WORKING PAPERS
Working Paper No 2011-38
July 2011
The geography of innovation
in the Luxembourg
metropolitan region: an
intra-regional approach
Vincent DAUTEL
Olivier WALTHER

CEPS/INSTEAD Working Papers are intended to make research findings available and stimulate comments and discussion.
They have been approved for circulation but are to be considered preliminary. They have not been edited and have not
been subject to any peer review.
The views expressed in this paper are those of the author(s) and do not necessarily reflect views of CEPS/INSTEAD.
Errors and omissions are the sole responsibility of the author(s).

1
The geography of innovation in the Luxembourg
metropolitan region: an intra-regional approach1
Vincent Dautel
Enterprise research unit, Centre for Population, Poverty and Public Policy Studies
(CEPS/INSTEAD), 3 avenue de la Fonte, L-4364 Esch/Belval, Luxembourg
E-mail: Vincent.Dautel@ceps.lu

Olivier Walther
Department of Geography, Centre for Population, Poverty and Public Policy Studies
(CEPS/INSTEAD), 3 avenue de la Fonte, L-4364 Esch/Belval, Luxembourg
E-mail: Olivier.Walther@ceps.lu
19 July 2011
Abstract
The main objective of the paper is to analyse the local determinants of
innovation in the Luxembourg metropolitan region. We are
particularly interested in the impact of the local milieu and
characteristics of firms. Our paper addresses two specific research
questions. Firstly, we examine the extent to which geographic space is
a determinant of innovation for five intra-regional units based on an
aggregation of municipalities. Secondly, we investigate whether
innovation is dependent on accessibility to the mean centre. In both
cases, we examine innovation propensity and innovation output using
microdata from the Community Innovation Survey (CIS 2006) carried
out in Luxembourg. The paper shows that space matters both in terms
of spatial units and accessibility within the intra-regional context of
Luxembourg. It provides, in particular, first evidence of a close link
between the effects on innovation at the intra-regional level of firms‟
profiles and agglomeration externalities. Both favour innovation for
firms from Luxembourg-City and, to a lesser extent, from the
Suburban Area.
Keywords: intra-regional innovation; firms‟ profile; location factors;
local polynomial regression; Luxembourg metropolitan region.
JEL classification codes: O31, O38, C14, R11, R12

1 This research is supported by core funding for CEPS/INSTEAD from the Ministry of Higher Education and
Research of Luxembourg. The authors would like to thank Cindy Lopes Bento, Sabine Dörry, Alessio Fusco,
Ludovic Halbert and Philippe van Kerm for their useful comments.

2
1. Introduction

It has been widely recognised in academic literature that geographic space, in addition to
characteristics specific to a particular firm or industry, is a key determinant of innovation and
technological change (Audretsch and Feldman 2004, Fagerberg 2005). The empirical
evidence suggests, for example, that being located in an agglomeration rich in knowledge
resources is more conducive to the growth of a firm than being located in a less well-endowed
region (Audretsch and Dohse 2007), because of positive externalities and spatial selection.

Location provides different types of externalities, such as knowledge spillovers, labour
pooling, and backward-forward linkages, which enhance innovation opportunities (Krugman
1991, Johansson and Lööf 2008). The sum of these externalities is defined by Shefer and
Frenkel (1998) as the „local innovation milieu‟. In addition, spatial selection has an important
role to play, firstly because firms may consider the effectiveness of potential spillovers when
deciding on a location (Audretsch and Feldman 1996), and secondly because local
competition may encourage innovation by forcing firms to innovate or fail (Porter 1990).
Firms in large urban regions are therefore expected to have higher rates of innovation, to
adopt innovations more rapidly (Glaeser 1999, Feldman and Audretsch 1999) and to be more
productive than firms located elsewhere.

Such externalities and spatial selection have been extensively studied at the international and
inter-regional levels, both between urban regions within a country (Andersson and Löof 2009,
Antonietti and Cainelli 2009, Czarninski and Hottenrott 2009, Lejpras and Stephan 2009,
Broekel and Brenner 2010), and between urban regions in a selection of EU countries
(Sternberg and Arndt 2001, Simmie 2003, Gössling and Rutten 2007, Copus, Skuras and
Tsegenidi 2008, Heidenreich 2009). Spatial selection has also been recently reconsidered
through theoretical approaches (Baldwin and Okubo 2006, Melitz and Ottaviano 2008).

However, little attention has been paid to the intra-regional level, especially with regard to
metropolitan regions. Notwithstanding the important contributions of Frenkel (2001), Oort
(2002), Smith, Broberg and Overgaard (2002), and Teirlinck and Spithoven (2008), who have
conducted research on the relation between urban cores and their peripheries, and Kaufmann
(2007), Arauzo-Carod and Viladecans-Marsal (2009) who have conducted research on the
Vienna and Spain regions using a metropolitan approach, this level of analysis appears to
have been rather neglected in Europe. This is surprising considering the important role of
metropolitan regions in the European economy, as well as their relative internal spatial
heterogeneity (Krätke 2007).

This paper is an attempt to fill this gap by analysing the local determinants of innovation at
the intra-regional scale in the Luxembourg metropolitan region. We are particularly interested
in the impact of the location of firms within local milieu and the characteristics of firms or
sectors on innovation activities and innovation impact at the intra-regional level. Two
measures of innovation activities will be examined through two different spatial variables.

Firstly, we examine whether the propensity to innovate and innovation output depend on the
regions where firms are located at intra-regional level. Our hypothesis is that firms located in
Luxembourg-City tend to be more innovative than firms located in the periurban areas
because they benefit from agglomeration economies, face higher competition, need to
compensate for agglomeration diseconomies (notably high land rents), and have adequate
characteristics to innovate. The composition of Luxembourg-City and other parts of the

3
regional system are not random. With the exception of Combes et al.‟s (2009) study on
productivity advantages, previous literature does not clearly separate between location effects
and the effects of firms‟ characteristics. We expect there to be a cumulative effect from both.

Secondly, we investigate whether the propensity to innovate and innovation output of firms
also depend on their accessibility to the centre of gravity (or mean centre). Studies of
functional regions, i.e. integrated labour market areas organised by intra-regional commuting,
usually assume that location characteristics should affect all firms in a given functional unit in
the same way. By looking at accessibility at the level of the firm, we would like to challenge
this perspective and provide an original view of the Luxembourg regional innovation system.
Such approach is rare in the literature, Anderson, Quigley and Wilhelmson (2004), Rosenthal
and Strange (2005) and Arzaghi and Henderson (2008) being the exceptions.

This paper is organised as follows. In the next section, we examine the scholarly debate on the
geography of innovation, focusing particularly on local external economies and on spatial
attractiveness in relation to the location of innovative firms. In section 3, we set out our
methodology and discuss the characteristics of Luxembourg. Section 4 provides some stylised
facts about the Luxembourg functional regions organised around Luxembourg-City. Section 5
presents the results, focusing on the propensity of firms to innovate and, innovation output in
relation to, in a first instance, the different functional regions and, in a second instance, to the
accessibility of the centre of gravity. In the last section, we conclude with a summary of the
key findings.

