The Role of Exports in Regional Economies
A System Dynamics model of the Western Australian
economy
William Grace
Australian Urban Design Research Centre
University of Westem Australia
PO Box 2729
Cloisters Square PO
PERTH WA 6850
+61 8 6318 6200
bill.grace@uwa.editau
Abstract
The U.S. cluster mapping project has highlighted the importance of export industries to
the performance of regional economies. Understanding these traded clusters and their
linkages with the local clusters that mainly serve local demand is an important element
in developing intervention strategies that seek to improve economic performance at the
regional scale. The economy of Western Australia is an example of a regional economy
that is very dependent on the export of resource commodities, and has benefitted
strongly from the massive expansion in the China economy since the early 2000s,
representing the state’s Black Swan moment. A simple system dynamics model of the
traded and local elements of the economy derived from input-output analysis has been
developed to gain insights into possible trajectories for the state’s economy and income
per capita. The results exemplify the situation of many norm-renewable resource based
export dominated economies that face long term challenges in replacing the value of
those exports.
Introduction
There is increasing interest in the policy treatment of regional economies in general and
of clusters in particular. This can be attributed in no small way to the influence of
Michael Porter’s The Economic Performance of Regions (Porter, 2003), itself a child of
his earlier book The Competitive Advantage of Nations (Porter, 1990). Porter and his
team at the Harvard Business School, in conjunction with MIT Sloan, the Temple Fox
School of Business and the U.S. Economic Development Administration manage the
USS. Cluster Mapping Portal’ which ‘provides a broader data infrastructure that covers
cluster presence and performance, as well as several types of data about regional
economic performance, business environment quality, and regional characteristics’.
Clusters are concentrations of related industries in a particular location or region
consisting of firms and their supply chains, complemented by other actors that influence
the productivity of the cluster including research, educational and training organisations
and govemment agencies.
* http://www.clustermapping.us/
The US cluster mapping project has developed an algorithm to group U.S. industry
codes into clusters groupings that reflect ‘inter-industry linkages based on input-output
measures, labor occupations, and the co-location pattems of employment and
establishments’. An important element of the method of grouping is the disaggregation
of clusters into traded and local categories (Porter, 2003). Traded clusters are groups of
firms that provide goods and services predominantly for export from the region of
interest, while local clusters provide goods and services predominantly to local
consumers including traded clusters. While the performance of both is important for
regional economies, it is traded clusters that make the greatest contribution to income
per capita. According to the U.S. Cluster Project:
While local clusters account for most of the employment and employment growth in
regional economies, traded clusters register higher wages, and much higher levels of
innovation. Local clusters provide necessary services for the traded clusters in a region,
and both are needed to support a healthy and prosperous regional economy.
While there is little doubt about the ability of cluster mapping in a regional economy to
provide clarity about the nature and magnitude of economic activity, there is more
controversy about how that information can be used to foster expansion or productivity
gains. In their article in the Oxford Review of Economic Policy (Nathan & Overman,
2013), the authors argue that using this knowledge to development economic policy is
not only fraught with difficulty, but that such policies have been shown not to work.
They favour ‘horizontal policies’ that focus on locations rather than clusters and
emphasise the benefits of agglomeration. Although Marshall (1890)is first credited with
the identification of agglomeration economies, many others have contributed to the
debate about their cause. Duranton and Puga (2004)review the literature and refine the
reasons for economies (i.e. increased productivities) that arise from agglomerations of
activity due to sharing (including indivisible goods and facilities); matching (of buyers,
sellers and labour in the supply chain); and learning (knowledge spillovers due to the
proximity of firms).
Agglomeration theory has attracted the interest of system dynamics practitioners
(Buendia, 2005), (Fratesi, 2002), (Lin, Tung, & Huang, 2006), (Dangelico, Garavelli, &
Petruzzelli, 2010), all of whom have modelled aspects of agglomeration and its role in
regional economies.
Clusters and their agglomeration benefits are subjects of interest in the Australian state
of Westem Australia, where policy makers are tuming their mind to economic
diversification in a period where the state has reached the end of a massive recent
expansion of capacity in the mining sector, and commodity prices have fallen
dramatically. The state is heavily reliant on bulk commodity exports and the supply
chains associated with these industries plays a critical role in the state’s economy. A
recent cluster analysis of the state’ s economy by local economics firm Pracsys” provides
the opportunity to examine the role of traded and local clusters as an initial step in
understanding the dynamics of the regional economy.
Western Australia - a profile
Westem Australia is Australia’s largest state, covering 2.5 million square kilometres or
33% of Australia some six times the size of Califomia The population is
* http://pracsys.com.auy/
approximately 2.5 million (about ten percent of Australia), of which some three-quarters
live in the capital city of Perth and a further ten percent in the adjacent Peel and south
west regions. Much of the state’s east and north is very sparsely populated, although
home to most of the state’s mineral and energy resources. The population has been
growing strongly (averaging 3% per annum) over the last 15 years, driven significantly
by migration from interstate and overseas.
The state’s Gross State Product in 2012-13 was A$253bn, representing about 17% of
Australia’s Gross Domestic Product (GSP), which translates to around A$100,000 per
capita (approximately US$70,000 at present exchange rates). This is some 50% higher
than Australian GDP per capita and 30% higher than equivalent U.S. GDP per capita.
The largest contributing industry is the mining sector (including and oil and gas),
representing almost a third of GSP (Figure 1).
BMining
m Construction
= Ownership of dwellings
m Manufacturing
Transport, Postal and Warehousing
Professional, Scientific and Technical Services
= Health Care and Social Assistance
= Other (incl taxes less subsidies)
Figure 1 Contributions to Westem Australian GSP (2013-14)
Source: (BankWest Curtin Economics Centre, 2015)
The economy has grown significantly over the last fifteen years (Figure 2), predicated
largely on growth in the commodity export sector. Average annual growth of around
8% per annum has occurred over this period. Population growth over the same period
has been around 3% pa, indicating how rapidly income per capita has increased.
