Regional Studies

ISSN: 0034-3404 (Print) 1360-0591 (Online) Journal homepage: www.tandfonline.com/journals/cres20

Smart specialization policy in the European Union:

relatedness, knowledge complexity and regional
diversification

Pierre-Alexandre Balland, Ron Boschma, Joan Crespo & David L. Rigby

To cite this article: Pierre-Alexandre Balland, Ron Boschma, Joan Crespo & David L.
Rigby (2019) Smart specialization policy in the European Union: relatedness, knowledge
complexity and regional diversification, Regional Studies, 53:9, 1252-1268, DOI:
10.1080/00343404.2018.1437900

To link to this article: https://doi.org/10.1080/00343404.2018.1437900

© 2018 The Author(s). Published by Informa View supplementary material

UK Limited, trading as Taylor & Francis
Group

Published online: 06 Mar 2018. Submit your article to this journal

Article views: 36834 View related articles

View Crossmark data Citing articles: 244 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=cres20
REGIONAL STUDIES
2019, VOL. 53, NO. 9, 1252–1268
https://doi.org/10.1080/00343404.2018.1437900

Smart specialization policy in the European Union: relatedness,

knowledge complexity and regional diversification
Pierre-Alexandre Ballanda , Ron Boschmab, Joan Crespoc and David L. Rigbyd

ABSTRACT
The operationalization of smart specialization policy has been rather limited because a coherent set of analytical tools to
guide the policy directives remains elusive. We propose a policy framework around the concepts of relatedness and
knowledge complexity. We show that diversifying into more complex technologies is attractive but difficult for European
Union regions to accomplish. Regions can overcome this diversification dilemma by developing new complex
technologies that build on local related capabilities. We use these findings to construct a policy framework for smart
specialization that highlights the potential risks and rewards for regions of adopting competing diversification strategies.
KEYWORDS
smart specialization; knowledge complexity; technological relatedness; regional diversification; European Union Cohesion Policy

JEL O25, O38, R11

HISTORY Received 23 June 2017; in revised form 9 January 2018

INTRODUCTION specialized (i.e., less diversified), but instead to leverage

existing strengths, to identify hidden opportunities and to
In a market economy, the rapidity of technological change generate novel platforms upon which regions can build
forces continuous adaptation as innovation and economic competitive advantage in high value-added activities.
crises reshuffle the competitive standing of firms and Smart specialization emanated from the idea that regions
regions alike. Within this environment, the capacity to across the EU have different economic and institutional
develop new ideas, to recombine existing knowledge assets structures that shape possibilities for their future develop-
and blaze new technological trajectories is imagined to offer ment (Kroll, 2015). The result was a clear denunciation
at least temporary respite to the relentless pressure of falling of the top-down ‘one-size-fits-all’ policy that had led the
costs. Facing a variety of technological possibilities and an EU to fund nation-states rather than individual regions,
uncertain future, however, few economic agents are in a and fashionable sectoral targets rather than realistic indus-
position to identify the best way forward. It was against trial foundations (Asheim, Grillitsch, & Trippl, 2016;
this backdrop that the European Union (EU) forged the Tödtling & Trippl, 2005).
ambitious growth strategy called Europe 2020. The question remained: how does one identify the tar-
At the core of this development strategy is smart gets of place-based development policy? An answer was
specialization, a vision of regional growth possibilities readily available in the recommendations of the Knowledge
built around existing place-based capabilities (Barca, for Growth Expert Group (Foray et al., 2009). Their call
2009; Foray, David, & Hall, 2009, 2011; McCann & for smart specialization focused on building competitive
Ortega-Argilés, 2015). The goal of smart specialization is advantage in research domains and sectors where regions
not to make the economic structure of regions more possessed strengths and leveraging those capabilities

CONTACT
a
[email protected]
Department of Economic Geography, Utrecht University, Utrecht, the Netherlands.
b
(Corresponding author) [email protected]
Department of Economic Geography, Utrecht University, Utrecht, the Netherlands; and UiS Business School, University of Stavanger, Stavanger,
Norway.
c
[email protected]
LEREPS, Institute of Political Studies, University of Toulouse, Toulouse, France.
d
[email protected]
Department of Geography, University of California – Los Angeles, Los Angeles, CA, USA.
Supplemental data for this article can be accessed https://doi.org/10.1080/00343404.2018.1437900

© 2018 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-
nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built
upon in any way.
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1253