2. Literature

The determinants of a firm‟s innovation behaviour are usually described both as consisting of
both internal and external factors (Sternberg and Arndt 2001, Audretsch and Feldman 2004).
The growing complexity of knowledge has made it increasingly difficult for firms in a
dynamic environment to capitalise internally on all types of knowledge (Lane, Koka and
Pathak 2006). Firms are therefore encouraged to specialise (Powell, Koput and Smith-Doerr
1996) and to focus on the acquisition of external knowledge. Different options are available to
firms seeking to capitalise on external knowledge: focusing on knowledge spillovers,
developing cooperative agreements and acquiring knowledge. Economic geography focuses
on these sources of external economies by dividing them into three different types: location
factors linked to the general characteristics of the region, which all affect the intra-regional
performance of firms; local technology and innovation policies, which may give an incentive
to firms to innovate; and the overall firm environment at the extra-regional level, which
influences firms‟ development.

Previous studies have also emphasised the role of firms‟ environments in stimulating
innovation activity (e.g. Audretsch and Feldman 1996, Autant-Bernard 2001) by suggesting
interactions between environmental factors and firms‟ internal factors. Indeed, in line with the
notion of absorptive capacities introduced by Cohen and Levinthal (1989, 1990) and further
developed by Cockburn and Henderson (1998), spillover effects may be more readily
assimilated if a plant has sufficient in-house capabilities. Firms are likely to exploit these
opportunities in the selection of their location, in an example of a spatial attractiveness
process. In addition, those who do not take advantage from these opportunities may have to
move or may fail, being disadvantaged locally in competitive terms. This section focuses
firstly on intraregional external economies and secondly on spatial attractiveness and spatial
selection.

4
2.1. Local external economies

Urban regions provide a favourable environment in terms of external factors, especially for
highly innovative firms. In the US, for example, Audretsch and Feldman (1996) found
innovative activity to cluster to an even greater extent than other activities. They also found
that this propensity to cluster geographically tends to increase for industries where new
economic knowledge plays a more important role. Patent citations are also highly localised
(see Jaffe, Trajtenberg and Henderson 1993, Almedia and Kogut 1997). In addition,
Audretsch and Feldman (2004) found that, in the US, the majority of new product innovations
were located in cities.

The reasons why innovation tends to be mostly associated with urban regions and to diffuse
faster within urban regions are well known (Gordon and McCann 2000): (1) urban regions
limit transaction costs and provide better access to markets, resources and/or land rents; (2)
urban regions provide a dense institutional context which favours the establishment of
business relations and trust among economic actors; and finally (3) urban regions provide
firms with Marshall agglomeration economies, such as a qualified pool of labour, knowledge
spillovers and backward-forward linkages (Krugman 1991). The size and qualification of the
labour market is indeed recognised as a key factor both for firms and for workers located in
urban regions, which lowers their risk of labour shortage and unemployment. Agglomeration
economies also arise from the possibility of learning from the experiences and innovation of
other firms. Well-integrated firms have better access to non-codified information than isolated
firms. Finally, agglomeration economies are also linked to the possibility of similar firms
sharing suppliers. Interestingly, each of these externalities seems to have a different spatial
range: more than 50 km for forward-backward linkages, less than 50 km for labour market
pooling and less than 10 km for information spillovers, the externalities most sensitive to
distance (Duranton and Overman 2006). Rosenthal and Strange (2005), analysing
entrepreneurship in the New York metropolitan area, also found that if agglomeration effects
exist they quickly decrease with distance (over five miles, in their case study) highlighting
that close proximity matters. These mechanisms suggest, therefore, that the concentration of
firms and workers in space favours productivity, because it improves the likelihood and
effectiveness of interactions between social actors and hence the propensity to innovate.

The question of why proximity is important for innovative activities has received a great deal
of attention over recent decades. Dosi (1988), for example, identifies the following five
characteristics of the innovation process: uncertainty, the increasing role of scientific input,
complexity, „tacit knowledge‟ that can only be obtained by special learning processes, and the
cumulative nature of innovation. Building on this, Feldman (1994a and 1994b) developed the
theory that location mitigates the inherent uncertainty of innovative activity: proximity
enhances the ability of firms to exchange ideas and be cognisant of important incipient
knowledge, hence reducing uncertainty for firms that work in new fields. This reduces the
costs of scientific discovery and commercialisation. Innovation activities depend on external
opportunities including knowledge spillovers and location, and proximity is important for
their exploitation. Spillovers are indeed spatially mediated (Audretsch and Feldman 1996).
Some studies consider even that knowledge spillovers tend to be geographically bounded
within the region where new economic knowledge was created (Agrawal 2002, Anselin,
Varga and Acs 1997, Autant-Bernard 2001, Black 2005, Orlando 2000).

5
2.2 Spatial attractiveness and spatial selection of firms

In addition, Feldman (1994a) suggests that innovative firms tend to be located in areas where
resources resulting from previous innovation success are available, in what can be viewed as a
path-dependent process. In this way, spatial attractiveness affects the location of innovative
firms.

Firstly, firms may consider the effectiveness of potential spillovers when choosing a location,
especially in industries where the generation of new economic knowledge is of relatively high
importance. This line of argument has also drawn attention to the presence of universities as a
locational factor (see Anselin, Varga and Acs 1997, Fischer and Varga 2003, Varga 2000, or
Huffman and Quigley 2002). Audretsch, Lehmann and Warning (2003) link locational choice,
as a strategic firm decision, to knowledge externalities in general and spillovers from
universities in particular. They conclude that geographic proximity may be important to
accessing the human capital embodied in university graduates, who may serve as a spillover
mechanism. Duranton and Puga (2001) highlight in addition that spatial attractiveness may
depend on firms‟ product lifecycles. Their study shows that diversified cities may serve as
nurseries for firms, by providing a fertile experimentation environment favouring new
products development, before these firms move to a more specialised cities with lower costs.

Secondly, spatial selection between firms also contributes to the location process of
innovative activities. Porter (1990) argues, based on case evidence, that local competition
encourages innovation by forcing firms to innovate or fail. Melitz and Ottaviano (2008) show
also that larger markets attract more firms, which makes competition tougher, causing less
productive firms to exit.

The fact that a number of theoretically-sound location factors help explain why firms are
located in particular area on the basis of their involvement in innovation activities does not,
nevertheless, mean that firms will automatically move to their ideal location. As Lejpras and
Stephan (2010) show, firms are faced with the importance of sunk costs, which tend to reduce
mobility. Finally, founder preferences or other factors, such as grants offered by the (local)
government (see Devereux, Griffith and Simpson 2007), have a role to play. However, as
highlighted by Audretsch and Dohse (2007) in their examination of firms‟ growth across
regions, the non-randomness of firms has to be controlled in order to examine the specific
effect of agglomeration economies. This leads us to specifically distinguish between firms‟
profiles, favouring innovation activities and agglomeration economies in our examination of
intra-regional innovativeness. In the selection of these firms‟ characteristics we refer to the
innovation literature, which highlights different factors such as firm size, economic sector,
competition, internal skills and group membership enhancing innovation activity.

