance in Australia
skill upgrading hypothesis – no distinction is made between the effects of skill upgrading at
the individual level and at the macroeconomic level. As a result, the heterogenous labour
market impacts of educational expansion have been largely ignored by policy makers.
These impacts are best understood by considering evidence on the links between education
and the structure of the labour market.
Education and labour market structure
Educational expansion has had a varied impact on the structure of the Australian labour
market. In the following, we discuss the impact of educational expansion on earnings
(Borland, 1996 and 1999; Sheehan and Esposto, 2001), occupational status (Vella and
Karmel, 1999) and the returns to education (Vella and Gregory, 1996).
The impact of educational expansion on earnings inequality in Australia can be viewed in a
number of ways. First, it is clear that, other things being equal, educational expansion has
not prevented a rise in earnings inequality since the 1970s. For example, Borland (1999)
finds that the dispersion between the top and bottom percentiles of the earnings distribution
increased between 1975 and 1997.1 Second, educational expansion has played a complex
role in the evolution of the earnings distribution. Borland and Kennedy’s (1998) application
of the Juhn, Murphy and Pierce (1993) decomposition finds that unobservable factors have
exerted the largest influence on the dispersion of earnings. In terms of human capital,
changes in the distribution of educational attainment and experience appeared to have had
a neutral effect on earnings dispersion. In contrast, changes in the returns to education and
experience tended to reduce earnings dispersion.
Borland relates this finding to changes in the relative demand and supply of skilled workers.
In particular, he reconciles the increase in the relative demand for skilled workers with the
rising incidence of low wage employment (Borland, 1999, p. 191):
What is happening is that the shift in demand has two opposing effects. The
number of low-skill workers declines, but the relative earnings of low skill
workers also fall. The proportion of workers in low-pay jobs will increase
where the effect on earnings of low-skill workers ‘dominates’ the effect on
employment.
This does not imply that educational expansion exerted a negative impact on labour
markets. Rather, educational expansion was essential for accommodating the increase in the
relative demand for skilled workers in the economy. However, the labour market success of
high skill workers (in terms of relative earnings) also had implications for the incidence of
low paid employment. This illustrates the difficulties that exist in using skill upgrading
policies as a tool to improve labour market outcomes.
Another dimension of the high-wage, high-skill argument relates to the incidence of
employment in the upper end of the occupational hierarchy. Vella and Karmel (1999)
examine changes in the occupational status of two successive groups of males in the mid
1
Specifically, Borland (1999, p. 177) notes that ‘between 1975 and 1997 real weekly earnings of a male employee at the 25th
percentile increased by 1.3%, whereas earnings of an employee at the 75th percentile increased by 19.3%’. For females, real
earnings increased by 15.8% and 32.1% for the 25th and 75th percentiles respectively.
– 158 –
Human Capital Investment and Economic Growth in the Australian Economy
1980s from the longitudinal Youth in Transition surveys. Their results indicate that, despite
a substantial increase in the education levels of the second group, the occupational
distribution of workers in this group was almost identical to those in the first group. This
indicates that, in the case they consider,2 educational expansion had moved all individuals
up the educational ladder without altering their relative position on the occupational ladder.
For example, the proportion of individuals in high skill occupations did not increase despite
a substantial increase in the educational composition of the labour force.
Furthermore, Vella and Karmel (1999) find no evidence that increased education was
manifested in higher real wages or served as a response to changes in the overall structure
of labour demand. This raises doubts about the efficacy of policies that assume there is an
automatic link between skill upgrading and the incidence of high skill employment. Their
findings are also interesting in light of Borland’s (1996) evidence on relative demand shifts
for educated workers.
Employing a similar approach, Vella and Gregory (1996) focus on the variations in labour
market outcomes within educational groups. Specifically, they analyse the effect of overachievement and under-achievement (defined relative to the expected educational
achievement based on individual socioeconomic background) and the influence of personal
and job characteristics on income for different levels of education. They find that each of
these factors has a different and substantial effect on earnings at each education level.
In turn, they emphasise the heterogenous impact of educational expansion (Vella and
Gregory, 1996, p. 217):
The findings indicate that the way in which education opportunities are
expanded is important. They suggest that an expansion of the education system
which proceeds evenly by moving everyone up the education ladder will have
different effects than an expansion which proceeds mainly by increasing the
extent of education over-achievement or reducing under-achievement.
Gender differences and human capital investment
The evidence on educational expansion reviewed above is ambivalent on the productivity
benefits of increased educational attainment in the labour force. An unnerving feature of this
research is the limited success of educational expansion as a vehicle for transforming labour
market outcomes. In this section we discuss the role of human capital investment in
reducing the gender wage gap since 1973. This illustrates the capacity that educational
expansion has to influence labour market outcomes and serves as a counterpoint.
Kidd and Shannon (2001) decompose the gender wage gap during the 1980s to distinguish
between the role of wage-structure and gender-specific effects in reducing the difference in
male and female earnings. They find that the improvement in female human capital can
account for 81% of the change in the wage gap in their basic model. An extended model –
inclusive of occupational and industry effects – indicates that human capital accounts for a
‘still considerable’ 44% of the gap.
2
Namely, full-time employed males from the 1985 wave of the Longitudinal Survey of Australian Youth.
– 159 –
Productivity and Regional Economic Performance in Australia
Australian research on the gender wage gap (Preston, 1997; Karmel, 1996) has regularly
identified education and experience as decisive factors influencing the path of this gap.
Indeed, Kidd and Shannon (2001) note that this is a recurrent theme internationally and cite
studies from the United States, the United Kingdom and Canada. While male–female wage
convergence was much slower in the 1980s compared with the 1970s, the pivotal role of
human capital investment for females in the 1980s indicates that educational expansion can
help to transform labour outcomes. In this case, while the female market was exposed to the
same demand shifts identified by Borland (1999), the effects relating to low wage
employment stand in relief to the bigger picture of gender wage convergence.
However, this does not imply that educational expansion can be treated as a simple policy
lever in the Keynesian fashion. For example, while human capital investment can explain
changes in the gender wage gap, the nature of the causal mechanisms at work is unclear.3
Projections of the future path of the gender wage gap by Kidd and Shannon (2002) also
indicate that wage convergence in the next three decades will hinge more on changes in the
wage structure – that is, relative returns to education and experience – than gains in female
human capital. Future policies that are intended to assist male–female wage convergence will
therefore need to focus on the links between human capital investment and the wage structure.
Human Capital Formation in the Australian Economy
The review above highlights the difficulties that arise in tracking changes in the educational
composition of the labour force as well as the economic impacts of those changes. The entry
and exit of different age groups from the labour force has long lags. In turn, this affects the
measurement of average labour quality and the evolution of the wage structure. Gregory’s
(1995) study found that the increase in labour quality between 1967 and 1990 was quite
small – approximately 0.5% per year (or 12% for the total period). He also notes that a
significant proportion of the increase in the educated labour force between 1967 and 1979
was due to the reduced employment of less educated full-time workers. As a result, labour
force quality increased more rapidly between 1967 and 1979 than in the following period.
These results emphasise the need to examine the quantitative and qualitative dimensions of
educational expansion in Australia. This is necessary as part of a more general
macroeconomic policy for human capital investment. In the following we review Australia’s
stocks of educational attainment relative to the rest of the OECD and identify the
determinants of Australia’s educational expansion since the 1970s via two growth
accounting exercises.
Australia and the OECD
Relative to OECD trends, Australia has had a varied performance in educational attainment.
Table 7.1 shows the distribution of the prime age population and labour force by level of
educational attainment across the OECD. Australia has above average outcomes at the
tertiary level but has poorer outcomes for completion at the secondary level.
3
For example, increased female human capital investment does not appear to have decisively affected the occupational
distribution of the female labour force. Kidd and Meng’s (1997) study indicates that the increased acquisition of female human
capital in the 1980s should have led to a greater decrease in occupational segregation than actually occurred.
– 160 –
Human Capital Investment and Economic Growth in the Australian Economy
Table 7.1: Distribution of Educational Attainment in the OECD,
25-64 years age group, %
Population
Country
Labour Force
Below
Nonupper
Upper university University
secondary secondary tertiary
level
education education education education
Total
Below
upper
secondary
education
NonUpper
university
secondary tertiary
education education
University
level
education Total
Australia
43
32
10
15
100
37
35
11
17
100
Austria
29
63
2
6
100
23
68
2
7
100
Belgium
47
30
13
11
100
37
33
16
14
100
Canada
24
29
31
17
100
18
29
33
20
100
Czech Republic
16
74
-
10
100
12
76
-
12
100
Denmark
34
44
7
15
100
29
47
8
17
100
Finland
33
46
9
12
100
29
48
10
14
100
France
40
41
9
10
100
34
44
11
11
100
Germany
19
60
9
13
100
14
61
10
15
100
Greece
56
25
7
12
100
50
26
9
15
100
Hungary
37
50
-
13
100
24
59
-
17
100
Ireland
50
28
12
11
100
43
29
14
14
100
Italy
62
30
-
8
100
54
34
-
11
100
Korea
39
42
-
19
100
38
41
-
21
100
Luxembourg
71
18
-
11
100
63
21
-
16
100
Netherlands
37
40
-
23
100
29
43
-
27
100
New Zealand
40
35
14
11
100
35
38
15
13
100
Norway
18
55
11
16
100
15
56
12
17
100
Poland
26
61
3
10
100
21
64
4
12
100
Portugal
80
9
3
7
100
76
11
4
9
100
Spain
70
13
5
13
100
62
15
6
17
100
Sweden
26
47
14
13
100
23
48
15
14
100
Switzerland
20
58
12
10
100
17
58
14
10
100
Turkey
83
11
-
6
100
78
13
-
9
100
United Kingdom
24
55
9
13
100
19
57
10
15
100
United States
14
52
8
26
100
11
52
9
28
100
Country mean
40
40
10
13
100
34
43
11
15
100
Source: OECD (1998a), Education at a Glance
This has implications for the dynamics of Australia’s qualification profile. Although there
have been significant improvements in Australia’s school completion rates and in vocational
training, this has not eliminated historical shortfalls in these areas of attainment. Other
OECD economies appear to have been expanding qualifications more rapidly and evenly
while Australia’s expansion has been concentrated on mass higher education. The OECD
(1998a) has also noted that Australia’s recent experience has been characterised by low
levels of completion but relatively high levels of educational participation.
– 161 –
Productivity and Regional Economic Performance in Australia
Educational expansion and labour force growth
We analyse two aspects of human capital accumulation in Australia since 1970: the role of
labour force growth and the shifting economic value of the total stock of human capital.
The role of labour force growth can be explored using Gemmell’s (1996) methodology for
measuring aggregate human capital. Gemmell (1996) defines a specialised indicator of the
human capital stock that separates the influence of labour force growth and changing
enrolment rates in driving human capital accumulation. He uses an explicit flows-based
approach to take account of the labour force entry and exit of the different age groups. Our
interest here lies in what the Gemmell measure reveals about the structure of human capital
accumulation in Australia.