through diversification into related activities (Foray et al., Technological relatedness and regional
2012). In important ways, this work built a series of policy diversification
prescriptions around the concept of heterogeneity in Competition in today’s global economy rests heavily on
regional knowledge bases and path dependence in their innovation as many of the standard price-based forms of
evolution (Rigby & Essletzbichler, 1997) and the related economic advantage have been flattened. As the overall
variety concept (Frenken, Van Oort, & Verburg, 2007). stock of knowledge has expanded, a division of labour dis-
However, the operationalization of smart specialization tributed its parts widely across agents and different regions.
has been criticized as a ‘perfect example of policy running Thus, regional economies are increasingly understood as
ahead of theory’ (Foray, David, & Hall, 2011, p. 1), lacking localized communities of practice that reflect place-bound
an ‘evidence base’ (Morgan, 2015; Unterlass et al., 2015), sets of technological capabilities, routines and institutional
and building on ‘anecdotal evidence rather than the appli- arrangements (Storper, 1997). With knowledge production
cation of theoretically grounded methodologies’ (Iacobucci increasingly conceived as a process of recombining existing
& Guzzini, 2016, p. 1511; see also Santoalha, 2016). ideas (Weitzman, 1998), so regional economies are ima-
This paper aims to contribute to this debate in three gined to move along place-based technological trajectories,
ways. First, it provides a theoretical framework for smart where processes of search and exploration are guided by
specialization around the concepts of relatedness (Hidalgo, existing knowledge capabilities and well-established rou-
Klinger, Barabassi, & Hausmann, 2007; Neffke, Henning, tines (Dosi, 1982; Rigby & Essletzbichler, 1997). This is
& Boschma, 2011) and knowledge complexity (Balland & not to argue that knowledge spillovers and other forms of
Rigby, 2017; Hidalgo & Hausmann, 2009). Second, it knowledge sharing are always localized (Bathelt, Malm-
assesses this policy framework empirically and provides berg, & Maskell, 2004). However, the tacit nature of
new evidence on how EU regions develop new technologi- much knowledge means that geography continues to play
cal fields. To achieve this goal, we identify existing knowl- a critical role in the emergence and evolution of technology,
edge bases of EU regions, develop a measure of relatedness especially that which is complex and more valuable, for
between technological categories using data from the Euro- many kinds of knowledge do not travel well (Balland &
pean Patent Office (EPO) and calculate a knowledge com- Rigby, 2017; Maskell & Malmberg, 1999; Sorenson, Riv-
plexity measure of technology classes using network-based kin, & Fleming, 2006).
techniques. Third, it combines the relatedness and com- The idea that new technology is born of existing ideas
plexity measures in a smart specialization policy tool that has rekindled the debate on the costs and benefits of regional
assesses the costs and benefits of alternative technological diversity (Balland et al. 2015; Essletzbichler, 2015). Fren-
trajectories in each region. ken et al. (2007) suggest that more important than the over-
The paper is organized as follows. The next section pre- all variety of sectors found in different regions is the extent
sents the theoretical concepts of relatedness and knowledge to which elements of that variety are related. The relatedness
complexity as key building blocks of smart specialization concept rests on the idea that knowledge has an architecture
policy. The third section shows how these may be operatio- that is based upon similarities and differences in the way that
nalized using patent documents and network-based tech- different types of knowledge can be used. When knowledge
niques. The fourth section presents new econometric subsets are close substitutes for one another, or when they
evidence on how relatedness and knowledge complexity demand similar sets of cognitive capabilities and skills for
have shaped diversification patterns in European regions their use, we think of them as being related or proximate
between 1990 and 2009. The fifth section combines the to one another in some form of ‘knowledge space’ (Breschi,
relatedness and complexity measures in a smart specializ- Lissoni, & Malerba, 2003). Organizations compete
ation policy framework. The sixth section applies this fra- through extending their knowledge domains and their
mework to examine diversification prospects across a capacity to use more components of the knowledge space.
series of EU regions. The seventh section provides a brief They do this by search and exploration of different parts
conclusion and discusses some remaining questions related of that space. Search costs rise rapidly around the boundaries
to smart specialization policy. of existing expertise and thus the cost of diversification
hinges critically on the distance between knowledge com-
BUILDING BLOCKS OF SMART ponents that are known and those that remain as yet
SPECIALIZATION unknown (Atkinson & Stiglitz, 1969; Chatterjee & Wer-
nerfelt, 1991; Webber, Sheppard, & Rigby, 1992).
This section connects the smart specialization literature For Frenken and Boschma (2007), diversification is
(Foray et al., 2009, 2011) with the regional diversification imagined as a branching process that gives rise to new
(Hidalgo et al., 2007; Neffke et al., 2011) and economic activities within regions. Related diversification of cities
complexity literature (Hidalgo & Hausmann, 2009). We and regions is depicted as a higher-order reflection of
argue that relatedness and knowledge complexity are key micro-level dynamics in which individuals and organiz-
building blocks of smart specialization, which we envisage ations extend the scope of their activities around the tech-
as a place-based policy in which regions aim at renewing nological competencies and the behavioural routines that
and upgrading their economic structure by building on they accumulate over time (Balland, 2016). Thus, the
their existing capabilities (Boschma, 2014). emergence of new technologies and new sectors within

REGIONAL STUDIES
1254 Pierre-Alexandre Balland et al.

regions is not random, rather it reflects the existing collec- So far it has proven difficult to answer this question, at least
tive capacity of agents that produce regions with distinctive in part because there are no readily available measures of tacit
technological and industrial profiles. A large and expanding knowledge, let alone its spatial distribution. Kogut and Zan-
volume of empirical studies provides supporting evidence der (1993) argue that complexity is a critical dimension of
(Boschma, 2017). Hidalgo et al. (2007) explored how what makes some knowledge tacit. After Simon (1962),
countries expand their mix of exports around the products the complexity of a technology, a subset of knowledge, is
in which they already established a comparative advantage. understood as a function of the number of components out
Neffke et al. (2011) used product portfolios of manufactur- of which it is constructed and the interdependence of those
ing plants to understand sectoral relatedness and exploited components (Fleming & Sorenson, 2001). Hidalgo and
this methodology to examine industrial diversification in Hausmann (2009) developed an idea of product and place
Swedish regions. Along similar lines, Rigby (2015) and complexity based on the product-level diversity of national
Boschma, Balland, and Kogler (2015) used measures of economies and the ubiquity (or range) of countries across
relatedness between patent classes to predict the evolution which individual products are produced. They argue that
of technological change within US cities. At a more global countries develop different core competences: countries that
scale, Petralia, Balland, and Morrison (2017) analyzed amass larger sets of capabilities tend to produce more special-
technological branching using patent applications by ized products that are hard to copy or imitate by others. The
inventors in 65 countries. complexity of an economy is embodied in the wide range of
Based on this discussion, we formulate two hypotheses knowledge or capabilities that are combined to make pro-
on relatedness that will be tested for European regions, and ducts: less ubiquitous products are more likely to require a
which will be used in the fifth section to formulate a smart greater variety of capabilities. These specialized (complex)
specialization policy framework in which relatedness goods tend to be produced by relatively few national econom-
reflects the difficulty or cost of moving from one technol- ies and form the basis for long-run competitive advantage.
ogy to another. Complex goods also tend to be in their early stages of devel-
opment, which enhances further their growth potential. Bal-
Hypothesis 1: Regions are more likely to develop new specializ- land and Rigby (2017) report wide variations in the
ations in technological activities that are related to their knowl- complexity of knowledge produced across US cities that cor-
edge bases. relate highly with longer-run patterns of economic perform-
ance, with only a few metropolitan areas capable of producing
Hypothesis 2: Regions are more likely to experience technological the most complex technologies.
growth in technological activities that are related to their knowl- It should be clear that regions benefit from building
edge bases. comparative advantage in complex technologies. Regions
that are early innovators lock in growth around new tech-
nologies based on cumulative technological advantages
Knowledge complexity and regional and quasi-monopolistic rents. However, complex technol-
diversification ogies are relatively scarce and it is therefore difficult for
Knowledge bases of regions vary in terms not only of their agents that comprise regional economies to develop com-
technological composition but also of their value. The value petences in these fields. These two tendencies give rise to
of knowledge types or technologies, like most goods, a ‘diversification dilemma’. On the one hand, the search
reflects a balance of supply and demand. Technologies for technological rents pushes regional actors to seek out
that are simple to copy, and which can be moved easily complex knowledge possibilities. On the other hand, com-
over space, tend to be of little value and thus do not provide plex technologies remain out of reach for most because they
a source of long-run rents. Technologies that are more lack the diversity of capabilities out of which complex tech-
complex and difficult to imitate are more sticky in space: nologies are derived. The general solution to the ‘diversifi-
they hold promise as sources of competitive advantage for cation dilemma’ is for regional economies to develop their
the firms and regions in which they are generated. A stan- existing knowledge cores and to expand their technological
dard distinction separates forms of knowledge that are repertoires along related trajectories that lead toward more
codified from that which is tacit (Maskell & Malmberg, complex technologies.
1999). Because tacit knowledge is built around interperso- Based on this discussion, we formulate four hypotheses
nal contact and collections of routines embodied in individ- on complexity that will be tested for European regions, and
ual firms and localized networks, it tends to be rooted in which will be used in the fifth section to formulate a smart
place. Tacit knowledge is viewed as a primary source of specialization policy framework in which complexity is
competitive advantage for firms (Kogut & Zander, 1992) associated with the benefits of moving from one technology
and regions (Asheim & Gertler, 2005; Lawson & Lorenz, to another:
1999). Though competition renders much that is tacit
codifiable over time, continuous learning and recombinant Hypothesis 3: Regions are less likely to develop new specializ-
innovation provide sustained economic advantage to more ations in complex technological activities.
technologically dynamic regions.
Which regions hold the most valuable knowledge, Hypothesis 4: Regions are more likely to experience technological
especially that which is tacit and therefore difficult to access? growth in complex technological activities.

