Integr ation and Industr ial Specialization in the
Eur opean Union.
K.H. Midelfar t-Knar vik
Norwegian School of Economics and Business Adminitsration and CEPR
H. G. Over man
London School of Economics and CEPR
Stephen Redding
London School of Economics and CEPR
A.J . Venables
London School of Economics and CEPR
* This paper draws on ‘The location of European Industry’, by the same authors, Economic
Papers no.142 (2000), European Commission, Directorate-General for Economic and Financial
Affairs, and ‘Comparative advantage and economic geography: estimating the location of
production in the EU.’ by K.H. Midelfart-Knarvik, H.G. Overman and A.J. Venables, CEPR
Discussion paper no. 2618
1. Intr oduction
European integration has caused substantial increases in trade flows between member states. The
growth in trade can be thought of as comprising two elements. One is ‘intra-industry’ trade.
This takes the form of countries importing and exporting similar products as firms from each
country penetrate the markets of other member states. Intra-industry trade can occur even if
countries have identical industrial structures, and its economic benefit derives from the fact that
it is pro-competitive, intensifying competition between firms located in different countries. The
other element is inter-industry trade: economic integration is likely to lead to growing trade
between sectors, as economies come to specialize more. The objective of this paper is to assess
the extent to which this process of specialization has occurred, what its determinants are, and
whether it is likely to continue.
A process of specialisation can occur for several different reasons. One is traditional
comparative advantage. Countries differ in their endowments of labour skill and other inputs to
production, or in their technological capacities. Trade allows sectors to expand and contract in
response to these advantages, to better match production to each country’s supply potential.
Comparative advantage can also derive from geography. Proximity to markets might be
particularly important for some sectors, in which case these sectors will relocate towards regions
with good market access – the centre rather than the periphery of Europe. A further driver of
specialisation is a process of agglomeration or clustering of activities. It might be that no country
or region has any inherent comparative advantage in a sector, but that firms derive advantage
from the presence of other firms in the sector. The advantage could be due to knowledge
spillovers, sharing a pool of common labour skills, or just forward and backward linkages
between supplier and customer firms. Any of these forces will tend to induce firms to cluster
together in particular locations, although the choice of location might be a matter of historical
accident. Integration facilitates this sort of clustering, and it will show up as a process of
increasing national (and regional) specialisation.
Specialization brings with it both benefits and costs. The benefits are that gains from
comparative advantage depend on countries actually specialising according to their advantages.
Furthermore, clustering usually brings with it efficiency benefits, as firms can interact more
efficiently with neighbouring firms engaged in similar or related activities. But a possible cost
is that specialisation increases the vulnerability of countries or regions to economic shocks – for
example, the contraction of a particular industry in which a region is highly specialised. The
magnitude of this potential problem depends on, amongst other things, the effectiveness of
international (or inter-regional) adjustment mechanisms, and might be greater in monetary union,
when adjustment via the exchange rate has been given up.
In this paper we argue that a process of specialisation has been occurring in the EU and that – on
the basis of comparison with the US – the process may have some way further to go. However,
the process has been gradual, and has not been dramatic enough either to cause major adjustment
problems, or to significantly increase the vulnerability of regions to industry specific shocks.
1
2. The specialization of countr ies
Our focus will be on specialisation within the manufacturing sector, but it is useful to first
sketch out the overall distribution of manufacturing in the EU. Table 1 shows the location of
the European manufacturing sector as a whole. At the beginning of the 1970s, more than
63% of manufacturing was located in the UK, France and Germany (countries accounting for
around 52% of Europe’s population), and over the last three decades this share has fallen,
standing at 59% in 1994/97. Southern European countries (Italy, Greece, Portugal and Spain)
raised their share gradually, from 19.9% in the early 70's to 24.6% in 1988/91 (compared to a
population share of 32%). The smaller countries – Austria, Finland and Ireland -- have seen
a steady increase in their share of European manufacturing. The changes are due to different
overall growth rates in each economy, and to changes in the share of manufacturing in each
country.