300,000
18 Other (incl taxes less subsidies)
250,000
= Health Care and Social Assistance
E 200,000 ™ Professional, Scientific and Technical Services
3
acne m= Transport, Postal and Warehousing
3 m= Manufacturing
§ 100,000
* = Ownership of dwellings
50,000 -— -— = Construction
mn , , Mining
1990-91 2000-01 2013-14
Figure 2 Westem Australian GSP
Source: (BankWest Curtin Economics Centre, 2015)
Page3
The increase in GSP over this period correlates strongly with exports which have grown.
from 30 to 50% of GSP (Figure 3).
160,000 60%
_——Exports
140,000 OSA soe
120,000 ————Exports/GSP N
100,000 V4 A 40%
80,000 30%
60,000 oR
40,000
20,000 10%
A$m current prices
3
R
3 a eng
RHRSRKESSSBSBRR
AAKAHRSSESSESESS
SA2AS8AA8RRRRKRARRA
Seago ti aang agaa
SSSSESSEESESEESES
8S&Sssessseags
2
Figure 3 Westem Australian export growth
Sourve: (Australian Bureau of Statistics, 2015a)
In 2012-13, Westem Australia accounted for 14% of Australia's value of imports and
47% of Australia's value of exports (ABS 2015). Exports are dominated by three
particular commodities: iron ore, gold and hydrocarbons, representing nearly 70% of
Westem Australian exports in 2012-13 (ABS 2015). Figure 4 illustrates the volumes of
those commodities for Australia in total. Iron ore and liquefied natural gas (LNG) have
grown strongly, in particular since around 2000. This growth phase coincides with the
huge expansion in China’s economy over that period (Figure 5), particularly in respect
of imports. China (47%), Japan (19%) and South Korea (8%) are Westem Australia’s
largest export destinations by value.
Iron ore exports in Mt (LHS) 30
600 Gold in t (RHS)
——NG exports in Mt (RHS)
7
wo \\ ee |
a ao y= 5
ton Saal .
oO T T f T T T T T T It T 0
PPP EI I SLES PS
Figure 4 Australian primary exports
Source: (Australian Govemment, 2014)
Page 4
—GoP at market prices
(constant 2005 USS)
— Imports of goods and services
(constant 2005 uss)
Figure 5 China GDP and imports
Source: (World Bank, 2016)
The Westem Australia economy is presently affected by low intemational prices for
both iron ore and petroleum products, although the impact has been somewhat offset by
an Australian dollar falling against the currencies of its major partners (Figure 6).
160
140 yo
0 | f\
100
80 +
60
6 — Australian Dollar Trade-weighted Index
- Index of bulk commodities prices — AS
0 : +
Sssgegegsgegsggagaagasee 3 e9aa8
SSSRAARKARRKREKREKREKRARKRSSESB
sggggggagggggaggegcgga
sessss sss sss sssssE
See a eee eed eee dda
Figure 6 Indices of commodity prices and exchange rates
Source: (Reserve Bank of Australia, 2016)
Notwithstanding current commodity prices, the Westem Australian economy is set for a
major expansion in exports over the medium term. New LNG capacity is presently
being planned / constructed which will ramp up production from around 25 Mt pa to at
least 60 Mt pa in the early 2020s. Subject to market demand, remaining resource stocks
can maintain that level of production for some decades to come. Even if prices remain
depressed, this expansion in volume will add significant export value to the economy.
Tron ore is ina similar position. Even though the rate of growth in China’s economy will
likely continue to slow, demand for steel worldwide will ensure that Westem Australia
increases production over the medium tem.
Page S
However over the longer term the outlook is not as rosy. The value of all of Westem
Australia’s main export commodities will likely be severely diminished within 50 years
or so, either due directly to resource depletion or the economic and / or environmental
cost of ongoing extraction (Grace, 2015).
In the following analysis, a simple system dynamics model is used to explore possible
trajectories for the Westem Australian economy, population and income per capita over
arange of timeframes.
Input output analysis
Using the national input-output data compiled by the Australian Bureau of Statistics, the
economics firm Pracsys has produced a Westem Australian version of the US cluster
dataset. This was achieved through a process of concordance between the Australian
and New Zealand Standard Industrial Classification (ANZIC). While the full range of
clusters has been identified (see Appendix A), it is the basic disaggregation of the
economy into traded and local clusters that is the focus of the analysis reported here.
The following table represents 2012-13 data escalated to reflect 2015 prices.
Table 1 Westem Australia Input-Output Table (all figures in A¢m)
Industry uses Final Exports Total
domestic supply
Traded Local Total ‘uses
clusters clusters
Intermediate use
- Traded clusters 86,274 25,142 111,416 71,102 112,698 295,215
- Local clusters 31,307 20,017. 51,323 84,198 2,719 138,241
Total 117,580 45,159 162,739 155,299 115418 433,456
P1 Compensation of 51,557 41,600 93,157 93,157
employees
P2 Gross operating surplus 88,608 35,747 124,354 124,354
& mixed income
P3__ Taxes less subsidies on 1,338 785 2,123 15,416 17,539
products
P4 Othertaxes less 3,291 3,632 6,923 6,923
subsidies on production
Competing imports 32,841 11,319 44,160 24,041 68,201
WA Production 295,216 138,241 433,457 194,757 115418 743,631
Value Added. 143,457 80,978 224,435
(P1+4P2+P4)
Gross State Product 241,974
Source: (Australian Bureau of Statistics, 2015b), Pracsys 2016
Notes:
Final domestic uses include: Final Consumption Expenditure by households and govemment,
Gross Fixed Capital Formation by households and goverment, Changes In Inventories.
Table 1 and Figure 7 identify the dominant role of traded industries in the Westem
Australia economy.