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1255

Hypothesis 5: Regions are more likely to develop new specializ- classes), and the links between them indicate their degree
ations in complex technological activities when related to their of relatedness. We compute relatedness (fi,j,t ) between
knowledge bases. each pair of technologies i and j for five different non-over-
Hypothesis 6: Regions are more likely to experience technological lapping periods: 1985–89, 1990–94, 1995–99, 2000–04
growth in complex technological activities when related to their and 2005–09. Figure 1 shows the relatedness network for
knowledge bases. the EU for the period 2005–09. Individual nodes are
coloured according to the five aggregate technological fields
defined by Schmoch (2008).4
DATA AND METHODS While Figure 1 displays the relatedness between tech-
nology pairings for the EU as a whole, it is also possible
A key challenge for smart specialization policy is to develop to identify the knowledge structure of individual regions
a framework to identify systematically technological oppor- within the EU. We are particularly interested in exploring
tunities in regions. We define technological opportunity as the knowledge cores of regions (Heimeriks & Balland,
the potential to develop critical capacity in a technology 2016), or how much of the technology produced within
that (1) draws on the specific knowledge bases of the region each NUTS-2 region (as captured by number of patents)
and (2) leads to technological upgrading. Technological tends to cluster around individual technological fields.
opportunities can be identified as those technologies in Thus, for each region r, we calculated the density of tech-
which a region does not yet possess critical capacity but nology production in the vicinity of individual technologies
that have a high degree of relatedness with the region’s i. Following Hidalgo et al. (2007) and Boschma et al.
existing knowledge base, and which will enhance the (2015), the density of knowledge production around a
knowledge complexity of the region. We use patent data given technology i in region r at time t is derived from
to measure relatedness between different sets of knowledge the technological relatedness fi,j,t of technology i to all
(captured by patent classes) and the complexity value of other technologies j in which the region has relative tech-
technologies. The relatedness measure provides an indi- nological advantage (RTA), divided by the sum of techno-
cator of the cost of moving from one technology to another, logical relatedness of technology i to all the other
while complexity provides a way of assessing the potential technologies j in the reference region (Europe) at time t:5
benefits of such movement.

We use the OECD-REGPAT1 database to identify
fij
j[r,j=i
technological fields and compute measures of relatedness RELATEDNESS DENSITY i,r,t =
∗100
fij
and knowledge complexity. The OECD-REGPAT derives j=i
from PATSTAT: it contains all patent applications to the
EPO between 1977 and 2011. Patent applications are RTA is a binary variable that assumes the value 1 when a
regionalized at the NUTS-2 level by inventor address.2 region possesses a greater share of patents in technology
Although the geographical coverage of the data spans all class i than the reference region (the EU as a whole); and
OECD countries, the focus is restricted to the EU-28 0 otherwise. A region r has RTA in production of techno-
plus Iceland, Norway and Switzerland. Patent applications logical knowledge i (r ¼ 1, … , n; i ¼ 1, … , k) such that
are classified in technological domains according to the RTAtr, i = 1 if:
International Patent Classification (IPC). We use Cam-

patentstr, i / i patentstr, i
bridge Econometrics to measure regional population, popu-


t .1
r patentsr, i /
t
r i patentsr, i
lation density and gross domestic product (GDP) per capita.
Figure 2 shows for all European regions the average relat-
Measuring relatedness between technologies edness density between existing technologies in a region
To measure technological relatedness between patent and all potential alternative technologies for the period
classes, we use the distribution of knowledge claims by 2005–09. The higher the region’s score on relatedness den-
IPC class on each patent across the EU as a whole, follow- sity, the closer, on average, its existing set of technologies to
ing Boschma et al. (2015) and Rigby (2015). This is done those technologies that are missing in the region. In other
by counting the number of EU patents for a given period words, it reflects an overall average score of the potential of
that contains a co-class pair, say i and j, and then standar- a region to develop new technologies. Figure 2 shows there
dizing this count by the total number of patents that record are huge differences in branching potential between
knowledge claims in IPC classes i and j. Relatedness is, countries and between regions in Europe. In general, Cen-
therefore, a standardized measure of the frequency with tral Europe (north Italy, south Germany, Austria) shows
which two IPC classes appear on the same patent. We high potentials to develop new technologies, in contrast
use a standardization method (Steijn, 2017) based on to many regions in southern and eastern Europe where
Van Eck and Waltman (2009), as implemented in the branching opportunities are much lower.
relatedness function of the EconGeo R package (Balland,
2017).3 The relatedness between technologies can be
formalized as a network, the knowledge space. The knowl- Measuring knowledge complexity
edge space is an n*n network where the individual nodes Which technologies and regions hold the most valuable
i (i ¼ 1, … , n) represent technological categories (IPC knowledge, especially that which is tacit and thus difficult

REGIONAL STUDIES
1256 Pierre-Alexandre Balland et al.

Figure 1. European knowledge space.

to access? A long literature explores the characteristics and renewals and litigation (Castaldi, Frenken, & Los, 2015;
value of innovations (Abernathy & Clark, 1985). Using Harhoff, Scherer, & Vopel, 2003). However, it remains
patent data, it is common to assess the quality and value unclear whether such measures accurately capture the
of knowledge using forward citations (Trajtenberg, 1990) characteristics of knowledge stocks that generate value.
or composite indicators including citations, family size, Maskell and Malmberg (1999) suggest that tacit forms of

Figure 2. Branching opportunities of European regions.