The incidence of these factors can be summarised as follows. Large firms typically have more
resources for product innovation and are also led to innovate in terms of processes in order to
decrease costs or to improve products. These incentives reflect large firms‟ commitment to
multiple projects (Veugelers and Cassiman 1999). Innovation is also found to vary
substantially across sectors (see Patel and Pavitt 1995). Technological opportunities (Nelson
and Wolff 1997), appropriation conditions and demand (Arvanitis and Hollenstein 1994)
which vary across sectors all favour innovation activities. In addition, internal skills favour in-
house capabilities and absorptive capacities, i.e. the ability to acquire external knowledge, to
assimilate it and to exploit it (Cohen and Levinthal 1989 and 1990). The increasing
complexity of the knowledge needed to innovate implies that innovation depends increasingly

6
on external knowledge (Fagerberg 2005). Firms operating mainly in foreign markets face
higher costs than those operating only within their national market, due to the additional cost
of developing their distribution or supply networks. Therefore, they need to have specific
advantages in order to deal with this situation. Innovations are likely to provide these needed
advantages. Finally, group membership is expected to provide absorptive capacity and
therefore opportunities to innovate. The main reason for the existence of multinationals is
their ability to transfer and exploit knowledge more efficiently than the market (Hymer 1976).

3. Data and case study

3.1 The community innovation survey (CIS)

The microdata used are taken from the Community Innovation Survey (CIS2006) carried out
in Luxembourg by CEPS/INSTEAD on behalf and under the methodological responsability of
Statec. This survey provides information about types of innovation implemented, innovation
success and firms‟ profiles. In this study we consider as “innovation firms” those who were
involved in product or process innovation in 2004-2006, as defined by the guidelines of the
OSLO-Manual (OECD and Eurostat 2005). We will also use, as in standard literature, the
percentage of turnover from product innovation as a measure of innovation success. Different
studies have already used this survey to examine firms‟ innovation success in the context of
Luxembourg, without however considering the potential impact of firms‟ locations (e.g.
Dautel 2008 and 2010).

The target population of this survey is firms with at least ten employees, belonging to the
following sectors: the „manufacturing sector‟, „electricity, gas and water supply‟, „wholesale
trade‟, „transport, storage and communication‟, „financial intermediation‟, „computer and
related activities‟, „research and development‟, „architectural and engineering activities‟, and
„technical testing and analysis‟. The final sample obtained for this survey, resulting from a
high response rate (around 90%), covered around 40% of this target population.

As noted previously, firms‟ characteristics favouring innovation activities are expected not to
be randomly distributed over space. This survey allows us to consider the following
characteristics over space: firm size, economic sector, group membership, nature of the
market (mainly international or not), reflecting the competition faced by the firm, and, finally,
percentage of employees with higher education, reflecting internal skills.

3.2 Spatial variables

In order to investigate the geography of innovation in Luxembourg, we follow Cheshire and
Gordon (1998) and Johansson, Klaesson and Olsson (2002) by defining a functional urban
region as a set of municipalities between which there is intense labour market commuting.
Despite a lack of comparative studies, we believe that these regions provide a coherent
framework for analysing the development of innovative processes among firms because
proximity externalities occur mainly within functional regions (Karlsson and Manduchi
2001).

The geography of innovation is investigated in terms of two different spatial factors. First, we
divide space into functional regions, which are organised around Luxembourg-City, as
previously defined by Sohn and Walther (2008) according to morphological and functional
criteria and recently used by Walther and Dautel (2010) to analyse intra-regional employment

7
growth. In this study, we use geographical divisions based on a monocentric vision of the
nation state. Five intra-regional units based on an aggregation of municipalities (NUTS 5
level) allow us to study centre-periphery dynamics: (1) Luxembourg-City (2) The
Luxembourg Urban Area, not including Luxembourg-City. (3) A Suburban Area in which the
proportion of commuters working in the Agglomeration was higher than 40% of the active
population in 2002. (4) A South Area in which the proportion of commuters working in the
Agglomeration was lower than 40% of the active population in 2002, located in the country‟s
former southern industrial basin. (5) A Commuter Area in which the proportion of commuters
working in the Agglomeration was between 8.3 and 39.9% of the active population in 2002.

Map 1. Surveyed firms in the Luxembourg metropolitan area

Source: Authors.

8
This division of space allows us to take into account the different economic dynamics existing
between the centre and periphery. The regions thus created remain, however, very rough. It is
possible to refine the analysis by calculating the accessibility of each firm. Accessibility is
defined as the distance of each firm to the centre of gravity of the spatial distribution of all
firms.

Accessibility was calculated according to a two-step procedure. Using ArcView, a
Geographical Information System software, we first identified the centre of gravity (also
known as mean centre or geographic centre) of our set of firms. The centre of gravity is the
average x and y coordinates of all firms in the Luxembourg metropolitan area. It is located
north of the municipality of Luxembourg-City, in the district of Rollingergrund. This is an
illustration of the global centrality of Luxembourg-City at the national level. The fact that this
centre of gravity is located here, and not in one of the main business districts of the city, for
example, is explained by the strong dispersion of firms in the municipal territory. Since
economic activity in the country is so concentrated in the City and Urban Area of
Luxembourg, the centre of gravity calculated in this study can be considered as an
approximation of the theoretical most central location to economic activity, i.e. the point in
space that firms can reach by covering the shortest distance.

Then we calculated the accessibility between the centre of gravity and every surveyed firm,
expressed in an average number of minutes of travel. This value corresponds to the time
needed to cover the distance between the two points by car, using the existing road network
without disruption of traffic and observing the speed limits. It is an approximation of the
dispersion of firms in space and allows us to make the assumption that there are specific
benefits to being located at a close distance from that point. As the article will show, this does
not necessarily mean that the density of firms is highest at this location.

3.3 The Luxembourg metropolitan region

Contrary to what might be expected, given the small size of the Luxembourg national territory
(2586 km²), the spatial distribution of economic activities is highly heterogeneous. The
following sections discuss the extent to which this spatial heterogeneity could affect
innovation positively or negatively by focusing predominantly on the main functional regions
of the country.

On the one hand, as shown by Walther and Dautel (2010), Luxembourg-City still is by a wide
margin the main employment centre of the country, and most of the knowledge-intensive
activities are primarily concentrated within the city despite lower employment growth there
than in its surrounding regions. In 2008, more than 73% and 80% of employment linked to the
knowledge economy was located in Luxembourg City and the Urban Area respectively
(Walther 2011).

