Journal of International Economics 57 (2002) 273–297
www.elsevier.com / locate / econbase
A U-shaped Europe?
A simulation study of industrial location
Rikard Forslid a , Jan I. Haaland b , Karen Helene Midelfart Knarvik b , *
a
b
Stockholm University and CEPR, Department of Economics, S-106 91 Stockholm, Sweden
Centre for International Economics and Shipping, Norwegian School of Economics and Business
Administration and CEPR, Helleveien 30, N-5045 Bergen, Norway
Received 17 April 2000; received in revised form 15 July 2001; accepted 16 July 2001
Abstract
We use a large-scale CGE-model to simulate the effects of gradual economic integration
on the location of industrial production. Our results reveal large differences among
industries. Industries with high scale elasticities typically display a non-monotonous
relationship between trade liberalisation and concentration, with maximum concentration for
intermediate trade costs. Other industries, more driven by comparative advantage, become
monotonously more concentrated as trade costs fall. On the aggregate level we find an
(inverted) U-shaped relation between trade costs and concentration. The results also show a
close correlation between real income gains and growth in manufacturing production,
stemming from pecuniary externalities in the manufacturing sectors.  2002 Elsevier
Science B.V. All rights reserved.
Keywords: Economic integration; Agglomeration; Economic geography
JEL classification: C68; F10; F12; F15; R12
1. Introduction
A common worry in peripheral regions is that economic integration may lead to
loss of industries and jobs in the periphery. While traditional trade theory would
*Corresponding author. Tel.: 147-55-959-510; fax: 147-55-959-350.
E-mail
addresses:
rf@ne.su.se
(R.
Forslid),
jan.haaland@nhh.no
karenhelene.knarvik@snf.no (K.H. Midelfart Knarvik).
(J.I.
0022-1996 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved.
PII: S0022-1996( 01 )00155-6
Haaland),
274
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
not give rise to such worries, a number of articles in the ‘new economic
geography’ literature (e.g. Krugman, 1991; Krugman and Venables, 1995) suggest
that economic integration may indeed lead to concentration and unequal regional
development. These theoretical studies, however, make their argument in highly
stylised models — normally a 2 3 2 3 2 framework.1 This is necessary because of
the complexity of the imperfect competition and industry-linkages framework. A
question then is whether the results and intuitions from simple theoretical
economic geography models go through in richer models.2
The purpose of this paper is to investigate if the results from small and stylised
models hold in a model of larger dimensions. We therefore simulate the effects of
trade liberalisation on the location and concentration of manufacturing industries
using a large-scale CGE-model. Hence, the paper aims at obtaining numeric
intuition of higher order properties of standard trade and location models — an
exercise that may also allow for insights beyond those obtained in simpler models.
An important insight from the theoretical literature is that industrial concentration can arise because of self-reinforcing backward and forward linkages.
These stem from a combination of increasing returns to scale (IRS), trade costs,
and the fact that firms are linked via their input–output structures (see e.g. Fujita et
al., 1999). Downstream firms use an aggregate of upstream varieties as an
intermediate input. When trade across borders incur costs, a larger number of
upstream firms in your region implies a lower price level for intermediate inputs.
This mechanism constitutes the forward link. More downstream firms, however,
also imply a larger home market for upstream firms, which increases their sales
and profits. This is the backward link.
In order to provide intuition for how these agglomeration forces interact with
other general equilibrium forces in determining location we will discuss two ‘pure’
cases. Consider first a two-country, three-sector, two-factor economy in which
countries have identical relative endowments but differ in size. Two sectors are
characterised by IRS and trade costs, and are linked via their input–output
structures, which create agglomeration forces due to backward and forward
linkages. The IRS-sectors have identical factor intensities. The third sector
produces a homogenous good under constant returns and perfect competition. This
sector has a different factor intensity. The profit maximising location of a firm in
any of the IRS sectors will depend on product-market considerations and on
factor-market considerations.
There are two different product-market considerations. On the one hand,
agglomeration forces associated with supplier proximity, draw firms to the larger
1
An exception is Puga and Venables (1996), who use a framework of multiple sectors with
inter-sectoral input–output linkages.
2
For instance, Davis (1998) has challenged the robustness of the home-market effects appearing in
such models. He shows that the introduction of equal trade costs for both goods in a two-sector model
takes away all agglomeration tendencies.
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
275
region. On the other hand, consumer-proximity considerations work against
concentration of firms in one country. That is, demand considerations encourage
firms to locate in proportion to demand rather than to concentrate. On the factor
market side, concentration of the two IRS-sectors in one country leads to a higher
relative price of the factor used intensively in the IRS sectors in this country,
which works as a force against concentration.
Consider now how the forces for and against concentration depend on trade
costs. For high trade costs export becomes prohibitively expensive, and firms
therefore locate according to demand. That is, consumer-proximity considerations
dominate firms’ location decision, and produce low concentration. In the opposite
case, when trade costs are low, trade in intermediates and final goods is cheap.
This implies that factor-market considerations dominate location, which also
produces low concentration since a concentration of the IRS sectors would drive
up the relative price of the factor used intensively in this sector. Finally, for
intermediate trade costs supplier-proximity considerations dominate the other two,
leading to high concentration. The combination of product-market and factormarket forces therefore makes concentration of the IRS industry non-monotonic in
trade costs, producing an inverted U-shape with maximum concentration for
intermediate trade costs.
Consider now instead a second ‘pure’ case — a traditional 2 3 2 3 2 model with
constant returns and perfect competition. Countries now differ in relative endowments. In this model there is no supplier-proximity effect. As before, at high trade
costs, the consumer-proximity effect dominates, which leads to a dispersed
production. For low trade costs factor market considerations dominate, but in this
case factor market competition dictates specialisation according to comparative
advantage, and leads production to concentrate geographically. This second pure
model, thus, produces a monotonic relationship with increasing industrial concentration as trade costs fall.
Note that the first model has no natural pattern of ‘comparative advantage’ other
than size differences, so at zero trade costs there is no concentration. The second
model does have a natural pattern of comparative advantage, leading to concentration at zero trade costs.
