EU integration and the
geographies of economic
activity: 1985–2019
Eleonora Cutrini
Department of Economics and Law, University of Macerata, Macerata, Italy
Ben Gardiner
Cambridge Econometrics, Cambridge, UK
Ron Martin
Department of Geography, University of Cambridge, Cambridge, UK
Abstract
One of the central predictions of the new economic geography (NEG) is that the removal of trade
barriers and other such frictions should lead to the geographical concentration and specialisation
of economic activity, both between and within nations. This prediction has been used to argue that
as the European Union becomes more integrated, economic activity would become more region-
ally concentrated and specialised. Using relative entropy measures applied to a new regional data
set for the period of 1985–2019, this paper finds that between 1985 and 2000 localisation and
specialisation between European countries increased to some extent but with a widespread fall
in specialisation and concentration within countries. After 2000, and particularly after 2007 and
the Global financial Crisis, the spatial distribution of economic activities in Europe appears to
have become more complex, with some degree of concentration, agglomeration and specialisation
both across and within countries. Overall, however, we find little long-run support for the NEG
prediction.
Keywords
Geographical concentration, geographical specialisation, relative entropy measures, European
Union
Introduction
A key focus of the so-called New Economic Geography (NEG) that has developed over the past
three decades has been to devise models of the spatial distribution of economic activity by
Corresponding Author:
Ron Martin, Department of Geography, University of Cambridge, Downing Place, Cambridge, CB2 3EN, UK.
Email: rlm1@cam.ac.uk
Original Article A Economy and Space
EPA: Economy and Space
2023, Vol. 55(2) 274–302
© The Author(s) 2022
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0308518X221127028
journals.sagepub.com/home/epn
combining ideas from ‘New Trade Theory’, Marshallian location theory, and the New Urban
Economics (e.g., Baldwin et al., 2003; Brakman et al., 2009; Fujita et al., 1999; Fujita and
Thisse, 2002; Krugman, 1991). One issue such models have been used to explore is what
happens to the geographical distribution of economic activity as the level and nature of economic
integration increases. Another is whether agglomeration increases national economic growth (see
Gardiner et al., 2011).
Economic integration between countries can be viewed as lowering the transaction costs
between the countries and the regions within them. Transaction costs can take the form of transpor-
tation costs, tariffs, differences in regulatory arrangements, quality standards, restrictions on the
movement of labour and capital, exchange rate controls, or other such barriers and frictions.
According to NEG theory, what happens to the spatial distribution of the industry if there is a
move towards economic integration of the countries in question depends on three key issues: the
extent and pattern of cost differences in the industry between regions across the countries, the pre-
existing pattern of regional specialisation in the industry, and the extent to which the industry in
question is subject to location-specific external economies.
NEG models typically assume a prior geographical distribution to be an equilibrium one. Now
assume that the process of economic integration lowers the transaction costs between countries and
hence between regions across those countries. If there were no cost differences amongst regions for
a given industry in question, then a geographically dispersed pattern of that industry will be a stable
(assumed to be ‘equilibrium’) state. If some regions enjoy cost or related advantages, then the
reduction of transaction costs, such as removal of tariffs, will encourage a shift in the geography
of the industry towards those regions with a pre-existing advantage and away from those with a
disadvantage. So, according to this theory, increasing economic integration should ordinarily
lead to a divergence in economic structure (i.e., increased specialisation) between regions in
terms of their industrial structure, and to an increasing geographical concentration of the industry
in particular regions, which then becomes a new stable (‘equilibrium’) pattern.
Krugman (1993) famously used this line of argument to suggest that as the countries belonging
to the European Union become progressively integrated economically, so – other things being equal
– they should become increasingly specialised in particular industries, with those industries concen-
trated in those regions in which the external economies benefits of localisation were greatest. His
argument was based not only on theoretical reasoning, but also on appeal to the United States. The
fact that the United States had a higher degree of internal spatial economic differentiation and
regional specialisation than the European Union at that time indicated, in his view, what to
expect in the European Union as it moved closer towards the sort of unified integrated economy
exemplified by the United States, and especially as the EU moved towards monetary union.
Since Krugman’s suggestive contribution, various NEG-type studies have sought, both theoret-
ically and empirically, to explore the implication of increasing international integration for the
internal economic geography of countries, particularly in the case of Europe. Some have argued
that when transaction costs fall below a certain threshold, international openness should lead to
the agglomeration of industries within countries (e.g., Crozet and Koenig-Soubeyran, 2004;
Monfort and Nicolini, 2000; Monfort and van Ypersele, 2003). Other, competing models,
however, based on different assumptions on labour mobility and congestion-related costs, have sug-
gested that trade liberalisation would favour the internal dispersion of economic activities, thus
reducing the level of spatial concentration (Alonso-Villar, 2001; Behrens et al., 2007; Krugman
and Livas, 1996; Puga, 1999). More recent work suggests that predictions need to take into
account other trends in the spatial organisation of economic activity that are both an outcome
and a source of increasing integration, particularly the rise of complex, geographically dispersed
production networks and supply chains (see Baldwin, 2016). Such spatial webs or networks of
supply and functional specialisation are now an important feature of the European economy.
Cutrini et al. 275
One of the most recent, and detailed, studies of regional trends in specialisation and concentra-
tion of economic activity across the European Union is that by Cutrini (2010) who used a new
version of the Theil Index to examine trends in both specialisation and concentration of manufac-
turing employment across some 145 NUTS2 regions for 10 European countries at three points in
time, 1985, 1993 and 2001. Her analyses suggested that a process of regional de-agglomeration
and de-specialisation of manufacturing activity had taken place over this period, a finding in con-
trast to the cross-national findings found in other studies, and not consistent with the basic economic
integration thesis. That analysis is now somewhat out of date. One of the novel contributions of this
paper is that it conducts an analysis of the changing geography of economic activity over the longer
period of 1985–2019, a period of nearly four decades. This not only gives a more extensive time
frame over which the process of European economic integration has been evolving, and hence
better to detect any trends and changes in regional specialisation and concentration that may
have occurred over this time, the extended period also includes some major events, most especially
the Global Financial Crisis of 2007–2008 and the deep recessionary shock this generated, which
may have disrupted both the integration process and the geographical organisation of economic
activity. The Financial Crisis and associated Great Recession, represented the worst economic
downturn and credit squeeze since the Great Depression of the early-1930s, and sent a shock
wave of historic proportions across European economies. A second novel feature of the paper,
therefore, is that our longer span of data allows an analysis of the possible impact of that Crisis
on the geographical distribution of economic activity. A third novel feature is that for the period
of 2003–2019, our analysis includes a range of private sector service activities in addition to manu-
facturing. The inclusion of services is important for a number of reasons. Not only is the European
Union economy now dominated by services of various kinds, the locational dynamics of services,
especially in response to increasing economic integration, may be different from those of manufac-
turing activities, and thus services deserve their own analysis. A fourth advance of this paper over
Cutrini’s earlier analysis is that the number of countries analysed increased from 10 to 13, and from
2003 the number of regions from 133 to 215.
Our main findings can be summarised as follows. Briefly, as regards the period of 1985–2007, our
analysis based on employment data shows that the overall convergence in regional economic struc-
tures and the spatial spreading of industries found up to 2001 in Cutrini’s earlier work, continued up
until 2007, just prior to the global financial crisis. The same analysis, when replicated for the more
recent period, 2003–2019, suggests that in the post-global financial crisis period, 2007–2019 the
process of overall spatial dispersion not only moderated, but that there is some evidence of an increase
in the agglomeration (or co-location) of economic activities accompanied by the geographical concen-
tration of individual manufacturing industries and services. For knowledge-intensive services,
however, the trend is towards greater spatial dispersion and co-location with the rest of the economic
activities. Interestingly, we found that concentration and agglomeration do not coincide with special-
isation. Summarising, therefore, our detailed empirical work suggests an overall trend towards less
regional specialisation and geographical concentration as the countries and regions of the EU have
become increasingly integrated – the reverse of standard NEG theory – and more in line with what
Krugman (2009) himself has more recently has argued is happening. Of course, the impact of the
COVID-19 pandemic and the war in Ukraine may well have additional and complex implications
for the geographical distribution of economic activities across the EU, and suggested that further
research on this issue will continue to be needed in the future.
Previous empirical research: a brief survey
Much of the empirical work on the spatial agglomeration of economic activities and specialisation
patterns in the European Union refers to the 1980s and 1990s. In the first wave of empirical studies,
276 EPA: Economy and Space 55(2)
specialisation and concentration in Europe were analysed mainly using countries as the basic unit of
analysis. As for specialisation, earlier studies employed export data to assess country specialisation
patterns (see, among others, e.g., Aquino, 1978; Sapir, 1996), while other studies analysed country
specialisation using production data, such as employment, value-added, or output (see Combes and
Overman, 2004 for a comprehensive survey). While Amiti (1999) reports a general increase in rela-
tive specialisation in EU countries from 1968 to 1990, Midelfart-Knarvik et al. (2004) tend to
suggest that the process of relative specialisation is more complex at least from the early-1980s
until the end of the 1990s. Recent studies, focussed on EU-27 countries, suggest increasing special-
isation of the new member states (see e.g., Bagoulla and Péridy, 2011 and Vechiu and Makhlouf,
2014). Specialisation patterns at the sub-national level have been less analysed empirically. Results
from the few studies available indicate that there is no evidence of increasing regional specialisation
over the pre-EMU period (e.g., Cutrini, 2010; Paluzie et al., 2001).