We calculate the Gemmell index for Australia in Table 7.2. UNESCO data on enrolment
rates are used together with Australian Bureau of Statistics labour force data. The UNESCO
data are available from 1970 and represent the most consistent enrolment series available
for Australia. The regular reclassification of many qualifications undermines the accuracy
of stock data for educational attainment in Australia over the long term. Secondary and
tertiary enrolment rates therefore offer a more consistent approach to capturing the
generational structure of human capital accumulation. Full details of the implementation of
the Gemmell index are given in Appendix 7A.
Table 7.2: Decomposition of Gemmell (1996) Human Capital Index for Australia
Secondary qualifications
Tertiary qualifications
Increase in
Gemmell
index
%
Labour force
growth only
%
Enrolment
growth only
%
Increase in
Gemmell
index
%
Labour force
growth only
%
Enrolment
growth only
%
1970-1980
64.4
73.0
27.0
50.7
76.0
24.0
1980-1990
26.5
98.8
1.2
48.4
54.1
45.9
1990-1995
11.7
56.4
43.6
23.0
28.6
71.4
1970-1995
132.2
73.9
26.1
175.1
49.3
50.7
Year
Sources: Enrolment growth rates based on UNESCO data; ABS Labour Force, Cat. no. 6203.0
Table 7.2 shows the scale of the increase in human capital that occurred between 1970 and
1995. The total stock of upper secondary qualifications increased by 132.2% over this
period while the stock of tertiary qualifications increased by 175.1%. This increase is in line
with the scale of educational expansion experienced in other advanced countries.
Importantly, Table 7.2 also decomposes these increases in the Gemmell index to measure the
relative contributions of labour force growth and enrolment rates to raising the human capital
stock. This decomposition uncovers two major structural features of human capital
accumulation in Australia. First, it establishes that labour force growth has driven the overall
increase in human capital, particularly in the 1980s. Almost all (98.8%) of the growth in
secondary qualifications during the 1980s was due to an expansion in the labour force.
However, the accumulation of tertiary qualifications was more evenly based – approximately
half of the growth (49.3%) resulted from labour force expansion. This labour force effect
was much more subdued in the early 1990s when growth due to labour force expansion
comprised only 56.4% of total growth in secondary qualifications.
– 162 –
Human Capital Investment and Economic Growth in the Australian Economy
Second, the decomposition in Table 7.2 indicates where the effects of various policy actions
have been felt. While labour force growth has clearly dominated changes in enrolment rates
in the secondary sector, this effect is weaker for tertiary qualifications. The move to a
system of mass higher education is reflected in the strong role that enrolment rates play in
driving tertiary qualification growth in the 1980s and early 1990s. The figures for the 1990s
in the last column indicate that 71.4% of tertiary qualification growth in this period can be
explained by changes in the enrolment rate.
The labour–income index of human capital calculated in Table 7.2 offers an alternative
perspective on the large increases in human capital recorded by the Gemmell index. Based
on Fernandez and Mauro’s (2000) methodology, the index in Table 7.3 weights educational
attainment by wages to calculate the economic value of human capital. By weighting
different segments of the labour force in this way, the Fernandez–Mauro index can capture
the ‘learning by doing’ effects in human capital accumulation. That is, as groups progress
through different age bands they receive higher wages. These higher wages are a reflection
of the extra productivity that is embodied in experienced workers. Full details of the
implementation of the Fernandez–Mauro index are given in Appendix 7A.
Table 7.3: Labour–Income Index of Human Capital for Australia a
Labour–income index
Rate of change (%)
1970
1975
1980
1985
1990
1995
1.75
1.83
1.93
2.02
2.17
2.26
1970-75
1975-80
1980-85
1985-90
1990-95
1970-95
4.7
5.0
4.9
7.3
4.3
4.7
na
a Based on the methodology developed by Fernandez and Mauro (2000).
Measuring aggregate human capital according to its economic value produces a different
pattern of growth to that exhibited by the Gemmell index. As Table 7.3 indicates, the
labour–income value of aggregate human capital grew at an average rate of 4.7% for the
five-year intervals between 1970 and 1995. This is substantially less than the large increases
recorded by the Gemmell index and reflects the fact that Fernandez and Mauro’s (2000)
approach is a qualitative one that measures changes in the value of the human capital stock.
In comparison, the Gemmell index measures the structure and dimensions of the
quantitative expansion in human capital.
Overall then, the results in Tables 7.2 and 7.3 have some implications for the structural
composition of human capital investment in Australia. As the decomposition for the 1990s
shows, Australia will not be able to rely on labour force growth alone to drive its human
capital accumulation in the future. This will affect Australia’s capacity to converge with the
levels of educational attainment that prevail in leading OECD economies such as the United
States. Although young, highly educated groups will move through the age structure and
begin to dominate the labour force, the rate of increase in the levels of educational
attainment present in the labour force will be slower than in previous decades. While this
will be a general trend in the OECD, Australia’s reliance on labour force growth will make
this effect more pronounced. This indicates that convergence with levels of educational
attainment that exist in leading OECD economies is not assured. Explicit education
planning policies may be needed to secure full convergence.
– 163 –
Productivity and Regional Economic Performance in Australia
Finally, the rates and source of growth in educational attainment in Table 7.2 corroborate
with the analysis in Gregory (1995). In particular, the pivotal role of labour force growth in
driving educational expansion is consistent with the pattern of labour quality and real wage
growth described in Gregory.
Human Capital Formation and State Economic Performance
The relationship between educational expansion and economic performance lies at the core
of the skill upgrading hypothesis outlined earlier. As discussed, Gregory (1995) found little
evidence that changes in the educational composition of the labour force could explain
changes in real wages and productivity at the national level. However, it is possible that a
range of intervening factors are obscuring the relationship between human capital
investment and economic performance.
In this section we examine the relationship between human capital investment and
economic performance at the state level. Specifically, we employ Bhatta and Lobo’s (2000)
cross-sectional approach to decompose variations in GSP per capita and human capital
among Australia’s states. This allows us to analyse the impact of human capital on
economic performance without the problems that arise in time series studies of this
relationship.4 In the following sections we survey the differences in human capital among
the states and then report the results of the Bhatta–Lobo decomposition.
State level differences in human capital
Compared to inter-country variations in human capital, the variations that exist between
Australia’s states are relatively small. State level differences in human capital can be
measured in terms of completion rates or the highest level of educational attainment. Figure
7.1 shows the rates of educational completion by state. Three categories are shown: the
proportion of the population with degree level qualifications, the proportion of the
population that has not completed year 12,5 and the proportion that has completed year 12.
Note that this latter category overlaps with degree completion such that the total figures
exceed 100%.
The figures for each category are reported in Table 7.4.6 These figures indicate that New
South Wales and Victoria have major advantages in the area of degree completion. Their
degree completion rates are approximately 5 percentage points higher than the rates for
Queensland and South Australia, 3 percentage points higher than Western Australia, and
over 6 percentage points higher than Tasmania. There is greater variation in the rates of year
12 non-completion. Interestingly, Victoria (36.2%) has a higher non-completion rate than
New South Wales (33.2%). Queensland and South Australia both have year 12 noncompletion rates of approximately 39% while Tasmania has the highest rate at 44.3%.
Western Australia has the second lowest non-completion result at 34.3%.
4
For example, Wolff (2000) notes that auxiliary, investment-related variables can cancel out the effects of human capital in
international growth regressions. Also, see Durlauf (2000) for a detailed discussion of econometric issues in growth
accounting.
5
Hereafter, we refer to this as the rate of year 12 non-completion.
6
Note that, as completion rates, the figures in Table 7.4 overlap and produce a total that is greater than 100%.
– 164 –
Human Capital Investment and Economic Growth in the Australian Economy
Figure 7.1: Educational Completion by State, 2000,
persons aged 15-64 years
< Year 12
Degree
Year 12
100
90
80
Per cent
70
60
50
40
30
20
10
0
NSW
Qld
Vic
SA
Tas
WA
Source: ABS, Education and Work, Australia, May 2000, Cat. no. 6227.0, unpublished data
Table 7.4: Educational Completion Rates by State and Territory, 2000,
persons aged 15-64 years, %
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
< Year 12
33.2
36.2
38.7
39.3
34.3
44.3
34.8
19.2
35.6
Year 12
66.8
63.8
61.3
60.7
65.7
55.7
65.2
80.8
64.6
Degree
19.8
19.7
14.7
14.7
16.5
13.2
17.4
32.6
18.1
Source: ABS, Education and Work, May 2000, Cat. no. 6227.0, unpublished data
Table 7.5 gives detailed results on the highest level of educational attainment for the
working age population in each state and territory. This uncovers further differences in the
distribution of qualifications. New South Wales has the largest proportion of higher degree
qualifications (3.2%), 0.8 to 1.6 percentage points higher than the other states. Queensland,
Western Australia and South Australia all perform better than New South Wales and Victoria
in the area of vocational qualifications.
– 165 –
Productivity and Regional Economic Performance in Australia
Table 7.5: Educational Attainment by State and Territory, 2000,
persons aged 15-64 years, %
NSW
Vic
Qld
SA
WA
Tas
NT
ACT
Aust
Higher degree
3.2
2.4
2.1
2.0
2.0
1.6
1.8
5.2
2.5
Postgraduate diploma
2.5
3.7
1.7
2.4
2.2
2.1
2.4
4.9
2.7
Bachelor degree
14.0
13.7
10.9
10.3
12.3
9.5
13.2
22.5
12.9
Total with degree
18.1
Post-school qualifications a
19.7
19.8
14.7
14.7
16.5
13.2
17.4
32.6
Undergraduate diploma
5.9
6.1
5.3
5.6
6.8
5.7
8.1
6.6
5.9
Associate diploma
3.8
3.8
2.8
2.9
2.4
2.8
4.3
4.2
3.4
Total with diploma
9.7
9.9
8.1
8.5
9.2
8.5
12.4
10.8
9.3
Skilled vocational qualification
12.1
13.1
15.1
13.3
14.9
13.8
14.4
9.3
13.3
Basic vocational qualification
10.7
6.5
8.4
8.5
8.8
7.9
6.9
10.4
8.8
Total with vocational qualification
22.8
19.6
23.5
21.8
23.7
21.7
21.3
19.7
22.1
Total with post-school qualifications
52.2
49.3
46.4
45.0
49.4
43.4
51.0
63.0
49.5
Without post-school qualifications b
Attending tertiary institution in May 2000
1.6
1.0
1.5
1.1
1.6
0.8
2.2
2.3
1.4
Not attending tertiary institution in May 2000 c 12.9
13.5
13.4
14.3
14.6
11.2
11.6
15.3
13.4
Total completed highest level of school
14.5
14.4
14.9
15.4
16.1
12.0
13.8
17.6
14.8
1.0
1.1
1.5
1.1
0.9
1.3
2.1
0.8
1.1
Not attending tertiary institution in May 2000 c 32.2
35.0
37.2
38.2
33.4
43.0
32.7
18.4
34.5
Total did not complete highest level of school
33.2
36.2
38.7
39.3
34.3
44.3
34.8
19.2
35.6
0.0
0.0
0.1
0.1
0.0
0.2
0.1
0.1
0.0
47.8
50.7
53.6
54.9
50.5
56.4
48.8
36.9
50.4
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
Attending tertiary institution in May 2000
Still at school
Total without post-school qualifications
TOTAL
a As defined under the ABS (1993), Classification of Qualifications.
b Includes persons who never attended school.
c Includes persons whose study was not intended to result in a recognised educational qualification.
