is innovation in cities a matter of knowledge intensive services

25
This article was downloaded by: [Universiti Teknologi Malaysia] On: 04 January 2014, At: 19:39 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Innovation: The European Journal of Social Science Research Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ciej20 Is innovation in cities a matter of knowledge-intensive services? An empirical investigation Roberta Capello a , Andrea Caragliu a & Camilla Lenzi a a Department of Building, Environment, Science and Technology , Politecnico di Milano , Piazza Leonardo 32, Milan , 20133 , Italy Published online: 19 Apr 2012. To cite this article: Roberta Capello , Andrea Caragliu & Camilla Lenzi (2012) Is innovation in cities a matter of knowledge-intensive services? An empirical investigation, Innovation: The European Journal of Social Science Research, 25:2, 151-174, DOI: 10.1080/13511610.2012.660326 To link to this article: http://dx.doi.org/10.1080/13511610.2012.660326 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms- and-conditions

Upload: siti-khatizah-aziz

Post on 17-Aug-2015

135 views

Category:

Leadership & Management


0 download

TRANSCRIPT

This article was downloaded by: [Universiti Teknologi Malaysia]On: 04 January 2014, At: 19:39Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Innovation: The European Journal ofSocial Science ResearchPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/ciej20

Is innovation in cities a matter ofknowledge-intensive services? Anempirical investigationRoberta Capello a , Andrea Caragliu a & Camilla Lenzi aa Department of Building, Environment, Science and Technology ,Politecnico di Milano , Piazza Leonardo 32, Milan , 20133 , ItalyPublished online: 19 Apr 2012.

To cite this article: Roberta Capello , Andrea Caragliu & Camilla Lenzi (2012) Is innovation in citiesa matter of knowledge-intensive services? An empirical investigation, Innovation: The EuropeanJournal of Social Science Research, 25:2, 151-174, DOI: 10.1080/13511610.2012.660326

To link to this article: http://dx.doi.org/10.1080/13511610.2012.660326

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Is innovation in cities a matter of knowledge-intensive services?An empirical investigation

Roberta Capello*, Andrea Caragliu and Camilla Lenzi

Department of Building, Environment, Science and Technology, Politecnico di Milano, PiazzaLeonardo 32, 20133 Milan, Italy

(Received 15 March 2011; final version received 10 October 2011)

The presence of large cities in a region represents a potential for regionalinnovation capacity: cities are in fact expected to generate dynamic agglomerationeconomies and knowledge spillovers. The paper adds to previous analyses onthis topic by investigating whether the linkage between the presence of cities inthe region and the innovative performance is mediated by the urban industrialstructure. In fact, a positive correlation is likely to exist between the presenceof large cities in a region and its innovative performance. Such a relationshipcould also depend on the presence of knowledge-intensive service, rather than onadvanced manufacturing activities. In order to verify this statement, we classifyEuropean NUTS2 regions both from an industrial perspective, as well as byspatial typologies. We integrate this classification with a novel data set on regionalinnovation, based on the Community Innovation Survey. On this basis, geo-graphical and descriptive analyses of regional innovation patterns are developedand explained. The descriptive results support our expectations. Regions host-ing large urban areas are the most innovative, and this statement is reinforcedin regions characterized by specialization in knowledge-intensive services. Thesimultaneous presence of advanced manufacturing and knowledge-intensiveservice activities generates synergic effects, fostering innovative performance.

Keywords: innovation; cities; knowledge-intensive services

Introduction

Technologically advanced and science-based sectors represent a major driver of

economic development. New jobs are expected to arise mainly in these new sectors,

while more traditional sectors are subject to restructuring or off-shoring, potentially

causing serious tensions in local labor markets.This stylized fact is certainly not new in the literature and links back to the

traditional and oldest theory explaining innovation through the presence of ‘‘science-

based’’ (Pavitt 1984) or high-technology sectors; regions hosting these sectors were

considered as ‘‘advanced’’ regions, leading the transformation of the economy.

Despite its early birth, this view is still well maintained in scientific discourse, and

it still inspires policy debate, as attested by the inclusion of employment data in high-

tech sectors in score boarding and benchmarking exercises aimed at measuring

innovation performance, also at the regional level (see among the many contributions

*Corresponding author. Email: [email protected]

Innovation � The European Journal of Social Science Research

Vol. 25, No. 2, June 2012, 151�174

ISSN 1351-1610 print/ISSN 1469-8412 online

# 2012 ICCR Foundation

http://dx.doi.org/10.1080/13511610.2012.660326

http://www.tandfonline.com

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

the Regional Competitiveness Index, 2010, the Regional and European Innovation

Scoreboard, 2009, and the 5th Cohesion Report).

The focus of studies on innovation has progressively shifted from the sector-

based approach towards what has been called a function-based approach (Camagni

and Capello 2009), which stresses the importance of pervasive and horizontal

functions like R&D and higher education as the main determinants of innovation

capability. ‘‘Scientific’’ regions, hosting large and well-known scientific institutions,

were studied deeply and relationships between these institutions and the industrial

fabric were analyzed, with some disappointment as far as an expected but not often

visible direct linkage was concerned (MacDonald 1987, Massey et al. 1992, Monk

et al. 1988, Storey and Tether 1998). Indicators of R&D inputs (like public and

private research investment and personnel) and increasingly indicators of R&D

output (like patenting activities) were used in order to understand the engagement of

firms and territories on knowledge, intended as a necessary long-term precondition

for continuing innovation (Dasgupta and Stiglitz 1980, Antonelli 1989, Griliches

1990). The most recent step forward made in the literature was including nonmaterial

drivers of innovation as an explanation of innovation capacity, such as in the

learning region theory (Lundvall and Johnson 1994) and the milieu innovateur

approach (Aydalot 1986, Aydalot and Keeble 1988, Camagni 1991).

At the urban level, the same conceptual development in the theoretical iden-

tification of urban innovative determinants has taken place, and for this reason

the industrial structure of urban economies has recently been a relatively neglected

factor of urban innovation. However, the recent idea of ‘‘smart cities’’, generally

referring to the presence of e-economy and e-society, calls once again for a sector-

based approach to explain urban efficiency. An e-society or e-economy, in fact,

implies the presence (and the innovative use) of knowledge-intensive services (KIS)

in cities, generating higher efficiency rates. Addressing the sector based-approach at

the urban scale is also of interest since many of the studies undertaken in this frame

have focused on manufacturing clusters rather than urban agglomerations.

In this paper we claim that an industrial approach is still worth analyzing.

Modern urban economies may benefit from a favorable industrial mix, supporting

innovation; in particular, we expect that metropolitan regions hosting knowledge-

intensive activities will generate higher innovation performance. The capacity to

innovate is expected to be even stronger as metropolitan areas host a combination of

advanced manufacturing and service activities, that synergically foster the emergence

of a creative atmosphere.