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1257

knowledge are valuable because they are difficult to repli- at the three-digit level for the empirical analysis. Table
cate. But this raises the problem of measuring the tacit A1 shows that the most complex technologies are related
dimension of knowledge. Kogut and Zander (1993) to digital communication, computer technology, biotech-
argue that complexity is a critical component of tacit nology and semiconductors in the period 2005–09. The
knowledge and thus propose complexity as a proxy measure least complex ones belong to textiles, mechanical elements,
of tacit knowledge. materials and metallurgy.
How then to measure knowledge complexity? From
Simon (1962), complexity is understood as a function of RELATEDNESS, KNOWLEDGE
the number of components out of which knowledge is con- COMPLEXITY AND DIVERSIFICATION IN
structed and the interdependence of those components. EUROPEAN UNION REGIONS
Fleming and Sorenson (2001) develop these ideas in a
measure of complexity for individual patents that captures This section presents the results of two econometric models
the difficulty of combining knowledge subsets represented that analyze: (1) the probability that a region develops a
by different technology subclasses in patent data. Hidalgo new RTA in a given technology (entry); and (2) the growth
and Hausmann (2009), in their analysis of product exports rate of technologies in regions, as measured by the growth
by country, offer an alternative measure of complexity that of patents in a given technology (growth). Following the
reflects the difficulty of mastering the capabilities required theoretical framework, the main variables of interest are
to export a particular commodity (indexed by the rarity of relatedness density between a given technology and the over-
exports of a given type), the diversity of capabilities held all technological portfolio of a region (i.e., proximity to
by different countries and the relatedness between them. existing technologies) and the knowledge complexity of tech-
An analogy with the board game Scrabble is often used nologies (i.e., potential upgrading of technological
to explicate understanding of this measure. Thus, those structure).
countries (players) that have rare capabilities (letters) tend We also include control variables: (1) four variables at
to produce more complex outputs (words), especially the regional level, i.e., population (log), GDP per capita,
when those countries possess broader sets of related capa- population density and technological stock (the total num-
bilities (a greater variety of letters that may be combined ber of technological claims in a region); and (2) one variable
to produce many different words). at the technological level, i.e., technological size (the total
Using patent data, Balland and Rigby (2017) explore number of technological claims of a technology). Popu-
the technological complexity of US metropolitan regions lation allows one to account for different sizes of regions.
between 1976 and 2005. They outline an eigenvector The level of economic development (GDP per capita) is
reformulation of the method of reflections proposed by an important driver of technological diversification (Petra-
Hidalgo and Hausmann (2009), that we employ below. lia et al., 2017). We included population density as a proxy
The starting point of the knowledge complexity index is for agglomeration economies (Boschma et al., 2015), and
the binary-valued network that connects regions to the technological stock as a proxy for the number of ideas
knowledge classes in which they have an RTA, rep- that could potentially be combined in a given region or
resented as an n × k two-mode binary adjacency matrix for a given technology. Table A2 in Appendix A in the
(M). The matrix (M) has dimension n ¼ 282 regions supplemental data online provides summary statistics for
(NUTS-2) by k ¼ 33 technological domains (two-digit the variables.
level),6 as proposed by Schmoch (2008). Following our All specifications are estimated at the region-technol-
calculations of relatedness, we divide the years for which ogy level. We use a linear probability model (LPM) to
we have patent data into six periods of five years each, assess the probability that a region specializes in a new
and we construct the two-mode binary adjacency matrix technological field (entry) using the following specification
(M) for each of the periods: 1985–89, 1990–94, 1995– (a similar econometric model is used for the growth
99, 2000–04 and 2005–09. models):
To construct our index of knowledge complexity, the
matrix (M) is row standardized along with its transpose ENTRY r,i,t = b1 RELATEDNESS DENSITY r,i,t−1
(M T ). The product matrix (B) = (M T ∗M) is a square + b2 KNOWLEDGE COMPLEXITY i,t−1
matrix with dimension equal to the number of technologi- + b3 REGIONSr,t−1 + b4 TECHSi,t−1
cal classes (33). The technological complexity index (TCI) + wr + at + 1r,i,t
for each of these classes is given by the elements of the
second eigenvector Q  of matrix (B). These elements are We expect a positive coefficient for relatedness density in
standardized as: both the entry model (hypothesis 1) and the growth
model (hypothesis 2), and a negative coefficient for com-
 − kQl
Q  plexity in the entry model (hypothesis 3) and a positive
TCIi =
stdev (Q) one in the growth model (hypothesis 4). The baseline spe-
cification is a two-way fixed-effects model where wr is a
The technological classification shown in Table A1 in region fixed effect, at is a time fixed effect and 1i,c,t is a
Appendix A in the supplemental data online corresponds regression residual. Since errors are correlated within
to the two-digit level of IPC, and is further broken down regions and technologies, the regression results presented

REGIONAL STUDIES
1258 Pierre-Alexandre Balland et al.

in all regression outputs are adjusted for clustering at the technologies and tends to solve the ‘diversification
region and technology level. Our panel consists of data dilemma’: when relatedness is high, regions are more likely
for 282 NUTS-2 regions and 617 technologies (IPC) to diversify into complex technologies, confirming hypoth-
over the period 1985–2009. We average the data over esis 5.
non-overlapping five-year periods, denoted by t. To dam-
pen potential endogeneity issues, all the independent vari- Growth models – full sample
ables are lagged by one period, denoted by t – 1. To investigate further the impact of technological related-
ness and knowledge complexity on sustainable specializ-
Entry models – full sample ations, we now turn to technological growth models in
In all entry7 models in Table 1,8 we find that relatedness Table 3.
density has a positive and significant effect on the probability We find that both relatedness and complexity have a
that a region specializes (RTA > 1) in a new technological positive, strong and statistically significant impact on tech-
field, which is consistent with other findings (Balland, nological growth at the regional level, confirming hypoth-
2016; Boschma et al., 2015; Rigby, 2015). The effect of eses 2 and 4. This pattern holds across all econometric
relatedness is also strong: an increase of 10% in relatedness specifications. We find that an increase of relatedness den-
density is associated with a 23–26% relative increase in the sity by 10 points is associated with an increase in techno-
probability of entry.9 The effect of knowledge complexity logical growth by about 2.0–4.64%. In contrast to the
on entry is more ambiguous. The sign for knowledge com- entry models, the impact of knowledge complexity is now
plexity is positive and significant in the simplest specification in the same order of magnitude. An increase in complexity
(model 2), still positive but insignificant when regional and by 10 points is now associated with an increase in techno-
technology controls are added (model 4), and negative and logical growth by about 1.2–2.0%. This means that once
significant in the fixed-effect model (model 5). Overall, the regions manage to diversify into more complex technol-
effect is always small: when the complexity of a technology ogies, they tend to experience higher technological growth.
goes up by 10 points (on a scale of 0–100), the relative like- Apparently, the most difficult step for regions is to enter
lihood that a region specializes in this technology increases by complex technological fields, but the rewards of techno-
0.30% in the simplest specification and decreases by 0.35% in logical upgrading seem to be fairly strong.
the fixed-effects specification. This result might indicate
the ‘diversification dilemma’ mentioned above: complex
Growth models for high and low levels of
knowledge is more attractive (a positive effect on entry) but
relatedness
at the same time also more difficult to produce (a negative
As for the entry models, to test hypothesis 6 we now split
effect on entry). Therefore, the relationship between knowl-
the sample into observations with high or low levels of
edge complexity and new specialization is not linear and
relatedness. The results are presented in Table 4.
might be region specific (Petralia et al., 2017; Pinheiro,
We find results in line with the entry models. Regions are
Alshamsi, Hartmann, Boschma, & Hidalgo, 2018).
more likely to experience technological growth if they
specialize in complex technologies that are related to their
Entry models for high and low levels of
capabilities, confirming hypothesis 6. When relatedness is
relatedness
high, i.e., when models include only the top 10% of
To investigate this further and test hypothesis 5, we split
region-technology observations in terms of relatedness den-
the sample between observations with high levels of relat-
sity, complexity has a positive, significant impact on techno-
edness and observations with a low level of relatedness.
logical growth. When the complexity of a technology goes up
The results are presented in Table 2.
by 10 points, technological growth increases by about 2.27–
The main result is that the effect of complexity on entry
3.25%. When relatedness is low, however, i.e., when models
appears to be conditional on the level of relatedness. When
include only the bottom 10% of region-technology obser-
relatedness is high, i.e., when models include only the top
vations in terms of relatedness density, complexity tends to
10% of region-technology observations in terms of related-
have a negative impact, though it is non-significant. In the
ness density,10 knowledge complexity has a positive, sig-
fixed-effect specification (for low complexity), the negative
nificant impact on the development of new technologies.
coefficient is statistically significant, though its impact is
When the complexity of a technology goes up by 10 points,
small (–0.3% growth for a 10-point increase). Therefore,
the likelihood that a region will develop this technology
relatedness conditions technological growth in regions,
increases by 1.3% in the simplest specification, which is
suggesting that regions need to upgrade their technological
five times higher than what was found in the specifications
structure by building on related, pre-existing capabilities.
unconditional to relatedness. In the more conservative
fixed-effects specification, complexity is still positive and
significant. What is interesting is that when relatedness is FRAMEWORK FOR SMART
low, i.e., when models include only the bottom 10% of SPECIALIZATION
region-technology observations in terms of relatedness
density, complexity never has a significant impact on the We now formalize the implications of these results into a
development of new technologies. What these results indi- framework for smart specialization. We present a frame-
cate is that relatedness conditions access to complex work to identify systematically technological opportunities