A number of scholars find a pattern of increasing geographical concentration of manufacturing
across European countries during the 1980s (Amiti, 1999; Brülhart, 1998; Midelfart-Knarvik et al.,
2004), followed by a process of dispersion (deconcentration) during the following decade (Aiginger
and Pfaffermayr, 2004; Midelfart-Knarvik et al., 2004). Using production data for Belgium, France,
Italy, Germany and the United Kingdom, Amiti (1999) reported a significant increase in relative
concentration for a majority of manufacturing industries between 1976 to 1989. Moreover,
Brülhart (1998) found out that, during the 1980s, localisation increased in 14 of the 18 European
industries considered, especially labour-intensive industries and increasing returns to scale indus-
tries. In contrast, analysing the period 1970–1997; Midelfart-Knarvik et al., 2004 suggested that
the upward trend of the 1980s was a brief interlude before returning to decreasing localisation
across European countries. Focusing on the spatial concentration of manufacturing value added
between 1985 and 1998, Aiginger and Pfaffermayr (2004) suggested that an upward trend occurred
in the pre-Single Market period while dispersion dominated afterwards, and the changes are signifi-
cant for each period.
More recently, the growing availability of disaggregated data has allowed analysis of the con-
centration of manufacturing industries across EU subnational regions (Aiginger and Leitner,
2002; Brülhart and Traeger, 2005; Cutrini, 2010; Ezcurra et al., 2006; Hallet, 2000). However,
the findings from these studies, all covering the period before European Monetary Union, have
varied. Some have found a declining regional concentration of manufacturing employment
across NUTS1 regions (Aiginger and Leitner, 2002), while others find evidence of an increasing
concentration of value added during the 1990s (Aiginger and Davies, 2004; Aiginger and
Pfaffermayr, 2004). On the basis of regional gross value added, Hallet (2000) found inconclusive
results, while Brülhart and Traeger (2005) found mixed evidence for the interregional concentration
of value added in manufacturing industries. While Ezcurra et al. (2006) find empirical support for
the “Krugman hypothesis” with regional data, other studies report decreasing localisation across
Spanish provinces during the 1980s (Paluzie et al., 2001), across Italian regions from the
early-1970s to the late-1990s (De Dominicis et al., 2013; De Robertis, 2001; Rombaldoni and
Zazzaro, 1997) and within Germany after 2000 (Dauth et al., 2016). Overall, while results from
the early regional studies are in favour of a downward trend in specialisation and dispersion of
manufacturing activities before the implementation of the single currency, polarisation may have
emerged in the post-EMU period before the global financial crisis, at least for some industries.
For example, Schweizog and Collins (2015), focussing on four key manufacturing industries,
provide some evidence of an increase in geographical concentration between 1999 and 2007.
Empirical research on industry location and agglomeration in Europe has focussed nearly exclu-
sively on manufacturing. Since services are an important part of the European economy, it is
important to consider them in any recent analysis of the EU geography of economic activities.
Recent studies, including services in the analysis, indicate heterogeneous evolutions across
Cutrini et al. 277
sectors in Italy and dispersion of industry and services in Germany (e.g., De Dominicis et al., 2013
for Italy, and Dauth et al., 2016 for Germany). Moreover, given the European institutional architec-
ture, it is essential to characterise the spatial distribution of economic activities both across and
within countries.
Geographical specialisation and concentration
Four major dimensions, or aspects, of the distribution of economic activity across geographic space,
can be identified from the vast literature that now exits on the subject: regional specialisation,
regional concentration, regional agglomeration and local clustering. Although both concentration
and agglomeration deal with the location of economic activity, their focus is somewhat different.
According to Brülhart, typically concentration analyses the location across space of an individual
sector of activity, whereas agglomeration analyses the co-location across space of several sectors,
for example, several different manufacturing industries, or several different types of services, or the
co-location of manufacturing industries and services. In the view of Brakman et al. (2009), studies
of agglomeration show how far a number of industries are co-distributed across geographic space;
whereas, the analysis of concentration does the same only for a particular type of economic activity,
says the production of motor vehicles.
Concentration and agglomeration can in turn be distinguished from specialisation (see Brakman
et al., 2009; Gorter, 2002; Hallet, 2000), even though they may be related in certain circumstances.
At this point, it is important to draw a distinction between relative specialisation and absolute spe-
cialisation, in parallel with the distinction between absolute and relative concentration that is well-
established in the literature (e.g., Bickenbach and Bode, 2008; Brülhart and Traeger, 2005).
Absolute specialisation deals with the question of whether or not a location’s (nation, region,
city, etc.) economic structure (in terms of shares of employment or output) is dominated by one
or just a few related sectors, for example, motor vehicles or textiles. It is usually measured by com-
paring the location’s economic structure (composed by N industries) against the uniform distribu-
tion, that is an economic structure with equally sized industries. Relative specialisation deals with
the question of whether and to what extent a location’s economic structure is dissimilar compared to
a relevant benchmark, for example, in the case of, say, a member state of the European Union, with
the aggregate EU’s economic structure, or in the case of a region or city, with the economic struc-
ture of the nation of which the region or city is a part.
The issue of specialisation has long been a key notion in trade theory, where it is argued that
nations, cities and regions will tend to specialise in those activities in which they have a compara-
tive advantage. Thus, the degree of relative specialisation in a city or region is typically taken to be
an indication of that city’s or region’s ‘revealed’ comparative advantage in the sector(s) concerned.
In certain circumstances, however, specialisation and concentration can be the two sides of the same
coin (Aiginger and Davies, 2004; Cutrini, 2009a), since a very high level of concentration of an
industry in a particular region may well mean that the region is in fact specialised in that sector,
in as much that it dominates the region’s employment structure.
Thus, as Brakman et al. (2009) point out, concentration, specialisation and agglomeration may or
may not coincide, depending on contingent circumstances. Further, when considering the regional
effects of the economic integration within the European Union, it is possible to have specialisation
between nations (that make up the EU), but not necessarily between regions within those nations.
Also, it is possible to have concentration of an industry between nations, but no industry concen-
tration within nations. Some examples are illustrated stylistically in Figure 1, in which there are two
industries and two countries each with four regions.
For simplicity, assume a pre-integration pattern in which both industries are equally distributed
across the four regions in each country (Figure 1(a)).1 Following integration of the two countries,
278 EPA: Economy and Space 55(2)
Figure 1. Some stylised geographical outcomes of economic integration: two countries, each with four
regions, two industries.
Cutrini et al. 279
and assuming this reduces transport costs, removes barriers to trade and factor movements, and so
on, then a variety of outcomes are possible, depending on the specifics of the industries, regions and
countries. In one case (Figure 1(b)), integration leads to a shift in both industries such that industry 1
becomes concentrated in country A, which is thus specialised in that industry, while industry 2
shifts to Country B, which thus becomes specialised in that industry. But note there need be no
regional concentration or specialisation in either country. In fact, this case corresponds to a scenario
of complete localisation between countries (Lb ) with perfect regularity (complete lack of localisa-
tion), within countries (Lw), that is Lb = L, Lw = 0. National comparative advantages drive these
location patterns and there is no role for short-distance agglomeration factors such as the advantages
of localisation economies within industries.
Other possible outcomes are illustrated in Figure 1(c–e). In Figure 1(c), country A becomes
specialised in industry 1, which is also concentrated in one of its regions (which thereby spe-
cialises in that industry), while country B becomes specialised in industry 2, which is also con-
centrated in one of its regions. This is a case in which localisation is particularly high, due to
both cross-country divergence and also within-country specialisation associated with a spatial
concentration of industries in different regions. In Figure 1(d) there is a concentration of both
industries 1 and 2 in country 1, and an agglomeration of the two industries in one particular
region of that country. Both countries have no relative specialisation since their economic
structures are the same as the benchmark (the sum of the two) (Lb = 0, L = Lw). As for the loca-
tion patterns within countries, regions are more specialisation in country 2 than in country 1. In
Figure 1(e), both industry 1 and industry 2 are concentrated in country A, there is regional spe-
cialisation (of both industries in country A, and of industry 1 in country B), and regional
agglomeration in country A.
Notice that with the agglomeration of the two industries in one region in country A (as in
Figure 1(d and e)), overall localisation – a weighted sum of relative concentrations of each industry
in the methodology of the present analysis – will be lower compared to the case in which the indus-
tries are mostly localised in different regions as in the case illustrated in Figure 2(c).
Figure 2. The growth of EU trade, 1970–2018. Source of data: Cambridge Econometrics, based on Eurostat
Data.
280 EPA: Economy and Space 55(2)
These examples merely illustrate the complex range of outcomes that are possible and which
might follow integration, and the different geographical levels at which such outcomes can be ana-
lysed. They also suggest that decomposability would be a desirable feature of any measure used to
assess the geographical distribution of economic activity in an integrated multi-country, multi-
region system such as the European Union.
Data and methodology
To evaluate agglomeration, specialisation and concentration and their changes over time, we used
employment as the main variable of interest (output data, such as Gross Value Added, are less
detailed and more restricted in coverage; further, output can change simply by changes in
product prices, which may have nothing to do with employment). We collected employment
data from two sources available in the Eurostat Region-SBS database to obtain an adequately
long period of time. An immediate challenge is that relevant data are available in discontinuous
time series due to different sectoral classifications for different time periods, from NACE Rev
1.1 to NACE Rev 2. To first extend the initial analysis based on the NACE Rev 1.1 classification
to 2007 some data manipulations were required in order to conduct the analysis with the same
sample of regions and thus avoid inconsistencies that would lead to possible aggregation bias.