Source: ABS, Education and Work, May 2000, Cat. no. 6227.0, unpublished data
Overall these data indicate that New South Wales has the highest human capital stock
defined in terms of the distribution of educational qualifications. Victoria is close behind,
and out of the remaining states Western Australia seems to be the best performer.
Cross-sectional growth accounting
Over time, small differences in human capital can lead to persistent differences in
economic performance. Bhatta and Lobo’s (2000) cross-sectional growth accounting for
the United States illustrates the strong impact of human capital on GSP per capita. Using
New York State as the representative rich state, they found that human capital could
account for approximately 50-80% of the differences in GSP per capita among the US
states. It was also found that secondary qualifications contributed more to these differences
than degree attainment alone and that the results were not sensitive to using a different
representative state.
The Bhatta and Lobo decomposition computes a lower bound for the contribution of a given
factor of production to differences in GSP. It does this without the need for explicit
knowledge concerning the dimensions of other factors of production. In effect, this
approach measures the extent to which variations in labour quality are able to account for
– 166 –
Human Capital Investment and Economic Growth in the Australian Economy
differences in GSP per capita. The practical advantage of this strategy is that it allows us to
investigate structural differences in economic growth among the states without estimating a
full production function for each state.7
The Bhatta and Lobo approach assumes constant returns to scale and substitutability
between skill levels. Educational attainment and age are used as proxies for human capital
and experience respectively. Our estimates use New South Wales as the base case for
comparison. Full details of the empirical implementation of the model are given in
Appendix 7B. We use the data on labour force structure available in the 1996 Census. The
data for GSP per capita are taken from the Murphy model and are also reported for 1996.
The findings reported in Table 7.6 indicate that human capital plays a significant role in
explaining the relative economic performance of different states. The results in column (4)
indicate that 68.5% of the difference in GSP per capita between New South Wales and
Victoria can be explained by differences in human capital. Similarly, human capital can
explain 87.0% of the difference in GSP per capita between Queensland and New South
Wales. Sampling problems prevented the calculation of robust figures for Tasmania. Western
Australia is also a special case where the outcomes for GSP per capita are heavily influenced
by the capital-intensive resource industries that dominate economic activity in that state. This
makes it difficult to separate the effects of labour quality from overall GSP per capita.
Table 7.6: Human Capital and Economic Performance by State, 2000
(1)
GSP
per
capita
(2)
State as
proportion
of NSW
(from (1))
(3)
Difference
between state
and NSW
(from (1))
(4)
Difference
explained
by human
capital
(5)
Difference
explained
by degree
alone
$
%
%
..
..
(6)
(7)
Maximum
Maximum
achievable
achievable
GSP per capita GSP per capita
(equal human (proportional
capital)
to NSW)
$
%
(8)
Aggregate
gain in
state GSP
State
$
%
NSW
42,142
100.0
Vic
39,972
94.9
2,170
68.5
9.8
41,459
98.4
5.37
Qld
35,647
84.6
6,495
87.0
21.0
41,299
98.0
14.65
SA
33,725
80.0
8,417
45.1
19.3
37,518
89.0
4.45
WA
44,068
104.6
-1,925
..
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
..
..
..
..
..
..
$ billion
..
..
Gross state product per capita. Data represent 1996 values based on the Murphy model dataset.
Ratio of state GSP per capita (where each state is compared with New South Wales).
Differences in state GSP per capita (where each state is compared with New South Wales).
Indicates the proportion of the difference in GSP per capita that is explained by human capital per capita (where each state
is compared with New South Wales). That is, it is not explainable in terms of any other factor of production.
Indicates the proportion of differences in GSP that can be explained by differences in the stock of degree level and above
human capital (where each state is compared with New South Wales).
The maximum feasible GSP per capita if the level of human capital is made equal with New South Wales.
Ratio of maximum feasible GSP in each state compared with New South Wales.
Aggregate gains from maximum feasible GSP based on equal human capital.
Column (5) disaggregates this result in order to examine the role of human capital at the
degree level and above in explaining variations in GSP per capita. The contribution of
degree qualifications is not as strong as some might expect. Degree level qualifications
explain 9.8% of the variation in GSP per capita in Victoria and approximately 19-21% in
Queensland and South Australia. This indicates that the economic benefits of human capital
may reside in the general labour force in medium skill occupations. High skill qualifications
7
While data on state level human capital can be obtained from the census, comparable data on the different components of state
capital stocks are more difficult to estimate.
– 167 –
Productivity and Regional Economic Performance in Australia
and jobs are important but it is not necessary to have them dominate the labour force to
achieve good economic performance.
Columns (6) and (7) in Table 7.6 report the potential GSP per capita values that would result if
the levels of human capital in Victoria, Queensland and South Australia were made equal to that
in New South Wales. This is calculated by adjusting the GSP per capita gap in column (3) by
the information in column (4). The results show that Victoria could increase its GSP per capita
by approximately $1,487 if its level of human capital was equalised in this way. Interestingly,
Queensland has the potential to make major gains in GSP per capita by increasing its human
capital stock. Queensland would gain up to $5,652 per capita if it equalised its stocks with New
South Wales.
Column (8) shows how these figures translate into aggregate benefits. It is calculated as the
difference between column (6) and column (1) multiplied by the total population in each
state. These can be interpreted as hypothetical educational expansions that bring each state
up to a benchmark based on New South Wales levels of human capital. Again, the potential
benefits for Queensland are substantial – approximately $14.65 billion in additional GSP for
the state. In comparison, the potential benefits for Victoria and South Australia range
between $4.5 billion and $5.4 billion. It must be noted that these are calculated as
comparative static exercises. As a result they do not consider how the hypothesised
educational expansion would affect the overall demand and supply of qualifications.
The results in Table 7.6 are also contingent upon the development of appropriate industries
in which to employ human capital and, in this regard, they could be viewed as upper limits.
Thus, our results enable us to classify the growth potential and economic structure of each
state. The relatively low contribution of human capital to GSP per capita in South Australia
suggests that future growth in that state will need to be driven by the other factors that exist
on the capital side of the production function. In contrast, Queensland’s future growth
would seem to depend upon labour side developments. This is a surprising finding given
that, similar to Western Australia, Queensland’s economy has a large resource sector that
could be expected to dominate in explaining GSP per capita. The fact that labour quality
emerges as such a strong explanatory factor reinforces the need for Queensland to intensify
its efforts in human capital accumulation.
As discussed, these results are significant in the context of existing evidence on educational
expansion in Australia since 1970. They suggest that human capital can explain variations
in economic performance across different parts of the Australian economy. These estimates
also indicate that a number of state economies could benefit from educational expansion.
As such, there is a need to model the labour market effects of changing the aggregate demand
and supply of qualifications. Following our discussion in the first section, this needs to take
account of the effect of educational expansion on earnings (Borland, 1996) and occupational
structure (Vella and Karmel, 1999; Kidd and Meng, 1997). In future research this could be
addressed by applying decomposition techniques traditionally used to examine gender wage
differences to variations in state wage structures. These decompositions could be used to
separate the effects of educational attainment, occupational structure, the cost of living and
other characteristics on interstate variations in average wages.8
8
For example, it is possible that the techniques outlined by Juhn, Murphy and Pierce (1993) and Gomulka and Stern (1990) could
be adapted to the analysis of variations in interstate wage structures.
– 168 –
Human Capital Investment and Economic Growth in the Australian Economy
Conclusion
The need for a human capital investment policy
This chapter has explored the role of human capital and labour quality in influencing
economic growth in Australia. In particular, we have discussed the credibility of the skill
upgrading hypothesis as a guide to educational and economic policy development. This
hypothesis posits strong links between skill upgrading and labour market outcomes. In
particular, it holds that ‘a better educated labour force will produce more jobs at higher rates
of pay’ (Gregory, 1995). Our discussion above indicates that this hypothesis needs to be
substantially revised if it is to guide future education policy development. Specifically, the
assumption that educational expansion has the capacity to deliver uniform improvements in
conditions at all levels of the labour market needs to be replaced with a more realistic
analysis of labour market dynamics.
In turn, this new analysis can be used as the basis for a human capital investment policy as
suggested by Heckman (1998) for the United States. By this we mean a set of policies for
educational expansion and institutional change that directly consider what can be achieved
through education investment. While Australia has aligned its education policy to economic
needs, in the past this has been done in an arbitrary fashion. In the language of
macroeconomics, the targets and instruments were not identified with enough rigour.
Following this approach, we can identify two prerequisites for the development of a human
capital investment policy at the state or national level.
First, the use of educational expansion as a policy tool needs to be revised in line with a
more analytical treatment of labour market dynamics. Skill upgrading has differentiated
impacts on labour market outcomes. In turn, this undermines the effectiveness of arbitrary
expansions of educational attainment.
Put differently, the composition of educational expansion – that is, the relative emphasis on
qualifications by level (secondary or tertiary) or type (general or vocational) – can
decisively affect the flow-on to labour market outcomes. However, we must stress that this
does not represent a call to reinstate a set of deterministic manpower planning policies
targeting the demand and supply of particular qualifications. This tradition can be
summarised as the labour market equivalent of picking winners and continues to influence
policy thinking, particularly at the level of vocational training.
Instead, educational expansion needs to be motivated by a range of strategic questions
concerning the links between educational attainment, skill development and labour market
outcomes. How can skill upgrading policies be used to improve conditions in the low wage
sector? What role can further human capital investment play in closing the gender wage gap?
How can occupational status and earnings across the labour market be improved by educational
expansion? Is it possible to combat the rise in within-group inequality through education
investment? What levels of educational participation are needed to ensure that Australia
converges with the levels of human capital apparent in leading OECD economies? These are a
different set of questions to that put by the manpower planning approach, which continues to
inspire politicians on both the left and right. Rather than trying to re-fashion the microeconomic
– 169 –
Productivity and Regional Economic Performance in Australia
structure of the labour market piece by piece, policy makers need to adopt a systems
perspective on the links between education, skill development and labour market outcomes.9
Second, a focus on investment as opposed to financing problems needs to be adopted in the
development of new policies for human capital investment. The focus on education finance
in Australian policy debate is a legacy of the Higher Education Contribution Scheme
(HECS). In the 1980s, Australia led policy innovation in the area of education by using
aspects of human capital theory and applied research in labour economics to construct the
novel HECS scheme. As Chapman (1997) explains, HECS sought to balance the private and
public benefits of higher education. It also addressed equity arguments through a carefully
designed income-contingent student loans scheme.
Since its introduction, the HECS scheme has been adjusted with varying degrees of success
and has been a regular focus of education policy debate in Australia. Most recently, it has
inspired a new generation of proposals in the area of education finance. These proposals
have attempted to continue the policy innovation of HECS by exploring the interface
between private and public funding.
