The economic geography

of Europe
Joeri Gorter*

Agglomeration economies
Agglomeration economies are positive spillover effects
between agents that locate close to each other. What comes
readily to mind are knowledge spillovers. In the course of
business activity, firms contribute to and benefit from a local


stock of knowledge that allows them to produce at lower

This paper describes trends in specialisation of regions and

costs. Other examples are reduced search efforts on pooled

concentration of industries within the European Union. It

labour markets, and the emergence of specialised markets for

complements existing empirical studies that tend to focus on

intermediate goods. The influential ‘new economic geogra-

country data. Both specialisation and concentration have

phy’ literature (Fujita et al., 1999) highlights backward and

declined, indicating that at the regional level there is a conver-

forward linkages. Firms benefit from geographical proximity

gence of economic structures and a dispersion of economic

because, in addition to having access to pooled markets for

activities of industries. This is surprising, since country

labour and final products, they save on transport costs in local

studies tend to find the opposite. A tentative explanation for

markets for intermediate goods.

cpb Report 2002/4

this empirical paradox is that European Integration allows


A salient quality of agglomeration economies is that they

specialisation and concentration forces to work at the interna-

are self-enforcing. If a firm decides to locate close to other

tional level, where they had before been confined to regions

firms it becomes more attractive for the next firm to do so as

within countries.

well. A snowball effect arises which may lead to different
production structures of a priori similar regions.
Localisation economies (or MAR economies, after


Marshall (1890), Arrow (1962) and Romer (1986)) are agglom-

Changes in the location of economic activity have always been a

eration economies that are confined to firms of the same

policy concern. The EU Member States and the European

industry. It implies that although it pays off for banks to

Commission alike acknowledge advantages, such as gains from

locate close to each other, there is no reason for textile plants

trade through exploitation of comparative advantages, or

to also be where they are. On the contrary, centrifugal forces

increased productivity through exploitation of agglomeration

such as congestion and high living expenses give an incentive

externalities. These advantages constitute the economic rationale

to locate elsewhere. Thus, localisation economies drive the

of the Single Market Program.

European economy towards a possibly large set of economi-

Changes in the location of economic activity may, however,
also carry costs. They thwart economic convergence if comparative advantages prompt lagging regions to specialise in technolog-

cally significant regions, each hosting a limited range of
Urbanisation economies (or Jacobs economies after Jacobs

ically regressive industries, or if agglomeration externalities

(1969)) extend to firms from different industries. This implies

enforce a wedge between core regions attracting economic

that it pays off for a variety of firms to locate close to each

activity, leaving behind a periphery of depressed regions with

other. Thus, urbanisation economies drive the European

limited employment opportunities. Indeed, European Cohesion

economy towards a situation in which core regions attract

Policy is geared towards mitigating the negative impact of the

business activity at the cost of the periphery.

Single Market on regional equality. In addition, specialisation
increases the probability of asymmetric shocks.
European integration gives prominence to economic geography. As Brülhart and Traeger (2002) note, many perceive that the

“scope for spatial relocation of economic activity is inversely
related to the costs of distance. As policy initiatives and technological advances have conspired over the last half century to reduce
the costs of economic transactions across region and country bor-

* For more information, contact Joeri Gorter (tel: +31-70-3383382;

ders, economic activities are thought to have become increasingly


footloose” (p.1).

b Share of gross value addeda Groningen 0. Among the empirical studies. This general belief is fed by considerable differences between the relatively fine grid of output data of manufacturing industries US and the EU. forestry.09 23 Pays de la Loire 0. grated than the EU.30 Recovery. industry growth rates explain most of the variation in industry however.10 Centre 0.09 Nord-Pas-de-Calais 0.08 a Mean over 1980-1995.Table 1 Top ten and bottom ten regions ranked according to their specialisation index Region Specialisation index a Industrya. repair. repair. been largely confined to theoretical contributions. lodging and catering services 33 Baleares 0. Haaland et al. trade. on the other hand. which row range of regions. WIFO 1999. that random differences in interest in the location of economic activity.. (1999) find a strong correlation between economic activity as industries would seek to concentrate in a nar. concentration. They tend to conclude that their European counterparts (Braunerhjelm et al.24 Services of credit and insurance institutions 17 Abruzzo 0. and fishery products 17 Madeira 0.09 Midi-Pyrenees 0.25 Agricultural. 2002. and US regions are more specialised than Middelfart-Knarvik et al. Haaland et al.36 Non market services 40 Alentejo 0.09 Austria 0. In particular.31 Agricultural. lodging and catering services 24 Luxembourg 0. b The industry with the highest difference between its share in the region’s gross value added and its mean share in the other 118 regions. forestry. trade. the US is economically more inte- (Barrios and Strobl.. 2000).industry concentration and the location of final demand. the EU may be on the eve of a catching-up process in terms of specialisation. repair. tobacco 25 Algarve 0.09 South West 0. Barrios and Strobl literature (see box) has done much to increase also academic (2002) claim. and fishery products 23 La Rioja 0. and fishery products 15 Flevoland 0.28 Agricultural. 2000). beverages. Progress has.28 Recovery.10 Scotland 0.30 Food. forestry.25 Recovery.40 Fuel and power products 43 Ceuta y Melilla 0.10 Rhone-Alpes 0. In other countries specialise and industries concentrate.09 cpb Report 2002/4 : Yorkshire and Humberside 0. 1999.. lodging and catering services 41 Açores 0. which inevitably involves relocation of There is no consensus on the explanation of the these trends. words. The formal framework of the ‘new economic geography’ points in the direction of the agglomeration externalities stressed by the new economic geography literature. trade. the majority use country data with a This paper contends that before inclining towards either .