Limited evidence is currently available at the European regional and urban scales

on the interplay between high-tech manufacturing, advanced services and innovation

activities. So far, a lack of data on innovation processes and different types of

innovations at the regional scale has unfortunately impaired similar exercises. Also,

this limitation is even more pronounced when the regional settlement structure

has to be introduced as a further dimension of analysis. The present paper moves

in this direction, by providing a systematic empirical analysis at regional level on

31 European countries. Our contribution is therefore mainly on empirical grounds

and based on descriptive evidence drawn from an original dataset covering EU27

plus European Free Trade Association (henceforth, EFTA) countries combining

employment data in high-tech manufacturing and services derived from EUROSTAT

with a novel data set on different types of innovation. This data set has been built by

152 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

the authors on the basis of data from the Community Innovation Survey (CIS)

EUROSTAT database.

The paper also describes spatial trends of different types of innovation (e.g.

product vs process innovation) and shows whether the presence of different industrialspecialization in metropolitan regions is related to a particular kind of innovation

activity. We speculate that product and marketing innovation � being dependent

either on a highly qualified workforce or on service sectors � are inclined to be more

intense in metropolitan regions. Finally, the most innovative performance, measured

by the simultaneous presence of firms able to carry out product and process

innovation in a region, is expected to be higher in metropolitan regions.

Research questions and data description

The general aim of the paper is to describe the relationship between the presence of

large cities in regions and the regional innovation performance. In fact, the presence

of large cities in a region represents per se a potential for its innovation capacity,

since cities are expected to generate dynamic agglomeration economies and knowl-

edge spillovers (Anselin et al. 2000, Capello 2001, Frenkel 2001, Simmie 2001).

This paper moves a step beyond previous analyses on this topic by describing

whether this linkage is mediated by the urban industrial structure. In fact, a positivecorrelation between the presence of large cities and the innovative performance in a

region is likely to exist. However, this relationship is expected to depend also on

the presence of knowledge-intensive service more than on advanced manufacturing

activities. This correlation can be even stronger for some specific types of innovation,

like product and marketing innovation, that either require a highly skilled labor

market (the former) or a widespread diffusion of service activities (the latter).

The research questions that are addressed in the empirical analysis are the

following:

(1) Is it true that agglomerated regions, hosting large urban areas, are the most

innovative ones?

(2) Is it true that, among all types of innovation, product and marketing

innovations are the most developed in agglomerated regions?

(3) Is it true that agglomerated regions specialized in knowledge intensive

service are more innovative than those specialized in manufacturing?

In order to answer these questions, this paper relies upon original data being

collected and developed in the frame of an ongoing ESPON (European Spatial

Observation Network) project.1 Data collection is based on the EUROSTAT

NUTS2 classification, and entails three main sources of data: data on the share of

employment in high tech manufacturing and service sectors; indicators on the

settlement structure; and an array of innovation indicators based on national CIS

figures developed at the NUTS2 level. The choice of the administrative areas used in

empirical analyses is a long disputed debate. Our work is based on EUROSTAT’sNUTS2 level for two main reasons. Conceptually, NUTS3 regions are often too

small to encompass functional urban areas; NUTS1 regions, on the contrary, tend to

be too large to contain local effects within their boundaries. In addition, this study is

related to an ongoing project financed by ESPON, whose research is mostly based on

the NUTS2 classification.

Innovation � The European Journal of Social Science Research 153

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

The high level of complexity of modern production systems makes defining high-

tech industries an awkward task. Because of the deep technological content of both

manufacturing and service activities, neither should be ex ante excluded from the

definition. The definition of high-tech industries is, however, partly arbitrary;therefore, we decided to choose a broad one, encompassing sectors with medium-

high and high-tech content so as to capture a wide range of sectors characterized

by considerable high-tech creation and deployment. High-tech sectors are iden-

tified according to the OECD definition (Organization for Economic Cooperation

and Development 2005, NACE revision 1.1). Such sectors include manufacturing

of aircraft and spacecraft, pharmaceuticals, office, accounting and computing

machinery, radio, TV and communications equipment, and medical, precision

and optical instruments; in the following, we will refer to these classification as‘‘HT manufacturing’’. As for high-tech services, we considered those defined by the

OECD as ‘‘Knowledge-Intensive Service (henceforth, KIS) Activities’’.2 We classi-

fied regions according to their specialization in high-tech manufacturing and/or

service sectors as compared with the EU average. In detail, specialization is here

captured by the location quotient (LQ) computed with respect to the EU31 average

value on the basis of regional employment data in HT manufacturing and KIS,

available in EUROSTAT. Specialization is computed for two years (2002 and 2007),

in order to detect possible time trends.For the settlement structure, urban areas are defined in this paper as metro-

politan regions. The settlement structure typology here adopted is the one developed

within ESPON projects. Agglomerated regions are defined as those with a city

of more than 300,000 inhabitants and a population density of more than 300

inhabitants per square kilometer, or a population density between 150 and 300

inhabitants per square kilometer. Urban regions are defined as those with a city of

between 150,000 and 300,000 inhabitants and a population density of between 150

and 300 inhabitants per square kilometer (or a smaller population density � 100 and150 inhabitants per square kilometer with a bigger center of more than 300,000).

Rural regions have a population density lower than 100 per square kilometer and a

center of more than 125,000 inhabitants, or a population density lower than 100 per

square kilometer with a center of less than 125,000. Therefore, agglomerated regions

turn out to be those with a high population density and hosting at least one large

city, while urban regions are characterized by the presence of medium-sized cities.

Lastly, innovation indicators are based on national CIS4 wave figures (covering

the 2002�2004 period), redistributed at the NUTS2 level. In particular, we focus onthe following types of innovation activities, captured by different questions of the

CIS: only product innovations, only process innovations, product and process

innovations (both types of innovation simultaneously as well as all the first three

main typologies together), and marketing and/or organizational innovations.

Throughout the paper, maps are drawn according to the natural breaks method

developed by George Jenks. This method is probably the most widely used system

of definition of classes in choropleth maps, in particular because it allows the

minimization of within-classes variance and the maximization of across-classesvariance levels.

For the sake of clarity, however, it is useful to introduce here the following

definitions. In particular, following the CIS, we consider as product innovation new

and significantly improved goods and/or services with respect to their fundamental

characteristics, technical specifications, incorporated software or other immaterial

154 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

components, intended uses, or user friendliness. Process innovation relates instead

to the implementation of new and significantly improved production technologies

or new and significantly improved methods of supplying services and delivering

products. A third category is available, called product and process innovation, whichcaptures the capacity of firms to develop both product and process innovation at

the same time. This category measures the best innovative performers since the

introduction of product and process innovation requires the simultaneous presence

of markedly different innovation skills. Lastly, in the CIS, another type of innovation

is captured, i.e. product and/or process innovation, which represents a general category

entailing either product or process or both of them. Marketing innovation is defined

as the introduction of ‘‘significant changes to the design or packaging of a good or

service’’ or ‘‘new or significantly changed sales or distribution methods, such asinternet sales, franchising, direct sales or distribution licenses’’. Finally, an

organizational innovation is defined as the introduction of either ‘‘new or significantly

improved knowledge management systems to better use or exchange information,

knowledge and skills within your enterprise’’, ‘‘a major change to the organization of

work within your enterprise, such as changes in the management structure or

integrating different departments or activities’’ or ‘‘new or significant changes in

your relations with other firms or public institutions, such as through alliances,

partnerships, outsourcing or sub-contracting’’. The Appendix fully details themethodology followed to obtain innovation figures at NUTS2 level.