@ Traded clusters Exports 98% 29)
m Local clusters
Value added
G xed |
SE eae
income
Compensation 55% A5%
Output 68% 32%
Figure 7 Comparison of traded and local industries in Westem Australia
The dynamics of regional economies
The regional economy is comprised of enterprises producing goods and / or services
mainly for domestic use (local clusters) and mainly for export (traded clusters).
Consider first an econonty where all goods and services produced are used locally and
there is no net migration in or out of the region (Figure 8). The final demand for goods
and services associated with these local clusters will be some function of the population
and the net income (gross value added (GVA)) per capita of the local population. The
population will increase naturally (in line with net birth / death rates) and the size of the
economy will grow proportionally with population. If real costs rise or fall, this will be
reflected in final demand, output and GVA.
Natural growth
rate
Domestic final population.
demand
Local cluster ane per
= capita
a
Figure 8 Causal loop diagram of local cluster activity within a regional economy
Now consider the same economy operating in an environment of free exchange of
goods and services with other regions in the nation and world (Figure 9). In this case,
traded clusters employ local people and purchase goods and services from local clusters
to meet export demand (and also contribute to meeting domestic final demand). Export
growth will relate to demand from elsewhere and is therefore essentially an exogenous
factor, reflecting both volumes and prices for exports. Population changes will reflect
the natural growth rate but also the net migration to and from the region. This migration
rate is a function of the relative attractiveness of the location for employment, and will
therefore reflect changes in the growth of the economy (GVA).
Page7
Natural growth
Pi
ay Domestic Aan
Por Satan”
demand 4
Export demand tae
K ral cluster “q Local cluster on WA per
output output capita
Beat Migration
~ growth rate
Figure 9 Causal loop diagram of traded and local cluster activity within a regional economy.
These dynamics illustrate why the traded activity is sometimes referred to as the engine
of the economy. The export trade has a multiplicative effect on the local economy
which increases (or decreases) GVA, which in tum attracts (or dissuades) migration.
The model
A simple system dynamics model of the Westem Australian economic representing this
causality has been developed using the Vensim software. The inputs to the model are
derived from the input-output analysis described above. The model ignores inflation and
makes several other simplistic assumptions:
° Cumrent exports are modified by a growth fraction which is an exogenous input;
. The inputs to clusters retain their existing relativity to output; and accordingly
0 compensation, imports and GVA remain as a fixed proportion of output;
0 sales within and between traded and local clusters remain a fixed proportion
of output.
Domestic final demand (final consumption expenditure by households and goverment,
gross fixed capital formation and changes in inventory) varies in direct proportion to
population changes and GVA per capita (which is taken as the sum of employee
compensation, gross operating surplus and mixed income, taxes less subsidies on
production). This latter assumption simply says that next year’s domestic demand is
proportional to this year’s income. Consumption of fixed capital is ignored in
income (taken as GVA / capita) but, on the assumption that this is a
relatively constant proportion of output, this simplification does not affect the model
dynamics.
Population changes are disaggregated between the natural growth fraction and a net
(interstate and overseas) migration. Since the 1980s the former has slowly declined
while the latter has varied significantly (Figure 10). At its peak in 2011, around 1500
people per week were entering the state.
Page 8
3.5%
Natural increase
3.0%
Net migration
2.5%
1.5% |
M that Ait atl
0% MYAN ANY Ma tht
05% 4 aoe 1 1) id
: i y
Annual % increase
5
0.0% T T T T
BRBBRRAERESRHBSSBSNSE
BRBSRRARRFRISSESRAS
SFARSAGaRARRRRRRSER
goa aid 2 doc.8 S12.8 8 G8 8 a8 8
ssscaSssgeaassssgass
Sasa essssassadsadad
Figure 10 Westem Australia population changes
Source: (Australian Bureau of Statistics, 2015a)
The spikes in migration have been caused by the economic cycles of the Westem
Australian economy. One way this can be seen is by correlating migration changes with
changes in GSP over the same period (Figure 11). In this analysis the migration changes
are plotted with a one year lag.
This relationship is used in the model to vary the migration fraction in response to
cours ih oval GSP, which in tum reflect the combined effect of traded and local
As exports are modelled as an exogenous multiplier on existing demand, and the
intemal economic relationships are fixed by the input-output multipliers, the causality
relating population to migration is the key endogenous feedback in the model.
3.00%
e
250% =0.1259x
a” o¢ R? = 0.43!
5 2.00%
$ +
a +
5 1.50%
3 oe”?
£ Pg
2 1.00% +
& ¢¢ £9
2 + e
0.50% rs >
xe
0.00% 1
0% 5% 10% 15% 20% 25%
Change in GSP
Figure 11 Changes in annual migration and GSP
Model simulations
The model is established in equilibrium based on the 2015 input-output parameters. The
dynamics are created entirely by entering a value for the natural population growth
(assumed as a constant at the existing rate of 0.84% pa), and experimenting with the
export growth fraction.
Page 9
The influence of exports
Initially the model is used to demonstrate the influence of exports on the Westem
Australian economy. Consider first a situation where equilibrium is disturbed through
1% natural population growth and there is no net migration. In this case exports are
assumed to remain static at the values in Table 1, meaning growth occurs mainly in the
local clusters as domestic final demand (including investment) increases. As the traded
clusters also serve local demand, growth occurs in those clusters also but at a much
lower rate (Figure 12).
Selected Variables
141,000
298,300
139,600
296,800
138,200
295,200
012345678 90ll 213 4 15
Tine (Year)
Local cluster output : Natural growth 1%
‘Traded cluster ouput : Natural growth 1%
Figure 12 Local and traded cluster output
The effect of this is an initial increase in output and GVA due to the increased final
demand associated with population increases. However, without the contribution of
export growth, GVA per capita (GVA/c) falls which depresses future final demand per
person. The net result is the establishment of a new equilibrium in output and GVA
after 15 years but with continuously falling income in the economy (Figures 13a and b).