Table 1: Regional str uctur e of Eur opean manufactur ing
Austria
Belgium
Denmark
Spain
Finland
France
UK
Germany
Greece
Ireland
Italy
Netherlands
Portugal
Sweden
70/73
2.1 %
3.9 %
1.4 %
5.8 %
1.3 %
16.9 %
16.9 %
29.4 %
0.7 %
0.4 %
12.5 %
4.3 %
0.9 %
3.6 %
100.0 %
82/85
2.4 %
3.3 %
1.4 %
6.3 %
1.8 %
16.4 %
15.5 %
27.7 %
1.0 %
0.7 %
14.5 %
4.3 %
1.2 %
3.3 %
100.0 %
88/91
2.5 %
3.4 %
1.3 %
6.3 %
1.8 %
15.6 %
14.3 %
28.8 %
0.7 %
0.7 %
16.4 %
3.9 %
1.2 %
3.2 %
100.0 %
94/97
2.4 %
3.8 %
1.6 %
6.5 %
1.7 %
15.1 %
13.9 %
30.0 %
0.7 %
1.2 %
14.5 %
4.3 %
1.4 %
3.1 %
100.0 %
UK+GER+FRA
ESP+ITA+GRC+PRT
63.2 %
19.9 %
59.6 %
23.0 %
58.7 %
24.6 %
59.0 %
23.1 %
Against this background we now turn to look within manufacturing, and see if countries have
become more or less specialised in particular manufacturing industries. We have consistent
data for 36 manufacturing industries, across 14 EU countries and covering the period 197097. It is possible to construct an index of specialization that compares the share of a
particular country’s manufacturing in each industry with the share of industries in the rest of
the EU (see appendix for details). The index takes value zero if the country has industrial
structure identical to that of the rest of the EU, and is higher the more different is the
industrial structure.
Table 2 gives values of this index for each EU country at a series of dates. Two striking
points emerge. First, countries became more less specialised between 1970/73 and 1980/83,
2
as the index fell for 10 of the 14 countries. But from the early 80's onwards, the index
registered a steady and substantial increase for all countries except 1 (the Netherlands),
indicating a growing process of specialisation.
Table 2: How differ ent ar e countr ies fr om the r est of the EU?
70/73
80/83
88/91
94/97
Austria
Belgium
Denmark
Spain
Finland
France
G. Britain
Germany
Greece
Ireland
Italy
Netherlands
Portugal
Sweden
0.314
0.327
0.562
0.441
0.598
0.204
0.231
0.319
0.531
0.701
0.351
0.508
0.536
0.424
0.275
0.353
0.553
0.289
0.510
0.188
0.190
0.309
0.580
0.623
0.353
0.567
0.478
0.393
0.281
0.380
0.585
0.333
0.528
0.207
0.221
0.354
0.661
0.659
0.357
0.547
0.588
0.402
0.348
0.451
0.586
0.338
0.592
0.201
0.206
0.370
0.703
0.779
0.442
0.517
0.566
0.497
Weighted average
0.326
0.302
0.33
0.351
Note: Minimum values for each country in bold font.
Figure 1 combines this information for countries grouped according to their entry date -- so
EC1 is the original 6, EC2 is Denmark, Ireland and the UK, EC3 is Greece, Spain and
Portugal, and EC4 Austria, Finland and Sweden. The main point to note from this figure is
not the level of the curves, but their slopes. There has been a more or less steady increase in
the specialisation of the EC1 group over the entire period. For the other groups the index
starts to increase after their entry to the EU. This suggests that, as theory would suggest,
integration in the EU has led to increasing national specialisation.
3
EC3
EC2
EC4
EC1
A further application of this specialization index is informative. So far, we have constructed
the index to measure the difference between one country’s industrial structure and the
structure of the rest of the EU. We can also construct it bilaterally, to measure the difference
between the industrial structures of any two selected countries. Table 3 reports some of these
bilateral comparisons, just for France, the UK and Germany, and comparing 1994-97 with
1980-83. Looking along each row of the table we see that the large economies are most like
each other – the lowest values of the indices occur within the group France, Germany, UK.
However, these countries are most unlike Greece, Portugal, Ireland and Finland. (Although
we do not report the full results for these countries, Greece and Portugal are similar to each
other, although very different from Ireland and Finland).