To this end, some regions were dropped and others were aggregated so to build a final and consist-
ent set of 133 NUTS1-NUTS2 regions belonging to EU-15. For the following period 2003–2019
employment data are based on the NACE Rev 2 classification, and collected for 215
NUTS1-NUTS2 regions belonging to EU-28. The Data Appendix contains a detailed description
of data sources.
We acknowledge that, ideally, agglomeration and concentration should be assessed in continu-
ous space to avoid the checkerboard problem related to arbitrary geographical boundaries, and the
inherent aggregation and scale problems (See Arbia, 1989). However, distance-based measures
require micro-geographical data, which are not available. In any case, geographical boundaries
are expressions of economic institutions responsible for planning and economic policies. Thus, it
may be important to know the spatial scale at which agglomeration, concentration and specialisa-
tion occur to inform policy makers. To this end, some authors propose the use of relative entropy
indices (Bickenbach and Bode, 2008; Brülhart and Traeger, 2005; Cutrini, 2009a, 2010), which
have distinct advantages over the standard measures constructed over the discrete space. The
most relevant is decomposability. This feature permits the decomposition of localisation across dif-
ferent spatial and sectoral scales in order to identify the contributions of individual regions (sectors)
to the overall localisation of economic activity.
Entropy-based approaches are appropriate since they allow for the nested geographical perspec-
tive (region-country) typical of the European economy and are incorporated in New Economic
Geography models. For this reason, relative entropy measures are becoming frequently used in
such analyses (e.g., Bagoulla and Péridy, 2011; Gokan, 2010; Stierle-von Schultz and Stierle,
2013; Vechiu and Makhlouf, 2014; among many others). Further, by bootstrapping the measures,
and their components, we are able to assess whether geographical concentration, specialisation and
localisation changed significantly over time. The resampling process is repeated 10,000 times.
Given the bootstrap estimate of the sampling distribution, it is possible to derive standard errors,
compute confidence intervals, and conduct the following hypothesis testing:
H0: ΔTk= 0
H1: ΔTk ≠ 0
Where ΔTk is the change in the relevant Theil entropy index.
In a period of deindustrialization, when employment and value-added formation has shifted from
manufacturing to service sectors, it is important to evaluate how the spatial distribution of sectors
Cutrini et al. 281
and the economic structures of countries and regions have evolved. For these reasons, we use both
absolute and relative measures to analyse the evolution of concentration, specialisation and agglom-
eration in Europe during the past three decades characterised by deepening European integration,
especially following the implementation of the monetary union. In the Methodology Appendix, we
have included a detailed description of the set of complementary indicators we have used.
Results on concentration, agglomeration and specialisation: from an
old to a new pattern
Recent trends in EU integration
EU Single Market Integration can be measured in various ways, typically centred around the four
fundamental freedoms of movement which underpin it: goods, people, services and capital. Since
much of NEG theory on the effects of economic integration on the geographical distribution of
industry emphasises the increased trade that integration is alleged to promote, it is useful, by
way of context to look briefly at trends in trade within the EU.
Figure 2 plots EU trade, as a proportion of total GDP, since 1970. In terms of overall importance,
and with the exception of the Great Recession of 2008–2010, trade has increased steadily as a pro-
portion of GDP since the early-1980s. A marked acceleration can be seen in the 1990s and 2000s
due to trends in both globalisation and increasing EU integration, with many new members joining
the EU in this period, including the Mediterranean enlargement of Spain and Portugal in the
mid-1980s, Austria, Finland and Sweden in mid-1990s, and many countries from the former
eastern European block countries from 2004 onwards; the EU12 of the mid-1980s had become
the EU28 by the end of the period in Figure 2. By this time, the process of monetary union that
began in 1990, and accelerated after 1999, had brought 19 EU member states into a common cur-
rency area (Eurozone), further integrating their economies.
What is also evident from Figure 2 is that at least up to the Great Recession of 2008–2010, trig-
gered by Global Financial Crisis of 2007–2008, intra-EU trade grew slightly faster than extra-EU
trade (as a share of GDP), so that by 2007, on the eve of the financial crisis, intra-EU trade
accounted for 63% of total EU trade. It then more or less stabilised thereafter. As measured by
trade, the economies of the European Union had become significantly more integrated over our
study period of 1985–2019.
Overall trends in economic localisation
From a systematic analysis of location patterns, we identified major long-run trends in localisation,
concentration and specialisation in the European Union. In a nutshell, as regards the period 1985–
2007 we basically confirmed the previous evidence that pointed to an overall convergence in
regional economic structures and a spatial spreading of industries. The same analysis, when repli-
cated for the more recent period, 2003–2019, suggests that moderation of the process of spatial dis-
persion and of convergence occurred, starting from the beginning of the new millennium.
In fact, over the 2003–2019 period, there is some evidence to suggest that economic activities
have become slightly more geographically concentrated. Figure 3 plots the within-country and
between-country contribution of overall localisation in the entire period of our analysis. Overall
localisation of manufacturing and its component have increased marginally from 2003 onwards,
perhaps starting to delineate a ‘U-shaped’ pattern whereby individual industries are diverging
from the spatial distribution of total manufacturing employment.
This first evidence is further investigated with the next stage of analysis, which focuses on
changes in patterns of geographical concentration and agglomeration, where we complement
282 EPA: Economy and Space 55(2)
relative entropy measures with absolute measures of concentration for a more comprehensive ana-
lysis of both manufacturing industries and services.
We found that low-tech industries are diverging from manufacturing as a whole over the period
2003–2019 and in some cases – Food and beverages, Rubber and plastic products, and Basic metals
– the changes are significant (Table 1, column 2). This result is robust to the change in the bench-
mark distribution (manufacturing employment or total employment in manufacturing and services)2
(Table 1, column 3). In other words, the same industries are also deviating from the spatial distri-
bution of total employment in manufacturing and services. Interestingly, these manufacturing
industries are also clustering in space as the results on absolute concentration suggest (Table 1,
column 1). Moreover, a significant geographical concentration of industries arises more often
across countries rather than within countries as the results of the testing procedure performed by
spatial scales suggest (Tables A3 and A4).
Service sectors are becoming more geographically concentrated too, except for scientific
research and development which is becoming more evenly distributed across Europe. The spatial
concentration of knowledge-intensive services has decreased significantly resulting in increasing
co-location with the rest of economic activities, that is, relative to total employment. The remaining
services –wholesale and retail trade, accommodation and food services, transport – are clustering in
some regions while converging to the spatial distribution of total employment (Table 1, columns 2
and 4). Finally, it is worth noting that both our benchmark distributions in the present study (manu-
facturing and total employment), are also becoming slightly more geographically concentrated, as
reported at the bottom of Table 1.
Figure 3. Evolution of the entropy index of overall localisation (L-index) and its components, 1985–2007;
2003–2019: manufacturing employment. Notes. The lines in the graph refer to the entropy index of overall
localisation (L-index), and the respective components (Lb, L-index between country; Lw, L-index within
country). The plain line refers to the analysis over the period 1985–2007 and is based on the sample of 133
regions and the NACE Rev 1.1 classification of sectors. The dashed lines refer to the analysis over the period
2003–2019 and are based on the sample of 215 regions and the NACE Rev 1.1 classification of sectors.
Cutrini et al. 283
Taken together, the results of Table 1 suggest that industries and services are clustering in
different places/regions. In fact, in this period, urban areas or capital regions have increasingly
attracted services while manufacturing employment has remained or clustered in extra-urban
regions. Based on this evidence, we suggest that, after 2000, the actual spatial distribution of
industries and services exhibits some degree of concentration, agglomeration, in some cases
combined, with specialisation thus suggesting that the European economic geography has
been changing towards a more complex pattern that resembles panels (d) and (e) of Figure 1,
depending on the country.
Table 2 reports the changes in relative specialisation, and its components, with their signifi-
cant level. Despite the general clustering of industries and services accompanied by their
agglomeration for the period 2003–2019 (Table 1), our results do not provide the support
that, as EU integration has progressed, all Members States and their regions are becoming
more economically specialised in recent years. Our analysis of recent data confirms a signifi-
cantly increasing regional specialisation only in Greece, France and within some other
Member States – such as Belgium and Germany while the increase in cross-country specialisa-
tion is positive and significant only in Sweden and Poland.
Hence, we find that in the European Union, the evolution of concentration and agglomeration
does not coincide with specialisation trends after the introduction of the Euro. Our results
confirm that concentration and agglomeration should be distinguished from specialisation, as sug-
gested by Brakman et al. (2009). Moreover, our results should also be interpreted in light of the
Table 1. Evolution of spatial concentration across 215 EU regions, 2003–2019.
Absolute
concentration
Relative concentration
Benchmarch:
manufacturing
Benchmarch:
manufacturing
and services
Obs.
diff. Sig.
Obs.
diff. Sig.
Obs.
diff. Sig.