Specialization in high-tech sectors in EU metropolitan regions

Location quotients for both HT manufacturing and KIS, on the basis of employment

data, point to a slight decrease in HT manufacturing activities that has taken place

in major industrial countries in the last decade. Regional specialization in the HT

industry markedly declined between 2002 and 2007 in most French, Polish, British,Bulgarian and Greek regions; at the same time, a relative positive shift occurred

in most regions belonging to two belts, one running north�south and the other

stretching west�east on the continent. On average, the location quotient for the HT

industries declined by 0.02 in the EU15 regions and increased by 0.09 in New

Member States (hereafter NMS).

Data also indicate that, although regions characterized by an agglomerated

settlement structure do not display remarkably higher specialization levels in the

HT manufacturing, they do so in terms of KIS. In time, however, a decrease in high-tech manufacturing specialization has taken place in the EU27, with rural regions

showing over the period 2002�2007 an increasing specialization in HT manufactur-

ing, which is mirrored by a simultaneous decrease in more urbanized areas.

The loss of high-tech manufacturing is not necessarily matched by a simultane-

ous process of increasing specialization in advanced services (Figure 1). In fact, on

average EU15 regions show zero variation in the KIS location quotients, whilst

NMS show a slight increase (0.01). Remarkable country effects characterize the

data, with three countries registering significant correlations between the changein HT manufacturing and the change in KIS specializations. In particular, nega-

tive correlation can be found only for Greece, Italy and Sweden, where regions

apparently switched regime, swapping a focus on advanced manufacturing with a

specialization in advanced services. Elsewhere, insignificant relations suggest that

manufacturing jobs flowing to NMS or outside Europe are not necessarily replaced

Innovation � The European Journal of Social Science Research 155

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

with similarly advanced functions. Country-wise, however, the negative correlation

between specialization levels in HT manufacturing and KIS industries is remarkable,

although not significant. At the country level, therefore, the shift of modern EU27

economies towards advanced services seems to characterize countries previously

specialized in HT manufacturing.

By examining top 10 performers over the analyzed time span (Table 1), the

picture displays a less dynamic behavior. In fact, from this perspective the situation

seems much more stable, with only one change taking place between 2002 and 2007

for HT manufacturing (Franche-Comte being substituted by Severovychod), while

more changes take place in KIS (five out of 10 regions in the 2007 top 10 table would

not be listed in 2002). The hierarchy of HT manufacturing seems therefore quite

hysteretic, with more change taking place in KIS, where in particular a strong

specialization of capital city-regions seems to take place.

In order to better understand the interplay and the possible synergic effects of

specialization in manufacturing and services, we classified regions according to their

level of specialization in both sectors (in comparison with the EU average value).

!

!

!

!

!

!!

!!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

Paris

Praha

Minsk

Tounis

Lisboa

Skopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km

© Politecnico di Milano, ESPON KIT Project , 2011

Change in LQ knowledge intensive services 2002-2007NA

-0.15 - -0.12

-0.11 - -0.05

-0.04 - -0.02

-0.01 - 0.01

0.02 - 0.04

0.05 - 0.09

0.10 - 0.16

0.17 - 0.46

Figure 1. Change in the location quotient in knowledge intensive services (2002�2007).

Source: authors’ calculation on EUROSTAT employment data.

156 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

We can thus identify four main types of region, as summarized in Figure 2: regions

with specialization simultaneously in HT manufacturing and KIS are labeled as

Technologically Advanced Regions (TARs); regions without any advanced specia-

lization are labeled low-tech regions; while regions with a specialization in either KIS

or HT manufacturing are called, respectively, KIS regions and HT manufacturing

regions.

Figure 3 displays (according to classes of Figure 2) the geography of European

regions according to the proposed classification. Twenty-one regions identified as

TAR are German, 13 British, eight French, five Belgian, four Swiss, three Swedish,

two Finnish and Danish, and one each for Italy, Norway, Slovenia and Slovakia.

The geography of technology in Europe is indeed highly concentrated, although

peripheral regions and regions with capital cities in NMS do play a major role. Over

time (although the time span considered may be too short to draw safe conclusions),

no region acquired or lost the status of TAR.The productive fabric of Europe shows therefore a remarkable concentration

of technology, related either to the advanced manufacturing or services activities.

Table 1. Top 10 regions in terms of location quotients in high-technology (HT) manufactur-

ing and knowledge-intensive services (KIS), 2002�2007.

HT manufacturing Knowledge intensive services

Location

quotient 2002 2007 2002 2007

Region 1 Stuttgart Stuttgart Inner London Inner London

Region 2 Tubingen Braunschweig Stockholm Stockholm

Region 3 Braunschweig Karlsruhe Oslo og Akershus Oslo og Akershus

Region 4 Franche-Comte Tubingen Outer London Hovedstaden

Region 5 Kozep-Dunantul Rheinhessen-Pfalz Brussels Aland

Region 6 Karlsruhe Unterfranken Hovedstaden Zurich

Region 7 Niederbayern Freiburg Ovre Norrland Berlin

Region 8 Unterfranken Severovychod Mellersta Norrland Noord-Holland

Region 9 Rheinhessen-Pfalz Kozep-Dunantul Ile de France Utrecht

Region 10 Freiburg Niederbayern Surrey and Sussex Ovre Norrland

Source: authors’ calculation from EUROSTAT data.

HT manufacturing regions Technologically-Advanced Regions

Low-tech regions

Specialization in KIS

Specialization in high-tech manufacturing

EU average

KIS regions

Figure 2. Sector-based typology of regions.

Innovation � The European Journal of Social Science Research 157

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

Moran’s I index, measuring the degree of spatial autocorrelation among regions and

calculated on the basis of a Rook contiguity matrix of second order, is in fact equal

to 0.18, and significant at all conventional levels, for the categorical variable

‘‘Technologically-Advanced Region’’ depicted in Figure 2.

This statement, however, needs qualification. In fact, while specialization in

high-tech manufacturing seems to be much more diffused across the European

space, specialization in KIS displays impressive concentration rates. It is finally

worth stressing that some countries present no specialization type � neither in HT

manufacturing, nor in KIS industries.

Tables 2 and 3 offer a summary of the spatial distribution of differently special-

ized regions by settlement structure. They are based on the same raw numbers

and differ only in the way percentages are calculated: in Table 2, percentages are

calculated by column (i.e. we show how many agglomerated, urban or rural are TAR,

HT manufacturing, KIS, and low-tech, respectively). In Table 3, instead, we show the

!