Luxembourg-City is also the only geographical unit undergoing specialisation of its economic
activities, and the only one which is specialised in highly job-creating sectors such as
financial knowledge-intensive services, and high-tech, market-intensive and other business
services which tended to support the finance industry (Dautel and Walther 2009). The
increasing concentration of such economic activities in the central and Kirchberg areas in
Luxembourg-City has over time contributed to the development of a highly interconnected
cluster where non-codified information is exchanged between firms, the national regulator
and the political authorities (Pieretti, Bourgain and Courtin 2007, Walther and Schulz 2009,

9
Schulz and Walther 2010). Such concentration is very likely to favour agglomeration
economies in Luxembourg-City.

On the other hand, the spectacular development of these activities, compared to the size of the
city population – Luxembourg has 1.5 more jobs than residents – has generated strong
competition between firms for access to the central city, resulting in a process of employment
suburbanisation. As a consequence, the Luxembourg Urban Area has experienced strong
employment growth since the 1990s and has developed as a prime location for manufacturing
industries or less knowledge-intensive firms which could not compete with rising land and
rent prices. Further from the capital, the South and Commuter Areas have experienced lower
employment growth and have been increasingly integrated into the metropolitan economy
over time, providing low-skilled services in the case of the former and a diversified mix of
services and manufacturing activities in the case of the latter. Very recently, the number of
high-technology and knowledge-intensive jobs grew significantly in the South Region which
is explained by the establishment of back office financial services and, to a lesser extent, by
the relocation of a public research centre previously located in the capital. This process could
potentially offer new opportunities for firms willing to develop new products outside the
capital city, but no clear evidence of such a pattern can yet be observed.

Such a structure, in which a central urban area reinforces its attractiveness, is consistent not
only with previous studies conducted in large European metropolitan centres (such as London
or Paris) but also with smaller specialised financial centres such as Zurich (Thierstein et al.
2008). It suggests that, even though the size of a urban region is usually regarded as a
prominent factor in explaining the propensity of firms to innovate (Andersson and Karlsson
2004), smaller urban regions are also likely to develop competitive regional innovation
systems, especially when these regions are highly specialised in high-tech or knowledge-
intensive activities (Camagni and Capello 2004).

4. Stylised facts on the Luxembourg functional regions

The propensity to innovate can be seen to vary greatly between the Luxembourg functional
regions. As shown in Table 1, firms from Luxembourg-City, the Luxembourg Urban Area and
the Suburban Area innovate more often than in the South and Commuter Areas. Those from
Luxembourg-City (60%) tend to innovate twice as much as those in the South (31%) and
Commuter Areas (25%). Large differences can also be seen between spatial units when
considering innovation output, with the commuter area being the less favoured. These results
highlight the impact of knowledge externalities, agglomeration economies and spatial
selection on innovation commitment and innovation output.

Table 1: Propensity to innovate and innovation output across functional regions

Luxembourg-
City (%)
Luxembourg
Urban Area
(%)
Suburban
Area (%)
South Area
(%)
Commuter
Area (%)
Propensity to innovate 60 47 52 31 25
Turnover from innovative
products
4.9 3.9 5.6 3.9 2.0
No. of observations 182 74 134 86 86
Source: CIS 2006. Authors‟ calculations.

10
In order to examine the specific impact of learning, the next section examines firms‟
characteristics across functional regions. Firms‟ profiles, which differ across regions, are
likely to explain most of these discrepancies.

4.1 Firms’ profiles across functional regions

Firms from Luxembourg-City have a very specific profile: they are highly specialised in
financial activities (38.1% of them) and IT services (13.4%), operate predominantly in
international markets (65.8%), and belong mainly to European groups (42.1%). The level of
education of employees is also higher than in other functional regions. Finally, large firms are
more numerous in Luxembourg-City than in the other functional units (Table 2).

Table 2: Firms’ profiles across functional regions

Luxembourg-
City (%)
Luxembourg
Urban (%)
Suburban
Area (%)
South Area
(%)
Commuter
Area (%)
Economic activity
High and medium high-tech 1.9 5.6 6.4 5.1 4.1
Medium low-tech 3.5 4.8 9.9 20.7 8.9
Low-tech 3.2 14.3 4.8 16.9 15.8
Gas and electricity 0.8 0.4 0.7 1.5 1.4
Wholesale and commission
trade
15.3 15.7 20.0 21.5 29.8
Transport and communication 14.4 16.5 25.5 24.5 30.7
Financial intermediation 38.1 27.6 14.7 0.0 3.0
Computer activities 13.4 11.3 11.3 8.0 0.0
R&D – Engineering and
consultancy – Technical
testing and analysis
9.5 3.8 6.8 1.9 6.4
Group membership
No 39.3 43.3 54.4 63.6 65.7
National group 13.4 6.9 17.9 15.8 17.0
European group 42.1 35.8 19.9 18.3 15.2
Non-European group 5.2 14.0 7.8 2.4 2.1
Size
Small (10-49 employees) 69.4 72.4 73.3 74.1 80.7
Medium (50-249 employees) 23.1 21.9 20.9 22.0 18.1
Large (>249 employees) 7.5 5.8 5.8 3.9 1.2
Market mainly foreign
No 34.2 30.9 35.9 58.4 39.3
Yes 65.8 69.1 64.1 41.6 60.7
% of employees with higher
education
43.0 36.1 35.1 23.2 23.2

No. of observations

182

74

134

86

86
Source: CIS 2006. Authors‟ calculations.

Firms from the Urban and Suburban Areas are less specialised in financial activities (27.6%
and 14.7%) than those in Luxembourg-City. They are more oriented than the latter towards

11
high and medium high-tech2 (5.6% and 6.4%). Some of them are large (5.6%), but this is
fewer than in Luxembourg-City. Most of them operate mainly in international markets (69.1%
and 64.1%). Firms from the Suburban Area also often belong to European groups (35.8%). By
contrast, firms from the South and Commuter Areas are unlikely to be large (3.9% and 1.2%)
or belong to a group (36.4% and 34.3%). These firms often operate in transport (24.5% and
30.7%), wholesale trade (21.5% and 29.8%) or low-tech manufacturing.

Firms‟ profiles appear therefore to differ substantially across functional units. These profiles
favour the propensity to innovate in Luxembourg-City and, to a lesser extent, in the Urban
and Suburban Areas. At the opposite end of the scale, the propensity to innovate is lower in
the South and Commuter Areas. These results confirm previous studies which show that
Luxembourg national territory is very heterogeneous despite its small size, with Luxembourg-
City and Urban Area being the main location for knowledge-intensive and high-tech
industries (Sohn and Walther 2008, Walther and Dautel 2010).

4.2 Accessibility according to functional region

As we analyse functional units and accessibility, it is highly important to examine to what
extent these two variables are related. In order to examine this issue we consider, using a
kernel density function, firms‟ accessibility density by functional region. As expected, the
accessibility density across functional regions (Figure 1) shows that Luxembourg-City has on
the whole the best accessibility to the centre of gravity, followed by the Urban, Suburban and
South Areas.

Figure 1: Accessibility density across functional regions

Source: CIS2006; Authors‟ calculation.

The results also show that some of these densities are bimodal, especially for Luxembourg-
City and the Commuter Area. Considering these, it appears that the second mode for
Luxembourg-City (13 minutes) does not differ much from the mode for the Urban Area (14
minutes). It appears also that the first mode for the Commuter Area (29 minutes) does not

2 The distinction between high and medium high-tech, medium low-tech, and low-tech is based on an OECD-
Eurostat knowledge-based classification (see OECD 2006 and Eurostat 2006).