Food products and beverages 0.072 ∗∗ 0.011 ∗∗ 0.008
Wood 0.002 0.057 ∗∗∗ 0.108 ∗∗∗
Paper 0.046 −0.022 −0.006
Rubber and plastic products 0.059 ∗ 0.029 ∗∗ 0.070 ∗∗∗
Manufacture of other nonmetallic mineral products 0.035 −0.019 ∗ −0.013
Manufacture of basic metals 0.130 ∗∗∗ 0.063 ∗∗∗ 0.042
Wholesale and retail trade and repair of motor vehicles
and motorcycles
0.009 0.011 ∗
Wholesale trade, except for motor vehicles and
motorcycles
0.016 −0.040
Retail trade, except for motor vehicles and motorcycles 0.047 ∗∗ −0.003
Accommodation and food service activities 0.082 −0.006
Land transport and transport via pipelines 0.021 −0.006
Scientific research and development −0.106 ∗∗ −0.166 ∗∗∗
Manufacturing as a whole 0.044
Manufacturing and services 0.046 ∗∗
Overall localisation 0.005 −0.004
Notes. ∗/∗∗/∗∗∗ denotes rejection of the null hypothesis that ΔTk= 0 at the 90%, 95% or 99% significance level. Increases over
time are reported in bold.
284 EPA: Economy and Space 55(2)
T
ab
le
2.
Ev
ol
ut
io
n
of
sp
ec
ia
lis
at
io
n
pa
tt
er
ns
ac
ro
ss
21
5
EU
re
gi
on
s,
20
03
–
20
19
.
R
el
at
iv
e
sp
ec
ia
lis
at
io
n
(a
R
Si
)
W
ith
in
-c
ou
nt
ry
Be
tw
ee
n-
co
un
tr
y
(T
i)
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
an
d
se
rv
ic
es
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
an
d
se
rv
ic
es
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
Be
nc
hm
ar
ch
:
m
an
uf
ac
tu
ri
ng
an
d
se
rv
ic
es
O
bs
.d
iff
.
Si
g.
O
bs
.d
iff
.
Si
g.
O
bs
.d
iff
.
Si
g.
O
bs
.d
iff
.
Si
g.
O
bs
.d
iff
.
Si
g.
O
bs
.d
iff
.
Si
g.
A
us
tr
ia
−
0.
00
2
−
0.
01
3
−
0.
01
6
0.
00
0
0.
01
4
−
0.
00
3
Be
lg
iu
m
−
0.
02
5
−
0.
00
7
−
0.
02
2
0.
01
4
∗∗
−
0.
00
3
−
0.
01
2
G
er
m
an
y
0.
01
5
−
0.
00
7
0.
01
9
∗
−
0.
00
5
−
0.
00
4
−
0.
00
4
G
re
ec
e
0.
09
7
∗∗
∗
0.
06
6
−
0.
00
2
0.
00
6
0.
09
9
0.
03
6
Sp
ai
n
−
0.
00
8
−
0.
01
6
∗∗
−
0.
01
3
∗
−
0.
01
0
0.
00
5
−
0.
00
1
Fr
an
ce
0.
05
2
∗∗
0.
01
1
0.
03
2
0.
00
7
0.
02
0
0.
00
4
H
un
ga
ry
−
0.
01
4
−
0.
00
4
−
0.
00
5
0.
00
0
−
0.
00
9
−
0.
00
9
Ita
ly
−
0.
01
7
∗∗
−
0.
01
6
−
0.
00
3
−
0.
01
0
∗∗
∗
−
0.
01
4
∗
−
0.
00
1
∗∗
N
et
he
rl
an
ds
0.
03
3
0.
00
5
0.
01
0
−
0.
01
9
0.
02
4
0.
01
5
Po
la
nd
−
0.
01
3
0.
02
4
−
0.
00
8
−
0.
01
1
−
0.
00
5
0.
02
7
∗
Sl
ov
ak
ia
0.
01
4
−
0.
24
8
∗∗
∗
−
0.
02
2
−
0.
09
3
∗∗
0.
03
5
−
0.
14
7
∗∗
Sw
ed
en
0.
03
5
−
0.
03
1
∗∗
∗
0.
01
5
−
0.
00
7
0.
02
1
∗∗
∗
−
0.
02
1
∗∗
∗
U
ni
te
d
K
in
gd
om
0.
00
1
−
0.
00
7
0.
00
8
0.
00
6
−
0.
00
7
−
0.
00
7
O
ve
ra
ll
lo
ca
lis
at
io
n
0.
00
6
−
0.
00
4
0.
00
5
−
0.
00
6
0.
00
4
0.
00
2
N
ot
es
.∗
/∗
∗ /
∗∗
∗
de
no
te
s
re
je
ct
io
n
of
th
e
nu
ll
hy
po
th
es
is
th
at
Δ
T
k
=
0
at
th
e
90
%
,9
5%
or
99
%
si
gn
ifi
ca
nc
e
le
ve
l.
In
cr
ea
se
s
ov
er
tim
e
ar
e
re
po
rt
ed
in
bo
ld
.
Cutrini et al. 285
growing evidence of the fragmentation of production processes across borders with countries and
regions trading tasks and functions rather than products. That is, regional sectoral (‘horizontal’) spe-
cialisation may be giving way to regional functional (‘vertical’) specialisation (Timmer et al., 2019;
see also, e.g., French, 2017; Yi, 2003). With cross-border production sharing, countries and regions
can specialise in particular stages of the production process – such as R&D, fabrication, marketing,
and management. Data classified by sector will not necessarily reveal this type of specialisation:
different types of data are required. Such forces are also linked to the globalisation of production
processes, which has been progressing at a rapid rate during the period under analysis and has
encouraged geographically dispersed production networks. It is quite possible, therefore, that the
economic forces caused by EU integration are relatively smaller on the latter than on the former.
It is possible that some degree of regional specialisation of tasks or functions is replacing regional
sectoral specialisation, but the detailed data required to investigate this are not available on a con-
sistent basis.
We acknowledge that those recent results should be treated as provisional not only because, due
to data availability, they encompass a limited number of manufacturing industries – typically
low-tech industry – and services. In the next section, we focus on manufacturing and describe
the detailed results for the period prior to the global financial crisis.
Relative concentration and specialisation before the Global Financial
Crisis: another look
Our results suggest that during the early phases of EU economic integration, corresponding to the
construction of the single market and the implementation of the single currency, national compara-
tive advantages, dispersion of industries and de-specialisation within countries have driven the
early adjustments in location patterns.
Table 3. Relative concentration of manufacturing industries across 133 EU regions.
Absolute change
Taxonomya
Average 1985–
2007 1985–2001b 1985–2007
Food LT 0.128 −0.054 ∗∗∗ −0.057 ∗∗∗
Textiles LT 0.291 0.034 0.072 ∗
Wood LT 0.197 −0.130 ∗∗ −0.118 ∗
Paper LT 0.132 −0.014 0.014
Chemicals HT 0.170 −0.020 −0.010
Rubber and plastic products LT 0.093 −0.056 ∗∗ −0.050 ∗
Other non-metallic mineral products LT 0.178 −0.032 −0.040
Basic metals and fabricated metal
products
LT 0.094 −0.083 ∗∗∗ −0.090 ∗∗∗
Machinery and equipment nec HT 0.098 −0.025 −0.021
Electrical and optical equipment HT 0.102 −0.046 ∗∗∗ −0.047 ∗∗∗
Transport equipment HT 0.161 0.021 0.024
Manufacturing nec HT 0.138 −0.125 ∗∗∗ −0.105 ∗∗∗
Overall localisation 0.123 −0.042 ∗∗∗ −0.042 ∗∗∗
Notes. ∗/∗∗/∗∗∗ denotes rejection of the null hypothesis that ΔTk= 0 at the 90%, 95% or 99% significance level. Increases in
relative concentration are reported in bold.
aOECD technology classification: L= low-tech, M-L=medium to low-tech; M-H=medium to high-tech.
bResults for the period 1985–2001 are drawn from Cutrini (2010) on 145 NUTS2-NUTS1 regions.
286 EPA: Economy and Space 55(2)
In other words, the economic geography of Europe tended to move toward a scenario
described in Figure 1(b) that is specialisation across countries without clustering of individual
industries.
Table 3 shows the results for the relative concentration of manufacturing industries with their
average values, observed difference and significant levels over the periods 1985–2001 and
1985–2007, the latter computed on the basis of the 133 NUTS2 regions and the NACE rev 1.1 clas-
sifications of industries.
The comparison of the results with the sample used in Cutrini (2010) lends further support to the
general convergence in location patterns of manufacturing industries, also supported by decreasing
levels of clustering of most of the sectors (Cutrini, 2009b). The observed differences in relative con-
centration measures and their significance level are reported in Table 1 while detailed results of the
bootstrap tests for significance are provided in Table A2 (Results Appendix).
Changes over time are usually negative irrespective of the database used and they are also com-
parable in terms of significant level. There are two notable exceptions as having a tendency towards
Figure 4. Evolution of the two components of relative concentration, 1985–2007.
Notes. The dashed lines in the left panels of the graph refer to the between-country component of the
localisation index. The dashed lines in the right panels of the graph refer to the within-country component of
the localisation index.
Cutrini et al. 287
spatial concentration, textile and transport equipment. The industry producing paper and paper pro-
ducts also experienced increasing concentration.
Figure 4 provides an overview of the decomposition of entropy indices that has allowed a dis-
tinction between inner-country and cross-country localisation. For illustration, the dashed lines in
Figure 4 represent the evolution of localisation within and between countries, respectively. These
reference lines are weighted indices that allow identifying those sectors persistently above the
average, such as textiles and transport equipment, in both components. Additional industries
exhibit diverging location patterns relative to total manufacturing from 2001 onwards, namely che-
micals, machinery and equipment and other no-metallic mineral products, but changes over time are
significant only for the latter (Table A2).