!

!!!!!!!!!

!!

!!

!!!!!!!!!

!!!!!

!!!!

!!

!!!

!

!

! !

!

!

!!!

!

!!!!!!

!

!!

!

!

!

!

!

!!

!

!!

!!!!!!!!!

!!!!!!!

!!

!

!

!!

!!!

!

!

!

!

!

!!!!!!!!

RRRRRooommmaaaaaaaaaa

RRRiiiiiiggggggga

OsssOssOsOOOsOsOOssOOO lllooo

BBBBBBBBernnnnnnnnnnnnnnnnnnnnn

WWWiieeennnnnnnnnnn

Kyiv

VVVVaVV ddddddddddddddddddddddddddddduuuuuuuuuuuuuzzzzzzzzzzzzzz

PPPPPPPPPPaaaarrrrrrrrriiiiiissss

PPPPPPPPPPPPPPPPPrrrrrrraahhahaaaahhhaaaaaa

Minsk

T uouuuTT nniiiissss

Liiissssssbbboooooooooooooooooooooooooa

SSSSkkkkopopopopopooopopoooooooo jjjjeee

ZZZZZaaaaaaaaaagreb

Ankara

MaddMaMadMM rrrrrrrriiiidd

Tiiiirrana

SSSSSSSSooffiiyyyaa

Loooooooooooonnnnnnddddddddddddddooooooonnnn BBBBBeeeeeeeeerrliill nn

DDDDDubbbuubbbblin

AAtttthhhhhhhhhhhhhhiiiiiiiiiiiiiiiiiiinnnnnnnnnaaaaaaaaiii

TaaaaaaTT llillllillillll nnnnnnn

NNNNNNiiiiiiccccccooooooooossssssiiiiiiaaaaaaaa

Beog!

rad

VViillnniuuuuuuuuuussss

Ar RRRRiiiibbbat

Kishineeevvvv

Sarajevoooo

Helsinnnnnnkkkkkkkiiiiii

BBBBBBBBBBBBBBBBBBBB ddddududdduudddddddddddddddddududddddddddudddduddududdaaaaapepeppeppppppeeeeeepepeeeeeesssssssttttt

WWWWaaarrssszzzaaawwwaaa

PPPPPodgo!!

rrrricccccccccccccaaa

El-Jazair

LLLLLLLLLLLjjjjjjjjjuuuuuuubljljlljaaannnnnnaaaaaaa

Stooooockkckkkkkkkcc hhhhhhhhhhhhhhhhooooooolm

RRReykjykjjykjykjkkykjykjyyyy jjkjkkkyky jjjkjjaaaaavviiikk

KKKøøøøøøøøøøøbbbbbbbbbbbbbbbbbenhaeneenennnnhanhahanhnhnn vvvnn

BBBuuccccuuuuurrrrrreeessttii

Amsssttttttteeerrrrrrddddddddddaaaammmmmmmmmmmmmmmm

BBBBBBBBBBBBBBBBBBBBBrrrrrrrrraaatislslslaaaaaaaaaavvvvvvvaaa

LuLLuLLLLLLuLLLLLLLLLLLLLLL xxxeeeeeemmmmmmmmmmmmmmmmmmmmbouuuuuurggggggg

BBBBBBBBBBrrrrruuuxxxeeeeeeeeeeelleeeeeeessssssssss////BBBBBBBBBBBBBBBBBBrrrrrruuu ssssssssssssssssssssssssssssssssss eeeeeeeeeeeeeeeeeeeeeeeeeeeel

VaVV lletta

Acores

GGGGGGuyyane

Madeira

Réunion

CaCaCaCannnnarias

Martrr iniquqq eGuadeloupe

0 5002522 0kmkk

© Politecnico di Milano, Project KIT, 2011

NA

Low tech regions

Advanced manufacturing regions

KIS regions

TeTT chnologically-advanced regions

Figure 3. Geography of European regions according to the sector-based classification, 2007.

Source: authors’ calculations from EUROSTAT data.

158 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

reverse, (i.e. how many TAR, HT manufacturing, KIS and low-tech are respectively

agglomerated, urban or rural).

In Table 2, column one clearly points out that the largest group is composed

of KIS regions, whereas the other three groups are similarly large. Agglomeratedregions are characterized by both advanced manufacturing and service activities and

by KIS specialization. Urban regions are typically advanced manufacturing regions,

while rural regions are characterized, as expected, by low-tech activities. Each

typology of region is significantly different from both the others, and the significance

of such difference is very strong. Two-tailed t-tests (not displayed here) always

rejected the null of equality of the two distributions at the 1% significance level.

Table 3 shows that TARs are mostly characterized by a dense, or hyper-dense,

urban structure (more than 80% of TARs are either agglomerated or urban). Similarconclusions apply to KIS regions, while HT manufacturing regions present mainly

an urban and rural settlement structure. Consequently, almost three-quarters of

low-tech regions are characterized by a rural structure. Once again, each type of

settlement structure characterizes a markedly different distribution of regions, with

the usual two-tailed t-tests (not displayed here) always rejecting the equality null at

the 1% significance level.

An interesting question is whether the innovative performance of agglomerated

regions changes according to their industrial specialization. This is the subject matterof the next section.

Innovation trends in EU metropolitan regions

The fundamental relevance of innovation process in contemporary economies is not

in general matched by quality data. CIS represents one of the best and most updated

Table 3. Advanced sector-based typology of regions by settlement structure.

Typology of regions

All Agglomerated Urban Rural Total

Technologically advanced regions 22% 38.60% 42.11% 19.30% 100%

HT manufacturing regions 21% 15.52% 48.28% 36.21% 100%

KIS regions 33% 36.59% 35.37% 28.05% 100%

Low-tech regions 24% 11.76% 14.71% 73.53% 100%

Source: authors’ calculation from EUROSTAT data.

Table 2. Advanced sector-based typology of regions by settlement structure.

Typology of regions

All Agglomerated Urban Rural

Technologically advanced regions 22% 37.64% 29.97% 12.28%

HT manufacturing regions 21% 15.16% 34.38% 23.05%

KIS regions 33% 35.71% 25.19% 17.86%

Low-tech regions 24% 11.50% 10.46% 46.80%

Total 100% 100% 100% 100%

Source: authors’ calculation from EUROSTAT data.

Innovation � The European Journal of Social Science Research 159

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

attempts to measure innovation activities. As pointed out above, in fact, the CIS

precisely allows the distinction between different types of innovation activities. CIS

data, however, are unequally stratified across space. Since in some EU countries data

are not stratified at NUTS2 level, such spatial detail is not publicly made available.

The authors offer a major improvement in this direction, by providing a robust

methodology to estimate regional CIS data (see the Appendix for a detailed

description of the methodology). This section presents the first descriptive results

drawn from this new database on regional innovation data.