In summary, the two pure cases produce quite different pictures of concentration
and trade costs. The first an inverted U, and the second a monotonic curve with
negative slope. Both scenarios will be relevant when we turn to our large
simulation model.
This paper simulates the effects of economic integration using a full-scale
CGE-model — the EURORA model with 14-industries and 10-regions (Forslid et
al., 1999b) — which is calibrated on actual 1992 data. This model captures
comparative advantage due to differences in endowments and technology, imperfect competition and scale economies, as well as backward and forward
linkages through a complete input–output structure.
Our results show that the locational effects of economic integration are highly
276
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
region- and sector-specific with some sectors being driven primarily by comparative advantage and others by agglomeration forces associated with scale economies
and input–output linkages. However, the results for the overall increasing returns
to scale manufacturing sector reveal an (inverted) U-shaped relationship between
trade liberalisation and concentration of the manufacturing sector. Dual to this we
report movements in factor prices and welfare effects. We show that welfare is
positively associated with the location of the increasing-returns-to-scale (IRS)
manufacturing. Finally, factor price movements tend to co-vary and are positively
related to the pattern of industrial location. Relative factor price changes, on the
other hand, show clear traces of Stolper–Samuelson effects.
Section 2 describes the model, while Section 3 presents the results on industrial
relocation and specialisation following a process of economic integration. Section
4 discusses the effects of integration on factor prices and welfare, and Section 5
offers some concluding remarks.
2. The model
The model has 10 regions, and we will conduct our integration experiment
among four of these (the four Western European regions). We call them Central,
North, South, and West.3 In each region there are 14 production sectors. Of the 14
sectors, two are assumed perfectly competitive (energy and agriculture), while
there are 12 imperfectly competitive sectors. Two of these are non-traded services
sectors while the remaining 10 are traded manufacturing sectors.4 Trade in
manufactured goods incurs trade costs, while the perfectly competitive, resourcebased sectors are modelled with free trade. We have kept the resource-based
sectors as simple as possible, since the emphasis of the model is on
manufacturing.5 The basic industrial structure of the model is shown in Table 1.
The model we use builds on the CGE model developed by Haaland and Norman
(1992), but with significant modifications with respect to linkage structure, various
types of trade costs and market structure.6 An important feature of the model is
3
See Appendix A for details on the regions.
It might be argued that private services should be modelled as internationally traded. However,
although trade in services constitutes a significant share of international trade, it is still the case that
there is a strong dominance of domestic supply in most services sectors (see EFTA, 1994). Hence, to
simplify and at the same time focus on the potential importance of domestic supply of services as
intermediate inputs, we have chosen to treat services sectors as non-traded.
5
The model includes production subsidies to agriculture and energy, but in the scenarios presented
here, these policies are kept unaltered.
6
Other related CGE model-based analyses of integration and trade liberalization are e.g. Allen et al.
(1998,Baldwin et al. (1996, 1997,Brown et al. (1992, 1995,Francois et al. (1995,Gasiorek et al. (1991,
1992,Harrison et al. (1995, 1996,Keuschnigg and Kohler (1996,Lopez-de-Silanes et al. (1994,RolandHolst et al. (1994).
4
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
277
Table 1
Industries
Set
Industry
Description
NT
Public services
Private services
Non-traded monopolistically competitive sector linked to
all other sectors through the input–output structure
PC
Agriculture
Energy
Traded perfect competitive sectors without trade costs.
Each sector has a specific factor, which creates an element
of decreasing returns to scale
ITG
Textiles
Leather and products
Wood products
Metals
Minerals
Chemicals
Food products
Transport equipment
Machinery
Other manufacturing
Traded sectors with monopolistic competition.
Transport costs of iceberg type, plus tariffs and export
taxes or subsidies.
Linked to all other sectors through the input–output
structure
that it has a complete input–output structure, i.e. all linkages across the 14 sectors
in the model are taken into account and are modelled in detail, using regionspecific input–output matrices.
Hence, sectors are linked via demand for intermediate inputs, which creates
agglomeration forces a` la Fujita et al. (1999). However, contrary to the typical
theoretical model where there is just one industry and thus only intra-industry
linkages, the simulation model includes both intra- and inter-industry linkages.
This implies that agglomeration forces are not only created within industries but
also between different kinds of economic activity.
There are three primary factors of production — capital, skilled labour and
unskilled labour; these are mobile between industries within a region, but
immobile between regions. Factor demand derives from the 14 producing sectors.
In addition to the three mobile factors, two of the sectors — energy and agriculture
— use sector-specific natural resources. Hence, these two sectors show decreasing
returns to scale with respect to the mobile factors.
2.1. Basic model equations
Consumers have Cobb–Douglas preferences over a set of all goods (AG),
implying that they, in each market (m), will spend a fixed share (a ) of their income
(Y) on each good (i):
Ym
Cim 5 aim ], i [ AG
Pim
(1)
278
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
where Cim , and Pim are consumption and price of good i in market m. For perfectly
competitive goods prices are world market prices given by world market clearing
conditions for the respective goods. One of these goods (energy) is chosen as
numeraire. Imperfectly competitive goods (the set I) are differentiated, and
consumers consume a CES-composite of the individual varieties. Dual to this
consumption composite is the price index
SO
D
R
Pim 5
s12 sid
Nij a ijm PI ijm
j51
1 / (12 si )
, i [I
(2)
where PIijm is the price of a variety of good i produced in country j and sold in
market m, Nij the number of varieties of good i from country j, a ijm the demand
share of good i from country j sold in market m, and si the elasticity of
substitution between various varieties of good i. For non-traded, differentiated
goods a ijm 5 0 for all m ± j, since by assumption only domestically produced
varieties are consumed.
The imperfectly competitive sectors are characterised by monopolistic competition a` la Dixit and Stiglitz with free entry and zero profits. The producer price
(PPIij ) of good i produced in j is given as a mark-up over firms’ marginal costs
(MC):
si
PPIij 5 ]]MCij , i [ I
(3)
si 2 1
while the consumer price in market m (PIijm ) for imperfectly competitive traded
goods (ITG) is subject to trade costs 7
PIijm 5 PPIijs1 1 T ijmd i [ ITG
(4)
Demand in market m for each variety of good i produced in j may now be derived
as:
S D
Pim
Xijm 5 a ijm ]]
PIijm
si
Cim
i [ ITG
(5)
Prices and demand for non-traded differentiated goods are derived in the same way
as for traded goods, but with no need to distinguish between producer and
consumer prices since there is only domestic consumption of these goods.