The specialisation index The specialisation index of region i equals Table 2 Top ten and bottom ten regions ranked according to the change in their specialisation index Region Change of specialisation index a where yik denotes the share of industry k in the gross value added of region i.05 Overijssel 0.13 Algarve – 0. It indicates that specialisation is at its maximum. At one extreme. the index the EU. Adding up the absolute values of these differences indicates the degree of specialisation of this region. with takes the value zero. and the industry data are confined to manufacturing.06 value added.02 : where the weight wi denotes the share of region i ‘s gross Friuli Venezia Giulia – 0.03 Limburg (B) 0.08 in the gross value added of all other regions. A weighted average constitutes the specialisation index for cpb Report 2002/4 a set of regions: 24 Friesland 0.09 Centro – 0. At the other extreme. The trend Norte – 0. It is the period in which Single The mean value of the EU-wide index over the sample period Market Program – which constitutes the heart of the Maastricht is approximately 0. Specialisation one region can’t be found anywhere else.15. other than radioactive) made up a mere 2. The largest industry (Other market services) accounted for as much as 25.07 Northern Ireland – 0. Vlaams Brabant.07 yik – yik is the extent of over.9 percent.03 Bayern 0. To give an in 119 European regions.04 Antwerpen 0. for example.15 Groningen – 0.03 Brabant Wallon 0.07 Piemonte – 0.19 remains possible that at the regional level other dynamics prevail. in 1986 the smallest industry (Ferrous and non-ferrous ores industries corresponds to a region’s economic structure. It corresponds to the gross value added of a comprehensive set of seventeen industries1 economic structure of. The difference Ceuta y Melilla 0.04 Canarias 0. tion between 1980 and 1995.5 per mill of A commonly used specialisation index (see box) measures the degree of dissimilarity between regional economic structures. and metals. where the relative importance of any of the its gross value added.08 try level. It Alentejo – 0. where any industry present in dominance of the industry ‘Fuel and power products. services. yik and the mean of the shares of industry k Noord-Brabant 0.06 try k in region i. Rhone-Alpes – 0. one must get the stylised facts straight. and that services exhibit a different pattern.2 The relative importance of these idea. This provides the motivation for describing trends of regional specialisation and concentration with data including both manufacturing and a Difference between specialisation indices in 1995 and 1980. In the absence of theoretical guidance.07 explanation. the index takes the value This paper follows Hallet (2000) in describing regional specialisa- one. The Dutch province Groningen is most specialised.08 towards specialisation and concentration is observed at the coun- Luxembourg (B) – 0. this Treaty – was prepared and implemented. Table 1 lists the top ten and bottom ten specialised regions in seventeen industries is exactly the same in all regions. It indicates that there is no specialisation the large natural gas deposits at Slochteren accounting for the whatsoever.’ The . The panel comprises the could be considered as a normal value.or under representation of indus- Limburg (NL) 0.