The spatial distribution of innovation activity in European regions does indeed

display consistent autocorrelation, as pointed out above. Table 4 shows the value

of Moran’s I statistic calculated on the basis of a distance weight matrix on the

subsample of 265 NUTS2 regions for which innovation rates are available and

coordinates do not alter significantly the magnitude of calculated statistics.3 All

three variables are characterized by a highly significant pattern of positive spatial

autocorrelation, which implies that regions with high levels of innovation rates tend

to be surrounded by regions with similarly high levels, and vice versa.

Product innovation trends

For product innovation only (depicted in Figure 4), spatial concentration seems to be

prominent. This variable in fact displays considerable concentration in selected

countries,4 the core being in German, Scandinavian, Swiss and British regions, with a

few notable exceptions outside these areas. EU15 regions tend on average to innovate

more, and significantly so, than Eastern ones; the same applies to denser regions,

while rural regions display a relatively lower product innovation rate. In general,

spatial concentration looks pronounced regardless of the country being a strong

or weak product innovator. In particular, the latter is the case for Portugal, where

Lisbon is the only area with some product innovation activity, Spain, with Madrid,

Barcelona and a few Pyrenean regions, Greece, and some NMS. Italy represents

an exception to this pattern, since several regions in the northern and central part

of the country display similar product innovation rates. Overall, spatial patterns

characterize the variable not only across the country, but also within countries;

in fact, capital regions tend to display higher product innovation rates, with some

notable exceptions of regions also registering consistent innovation performance

despite not hosting the capital city (e.g. Rhone-Alps and Toulouse in France).

Spatial concentration can be technically verified by mapping the clusters of high

levels of the variables mapped surrounded by similarly high levels, low�low, high�low and low�high levels as defined with the use of Local Indicators of Spatial

Association (LISA; Anselin 1995). Maps are available in Appendix 2; Table 5 shows

that, among the possible urban settlement structures, agglomerated regions are the

ones that have the highest percentage of firms innovating in product only, belonging

Table 4. Moran’s I statistic for different regional innovation rates.

Type of innovation Moran’s I

Product innovation only 0.17***

Process innovation only 0.21***

Marketing and/or organizational innovation 0.32***

160 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

!

!

!

!

!

!!

!!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

Paris

Praha

Minsk

Tounis

Lisboa

Skopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km© Politecnico di Milano, Project KIT, 2011

Share of product innovation onlyNA0 - 3.263.27 - 5.925.93 - 9.129.13 - 12.8012.81 - 17.3017.31 - 23.4323.44 - 33.4533.46 - 44.42

Figure 4. Share of firms developing product innovation only.

Source: authors’ estimations from CIS national EUROSTAT data.

Table 5. Share of firms developing product innovation only by LISA cluster and regional

settlement structure.

LISA cluster Total regions

Agglomerated

regions Urban regions Rural regions

High�high 66 0.47 0.48 0.05

Low�Low 96 0.21 0.20 0.59

Low�high 22 0.18 0.27 0.54

High�low 2 0.50 0.50 0.00

No spatial association 79 0.16 0.42 0.42

Source: authors’ calculations.LISA, Local Indicators of Spatial Association.

Innovation � The European Journal of Social Science Research 161

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

to the cluster displaying positive spatial autocorrelation (high�high). Rural areas, onthe contrary, are characterized by a negative spatial association.

Process innovation trends

Process innovation shows a more dispersed pattern than product innovation

(Figure 5). Countries such as Portugal, Spain, France, Germany and the UK do not

display a remarkable concentration of process innovation within their boundaries.

The variance associated with this variable is much lower than the same measureassociated with product innovation. This finding further strengthens the case for

a more evenly distributed practice. In fact, this is also reflected in the case of

NMS, that are unexpectedly characterized by homogeneous spatial trends. Process

innovation takes place more frequently in densely populated regions and in

metropolitan areas. A relevant dichotomy shows up between western and eastern

!

!

!

!

!

!!

!!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

Paris

Praha

Minsk

Tounis

Lisboa

Skopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km© Politecnico di Milano, Project KIT, 2011

Share of process innovation onlyNA0 - 5.405.41 - 8.098.10 - 10.0910.10 - 12.3212.33 - 14.7114.72 - 18.0118.02 - 25.9225.93 - 55.08

Figure 5. Share of firms developing process innovation only.

Source: authors’ estimations from CIS national EUROSTAT data.

162 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

countries, the former averaging process innovation rates higher by about 5% than

regions in NMS. Given the softer nature of process innovation, however, on average

innovation rates are in the case of this variable consistently higher than product

innovation. Overall, process innovation displays an average value, across Europeancountries, higher by just 1 percentage point than product innovation. In particular,

it is worth stressing that process innovation displays on average higher values in

southern European countries than in the rest of Europe, by about 2 percentage

points. However, product and process innovation display remarkable levels of co-

variation. On average, regions displaying large levels of product innovation are also

simultaneously proficient in process innovation. A notable exception is represented

by southern European countries, notably Spain, Greece, Italy and Portugal, where

relatively insufficient performance in terms of product innovation is matched bysuperior performance in process innovation.

Table 6 shows the LISA clusters associated with process innovation and a picture

of spatial association of high values in different spatial typologies, strengthening

the above-mentioned result of a relatively evenly distributed innovation activity.

A remarkable result of this analysis is also the absence of any type of spatial associa-

tion in 65% of all observations analyzed, which further confirms that process

innovation is a more evenly distributed activity than product innovation.

Marketing and/or organizational innovation trends

Differently from product and process innovation, marketing and/or organizational

innovations capture nontechnological elements of innovation progress such as qual-

ity improvements, reductions of environmental damages stemming from firms’ pro-duction, reductions of energy consumption, creation of new markets, reduced labor

costs, reductions of amount of materials required for production, and conformance

to regulations.

Figure 6 highlights a significant concentration of marketing and/or organiza-

tional innovation in regions in the EU15 countries, with particularly high values in

German and Austrian regions. However, the spatial distribution of this soft form of

innovation seems much more even across the European space. The relatively even

distribution is in particular remarkable when observed within countries, witnessing asimilar innovative capability among regions.

This even distribution notwithstanding, spatial patterns characterize marketing

and/or organizational innovation, with a consistently higher tendency to introduce

such improvements in capital regions, and higher innovation rates also as region

Table 6. Share of firms developing process innovation only by LISA cluster and regional

settlement structure.

LISA cluster Total regions

Agglomerated

regions Urban regions Rural regions

High�high 76 0.28 0.32 0.41

Low�Low 0 0.00 0.00 0.00

Low�high 14 0.14 0.21 0.64

High�low 3 0.00 1.00 0.00

No spatial association 172 0.27 0.35 0.38

Source: authors’ calculations.

Innovation � The European Journal of Social Science Research 163

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

!

!

!

!

!

!!

!!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

ParisPraha

Minsk

Tounis

LisboaSkopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km© Politecnico di Milano, Project KIT, 2011

Share of marketing and organizational innovationNA0 - 9.059.06 - 15.2415.25 - 19.8119.82 - 23.5323.54 - 29.5629.57 - 37.5037.51 - 48.0548.06 - 78.36

Figure 6. Share of firms developing marketing and/or organizational innovation. Source:

authors’ estimations from CIS national EUROSTAT data.