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1259

Table 1. Entry models – full sample.

Dependent variable: Entry (¼ 1) | 1990–2009
Baseline Complexity Controls Full model Full model (FE)
(1) (2) (3) (4) (5)
Constant 0.1632872*** 0.1632945*** 0.1498963*** 0.1639320*** –0.0117608
(0.0005543) (0.0005543) (0.0005242) (0.0005722) (0.0255653)
Relatedness Density 0.0042477*** 0.0042494*** 0.0041635*** 0.0037696***
(0.0000388) (0.0000388) (0.0000419) (0.0000449)
Knowledge Complexity 0.0000459* 0.0000354 –0.0000575**
(0.0000199) (0.0000211) (0.0000215)
Population (log) 0.0322163*** 0.0172538*** –0.1155466***
(0.0008129) (0.0008150) (0.0148724)
GDP per capita 0.0000020*** 0.0000005*** 0.0000017***
(0.0000001) (0.0000001) (0.0000003)
Population Density –0.0000090*** –0.0000030*** 0.0000198
(0.0000007) (0.0000007) (0.0000122)
Technological stock –0.0000022*** –0.0000022*** –0.0000023***
(0.0000001) (0.0000001) (0.0000002)
Technological size 0.0000004** 0.00000005 0.0000013***
(0.0000002) (0.0000002) (0.0000002)

Region fixed effects No No No No Yes

Time fixed effects No No No No Yes
Observations 498,785 498,785 466,814 466,814 466,814
R2 0.0303005 0.0303106 0.0040004 0.0306804 0.0371538
Adjusted R 2 0.0302985 0.0303068 0.0039897 0.0306659 0.0366399
Notes: The dependent variable entry equals 1 if a region r gains a new relative technological advantage (RTA) in a given technology i during the correspond-
ing five-year window; and 0 otherwise. All independent variables are mean centred and lagged by one period. Heteroskedasticity-robust standard errors
(clustered at the region and technology level) are shown in parentheses.
Coefficients are statistically significant at the *p < 0.05, **p < 0.01 and ***p < 0.001 level.

for regions, based on the relatedness density and knowledge technologies in which the region does not have an RTA.
complexity of individual technologies. This framework Thus, for each region, the quadrants in the x–y space of
reflects the fact that regions have different sets of core com- Figure 3 represent the relatedness between the region’s
petences, and thus a top-down ‘one-size-fits-all’ model knowledge base and each technology for which the RTA
built around picking champions is unlikely to be successful. does not exist in the region, and a measure of how the
Instead, we offer a model for smart specialization built development of the RTA in each of these technologies
around relatedness that demands a bottom-up approach. would shift the region’s overall knowledge complexity.
This framework could be used to identify hidden techno- Hausmann et al. (2014) proposed a similar strategy to
logical opportunities, avoid unrealistic investments and identify future economic opportunities for countries.
assess how entrepreneurial discoveries that emerge from
regional actors fit in the overall economic landscape of
the region.
This framework is presented in Figure 3. Operationali-
zation requires identification of the knowledge base of
regions to which a specific set of technological possibilities
is tailored. The framework uses the relatedness measure to
map technologies in which the region does not yet possess
an RTA but which are relatively close to the region’s exist-
ing technology core. Relatedness measures allow precise
mapping of the accessibility of new regional growth paths
and thus an accounting of the costs of their deployment.
An index of the relative ease with which a region might
be able to develop an RTA in a new technology is defined
along the x-axis. The y-axis reflects the complexity of all Figure 3. Framework for smart specialization.

REGIONAL STUDIES
1260 Pierre-Alexandre Balland et al.

Table 2. Entry models by level of relatedness.

Dependent variable: Entry (¼ 1) | 1990–2009
High Low High Low High Low
relatedness relatedness relatedness relatedness relatedness relatedness
(1) (2) (3) (4) (5) (6)
Constant 0.3669312*** 0.0309562*** 0.3614363*** 0.0405249*** 0.2306594 0.0903739**
(0.0023488) (0.0006430) (0.0026666) (0.0009141) (0.1847726) (0.0327663)
Knowledge 0.0004628*** –0.0000389 0.0002671* –0.0000062 0.0002526* –0.0000359
Complexity (0.0001007) (0.0000272) (0.0001127) (0.0000395) (0.0001124) (0.0000419)
Population (log) 0.0433384*** 0.0224518*** –0.0657516 0.0488210*
(0.0044990) (0.0014247) (0.0934813) (0.0200723)
GDP per capita 0.0000004 0.0000015*** 0.0000016 0.0000002
(0.0000004) (0.0000001) (0.0000016) (0.0000005)
Population 0.0000016 –0.0000057*** 0.0000252 –0.0000202
Density (0.0000034) (0.0000015) (0.0000569) (0.0000281)
Technological –0.0000026*** 0.0000002 –0.0000036*** 0.0000003
stock (0.0000004) (0.0000002) (0.0000007) (0.0000004)
Technological 0.0000088*** 0.0000021** 0.0000139*** 0.0000018*
size (0.0000012) (0.0000007) (0.0000013) (0.0000007)

Region fixed No No No No Yes Yes

effects
Time fixed No No No No Yes Yes
effects
Observations 42,164 72,557 34,309 47,029 34,309 47,029
R2 0.0005119 0.0000281 0.0053447 0.0127176 0.0584039 0.0334063
Adjusted R 2 0.0004882 0.0000143 0.0051707 0.0125916 0.0515479 0.0282820
Notes: High relatedness models only include the top 10% region–technology observations in terms of relatedness density. Low relatedness models only
include the bottom 10% region–technology observations in terms of relatedness density. The dependent variable entry equals 1 if a region r gains a
new relative technological advantage (RTA) in a given technology i during the corresponding five-year window; and 0 otherwise. All independent variables
are mean centred and lagged by one period. Heteroskedasticity-robust standard errors (clustered at the region and technology level) are shown in parenth-
eses.
Coefficients are statistically significant at the *p < 0.05, **p < 0.01 and ***p < 0.001 level.