The price index for differentiated intermediate goods (Q hm ) is industry-specific
by purchasing industry (h) and region (m). The industry uses all goods as inputs,
where the share of each good is given by the parameter gihm .
7
The model is actually calibrated for three types of trade costs: export taxes, transport costs, and
tariff equivalents of import barriers. The transport costs are of the iceberg type, while export taxes and
import tariffs are transfers (to the representative consumer). The experiment of reduced trade costs
implies an equiproportionate reduction in all three of these.
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
Q hm 5
SO g
ihm
D
12sqd
s
P im
;i [I
1 / (12sq)
, ;h [ AG
279
(6)
where sq is the elasticity of substitution among imperfectly competitive goods
used as intermediates. Observe that we use the same price index (Pim ) for industry
i here as for consumer demand; hence, we assume that intermediate demand and
final demand use different varieties of good i in the same proportions. The price
indices for perfectly competitive goods (the set PC) as intermediates are
constructed in the same way
QPChm 5
SO
D
12sqd
gihm PPC is
;i [PC
1 / (12sq)
;h [ AG.
(7)
PVij is a price aggregate for all primary factors used in the production in sector i
in region j. The share of each individual factor (k) is industry- and countryspecific, and is given by the parameter bijk
SO
K
PVij 5
k 51
bijkW jk12s i
D
1 / (12s i )
i [ AG.
(8)
Finally, the marginal cost for industry i in country j is specified as a nested
CES-function, with primary inputs, differentiated intermediates, and homogenous
intermediates in one second-level nest each, and with Stop as the elasticity of
substitution between the nests at the top level. Using the price indices above, the
marginal cost function may be written
MCij 5fBVijsPVijd 12S top i 1 BZijsQ ijd 12S top i 1 BZPCijsQPCijd 12S top ig 1 / ( 12S top i )
(9)
where BVij , BZij , and BZPCij are all calibrated parameters. From (9), using (6)–(8)
and market clearing conditions for each good, we find the demand for primary
factors and intermediate goods from each sector. Together with supply conditions,
these form the general equilibrium system. As the imperfectly competitive sectors
are characterised by monopolistic competition and zero profits, the scale of each
firm is fixed, and any output expansion is entirely reflected through an increase in
the number of firms (varieties) in the respective sector. In equilibrium there is
moreover a one-to-one, inverse relationship between elasticity of substitution (si )
and scale elasticity.
The use of intermediates from own as well as other industries implies the
existence of inter- and intra-industry cost linkages. The presence of these linkages,
together with trade costs, means that the number of firms producing in the region
affects each firm’s costs. This can be seen from (9) together with (6), (4) and (2):
imported varieties bear trade costs, and the more varieties of a good that have to be
imported, the more costly is the good as an intermediate input (a higher Q in (9)).
The supplier-proximity considerations will, thus, encourage firms to locate in a
280
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
region with a large number of suppliers of important intermediates, as this leads to
lower marginal costs and makes them more competitive.
At the same time consumer-proximity considerations pull in the opposite
direction. This follows from (5), (4) and (2). A higher T increases the relative
price of imported varieties, and raises the demand for locally produced varieties.
This encourages producers to locate according to demand.
Finally, factor market effects work through Eq. (8). In the first ‘pure’ case with
identical relative factor endowments, concentration of all the industry in one
country drives up the price of primary factors (PV ) relative to the other countries.
Factor market competition is thus a force for de-concentration of industrial
activity, but does not necessarily imply the dispersion of individual industries.
With differences in relative factor endowments (the second pure case) an equal
division of industry across countries produces unequal factor prices (PV ). In this
case factor market competition dictates the concentration of individual industries.
That is, specialization according to comparative advantage.
The relative strength of concentration and de-concentration forces depends, as
discussed in Section 1, on the level of trade costs. For low trade costs factormarket considerations dominate, for high trade costs consumer-proximity considerations determine location, and for intermediate trade costs supplier-proximity
becomes the dominating factor.
Agglomeration forces do not directly affect the perfectly competitive sectors.
These sectors, however, expand or contract as a consequence of competition for
factors with the other sectors. The decreasing returns in these sectors (due to a
specific factor) act to dampen the expansion of the ITG sectors, as their presence
implies that factors are drawn into the ITG sectors at increasing cost.
2.2. Data and calibration
To calibrate the model we use actual 1992 data from Eurostat, GTAP and
NBER World Trade Flows (see Feenstra, 1997) for input–output tables, trade
flows and factor shares, of which more details are provided in Appendix A. With
respect to other key features, such as market structure, and demand and technology
parameters, on the other hand, we have to rely on secondary sources or pure
assumptions. In our model with large-group monopolistic competition and free
entry / exit, there is in equilibrium a one-to-one (inverse) relationship between
elasticity of substitution between varieties and scale elasticity. Consequently, we
can use data on scale elasticity to calculate the elasticity of substitution between
varieties.
The calibration procedure essentially solves the model backwards. That is, it
uses data for all variables and some parameters and solves the model for remaining
parameters. Table 2 provides an overview of the parameters in Eqs. (1)–(9).8
8
For the complete model and the complete set of parameters, see Forslid et al. (1999b).
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
281
Table 2
Overview of parameters
Parameters calibrated on data of production
trade, and input–output tables
aim , a ijm , gihm , bijk , BVij , BZij , BZPCij
Parameters based on secondary data sources
si , T ijm
Assumed parameter values
sq, s i , Stop i
Without proper testing of all parameters, the empirical applicability of the
simulations may be limited. Hence, the model exercise should be viewed as
‘theory with numbers’, rather than empirical results with direct policy implications.