If we turn to specialisation we also confirm the prevailing convergence of regional industrial
structures, especially within countries, with the Greek internal geography being the only exception
to the rule. The general convergence in countrys’ structures to the EU-15 average is also confirmed
with more recently available data, with only Germany and Italy still arise as the countries becoming
more specialised over time (Table 4).
While the limitations of data have to be borne in mind, it does not appear to be the case that
agglomeration forces have been sufficient to produce any significant increase in the geographical
concentration of employment within industries, at least at the spatial scale for which analysis has
been possible.
Discussion of results, possible explanations and some concluding
remarks
The results of the relative concentration and specialisation before the Global Financial Crisis:
another look section present some evidence to suggest that, as EU integration has progressed, dis-
persion and de-specialisation of the early stages have been followed by an increase in concentration
and agglomeration, sometimes coupled by an increase in specialisation after 2000. Since 1985 and
until the Global Financial Crisis, localisation and specialisation between countries increased to
Table 4. Specialisation indices and components, absolute change, 1985–2001 and 1985–2007
aRSi aRSwi T
b
i
Diff
1985–2001
a
Diff 1985–
2007
Diff 1985–
2001
Diff 1985–
2007a
Diff 1985–
2001a
Diff 1985–
2007
Greece −0.13 ∗∗∗ −0.128 0.02 0.017 −0.14 ∗∗∗ −0.145 ∗
Belgium and
Luxembourg
−0.13 ∗∗∗ −0.123 ∗∗∗ −0.06 ∗∗∗ −0.043 ∗∗ −0.07 ∗∗∗ −0.080 ∗∗∗
Spain −0.08 ∗∗∗ −0.095 −0.04 ∗∗∗ −0.049 ∗∗∗ −0.04 ∗ −0.046 ∗
United Kingdom −0.06 −0.065 ∗∗ −0.04 −0.048 ∗∗ −0.02 −0.018
Netherlands −0.09 −0.094 ∗∗∗ −0.02 −0.022 ∗∗ −0.07 −0.072 ∗∗∗
Italy −0.01 −0.014 −0.04 ∗∗ −0.039 ∗∗ 0.02 0.025 ∗∗
France −0.01 −0.014 −0.01 −0.012 0 −0.002
Germany −0.02 −0.015 −0.03 ∗∗∗ −0.035 ∗∗∗ 0.02 0.020
Overall
localisation
−0.042 ∗∗∗ −0.042 ∗∗∗ −0.04 ∗∗∗ −0.043 ∗∗∗ −0.007 0.001
Notes. ∗/∗∗/∗∗∗ denotes rejection of the null hypothesis that ΔTk= 0 at the 90%, 95% or 99% significance level. Increases in
the relative specialisation are reported in bold.
a
Results for the period 1985–2001 are drawn from Cutrini (2010).
288 EPA: Economy and Space 55(2)
some extent but with a widespread fall in specialisation and concentration within countries. After
2000, the actual spatial distribution of economic activities in Europe appears to have changed
towards a more complex economic geography, with some degree of concentration and agglomer-
ation, but only sometimes coupled with regional and country specialisation.
From the systematic analysis presented above, it is possible to identify some major trends. First,
extending the period of analysis to 2007 with a sample of consistent NUTS2 regions (NACE rev.
1.1 classification), it would be difficult to argue that increasing integration has led to increasing spe-
cialisation and concentration before the global financial crisis. Our results for the post-EMU period
and post-Crisis years (2003–2019) point to a slight increase in the geographical concentration of
industries and agglomeration among economic activities. Nevertheless, overall, our results still
do not lend support to the line of argument put forward by Krugman (1993) for the European
Union. In fact, after 2000 we observe mixed evidence in terms of the evolution of economic struc-
tures, with decreasing regional and country specialisation (Tables 2, A6 and A7).
It is likely that the historical shock associated with the Covid-19 pandemic, and the recession this
caused, will also have an impact. Indeed, the dramatic changes in the global economy, in global
trade patterns, and the reconfiguration of supply chains including their reshoring, caused by the pan-
demic all render the relationship between national and regional ‘openness’ and connectivity, on the
one hand, and the locational dynamics of economic activity, on the other, much more complex than
in the past. At the time of writing, the lockdowns in China are causing major blockages in supply
chains, as is the war in Ukraine. In addition, future developments – such as the gathering pace of
artificial intelligence and its impact on employment and production, the mounting pressure for shift-
ing to a low-carbon or carbon-neutral economy, and the uncertainty of trade arrangements, and
agreements between Europe and the United States and China – will all leave some imprint on
the economic geography of the European Union.
Finally, what, then, are the potential implications of our empirical findings for regional policy?
The overall trend towards less geographical concentration of economic activity and less regional
sectoral specialisation within the EU countries studied here would seem to raise questions for
those policy approaches that advocate regional specialisation and spatial agglomeration (both
claimed to raise regional productivity). At the very least, it is possible that such policies may
well be running counter to what are the long-run evolutionary trends underway in the economic
geographies of these countries. To add to this ambiguity, Paul Krugman himself has more recently
voiced some doubt as to whether the effects of increasing returns he and other spatial economists
claimed for spatial agglomeration are as important as they once were:
There’s good reason to believe that the world economy has, over time, actually become less character-
ized by the kinds of increasing returns effects emphasised by new trade theory and new economic geog-
raphy. In the case of geography, in fact, the peak of increasing returns occurred long before the theorists
arrived on the scene. (Krugman, 2009: 569)
Our analysis of the regions of the EU suggests that he is right.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publi-
cation of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Cutrini et al. 289
Notes
1. The pre-integration pattern corresponds to a theoretical scenario of perfect regularity (a nil value of the
L-index). In our methodological approach which is described in the Appendix, this location pattern is asso-
ciated with the absence of specialization of regions (Tij, for all ij) and the absence of geographic concen-
tration of industries (Tk = 0, for all k). See Cutrini (2021) and the Appendix for details. In this case, all
regions in the area have the same economic structure, also countries are characterised by the same economic
structures than the benchmark. The spatial distribution of industries is inevitably symmetrical, since all the
industries exhibit perfectly overlapping distributions across spatial units (regions, countries). Notice that
perfect regularity (L = 0) implies that both Lb = 0 and Lw = 0.
2. Detailed results on the bootstrap tests over the period 2003–2019, with different benchmark distributions,
performed as robustness checks, are provided in the Results appendix (Tables A3 and A4).
3. Contract JRC/SVQ/2017/B.5/0034/NC, EC responsible officer Dr Robert Marschinski. See https://publica-
tions.jrc.ec.europa.eu/repository/handle/JRC119363 for final report online.
4. See https://ec.europa.eu/eurostat/web/structural-business-statistics.
5. See https://www.lisa.nl/.
6. See https://www.bvdinfo.com/en-gb.
References
Aquino A (1978) Intra-industry trade and inter-industry specialization as concurrent sources of international
trade in manufactures. Weltwirtschaftliches Archiv 114: 275–296.
Aiginger K and Davies SW (2004) Industrial specialisation and geographic concentration: Two sides of the
same coin? Not for the European Union. Journal of Applied Economics 7(2): 231–248.
Aiginger K and Leitner W (2002) Regional concentration in the United States and Europe: Who follows
whom? Weltwirtschaftliches Archiv 138(4): 652–679.
Aiginger K and Pfaffermayr M (2004) The single market and geographic concentration in Europe. Review of
International Economics 12: 1–11.
Alonso-Villar O (2001) Large metropolises in the third world: An explanation. Urban Studies 38: 1359–1371.
Amiti M (1999) Specialization patterns in Europe. Review of World Economics 135: 573–593.
Arbia G (1989) Spatial Data Configuration in Statistical Analysis of Regional Economic and Related
Problems. Springer Netherlands.
Bagoulla C and Péridy N (2011) Market access and the other determinants of North–South manufacturing loca-
tion choice: An application to the Euro-Mediterranean area. Economic Systems 35: 573–561.
Baldwin R (2016) The Great Convergence: Information Technology and the New Globalisation. Cambridge,
Mass: Belknap Press.
Baldwin R, Forslid R, Martin P, et al. (2003) Economic Geography and Public Policy. Princeton, NJ:
Princeton University Press.
Behrens K, Gaigne C, Ottaviano GI, et al. (2007) Countries, regions and trade: On the welfare impacts of eco-
nomic integration. European Economic Review 51: 1277–1301.
Bickenbach F and Bode E (2008) Disproportionality measures of concentration, specialization and localiza-
tion. International Regional Science Review 31(4): 359–388.
Brakman S, Garretsen JH and Marrewijk C (2009) The New Introduction to Geographical Economics.
Cambridge: Cambridge University Press.
Brülhart M (1998) Trading places, industrial specialisation in the European Union. Journal of CommonMarket
Studies 36(3): 319–346.
Brülhart M and Träger R (2005) An account of geographic concentration patterns in Europe. Regional Science
and Urban Economics 35(6): 597–624.
Combes P-P and Overman H (2004) The spatial distribution of economic activities in the European Union.
In: Henderson V and Thisse JF (eds) Handbook of Urban and Regional Economics. Amsterdam:
Elsevier-North Holland, pp.2845–2909.
Crozet M and Koenig-Soubeyran P (2004) EU Enlargement and the internal geography of countries. Journal of
Comparative Economics 32: 265–279.
290 EPA: Economy and Space 55(2)
Cutrini E (2009a) Using entropy measures to disentangle regional from national localization patterns. Regional
Science and Urban Economics 39(2): 243–250.