Table 7. Share of firms developing marketing and/or organization innovation by LISA

cluster and regional settlement structure.

LISA cluster Total regions

Agglomerated

regions Urban regions Rural regions

High�high 58 0.28 0.52 0.21

Low�Low 77 0.34 0.26 0.40

Low�high 25 0.12 0.36 0.52

High�low 4 0.50 0.25 0.25

No spatial association 101 0.22 0.31 0.48

Source: authors’ calculations.

164 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

density increases, as well as in regions with large cities, bearing the diversified and

creative environment leading to innovative behavior. NMS innovate in marketing

and/or organization less than EU15 regions, on average by about 9 percentage points.

Similar patterns affect Nordic and Mediterranean countries, regions in the lattersample innovating less in marketing and/or organization by about 5 percentage

points. Marketing and/or organizational innovation, however, is not an activity apart

from product and process innovation. In fact, pure correlation between marketing

and/or organizational innovation, on the one hand, and product and/or process

innovation, on the other, is remarkably high (equal to 0.71 and significant at all

conventional levels).

Table 7 shows mean values for each LISA cluster in agglomerated, urban and

rural regions. Interestingly, 52% of all regions in the high�high cluster are located inurban areas, suggesting the concentration of marketing innovation activities in

medium-size cities. Furthermore, about 40% of the whole sample displays no spatial

association at all.

To reply to the first two research questions in a descriptive way, we present the

innovation rates of agglomerated regions with respect to all other regions by type of

innovation activities. Overall, agglomerated regions show superior innovation rates

to urban and rural regions in all five fields. It seems likely indeed that metropolitan

regions tend to outperform the rest of the EU territory in all respects, the differencebeing strongly statistically significant as shown by the t-tests (Table 8). Moreover,

Table 8 also shows that the difference in innovation performance of regions is

statistically significantly higher in agglomerated for what concerns product innova-

tion and marketing and/or organizational innovation. Our impression that these two

kinds of innovation � being dependent either upon high-level quality workforce or on

service sectors � are those calling for the presence of large urban areas to take place

turns out to be true.

Industry innovation trends in EU metropolitan regions

Since agglomerated regions show greater innovation rates, it is of interest to

understand whether this relates to their industrial specialization. Table 9 gives some

insights in this direction. In particular, Table 9 replies to our third research questionpresented above by showing that KIS regions innovate more than advanced

manufacturing regions, but this is true for product innovation only and for marketing

innovation. Moreover, KIS regions are more inclined to innovate more than other

regions in product rather than in process innovation. This result is rather interest-

Table 8. Innovation rates in agglomerated regions by type of innovation activities.

Typology

Share of

product

innovation

only

Share of

process

innovation

only

Share of both

product and

process

innovation

Share of

product and/

or process

innovation

Share of

marketing and/or

organizational

innovation

Agglomerated 14.88 12.41 15.76 40.94 28.85

Others 8.69 10.57 14.70 33.40 24.83

t-Test 3.65*** 2.97*** 2.03** 3.28*** 16.38***

***, **, * Significant at the 1, 5 and 10% level, respectively.Source: authors’ calculation from CIS data estimations.

Innovation � The European Journal of Social Science Research 165

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

ing, witnessing that innovative activities are related to the production of new and

advanced services rather than to new processes. If this is related to the emergence

of smart cities through e-society, we can claim that the interpretation given by

agglomerated regions of the e-society is the production of new and advanced

services, rather than the mere supply of old services through new ICT networks.

Another interesting result contained in Table 9 is that synergic effects in innovation

performance exist if the region is specialized in advanced manufacturing and service

activities at the same time. Technologically advanced regions, in fact, look more

innovative than the other regions, regardless of the type of innovation being considered

(technological, such as product or process, or nontechnological, such as marketing

and/or organizational) and the settlement typology. These regions systematically

innovate more than the average of the whole sample, but also with respect to the

average innovation rate registered by either high-tech manufacturing or services

regions. This result suggests that the simultaneous presence of advanced manufactur-

ing and KIS may stimulate innovation activities, probably exploiting urbanization (i.e.

diversity) externalities, as suggested in Feldman and Audretsch (1998). In order to

provide support for this statement, however, more in-depth analysis is required.

Conclusions

In this paper we provide empirical evidence on the relationship between specializa-

tion in high-tech sectors and innovation performance at the regional level. We claim

that a sector-based approach is still worth analyzing as it is still shaping both the

scientific discourse as well as inspiring the current policy debate. Looking at urban

innovation process from this standpoint is also of paramount interest, since much of

Table 9. Innovation rates by industrial specialization and types of innovation activities in

agglomerated regions.

Typology

Share of

product

innovation

only

Share of

process

innovation

only

Share of

both product

and process

innovation

Share of

product and/

or process

innovation

Share of

marketing

and/or

organizational

innovation

Technologically advanced regions

Agglomerated 21.05 13.31 19.02 49.48 35.30

Others 13.54 12.34 17.60 42.52 30.87

t-Test 5.48*** 3.88*** 1.91** 4.65*** 2.75***

HT manufacturing regions

Agglomerated 10.46 13.53 16.24 40.24 25.63

Others 9.23 10.37 15.97 35.57 24.98

t-Test 1.78** 1.99** 0.28 0.94 0.04

KIS regions

Agglomerated 14.54 11.80 14.55 38.86 27.06

Others 10.37 10.60 13.29 33.25 23.67

t-Test 2.90** 0.84 0.71 1.70** 1.76**

***, **, * Significant at the 1, 5 and 10% level, respectively.Source: authors’ calculation from EUROSTAT data.

166 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

previous studies undertaken in this frame primarily focused on cluster-like

agglomeration environments rather than urban settlements.

The results of our analysis point to a picture of spatial concentration of both

high-tech manufacturing and services sectors, that have partly increased over time.

Interestingly, most European regions show some degree of specialization in either

services or manufacturing or both (being thus labeled as Technologically Advanced

Regions). The productive fabric of Europe shows therefore a remarkable concentra-

tion in advanced activities, related to either advanced manufacturing or services. This

statement, however, needs qualification. In fact, while specialization in manufac-

turing high-tech seems to be much more diffused across the European space,

specialization in KIS displays impressive concentration rates. Interestingly enough,

this concentration increases within agglomerated regions. Nearly 36% of agglomer-

ated regions are in fact specialized in KIS, while 37% are specialized in both

advanced manufacturing and service activities.

These patterns describe also the regional innovative performance and the type

of innovation activities carried out. Product innovation shows quite a prominent

spatial concentration, especially in central and northern European regions. On the

other hand, the geography of process innovation is less concentrated, and southern

regions seem to perform especially well in this regard. Lastly, marketing and/or

organizational innovation is relatively more evenly distributed across the European

space. Importantly, agglomerated regions are the best performing in all types of

innovation activities, but they look a particularly favorable setting for product and

marketing innovations. Moreover, our descriptive analysis shows that KIS regions

register a higher innovation performance than HT manufacturing regions.