12
differ much from the mode for the South Area (23 minutes). The accessibility exam will
therefore to some extent mix firms from different functional units. It must also be noted that
most of the firms have adequate accessibility to the centre: 94% of them needed less than 40
minutes to access the centre.

5. Empirical results

5.1. Econometric approach

In line with the aim of the article, we use the following model:

Y= f(X, loc, u)

where Y represents the propensity to innovate or innovation output, X the independent or
control variables, „loc‟ the functional regions or the accessibility measure and „u‟ the
unobserved effects or measurements errors. The functional regions are dummy variables and
accessibility, calculated by the number of minutes to access the centre of gravity of all firms
located in Luxembourg, is a discrete variable. The independent variables are firm sector, firm
size, belonging to a group, main geographical market and percentage of employees with
higher education. The two control variables refer to the establishment of the firm during the
reference period and an increase in turnover of at least 10% due to a merger.

This model is estimated by two types of functions according to the dependent variable, i.e.
probit for propensity to innovate and tobit for innovation output (including a logistic
transformation of the dependent variable). Indeed, as some firms did not introduce new or
improved products to the market, their innovation output are left-censored. We take this
restriction into account by estimating tobit models3 (see Greene 1997 or Gourieroux 2000).

Based on this model, we compute predicted probabilities (propensity to innovate) or linear
predictions (innovation output) in order to estimate profile effects and agglomeration effects.
These are computed according to „average‟ enterprise characteristics, which are firms with the
average characteristics of the overall economy (i); or firms with the average characteristics of
a given functional region or a given accessibility (j). The effect of a given spatial unit or given
accessibility is also considered (loc).

More precisely, our predicted probabilities or linear predictions are computed as follows:

- Y1
*
corresponds to the expected propensity to innovate or the expected innovation output for
a firm with the overall average characteristics.
- Y1
*
loc is the expected result for a firm with the average characteristics of a given spatial unit
or accessibility.
- Y2
*
loc is the expected result for a firm with the average characteristics of a given spatial unit
or accessibility and including the spatial unit or accessibility effect.

Based on these results, profile and agglomeration effects are estimated as follows:

(1) Profile_effect=Y1
*
loc /Y1
* with Y1
*= a.Xi ; Y1
*
loc =a.Xloc,j


3 It has to be noticed that the presence of heteroscedasticity has to be rejected for the overall tobit model. LR
Chi2 (9) = 3.58; Prob> Chi2 = 0.937.

13
(2) Agglo_effect=Y2
*
loc/Y1
*
loc with Y2
*
loc =a2.Xloc, j + b.locj

The accessibility estimations require, however, additional specifications due to the small
number of enterprises at some levels of accessibility. Firstly, we retain a „sliding bandwidth‟
of five minutes for estimating Y*. Secondly, we smooth (1) and (2) by regressing these
measures on accessibility using a local polynomial regression (see Fan and Gijbels 1996). The
local polynomial regression offers, in addition, a way to reduce the inconsistency resulting
from the eventual presence of heteroscedasticity in the tobit model estimations. The influence
of outliers is indeed downweighted in the estimation of a local polynomial regression. The
estimates derived from these additional specifications are represented in Figures 3 and 5.

5.2 Propensity to innovate by functional region and accessibility

Controlling for size, sectors, corporate structure, market, human capital and additional factors,
the propensity to innovate differs across functional units, as shown by Table 3. Firms located
in the South and Commuter Areas, and to a lesser extent in the Urban Area, appear less likely
to innovate than those in Luxembourg-City. Distance to the centre of gravity, which acts as a
proxy for accessibility, also appears to decrease the propensity to innovate. These two results
suggest, therefore, the presence of agglomeration effects. However, firms from the Suburban
Area do not appear to differ significantly from those in Luxembourg-City.

Table 3: Probit estimates
Propensity to innovate Functional regions Accessibility
Coef. P>z Coef. P>z
Minutes - - -0.014 0.01 ***
Urban Area -0.318 0.09 * - -
Suburban Area -0.094 0.561 - -
South Area -0.611 0.002 *** - -
Commuter Area -0.348 0.074 * - -
High and medium high-tech 1.000 0 *** 0.944 0.001 ***
Medium low-tech 0.341 0.199 0.268 0.303
Low-tech 0.353 0.211 0.305 0.276
Wholesale and commission trade 0.193 0.444 0.175 0.484
Transport and communication -0.332 0.196 -0.319 0.212
Financial intermediation 0.254 0.327 0.225 0.382
Computer activities -0.015 0.957 -0.062 0.817
Size 0.257 0 *** 0.253 0 ***
Group membership 0.207 0.114 0.208 0.111
International market 0.179 0.188 0.233 0.084 *
% of employees with higher education 0.861 0.001 *** 0.858 0.001 ***
Creation in 2004-2006 0.207 0.459 0.266 0.345
Merger 0.640 0.057 * 0.673 0.052 *
constant -1.428 0 *** -1.356 0 ***
Pseudo R2 0.178 0.168
No. of observations 562 562
*, **, ***: statistically significant at the 0.1, 0.5, 0.01 level.
Source: CIS2006. Authors‟ calculations.

14
As expected, other aspects, which were found not to be randomly localised (see 4.1), also
impact the propensity to innovate: firm size, belonging to a group, employees‟ level of
education, and operating in high and medium high-tech industries increase the propensity to
innovate in both regressions. In addition, the fact that a firm operates mainly in foreign
markets increases its propensity to innovate in the second estimation. It is therefore expected
that profile effects operate.

But what about the size of both these effects across functional regions? The calculation of
ratios described in the previous section, based on predicted probabilities or linear predictions,
allows this to be done. The results show close links between profile and agglomeration
effects. Agglomeration effects appear favouring innovation commitment in Luxembourg-City
and the Suburban Area in comparison to the South and Commuter Areas (Figure 2). In
addition, profile effects benefit firms from Luxembourg-City, the Luxembourg Urban Area
and the Suburban Area in comparison to firms from the South and Commuter Areas.

This result is in line with previous findings for larger spatial units. Johansson and Lööf
(2008), for example, have found that, after controlling for firm profile, location of the firm
across Swedish regions impacts upon innovation propensity, favouring firms from the
metropolitan Stockholm region. Their descriptive results also show that firms from the
Stockholm region benefit from characteristics advantageous to innovation.

As noted above, such an approach assumes that location characteristics affect all firms in a
given functional unit in the same way. In order to overcome this limitation, accessibility was
examined.

Figure 2: Profile and agglomeration effects by functional region

Source: CIS2006; Authors‟ calculation.

The accessibility results show lower profile effect for firms very close to the centre of gravity
but still mainly operating in Luxembourg-City (Figure 3). In addition, higher profile effects,
in comparison to nearby firms, are found for firms 40 minutes from the gravity centre, and
lower effects for those 55 minutes away. In both of the latter cases, firms are mainly operating
in the South and Commuter Areas. This suggests, therefore, some diversity within functional
units regarding firms‟ profile effects. Some diversity is also found in terms of agglomeration

15
effects, especially between firms from Luxembourg City. Our results indicate that
agglomeration effects increase with firm density in Luxembourg-City, and decrease with
accessibility. Some variability is also found among the small number of firms with low
accessibility. It must however be noted that this latter variability may be due to the small
number of observations considered.