TUGVA TU GVA percapita
227,600 = 6600
(226,800 83,850
226,000 81,110
225200 73370
224,400 75,630
123 4 5 6 7 8 9 10 11 12 13 14 15 o12 3 4 5 6 7 8 9 10 11 12 13 14 15
‘Tine (Year) hime (Y¢
Tuc 1% THGVA
Figure 13 a) GVA without exports Figure 13b) GVA/c without exports
If export growth is now invoked in the model in addition to natural population growth
(but without any net migration) it can be established that approximately 1.125% pa
export growth is required to maintain GVA per capita (i.e. income) at or near the initial
values (Figure 14 a and b).
Page 10
TUGVA TIGVA percapita
261,700 an SOO ee
252,400 84,130
243,000 81,300
233,700 78460
224,400 sre et 75630
123456786902 14 15 T234567890l 23 4
Tire (Yea) Tine (Yea)
TUQVA : Natur growth re
TG 1 fs
Figure 14a) GVA with exports but no Figure 14b) GVA/c with exports but no
migration migration
However, if the migration effect in the model is invoked (i.e. migration responds to
economic growth) then slightly higher export growth (1.25% pa) is required to maintain
income per capita (Figures 15a and b).
TUGVA TU GVA percapita
266,400 86970 ac a
255,900 84,130
205,400 81,300
234,900 aan
20,400 | ee | .
: 75,630
he Bh ER RB eS 0 UE 2 ed Oe 1234567890213 4 15
or Tine (Yea)
TIGIA : port gowth at L125¥6n0 nigeaion a ee
TIGIA : por growth at 125% ith mgation 7 2
Figure 15a) GVA with exports and migration Figure 15b) GVA/c with exports and
migration
The effect of these changes in population is set out in Figure 16. This indicates that
export growth of 1.25% pa would likely increase the average annual population growth
rate over 15 years from 1% to about 1.14%.
Population
2.592M
012 3 4 5 6 7 8 9 10 11 12 13 14 15
‘Tine (Year)
growth at 1.12
Population : Export growth at: 1.25% with migration
Figure 16 Population with and without exports
Page 11
Assuming a constant 1% pa natural population growth rate, and migration that reflects
economic growth, the influence of export growth on GVA per capita over 15 years can
be established (Figure 17).
3
a export growth rate
0%
—1%
27> —3%
—_—%
—10%
1 =
GVA per capita index
0 5 10 15
Years
Figure 17 Effect of export growth in GVA/c
The experiments establish the importance of export growth to maintaining income per
capita in the economy of Westem Australia In reality, as the following sections
illustrate, export growth has been well in excess of that required to achieve this
objective.
The trajectory of the Westem Australian econony
The following set of runs seeks to explore the immediate future of the state’s economy,
i.e. over the next 15 years.
Model case 1 - return to recent historical export growth
This case assumes the current downtum (which is price related and essentially due to
imbalances in intemational commodity supply and demand) is short lived and that
export growth recovers from 5% (the average value from 2010-2015) to levels observed
during the last twenty five years (Figure 18), i.e. around 10% per annum over the next
15 years.
40%
——5P growth
30% +—— § a Tl
——Export growth
sont AVA
10%
0%
Annual growth rate
-10%
-20%
> PP oh
RRR
SP Pry
oF FSS Pr SP
SHG HM HF GF O_o”
PPG IG HW" HW
Figure 18 GSP and export growth rates
Source: (Australian Bureau of Statistics, 2015a, 2015c)
Page 12
Model cases 2 - 4 declining export growth
These cases assume that the cument downtum in exports persists, either because
production levels are lower than trend, prices remain depressed or some combination of
both factors.
Under these assumptions the model commences with export growth of 5% pa (the
average value from 2010-2015) but growth declines linearly over fifteen years:
Model case 2 - export growth declines to 2.5% pa
Model case 3 - export growth declines to 0% pa
Model case 3 - export growth declines to -5% pa
The results of these simulations are shown in Figures 19a, b and c.
e
e
e
600,000
—bata
500,000 -- ——Miodel Case 1
——Model Case 2
400,000
= ——Model Case 3
& 300,000 - —Model Case 4
=
$
fc
200,000 (~~
100,000 |
0
1998 2005 2012 2019 2025 2032
Figure 19a) Future projections - Gross Value Added
3,500,000
3,000,000 a
& 2,500,000
3 Data
5 Model Case 1
3
4 2,000,000 —— Model Case 2
ES
ae —— Model Case 3
4,500,000 ——Model Case 4
1,000,000 T T 1
1998 2005 2012 2019 2025, 2032
Figure 19b) Future projections - Population
Page 13
260,000
240,000 }|- ——pata
220,000 +
—Model Case 1
200,000 +-
180,000 |. —=Model Case 2
5
@ 160,000 -- ——Model Case 3
Fy L
= SG O08 Model Case 4
& 120,000
% 100,000
zu Pal
80,000 {~~
60,000 ea
40,000
20,000
1998 2005 2012 2019 2025 2032
Figure 19b) Future projections - GVA per capita
The long tem
Next, a series of experiments were undertaken that examine the implications of
declining resource stocks over the longer term (fifty years).
Iron ore
The term Economic Demonstrated Resources (EDR) is defined by Geoscience Australia
as ‘resources for which profitable extraction or production under defined investment
assumptions is possible’. The world’s EDR of iron ore are estimated at 190,000 Mt
(Australian Govemment, 2015) of which Australia holds some 54,000 Mt (29%). These
resources are presently being mined globally at a rate of around 3,220 Mt pa (2014
figures (Australian Govemment, 2015)). Production is driven by steel production which,
according to the World Steel Association (2016), is set to increase by 150% by mid-
century, i.e. an average annual growth of 1.12% pa. Simplistically assuming a pro-rata
increase in global iron ore production, global iron ore stocks would be depleted by
around 2060.