Knowledge-intensive service regions also show their outstanding performance

in product innovation rather than process innovation with respect to other regions.

This is particularly important, since it witnesses that innovation in service is not the

mere supply of old services through new technological means (like e-services), but the

production and creation of new services (e.g. online data base management and data

mining), in which the real efficiency gains lie.

Interpretative analyses of the geographical patterns of innovation will be

developed, where other territorial specificities, playing the role of co-determinants,

shall be taken into account. Our results suggest that innovation in urban areas is still

an interesting research field. In particular, economic modeling of the interaction

between innovation and urban performance is the object of our future research.

Notes

1. The project is called KIT (‘‘Knowledge, Innovation and Territory’’) and is led by thePolitecnico of Milan, and in particular by the authors of this paper. For further details onthe project and the consortium members, see http://www.espon.eu/main/Menu_Projects/Menu_AppliedResearch/kit.html.

2. Medium-high and high-tech manufacturing industries include chemicals (NACE 24),machinery (NACE 29), office equipment (NACE 30), electrical equipment (NACE 31),telecommunications and related equipment (NACE 32), precision instruments (NACE 33),automobiles (NACE 34) and aerospace and other transport (NACE 35); KIS include watertransport (NACE 61), air transport (NACE 62), post and telecommunications (NACE 64),financial intermediation (NACE 65), insurance and pension funding (NACE 66), activitiesauxiliary to financial intermediation (NACE 67), real estate activities (NACE 70), rentingof machinery and equipment (NACE 71), computer and related activities (NACE 72),research and development (NACE 73) and other business activities (NACE 74).

Innovation � The European Journal of Social Science Research 167

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

3. Statistics are calculated with the exclusion of the French Overseas Department, Iceland,Switzerland and Norway.

4. Additional information on spatial patterns in innovation data can be found in Appendix 2.

References

Anselin, L., 1995. Local indicators of spatial association � LISA. Geographical analysis, 27 (2),93�115.

Anselin, L., Varga, A., and Acs, Z., 2000. Geographic and sectoral characteristics of academicknowledge externalities. Papers in regional science, 79 (4), 435�443.

Antonelli, C., 1989. A failure inducement model of research and development expenditure:Italian evidence from the early 1980s. Journal of economic behaviour and organization, 12 (2),159�180.

Aydalot, P., ed., 1986. Milieux innovateur en Europe. Paris: Groupe de Recherche European surles Milieux Innovateurs (GREMI).

Aydalot, P. and Keeble, D., eds., 1988. High technology industry and innovative environments:the European experience. London: Routledge.

Camagni, R., 1991. ‘Local milieu’, uncertainty and innovation networks: towards a newdynamic theory of economic space. In: R. Camagni, ed. Innovation networks: spatialperspectives. London: Belhaven Press.

Camagni, R. and Capello, R., 2009. Knowledge-based economy and knowledge creation: therole of space. In: U. Fratesi and L. Senn, eds. Growth and innovation of competitive regions:the role of internal and external connections. Berlin: Springer, 145�166.

Capello, R., 2001. The determinants of innovation in cities: dynamic urbanisation economiesvs. milieu economies in the metropolitan area of Milan. In: J. Simmie, ed. Innovative cities.London: Spon, 95�128.

Dasgupta, P. and Stiglitz, J., 1980. Uncertainty, industrial structure and the speed of R&D.Bell journal of economics, 11 (1), 1�28.

Feldman, M. and Audretsch, D., 1998. Innovation in cities: science-based diversity, speciali-zation and localized competition. European economic review, 43 (2), 409�429.

Frenkel, A., 2001. Why high-technology firms choose to locate in or near metropolitan areas.Urban studies, 38 (7), 1083�1101.

Griliches, Z., 1990. Patent statistics as economic indicators: a survey. Journal of economicliterature, 28 (4), 1661�1707.

Lundvall, B.-A. and Johnson, B., 1994. The learning economy. Journal of industry studies,1 (2), 23�42.

MacDonald, S., 1987. British science parks: reflections on the politics of high technology.R&D management, 17 (1), 25�37.

Massey, D., Quintas, P., and Wield, D., 1992. High tech fantasies: science parks in society,science and space. London: Routledge.

Monk, C.S.P., Porter, R.B., Quintas, P., Storey, D., and Wynarczyk, P., 1988. Science parks andthe growth of high technology firms. London: Croom Helm.

Pavitt, K., 1984. Sectoral patterns of technical change: towards a taxonomy and a theory.Research policy, 13 (4), 343�373.

Organization for Economic Cooperation and Development, 2005. Science, technology andindustry scoreboard 2005. Paris: OECD.

Simmie, J., 2001. Innovative cities. London: Spon.Storey, D.J. and Tether, B.S., 1998. Public policy measures to support new technology-based

firms in the European Union. Research policy, 26 (9), 1037�1057.

Appendix 1. Innovation data: estimation methodology

The CIS is designed to obtain information on innovation activities within enterprises with 10or more employees. National CIS data are available in EUROSTAT. We estimated regionaldata (i.e. NUTS2 level) starting from the national data (i.e. NUTS0 level) in order to ensurecomparability across countries.

168 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

The estimates at NUTS2 level were obtained with a two-step procedure. The first step callsfor the disaggregation of the national data according to specific weights, especially defined foreach category of innovation. The weights aim to capture both a functional as well as anindustrial dimension. The former is captured by looking at the share of professions, the latterby looking at the industrial specialization. In absence of any a priori assumption on differentrelevance of the functional vs the industrial dimension, we attributed equal importance to theselected weights. Table A1 shows the selected weights.

The second step of the estimate requires the robustness of the estimates. To check therobustness of our estimates we implemented a series of benchmark exercises. In detail, weimplemented three types of tests, namely on the equality of means, on the equality of standarddeviation, and of Kolmogorof�Smirnoff, to assess whether our estimates diverge from theoriginal sample distribution.

We performed two sets of comparisons. First, we compared our estimates of the share ofonly product innovators, the share of only process innovators and the share of product andprocess innovators with regional data from National Statistical Offices. These latter wererescaled at the National value available from EUROSTAT, since the national figures availablefrom EUROSTAT and National Statistical Offices may differ according to different strataweighting procedures. The tests could be implemented only on limited set of countries, namelyItaly, Romania and Czech Republic, that publicly release these data on their websites.

Next, to support further our estimates, we made use of data on product and/or processinnovators from Regional Innovation Scoreboard (RIS). In particular, we compared ourestimates of product and/or process innovators, obtained as the sum of the first threecategories of innovators (i.e. only product innovators, only process innovators, product andprocess innovators), with RIS data. The tests could be implemented only on those countrieswhose data are available in the annex to the RIS methodology report.

Still, some problems of comparability remain. For example, the France NUTS0 dataavailable from RIS on the share of product and/or process innovators is different from theFrance NUTS0 data available from EUROSTAT (in particular, the former is smaller than thelatter), which may affect the mean value of our estimates.