In contrast, we argue that this focus on financing has misinterpreted the innovative
contribution of HECS as an example of sophisticated policy design. Specifically, it has been
forgotten that the sophistication of HECS was a product of its origins in human capital theory
and applied labour economics. This rigorous background provided the impetus for the pivotal
income-contingent loans component of the scheme. The innovative contribution of HECS as
an example of sophisticated problem solving and policy design is what needs to be recognised
in current discussion rather than extensions of the scheme itself. In the next section we provide
two examples of this approach that focus on the composition of human capital investment.
Policies for human capital formation
The examples that follow are intended to illustrate the opportunities that exist to apply
insights from human capital theory and applied labour economics to education investment
problems. By focusing on the composition rather than the financing of human capital
investment, we aim to highlight the novel features of the HECS approach outlined above.
Our first example relates to the growth accounting results obtained in the second section
above. The decomposition of GSP per capita and human capital differences reported in this
section indicates there is scope to develop human capital investment policies at the state
level. This has major implications for state development strategies as they currently exist.
The direction of state and regional development policy has been dominated by strategies
that hinge on attracting physical and financial capital to particular geographical areas. The
value of such policies for restructuring the economy and boosting productivity can be
limited, particularly when they emphasise competition against other regions in attracting
existing economic resources.
These limitations highlight the need for state and regional governments to adopt explicit
policies for encouraging human capital formation. This is a viable alternative to existing
9
Metcalfe (2001) provides a precedent for such a systems perspective in his discussion of innovation policy. Foster (1999) also
offers an evolutionary macroeconomic perspective on the rise of unemployment and working poverty in OECD labour markets.
– 170 –
Human Capital Investment and Economic Growth in the Australian Economy
state development policies. State governments regularly implement business incentives and
subsidies that operate on the capital component of the production function. Yet, as our crosssectional growth accounting results indicate, the labour component of the production
function can explain variations on economic performance, suggesting that there is scope to
introduce parallel policies to attract, build and enhance human capital. As discussed, these
policies will need to take account of the labour market effects of increasing the supply of
qualifications. That is, policy development will need to be guided by empirical research on
the labour market structure that prevails in different states.
Our second example relates to the development of a comprehensive investment program for
the improvement of primary and pre-primary education. Heckman (1999) outlines a strong
argument for early intervention and investment in the process of human capital formation.
He argues for an investment strategy that lays strong foundations for the formation of
cognitive and non-cognitive skills in early childhood as a pathway to economic and social
success in adult life. Specifically, Heckman argues that early human capital investment
promotes later investment: ‘Non-cognitive skills and motivation are important determinants
of success and can be improved upon more successfully and at later stages than basic
cognitive skills’.
This argument for early investment in education has clear advantages from the perspective
of cost benefit analysis. By intervening to alleviate deficits created in the early part of the
life cycle, policy makers can avoid the large costs that are incurred in later years. That is, it
is relatively more expensive to correct skill deficits and restore lost capacities later in life
than it is to invest in prevention. This occurs for two reasons. First, early intervention
creates the prospect of recouping the economic costs of programs over a longer period of
time. Second, later interventions such as labour market and skill remediation programs face
the challenge of overcoming entrenched cognitive and non-cognitive skill deficits.
We believe that the potential of early intervention has been overlooked in recent discussions
of lifelong learning. These discussions have lacked an effective context and have
emphasised the extension of adult learning programs in an arbitrary fashion. If lifelong
learning policies are to be meaningful they must be based on an understanding of the
relationship between age, earnings and education. Practically, these policies must begin
with early intervention programs that will maximise the potential for ongoing learning and
skill upgrading through later stages in life. In this sense an intertemporal or labour
economics perspective is crucial.
A number of specific initiatives can be envisaged to capitalise on the benefits of early
intervention. Investment in selective and universal preschool programs have been
successful in the United States. Measured through to the age of 27, the Perry Pre-School
program has reported returns of $US5.70 for every dollar spent. Outcomes for the
universally based Headstart program have been less decisive and indicate that school
quality can play a role in sustaining the gains made at the preschool level (Heckman,
1998).10
10
The Queensland government’s recent Education and Training Reforms for the Future (ETRF) white paper provides an example of
this approach. The government is trialling a preschool year in a selected number of schools in 2003. If this trial is successful a
preparatory year at schools will be introduced for every child, replacing existing preschool education.
– 171 –
Productivity and Regional Economic Performance in Australia
Another specific initiative relates to how investments in school quality are made. Large
financial commitments to achieving marginal improvements in class size or expenditure per
student can be less effective than investments to enhance school quality from low levels in
a smaller range of schools. In this respect, the Education Action Zone (EAZ) policy
currently being tested by the United Kingdom government deserves further attention as a
policy to assist both economic development and social outcomes.11
Both of these examples illustrate how aspects of human capital theory and empirical labour
economics can be brought to bear on education investment problems. Educational or human
capital investment has been associated with financial expenditure for too long. A prime
example of this is the use of government expenditure on education as a proportion of GDP
as a benchmark for discussions of education quality.12 While financing is a necessary part
of the debate, it should not obscure the discussion of other important issues where policy
innovation and creative thinking are needed.
11
A full description of the EAZ policy is available at the Department for Education and Skills (DfES) website,
http://www.standards.dfee.gov.uk/eaz/
12
For example, the federal opposition’s Knowledge Nation report contained a heavy emphasis on expenditure data in its analysis
of Australia as an ‘underperforming knowledge nation’ (Chifley Research Centre, 2001).
– 172 –
Human Capital Investment and Economic Growth in the Australian Economy
References
Australian Bureau of Statistics (1993), Classification of Qualifications, Cat. no. 1262.0,
Canberra.
Australian Bureau of Statistics (2000), Labour Force, Australia, Cat. no. 6203.0, Canberra.
Australian Bureau of Statistics (2000), Education and Work, Australia, Cat. no. 6227.0,
Canberra.
Bhatta, D. and Lobo, J. (2000), ‘Human capital and per capita product: A comparison of
U.S. states’, Papers in Regional Science, 79, 393-411.
Borland, J. (1996), ‘Education and the structure of earnings in Australia’, The Economic
Record, 72 (219), 370-380.
Borland, J. (1999), ‘Earnings inequality in Australia: Changes, causes and consequences’,
The Economic Record, 75 (229), 177-228.
Borland, J. and Kennedy, S. (1998), Earnings Inequality in Australia in the 1980s and
1990s, Discussion Paper No. 389, Centre for Economic Policy Research, Australian
National University, Canberra.
Chapman, B. (1997), ‘Conceptual issues and the Australian experience with income
contingent charges for higher education’, The Economic Journal, 107 (May), 738-751.
Chifley Research Centre (2001), Report of the Knowledge Nation Taskforce,
http://www.alp.org.au/kn/kntreport_index.html. Last accessed 13 May 2002.
Durlauf, S. (2000), Econometric Analysis and the Study of Growth: A Skeptical
Perspective, Mimeo, University of Wisconsin, Madison.
Fernandez, E. and Mauro, P. (2000), The Role of Human Capital in Economic Growth:
The Case of Spain, IMF Working Paper No. WP/00/8.
Foster, J. (1999), ‘The rise of unemployment and working poverty: An evolutionary
macroeconomic perspective’, in M. Setterfield (ed.), Growth, Employment and
Inflation: Essays in Honour of John Cornwall, St Martins Press, New York.
Gemmell, N. (1996), ‘Evaluating the impacts of human capital stocks and accumulation on
economic growth: Some new evidence’, Oxford Bulletin of Economics and Statistics,
58 (1), 9-28.
Gomulka, J. and Stern, N. (1990), ‘The employment of married women in the United
Kingdom 1970-83’, Economica, 57 (226), 171-199.
Gregory, R.G. (1995), ‘Higher education expansion and economic change’, The
Australian Bulletin of Labour, 21 (4), 295-322.
Heckman, J. (1998), ‘What should be our human capital investment policy?’, Fiscal
Studies, 19 (2), 103-119.
– 173 –
Productivity and Regional Economic Performance in Australia
Heckman, J. (1999), Policies to Foster Human Capital, NBER Working Paper No.
W8239, August.
Juhn, C., Murphy, K. and Pierce, B. (1993), ‘Wage inequality and the rise in returns to
skill’, The Journal of Political Economy, 101 (3), 410-422.
Karmel, T. (1996), Educational Choice and the Gender Wage Gap in the Australian
Labour Market, Department of Employment, Training and Youth Affairs, Canberra.
Kidd, M. and Meng, X. (1997), ‘Trends in the Australian gender wage differential over
the 1980s: Some evidence on the effects of legislative reform’, Australian Economic
Review, 30 (1), 31-44.
Kidd, M. and Shannon, M. (2001), ‘Convergence in the gender wage gap in Australia over
the 1980s: Identifying the role of counteracting forces via the Juhn, Murphy and Pierce
decomposition’, Applied Economics, 33 (7), 929-936.
Kidd, M. and Shannon, M. (2002), ‘The gender wage gap in Australia: The path of future
convergence’, The Economic Record, 78 (241), 161-174.
Metcalfe, J.S. (2001), Evolutionary and Equilibrium Foundations for Technology Policy,
CRIC Discussion Paper, University of Manchester.
OECD (1998a), Education at a Glance: Indicators 1998, Centre for Educational Research
and Innovation, OECD, Paris.
OECD (1998b), Human Capital Investment: An International Comparison, Centre for
Educational Research and Innovation, OECD, Paris.
Preston, A. (1997), ‘Where are we now with human capital theory in Australia?’, The
Economic Record, 73 (220), 51-78.
Sheehan, P. and Esposto, A. (2001), ‘Technology, skills and earnings inequality’, in J.
Borland, B. Gregory and P. Sheehan (eds), Work Rich, Work Poor: Technology and
Economic Change in Australia, Centre for Strategic Economic Studies, Victoria
University, Melbourne.
Vella, F. and Gregory, R. (1996), ‘Selection bias and human capital investment: Estimating
the rates of return to education for young males’, Labour Economics, 3 (2), 197-219.
Vella, F. and Karmel, T. (1999), ‘Evaluating the impact of educational expansion on the
occupational status of youth’, Australian Economic Papers, 38 (3), 310-327.
Wolff, E. (2000), ‘Human capital investment and economic growth: Exploring the crosscountry evidence’, Structural Change and Economic Dynamics, (11) 4, 433-472.
– 174 –
Appendices1
Appendix 4 – Technical Exposition of Shift-Share Analysis Variants
Some notation needs to be introduced to facilitate differentiation between the various
extensions and reformulations of the basic shift-share model. The superscripts r and n index
the region and nation (or reference area) respectively. The subscript i indexes the industrial
sector. Q and g represent level of output (or gross regional product) and the rate of change of
output over the given time period respectively. If we let t0 be the start of the study period and
t1 be the end of the study period then, for example, gi can be defined as (Qi(t1) - Qi(t0)) / Qi(t0).
Traditional comparative static shift-share analysis
The national growth effect (NS) for sector i (i = 1,...,I) in region r is found as:
r
NSi =
r
Qi ng
(1)
where ng is the national rate of output growth for all sectors combined.