Cutrini E (2009b) The changing location of the European industry: A twofold geographical perspective. In:
Salvadori N, Commendatore P and Tamberi M (eds) Geography, Structural Change and Economic
Development: Theory and Empirics. Cheltenham, UK and Northampton, MA, USA: Edward Elgar
Publishing, pp.88–124.
Cutrini E (2010) Specialization and concentration from a twofold geographical perspective: Evidence from
Europe. Regional Studies 44(3): 315–336.
Cutrini E (2021) Entropy, agglomeration and economic structures in spatial analysis. In: Reggiani A, Schintler
LA, Czamanski D and Patuelli R (eds) In Handbook on Entropy, Complexity, and Spatial Dynamics, a
Rebirth of Theory? Edward Elgar, pp.118–138.
Dauth W, Fuchs M and Otto A (2016) Long-run processes of geographical concentration and dispersion:
Evidence from Germany. Papers in Regional Science 97: 569–593.
De Dominicis L, Arbia G and De Groot HL (2013) Concentration of manufacturing and service sector
activities in Italy: Accounting for spatial dependence and firm size distribution. Regional Studies 47:
405–418.
De Robertis G (2001) European Integration and internal economic geography: The case of the Italian manu-
facturing industry 1971–1991. The International Trade Journal 15(3): 345–371.
Ezcurra R, Pascual P and Rapún M (2006) The dynamics of industrial concentration in the regions of the
European Union. Growth and Change 37(2): 200–229.
French S (2017) Revealed comparative advantage: What is it good for? Journal of International Economics
106: 83–103.
Fujita M, Krugman P and Venables AJ (1999) The Spatial Economy: Cities, Regions and International Trade.
Cambridge, Mass: MIT Press.
Fujita M and Thisse J-F (2002) Economics of Agglomeration: Cities, Industrial Location and Regional
Growth. Cambridge: Cambridge University Press.
Gardiner B, Martin RL and Tyler P (2011) Does spatial agglomeration increase national growth? Evidence
from the European Union. Journal of Economic Geography 11: 979–1006.
Gokan T (2010) On the usage of the measurements of geographical concentration and specialization with areal
data. In: Kuroiwa I (ed) Spatial Statistics and Industrial Location in CLMV: An Interim Report. Chiba, JP:
IDE-JETRO.
Gorter J (2002) The economic geography of Europe. CPB Report 2002/4. The Netherlands: Centraal
Planbureau, Bureau for Economic Policy Analysis.
Hallet M (2000) Regional specialisation and concentration in the EU, European Economy - Economic Papers
2008–2015, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.
Krugman P (1991) Geography and Trade. Cambridge, MA: MIT Press.
Krugman P (1993) Lessons of Massachusetts for EMU, ch 8. In: Torres F and Giavazzi F (eds) Adjustment and
Growth in the European Monetary Union. Cambridge: Cambridge University Press, pp.241–261.
Krugman P (2009) The new economic geography: Now middle aged. Regional Studies 45: 1–7.
Krugman P and Livas RE (1996) Trade policy and the third world metropolis. Journal of Development
Economics 49(1): 137–150.
Midelfart-Knarvik KH, Overman HG, Redding SJ, et al. (2004) The Location of European industry. In: Dierx
A, Ilzkovitz F and Sekkat K (eds) European Integration and the Functioning of Product Markets. Edwarg
Elgar.
Monfort P and Nicolini R (2000) Regional convergence and international integration. Journal of Urban
Economics 48: 286–306.
Monfort P and van Ypersele T (2003) Integration, regional agglomeration and international trade. CEPR dis-
cussion paper no. 3752. London: Centre for Economic Policy Research.
Paluzie E, Pons J and Tirado DA (2001) Regional integration and specialization patterns in Spain. Regional
Studies 35: 285–296.
Puga D (1999) The rise and fall of regional inequalities. European Economic Review 43: 303–334.
Rombaldoni R and Zazzaro A (1997) Localizzazione delle imprese manifatturiere e specializzazione regionale
in Italia: 1970–1990. Rassegna dei lavori dell’ISCO XIV: 151–181.
Cutrini et al. 291
Sapir AE (1996) The effects of Europe’s internal market program on production and trade: A first assessment.
Review of World Economics 132(3): 457–475.
Schweizog R and Collins A (2015) A simple location index plus some maps and no apologies: Back to basics
on the development of links between economic integration and spatial concentration of industries.
Tijdschrift voor economische en sociale geografie (Journal of Economic and Social Geography) 106(1):
17–35.
Stierle-von Schultz U and Stierle MH (2013) Regional specialisation and sectoral concentration in an enlarged
EU. INFER WP 2/2013. Cologne, Germany: International Network for Economic Research.
Timmer MP, Miroudot S and de Vries GJ (2019) Functional specialisation in trade. Journal of Economic
Geography 19: 1–30.
Vechiu N and Makhlouf F (2014) Economic integration and specialization in production in the EU27: Does
FDI influence countries’ specialization? Empirical Economics 46: 543–572.
Yi KM (2003) Can vertical specialisation explain the growth of world trade? Journal of Political Economy
111: 52–102.
Appendix
Data appendix. The dataset used in this paper was constructed as part of a project for DG
JRC-Seville entitled ‘Sectoral Analysis and Assessment of Geographical Concentration of EU
Industries (SECOAYA)3’, final report produced in December 2019. What follows is a brief descrip-
tion of the sources, dimensions, and choices made that led to the final database being established.
Data sources. Detailed sectoral-regional data with a pan-European coverage are in short supply. The
only official source of such data is the Structural Business Statistics database4 contained in Eurostat,
with employment used as the main metric of comparison. Although the coverage of the SBS
improves with time, it was found to be insufficient for purpose across certain countries and time
periods, and so was supplemented with a number of other sources which were checked against,
and blended into, the original SBS data to maintain consistency:
- National Statistical Office (NSO) data.
- Firm-level data (in particular the LISA database5 for the Netherlands, and the Bureau van Dijk
(BvD) Orbis database6.
- Cambridge Econometrics’ proprietary regional database (for the UK only).
Time periods. The SBS database has two vintages, one covering the period 1997–2007, the other
covering 2008–2019 (note there is no overlap between the two sets of data). On investigation of
geographical and sectoral coverage, while also wanting to cover a long period of structural
change, it was decided against attempting to construct a full time series and instead four periods
were selected as having the best coverage/time period trade-off: 2003, 2007, 2011, 2015 and 2019
Sectoral coverage. Aside from the lack of overlap of the two SBS database vintages, an additional
challenge was presented by a change in the sectoral classification used – from NACE rev 1.1 to
NACE rev 2. This meant that only a limited number of sectors could be matched between the
two classifications for any degree of (two-way) correspondence of the activities within them.
The final collection of sectors chosen is shown in the Table below.
Geographical coverage. Member States: In selecting those Member States to include in the database,
a number of considerations were made. Firstly, those countries with only a single NUTS2 region
were not of interest, as were countries where coverage did not extend over all four periods being
292 EPA: Economy and Space 55(2)
analysed. Ultimately, this led to the following sample of 13 countries being included: Austria,
Belgium, France, Germany, Greece, Hungary, Italy, the Netherlands, Poland, Slovakia, Spain,
Sweden, and the UK. In most cases, this meant that the proportion of EU sectoral employment
covered by the sample was in excess of 80%.
Regions: For certain countries, regional coverage (often small or island regions) data coverage
was poor or non-existent. Rather than remove the whole country for this reason, in some cases,
these regions were removed from the sample. This includes the following regions, for all
sectors, all FRA NUTS 2 regions (Départements d’Outre-Mer), and ES63 (Ceuta) and ES64
(Melilla) for Spain.
In addition, for two sectors, the NUTS2 data were replaced by NUTS1 aggregates to avoid their
removal from the analysis:
For the Manufacture of basic metals, French regions (FR81, FR82, FR83: Languedoc-
Roussillon, Provence-Alpes-Côte d’Azur, Corsica) were replaced by the FR8 aggregate.
For the manufacture of paper and paper products, AT11, AT12, and AT13 (Burgenland,
Niederösterreich, and Wien) were replaced by the AT1 aggregate, as were the FR81-83 NUTS2
regions for France.
Methodological appendix
Relative entropy measures of localisation, specialisation and concentration. The variable of main interest
is employment in the present case; subscripts i, j, k index country, region and industry, respectively.
Thus:
xijk = employment in manufacturing industry k (k= 1,…, n) of region j ( j= 1,…, ) belonging to
country i (i= 1, …, m)
xij = total manufacturing employment in region i of country j
xik = total employment in the manufacturing industry k of country i
xi = total employment in country i
xk = total employment in manufacturing industry k in Europe
x = total manufacturing employment in Europe
N is the number of manufacturing industries while R is the number of regions in the whole
economy.
Table A1. Final sector selection.
NACE rev 2
code Sector name
C10, C11 Manufacture of food products; manufacture of beverages
C16 Manufacture of wood and of products of wood and cork, except furniture; manufacture of
articles of straw and plaiting materials
C17 Manufacture of paper and paper products
C22 Manufacture of rubber and plastic products
C23 Manufacture of other non-metallic mineral products
C24 Manufacture of basic metals
G45 Wholesale and retail trade and repair of motor vehicles and motorcycles
G46 Wholesale trade, except for motor vehicles and motorcycles
G47 Retail trade, except for motor vehicles and motorcycles
I Accommodation and food service activities
H49 Land transport and transport via pipelines
M72 Scientific research and development
Cutrini et al. 293
The basic dissimilarity Theil index, to measure the relative concentration of one industry k is:
Tk =
∑m
i=1
∑ri
j=1
xijk
xij
ln
xijk
xk
xij
x
⎛
⎜⎜⎝
⎞
⎟⎟⎠ (1)
The two geographical components of the concentration index for each industry k can be derived by
factor decomposition (see Cutrini, 2009a for details on the formal decomposition of the concentra-
tion index defined in equation (1)).