2.3. Industry and region characteristics
Before we turn to model simulations, we present key characteristics of potential
importance for the results. In particular we focus on features that are expected to
influence the location of sectors as trade costs are lowered. Five factors affect the
strength of the backward and forward linkages in this model: trade costs,
elasticities of substitution, economies of scale, the input–output structure, and the
size of regions (home market effects). However, as noted above, there is a
one-to-one (inverse) relationship between elasticity of substitution between
varieties and scale elasticity. Location of industries is moreover affected by
standard comparative advantage — especially for low trade costs — due to
differences in endowment and technology.
Column (a) in Table 3 ranks industries in descending order according to trade
distortions (i.e. the combined effect of transport costs, import barriers and export
Table 3
Key industry characteristics (average values for the four integrating regions)
Textiles
Leather and products
Wood products
Metals
Minerals
Chemicals
Food products
Transport equipment
Machinery
Other manufacturing
(a)
Trade
distortions a
(b)
Scale
elasticity a
(c)
Own input
share
(d)
Intermediate
share
(e)
Unskilled /
skilled ratio
(f)
Labour /
capital ratio
4
7
5
6
2
3
1
10
9
8
10
9
5
4
6
2
7
1
3
8
0.294
0.187
0.268
0.366
0.130
0.297
0.158
0.145
0.169
0.026
0.561
0.543
0.555
0.634
0.486
0.603
0.655
0.570
0.489
0.335
3.40
3.44
2.16
2.43
2.33
1.58
2.43
2.02
1.53
2.30
3.28
3.46
3.41
3.94
1.84
2.51
1.74
4.09
3.76
3.87
0.204
0.543
2.36
3.19
Average
a
Rank in decreasing order.
282
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
taxes or subsidies). In the trade liberalisation experiments we lower trade costs
equiproportionately in all sectors, which implies that we would expect more
‘action’ in sectors with initially high trade costs. Let us therefore note that food
products, minerals, chemicals and textiles are all sectors with relatively significant
trade distortions.
Column (b) in Table 3 ranks industries in descending order according to scale
economies, and shows that scale economies are most important for transport
equipment, chemicals, machinery and metals.9 According to theory (see Krugman,
1980; Krugman and Venables, 1995; Amiti, 1998) we would, ceteris paribus,
expect these to agglomerate the most.
Industries purchase intermediates from own sector as well as from other sectors.
Columns (c) and (d) of Table 3 give a summary of key characteristics regarding
the average intermediate use. Column (c) shows the use of input from own sector
as share of output value; column (d) gives total use of intermediates from all
sectors as share of value of output. The ratio (c) /(d) hence indicates the
importance of intra- relative to inter-industry intermediate inputs. These shares —
disaggregated at country and industry level — are used to calibrate the parameters
in the cost function (9). A higher share of own industry inputs means that
supplier-proximity to firms from the same sector becomes more important. An
expansion of the own sector at home comes in the form of more domestically
located firms, which leads to a lower price index Q and therefore to lower
marginal costs. Hence, the use of intermediates from own industry creates a
positive feedback and makes agglomeration self-reinforcing. However, the use of
intermediates from other sectors may work both for and against agglomeration
depending on the location of the supplying sectors. A strong dependence on
sectors that are rather dispersed across regions or alternatively concentrated in
another region than the purchasing sector, discourages agglomeration. In general,
we would ceteris paribus expect industries with a strong bias towards use of inputs
from own industries (high (c)), and with intra-industry linkages that are stronger
than inter-industry linkages, to be relatively more concentrated geographically.10
From Table 3 we can see that textiles, wood products, metals and chemicals are
industries with an above average use of inputs from own sector and which also
have stronger within than between industry linkages.
For low trade costs, agglomeration forces become weak. Instead comparative
advantage forces will tend to dominate. Industries have different factor intensities,
which opens up for location of production based on comparative advantage. This
does, however, not necessarily imply a greater geographical dispersion of
production in an industry; depending on the interaction between the total set of
forces determining location, comparative advantage may reinforce or discourage
9
Our ranking of industries according to economies of scale follows Pratten (1988).
See Fujita et al. (1999) for a discussion of the impact of inter- versus intra-industry linkages on
agglomeration.
10
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
283
geographical concentration of industries. The final two columns of Table 3 show
factor intensities (averages for the four integrating regions). Chemicals, transport
equipment and machinery are skill-intensive sectors. Textiles and leather use
unskilled labour intensively (and will hence be labelled labour intensive), while
food products and minerals are capital intensive.
Table A.2 in Appendix A provides some regional characteristics for the four
integrating regions. It should be observed that South is relatively abundantly
endowed with unskilled labour, while North is relatively abundant in skilled
labour. As for capital endowment, Central and North are relatively more capital
abundant than South and West. In terms of relative size of the regions, West and
Central are of about the same size, while South is considerably smaller, and North
is only around 1 / 7 of the large core regions. Since we know that home market
effects may have a strong impact on the location of production, the relative size of
the regions may play an important role.
3. Economic integration and the location of production
We now turn to the question of how the pattern of industrial production may
change as trade impediments are dismantled between the four integrating regions
in our model. We first discuss the relocation of individual manufacturing sectors
resulting from trade liberalisation. In a simple two-region model it is obvious what
increased industrial concentration means, while in our case with four integrating
regions it is less clear. We therefore proceed by analysing changes in locational
patterns using concentration indices. These indices provide us with an overall
picture of the degree of industrial concentration. We first study such a concentration measure for each manufacturing industry individually, and then we look
at the total geographical concentration for all traded manufacturing production.
This latter measure indicates whether industries tend to agglomerate in the same or
in different regions.
Our model experiments consist of successive lowering of all three types of trade
costs (transport costs, tariffs and export taxes) with 10% per step, starting from the
benchmark situation.11 We do, however, also show the result for a few steps of
increase in trade costs. We focus on the imperfectly competitive, traded goods
(ITG) sectors. Agriculture and energy are modelled with perfect competition and
free trade, which implies that agglomeration forces are absent in these sectors.
Still, decreasing returns to scale in these sectors — due to specific factors — act to
11
It is well known from theoretical work that there may be multiple equilibria in models like this. We
cannot rule out such possibilities in the present model. However, extensive model experiments have not
revealed such multiple equilibria. One possible reason why multiple equilibria do not occur, may be the
non-negativity restrictions that are placed on most economic variables in the model.