Turning to changes, we see that over the period UK economy becomes more different from
all others except France and the Netherlands: France becomes more different from all others
except the UK and the Netherlands: and Germany becomes more different from all except the
Netherlands. Once again then, this confirms the picture of divergence of industrial structures
from the early 1980s onwards.
4
Table 3: How differ ent ar e countr ies fr om each other ?
Aus
Bel
Den
Spa
Fin
Fra
UK
Ger
Gre
Ire
Ita
Net
Por
Swe
France
1980-83
1994-97
0.38 0.34 0.57 0.26 0.49 0.00 0.22 0.31 0.57 0.63 0.39 0.51 0.47 0.41
0.43 0.44 0.57 0.33 0.62 0.00 0.19 0.35 0.69 0.78 0.51 0.46 0.55 0.51
UK
1980-83
1994-97
0.32 0.42 0.56 0.37 0.54 0.22 0.00 0.25 0.61 0.67 0.40 0.53 0.55 0.39
0.39 0.48 0.58 0.38 0.58 0.19 0.00 0.36 0.72 0.77 0.47 0.46 0.59 0.51
Germany
1980-83
1994-97
0.33 0.43 0.65 0.40 0.66 0.31 0.25 0.00 0.73 0.75 0.43 0.64 0.64 0.42
0.46 0.61 0.72 0.43 0.66 0.35 0.36 0.00 0.86 0.82 0.49 0.61 0.74 0.49
3. National specialisation and industr ial concentr ation.
The previous section tells us that a process of specialisation is going on, but what sorts of
industries are going to what sorts of countries, and what underlying factors are driving these
changes? Simple description is not very helpful (since there are 36 x 14 changes to describe).
There are several ways in which we can try to organise the information to answer these
questions.
3.1 What sor t of industr ies do countr ies attr act?
First, we can identify selected characteristics of industries, and ask what countries are
attracting or losing industries with those characteristics. For example, industries are
classified by the OECD into high, medium and low technology groups, which can be scored
2, 1, 0. We can compute the average score for each country (industry scores weighted by the
share of the industry in the country), and compare these scores across countries and across
time. Several results come through very clearly. The technology scores are highest for
Germany, France, the UK and Sweden, although they decline slightly for all these countries
except Sweden. The largest changes are, as would be expected, for Ireland and Finland: these
both start low and increase dramatically, measuring the growing share of high technology
industries in these countries. The lowest levels are in Greece and Portugal, which start and
remain low.
Another classification of industries, is by their use of high skilled labour. Taking a fixed
measure of this for each industry, we can compute a score for each country. We find that the
Netherlands has the highest score on this characteristic, indicating an industrial composition
skewed towards skill intensive sectors. The Netherlands is followed by the UK, Germany,
France, Belgium and Sweden, although all these countries are overtaken by Ireland, rising
from near lowest to second highest from the early 80s to mid 90s.
5
3.2 Which industr ies ar e spatially concentr ated?
A second approach is to group industries according to their initial locational pattern, and then
investigate major changes that have occurred. Table 4 classifies industries according to the
extent to which they are spatially concentrated or dispersed across Europe. To measure
concentration we used the Gini coefficient. Details on calculation of the coefficient and a
complete table of indices for all industries are found in the appendix.
Industries in the category CC were the most spatially concentrated in both 1970/73 and
1994/97, and industries in category CD were initially concentrated but became dispersed by
1994/97.
Table 4: Industr ies gr ouped by levels and changes in concentr ation.
Concentrated industries that have remained
concentrated over time; (CC)
Concentrated industries that have become less
concentrated; (CD),
Motor Vehicles
Motor Cycles
Aircraft
Electrical Apparatus
Chemical Products NEC
Petroleum & Coal Products
Beverages
Tobacco
Office & Computing Machinery
Radio, TV & Communication
Professional Instruments
Machinery & Equipment NEC
Dispersed industries that have become more
concentrated over time; (DC)
Dispersed industries that have stayed
dispersed; (DD)
Textiles
Wearing Apparel
Leather & Products
Furniture
Transport Equipment NEC
Food
Wood Products
Paper & Products
Printing & Publishing
Metal Products
Non-Metallic Minerals NEC
Shipbuilding
(Residual group of 12 industries omitted).