Australia’s current share of global production (of which the vast majority is from
Westem Australia) is around 735 Mt or 23% (Australian Govemment, 2015). In
Westem Australia production has been expanded dramatically in recent years but this
expansion is expected to peak by 2020 at around 800 Mt pa (from 720 Mt in 2015)
(Govemment of Westem Australia, 2015a). New capacity additions after this appear
unlikely given that iron ore prices are not expected to retum to their previous highs as
China’s GDP growth rate slows. It is therefore possible that iron ore export values could
have already peaked in 2015.
However this simulation assumes that:
e prices recover somewhat by 2020 and export values stabilise at around $52
billion pa (800 Mt @ $65 per Mt); and
e production begins to fall in 2050 and by 2065 has reduced to 150 Mt pa.
LNG
Westem Australia’s current LNG exports are around 20 Mt pa Projects that are
currently under construction or in planning will increase this figure to around 60 Mt pa
Page 14
by the early 2020s (Govemment of Westem Australia, 2015b). If this production figure
then remains static, Westem Australia’s remaining stocks of gas (2,460 Mt) (Australian
Govemment, 2015) would be exhausted in around 40 years. As Westem Australia has a
15% domestic gas reserve, it can be assumed that exports will rise to around 50 Mt pa
over the next 10 years or so and remain at that level for 30 years or so. Of course,
additional discoveries of both conventional and unconventional gas are likely. It is also
possible that LNG exports could be adversely affected by a global shift away from
fossil fuels over this period.
In this simulation it is assumed that:
e Export production increases steadily to 52 Mt pa by 2020 and stabilises at that
leve;
e LNG prices recover by 2020 to previous levels (approximately A$720 per Mt);
e Production begins to fall in 2045 and by 2065 has reduced to 10 Mt pa.
Other exports
Intemational export growth in industries other than mining (which includes LNG) has
been minimal (Figure 20) at 0.6% average annual growth in the period 2006-07 to 2013-
14 (ABS 2015). Particularly noticeable is the decline of exports in the manufacturing
sector.
120,000
— Agriculture, forestry
100,000 ~ and fishing
—— Manufacturing
80,000 -
— Construction
& cov -
2 Wholesale trade
40,000 - i
Retail trade
20,000 ——Transport, postal and
A ——— ‘
0) x —Mining
SPP SW Pw
ION IA AV OS
PP Pr ah OY 48” 0 48”
Figure 20 Recent Westem Australian export values
Source: (Australian Bureau of Statistics, 2015d)
For the purposes of this simulation it is assumed that all other export values increase by
1% pa throughout the period to 2065 (a total increase of 62%).
The net impact of these assumptions on Westem Australian export value is shown in
Figure 21.
Page 15
a —ron ore
—LNG
§ 120 —other
= 100
3 =r SSS
_———— oes
60
40
20
SS
0 + T TTT T
Se rn gre See rar are
eeSeeegggeeee2e2 38
Figure 21 Export projections 2016-65
The results when this data is included in the model are depicted in Figure 22. This
shows that a slowing in export growth after 2020 will lead to a peaking in income per
capita at something like existing levels, and net reductions in exports around 2050 will
likely lead to reductions in GVA in both absolute and per capita terms.
400,000
Data Model results
350,000 ==
SS ae
300,000 ——__s=
Gow
250,000 —
fF ie —sva ($m) LHS
200,000 —GVA/c ($) LHS]
150,000 —— Population RHS
100,000 Zr = ==
50,000
0 T T
1998 2012 2025 2039 2053
2067
4,500,000
4,000,000
3,500,000
3,000,000
2,500,000
2,000,000
1,500,000
| 1,000,000
500,000
0
Figure 22 Economy projections 2016-65
Clearly if export growth outside the non-renewable resource sector is less than 1% pa
this outcome would occur somewhat sooner. It is also likely that over the medium term,
real prices for these commodities will increase as they become scarce. This would
increase the magnitude of the peaks in Figure 22 but probably also bring forward the
decline (as substitutions become more economically viable).
Discussion
Of course there are many assumptions included in the model which will undoubtedly
tum out to be incorrect. However, it is difficult to imagine that the eventual depletion of
the state’s globally valuable natural resources can be offset with substitutions of other
exports. The inevitable result of this will be a reduction in exports and hence output and
GVA and with it a net reduction in migration. Most importantly for living standards this
will also likely lead to a reduction in income per capita.
In this conference we are encouraged to consider the Black Swan effect in modelling the
real world. In his book of that name which has created such interest in the topic Nassim
Page 16
Nicholas Taleb (2007) describes a Black Swan event as having three characteristics,
being:
e an outlier, as it lies outside the realm of regular expectations, because nothing
in the past can convincingly point to its possibility.
e itcarries an extreme impact; and
e humen nature makes us concoct explanations for its occurrence after the fact,
making it explainable and predictable’.
The dramatic impact of China’s import growth in the period 2000-2012 (Figures 2, 4
and 5) is perhaps Westem Australia’s Black Swan event’, assuming a broad
interpretation of an ‘event’ as something that occurs in a certain place during a
particular interval of time. In fact understanding that it is an event rather than an
trend is a necessary precursor to addressing the long term structural
implications of an inevitable re-structuring of the economy over the coming decades.
Altematively of course, there could be another Black Swan.
3 Tn fact this human trait extends well beyond explaining Black Swan effects as is well established in the
psychology literature (Haid, 2006)
The irony here is that the black swan is the state’ s bird emblem and features on the coat of arms.
Page 17
References
Australian Bureau of Statistics. (2015a). Catalog 3101.0 Australian Demographic Statistics,
Table 4 Estimated Resident Population, States and Territories.