284
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
dampen the expansion and concentration of the ITG sectors as labour and capital
have to be drawn into the ITG industries at increasing cost.
3.1. Changing patterns of production
We shall here describe the simulated production patterns, while leaving further
analysis of geographical concentration to the next section. Fig. 1 shows how
production in different sectors changes as trade costs are lowered between the four
regions.12 The horizontal axis depicts trade costs relative to the base case, i.e. trade
cost51 in the base case (e.g. 0.5 means half of base-case trade costs).
Three sectors — textiles, leather, and food products — show the most dramatic
patterns in terms of changing locations. Textiles move out of Central and into West
and South. Leather expands in South, while contracting in all other regions. Food
production leaves South and Central, moving into North but particularly into West.
Consider first textiles. For very low trade costs production abruptly disappears
from Central and agglomerates in West and South. The possibility of abrupt
changes in location as trade costs are lowered, is well known from theory (e.g.
Krugman, 1991). Table 3 shows that within-industry linkages are relatively strong
in textiles production, which implies that self-reinforcing forces of agglomeration
are likely to be important for the location of production; thus, the sector is a
candidate for strong relocation effects. It should also be noted that textile
production is a relatively small industry, implying that large swings in this sector
can occur without causing much pressure in the factor markets.
The reason why textiles expand so substantially in South seems rather clear:
textile production is one of the most (unskilled) labour-intensive industries, and
South has a comparative advantage in the production of labour-intensive goods.
But why does production of textiles move out of Central and into West, and not
vice versa? Factor endowments cannot explain this change in production patterns.
The presence of agglomeration forces, however, implies that even small, initial
difference may suffice to tip the balance in favour of one location. In our case
West does have a slightly larger initial textiles production than Central.
Another small industry is the leather industry, which exhibits a locational
pattern similar to textiles — with low trade costs leading to a core-periphery
outcome. The difference is that the relocation of production is more continuous
and that agglomeration only takes place in one region: South. The characteristics
of the leather sector are similar to textiles. However, in the base case the leather
production of South is more than twice as large as in any other region, which
together with South’s comparative advantage in labour-intensive production, is
12
For clarity, two sectors have been left out of the presentation when discussing the individual
sectors. These are other manufacturing which is a fairly heterogenous group of industries and wood
products which are strongly affected by subsidies in the North. The locational effects for all sectors are
presented in the working paper version of this paper: Forslid et al. (1999a).
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
285
Fig. 1. Production (billion US $).
certainly the main explanation for the resulting agglomeration in this region. The
more continuous relocation of this sector is consistent with a relatively low own
input share, and thus less significant intra-industry linkages.
The large swings in production of food products are linked to this industry’s
286
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
initially high trade costs (cf. Table 3). One surprise, perhaps, is that this industry
starts to agglomerate in North for low trade costs, even though this region initially
has production that is only one third or less of the other regions’ production
volume. The explanation seems to be that food products are relatively capitalintensive, which gives North a comparative advantage. Food products are also
characterised by rather low (increasing) returns to scale and a low own input
share, which ceteris paribus make proximity to a large market less important for its
location, and further justify the movement into the northern periphery when trade
costs go down.
What about the remaining ITG industries? Most of these industries exhibit
relatively stable patterns of localisation. It should, however, be noted that they
generally display a non-monotonous relationship between trade liberalisation and
location. Among these industries are the four sectors with the most significant
increasing-returns-to-scale technology (see Table 3): metals, chemicals, transport
equipment and machinery. In the base case they are all rather concentrated in the
two largest regions: Central and West. Substantial increasing returns to scale and
the presence of intra-industry linkages suggest that proximity to markets and
self-reinforcing agglomeration forces are important determinants of the location of
production in these industries. As trade costs are reduced, the sectors remain
concentrated in the core of Europe — close to the larger markets. This may, on the
one hand, reflect very strong IRS and intra-industry linkages. However, it could
also be the case that market size and factor abundance effects pull in the same
direction — maintaining concentration in the core as trade costs are reduced. If a
capital intensive industry initially is concentrated in a large capital abundant
country, a fall in trade cost reduces the market size incentive to stay concentrated,
but increases the factor market incentive to do so — and the result may typically
be relative stable patterns of localisation.
Taking a more aggregated perspective, Fig. 2 displays the share of ITG-industry
Fig. 2. ITG shares across European regions.
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
287
in each of the regions. North, being by far the smallest region, shows a distinct
U-shaped pattern with a loss of ITG-industry for intermediate trade costs for which
agglomerative forces reach a maximum. The large region Central exhibits an
inverse relationship between ITG share and trade costs. The region’s dominant
position in the ITG sectors is reinforced for intermediate trade costs, while it may
decline as trade costs are further reduced. West shows a monotonous increase in
ITG-share as trade costs are lowered, while the ITG share of South actually
follows a bell-shaped pattern, although it is not very distinct.
To conclude this section, we see that comparative advantages as well as
economies of scale and intra-industry linkages interact in determining the location
of industry. Due to differing industry characteristics, there are, however, large
differences across industries as to which factors are relatively more important as
determinants of location.
3.2. Industrial concentration
So far we have investigated the reallocation of industry sectors triggered by
economic integration. It is, however, not clear from this analysis whether industrial
production becomes more or less concentrated in Europe as a consequence of
integration. To take only a few examples, it is not obvious from Fig. 1 whether the
production of transport equipment gets more or less concentrated as integration
takes place; it is also difficult to see what happens to the concentration of e.g.
production of chemicals.
In this and the next section we shall focus on industrial concentration using
summary indices of concentration. We start by analysing the concentration
tendencies of individual industries, using a measure for absolute industrial
concentration of the following form,13
œSNO s 2SO s D D /NsN 2 1d,
Ci 5
]]]]]]]]]
2
2
j ij
j ij
with s ij 5 Xij / o j Xij being the share of production in industry i taking place in
region j, and with N depicting the number of integrating regions (i.e. four). The
index of absolute concentration, Ci , is the standard deviation of the distribution of
s ij . A high value of this statistic indicates a highly concentrated industry. Fig. 3
illustrates how increased integration affects the degree of concentration of the ITG
industries.