trade. 1999. Figure 1 suggests that European regions have become more simi- Table 2 lists the top and bottom ten regions ranked according to lar in economic structure between 1980 and 1995. It makes clear that what is true at the index. It also demon- Figure 1 Regional specialisation in Europe strates that predictions based on rough and ready comparisons between the EU and the US are hazardous. and fishery products’ playing a dominant role in the where xik denotes share of region i in the gross value added of sparsely populated. in spite and 1995 specialisation gradually declines. . It shows that between 1980 national level is not necessarily true at the regional level. since its dynamics are often characterised by path dependencies. production of industry k. gross value added.155 cal accidents. while 85 moved away from it. or ures whether some regions tend to be over represented in the for ‘Services of credit and insurance institutions’ in Luxembourg. Figure 1 depicts its time series. Middelfart- 25 Is there a trend towards specialisation? The regional data show a wide variety of changes in economic structure: 34 regions Summing up the indices of the individual regions. forestry. for ‘Recovery.Spanish government’s involvement in its Moroccan enclaves The concentration index Ceuta y Melilla drives the prominence of the industry ‘Non mar- The concentration index of industry k is ket services’ there.160 economic geography in particular.. and despecialisation . WIFO 1999. It is comforting to note that the specialisation index yields further results that are by and large consistent with casual observation.14]. the degree to 1982 1984 1986 1988 S 1990 1992 1994 which industries are over-represented in some regions. where Si measures whether some vineyards and bodegas. and tobacco’ in La Rioja. xik the mean of the shares of this region for all ‘food. rural region Alentejo. giving prominence to histori- . then industries must necessarily concen- . It is hardly surprising to see ‘Agricultural. moved towards more specialisation.. the results do where the weight wk denotes the share of industry k in total performance. They note that is in line with the view that . Note. repair. The bottom ten specialised regions are in terms of economic structure closest to the European average. (2001) that stylised facts in economic geography are often of a fractal nature. Thus. If regions specialise in a limited number of industries.130 1980 The mirror image of specialisation is concentration. lodging and industries tend to be over represented in region i. but clashes with studies at the cpb Report 2002/4 not suggest a relation between specialisation and economic Member State level (Haaland et al. and under-represented in others. a region covered by other industries. that A weighted average constitutes the concentration index for a set of industries: among them are depressed ones such as the Nord-Pas-de-Calais as well as high performers such as Scotland. of the observation of Brakman et al. however.135 Concentration . This holds true for . yields the EU-wide Knarvik et al.145 “European integration opens scope for between-country specialisation which hitherto had existed only at the within-country level” . with the findings of Hallet (2000). weighted by the relative economic size of the regions. It is consistent the change in specialisation index. 2000). Ck meas- catering services’ in tourist destinations Algarve and Baleares. Thus.150 at the regional level. The same holds true for industry k. beverages. Brülhart and Traeger (2002) tentatively explain the paradox of specialisation at the country level.140 [p.

08 South East 5 Metal products.12 Île de France 7 services 0.145 cates that concentration is at its maximum. provided that size differ- Figure 2 Industry concentration in Europe ences between regions or industries are not too large.20 Bayern 8 Paper and printing products 0.21 Nordrhein-Westfalen 9 Non-metallic minerals and mineral products 0. where any region’s share in an industry’s total gross value added is exactly the same as this region’s share in all other industries’ total gross value added.140 1980 1982 1984 1986 1988 1990 1992 1994 cpb Report 2002/4 C 26 Table 3 Industries ranked according to their concentration index Industry Concentration indexa Regionab Share of gross value addeda Ferrous and non-ferrous ores and metals. and fishery products 0. . where any region with a positive share in one industry has a zero share in others. other than radioactive 0.17 Île de France 9 Transport and communication services 0. machinery.17 Baden-Württemberg 7 Services of credit and insurance institutions 0. beverages and tobacco 0. equal to 0. for exam- . repair. forestry. equipment. trade. At the other extreme. .165 extreme.18 Baden-Württemberg 8 and plastic 0.23 Nordrhein-Westfalen 10 Fuel and power products 0.11 Île de France 5 Building and construction 0. leather and footwear 0. At one .16 South East 8 Food. The mean value of the EU-wide concentration index is. It corresponds.15.160 .155 It indicates that there is no concentration whatsoever.34 Lombardia 9 Agricultural. the index takes the value one. b The region with the highest difference between its share in the industry’s gross value added and its mean share in the other 17 industries.33 Greece 6 Transport equipment 0. A concentration index for industries can be calculated along the same lines as the specialisation index for regions.19 South East 8 0.14 Sweden 6 Other market services 0.16 Niedersachsen 3 Non market services 0.trate on a limited number of regions. lodging and catering a Mean over 1980-1995.35 Nordrhein-Westfalen 13 Textiles and clothing. the index takes the value zero.170 . and electrical goods Products of various industries including rubber Recovery.23 Baden-Württemberg 8 Chemical products 0.150 . It indi- . just as the specialisation index.