The proportionality shift (PS) for sector i in region r is based on differences between the
national growth rate for this sector and the overall national growth rate, and is found as:
r
PSi
=
r
Qi ( ngi - ng)
(2)
The differential shift (DS) for sector i in region r is based on differences between the
national and regional growth rate for this sector and is found as:
r
DSi =
r
Qi ( rgi - ngi)
(3)
The total shift (TS) for sector i in region r is found as:
r
TSi
=
r
Qi ( ngi - ng) + rQi ( rgi - ngi)
(4)
For each effect or shift, the individual sector values are summed to obtain the total (or net)
effect for the region. Thus, a region that has a large share of output generated in industrial
sectors that are slow (fast) growing nationally will have a net negative (positive)
proportionality shift. Similarly, a region with a negative (positive) differential shift will
have experienced lower (higher) net output growth than would have been expected on the
basis of its industrial structure.
Dynamic shift-share analysis
Equations (1) to (4) above are used to calculate the annual national growth effect,
proportionality shift, differential shift and total shift respectively for each industrial sector
and region. The ‘total period’ national growth effect, for example, is then calculated as the
sum of the corresponding annual national growth effects. Other total period effects are
calculated similarly.
1
For references in appendices, see relevant chapter.
– 175 –
Productivity and Regional Economic Performance in Australia
Simple productivity-extended shift-share analysis
Our adaptation of this Rigby and Anderson (1993) method uses the following additional
notation:
r
qit
=
r
Qit / rEit
(5)
(where rEit is employment in sector i, in region r at time t)
= average labour productivity in sector i, region r at time t.
r
Ait
= (rEit+1 - rEit). rqit
(6)
= change in gross domestic product (GDP) in sector i, region r that would
have been observed over the given time period if productivity had remained
constant and employment changed as observed.
r
ait
=
r
Ait / rQit
(7)
= rate of GDP change in sector i, region r resulting from variations in
employment over the given time period with productivity held constant.
r
Bit
= (rEit+1. rqit+1) - (rEit+1. rqit)
(8)
= change in GDP in sector i, region r that would have been observed over
the given time period if employment had remained constant and
productivity changed as observed.
r
bit
=
r
Bit / rQit
(9)
= rate of GDP change in sector i, region r resulting from variations in
productivity over the given time period with employment levels held
constant.
Then rgi = rai + rbi and these rates of change can be defined at the level of the industrial
sector, region or nation, and the various shift-share components defined in equations (1) to
(4) are now recalculated as follows:
r
NSi = rNSi (a) + rNSi (b) = (rQi . na) + (rQi . nb)
r
(10)
= rPSi (a) + rPSi (b) = (rQi (nai - na)) + (rQi (nbi - nb))
(11)
DSi = rDSi (a) + rDSi (b) = (rQi (rai - nai)) + (rQi (rbi - nbi))
(12)
PSi
r
r
TSi
= rTSi (a) + rTSi (b) = [rPSi (a) + rDSi (a)] + [rPSi (b) + rDSi (b)]
= [rQi (nai - na) + rQi (rai - nai)] + [rQi (nbi - nb) + rQi (rbi - nbi)]
– 176 –
(13)
Appendices
Multifactor productivity-extended shift-share analysis
We can redefine qit = qiLt + qiKt where qiLt = _i Qit /Eit and qiKt = (1 - _i )Qit /Eit where _i
is the cost share of labour input for sector i. By substituting these into equations (6) to (9)
we get:
r
and
and
AiL = (rEit+1 - rEit ) . rqiLt
r
AiK = (rEit+1 - rEit ) . rqiKt
r
r
r
r
(14)
AiL / rQi
AiK / rQi
aiL =
aiK =
(15)
r
and
BiL = (rEit+1 . rqiLt+1) - (rEit+1 . rqiLt )
BiK = (rEit+1 . rqiKt+1) - (rEit+1 . rqiKt )
r
(16)
r
and
biL = rBiL / rQi
biK = rBiK / rQi
r
(17)
We can then rewrite the shift-share equations (10) to (13) to investigate the impact of
labour-related change as:
r
NSiL = rNSi(aL ) + rNSi(bL ) = (rQi . naL) + (rQi . nbL )
(18)
r
(19)
PSiL = rPSi(aL ) + rPSi(bL ) = (rQi (naiL - naL)) + (rQi (nbiL - nbL ))
r
r
DSi(aL ) + rDSi(bL ) = (rQi (raiL - naiL)) + (rQi (rbiL - nbiL ))
DSiL =
(20)
r
TSiL = rTSi(aL ) + rTSi(bL )
= [ rQi (naiL - naL) + rQi (raiL - naiL )]
+ [ rQi (nbiL - nbL) + rQi (rbiL - nbiL)]
(21)
and to investigate capital or other factor-related change as:
r
NSiK =
r
NSi(aK) + rNSi(bK) = (rQi . naK) + (rQi . nbK)
r
r
PSiK =
r
r
n
r
r
r
r
r
r
r
TSiK =
(22)
r
n
r
r
n
PSi(aK) + PSi(bK) = ( Qi ( aiK - aK)) + ( Qi ( biK - bK))
r
DSiK =
n
n
n
DSi(aK) + DSi(bK) = ( Qi ( aiK - aiK)) + ( Qi ( biK - biK))
(23)
(24)
TSi(aK) + TSi(bK)
= [ rQi (naiK - naK) + rQi (raiK - naiK)]
+ [ rQi (nbiK - nbK) + rQi (rbiK - nbiK)]
(25)
Since reliable capital stock data are not available at the regional and sectoral level for the
period covered in this study, we follow Haynes and Dinc (1997) and calculate the GDP
change resulting from the contribution of capital (or other factors) to total factor
productivity as a residual, that is as the difference between actual GDP change in total
(given by rNSi + rPSi + rDSi) and the actual GDP change attributable to labour (given by
r
NSiL + rPSiL + rDSiL) rather than via use of equations (22) to (25).
– 177 –
Productivity and Regional Economic Performance in Australia
Appendix 5A – Törnqvist Index Methodology
This appendix outlines the Törnqvist index number method to estimating multifactor
productivity (MFP) and the related assumptions and limitations of this approach. The
ABS (2000, chapter 27) also uses Törnqvist index numbers to calculate its estimates of
‘market sector’ MFP at the national level.
The Törnqvist methodology calculates MFP by establishing a theoretical link between the
production function and index numbers, first pioneered by Solow (1956). To illustrate,
assume a production function of the form of equation (1), where Yt, Lt, Kt and At denote
output, labour, capital and the state of technology respectively. Taking the logarithmic
differential of this production function in (2) illustrates that output growth is equal to
technological progress plus the sum of the growth rates in labour and capital, weighted by
their output elasticities:
Yt = Atf(Lt, Kt )
.
.
.
At
,Yt L t
Lt
Yt
.
.
=
+
Yt
At
,L t Yt
Lt
(1)
,Yt
+
,Kt
.
Kt
.
Yt
.
Kt
(2)
Kt
In (2), the output elasticities of labour and capital are the only variables that are not directly
observable from national or state accounts. Yet, they can be identified if perfect competition
is assumed. Under perfect competition, each input will be paid its marginal product, which
in turn is equal to the input price relative to the price of output. This is shown in (3), where
wt, rt and pt denote the prices of labour, capital and output respectively:
,Yt
,Lt
,Yt
wt
=
and
pt
rt
(3)
=
,Kt
pt
Substituting these relative prices for their marginal products in (4) shows that the output
elasticities of labour and capital are equal to their income shares under perfect competition,
with these income shares readily available from national or state accounts. Rearranging as
in (5) shows that technological progress can thus be calculated as the residual of output
growth minus the growth in labour and capital, weighted by their income shares:
.
.
.
.
At
wtLt
Lt
rtKt
Kt
Yt
.
.
(4)
=
+
+
At
ptYt
Lt
ptYt
Kt
Yt
.
.
.
.
At
Yt
Lt
Kt
L
K
(5)
=
- st .
- st .
At
Yt
Lt
Kt
Note that (5) is expressed in continuous time. Thus, the Törnqvist index number approach
approximates (5) in discrete time by using average between-period income shares for labour
and capital, along with the growth rates in output, labour and capital, as in (6):
Yt
At
log
= log
At-1
Yt-1
(s Lt + s Lt -1)
-
. log
2
Lt
Lt-1
– 178 –
-
(sKt + sKt -1)
2
. log
Kt
Kt-1
(6)
Appendices
Crucially, resulting estimates of the residual in (6) are in practice interpreted as MFP
growth, not technological progress. This is because the methodology in (1) through to (6)
makes several assumptions, including that of perfect competition, and thus allocative and
technical efficiency, and also constant returns to scale, since the income shares of labour and
capital in the national or state accounts necessarily sum to unity (Coelli, 1998, pp. 37-38).
The resulting residual will only represent technological progress when all of these
assumptions hold. In practice, markets are imperfectly competitive, inefficiencies often
exist, along with economies or diseconomies of scale. Thus, the residual in (6) is interpreted
as MFP growth, reflecting that the productivity of labour and capital will not only be
influenced by technological change, but also by changes in efficiency, scale economies and
other factors.
The main limitation in estimating MFP growth as a residual is that measurement errors on
the right-hand side of (6) will be fully reflected in measured MFP growth. These
measurement errors could include incorrectly estimating labour or capital inputs or their
income shares. For instance, overestimating (underestimating) capital stock growth will
.
.
.
lead to . underestimating
(overestimating)
MFP growth. Similarly,
imperfect competition,
scale economies, regulatory distortions, fluctuations in capacity utilisation, rigidities or
adjustment costs will cause prices for capital and labour to deviate from their marginal
products. This in turn means that the income shares of labour and capital will not reflect
their true output elasticities (Morrison, 1998). These measurement problems highlight one
reason why the resulting MFP estimates should be taken as indicative of trends rather than
precise estimates (see also Appendix 5B).
In Chapter 5, there was some evidence of measurement bias in Western Australia. This state
had a much higher estimate on the income share of capital (42%) relative to other states
(33%). Rather than truly reflecting a much higher output elasticity in Western Australia, this
result may have been due to a large investment surge over the sample causing the income
share of capital to overestimate the output elasticity of capital in this state. This was most
evident when viewing the resulting estimate of MFP in Western Australia. Given that the
capital stock forms a much larger input than labour hours in (6), using such a high
capital–income share leads to an estimated level of MFP in Western Australia well below
that in any other state.
Where possible, adjustments should be made to correct for such measurement errors
.
.
.
.
(Morrison, 1998). Yet, this usually requires detailed disaggregated data in order to study the
extent and direction of any resulting bias. Such data are not available for the current study.
However, given Western Australia was the only outlier in this respect, the approach taken
in this study
was
. to set the .income shares
. of labour and capital in each state equal to the
.
Australian average income shares over the sample (65% and 35%). Theoretically, this
approach is equivalent to assuming capital and labour mobility equalises returns to factor
inputs across regions. Empirically, this assumption is implicit in panel regression estimates
of MFP, where an average output elasticity for both labour and capital is applied across all
regions in the panel.