Our smart specialization policy framework (Figure 3) policy. An attractive smart specialization approach is
allows policy-makers to weigh the relative ease of develop- the one that supports potential technologies that occupy
ing a new technology in a region in relation to the gains in the north-east quadrant of Figure 3, for these technol-
knowledge complexity. Technologies with high relatedness ogies promise above-average returns and can be devel-
represent new technological possibilities that are closely oped at relatively low risk. We refer to that as the
connected to the current knowledge base of the region. ‘high road’ policy. In contrast, a policy focus on potential
These technologies might reasonably be developed at rela- technologies in the south-west quadrant represent tech-
tively low cost and thus represent a low-risk strategy for the nologies that are far removed from the existing knowl-
region. Technologies with a low level of relatedness in edge base of the region and their development is thus
relation to a region’s existing knowledge base represent risky. Alongside the risk, these technologies are unlikely
new techniques that are distant from the region’s existing to raise the complexity value of the regional knowledge
specializations and for which development is likely to be base. Therefore, we refer to such a policy as a ‘dead-
more costly, embodying greater risk. Each of these poten- end’ policy. The north-west quadrant reflects a poten-
tial technology types is associated with a change in the tially high-benefit strategy, but it is not rooted in regional
region’s complexity or the value of the knowledge gener- capabilities and therefore likely to fail due to high risks.
ated in the region. Potential new technologies that have This policy aims at developing new technologies from
high complexity are likely to add considerably to the com- scratch, which we therefore refer to as a ‘casino’ policy.
plexity of the region and the overall value of the region’s The fourth policy option focuses on technologies in the
knowledge assets, in contrast to potential new technologies south-east quadrant, which reflects a relatively low-risk
with low complexity. strategy because it builds on related capabilities. We
The four quadrants of Figure 3 highlight the cost– dub this the ‘slow-road’ policy because these potential
benefit trade-off that undergirds smart specialization technologies offer few expected benefits.

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1261

Table 3. Growth models – full sample.

Dependent variable: Technological growth | 1990–2009
Baseline Complexity Controls Full model Full model (FE)
(1) (2) (3) (4) (5)
Constant 13.7038900*** 13.7207500*** 13.5505400*** 13.5218000*** 73.6810700***
(0.1707395) (0.1707130) (0.1773364) (0.1767446) (7.4333840)
Relatedness Density 0.4642356*** 0.4650504*** 0.3519811*** 0.2038730***
(0.0101046) (0.0101016) (0.0113171) (0.0119662)
Knowledge Complexity 0.2083142*** 0.1811793*** 0.1236107***
(0.0079042) (0.0082521) (0.0079222)
Population (log) 15.6830000*** 13.9970400*** 57.4033100***
(0.2933130) (0.2957986) (4.4582030)
GDP per capita 0.0004739*** 0.0003251*** 0.0000061
(0.0000201) (0.0000205) (0.0000892)
Population Density –0.0039671*** –0.0033924*** –0.0031405
(0.0002246) (0.0002242) (0.0032979)
Technological stock –0.0004838*** –0.0005061*** –0.0051211***
(0.0000325) (0.0000326) (0.0001068)
Technological size 0.0010760*** 0.0007250*** 0.0016523***
(0.0000560) (0.0000565) (0.0000584)

Region fixed effects No No No No Yes

Time fixed effects No No No No Yes
Observations 556,721 556,721 521,175 521,175 521,175
R2 0.0039793 0.0055811 0.0072392 0.0103975 0.0671133
Adjusted R 2 0.0039775 0.0055776 0.0072297 0.0103842 0.0666674
Notes: The dependent variable growth corresponds to the rate of technological growth (growth in the number of claims) of a technology i in a region r from
period t to period t + 1. All independent variables are mean centred and lagged by one period. Heteroskedasticity-robust standard errors (clustered at the
region and technology level) are shown in parentheses.
Coefficients are statistically significant at the *p < 0.05, **p < 0.01 and ***p < 0.001 level.

Therefore, the core idea of the smart specialization fra- Capello, 2013): a core leading region (Île-de-France), a
mework is that region-specific capabilities define not only high-tech region (Noord-Brabant, the Netherlands), an
the set of opportunities to develop new growth paths but old industrial region (Lancashire, UK) and a lagging,
also the limits on those choices. By operationalizing this peripheral region (Extremadura, Spain). We present the
idea, we can eschew the one-size-fits-all policy model four cases in Figures 4(a–d), following the same structure
and offer place-based policy adapted to the particular con- of Figure 3, with relatedness density on the horizontal
ditions of regions (Tödtling & Trippl, 2005). In the past, axis and knowledge complexity on the vertical axis. The
regional technology policy was too often focused only on nodes (circles) represent potential technologies (or new
the y-axis of Figure 3. Such a policy of picking winners, growth paths) in which the region does not exhibit an
of building ‘cathedrals in the desert’, paid insufficient atten- RTA in the period 2005–09. The size of the nodes rep-
tion to whether host regions had the technological capacity resents the total number of patents in each technology
to support such ventures. Balancing the dimensions of class, though note that all such classes are smaller than
relatedness and knowledge complexity provides an evi- might be expected in the regions, based on the share
dence-based platform on which to develop smart specializ- of patents in these classes at the European level. Colours
ation, a platform that is more attuned to region-specific show the one-digit classification of the technological
capabilities and thus less likely to lead to duplication and classes. The distribution of the nodes in each figure
waste of scarce public and private resources. makes clear that not all regions are in the same situation
to build new growth trajectories: the size and content of
REGIONAL CASES TO ILLUSTRATE the list of choices, as well as their accessibility, vary from
FURTHER THE SMART SPECIALIZATION one region to another.
FRAMEWORK The Île-de-France is the region with most patent
applications. Its diversified portfolio of activities reveals
For illustration purposes, we apply the smart specializ- the existence of a large set of capabilities the region
ation framework to four types of NUTS-2 regions that can rely on to branch out towards new activities. Figure 4
can be considered representative cases (Camagni & (a) shows that the Île-de-France has a relatively high

REGIONAL STUDIES
1262 Pierre-Alexandre Balland et al.