Turning to the specialisation side, it is possible to define several measures, by varying the
observed region and/or the supra-regional references.
Tij =
∑n
k=1
xijk
xij
ln
xijk
xij
xk
x
⎛
⎜⎜⎝
⎞
⎟⎟⎠ (2)
The index Tij evaluates the dissimilarity between the economic structure of one region (composed
of N industries), compared to that of a supranational geographical unit (Europe, USA, world).
To measure the relative specialisation of region ij,with respect to the country i, or with respect to
the supranational area, the following indices are derived:
Twij =
∑n
k=1
xijk
xij
ln
xijk
xij
xik
xi
⎛
⎜⎜⎝
⎞
⎟⎟⎠ (3)
assesses the divergence between the regional manufacturing allocation among industries, and the
country structure; and,
Tbij =
∑m
i=1
xik
xi
ln
xik
xi
xk
x
⎛
⎜⎜⎝
⎞
⎟⎟⎠ (4)
assesses the dissimilarity between the national manufacturing structure and the supranational
benchmark. When a regional perspective is adopted, the average regional specialisation of a
country relative to a supranational geographical area, in our case Europe (aRSi) is composed of
two elements: the within-country component and the country bias. Specifically, the following rela-
tion holds:
aRSi = aRSwi + Tbi (4)
where:
- aRSt =
∑ri
j=1
Tij
xij
xi
is the average regional specialisation of the regions of country i relative to the
supranational manufacturing structure
and
294 EPA: Economy and Space 55(2)
- aRSwi =
∑ri
j=1
Twij
xij
xi
is the average regional specialisation of all the regions of country i relative to
the country’s manufacturing structure.
The entropy index of overall localisation (L-index) is a weighted sum of the logarithms of loca-
tion quotients, where the weights are the industry-region shares of aggregate manufacturing (xijkxi ).
The L-index is a summary statistic of both relative specialisation indices and relative concentration
ones:
L =
∑r
j=1
xij
x
Tij =
∑n
k=1
xk
x
Tk (5)
is a measure of the ‘distance’ between the industrial composition of the country, and the indus-
trial composition of the wider region (relative specialisation of region i, with respect to the supra-
national area).
When dissimilarity logic is adopted, country specialisation, relative to a supranational unit (Tbi ),
can be envisioned as a residual of the averaged regional specialisation, relative to the same bench-
mark – once the divergence of regional manufacturing structures related to that country has been
accounted for (See Cutrini, 2009a for details).
From the specialisation point of view, the aggregation gives an idea of the average dissimilarity
between the regional employment proportions across industries and the manufacturing structure of
the supranational economy selected as benchmark. Similarly, from a concentration standpoint, the
composite measure of localisation informs us about the average dissimilarity between the distribu-
tion of industries across geographical units and the spread of overall manufacturing across geo-
graphical units. Since the size of the regional units affects the degree of relative specialisation,
all of the raw specialisation indices are weighted by the region’s share in total manufacturing
employment (xijx ). Similarly, the difference in industry size is accounted for by weighting each rela-
tive concentration index by the industry’s share in total manufacturing employment (xkx ).
Absolute entropy measures. The importance of distinctive agglomeration forces (e.g., urbanisation,
localisation economies, market access considerations) can be better captured with the accurate com-
bination of absolute and relative indices. Therefore, for a thoughtful assessment of specialisation
and concentration in a period characterised by fundamental structural changes it may be useful
to complement the analysis based on relative indicators with the use of absolute measures.
The most widely used measure of absolute concentration is the normalised Shannon index,
adopted in Aiginger and Davies (2004) and Aiginger and Pfaffermayr (2004), among others.
The normalised Shannon index or absolute Theil index is defined as
Tabs = ln(R) −
∑R
r=1
qijk ln
1
qijk
( )
(8)
which equates to
Tabs = ln(R) −
∑R
r=1
qijk ln qijk
( )
) (9)
where qijk = xijkxk is the regional shares of the distribution of employment in sector k.
The index defined in equation (8) may be categorised as a dissimilarity index between the actual
distribution and the benchmark uniform distribution, the latter of which is associated to the highest
Cutrini et al. 295
degree of uncertainty and disorder. In other words, Tabs can be viewed as a deviation of the
observed entropy from the maximum entropy associate to the uniform distribution – that is, ln(R).
The higher the value of the index, the lower the uncertainty, possibly due to self-organisation and
path dependence in industry-specific location patterns.
For a compendium of entropy-based measures in geographical context, as well as their economic
interpretations, see Cutrini (2021).
Results appendix
Table A2. Bootstrap results for overall localisation, concentration and specialisation measures, absolute
change, 1985–2007; 2001–2007.
Localisation
1985–2007
Obs. Boot.
differ. Std. Err. z Sig.
L −0.042 0.011 3.81 ∗∗∗
Lw −0.043 0.007 6.48 ∗∗∗
Lb 0.001 0.007 0.2
Concentration
1985–2007
Obs. Boot.
differ. Std. Err. z Sig.
Food T −0.057 0.018 3.19 ∗∗∗
Tw −0.032 0.009 3.51 ∗∗∗
Tb −0.025 0.016 1.58
Textiles T 0.072 0.039 1.85 ∗
Tw −0.055 0.030 1.81 ∗
Tb 0.127 0.022 5.76 ∗∗∗
Wood T −0.118 0.064 1.86 ∗
Tw −0.084 0.023 3.67 ∗∗∗
Tb −0.034 0.053 0.65
Paper T 0.014 0.020 0.72
Tw 0.010 0.014 0.77
Tb 0.004 0.011 0.34
Chemicals T −0.010 0.019 0.53
Tw 0.013 0.018 0.72
Tb −0.023 0.012 1.89 ∗
Rubber and plastic products T −0.050 0.029 1.74 ∗
Tw 0.008 0.027 0.29
Tb −0.058 0.007 8.9 ∗∗∗
Other non-metallic mineral products T −0.040 0.037 1.08
Tw 0.028 0.030 0.94
Tb −0.068 0.018 3.78 ∗∗∗
Basic metals and fabricated metal products T −0.090 0.015 6.03 ∗∗∗
(continued)
296 EPA: Economy and Space 55(2)
Table A2. (continued)
Localisation
1985–2007
Obs. Boot.
differ. Std. Err. z Sig.
Tw −0.012 0.014 0.92
Tb −0.078 0.012 6.43 ∗∗∗
Machinery and equipment nec T −0.021 0.016 1.36
Tw −0.021 0.011 1.92 ∗
Tb 0.000 0.012 0.01
Electrical and optical equipment T −0.047 0.012 3.95 ∗∗∗
Tw 0.008 0.010 0.76
Tb −0.055 0.007 7.49 ∗∗∗
Transport equipment T 0.024 0.024 1
Tw 0.039 0.012 3.21 ∗∗∗
Tb −0.016 0.023 0.68
Manufacturing nec T −0.105 0.040 2.61 ∗∗∗
Tw −0.057 0.019 2.95 ∗∗∗
Tb −0.047 0.031 1.55
Specialisation
1985–2007
Obs. Boot.
differ. Std. Err. z Sig.
aRSi −0.123 0.021 5.72 ∗∗∗
Belgium and Luxembourg aRSi
w −0.043 0.021 2.04 ∗∗
Ti
b −0.080 0.027 2.97 ∗∗∗
aRSi −0.015 0.021 0.71
Western Germany aRSi
w −0.035 0.013 2.63 ∗∗∗
Ti
b 0.020 0.013 1.53
aRSi −0.095 0.035 2.69 ∗∗∗
Spain aRSi
w −0.049 0.017 2.87 ∗∗∗
Ti
b −0.046 0.027 1.72 ∗
aRSi −0.014 0.014 0.99
France aRSi
w −0.012 0.013 0.95
Ti
b −0.002 0.010 0.2
aRSi −0.128 0.086 1.5
Greece aRSi
w 0.017 0.026 0.63
Ti
b −0.145 0.083 1.74 ∗
aRSi −0.014 0.023 0.63
Italy aRSi
w −0.039 0.018 2.16 ∗∗
Ti
b 0.025 0.012 2.07 ∗∗
aRSi −0.094 0.022 4.23 ∗∗∗
Netherlands aRSi
w −0.022 0.009 2.43 ∗∗
Ti
b −0.072 0.022 3.33 ∗∗∗
aRSi −0.065 0.028 2.31 ∗∗
United Kingdom aRSi
w −0.048 0.020 2.34 ∗∗
Ti
b −0.018 0.012 1.52
Cutrini et al. 297
Table A3. Bootstrap results for overall localisation and concentration measures, absolute change, 2003–2019
benchmark: manufacturing.
2003–2019
Obs. Boot.
differ.
Std.
Err. |Z| Sig.
L 0.005 0.006 0.89
Lw 0.003 0.005 0.71
Lb 0.001 0.004 0.36
2003–2019
Obs. Boot.
differ. Std.
Err.
|Z| Sig.