Constraining the production function to be the same across all states will affect estimates of
MFP levels rather than estimates of MFP growth. Given this added measurement difficulty,
– 179 –
Productivity and Regional Economic Performance in Australia
the estimates of MFP levels should be regarded as less robust than the estimates of MFP
growth in this study. Having said this though, imposing the same production function across
all states influences the estimated level of MFP mainly in Western Australia. The MFP
levels in other states remain relatively unchanged, since the output elasticities of labour and
capital in these other states were closely gathered around the national average to begin with.
– 180 –
Appendices
Appendix 5B – Data Sources and Descriptions
This appendix outlines the sources and construction of the variables used in estimating MFP
at the state level and the variables used in the subsequent econometric analysis. All variables
were based in financial year terms over the period 1984-85 to 2000-01 (apart from business
expenditure on R&D, which was only available over 1984-85 to 1999-2000).
The calculation of state MFP required data on output, labour and capital inputs, and their
income shares. Output was defined as gross state product (ABS State Accounts 5220.0),
while labour was defined as total hours worked (ABS unpublished Labour Force 6202.0
statistics). The ABS only produces a national net capital stock series. Thus, state capital
stocks were constructed using the perpetual inventory method (PIM) in (1), where Kt
denotes capital stock at end of period, bt is the annual depreciation rate and It is annual real
investment expenditure:
Kt
=
(1 - bt)Kt-1 + It
(1)
However, there exist several measurement errors associated with constructing state capital
stocks that will be in turn reflected in resulting MFP estimates. This illustrates another
reason why MFP estimates should be taken as indicative of trends rather than exact
estimates (see also Appendix 7A), and also highlights the need for future work into
estimating state capital stocks. The current study’s approach and some of its limitations are
outlined below.
First, capital stocks were calculated on an aggregate basis rather than sector basis, given the
lack of disaggregated state level data. In this case, investment (It) in (1) was taken as
aggregate real investment (ABS State Accounts 5220.0), while the national depreciation rate
was used across all states, with this rate calculated by dividing the national consumption of
fixed capital into net national capital stock in the previous period (ABS National Accounts
5204.0). This type of aggregation cannot therefore filter out from resulting MFP estimates
the effect of any interstate differences in depreciation rates on capital stock estimates, or the
effect of shifts from less to more productive capital components within each state’s
aggregate capital stock.
Second, starting values for the capital stocks are required to initiate the PIM formula in (1).
These were calculated by making use of nominal investment data available by state over
1980-81 to 1983-84 (ABS SNA68 5204.0 and State Accounts 5220.0). In this case, a capital
value for 1983-84 in each state was estimated by taking each state’s share of nominal
investment nationally over 1980-81 to 1983-84 and then multiplying this by the national net
capital stock in 1983-84. This creates another possible measurement error, however,
whereby incorrectly estimating the initial capital value in any state will alter the level and
growth in that state’s capital stock.
Recursively solving (1) as in (2) shows that with a national average annual depreciation rate
of around 5.4% over the past 17 years, 39% of each initial state capital stock will remain
undepreciated in 2000-01 (0.390 = 0.94617). That is, any error in measuring the initial value
will still influence capital stock estimates 17 years later. In fact, investment series spanning
at least 50 years are required to depreciate more than 95% of the initial capital stock
(0.047 = 0.94655) and thus reduce the influence of any error in estimating the initial value
– 181 –
Productivity and Regional Economic Performance in Australia
to an acceptable size. Yet, such a time span on investment expenditure at the state level is
currently unavailable:
K2000-01 = (1 - b)17 . K1983-84 +
17
i
Y (1 - b) . It-1
(2)
I=0
Finally, the sum of the state capital stocks, YKI, derived from the PIM formula each year
was constrained to equal the national capital stock series by multiplying each state’s capital
stock, Ki, by the ratio KN/YKi. However, this made little change to the resulting capital stock
series, since the difference between the sum of the state capital stocks YKi originally derived
from the PIM formula and the national capital stock series was below 0.4% in every year of
the sample.
Income shares for labour (CL) and capital (CK) were calculated as the share of compensation
of employees and gross operating surplus in total factor income respectively, with an
adjustment made to extract the labour income of self-employed persons and employers
contained in gross operating surplus and add it to compensation of employees. The size of
this transfer was estimated by multiplying the share of self-employed persons and
employers in total employment to gross operating surplus (ABS State Accounts 5220.0;
ABS unpublished Labour Force 6202.0 statistics).
For the econometric analysis, the R&D stocks were also calculated using the PIM formula in
(1), with It in this case representing estimated annual real business R&D in each state (based
on ABS Research and Experimental Development, Businesses 8104.0). A depreciation rate
of 10% was chosen, following the Industry Commission’s (1995, QA. 25) finding in its
survey of R&D literature that this was the most often used rate of depreciation. The initial
value for the R&D stock in this case was calculated as in (3), where I0 represents business
R&D in the initial year, gi represents average annual growth in business R&D over the
sample and is the assumed depreciation rate (following Griliches, 1980, p. 345; Coe and
Helpman, 1993, p. 883):
K0
I0
(3)
=
(gi + b)
The tariff rate in each state was calculated by dividing customs duty (ABS unpublished
National Accounts 5204.0 statistics) into total imports (ABS State Accounts 5220.0). The
rate of industrial disputation was calculated as the number of working days lost due to
industrial disputes (ABS Industrial Disputes 6321.0) per 100 wage and salary earners (ABS
unpublished Labour Force 6202.0 statistics). The high school retention rate in each state
was calculated as the proportion of students in their first year of high school that remained
in high school until year 12 (ABS Schools 4221.0). The capacity utilisation variable was
calculated as the residual from a regression of the log of output in each state on a time trend
(following Dowrick and Nguyen, 1989, p. 1013; Chand, 1999, p. 33).
– 182 –
ADF(0) = -3.007
PP(3) = -8.073***
ADF(0) = -8.791***
PP(3) = -8.936***
ADF(0) = -7.437***
PP(3) = -7.433***
VIC/QLD
VIC/SA
ADF(0) = -8.752***
PP(3) = -8.694***
NSW/SA
NSW/TAS
ADF(0) = -7.808***
PP(3) = -7.808***
NSW/QLD
ADF(0) = -7.331***
PP(3) = -7.332***
ADF(0) = -8.584***
PP(3) = -8.530***
NSW/VIC
NSW/WA
Unit root tests
on residuals a
Pairwise case
– 183 –
F calc = 6.907***
F calc = 1.988
F calc = 8.856***
F calc = 1.949
F calc = 5.408**
F calc = 5.129**
F calc = 4.406*
Unit root joint F test:
null of pure unit
root process b
F calc = 9.823***
F calc = 2.877
F calc = 12.793***
F calc = 2.896
F calc = 7.311**
F calc = 7.691**
F calc = 6.505*
Unit root joint F test:
null of unit root
process with drift b
F calc = 12.350[0.001]***
f
Bootstrap critical values:
95% = 4.625 99% = 8.079
F calc = 1.378[0.245]
F calc =18.05[0.000]***
Bootstrap critical values: e
95% = 4.695 99% = 8.146
F calc = 1.384[0.244]
d
Bootstrap critical values:
95% = 3.901 99% = 5.446
F calc = 9.047[0.004]***
c
Bootstrap critical values:
95% = 4.927 99% = 8.064
F calc = 3.373[0.072]*
F calc = 3.290[0.075]*
Redundancy test on
time trend
Table 6A.1: Tests for Catching Up and Convergence,
State GSP Per Capita
Appendix 6A – Detailed Convergence Results
Long-run convergence
Long-run convergence
Catching up
Long-run convergence
Catching up
Long-run convergence
Long-run convergence
Conclusion
Appendices
Unit root tests
on residuals a
ADF(0) = -2.654
PP(3) = -7.615***
ADF(0) = -7.909***
PP(3) = -7.977***
ADF(0) = -8.594***
PP(3) = -8.623***
ADF(0) = -7.183***
PP(3) = -7.164***
ADF(0) = -7.523***
PP(3) = -7.535***
ADF(0) = -7.949***
PP(3) = -7.949***
ADF(0) = -2.967
PP(3) = -7.973***
Pairwise case
VIC/TAS
SA/QLD
SA/WA
SA/TAS
WA/QLD
– 184 –
WA/TAS
TAS/QLD
F calc = 4.314*
F calc = 3.051
F calc = 4.299*
F calc = 6.715***
F calc = 1.815
F calc = 2.534
F calc = 6.304**
Unit root joint F test:
null of pure unit
root process b
F calc = 6.027*
F calc = 4.415
F calc = 6.416*
F calc = 10.068***
F calc = 2.521
F calc = 3.045
F calc = 9.196***
Unit root joint F test:
null of unit root
process with drift b
F calc = 7.180[0.010]***
Bootstrap critical values: i
95% = 5.082 99% = 8.889
F calc = 4.596[0.036]**
Bootstrap critical values: h
95% = 3.687 99% = 5.216
F calc = 2.142[0.149]
F calc = 4.892[0.031] **
F calc = 1.551[0.218]
Bootstrap critical values: h
95% = 4.372 99% = 6.090
F calc = 2.601[0.112]
F calc = 10.95[0.002]***
Bootstrap critical values: g
95% = 4.746 99% = 8.234
Redundancy test on
time trend
Table 6A.1 (continued) : Tests for Catching Up and Convergence,
State GSP Per Capita
Catching up
Catching up
Long-run convergence
Catching up
Long-run convergence
Long-run convergence
Catching up
Conclusion
Productivity and Regional Economic Performance in Australia
n
k=1
= µ + _ (yi,t-1 - yj,t-1) + `t + Y bk *¨ (yi,t-k - yj,t-k) + ¡t
(1)
– 185 –
i Bootstrap critical values were calculated because the residuals from the estimated model failed the Breusch-Godfrey serial correlation LM
test (at low and moderate lag lengths) and the Ramsey reset (4) test at the 1% level of significance.
h Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test, the White
heteroscedasticity tests and the Ramsey reset (4) test at the 1% level of significance.
g Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test, the
Breusch-Godfrey serial correlation LM test (at low and moderate lag lengths) and the Ramsey reset (4) test at the 5% level of significance.
f Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 1%
level of significance.
e Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test and the
Ramsey reset (4) test at the 1% level of significance.
d Bootstrap critical values were calculated because the residuals from the estimated model failed the White heteroscedasticity tests activated
in the EVIEWS 3.1 Econometric Package. The Ramsey reset (4) test is also marginal at the 5% level of significance.
c Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test and the
Ramsey reset (4) test at the 1% level of significance.
It should be noted that the critical values for both joint F statistics were determined from simulation by the authors.
The second joint F test is based on a null hypothesis: Ho = (µ,_,`) = (µ,1,0), assuming k = 0 in (1). This equation depicts, under the null
hypothesis, a model of a unit root process with drift. The results of this test are shown in the fourth column.
The first joint F test is based on a null hypothesis: Ho = (µ,_,`) = (0,1,0), assuming k = 0 in equation (1) which, under the null hypothesis,
is essentially a model of a unit root process without drift. The results of this test are shown in the third column.
yi,t - yj,t
b The model framework adopted was:
a The lag length chosen for the ADF test corresponds to the lag length that gives the smallest values for the Akaike information criteria and
Schwarz criteria. The lag length for the Phillip–Perron (PP) on test was chosen according to the automatic selection criteria activated in
the EVIEWS 3.1 Econometric Package. Note further that the superscripts *, ** and *** presented after the calculated values of the test
statistic denote rejection of the respective null hypotheses at the 10%, 5% and 1% levels of significance respectively.