Amongst the CC industries, the motor vehicles and motorcycles sectors saw consolidation of the
German position relative to both France and the UK, although this is slightly offset by the
increases in shares of production occurring in Portugal, Austria, and Spain. For Aircraft,
Germany, the UK and France remain the dominant countries with a 78% share of EU Aircraft
production in 1997. In the electrical apparatus industry there was little relocation, although
Austria and Italy registered slight increases in their shares. In Chemicals the UK, Germany and
France remain dominant despite Spain and Ireland capturing around 6% of the industry.
The second group of industries, CD, started off concentrated but experienced dispersion over the
period. Some, although not all, of these are relatively fast growing industries. In particular,
Office & Computing equipment and Radio, TV & Communication Equipment experienced a
6
major decline in geographical concentration between 1991 and 1997. The increased geographical
dispersion is primarily driven by decreasing German dominance and reinforced by shrinking
shares of the UK and France. In Office & Computing Machinery, Machinery & Equipment,
Radio, TV & Communication Equipment and Professional Instruments, between 7% and 17%
of the EU production left Germany, France and the UK. Countries that strengthened their
positions in some, or all, of these industries, were small countries such as Austria, Finland,
Ireland and Sweden; and also the Southern European countries Italy, Portugal, and Spain.
The third group are industries that were initially dispersed and have become more spatially
concentrated. These are typically slow growing or declining industries. Thus, France, Germany
and the UK experienced reduced shares in Textiles, Wearing Apparel and Leather & Products,
while the Southern European countries showed growing shares.
The final group of industries, DD, are industries that have been relatively dispersed throughout
the period. These include a mixture of industries, many of which are highly dependent on local
markets or face high transport costs.
4. Deter minants of specialisation.
To this point we have been simply describing changes that have occurred. Is it possible to go
behind the description, and explain the changing patterns of industrial location? Economic
reasoning tells us that industrial location is determined by the interaction between characteristics
of industries and characteristics of countries. Thus, skilled labour intensive industries are likely
to be located in countries that are abundant in skilled labour. Industries that are intensive users
of intermediate goods are likely to locate where they have good access to supply of these goods,
and so on. It is possible to get data on some, if not all, of the relevant industry and country
characteristics, and then use econometric techniques to explore the hypotheses suggested by
theory. What do we learn about the fundamental determinants of industrial specialisation in
Europe when we do this?
The six pairs of interactions that are used in the econometric study are listed in table 5, together
with a final column summarising the importance of the interaction. It is worth discussing each
of these interactions in turn. The first is between countries’ endowments of skilled labour (the
share of the population aged 25-59 with secondary or higher education) and industries’ ratio of
non-manual to manual workers. This turns out to be a highly significant interaction, and one that
increases in importance over time. The order of magnitude of the effect is indicated by the fact
that increasing the share of a country’s labour force that has secondary or higher education by
10% will, other things being equal, increase output in Professional Instruments, Drugs and
Medicines, Office and Computing equipment, and Printing and Publishing by around 40%: most
other machinery and electrical equipment sectors would experience a rise of around 10%. Of
course, other sectors have to shrink to accommodate these expanding sectors. For example, the
estimates indicate that Motor Vehicles would contract by 4%, Non-Ferrous metals by around
10%, and Footwear, Textiles, Apparel, and Leather sectors by between 5 and 12%. These
numbers indicate how quite small changes in endowments will have magnified effects on
increasing output in some sectors, while leading to the contraction of others.
7
The second interaction is between the endowment of scientists and the R&D intensity of
industries, and this too is highly significant and of increasing importance. The third indicates that
industrial sectors that use agricultural inputs do, as expected, locate in countries with relatively
large agricultural sectors, although the effect is weaker than the previous two interactions.
Remaining interactions deal with countries’ geographical characteristics. Market potential is a
measure of the centrality of each country, based on its average distance from sources of demand.