Figure 3: Profile and agglomeration effects according to accessibility

Source: CIS2006; Authors‟ calculation

5.3 Output of innovation activities according to functional region and accessibility.

Considered by functional unit, the estimates are of lower innovation output for firms in the
South and Commuter Areas than for firms in Luxembourg-City (Table 4). Accessibility
estimates show a negative impact of minutes on innovation output. Both results suggest that
agglomeration effects are operating and that they decrease with distance.

Other variables impact positively innovation output: firm size, membership of a group,
employees‟ level of education, operating mainly in international markets and belonging to
high and medium high-tech or medium low tech industries. It is therefore expected that
profile effects will be seen.

Table 4: Tobit estimates
% of turnover from new
products
Functional regions Accessibility
Coef. P>t Coef. P>t
Minutes - - -0.025 0.09
Urban Area -0.595 0.218 - -
Suburban Area -0.361 0.378 - -
South Area -1.260 0.019 ** - -
Commuter Area -0.858 0.105 - -
High and medium high-tech 2.024 0.005 *** 2.010 0.005 ***
Medium low-tech 1.333 0.07 * 1.165 0.111
Low-tech 1.137 0.148 1.057 0.178

16
Transport and communication 0.877 0.218 0.822 0.249
Financial intermediation -0.409 0.573 -0.408 0.575
Computer activities 0.951 0.15 0.944 0.154
Transport and communication 0.337 0.645 0.281 0.701
Size 0.325 0.015 ** 0.339 0.011 **
Group membership 0.606 0.087 * 0.632 0.075 *
International market 0.974 0.008 *** 1.056 0.004 ***
% higher educated employees 1.720 0.006 *** 1.786 0.005 ***
Creation in 2004-2006 -0.717 0.37 -0.602 0.454
Merger -0.199 0.781 0.014 0.985
constant -8.801 0 *** -8.883 0 ***
Log likelihood -642.74 -644.06
No. of observations
(uncensored)
188 188
*, **, ***: statistically significant at the 0.1, 0.5 and 0.01 level.
Source: CIS2006. Authors‟ calculations.

The estimation of ratios described in 4.3 and presented in Figure 4 highlights the impact of
profile and agglomeration effects on innovation output. Specifically, agglomeration effects
tend to be lower in the Urban Area than in Luxembourg-City or the Suburban Area, and lower
in the South than in the more remote Commuter Area. Profile effects decrease more linearly
across functional units. Once more, our results suggest a close link between agglomeration
and profile effects. However, agglomeration effects seem to be higher than profile effects in
Luxembourg-City, the Suburban Area and the Commuter Area.

Figure 4: Profile and agglomeration effects by functional region

Source: CIS2006; Authors‟ calculation

The examination of accessibility shows higher variability of agglomeration effects when
considering innovation output (Table 5). However, due to firms‟ density close to the centre of
gravity, numerous firms benefit from high agglomeration effects. Profile effects appear to be
high only for firms even closer to the centre.

17
In order to fine-tune these results, we have estimated them once again removing firms from
given functional regions. The resultant findings show clearly that the second mode of
agglomeration effects (at around 30 minutes to the centre of gravity) depends from the
suburban area firms. This result suggests, therefore, that part of this area could offer more
favourable conditions for innovation than other locations with identical accessibility to the
centre.

Figure 5: Profile and agglomeration effects according to accessibility

Source: CIS2006; Authors‟ calculation

6. Conclusion

Our analysis of the geography of innovation within the Luxembourg metropolitan region
suggests that innovation varies at the intra-regional level.

With regard to functional regions, firms from Luxembourg-City appear to innovate more
often than those from other regions due to their own characteristics favouring innovation and
the existence of positive externalities. Firms from the Suburban Area take also advantage of
these effects, but to a lesser degree than firms from Luxembourg-City. At the opposite
extreme, firms from the South and Commuter Areas are less likely to innovate, with both of
these effects becoming negative. Our analysis of innovation output also highlights the
presence of these effects, which are found to vary to an even greater extent. Both of these
results provide first evidence of a close link between profile and agglomeration effects at the
intra-regional level.

With regard to accessibility, our results provide further evidence of the variability of
innovation at the intra-regional level, suggesting that profile and agglomeration effects may
differ within the functional units. The accessibility results in relation to the propensity to
innovate show, however, the organisation of innovation activities around one main centre,
located in Luxembourg-City close to the centre of gravity of the spatial distribution of firms.
This finding is less clear when innovation output is considered.

Our results suggest several directions for future empirical research on innovation within the
Luxembourg metropolitan region. Firstly, this paper, which highlights the importance of

18
agglomeration effects in the innovation process, does not provide clear empirical evidence
concerning their origins. Further analysis is therefore required of the spatial determinants of
innovation. This is a complex issue, as these determinants are likely to differ according to
location (Autant-Bernard 2001).

Secondly, our empirical evidence is, due to the availability of data, based on average firms,
which does not allow us to take into account the fact that firms from a given location might
belong to dissimilar clusters of activities. Since local spillovers are conditioned by
technological proximity, firms are likely to benefit in different ways from local externalities.
This is a particularly important issue in Luxembourg, as the country is currently creating a
new science city, Belval, which is expected to combine in a single area most of the national
research and higher education institutions with mostly low-tech or low knowledge-intensive
incumbent firms. The expected relocation in Belval of some national R&D providers and the
attraction of foreign firms, at least from bordering regions, should reinforce this issue.

Thirdly, as our models are static, our agglomeration effects are subject to endogeneity due to
firms with unobserved attributes. This means that if e.g. firms with overachieving
entrepreneurs would be disproportionately found in a specific area (Rosenthal and Strange
2003) such as Luxembourg-City, our agglomeration effects would be overestimated for this
area. This issue has to be resolved through a more dynamic approach.

Fourthly, as indicated in previous studies (Walther and Dautel 2010), Luxembourg is
undergoing a “metropolisation process” characterised by decreasing employment growth in
Luxembourg-City and increasing growth within the close periphery. More research could be
done to link these findings to those of this paper, in order to compare the geography of
employment with the geography of innovation and give a more complete picture of the
complex organisation of economic activities at the intra-regional level. Luxembourg-City
favours firms‟ innovation activity, while the Urban and Suburban Areas are the main
recipients of employment growth. Further analysis is required to investigate this important
topic.