Appendices
ADF(3) = -5.301***
PP(3) = -7.970***
ADF(0) = -8.455***
PP(3) = -8.502***
ADF(0) = -8.440***
PP(3) = -8.441***
ADF(0) = -7.894***
PP(3) = -7.924***
VIC/QLD
VIC/SA
VIC/WA
ADF(0) = -9.187***
PP(3) = -9.100***
NSW/SA
NSW/TAS
ADF(0) = -7.450***
PP(3) = -7.449***
NSW/QLD
ADF(3) = -3.841**
PP(3) = -7.147***
ADF(0) = -8.193***
PP(3) = -8.176***
NSW/VIC
NSW/WA
Unit root tests
on residuals a
Pairwise case
– 186 –
F calc = 2.422
F calc = 3.129
F calc = 4.748*
F calc = 6.102**
F calc = 2.205
F calc = 4.269*
F calc = 9.327***
F calc = 6.985***
Unit root joint F test:
null of pure unit
root process b
F calc = 3.633
F calc = 4.370
F calc = 7.110**
F calc = 9.026**
F calc = 3.298
F calc = 6.054*
F calc = 13.989***
F calc = 10.447***
Unit root joint F test:
null of unit root
process with drift b
F calc = 0.760[0.387]
Bootstrap critical values: g
95% = 3.857 99% = 5.478
F calc = 2.921[0.093]*
Bootstrap critical values: f
95% = 3.886 99% = 5.440
F calc = 0.589[0.446]
F calc = 7.349[0.009]***
Bootstrap critical values: e
95% = 4.685 99% = 8.215
F calc = 1.170[0.284]
Bootstrap critical values: d
95% = 3.867 99% = 5.458
F calc = 3.185[0.080]*
Bootstrap critical values: c
95% = 3.631 99% = 5.129
F calc = 2.167[0.146]
F calc = 0.003[0.960]
Redundancy test on
time trend
Table 6A.2: Tests for Catching Up and Convergence,
State GSP Per Person Employed
Long-run convergence
Long-run convergence
Long-run convergence
Catching up
Long-run convergence
Long-run convergence
Long-run convergence
Long-run convergence
Conclusion
Productivity and Regional Economic Performance in Australia
– 187 –
ADF(0) = -7.880***
PP(3) = -7.950***
ADF(8) = -3.889**
WA/TAS
TAS/QLD
PP(3) = -7.257***
ADF(6) = -7.547***
PP(3) = -7.549
ADF(0) = -8.914***
PP(3) = -9.002***
SA/WA
WA/QLD
ADF(0) = -8.399***
PP(3) = -8.503***
SA/QLD
ADF(0) = -6.075***
PP(3) = -6.424***
ADF(1) = -6.356***
PP(3) = -6.787***
VIC/TAS
SA/TAS
Unit root tests
on residuals a
Pairwise case
F calc = 4.194
F calc = 2.292
F-calc = 4.060
F calc = 7.168***
F calc = 1.694
F calc = 2.655
F calc = 6.201**
Unit root joint F test:
null of pure unit
root process b
F calc = 6.195*
F calc = 3.391
F-calc = 6.086*
F calc = 10.751***
F calc = 2.429
F calc = 3.712
F calc = 9.221***
Unit root joint F test:
null of unit root
process with drift b
Bootstrap critical values: l
95% = 3.516 99% = 4.994
F calc = 4.527[0.038]**
F calc = 1.645[0.205]
Bootstrap critical values: k
95% = 3.966 99% = 5.525
F calc = 2.373[0.129]
Bootstrap critical values: j
95% = 3.433 99% = 4.868
F calc = 3.055[0.086] *
Bootstrap critical values: i
95% = 3.140 99% = 4.498
F calc = 0.548[0.462]
Bootstrap critical values: h
95% = 4.618 99% = 6.315
F calc = 1.692[0.199]
F calc = 8.567[0.005]***
Redundancy test on
time trend
Table 6A.2 (continued): Tests for Catching Up and Convergence,
State GSP Per Person Employed
Catching up
Long-run
convergence
Catching up
Long-run
convergence
Long-run c
convergence
Long-run
convergence
Catching up
Conclusion
Appendices
n
(1)
– 188 –
g Bootstrap critical values were calculated because the residuals from the estimated model failed the White heteroscedasticity tests at the 1%
level of significance.
f Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 5%
level of significance and were marginal at the 1% level of significance.
e Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 5%
level of significance and the Ramsey reset (4) test at the 1% level of significance.
d Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 5%
level of significance.
c Bootstrap critical values were calculated because the residuals from the estimated model failed the White heteroscedasticity tests at the 5%
level of significance.
It should be noted that the critical values for both joint F statistics were determined from simulations undertaken by the authors.
The second joint F test is based on a null hypothesis: Ho = (µ,_,`) = (µ,1,0), assuming k = 0 in (1). This equation depicts, under the null
hypothesis, a model of a unit root process with drift. The results of this test are shown in the fourth column.
The first joint F test is based on a null hypothesis: Ho = (µ,_,`) = (0,1,0), assuming k = 0 in equation (1) which, under the null hypothesis,
is essentially a model of a unit root process without drift. The results of this test are shown in the third column.
k=1
yi,t - yj,t = µ + _ (yi,t-1 - yj,t-1) + `t + Y bk *¨ (yi,t-k - yj,t-k) + ¡t
b The model framework adopted was:
a The lag length chosen for the ADF test corresponds to the lag length that gives the smallest values for the Akaike information criteria and
Schwarz criteria. The lag length for the Phillip–Perron test was chosen according to the automatic selection criteria activated in the EVIEWS
3.1 Econometric Package. Note further that the superscripts *, ** and *** presented after the calculated values of the test statistics denote
rejection of the respective null hypotheses at the 10%, 5% and 1% levels of significance respectively.
Productivity and Regional Economic Performance in Australia
l Bootstrap critical values were calculated because the residuals from the estimated model failed the Breusch-Godfrey serial correlation LM
test (at low and moderate lag lengths) and the Ramsey reset (4) test at the 1% level of significance.
k Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test and the
Ramsey reset (4) test at the 1% level of significance.
j Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 5%
level of significance and the Ramsey reset (4) test at the 1% level of significance.
i Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test at the 5%
level of significance and were marginal at the 1% level of significance.
h Bootstrap critical values were calculated because the residuals from the estimated model failed the Jarques-Bera normality test and the
White heteroscedasticity tests at the 1% level of significance, and the Ramsey reset (4) test at the 5% level of significance.
Appendices
– 189 –
Productivity and Regional Economic Performance in Australia
Appendix 6B – Bootstrap Algorithm for Time Trend
Redundancy Test
The initial regression equation:
yi,t - yj,t = µ + _ (yi,t-1 - yj,t-1) + `t +
n
Y bk ¨ (yi,t-k - yj,t-k) + ¡t
(1)
k=1
was estimated for each pairwise combination of states. For the purpose of deriving the
redundant variable test statistics, equation (1) can be viewed as the unrestricted model. The
restricted model is given by:
yi,t - yj,t = µ + _ (yi,t-1 - yj,t-1) +
n
Y bk ¨ (yi,t-k - yj,t-k) + ¡t
(2)
k=1
The unrestricted model is estimated by OLS and the residuals are calculated. These
residuals will be used later in the bootstrap algorithm.
The restricted model is then estimated. Two tests for redundant variables are then
calculated.1 The first is the Wald test:
W = ((RSSr - RSSu) / (k-m)) / (RSSu / (T - k))
(3)
The second is the likelihood ratio test:
L = -2 (LRr - LRu)
(4)
where RSS denotes residual sum of squares and LR denotes log likelihood with subscripts u
and r denoting unrestricted and restricted models respectively. Note k, m and T represent the
number of estimated coefficients in the unrestricted and restricted models and sample size
respectively. In the current case, k= 3 and m= 2, implying that the number of restrictions
(k-m) equals 1.
Under the null hypothesis that the k-m restrictions are redundant – that is, coefficients
m+1,...,k in the unrestricted model are not significant – the above tests will be distributed
as F(k-m,T-k) and Chi-square (k-m) respectively. This distribution will be exact provided the
residuals are distributed as normal iid variates. If this is not the case, these distributions will
be asymptotic provided certain conditions as discussed in Hamilton (1994, chapter 8) are
fulfilled.
Bootstrap techniques are employed to assess the sensitivity of test outcomes to departures
from normality and homogeneity (independence and homoscedasticity requirements) implied
by the distributional conditions under which the above test statistics have an exact F and
Chi-square distribution. The bootstrap algorithm employed is based on the generation of
artificial data using re-sampling with replacement from the original set of residuals from the
unrestricted model. Before commencing re-sampling, the original set of residuals is
standardised by subtracting the mean from each ‘one step’ residual and then dividing by the
standard deviation. This operation ensures that the set of standardised residuals are centred
and have unit variance.
1
The tests were performed using the EVIEWS software package (see Quantitative Micro Software, 2001).
– 190 –
Appendices
Following this operation, the residuals are randomly re-sampled using a shuffle algorithm
that uses the RAN2 pseudo random number generator contained in Press et al. (1992, pp.
272-3). This shuffling algorithm essentially produces random permutations of the initial
indexing of the residuals, thereby implementing random sampling of the initial set of
residuals with replacement.
Once the set of residuals is randomly re-sampled, a new set of data for the dependent
variable was generated. The initial value for the lagged dependent variable is set to its
historical value. Then for the first time period, the following procedure is used to generate
the new value for the dependent variable:
y(1) = µ + _ * y(0) + `t + w(1)
(5)
where µ, _ and ` are the estimated coefficients from the original unrestricted regression and
t is the time trend. The term w(1) is the first element in array w of the re-sampled residuals.
Finally, y(0) is the initial value for the lagged dependent variable that is set equal to its
historical value. Data for other time periods are then generated using the relation:
y(i) = µ + _ * y(i-1) + `t + w(i)
(6)
for i = 2,..., T .
We then define the new dependent and lagged dependent variable series as y(2),...,y(T) and
y(1),...,y(T-1) respectively, and then re-estimate equations (1) and (2) and calculate a new
set of values for the redundant variable tests outlined in equations (3) and (4) using these
new data series. The completion of this process is termed a replication.
Bootstrap critical values can be computed after performing a large number of replications.
Often, the total number of replications used is in the range of 100,000 to 200,000, with this
figure being deemed necessary to adequately enumerate the tails of the empirical
distribution functions (EDF) of the test statistics. The values of the test statistics are
typically stored in arrays during the replications loop that have dimension equal to the total
number of replications. After the replications loop is finished, the test statistic values are
sorted in ascending order and desired quantiles of the empirical distribution implied by the
sorted values are calculated, typically the 90%, 95% and 99% quantiles. These quantiles
provide bootstrap estimates of the desired critical values of the EDF of the test statistics.