This is interacted with transport costs (interaction 5), with the expectation that countries with
good market potential might attract industries with high transport costs: however, no significant
relationship was found. Interaction 6 investigates the importance of ‘backwards linkages’ in
determining the location of industry. The hypothesis is that upstream industries – those with a
high share of their sales going to industry rather than to final consumers – will locate in countries
that have a relatively high market potential for sales of intermediate goods. The econometrics
reveals that these backward linkages are important, although diminishing in strength during the
period. Finally (interaction 6), are ‘forward linkages’. Do industries that are downstream (i.e.
heavily dependent on using intermediate goods) locate in countries that have good access to
industrial suppliers? In the 1980s this effect was extremely weak, although it increases in
importance, becoming a significant determinant of industrial location by the mid 1990s.
Table 5: Deter minants of industr ial str uctur e: countr y and industr y inter actions.
Countr y Char acter istic
Industr y Char acter istic
Effect
1
Secondary and higher education,
as % population
Non-manual workers
relative to manual
Significant,
small increase
2
Researchers and Scientists as %
of labour force
R&D, as % value added
Significant,
large increase
3
Agricultural production as %
GDP
Agricultural input as %
of total costs
Weak,
small increase
4
Market Potential
Transport costs
?
5
Market potential intermediates,
relative to mp final goods.
Sales to industry as %
total sales
Significant,
large decrease.
6
Access to suppliers
Intermediate goods as %
of total costs
Weak,
large increase
Summarising then, the econometrics paints a quite robust picture of the changing interaction
between factor endowment and economic geography determinants of location. The results
indicate an increasing importance of forward linkages and of the availability of skilled labour and
researchers in determining the location of industry from 1980 onwards.
8
5. An EU-US compar ison
We have seen that there is an ongoing process of industrial specialization in the EU, but how
much further is this process likely to go? A natural place to look for an answer to this question
is the US, a market that has always been tariff free and with a common currency, while
experiencing major reductions in internal transport costs as various transport revolutions
occurred.
Recent US experience has been quite different from that of the EU, as the specialisation of US
states has been decreasing. The US has a long time series of data which indicates increasing state
specialisation up to the 1940s, followed by a steady decline since. Only two out of 21 US
manufacturing industries do not record a decrease in spatial concentration between 70/73 and
94/97 (Tobacco products and Textile mill products).
More importantly for present purposes is a comparison of the absolute levels of industrial
specialisation in the two continents. This comparison is inherently difficult, as the continents
have quite different sizes and shapes, and there is no ‘correct’ way to aggregate US states to
mirror the geography of European countries. However, the two approaches we follow below lead
us to the conclusion that US industry remains somewhat more spatially concentrated than is
European.
5.1 The motor vehicle industr y: A US-EU compar ison
We start by making a comparison of the location of the motor vehicle industries in the two
continents. The basis of the comparison is to identify the European countries and US states that
are the main producers of motor vehicles, and then compare their shares of vehicle production
with their shares of manufacturing as a whole. If this share is high, then these countries/ states
are highly specialized in motor vehicles
In table 6 the row ‘EU2' gives the share of EU motor vehicle production undertaken by the two
largest producing countries (Germany and France). Over the period, we see that their share of
total EU vehicle production increased somewhat (from 58% to 62%), while their share of
manufacturing as a whole decreased slightly, indicating a small increase in the relative
concentration of the industry in these two countries. The second row gives the number of US
states that produced approximately the same share of US motor vehicles as Germany and France
produced of Europe’s. We see that this went from two states in 1970 to six states in 1996. And
the share of manufacturing as a whole produced in these matching states went from 13% to 33%.
This indicates considerable dispersion of US motor vehicle production relative to European, but
also indicates that it remains more concentrated: 63% of US vehicles are produced in states
producing just 33% of overall manufacturing, whereas in Europe 62% is produced by countries
that produce 45% of overall manufacturing.
The final two columns of the table reproduce the same argument, but using the top 4 European
motor vehicle producers as comparators. These top 4 countries produce over 80% of EU
vehicles, and the matching number of US states rises from 10 to 13 over the period. However,
the matching US states are again more concentrated in motor vehicles relative to manufacturing
as whole, than are the EU countries.