Australian Bureau of Statistics. (2015b). Catalog 5209.0.55.001 - Australian National Accounts:
Input-Output Tables, 2012-13.
Australian Bureau of Statistics. (2015c). Catalog 5220.0 Australian National Accounts: State
Accounts, Table 6. Expenditure, Income and Industry Components of Gross State
Product, Western Australia, Chain volume measures and current prices.
Australian Bureau of Statistics. (2015d). Catalog 5368055006 - Characteristics of Australian
Exporters, 2013-14, Table 7. Value of Goods Exports, by Industry of Exporter and State
of Origin of the Exported Commodity.
Australian Government. (2015). Australia's Identified Mineral Resources 2015.
Australian Government, D. o. |., Office of the Chief Economist. (2014). Resources and Energy
Statistics 2014.
BankWest Curtin Economics Centre. (2015). Tiger, Tiger, Burning Bright? Western Australia’s
trade and economic development with Asia. Perth: Curtin University.
Buendia, F. (2005). Increasing Returns to Economic Activity Concentration. Paper presented at
the 23rd International Conference of the System Dynamics Society, Boston,
Massachusetts, USA.
Dangelico, R. M., Garavelli, A. C., & Petruzzelli, A. M. (2010). A system dynamics model to
analyze technology districts’ evolution in a knowledge-based perspective.
Technovation, 30(2), 142-153. doi: 10.1016/j.technovation.2009.09.006
Duranton, G., & Puga, D. (2004). Micro-foundations of urban agglomeration economies
Handbook of regional and urban economics (Vol. 4): Elsevier.
Fratesi, U. (2002). Regional Economies and Innovative Performance as the Source of
Competitiveness and Agglomeration: a System Dynamics Representation. Paper
presented at the 20th International Conference of the System Dynamics Society,
Palermo, Italy.
Government of Western Australia. (2015a). 2015-16 Budget Fact Sheet - Iron Ore Royalty
Revenue.
Government of Western Australia. (2015b). Western Australian Mineral and Petroleum
Statistics Digest 2014-15.
Grace, W. (2015). Simulating sustainability: a resources perspective. Journal of Natural
Resources Policy Research, 1-30. doi: 10.1080/19390459.2015.1050202
Haidt, J. (2006). The Happiness Hypothesis: Finding Modern Truth in Ancient Wisdom: Basic
Books.
Lin, C.-H., Tung, C.-M., & Huang, C.-T. (2006). Elucidating the industrial cluster effect from a
system dynamics perspective. Technovation, 26(4), 473-482. doi:
10.1016/j.technovation.2004.11.008
Marshall, A. (1890). Principles of Economics. London: Macmillan.
Nathan, M., & Overman, H. (2013). Agglomeration, clusters, and industrial policy. Oxford
Review of Economic Policy, 29(2), 383-404. doi: 10.1093/oxrep/grt019
Porter, M. (2003). The Economic Performance of Regions. Regional Studies, 37(6-7), 545-546.
doi: 10.1080/0034340032000108688
Porter, M. E. (1990). The Competitive Advantage of Nations. New York: Free Press.
Reserve Bank of Australia. (2016). Statistical tables.
Taleb, N. N. (2007). The Black Swan: The Impact of the Highly Improbable: Random House.
World Bank. (2016). World Development Indicators.
Page 18
World Steel Association. (2016). Steel facts Retrieved 19/03/2016, from
https://www.worldsteel.org/Steel-facts.html
Page 19
Appendix A
Clusters in the Westem Australian economy
Agricultural Inputs and Services
Agriculture
Automotive
Business Services
Coal Mining
Construction Products and Services
Distribution and Electronic Commerce
Downstream Chemical Products
Downstream Metal Products
Education and Knowledge Creation
Electric Power Generation and Transmission
Environmental Services
Financial Services
Hospitality and Tourism
Household Goods
Infomation Technology and Analytical Instruments
Insurance Services
Jewelry and Precious Metals
Lighting and Electrical Equipment
Livestock Processing
Local Commercial Services
Local Community and Civic Organizations
Local Education and Training
Local Entertainment and Media
Local Food and Beverage Processing and Distribution.
Local Health Services
Local Hospitality Establishments
Local Household Goods and Services
Local Logistical Services
Local Motor Vehicle Products and Services
Local Personal Services (Non-Medical)
Local Real Estate, Construction, and Development
Local Retailing of Clothing and General Merchandise
Marketing, Design, and Publishing
Medical Devices
Meal Mining
Metalworking Technology
Music and Sound Recording
Nonmetal Mining
Other Livestock Farming
Paper and Packaging
Performing Arts
Personal Services
Plastics
Production Technology and Heavy Machinery
Recreational and Small Electric Goods
Science and Technology Services
Sheep, Beef Cattle and Grain Farming
Textile Manufacturing
Tobaoco
‘Trailers, Motor Homes, and Appliances
Transportation and Logistics
Upstream Chemical Products
Upstream Metal. Manufacturing
Utilities
‘Video Production and Distribution.
‘Vulcanized and Fired Materials
Water Transportation.
Wood Products
Page 20
Appendix A Model description
This appendix contains more details about the implementation of the model structure, which was
shown at an aggregated level in Figure 3.
The leaming-by-searching and learning-by-doing mechanisms are modelled with the two factor
learning curve formula (Kouvaritakis et al., 2000).
The formula for cost improvement is as follows:
CC. KS. _
spc =A. (Sy*. SF (1)
In equation (1), SPC stands for the investment cost per unit for the technology (specific cost ) and CC
is the cumulative installed capacity at a given time, which is divided by the cumulative capacity at
time 0 (CCo). —« is the learning factor, and A is the specific cost at time 0. KS stands for the knowledge
stock at a given time, measured by the R&D investments of that year. KSo is the initial knowledge
stock and # represents the learning by searching factor, which is the representation of the percentage
improvement on the investment cost coming from the learning-by-searching process, similar to —«
representing the percentage improvement of leaming-by-doing process. This formula is also adopted
for performance improvement of the capacity factor, with positive a and £ values.