– 191 –
Productivity and Regional Economic Performance in Australia
Appendix 7A – Human Capital Indicators
Educational attainment is the core component of aggregate measures of human capital. In
many economic models, educational attainment is assumed to embody the production of
relevant skills present in the labour force. Those individuals who have acquired a greater
quality and quantity of education are assumed to be higher in the ‘skills hierarchy’ that
progresses from the primary to tertiary level. The aggregation of these different levels of
skill is then achieved by weighting different segments of the labour force according to their
educational attainment.
We define two measures of human capital growth: (a) measure of the stock of educational
attainment in the economy and (b) a labour income based measure of the human capital
stock.
Educational attainment
First, following Gemmell (1996) we attempt to track changes in human capital stock based
on the secondary and tertiary enrolment rates of new groups entering the labour force. At
time t, aggregate human capital at time T can be defined as:
HT = H0 + Y _tdLt + Y (`t - _t) Rt
(1)
where: `t is the human capital embodied in the new entrants in period t
Nt is the number of new entrants to the labour force in period t
_t is the human capital embodied in retirees in period t
Rt is the number of retirees in period t
dLt is the change in the labour force (Nt-Rt) in period t.
For the purposes of compiling this index, for any time period, ` can be represented as the
relevant education enrolment rates for the group entering the labour market, while _ is the
enrolment rates present when current period retirees would have first entered the labour
market. A typical assumption is that the average working life is 40 years. Hence the
enrolment rates relevant to retirees are 40 year lagged enrolment rates (_t ~ `t-40).
Data on entry and exit (variables N and R) from the labour force are not readily available,
although information on net changes (dL) are available. Hence we have to approximate
human capital levels at time T as:
HT ~ H0 + Y `tdLt
(2)
UNESCO data on secondary and tertiary enrolment rates are used to determine the relevant
` values over time. Due to the lack of a yearly time series on enrolment rates, the Gemmell
index for Australia is calculated at discrete five year intervals from 1970 to 1995. Point
estimates of the age structure of the labour force (ABS Labour Force 6203.0) were used to
determine net changes to labour force size. Together this allows us to construct an index of
educational attainment for Australia over the period 1970 to 1995.
– 192 –
Appendices
Labour income-based index
A shortcoming of the previous approach is that it provides no direct quantification of the
economic value of education. Due to issues such as over-education and falling aggregate
returns to human capital, this may overstate the economic impact of educational expansion.
Labour income-based indices provide a direct link between the growth of educational
qualifications in the labour force and the changing aggregate returns to education.
Labour ‘products’ are classified by age and educational attainment, and weighted according
to their average wage rate, where it is assumed that the average wage rate is a reasonable
proxy for the marginal product of labour, and where all weights are computed with the
‘females 15-24 years no post-school qualifications’ category taken as the base value (e.g.
w31 = average wage ‘females 35-44 years no post-school qualifications’ / average wage
‘females 15-24 years no post-school qualifications’) (see Table 7A.1).
By definition then, w11 = 1.00
Using these weights in conjunction with data on labour force growth over the period 1970
to 1995, we can define the average level of human capital at time t as:
h(t) = Y wij . nij (t)
(3)
ij
where: nij (t) is the proportion of individuals with schooling i and in age group j, at time t
wij is the corresponding weight given by the 1996 earnings ratio in Table 7A.1.
Table 7A.1: Labour Force by Age and Educational Attainment
15-24
25-34
35-44
45-54
55-64
No post-school qualifications
w11
w21
w31
w41
w51
Post-school qualifications
w12
w22
w32
w42
w52
No post-school qualifications
w13
w23
w33
w43
w53
Post-school qualifications
w14
w24
w34
w51
w54
Females
Males
– 193 –
Productivity and Regional Economic Performance in Australia
Appendix 7B – Cross-sectional Growth Accounting
Bhatta and Lobo methodology
The Bhatta and Lobo (2000) methodology provides the basis for analysing the contribution
of human capital to differences in gross state product (GSP).
We define GSP as:
GSP = f (X0,X1,...,Xn)
(1)
where: GSP is gross state product
X0 is the size of the labour force
Xj is the quantity of the jth factor of production, j=1,...,n where n is the total
number of input factors.
It is assumed that (1) is a constant returns to scale function, but as explained by Bhatta and
Lobo (2000, p. 397), modifying this assumption to allow for spatial increasing returns (i.e.
agglomeration) would only strengthen the argument that human capital clustering in richer
regions/states is a major determinant of inter-regional productivity differentials.
Assumptions are as follow:
• a Cobb–Douglas production function (i.e. constant elasticity of substitution or CES),
although the CES property is not fundamental;
• the richer state is endowed with more of every factor of production per capita; and
• individuals with less than high school qualifications work as unskilled labour.
To implement this method, we first assume that Australian states’ production functions are
identical, and differences in state production are driven by differences in factor
endowments. On this basis, if we know the per capita levels and the marginal products of
human capital (proxied by average wage levels), we can compute the minimum difference
in GSP per capita that would result between two states as a result of disparities in human
capital stocks as:
m
1
2
y1 - y2 > + Y (MPi1 . xi - MPi2 xi )
i=1
where y1 is GSP per capita in region i.
Calculating this index requires the collection of age group, education and
wage data of the form displayed in Table 7B.1 (for each state).
– 194 –
Appendices
Table 7B.1: Labour Force by Age and Qualification
Age
High school
Diploma
Degree
15-19
Population
Wage
P11
W11
P21
W21
P31
W31
20-24
Population
Wage
P12
W12
P22
W22
P32
W32
25-34
Population
Wage
P13
W13
P23
W23
P33
W33
35-44
Population
Wage
P14
W14
P24
W24
P34
W34
45-54
Population
Wage
P15
W15
P25
W25
P35
W35
55-64
Population
Wage
P16
W16
P26
W26
P36
W36
Wij is the total state wages for education category i and age group j
Pij is the total state population for education category i and age group j.
Alongside this, GSP per capita figures are required. In this case, we have used the 1%
sample of the 1996 Census and the Murphy model GSP data respectively.
Splitting human capital into differing age and qualification groups explicitly measures
heterogeneous labour inputs.
These data can be used to approximate the marginal product labour for each age–education
category. Critically, persons with diploma or degree level education will also have the skills of
individuals with high school education. To compute the total marginal product of no post-school
education human capital, we must take the sum of the population fractions in every education
level for a given age group. Hence the total level of high school level education equals:
HighSchj =
Yij Pij
(2)
P
where HighSchj is the proportion of the total population (P) who are in age category j and
possess high school level human capital, that is i = 1.
For both diploma and degree based human capital, proportions are calculated as the number
of individuals in age category j and education category i divided by the total population.
The marginal product for each age–education category is simply the average wage for that
category:
MPij =
Wij
Pij
(3)
– 195 –
Productivity and Regional Economic Performance in Australia
Hence the marginal product for each age–education category, and its contribution towards
GSP per capita is:
MPij =
Wij
.x
Pij
(4)
where x is the proportion of the population in each age–education category.
For any two states this, together with the interstate difference in GSP per capita, can be used
to determine the minimum difference explained by human capital stocks.
m
(MPi .xi - MPi .xi )
Y
i=1
1
% explained difference =
1
y1
-
2
y2
where y refers to GSP per capita figures.
– 196 –
2
(5)
1SPEVDUJWJUZBOE3FHJPOBM
&DPOPNJD1FSGPSNBODFJO"VTUSBMJB
1SPEVDUJWJUZHSPXUIJTUIFNBJOGBDUPSDPOUSJCVUJOHUPJNQSPWFNFOUT
JONBUFSJBMMJWJOHTUBOEBSETBOETVTUBJOBCMFFDPOPNJDHSPXUIBDSPTT
FDPOPNJFT.VDIBUUFOUJPOIBTCFFOQBJEUPUIFBDDFMFSBUJPOJO
QSPEVDUJWJUZHSPXUIJO"VTUSBMJBPWFSUIFQBTUEFDBEFBOEBIBMGBOE
JOQBSUJDVMBSUIFA/FX&DPOPNZFSBBTTPDJBUFEXJUIUIFQFSJPETJODF
UIFNJET)PXFWFSMFTTDPOTJEFSBUJPOIBTCFFOHJWFOUPUIF
EJTUSJCVUJPOPGUIFTFOBUJPOBMQSPEVDUJWJUZHBJOTBDSPTT"VTUSBMJBOTUBUFT
1SPEVDUJWJUZBOE3FHJPOBM&DPOPNJD1FSGPSNBODFJO"VTUSBMJBBEESFTTFT
UIJTLFZJTTVFCZCSJOHJOHUPHFUIFSBDPMMFDUJPOPGQBQFSTCZBOVNCFS
PGFNJOFOU2VFFOTMBOEBDBEFNJDFDPOPNJTUTFYBNJOJOHUIFESJWFST
PGQSPEVDUJWJUZBOEFDPOPNJDQFSGPSNBODFBDSPTT"VTUSBMJBOTUBUFT
PWFSUIFQBTUEFDBEFBOEBIBMG5IJTCPPLSFnFDUTUIFSFTVMUTPGUIF
mSTUTUBHFPGUIF%SJWFSTPG&DPOPNJD(SPXUIQSPKFDUBDPMMBCPSBUJWF
VOEFSUBLJOHJOWPMWJOH2VFFOTMBOE5SFBTVSZUIF6OJWFSTJUZPG
2VFFOTMBOEBOE(SJGmUI6OJWFSTJUZ
5IF%SJWFSTPG&DPOPNJD(SPXUIQSPKFDUXBTJOJUJBUFEUPQSPWJEFBCFUUFS
VOEFSTUBOEJOHPGUIFGBDUPSTUIBUIBWFESJWFO2VFFOTMBOETGBTUFS
FDPOPNJDHSPXUISFMBUJWFUPUIFSFTUPG"VTUSBMJBPWFSUIFQBTUEFDBEF
BOEBIBMGBOENPTUJNQPSUBOUMZUPJEFOUJGZUIFLFZGBDUPSTUIBUXJMM
ESJWFUIF4UBUFTFDPOPNJDHSPXUIJOUIFGVUVSF
5IFSFTFBSDIJOUIJTCPPLDPOmSNTUIFJNQPSUBODFPGUIFQPMJDZ
BQQSPBDIFTPGUIF(PWFSONFOUJOFOIBODJOHQSPEVDUJWJUZHSPXUIBOE
MJWJOHTUBOEBSEToJOOPWBUJPOFEVDBUJPOBOETLJMMTUFDIOPMPHZBOE
SFTFBSDIBOEEFWFMPQNFOU*UXJMMUIFSFGPSFCFVTFGVMBOEJOUFSFTUJOH
UPBMMXIPIBWFBOJOUFSFTUJOUIFFDPOPNJDQSPHSFTTPG2VFFOTMBOE
XIFUIFSJOUIFQSJWBUFTFDUPSmOBODJBMNBSLFUTPSJOUIFQVCMJDTFDUPS
0GGJDFPG&DPOPNJDBOE
4UBUJTUJDBM3FTFBSDI
"QPSUGPMJPPGGJDFPG
2VFFOTMBOE5SFBTVSZ
XXXPFTSRMEHPWBV