9
Table 6 Eur opean and US motor vehicle pr oduction
1970
1982
Share
vehicle
Share
manuf
EU 2
58%
46%
US 2
56%
EU 4
US 10
1996
Share
vehicle
Share
manuf
Share
vehicle
Share
manuf
EU 2
59%
44%
EU 2
62%
45%
13%
US 4
61%
25%
US 6
63%
33%
86%
76%
EU 4
84%
74%
EU 4
82%
65%
87%
56%
US 12
84%
61%
US 13
82%
61%
5.2 Relative spatial dispersion.
The preceding example was based on looking at the share of motor vehicle production in a set
of regions (countries or states) relative to these regions production of manufacturing as a whole.
We can extend this principle, and construct an index of the location of each industry, relative to
the location of manufacturing as a whole (see the appendix for details on how the index of
location is constructed; referred to as spatial dispersion index). The ratio of EU to the US values
of these indices are presented, for each industry, in table 7. The index we construct is actually
one of dispersion, so that if a number in the table is greater than unity, it means that the EU
industry is more dispersed than its US counterpart.
Looking at table 7 we see that EU industry is relatively more spatially dispersed than its US
counterpart in 16 of the 21 industries reported (dropping in the mid 1990s to 15). In two
industries the EU is much more dispersed B Tobacco and Textile Mills products. In a few the
EU’s high level of dispersion relative to the US has declined somewhat B for example Motor
Vehicles and Apparel. In the majority of others the EU pattern was and remains more
dispersed than that of the US. While levels of the index appear quite modest, it can be shown
that the gap between the EU and the US is generally considerably larger than the changes that
have actually taken place in the EU since the early 80s. The analysis suggests then, that if
EU patterns of specialisation and relative dispersion are to attain US levels, there is still some
way to go.
10
Table 7: EU dispersion relative to US dispersion
Industry
413 Lumber and wood products
417 Furniture and fixtures
420 Stone, clay, and glass products
423 Primary metal industries
426 Fabricated metal products
429 Industrial machinery and equipment
432 Electronic and other electric equipment
435 Motor vehicles and equipment
438 Other transportation equipment
441 Instruments and related products
444 Miscellaneous manufacturing industries
453 Food and kindred products
456 Tobacco products
459 Textile mill products
462 Apparel and other textile products
465 Paper and allied products
468 Printing and publishing
471 Chemicals and allied products
474 Petroleum and coal products
477 Rubber and misc. plastics products
480 Leather and leather products
Manufacturing sector
Average
82/85
1.07
1.06
1.09
1.23
1.04
0.951
0.814
1.44
0.811
0.868
1.04
1.01
2.08
2.14
1.2
1.3
1.02
1.01
0.919
1.03
1.35
1
1.16
88/91
1.11
1.08
1.09
1.2
1.04
0.975
0.843
1.42
0.763
0.835
1.03
1.03
2.27
2.1
1.14
1.29
1.03
1.06
0.885
1.07
1.29
1
1.16
94/97
1.12
1.05
1.07
1.22
1.02
1.03
0.848
1.3
0.797
0.889
0.986
1.01
2.19
2.03
1.04
1.33
0.965
1.07
0.853
1.05
1.24
1
1.14
6. Conclusions.
It seems clear from the analysis of this paper that, from the early 1980s onwards, the
industrial structures of EU economies have become more dissimilar. This is as would be
predicted by trade theory (old and new) during a period of economic integration. The
changes are to be welcomed, particularly as the econometric study indicates that the changes
are driven – in part at least – by economic fundamentals. Production is moving to draw on
input supplies, a set of changes that is likely to bring gains in economic efficiency.
Several other characteristics of the change are noteworthy. First, it is slow. Over a fourteen
year period, most economies have only seen a few percent of their industrial production move
between sectors. Of course, more activity might be expected to show up if more disaggregate
data were available, but nothing in our results suggest that the process is particularly rapid.
This is good news, in so far as it suggests that the adjustment costs associated with structural
change are likely to be quite low.
Second, the process is in the opposite direction from the one we observe in the US. The US
saw states becoming increasingly dissimilar from 1860 until around 1940, but a considerable
amount of convergence has occurred since. Despite recent work in the area it is still not clear
what forces drive these trends for the US.