Manufacture of food products; manufacture of beverages T 0.011 0.005 2.14 ∗∗
Tw 0.002 0.003 0.58
Tb 0.009 0.004 2.35 ∗∗
Manufacture of wood and of products of wood and cork, except furniture;
manufacture of articles of straw and plaiting materials
T 0.057 0.022 2.60 ∗∗∗
Tw 0.040 0.014 2.91 ∗∗
Tb 0.017 0.014 1.21
Manufacture of paper and paper products T −0.022 0.016 1.37
Tw −0.002 0.013 0.18
Tb −0.019 0.009 2.27 ∗∗
Rubber and plastic products T 0.029 0.012 2.51 ∗∗
Tw 0.018 0.012 1.56
Tb 0.011 0.009 1.24
Manufacture of other nonmetallic mineral products T −0.019 0.011 1.75 ∗
Tw 0.005 0.008 0.60
Tb −0.023 0.010 2.35 ∗∗∗
Manufacture of basic metals T 0.063 0.031 2.02 ∗∗∗
Tw 0.037 0.027 1.38
Tb 0.026 0.020 1.33
Table A4. Bootstrap results for overall localisation and concentration measures, absolute change, 2003–2019
benchmark: manufacturing and services.
2003–2019
Obs. Boot.
differ.
Std.
Err. |Z|
L −0.004 0.005 0.861
Lw −0.006 0.004 1.384
Lb 0.002 0.002 0.807
Manufacture of food products; manufacture of beverages T 0.008 0.015 0.567
Tw 0.007 0.012 0.569
Tb 0.002 0.007 0.244
(continued)
298 EPA: Economy and Space 55(2)
Table A4. (continued)
2003–2019
Obs. Boot.
differ.
Std.
Err. |Z|
Manufacture of wood and of products of wood and cork, except
furniture; manufacture of articles of straw and plaiting materials
T 0.108 0.022 4.900 ∗∗∗
Tw 0.072 0.016 4.366 ∗∗∗
Tb 0.036 0.020 1.805 ∗
Manufacture of paper and paper products T −0.006 0.025 0.244
Tw −0.004 0.019 0.212
Tb −0.002 0.013 0.142
Rubber and plastic products T 0.070 0.018 3.782 ∗∗∗
Tw 0.026 0.019 1.393
Tb 0.044 0.017 2.576 ∗∗∗
Manufacture of other nonmetallic mineral products T −0.013 0.021 0.628
Tw −0.007 0.014 0.518
Tb −0.006 0.016 0.376
Manufacture of basic metals T 0.042 0.049 0.863
Tw 0.027 0.040 0.685
Tb 0.015 0.028 0.523
Wholesale and retail trade and repair of motor vehicles and
motorcycles
T 0.011 0.006 1.819 ∗
Tw 0.008 0.006 1.371
Tb 0.004 0.002 1.506
Wholesale trade, except for motor vehicles and motorcycles T −0.014 0.008 1.632
Tw −0.011 0.007 1.501
Tb −0.003 0.003 1.105
Retail trade, except for motor vehicles and motorcycles T −0.003 0.004 0.597
Tw 0.002 0.004 0.553
Tb −0.005 0.001 3.816 ∗∗∗
Accommodation and food service activities T −0.006 0.019 0.290
Tw −0.013 0.012 1.091
Tb 0.007 0.011 0.684
Land transport and transport via pipelines T −0.006 0.007 0.955
Tw −0.010 0.005 2.293 ∗∗
Tb 0.004 0.005 0.870
Scientific research and development T −0.166 0.044 3.750 ∗∗∗
Tw −0.106 0.036 2.924 ∗∗∗
Tb −0.060 0.026 2.312 ∗∗
Cutrini et al. 299
Table A5. Bootstrap results for absolute concentration measures, absolute change, 2003–2019.
2003–2019
Obs. Boot.
differ.
Std.
Err. |Z| Sig.
Manufacture of food products; manufacture of beverages 0.072 0.035 2.06 ∗∗
Manufacture of wood and of products of wood and cork, except furniture;
manufacture of articles of straw and plaiting materials
0.002 0.032 0.08
Manufacture of paper and paper products 0.046 0.033 1.39
Rubber and plastic products 0.059 0.033 1.79 ∗
Manufacture of other nonmetallic mineral products 0.035 0.070 0.50
Manufacture of basic metals 0.130 0.048 2.73 ∗∗∗
Wholesale and retail trade and repair of motor vehicles and motorcycles 0.009 0.016 0.59
Wholesale trade, except for motor vehicles and motorcycles 0.016 0.016 0.98
Retail trade, except for motor vehicles and motorcycles 0.047 0.020 2.38 ∗∗
Accommodation and food service activities 0.082 0.091 0.90
Land transport and transport via pipelines 0.021 0.015 1.38
Scientific research and development −0.106 0.049 2.16 ∗∗
Manufacturing as a whole 0.044 0.029 1.51
Manufacturing and services 0.046 0.021 2.13 ∗∗
Table A6. Bootstrap results for relative specialisation measures, absolute change, 2003–2019 benchmark:
manufacturing and services.
2003–2019
Obs. Boot.
differ. Std. Err. |Z|
Austria aRSi −0.013 0.009 1.493
aRSi
w 0.000 0.004 0.725
Ti
b −0.003 0.007 0.413
Belgium aRSi −0.007 0.009 0.870
aRSi
w 0.014 0.005 2.334 ∗∗∗
Ti
b −0.012 0.008 1.486
Germany aRSi −0.007 0.007 1.000
aRSi
w −0.005 0.005 0.836
Ti
b −0.004 0.004 0.918
Greece aRSi 0.066 0.057 1.161
aRSi
w 0.006 0.013 2.753
Ti
b 0.036 0.048 0.758
Spain aRSi −0.016 0.007 2.251 ∗∗
aRSi
w −0.010 0.043 0.021
Ti
b −0.001 0.039 0.023
France aRSi 0.011 0.013 0.859
aRSi
w 0.007 0.013 0.282
Ti
b 0.004 0.007 0.524
Hungary aRSi −0.004 0.013 0.345
(continued)
300 EPA: Economy and Space 55(2)
Table A6. (continued)
2003–2019
Obs. Boot.
differ. Std. Err. |Z|
aRSi
w 0.000 0.004 2.122
Ti
b −0.009 0.012 0.696
Italy aRSi −0.016 0.005 3.186 ∗∗∗
aRSi
w −0.010 0.003 0.276 ∗∗∗
Ti
b −0.001 0.005 0.195
Netherlands aRSi 0.005 0.010 0.536
aRSi
w −0.019 0.005 2.883 ∗∗∗
Ti
b 0.015 0.007 2.166 ∗∗
Poland aRSi 0.024 0.020 1.229
aRSi
w −0.011 0.008 3.437
Ti
b 0.027 0.014 1.942 ∗
Slovakia aRSi −0.248 0.051 4.886 ∗∗∗
aRSi
w −0.093 0.038 3.853 ∗∗
Ti
b −0.147 0.064 2.308 ∗∗
Sweden aRSi −0.031 0.009 3.408 ∗∗∗
aRSi
w −0.007 0.005 4.116
Ti
b −0.021 0.007 2.948 ∗∗∗
UK aRSi −0.007 0.007 1.040
aRSi
w 0.006 0.004 1.826
Ti
b −0.007 0.006 1.121
Table A7. Bootstrap results for relative specialisation measures, absolute change, 2003–2019 benchmark:
manufacturing.
2003–2019
Obs. Boot.
differ. Std. Err. |Z|
Austria aRSi −0.002 0.015 0.105
aRSi
w −0.016 0.012 1.359
Ti
b 0.014 0.014 1.029
Belgium aRSi −0.025 0.019 1.298
aRSi
w −0.022 0.014 1.626
Ti
b −0.003 0.013 0.212
Germany aRSi 0.015 0.013 1.180
aRSi
w 0.019 0.012 1.668 ∗
Ti
b −0.004 0.006 0.662
Greece aRSi 0.097 0.024 4.047 ∗∗∗
aRSi
w −0.002 0.012 0.195
Ti
b 0.099 0.024 4.190
Spain aRSi −0.008 0.021 0.392
aRSi
w −0.013 0.007 1.880 ∗
Ti
b 0.005 0.020 0.241
(continued)
Cutrini et al. 301
Table A7. (continued)
2003–2019
Obs. Boot.
differ. Std. Err. |Z|
France aRSi 0.052 0.022 2.323 ∗∗
aRSi
w 0.032 0.024 1.345
Ti
b 0.020 0.017 1.173
Hungary aRSi −0.014 0.024 0.559
aRSi
w −0.005 0.011 0.449
Ti
b −0.009 0.024 0.370
Italy aRSi −0.017 0.007 2.558 ∗∗
aRSi
w −0.003 0.006 0.533
Ti
b −0.014 0.008 1.788 ∗
Netherlands aRSi 0.033 0.026 1.261
aRSi
w 0.010 0.015 0.656
Ti
b 0.024 0.026 0.917
Poland aRSi −0.013 0.011 1.240
aRSi
w −0.008 0.007 1.136
Ti
b −0.005 0.011 0.481
Slovakia aRSi 0.014 0.069 0.202
aRSi
w −0.022 0.018 1.199
Ti
b 0.035 0.061 0.579
Sweden aRSi 0.035 0.030 1.177
aRSi
w 0.015 0.010 1.460
Ti
b 0.021 0.030 0.679 ∗∗∗
UK aRSi 0.001 0.016 0.068
aRSi
w 0.008 0.017 0.471
Ti
b −0.007 0.006 1.292
302 EPA: Economy and Space 55(2)