Affinity with wind turbines represents the probability of an actor to purchase wind turbines by
comparing it with the conventional technologies. Modelling affinity and familiarity is adopted from
Struben and Sterman’s (2008) work on electric vehicle adoption. The formula of affinity is based on
standard multinomial logit choice models, which is a commonly used framework in modelling
consumer choice among the different options in the consideration set. In this case, the consideration
set consists of wind turbine vs. conventional technologies:
aj = a*exp(—6 ae - 1p (2)
Equation (2) represents the affinity of wind turbines based on LCOE. aj is the affinity towards wind
turbines at a given time, where a* represents the reference affinity for the reference LCOE value
LCOE*. The reference value stands for a normal value that the adopter has an idea about. For
example, an actor decides whether the given LCOE of the available options is expensive or not by
comparing it with the reference LCOE*. If the given LCOE is more expensive than the reference
value, the affinity decreases and vice versa. The reference values are determined separately for
conventional technologies and for wind turbines. For conventional technologies, the average LCOE of
all times is taken as LCOE* and then affinity at this value is assigned as 1, because, on the average
price of electricity generation cost, the utilities will go for the conventional methods. After
determining reference values of conventional technologies, wind turbine reference values are
determined accordingly. Assuming that if wind turbine is competitive with the conventional
technologies, the affinity to the wind turbines is assigned to 1 with a lower reference LCOE* value,
since it is a relatively new technology and utilities will have questions in their mind for going for a
new technology. Besides there will be switching costs of the utilities for moving to a new technology,
due to limited experience and unknowingness of the new technology. This way affinity is modelled as
a decision making process of utilities for purchasing wind turbines. The formulation of LCOE is given
in (3):
21
EAC
LCOE, = =-+M, + Fe (3)
where
_ Iyr.Gtr)"
BAG (1tryP 41 (4)
EAC stands for equivalent annual cost representing the cost per year of owning and operating an asset
over its lifetime (Short et al., 1995). Io represents the overnight cost of the project meaning if the
project was completed overnight (no interest rate was taken into account). E; stands for the electricity
generation in the year t, r stands for discount rate, n represents the lifetime of the project (which is 20
years for wind turbines), M: represents the operation and maintenance cost in year t and F; represents
the fuel cost in year t. Utilities make their decision based on this LCOE in the model with the
corresponding affinity to the LCOE of that year.
However, to be able to consider wind turbines as an option of energy generation, the actor should be
familiar with it. Knowledge diffusion through networks is another important part of the diffusion
process, therefore familiarity should also be modelled. Modelling familiarity is also adopted from
Struben & Sterman’s (2008) study. The causal loops are shown in Figure A1.
total social exposure
Figure A 1. Stock-flow diagram for modelling Familiarity
Familiarity is a stock variable between 0 and 1, where 0 represents no familiarity with the technology
and 1 represents that all adopters and potential adopters are familiar with the new technology.
Familiarity gain is determined by the social exposure coming from marketing campaigns, individual
learning and social exposure coming from the users, and finally the social learning standing for the
social exposure coming from non-users, which is word-of-mouth. The equation is shown below:
ne =at ciF () + GF (1 - 2) (5)
In the familiarity gain equation which is illustrated in (5), a represents the social exposure gained by
marketing pre ci represents the effectiveness ratio of the users, F represents the
familiarity value at that time, W represents the installed MW of wind turbines, N represents the total
installed capacity for electricity generation in MW and finally c represents the effectiveness ratio of
non-users on adoption.
Familiarity also decays over time, when there is not enough social exposure. This decay occurs in a
non-linear way, because if the level of social exposure is low, familiarity decays very fast, but if the
22
level of social exposure is high it does not. This is modelled with the following exponential function
(Struben and Sterman, 2008):
_ exp(—4e(mp-n°))
9, = % T+ exp(—4e(n;- n")) (6)
In this function, which is a characteristic logistic function, n: represents the social exposure from users,
nonusers and awareness campaigns at time t. n* represents the reference rate of social exposure where
familiarity decreases at half of the normal rate. The greater the exposure, the slower the decay. & is
the maximum familiarity decay rate. Familiarity decreases fastest when n; is small. ¢ stands for the
slope of the decay rate at a given point. It is assumed that « is 1/n* which normalizes the elasticity of
the familiarity decay to exposure at 1.
Appendix B- Initial Settings for the Model
Although the active mechanisms for California and Denmark are the same in the model, there are
important differences in the initial settings triggering these mechanisms. These differences are
illustrated in the model by determining the initial values of variables as well as the values of
exogenous variables. The values of these variables an the reasoning behind those values are shown in
the Table below.
Variable CA DK | Comments
‘Alpha valuefor | 1.07 | 1.07 | Capacity factor leaming did not show significant changes between CA and DK, therefore in the
leaming by doing model, this variable is treated as a global value with the same values.
on capacity factor
‘Alpha value for 0.88 0.95 When we look at the investment cost at 1980 and investment cost at 1995, we see that CA had
leaming by doing a more impressive leaming curve compared to DK. (For investment costs, the lower the alpha
on investment cost value, the greater the leaming impact, since it represents the percentage reduction on the cost).