There appear to be two groups of industries. Consider first metals, chemicals,
transport equipment and machinery — a group of industries where concentration
displays an (inverted) U-shaped curve as the regions are integrated. These
industries all have high scale elasticities (cf. Table 3), indicating strong agglomera13
The differing sizes of the units (the regions) make relative indices a less attractive choice as a
measure of industrial concentration. See Haaland et al. (1999) and Midelfart Knarvik et al. (2000) for
discussion of the use of relative and absolute measures of concentration.
288
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
Fig. 3. Industrial concentration.
tion forces. When trade costs are reduced from a high level, concentration initially
increases. However, lower trade costs also decrease the agglomeration forces so
that, when a critical level of trade costs is reached, the process is reversed as factor
market pressures start to dominate. These pressures will tend to dampen the overall
tendency of industrial agglomeration, but whether this will make individual
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
289
industries concentrate more or less, depends on the interaction between industry
intensities and regional characteristics. We would expect industries with high
economies of scale to concentrate in large markets. Hence, it comes as no surprise
that initially they are all rather concentrated in the large regions Central and West,
with Central having the dominant position in all four industries. Metals and
chemicals moreover exhibit relatively strong intra-industry linkages. Even though
the changes in concentration that we observe are modest, there is a clear pattern:
when trade costs are reduced, agglomeration is first reinforced, confirming that the
forces of agglomeration are strongest for ‘intermediate’ levels of trade costs.
Declining concentration then follows this development. Metals experience the
most significant decline in geographical concentration (around 19%): Central’s
dominant position is reduced, while especially West increases its share of the
industry.
Textiles, leather, and food products constitute the second group of industries,
and these become increasingly concentrated as integration proceeds. Hence, the
industries where lower trade costs imply agglomeration, are exactly the same
industries as the ones that exhibit the most dramatic changes in location patterns.
As argued in the previous section, comparative advantage is a dominating factor
for textiles and leather, amplified by strong intra-industry linkages and the initial
pattern of production. Significant initial impediments to trade constitute an
important explanation for the large movements of the third sector in this group
(food products).
3.3. Overall concentration
Having studied concentration effects for individual industries, a natural question
is whether industries — to the extent that they become more concentrated — tend
to concentrate in the same or in different regions. In other words, will all
manufacturing activities tend to concentrate in the core, with de-industrialisation
of the periphery?
We measure the degree of overall industrial concentration by the following
index:
]]]]]]]]]
2
2
H5
N jhj 2
/NsN 2 1d,
j hj
œS
O
SO D D
with h j 5 o i Xij / o j o i Xij being the share of overall manufacturing production taking
place in region j, and with N depicting the number of regions subject to
integration.
Fig. 4 shows the concentration of the ITG sector. Agglomeration forces tend to
dominate for intermediate trade costs, while factor market effects become more
important for low trade costs. Although this U-shaped pattern is well known from
simple theoretical two-sector models, it is not obvious from theory that we should
get such a pattern in a multi-sector model. Even if agglomeration forces in each
290
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
Fig. 4. Overall industrial concentration.
sector would work like this, it could well be the case that the sectors would be
pulled in different directions. Whether they actually end up in the same or different
regions will depend on the trade-off between on the one hand agglomeration forces
through inter-industry linkages, and on the other hand general equilibrium factor
price effects. Fig. 4 indicates that — at least for intermediate trade costs —
agglomeration forces may yield increased overall concentration of manufacturing
activities as a consequence of economic integration.
4. Factor prices and welfare
Whereas the patterns of industrial concentration and specialisation are interesting phenomena in their own right, the main reason for the political interest
received by the new economic geography, is probably the theoretically based
presumption about a relationship between the pattern of production and specialisation and real national income. Moreover, from neoclassical trade theory there are
strong reasons to expect changes in national income to be unevenly distributed
among different factors of production. We therefore next investigate the effects of
economic integration on factor prices.
4.1. Factor prices
In a Heckscher–Ohlin framework, regional specialisation will have very
different impact on the factors of production in a region. The relatively abundant
factor will gain whereas the scarce factor will lose according to the Stolper–
Samuelson theorem. Indeed, it has been put forward as a serious problem of
economic integration that in regions well endowed with skilled labour — in our
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
291
Fig. 5. Real factor prices.
case North and West — integration tends to benefit skilled labour at the expense of
unskilled. Fig. 5 shows changes in real factor returns in the four integrating
regions relative to the benchmark situation (t 5 1).
Two features should be noted. Firstly, changes in factor prices are in general
rather correlated with the ITG-shares in the region (Fig. 2). Secondly, although
there is substantial co-variation in factor prices within a region, relative factor
prices do indeed mirror Stolper–Samuelson effects. For instance, for low trade
costs, skilled labour is a relative winner both in North and West, as skill-intensive
industries agglomerate in these regions, while low-skilled workers are the relative
winners in South.
4.2. Welfare
Traditional trade models would predict gains from specialisation according to
comparative advantages, but the theory would not predict that some industries are
‘more worth’ than other industries. New trade theory models allow industries to
differ — saying that there are potential gains from getting a larger share of
industries in which there are pure profits. In our model, free entry and exit of firms
in all industries eliminate pure profits. Yet, the interaction between the desire for
variety, imperfect competition, and industry linkages generates ‘agglomeration
rents’. These rents show up in the returns to factors of production and in the price
292
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
indices for imperfectly competitive goods sectors. For these sectors, trade costs
and the number of varieties matter for the price level of an industry aggregate (see
(2) and (4)); hence, the more local varieties there are, the lower is the price level
and the higher is real income and welfare. Since the magnitude of such rents
differs between industries, the location of manufacturing activities may have
important welfare implications. Ceteris paribus a region gains from getting more
of the industries with large agglomeration benefits relative to other industries.
Whether these gains are strong enough to outweigh other effects of relocation of
industry, is an empirical question; in this section we present some indicative
‘evidence’ of the importance of such rents for real income.