Table 4. Growth models by level of relatedness.

Dependent variable: Technological growth | 1990–2009
High Low High Low High Low
relatedness relatedness relatedness relatedness relatedness relatedness
(1) (2) (3) (4) (5) (6)
Constant 53.7695700*** –6.0870230*** 49.6868700*** –8.4233190*** –31.0214900 –10.3398200
(0.7609844) (0.1377738) (0.8003883) (0.2020142) (60.1171500) (5.9970820)
Knowledge 0.3256727*** –0.0096395 0.2581051*** –0.0107926 0.2276855*** –0.0305587**
Complexity (0.0340018) (0.0069824) (0.0361123) (0.0104075) (0.0344880) (0.0101187)
Population 28.4596000*** –3.3740520*** –26.7761900 –2.6496500
(log) (1.4859160) (0.3181761) (31.0434900) (4.0115730)
GDP per 0.0001111 –0.0003621*** 0.0037211*** 0.0001280
capita (0.0001073) (0.0000220) (0.0005146) (0.0000970)
Population –0.0048569*** 0.0006342* –0.0729367*** –0.0052360
Density (0.0009146) (0.0002874) (0.0144477) (0.0046896)
Technological –0.0020091*** –0.0001944* –0.0080889*** –0.0015154***
stock (0.0001021) (0.0000909) (0.0003040) (0.0002594)
Technological –0.0012314*** –0.0013100*** 0.0001989 –0.0011047***
size (0.0001257) (0.0002712) (0.0001165) (0.0002697)

Region fixed No No No No Yes Yes

effects
Time fixed No No No No Yes Yes
effects
Observations 63,797 74,199 54,992 48,659 54,992 48,659
R2 0.0017529 0.0000365 0.0115695 0.0155464 0.1329042 0.0509171
Adjusted R 2 0.0017372 0.0000230 0.0114617 0.0154250 0.1289760 0.0460551
Notes: High relatedness models only include the top 10% region–technology observations in terms of relatedness density. Low relatedness models only
include the bottom 10% region–technology observations in terms of relatedness density. The dependent variable growth corresponds to the rate of tech-
nological growth (growth in the number of claims) of a technology i in a region r from period t to period t + 1. All independent variables are mean centred
and lagged by one period. Heteroskedasticity-robust standard errors (clustered at the region and technology level) are shown in parentheses.
Coefficients are statistically significant at the *p < 0.05, **p < 0.01 and ***p < 0.001 level.

relatedness with a large number of technologies in which technologies with relatively high relatedness is smaller.
it is not specialized, many belonging to fields such as While for the Île-de-France the low-risk options are
electrical engineering and chemistry which are already in high- and low-complexity technologies, in Noord-
established in the region, and to a lesser extent to mech- Brabant they tend to be concentrated in high-complexity
anical engineering and instruments. Thus, the Île-de- technologies (such as communication) in the electrical
France has a wide range of options to diversify at rela- engineering domain. Investing in other complex
tively low risk. They involve both high- and low-com- technology domains such as chemistry (in the north-
plexity technologies, making the second criterion of our west quadrant) would imply a radical diversification
framework relevant when choosing the direction by dis- strategy that is very risky, given the current, relatively
criminating among the low-risk alternatives. Since more specialized, technological structure of the Noord-Brabant
complex technologies tend to have higher upgrading economy.
and growth potential, the Île-de-France could easily Lancashire, in the north-west of England, is a typical
implement a ‘high-road’ smart specialization policy that case of an old industrial region. Its patenting activity is far
targets the development of potential technologies in the below the EU average, and less diversified than the previous
north-east quadrant. two cases. Figure 4(c) shows that Lancashire, as Noord-
Noord-Brabant is a typical high-tech region, located Brabant, has many diversification options related to existing
around Eindhoven, the Netherlands. In spite of its rela- technologies. From the complexity angle, however, the story
tively small size, the region ranked as one of the top 10 is different. Almost all potential technologies in Lancashire
most innovative regions in the EU. Figure 4(b) shows are closely related to low-complexity domains mostly in
the technologies in which Noord-Brabant could diversify mechanical engineering, while complex domains, with
by redeploying its existing capabilities. The menu of high-growth potential, are difficult to access given the cur-
potential low-risk paths is less rich and more limited rent strengths of the region. Thus, for Lancashire, there
than for the Île-de-France, i.e., the number of potential are no low-risk–high-benefit smart specialization options.

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1263

Figure 4. Application of the framework: (a) Île-de-France (France, FR10); (b) Noord-Brabant (the Netherlands, NL41); (c) Lanca-
shire (UK, UKD4); and (d) Extremadura (Spain, ES43).

Targeting most electrical engineering technologies might Extremadura could go for a ‘casino’ strategy focusing on
reflect a ‘casino’ policy, while policy focus on most mechan- complex technologies because the expected benefits are
ical engineering technologies comes down to a ‘slow-road’ higher. However, complex technologies require a large set
approach. A smart specialization approach for Lancashire of capabilities that makes such successful long jumps for
could target technologies that are relatively complex peripheral regions near to impossible. In our framework,
(upgrading its local economy), but still related to existing every region can find a smart specialization policy by
capabilities, such as some instruments and a few chemical adjusting its ambitions to the local economic context. In
technologies in the centre of the graph. the case of Extremadura, the subset of chemical technol-
Finally, Extremadura, in western Spain, is a typical per- ogies can be a reasonable starting point to accumulate
ipheral region with a weak technological base (Figure 4(d)). knowledge and capabilities that can further be redeployed
This severely limits the ability of the region to develop new into increasingly complex technologies.
growth paths. In the terminology of our framework, the
region has many policy options in the high-risk quadrants, DISCUSSION AND CONCLUSIONS
with both high- and low-expected benefits, but none in the
low-risk quadrants, because most missing technologies This paper has attempted to tackle the perceived lack of
show low relatedness. Peripheral regions provide one of a strong theoretical and empirical foundation for smart
the most complicated cases to build an effective smart specialization policy in Europe. First, it constructed a
specialization policy, which is also echoed in writings on theoretical framework for smart specialization built
smart specialization policy (Foray, 2015; Morgan, 2015). around the concepts of relatedness and knowledge

REGIONAL STUDIES
1264 Pierre-Alexandre Balland et al.