Table A2 summarizes the results of these tests. Overall, they indicate that our estimates donot statistically differ in their mean, standard deviation and distribution from the official datareleased either by National Statistical Offices or by RIS. Although for some countries, the testsindicate that either the mean or the standard deviation can be statistically different, the outputof the Kolmogorov�Smirnoff test lends support to our estimates and indicates that thedistribution of the original sample does not statistically differ from that of our estimates.

Appendix 2. Spatial patterns in regional innovation patterns

Figures A1�A3 map LISA clusters for the product innovation only, process innovation only,and marketing and/or organizational innovation rates. Country effects are evident in the firstand the last case. While in terms of product innovation, German and British regions tend to

Table A1. Selected weights.

Type of innovation Weights

Only product Percentage scientists, percentage employment in high-tech

(employment in manufacture of electrical and optical equipment

(DL sector in NACE Rev 1.1. classification))

Only process Percentage employment in manufacturing, percentage technicians,

percentage managers

Both product and process Percentage scientists, percentage employment in high-tech (DL),

Percentage employment in manufacturing, percentage technicians,

percentage managers

Marketing and/or

organizational

Percentage managers, percentage employment in services

Innovation � The European Journal of Social Science Research 169

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

Table A2. Robustness checks.

Type of innovation

Sample

Mean

estimates

Mean benchmark

estimates Mean difference

Standard deviation

difference

Kolmogorov�Smirnoff test

(different distribution)

Product only

Italya 4.41 4.53 NS NS Not significant; p-value

equals 0.94.Romaniaa 1.95 1.69 NS �; pB0.05

Process only

Italya 14.27 14.00 NS NS Not significant; p-value

equals 0.95.Romaniaa 4.72 4.82 NS �; pB0.01

Product and process

Czech Republica 14.48 14.38 NS B; pB0.05 Not significant; p-value

equals 0.98.Italya 8.90 9.01 NS NS

Romaniaa 13.87 13.15 NS B; pB0.01

Product and/or process

Austriab 49.03 50.03 NS NS Not significant; p-value

equals 0.98.Belgiumb 42.37 46.61 NS NS

Bulgariab 15.03 15.21 NS NS

Czech Republicb 37.03 36.05 NS NS

Spainb 29.97 29.06 NS �; pB0.01

Finlandb 34.45 34.52 NS NS

Franceb 27.55 24.37 NS �; pB0.01

Greeceb 29.72 39.30 B; pB0.01 NS

Hungaryb 18.09 17.37 NS NS

Italyb 31.77 32.21 NS NS

Polandb 23.07 38.95 NS NS

Portugalb 39.40 38.95 NS NS

Romaniab 20.18 17.74 NS NS

17

0R

.C

ap

elloet

al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

Table A2 (Continued )

Type of innovation

Sample

Mean

estimates

Mean benchmark

estimates Mean difference

Standard deviation

difference

Kolmogorov�Smirnoff test

(different distribution)

Sloveniab 34.11 23.85 �; pB0.05 NS

Slovakiab 22.43 20.01 NS NS

United Kingdomb 25.80 42.08 NA NA

Italya 31.77 27.59 NS NS

Romaniaa 20.18 20.54 NS NS

Marketing and/or organizational

Austriab 80.52 80.52 NS NS Not significant; p-value

equals 0.51Belgiumb 80.33 70.36 NS NS

Bulgariab 0.76 0.94 NS NS

Czech Republicb 54.83 54.23 NS NS

Spainb 35.72 32.53 �; pB0.05 NS

Finlandb 69.13 72.81 NS NS

Franceb 55.78 56.04 NS �; pB0.05

Italyb 49.12 51.39 NS NS

Polandb 26.88 27.43 NS NS

Portugalb 64.49 67.43 NS NS

Romaniab 33.71 32.10 NS NS

Sloveniab 54.35 54.28 NS NS

Slovakiab 19.65 18.15 NS �; pB0.05

United Kingdomb 42.14 43.44 NS �; pB0.05

aSource of data used as benchmark: National Statistical Offices.bSource of data used as benchmark: Regional Innovation Scoreboard 2009.NS, not significant; NA, not available.

Inn

ovatio

n�

Th

eE

uro

pea

nJo

urn

al

of

So

cial

Scien

ceR

esearch

17

1

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

represent clusters of high values, in marketing and/or organizational innovation, Germanregions are matched by Finnish ones.

Figure A2 shows how instead process innovation departs significantly from this pattern,by identifying a belt of Southwestern EU countries, whose regions, from Portuguese andSpanish to French, from Italian to German and Austrian ones, tend to identify clusters ofconsistent innovative activity.

In the case of marketing and/or organizational innovation, it is worth stressing how capitalregions (Bucharest, Paris, Prague and Vienna) are characterized by a relationship of negativespatial association with the territory around them. This identifies metropolitan regionssurrounded by regions with relatively low values of marketing and/or organizationalinnovation, suggesting the role of regional attractors of human capital-intensive activitiestypical of this type of innovation.

Figure A1. LISA for the variable ‘‘Share of firms developing product innovation only’’.

Source: authors’ estimations from CIS national EUROSTAT data.

172 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

!

!

!

!

!

!!

!!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

Paris

Praha

Minsk

Tounis

Lisboa

Skopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km© Politecnico di Milano, Project KIT, 2011

LISA clusters

Process innovation onlyNAHigh-highLow-highHigh-low

Figure A2. LISA for the variable ‘‘Share of firms developing process innovation only’’.

Source: authors’ estimations from CIS national EUROSTAT data.

Innovation � The European Journal of Social Science Research 173

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14

!

!

!

!

!

!!

!!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

Roma

Riga

Oslo

Bern

Wien

Kyiv

Vaduz

Paris

Praha

Minsk

Tounis

Lisboa

Skopje

Zagreb

Ankara

Madrid

Tirana

Sofiya

London Berlin

Dublin

Athinai

Tallinn

Nicosia

Beograd

Vilnius

Ar Ribat

Kishinev

Sarajevo

Helsinki

Budapest

Warszawa

Podgorica

El-Jazair

Ljubljana

Stockholm

Reykjavik

København

Bucuresti

Amsterdam

Bratislava

Luxembourg

Bruxelles/Brussel

Valletta

Acores

Guyane

Madeira

Réunion

Canarias

MartiniqueGuadeloupe

0 500250km© Politecnico di Milano, Project KIT, 2011

LISA clusters

Marketing and organizational innovationNAHigh-highLow-lowLow-highHigh-low

Figure A3. LISA for the variable ‘‘Share of firms developing marketing and/or organiza-

tional innovation’’.

Source: authors’ estimations from CIS national EUROSTAT data.

174 R. Capello et al.

Dow

nloa

ded

by [

Uni

vers

iti T

ekno

logi

Mal

aysi

a] a

t 19:

39 0

4 Ja

nuar

y 20

14