Fig. 6 shows the real GDP in each of the four integrating regions, relative to the
benchmark level (t 5 1). Two things should be noticed. Firstly, for successive
steps of liberalisation for trade costs between 1 and 0.5 all regions gain, in spite of
increased concentration (see Fig. 4). The welfare effects are, however, modest for
these intermediate trade costs. For high and low trade costs, on the other hand,
there are winners and losers. Secondly, the results in Fig. 6 together with Fig. 2
reveal a close link between a region’s specialisation in imperfectly competitive
traded goods (ITG) industries and real income. Fig. 2 shows the aggregate share of
total value of production taking place in the ITG industries. In reality, the
ITG-share is only a crude measure of the agglomeration rents created in the
imperfectly competitive sector, as there is significant variation with regard to these
rents across the ITG industries. Nevertheless, the results are quite illustrative,
suggesting that there are tight links between specialisation in ITG industries and
real GDP.
Fig. 6. Real GDP relative to benchmark case (t 5 1).
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
293
5. Conclusions
This paper applies a full-scale CGE model to study the locational effects of
economic integration. Our experiment is to gradually reduce trade costs, and study
the pattern of industrial concentration in individual industries as well as for
manufacturing as a whole. This should be viewed primarily as an exercise to get
numeric intuition about higher order properties of simple theoretical trade and
location models (e.g. Fujita et al., 1999). We use the EURORA model (Forslid et
al., 1999b), which has a complete input–output structure in the sense that all
linkages across the 14 sectors in the model are taken into account and are
modelled in detail.
From stylised theoretical models we know that in a setting with imperfect
competition, trade costs and intra- and inter-industry linkages, there is a trade-off
between forces working for and against agglomeration. Location is determined by
the interaction of consumer-proximity considerations, supplier-proximity considerations, and factor-market competition. On the basis of the theoretical economic
geography models there is moreover one possible outcome of economic integration
that has been especially emphasised; a bell-shaped (inverted U-shape) relationship
between trade costs and geographical concentration of the imperfectly competitive
sector; where agglomeration forces dominate for intermediate trade costs. The
question, however, is to which extent this result comes through in a multi-sector,
multi-factor, multi-region framework — and if so, at what level do we typically
see it? At the level of the individual industry, or at the level of the manufacturing
sector?
Several results follow from our simulations. Firstly, on the industry level the
locational response to lower trade costs depends on industry characteristics. For a
number of industries we find an (inverted) U-shaped relationship between
integration and concentration — industrial concentration is low for high and low
trade costs, and higher for intermediate trade costs. A common feature for these
industries is that there are significant increasing returns to scale and important
intra-industry linkages. For other industries there is a monotonous increase in
concentration as trade costs are lowered. These industries are typically industries
in which scale economies are less important, but where initial trade costs have
prevented sufficient specialisation according to comparative advantage.
Secondly, on the aggregate level of the imperfectly competitive sectors, we find
that the (inverted) U-shaped relationship between economic integration and overall
concentration in manufacturing, well known from stylised theoretical models,
actually carries over to a more realistic higher-order setting with multiple product
and factor markets.
Thirdly, we investigate the effects on factor prices and welfare. Our simulations
show how agglomeration forces as well as standard comparative advantage forces
drive factor prices. On the one hand factor prices in a region tend to co-vary
because of agglomeration forces, but on the other hand relative changes go in the
294
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
Stolper–Samuelson direction. This implies that the concerns regarding intraregional income distribution associated with trade liberalisation in a traditional
trade model may possibly be muted, while the distribution of real income between
regions is a potential issue of concern. Our results show a correlation between real
income gains and growth in manufacturing production. However, in our simulations, contrary to stylised theoretical models of trade and location where real
wages and welfare fall sharply in the periphery when agglomeration occurs, all
regions may actually gain during a large part of the integration process in spite of
increasing concentration.
Acknowledgements
An earlier version of this paper has been presented at the SNEE conference,
¨
Molle,
and at the European Research Workshop in International Trade (ERWIT) in
Bergen, both in June 1999. Thanks to the participants, and in particular to Richard
Baldwin, Karolina Ekholm, Joe Francois, Michael Gasiorek, Jim Markusen, Victor
D. Norman and Diego Puga for comments and suggestions. We are also grateful to
two anonymous referees and the editor for very useful comments. Financial
support from the Research Council of Norway grant no. 124559 / 510 and from the
European Commission (TMR) is gratefully acknowledged.
Appendix A
Data
Data sources for input–output tables, trade flows and factor shares are
EUROSTAT (EU input–output tables), GTAP (Global Trade Analysis Project)
database and NBER World Trade Flows (see Feenstra, 1997). A detailed
description of data, data sources, and how the benchmark data set was constructed,
can be found in Forslid et al. (1999b) or obtained from the authors upon request.
Forslid et al. (1999b) also provides a descriptive analysis of the data material,
focusing on the distribution of production across regions, trade flows and trade
volume, differences in technology and factor use across industries. Our data on
economies of scale have been taken from Cawley and Davenport (1988), who base
their estimates on Pratten (1988). Transport costs and data on trade barriers are
from GTAP versions 3 and 4.
Regions and regional characteristics
Table A.1 gives details of the regions.
Table A.2 presents some key characteristics for the four integrating regions. As
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
295
Table A.1
Regions
Regions
Description
Central
North
South
West
Austria, Denmark, Germany, Switzerland
Finland, Iceland, Norway, Sweden
Greece, Italy, Portugal, Spain
BeNeLux, Ireland, France, UK
Europe East
Bulgaria, Hungary, Czech. Rep., Poland,
Romania, Slovakia, Slovenia
Former Soviet Republics including Estonia,
Lithuania, Latvia
China, India, Bangladesh, Bhutan, Maldives,
Nepal, Pakistan, Sri Lanka
South East Asia including Japan
USA and Canada
Other nations not elsewhere included
Former Soviet Union
China and South Asia
South East Asia
USA and Canada
Rest of the world
Table A.2
Relative factor endowments and relative size
Unskilled / skilled
labour force
Labour / capital
stock
Share of European
GDP a
Central
North
South
West
4.02
1.69
7.41
2.77
0.019
0.017
0.038
0.037
34.5%
5.8%
24.3%
35.5%
Mean
2.99
0.028
a
Base case (1992) model data.
we cannot separate factor prices and factor stocks in our benchmark data, we use
the factor endowments data provided by Maskus and Penubarti (1995).