Is the process of growing dissimilarity in the EU likely to continue, or is it reaching some
11
limit? We see no evidence that it is reaching a limit. In so far as any direct comparisons with
the US are possible, it is likely that EU industry remains more dispersed than that of the US.
The time series record for Europe indicates no evidence of a slow down. And as we have
seen, the process is slow; economies are nowhere near pressing against the limits of complete
specialization.
12
Refer ences
Balassa, Bela (1965): “Trade liberalisation and revealed comparative advantage”, The
Manchester School of Economics and Social Studies 33: 99-123
Barro, Robert and Jong-Wha Lee (1993): “International Comparisons of Educational
Attainment”, NBER Working Paper no. 4349
Brülhart, Marius and Johan Torstensson (1996): “Regional integration, scale economies and
industry location in the European Union”, CEPR Discussion paper no. 1435
CEP II (1997): “Trade patterns inside the single market”, in European Commission: The
single market review, subseries IV: Impact on trade and investment
Ellison, Glenn and Edward L. Glaeser (1999): “The geographic concentration of industry:
Does natural advantage explain agglomeration?”, American Economic Review 89, Papers and
Proceedings: 311-316
European Commission (1995): “Fifth survey on state aid in the European Union in the
manufacturing and certain other sectors ”
Eurostat (1996): Labour costs 1992, Luxembourg - Brussels, 1996
Eurostat (1992): Earnings – Industries and Services 1990, Luxembourg - Brussels, 1992
Haaland, Jan I., Hans J. Kind and Karen Helene Midelfart Knarvik (1999): “What determines
the economic geography of Europe?”, CEPR Discussion paper no. 2072
Kim, Sukkoo (1995): “Expansion of markets and the geographic distribution of economic
activities: the trends in U.S. regional manufacturing structure, 1860-1987”, Quarterly Journal
of Economics 110: 881-908
Krugman, Paul R. (1991): Geography and trade, MIT press, Cambridge, Massachusetts
OECD (1994): “Manufacturing Performance: a Scoreboard of Indicators”
Pratten, Cliff (1988): “A survey of the economies of scale”, in Commission of the European
Communities: Research on the “cost of non-Europe”, vol. 2: Studies on the economics of
integration
United Nations (1993): “Industrial Statistics Yearbook 1991”, Vol. 1: General Industrial
Statistics, UN, New York 1993
13
Appendix
Data
Our main data source is the OECD STAN database. This provides production data for 13 EU
countries and 36 industries, from 1970 to 1997. We combine this with production data for
Ireland from the UN UNIDO database, giving us data on a set of 14 EU countries (the EU 15,
excluding Luxembourg).
The comparison of the economic geography of Europe with that of the US, draws on US State
level data for manufacturing employment, 1970-97. The US regional data on manufacturing
employment have been kindly provided by Gordon Hanson.
Measures of specialization
The basic unit of analysis is the activity level (measured, when using the production data, by
k
the gross value of output) of industry k in country i at time t, which we shall denote xi (t) .
We usually want to work with this expressed as a share, either of activity in the country, or
k
k
total EU activity in the industry. We call these shares vi (t) and si (t) respectively. Thus
vi (t) xi (t) / Mk xi (t),
k
k
si (t) xi (t) / Mi xi (t)
k
k
k
k
k
Thus vi (t) is the share of sector k in the total activity of country i, which forms the basis of
k
our analysis of countries in Section 2. si (t) is the share of country i in the total activity of
industry k, which is the basis of the industry analysis of Sections 3 and 4.
Ki(t) Mk abs vi (t) v̄i (t)
k
k
We call this the Krugman specialisation index, or K-spec (see Krugman, 1991). It takes value
zero if country i has an identical industrial structure to the rest of the EU, and takes maximum
value two if it has no industries in common with the rest of the EU.
Spatial separation index
The concentration index employed so far provides information about the extent to which each
industry is concentrated in a few countries, but does not tell us whether these countries are
close together or far apart. Using this measure, two industries may appear equally
geographically concentrated, while one is predominantly located in two neighbouring
countries, and the other split between Finland and Portugal. Distinguishing such patterns will
provide additional insights on the geography of individual industries, about cross industry
differences and about the driving forces of economic geography.