19
References

Agrawal A. 2002. Innovation, Growth Theory and the Role of Knowledge Spillovers.
Innovation Analysis Bulletin 4(3): 3–6.
Almedia P, Kogut B. 1997. The Exploration of Technological Diversity and the Geographic
Localisation of Innovation. Small Business Economics 9: 21–31.
Anderson R, Quigley JM, Wilhelmson M. 2004. University Decentralization as Regional
Policy: The Swedish Experiment. Journal of Economic Geography 4(4): 371–388.
Andersson M, Karlsson C. 2004. The Role of Accessibility for the Performance of Regional
Innovation Systems. CESIS Electronic Working Paper 9.
Andersson M, Lööf H. 2009. Agglomeration and Productivity: Evidence from Firm-level
Data. The Annals of Regional Science 46(3): 601–620.
Anselin L, Varga A, Acs Z. 1997. Local Geographic Spillovers Between University Research
and High Technology Innovations. Journal of Urban Economics 42: 422–448.
Antonietti R, Cainelli G. 2009. The Role of Spatial Agglomeration in a Structural Model of
Innovation, Productivity and Export: A Firm-level Analysis. The Annals of Regional
Science 46(3): 577–600.
Arauzo-Carod J-M, Viladecans-Marsal E. 2009. Industrial Location at the Intra-Metropolitan
Level: The Role of Agglomeration Economies. Regional Studies 43(4): 545–558.
Arvanitis S, Hollenstein H. 1994. Demand and Supply Factors in Explaining the Innovative
Activity of Swiss Manufacturing Firms. Economics of Innovation and New Technology
3(1): 15–30.
Arzaghi M, Henderson JV. 2008. Networking Off Madison Avenue. The Review of Economic
Studies 75(4): 1011–1038.
Audretsch DB, Dohse D. 2007. Location: A Neglected Determinant of Firm Growth. Review
of World Economics 143(1): 79–107.
Audretsch DB, Feldman MP. 1996. R&D Spillovers and the Geography of Innovation and
Production. American Economic Review 86(3): 630–640.
Audretsch DB, Feldman MP. 2004. Knowledge Spillovers and the Geography of Innovation,
in Henderson V, Thisse J.-F (eds) Handbook of Regional and Urban Economics 4.
Amsterdam, Elsevier: 2713–2739.
Audretsch DB, Lehmann EE, Warning S. 2003. University Spillovers: Strategic Location and
New Firm Performance. London, Centre for Economic Policy Research, Discussion Paper
Series 3837.
Autant-Bernard C. 2001. The Geography of Knowledge Spillovers and Technological
Proximity. Economic of Innovation and New Technology 10: 237–254.
Baldwin R, Okubo T. 2006. Heterogeneous Firms, Agglomeration and Economic Geography:
Spatial Selection and Sorting. Journal of Economic Geography 6: 323–346.
Black G. 2005. The Geography of Small Firm Innovation. Boston, Kluwer Academic.
Broekel T, Brenner T. 2010. Regional Factors and Innovativeness: An Empirical Analysis of
Four German industries. The Annals of Regional Science 47(1): 169–174.
Camagni R, Capello R. 2004. The City Network Paradigm: Theory and Empirical Evidence,
in Capello R, Nijkamp P (eds) Urban Dynamics and Growth. Amsterdam, Elsevier: 495–
529.
Cheshire P, Gordon IR. 1998. Territorial Competition: Some Lessons for Policy. The Annals
of Regional Science 32: 321–346.
Cockburn I, Henderson R. 1998. Absorptive Capacity, Coauthoring Behavior, and the
Organization of Research in Drug Discovery. The Journal of Industrial Economics 46(2):
157–182.

20
Cohen WM, Levinthal DA. 1989. Innovation and Learning: The Two Faces of R&D.
Economic Journal 99: 569–596.
Cohen WM, Levinthal DA. 1990. Absorptive Capacity: A New Perspective on Learning and
Innovation. Administrative Science Quarterly 35: 128–152.
Combes PP, Duranton G, Gobillon L, Puga D, Roux S. 2009. The Productivity Advantages of
Large Cities: Distinguishing Agglomeration from Firm Selection. London, Centre for
Economic Policy Research, Discussion Paper 7191.
Copus A, Skuras D, Tsegenidi K. 2008. Innovation and Peripherality: An Empirical
Comparative Study of SMEs in Six European Union Member Countries. Economic
Geography 84(1): 51–82.
Czarnitzki D, Hottenrott H. 2009. Are Local Milieus the Key to Innovation Performance?
Journal of Regional Science 49(1): 81–112.
Dautel V. 2008. Research and Development Activities and Innovative Performance of Firms
in Luxembourg, in van Geenhuizen M, Trzmielak D, Urbaniak M (eds) Value Added
Partnering and Innovation in a Changing World. West Lafayette, Purdue University Press:
297–314.
Dautel V. 2010. Profil de l‟entreprise, stratégie d‟innovation et conditions de marché.
Luxembourg, CEPS/INSTEAD Cahier 13.
Dautel V, Walther O. 2009. Croissance de l‟emploi et mobilité des entreprises au
Luxembourg (1994-2005). Luxembourg, CEPS/INSTEAD Population & Territoire 15.
Devereux M, Griffith R, Simpson H. 2007. Firm Location Decisions, Regional Grants and
Agglomeration Externalities. Journal of Public Economy 91(3–4): 413–435.
Dosi G. 1988. The Nature of the Innovative Process, in Dosi G, Nelson R, Silverberg G,
Freeman C, Soete L. (eds) Technical Change and Economic Theory. London, Pinter: 221–
238.
Duranton G, Overman, H. 2006. Exploring the Detailed Location Patterns of UK
Manufacturing Industries Using Microgeographic Data. London, Centre for Economic
Policy Research, Discussion Paper 5858.
Duranton G, Puga D. 2001. Nursery Cities: Urban Diversity, Process Innovation, and the
Life-cycle of Products. American Economic Review 91(5): 1454–1477.
Eurostat 2006. High-technology Manufacturing and Knowledge-intensive Services Sectors.
Luxembourg, Eurostat.
Fan J, Gijbels I. 1996. Local Polynomial Modeling and Its Applications. London, Chapman
and Hall.
Fagerberg J. 2005. Innovation: A Guide to the Literature, in Fagerberg J, Mowery DC, Nelson
RR (eds) The Oxford Handbook of Innovation. Oxford, Oxford University Press: 1–26.
Feldman MP. 1994a. Knowledge Complementarity and Innovation. Small Business
Economics 6: 363–72.
Feldman MP. 1994b. The Geography of Innovation. Boston, Kluwer Academic.
Feldman MP, Audretsch DB. 1999. Innovation in Cities: Science-based Diversity,
Specialization and Localized Competition. European Economic Review 43(2): 409–429.
Fischer M, Varga A. 2003. Spatial Knowledge Spillovers and University Research: Evidence
from Austria. The Annals of Regional Science 37(2): 303–322.
Frenkel A. 2001. Can Regional Policy Affect Firms‟ Innovation Potential in Lagging
Regions? The Annals of Regional Science 34(3): 315–341.
Glaeser EL. 1999. Learning in Cities. Journal of Urban Economics 46(2): 254–277.
Gordon IR, McCann P. 2000. Industrial Clusters: Complexes, Agglomerations and/or Social
networks? Urban Studies 37(3): 513–532.
Gössling T, Rutten, R. 2007. Innovation in Regions. European Planning Studies 15(2): 253–
270.