References
Allen, C., Gasiorek, M., Smith, A., 1998. The competition effects of the single market in Europe.
Economic Policy 27, 439–486.
Amiti, M., 1998. Inter-industry trade in manufactures: does country size matter? Journal of International Economics 44, 231–255.
Baldwin, R.E., Forslid, R., Haaland, J.I., 1996. Investment creation and investment diversion in Europe.
The World Economy 19, 635–659.
Baldwin, R.E., Francois, J.F., Portes, R., 1997. The costs and benefits of eastern enlargement: the
impact on the EU and central Europe. Economic Policy 24, 125–176.
296
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
Brown, D.K., Deardorff, A.V., Stern, R.M., 1992. A North American Free Trade Agreement: analytical
issues and a computational assessment. The World Economy 15, 11–30.
Brown, D.K., Deardorff, A.V., Fox, A.K., Stern, R.M., 1995. Computational analysis of goods and
services liberalization in the Uruguay Round. In: Martin, W., Winters, L.A. (Eds.), The Uruguay
Round and the Developing Economies. Cambridge University Press, Cambridge, pp. 365–380.
Cawley R., Davenport M., 1988. Partial equilibrium calculations of the impact of internal market
barriers in the European Community. In Commission of the European Communities: Research on
the cost of non-Europe, Vol. 2: Studies on the economics of integration, Luxembourg.
Davis, D.R., 1998. The home market, trade and industrial structure. American Economic Review 88,
1264–1276.
EFTA, 1994. European Economic Integration. Effects of ‘1992’ on the Services Sectors of the EFTA
Countries. EFTA, Occasional papers, no. 49.
Feenstra, R.C., 1997. World Trade Flows, 1970–1992, with Production and Tariff Data. NBER,
Working paper no. 5910.
Forslid, R., Haaland, J.I., Midelfart Knarvik, K.H., 1999a. A U-shaped Europe? A simulation study of
industrial location. CEPR, Discussion paper no. 2247.
Forslid, R., Haaland, J.I., Mæstad, O., Midelfart Knarvik, K.H., 1999b. Integration and Transition:
Scenarios for Location of Production and Trade in Europe. NHH, Discussion paper no. 13 / 99.
Francois, J.F., McDonald, B., Nordstrom, H., 1995. Assessing the Uruguay Round. In: Martin, W.,
Winters, L.A. (Eds.), The Uruguay Round and the Developing Economies. Cambridge University
Press.
Fujita, M., Krugman, P.R., Venables, A.J., 1999. The Spatial Economy: Cities, Regions and International Trade. MIT Press.
Gasiorek, M., Smith, A., Venables, A.J., 1991. Completing the internal market in the EC: factor
demands and comparative advantage. In: Venables, A.J., Winters, L.A. (Eds.), European Integration:
Trade and Industry. Cambridge University Press.
Gasiorek, M., Smith, A., Venables, A.J., 1992. ‘1992’: trade and welfare — a general equilibrium
analysis. In: Winters, A. (Ed.), Trade Flows and Trade Policy after ‘1992’. CEPR and Cambridge
University Press.
Haaland, J.I., Norman, V.D., 1992. Global production effects of European integration. In: Winters, A.
(Ed.), Trade Flows and Trade Policy after ‘1992’. CEPR and Cambridge University Press.
Haaland, J.I., Kind, H.J., Midelfart Knarvik, K.H., Torstensson, J., 1999. What determines the
economic geography of Europe? CEPR, Discussion paper no. 2072.
Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1995. Quantifying the Uruguay Round. In: Martin, W.,
Winters, L.A. (Eds.), The Uruguay Round and the Developing Economies. Cambridge University
Press, pp. 215–284.
Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1996. Increased competition and completion of the market
in the European Union: static and steady state effects. Journal of Economic Integration 11, 332–365.
Keuschnigg, C., Kohler, W., 1996. Austria in the European Union: dynamic gains from integration and
distributional implications. Economic Policy 22, 155–212.
Krugman, P.R., 1980. Scale economies, product differentiation, and the pattern of trade. American
Economic Review 70, 950–959.
Krugman, P.R., 1991. Increasing returns and economic geography. Journal of Political Economy 99,
483–499.
Krugman, P.R., Venables, A.J., 1995. Globalization and the inequality of nations. Quarterly Journal of
Economics CX, 857–879.
Lopez-de-Silanes, F., Markusen, J.R., Rutherford, T.F., 1994. The auto industry and the North
American Free Trade Agreement. In: Francois, J.F., Shiells, C.R. (Eds.), Modeling Trade Policy:
Applied General Equilibrium Assessments in North American Free Trade. Cambridge University
Press, pp. 223–255.
Maskus, K.E., Penubarti, M., 1995. How trade related are intellectual property rights? Journal of
International Economics 39, 227–248.
R. Forslid et al. / Journal of International Economics 57 (2002) 273 – 297
297
Midelfart Knarvik, K.H., Overman, H., Redding, S., Venables, A.J., 2000. The Location of European
Industry. Report Prepared for the Directorate General for Economic and Financial Affairs, European
Commission. European Commission, Brussels, Economic papers no. 142.
Pratten C. 1988. Partial equilibrium calculations of the impact of internal market barriers in the
European Community. In Commission of the European Communities: Research on the cost of
non-Europe, Vol. 2: Studies on the economies of integration, Luxembourg.
Puga, D., Venables, A.J., 1996. The spread of industry: spatial agglomeration in economic development.
Journal of the Japanese and International Economies 10, 440–464.
Roland-Holst, D.W., Reinert, K.A., Shiells, C.R., 1994. A general equilibrium analysis of North
American economic integration. In: Francois, J.F., Shiells, C.R. (Eds.), Modeling Trade Policy:
Applied General Equilibrium Assessments in North American Free Trade. Cambridge University
Press.