Hence, as a complement to the traditional concentration indices, we propose an index of
spatial separation, that can be thought of as a supra-national index of geographical location.
We define the spatial separation of industry k, (SP k) as follows:
SP
k
C Mi Mj (s i sj δij)
k
14
k
k
where δij is a measure of the distance between i and j, si is the share of industry k in location
k
i, and C is a constant. For a given location i, Mj (s j δij) is the average distance to other
production in industry k. The first summation adds this over all locations i, weighted by their
k
k k
share in the industry, si . The interpretation of Mi Mj (si s j δij) is therefore a production
weighted sum of all the bilateral distances between locations. The measure is zero if all
production occurs in a single place, and increases the more spatially separated is production.
Gini coefficient of concentration
To measure the degree of concentration, we use the Gini coefficient of concentration. If all
countries have the same amount of manufacturing this measure is zero; if all manufacturing is
in a single economy it would take value 1.The Gini coefficient of concentration measures the
k
dispersion of a distribution of absolute production shares, si (t) , across countries for a given
k
industry. The Lorenz curve associated with the coefficient has cumulated si on the vertical
(as before), cumulated number of locations on the horizontal (each interval with the same
k
width, 1/N). Locations are ranked by si (the gradient of the Lorenz curve).
15
Table A1: Gini Coefficient of Concentration
NO
NAME
ISIC
70/73
82/85
88/91
94/97
1
Food
3110
0.503
0.471
0.464
0.46
2
Beverages
3130
0.647
0.592
0.576
0.557
3
Tobacco
3140
0.662
0.622
0.624
0.592
4
Textiles
3210
0.554
0.561
0.589
0.566
5
Wearing Apparel
3220
0.575
0.587
0.61
0.613
6
Leather&Products
3230
0.547
0.62
0.668
0.685
7
Footwear
3240
0.594
0.641
0.672
0.669
8
Wood Products
3310
0.533
0.477
0.482
0.498
9
Furniture & Fixtures
3320
0.568
0.584
0.59
0.596
10
Paper & Products
3410
0.504
0.483
0.488
0.479
11
Printing & Publishing
3420
0.539
0.524
0.514
0.515
12
Industrial Chemicals
3510
0.613
0.582
0.571
0.546
13
Pharmaceuticals
3522
0.597
0.572
0.553
0.519
14
Chemical Products nec
3528
0.658
0.615
0.629
0.622
15
Petroleum refineries
3530
0.631
0.541
0.586
0.621
16
Petroleum & Coal Products
3540
0.673
0.7
0.658
0.682
17
Rubber Products
3550
0.619
0.608
0.616
0.624
18
Plastic Products
3560
0.602
0.591
0.598
0.6
19
Pottery & China
3610
0.624
0.699
0.728
0.695
20
Glass & Products
3620
0.616
0.601
0.611
0.569
21
Non-Metallic minerals nec
3690
0.576
0.537
0.532
0.542
22
Iron & Steel
3710
0.625
0.6
0.622
0.611
23
Non-Ferrous Metals
3720
0.581
0.607
0.609
0.623
24
Metal Products
3810
0.576
0.555
0.569
0.567
25
Office & Computing Machinery
3825
0.68
0.634
0.631
0.608
26
Machinery & Equipment nec
3829
0.663
0.609
0.619
0.592
27
Communication equipment
3832
0.654
0.625
0.623
0.589
28
Electrical Apparatus nec
3839
0.668
0.64
0.655
0.645
29
Shipbuilding & Repairing
3841
0.467
0.452
0.457
0.445
30
Railroad Equipment
3842
0.639
0.618
0.559
0.591
31
Motor Vehicles
3843
0.694
0.689
0.686
0.703
32
Motorcycles & Bicycles
3844
0.642
0.689
0.64
0.671
33
Aircraft
3845
0.677
0.704
0.704
0.693
34
Transport Equipment nes
3849
0.551
0.567
0.582
0.628
35
Professional Instruments
3850
0.665
0.634
0.636
0.597
36
Other Manufacturing
3900
Unweighted Average
16
0.577
0.567
0.572
0.552
0.605
0.594
0.598
0.593