Also, as Hekkert et. al mentions, leaming by doing is hugely affected from entrepreneurial
activities (2007). In Denmark the entrepreneurs were producing agricultural equipment before,
therefore they leamed slowly with trial and error (Kamoe & Garud, 2001)
Bela value for 1.04 | 1.04 | Since capacity factor is treated as a global value, this leaming effect is also the same. The
leaming by reason it is lower than alpha value is based on literature (Kamp, 2002).
searching on
capacity factor
Beta value for 0.90 | 0.96 | The reason to have lower value for CA which results in better cost reduction is based on
leaming by available data. Note that these beta values are also less effective compared to alpha values
searching on cost which are based on literature (Kamp, 2002)
Effectiveness of | 0.38 | 045 | Since the communication among potential adopters in DK was higher than CA due to
contacts of nonusers published Naturtig Energi magazine where the performances of wind turbines was made public
(Kamp, 2004). This magazine helped them to increase the knowledge of nonusers. For this
reason, the effectiveness of contacts of non-users is assumed to be 7% less in CA
Effectiveness of | 0.68 | 0.8 | Communication between the users of wind turbines were also higher in DK due to Wind
contacts of users Meetings where knowledge and experience were shared between manufacturers, owners and
researchers. They also established Danish Windmill Owners Association (Kamp, 2004), For
this reason, the effectiveness of contacts of users are assumed to be 12% less in CA.
Initial familiarity 0.25 0.25 Initial familiarity with the wind turbines was low but not zero for both cases. Both CA and DK
had historical experiences with wind turbines (see Section 3) and they were familiar with the
windmills. There was no real indication of familiarity difference between two cases in the
literature, therefore they are assumed to be the same.
Initial installed 55000 7072 | This numberis based on EIA data, reflecting the real values (in MW).
capacity for
electricity
generation
Initial investment |~ 2500 |” 1322 | This data is taken from the literature and converted to 1980°s dollar value. (Sawin, 2001; Lantz
cost of wind etal 2012).
turbines per kW
Interest rate 0.66 0.77 | The interest rates are also taken from the literature (Sawin, 2001).
(mean) | (mean)
0.0265 | 0.0172
(stdev) | _(stdev)
Maximum decay 042 | 042 | Maximum decay rate for both cases are assumed to be same, because this value represents the
rate reference value for forgetting rate. Due to differences in cultures this number could differ, but
in general, people tend to foryet the new technology when the exposure is not frequent enough
23
(Struben & Sterman, 2008). Since this situation is valid both for CA and DK the same value is
used in the simulation.
Nomnal social 0.20 | 0.20 | Similarto maximum decay rate, this value represents the reference value for forgetting rate.
exposure When itis 0.2 it means that familiarity decays with the half of the maximum decay rate. Since
maximum decay rate is assumed to be the same for both cases, it is reasonable to take the same
reference value for normal social exposure, ensuring the decay behaves the same for both
cases.
Operation costof | 14.19 | 12.73 | These costs change over time, therefore their mean and standard deviation is given in the table.
wind turbines (mean) | (mean)
sa) aso
(stdew)_|_ (dew)
Percentage increase | 2.5% | 2.5% | These values are also calculated on average, by looking at the net changes of installed capacity
of inst between 1980 and 1995 (EIA, 2012). The average capacity increase per year for both cases
electricity capacity tumed out to be the same
Sensitivity value for T T.8 | The reason for taking Danish utilities’ sensitivity values higher than Califomia is due to
wind turbines market's results. When weighted average cost of conventional methods and LCOE of wind is
examined, it is observed that standard deviation of the prices is much higher in Denmark
compared to Califomia. This situation implies an insecure market structure with more sensitive
buyers to price. The numbers are calibrated with the fit to historical data, For both values DK
Sasi valetor | 054 J values are 1.8 times higher than CA,
conventional
technologies
Weighted average | 24.87 | 61.61 | These values are based on historical data, Since the value changes over time the mean and the
cost of conventional | (mean) | (mean) | standard deviation is given in the table. As can be seen, the prices are more stable in California.
methods (Average 2.607 11.63
LCOE) (stdev) | (stdev)
LCOE of wind | 31.75 | 56.83 | These values are calculated by the model, but to show the changes in the price over time itis
6.95 | 19.68 | added to the table.
24
Appendix C — R2 and MAE/Mean Tests Results
The tables below show the real data, model data and the results of these tests.
Denmark R2 and MAE/Mean tests
jent cost yearly i
real model real model real model
1980 1322 1322.00 2 5.00 2 1.40
1981 1322 1248.06 6 6.09 4 241
1982 1360 1195.01 10 9.12 4 4.56
1983 1284 1145.09 14 16.19 4 14.32
1984 1265 1091.72 20 31.04 6 13.76
1985 1209 1042.04 47 48.62 27 22.28
1986 1133 998.38 n 73.75 25 25.57
1987 1058 959.49 112 104.08 40 26.80
1988 982 925.85 190 135.94 8 57.84
1989 1095 895.92 246 185.26 56 55.51
1990 1076 868.11 325 242.24 79 60.71
1991 1020 843.12 392 300.26 67 62.08
1992 982 820.78 435 358.17 43 39.31
1993 906 800.83 467 407.81 32 62.75
1994 831 783.00 520 470.23, 53 82.81
1995 793 766.64 599 552.86 79 96.69
R2 0.87 0.98 0.76
MAE/Mean 10.95% 15.09% 27.56%
California R2 and MAE/Mean tests
investment cost
re model real model i real
1980 2500.00 2500.00 0 x 0
1981 2297.00 2189.47 10 7.87 0
1982 1847.19 1990.29 70 11.03 10
1983 1466.00 1792.83 240 22.77 60
1984 1404.00 1374.73 617 300.35 170
1985 1350.00 1063.98 911 682.08 377
1986 1044.07 906.82 1235 973.06 398
1987 750.00 813.02 1304 1192.55 275
1988 701.75 750.37 1202 1279.71 154
1989 698.50 707.12 1302 1291.80 59
1990 681.37 675.06 1454 1297.51 64
1991 618.11 649.39 1679 1369.92 161
1992 621.00 627.24 1655 1443.12 165
1993 632.00 607.96 1608 1481.98 17
1994 599.50 590.91 1609 1510.12 9
1995 567.00 575.53 1523 1527.64 54
R2 0.96 0.96 7083
MAE/Mean__| 7% 13% 139%
25