21
Gourieroux C. 2000. Econometrics of Qualitative Dependent Variables. Cambridge,
Cambridge University Press.
Greene W H. 1997. Econometric Analysis. New York, Macmillan.
Heidenreich M. 2009. Innovation Patterns and Location of European Low- and Medium-
Technology Industries. Research Policy 38: 483–494.
Huffman D, Quigley M. 2002. The Role of the University in Attracting High-tech
Entrepreneurship: A Silicon Valley Tale. The Annals of Regional Science 36(3): 403–419.
Hymer S. 1976. The International Operations of National Firms: A Study of Direct
Investment. Cambridge, The MIT Press.
Jaffe AM, Trajtenberg M, Henderson R. 1993. Geographic Localization of Knowledge
Spillovers as Evidenced by Patent Citations. Quarterly Journal of Economics 63: 577–598.
Johansson B, Klaesson J, Olsson M. 2002. Time Distances and Labor Market Integration.
Papers in Regional Science 81: 305–327.
Johansson B, Lööf H. 2008. Innovation Activities Explained by Firms Attributes and
Location. Economics of Innovation and New Technology 17(6): 533–552.
Karlsson C, Manduchi A. 2001. Knowledge Spillovers in a Spatial Context – A Critical
Review and Assessment, in Fischer MM, and Fröhlich J. (eds) Knowledge, Complexity and
Innovation Systems. Berlin, Springer: 101–123.
Kaufmann A. 2007. Patterns of Innovation Relations in Metropolitan Regions: The Case of
the Vienna Urban Region. The Annals of Regional Science 41: 735–748.
Krätke S. 2007. Metropolisation of the European Economic Territory as a Consequence of
Increasing Specialisation of Urban Agglomerations in the Knowledge Economy. European
Planning Studies 15(1): 1–27.
Krugman P. 1991. Increasing Returns to Scale and Economic Geography. Journal of Political
Economy 99(3): 483–499.
Lane PJ, Koka BR, Pathak S. 2006. The Reification of Absorptive Capacity: A Critical
Review and Rejuvenation of the Construct. Academy of Management Review 31(4): 833–
863.
Lejpras A, Stephan A. 2009. Locational Conditions, Cooperation, and Innovativeness:
Evidence from Research and Company Spin-offs. The Annals of Regional Science 46(3):
543–575.
Melitz MJ, Ottaviano GIP. 2008. Market Size, Trade, and Productivity. Review of Economic
Studies 75(1): 295–316.
Nelson RR, Wolff EN. 1997. Factors Behind Cross-industry Differences in Technical
Progress. Structural Change and Economic Dynamics 8(2): 205–220.
OECD 2006. Innovation and Knowledge-Intensive Service Activities. Paris, Organisation for
Economic Co-operation and Development.
OECD and Eurostat. 2005. Oslo Manual. Guidelines for Collecting and Interpreting
innovation data. Paris, OECD and Luxembourg, Eurostat.
Oort F van. 2002. Innovation and Agglomeration Economies in the Netherlands. Tijdschrift
voor Economische en Sociale Geografie 93(3): 344–360.
Orlando M. 2000. On the Importance of Geographic and Technological Proximity for R&D
Spillovers: An Empirical Investigation. Research Division, Federal Reserve Bank of
Kansas City.
Patel P, Pavitt K. 1995. Patterns of Technological Activity: their Measurement and
Interpretation, in Stoneman P. (ed.) Handbook of the Economics of Innovation and
Technological Change. Oxford, Blackwell: 14–52.
Pieretti P, Bourgain A, Courtin P. 2007. Place financière de Luxembourg. Analyse des
sources de ses avantages compétitifs et de sa dynamique. Brussels, De Boeck.
Porter ME. 1990. The Competitive Advantage of Nations. London, Macmillan.

22
Powell W, Koput K, Smith-Doerr L. 1996. Interorganizational Collaboration and the Locus of
Innovation: Networks of Learning in Biotechnology. Administrative Science Quarterly
41(1): 116–145.
Rosenthal SS, Strange WC. 2003. Evidence on the Nature and Sources of Agglomeration
Economies, in Henderson J.V, Thisse J.-F. (eds) The Handbook of Regional and Urban
Economics 4. Amsterdam, Elsevier: 2119–2172
Rosenthal SS, Strange WC. 2005. The Geography of Entrepreneurship in the New York
Metropolitan Area. New York, Federal Reserve Bank, Economic Policy Review: 29–53.
Schulz C, Walther O. 2010. Adaptation Strategies of Luxembourg‟s Financial Centre Under
Pressure. Luxembourg. Luxembourg, CEPS/INSTEAD Working Paper 26.
Shefer D, Frenkel A. 1998. Local Milieu and Innovations: Some Empirical Results. The
Annals of Regional Science 32(1): 185–200.
Simmie J. 2003. Innovation and Urban Regions as National and International Nodes for the
Transfer and Sharing of Knowledge. Regional Studies 37(6-7): 607–620.
Smith V, Broberg A, Overgaard J. 2002. Does Location Matter for Firms‟ R&D Behaviour?
Empirical Evidence for Danish Firms. Regional Studies 36(8): 825–832.
Sohn C, Walther O. 2008. Géographie des activités économiques métropolitaines au
Luxembourg. Luxembourg, CEPS/INSTEAD Population & Territoire 14.
Sternberg R, Arndt O. 2001. The Firm or the Region: What Determines the Innovation
Behavior of European firms? Economic Geography 77: 364–382.
Teirlinck P, Spithoven A. 2008. The Spatial Organization of Innovation: Open Innovation,
External Knowledge Relations and Urban Structure. Regional Studies 42(5): 689–704.
Thierstein A, Lüthi S, Kruse C, Gabi S, Glanzmann L. 2008. Changing Value Chain of the
Swiss Knowledge Economy: Spatial Impact of Intra-firm and Inter-firm Networks within
the Emerging Mega-City Region of Northern Switzerland. Regional Studies 42(8): 1113–
1131.
Varga A. 2000. Local Academic Knowledge Spillovers and the Concentration of Economic
Activity. Journal of Regional Science 40(2): 289–309.
Veugelers R, Cassiman B. 1999. Make and Buy in Innovation Strategies: Evidence from
Belgian Manufacturing Firms. Research Policy 28(1): 63–80.
Walther O. 2011. Regional Growth of the Knowledge Economy in Luxembourg, in Sohn C.
(ed.) Luxembourg: An Emerging Cross-border Metropolitan Region. Brussels, Peter Lang
(in press).
Walther O, Dautel V. 2010. Intra-regional Employment Growth in Luxembourg (1994-2005).
Geografiska Annaler: Series B, Human Geography 92(1): 45–63.
Walther O, Schulz C. 2009. Finanzplatz Luxemburg – vom „Steuerparadies‟ zur
Investmentfonds-Kapitale. Geographische Rundschau 61(1): 30–35.

3, avenue de la Fonte
L-4364 Esch-sur-Alzette
Tél.: +352 58.58.55-801
www.ceps.lu