machinery.furnaces of the industry Does the trend in concentration confirm the downward trend in specialisation? Table 4 lists the industries ranked accord- ‘Ferrous. equipment and electrical goods – 0.01 Food. it produces less than one per mill in Ceuta and footwear’ concentrates on the Italian region Lombardia. other than radio active’ ing to the change in concentration index.00 Transport and communication services – 0. other than radioactive – 0.00 Recovery. the ing away from it. contributes most to the high concentration index. This more or less equal distribution suggests new economic geography literature (Fujita et al.03 Transport equipment 0.01 Products of various industries. the industry on average Westfalen’s mean share in all other industries is the highest of all generates 100 / 119 = 8 per mill of its output in a single region. leather and footwear 0. 27 ple. beverages. The industry’s industries moved towards more concentration against eight mov- first rank should therefore not come as a surprise. forestry. repair. it is the importance of fixed costs for clustering of economic activity.01 Non market services – 0. other Size differences between industries or regions may.02 Fuel and power products – 0. Table 3 lists the seventeen industries ranked according to their concentration index.03 a Difference between concentration indices in 1995 and 1980. not pronounced. This is Melilla.08 Services of credit and insurance institutions 0. At 119 regions. lodging and catering services 0. Figure 2 displays the EU-wide concentration index. The blast. including rubber and plastic – 0. Many associate ‘Ferrous and non-ferrous ores and metals.01 Non-metallic minerals and mineral products 0. It shows that nine imply a production process with high fixed costs. beverages. The industry ‘textiles and clothing. After all.01 Ferrous and non-ferrous ores and metals.00 Other market services 0. also consistent with casual observation. tobacco 0. and non-ferrous ores and metals. 1999) stresses that if there is a downward trend in EU wide concentration. Since there are 119 regions. Indeed. and the difference between this share and Nordrhein- industry. than radioactive’ with the German Ruhrgebiet.03 Paper and printing products 0. tip the balance in favour of either concentration or deconcentra- Nordrhein-Westfalen – the region that hosts this industrial area – tion. No less than thirteen percent of the industry’s gross value added is generated cpb Report 2002/4 Change of concentration indexa Industry ..01 Metal products.Table 4 Industries ranked according to the change in their concentration index Textiles and clothing. and fishery products 0. to the concentration of the ‘Food.00 Agricultural. leather and the extremes. trade.00 Building and construction 0. and six percent in Nordrhein-Westfalen. however.01 Chemical products 0. and tobacco’ there.

They tend to report that countries become more tions force usage of larger NUTS 1 regions for some Member States.M. and E. 1986. Strobl. 2001. Specialisation and (geographic) concentration of European manufacturing. 29. Kind. Braunerhjelm. The spatial economy: cities.J. This is System of Integrated Economic Accounts (ESA). 1890. Data limita- country studies. CEPR discussion paper No. just as regions have become more specialised.. 1969. Norman. have become more concentrated. FEDEA working document. 2000. and A. New York: Vintage. A tentative account of the paradox is that European integration allows specialisation and concentration forces to work between countries. K.J. Midelfart-Knarvik. This empirical paradox implies that the stylised facts that are to be explained are more complex than a one-dimensional trend towards either specialisation and concentration. 1999. Thus.. and J. although the trend towards con- 94. which obeys the European specialised and industries become less concentrated.. What determines the economic geography of Europe?. Torstensson. Review of Economic Studies.J. An account of geographic concentration patterns in Europe. Brakman. 2000. S. Cambridge University Press: Cambridge. 141. 28 References Arrow. The . Increasing returns and long run growth. Regional specialisation and concentration in the EU. 1999. P. van Marrewijk. H. K. and A. and Financial Affairs of the European Commission. Journal of Political Economy. where they had before been confined to regions within countries. Ruane. Venables. Fujita. J... of Brülhart and Traeger (2002). Traeger. 155-73. S. regions. Hallet. J.: MIT Press.. Krugman. or the reverse.G.. cpb Report 2002/4 specialised and that industries become more concentrated. CEPR: London. Jacobs. London: Macmillan Press. F. S. Venables. 1002-37. 2002. Brussels: European Commission.The index shows a more or less constant degree of concentration location of European industry. Report prepared for the Directorate General for Economic during the 1980s. 2000.. P.. Economic Papers.. Directorate General of the European Commission.J. and a sharp drop at the beginning of the 1990s. H. M. H. The economy of cities. Seabright. Cambridge. The economic implications of learning by doing. Marshall. Redding. Garretsen and C. 2002. 1999. industries Romer. Overman. V. K. Principles of economics. centration is less pronounced. H. P. and international trade. M. Brülhart. and R.J. R. Faini.I. Midelfart-Knarvik. This is consistent with the findings WIFO. 2072. Conclusion Notes Agglomeration economies appear to push the economic geography of Europe towards a situation in which regions become less 1 The industries are classified according to NACE-CLIO RR17. An introduction to geographical economics. 1962. and P. Haaland.H. Industry mobility and concentration in the European Union. Integration and the regions of Europe: how the right policies can prevent polarization. contrary to what one might expect on the basis of the results of 2 Most regions are NUTS 2 (Nomenclature of Statistical Territorial Units). M. Barrios. but again points in another direc- background paper for “the competitiveness of European industry” for the Enterprise tion as the country studies. Mimeo. MA. A.