complexity. Second, it operationalized this framework Cohesion Policy that aims at reducing disparities between
empirically by calculating relatedness between technol- EU regions, and which has repeatedly been mentioned as
ogies using EPO patent data and measuring the knowl- one of the key challenges in EU regional innovation
edge complexity of technologies using network-based policy (McCann & Ortega-Argilés, 2013, 2015; Mor-
techniques. Third, it used these measures to provide gan, 2015). This is because non-technological knowledge
new evidence on how EU regions develop new technol- (of both high and low complexity) also provides diversifica-
ogies. We find that relatedness has a positive effect on tion opportunities for regions, especially in the more
technological diversification within regions. Diversifying peripheral parts of the EU (Asheim, Boschma, & Cooke,
into complex technologies is difficult for many regions, 2011).
though it is easier when such technologies are more clo- We need to think more deeply about how to tackle the
sely related to the existing knowledge core of the region. inherent tension within smart specialization policy between
And regions tend to grow more if they specialize in prioritizing and selecting activities based on regional
complex technologies related to existing technologies in potentials (as in the proposed framework), on the one
the region. hand, and reliance on the entrepreneurial discovery process
Based on these theoretical and empirical underpin- in which this selection process is completely decentralized,
nings, we proposed a policy framework for smart specializ- bottom-up and process-led, on the other hand (Coffano &
ation that highlights the potential risks and rewards for Foray, 2014). One potential way to solve this tension is,
regions of adopting alternative diversification strategies. first, to identify diversification opportunities in each region
We showed how potential risks of smart specialization based on their scores on relatedness and complexity, after
may be assessed using the concept of relatedness, and which, within that range of opportunities, the entrepre-
how potential benefits can be derived from estimates of neurial discovery process will unfold, in which a range of
the complexity of technologies. Our policy framework local actors will decide which activities to target and assist
incorporates the key logic of smart specialization, that is, (Boschma & Gianelle, 2013). Or the other way around, a
a bottom-up policy approach based on the idea that regions range of activities will first be selected through the entre-
should leverage their existing capabilities to develop and preneurial discovery process, which are then assessed
secure comparative advantage in related high-value-added within the diversification opportunities of each region
activities. We underlined further the relevance of such a identified by our framework.
policy framework and the need for a region-specific focus Our smart specialization framework still has to be
by identifying the diversification opportunities in different framed in relation to territorial terms (Iacobucci & Guz-
types of regions. zini, 2016; Sörvik, Midtkandal, Marzocchi, & Uyarra,
Although this paper provides some important pieces to 2016; Thissen, van Oort, Diodato, & Ruijs, 2013) to
the puzzle of smart specialization, we are still far from a account for non-local linkages to avoid regional lock-in
comprehensive framework. We have not focused on the (Bathelt et al., 2004), as these give regions access to
design and implementation of smart specialization policy, complementary capabilities elsewhere (Miguelez & Mor-
and what exact role public agents need to play (Capello eno, 2017; Tavassoli & Carbonara, 2014). One example
& Kroll, 2016; Iacobucci, 2014; Kroll, 2015; Matti, Con- is the recent focus on trade networks and value chains
soli, & Uyarra, 2017; Moodysson, Trippl, & Zukauskaite, and how to incorporate these into smart specialization
2016; Valdaliso, Magro, Navarro, Aranguren, & Wilson, strategies (Radosevic & Ciampi Stankova, 2015).
2014). Moreover, our framework focuses primarily on the Another challenge is how smart specialization policy
supply side, and less so on the demand side (such as internal can assist in linking peripheral regions to the outside
or external demand, and market access), at least not expli- world to compensate for lack of local capabilities (Fitjar
citly (Montresor & Quatraro, 2017; Tanner, 2014). We & Rodríguez-Pose, 2011; Grillitsch & Nilsson, 2015).
now discuss some other important issues that still need to And we need to think how smart specialization policy
be tackled. can incorporate the crucial importance of inflows of
This paper has focused on technological knowledge, skilled labour for related diversification and structural
and only that part of knowledge embodied in patent change in regions (Neffke, Hartog, Boschma, & Hen-
activity. It is crucial to emphasize that smart specialization ning, 2018).
policy is about diversifying from regional capabilities in Another building block of smart specialization policy
general, not only from knowledge captured by patents. is the need for the right institutional governance structure
The proposed smart specialization policy framework to be in place at the local level (Grillitsch, 2016;
could and should, therefore, assess diversification options McCann, van Oort, & Goddard, 2017; Morgan, 2017).
for regions that include other forms of knowledge and capa- When the quality of institutional governance differs
bilities not captured by patents, such as products, industries, greatly between regions, as it does in Europe (Rodrí-
scientific disciplines and jobs (Cortinovis, Xiao, Boschma, & guez-Pose, di Cataldo, & Rainoldi, 2014), it is also likely
van Oort, 2017; Unterlass et al., 2015). It could also be to affect their ability to move into new and more complex
applied to tacit and symbolic forms of knowledge technologies. This might require a local institutional con-
through targeted educational programmes and college text (governance, social capital) that facilitates the coordi-
degrees. This is especially important for bringing smart nation and combination of a wide range of capabilities
specialization policy more in line with the objective of (Cortinovis et al., 2017). Especially, peripheral regions

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1265

suffer from institutional weaknesses that might limit NOTES

smart specialization strategy (Karo & Kattel, 2015;
Rodríguez-Pose & Wilkie, 2015), such as a traditional 1. September 2015 Edition. OECD ¼ Organisation for
approach to governance (no experimentation culture), Economic Co-operation and Development.
weak administrative capabilities, a low quality of govern- 2. Fractional counts cleaned by inpadoc families of PAT-
ment, tight connections with local vested players, weak STAT. NUTS ¼ Nomenclature of Territorial Units for
entrepreneurial capacities and the absence of a local cul- Statistics.
ture of collaboration. 3. In this paper we used EconGeo, version 1.3 (see
Lastly, there is an ongoing discussion whether smart https://github.com/PABalland/EconGeo).
specialization policy should focus on disruptive activities, 4. Source: WIPO IPC Technology Concordance Table.
as echoed in recent pleas for mission-oriented policies 5. Our notion of relatedness is inspired by the related var-
such as curing cancer or greening the economy (Frenken, iety literature (Frenken et al., 2007), but we apply it to the
2016; Mazzucato, 2013). While there is little under- topic of regional (related) diversification, instead of regional
standing about how mission-oriented policies should growth. Some papers on (related) diversification have used
actually be designed (except for setting specific targets related variety as an explanatory factor for diversification in
and directions for future development), our framework regions (e.g., Xiao, Boschma, & Andersson, 2018). How-
shows how risky such demand-led policy targeting very ever, a positive correlation between related variety and the
complex and unknown technologies will be when formu- entry of new technologies at the regional level does not
lated and implemented without embedding it in the necessarily imply that the new technologies are related to
appropriate technological and institutional context in existing regional technologies, as measured by related var-
countries and regions. When ignoring that, it is not dif- iety. Therefore, we make use of relatedness density, as
ficult to foresee that such mission-oriented policies will this indicator measures directly the degree of relatedness
fall into the same traps that smart specialization policy between existing technologies in the region and missing
aims to avoid, such as picking fashionable activities technologies in the region.
from scratch, and duplicating major research and inno- 6. The two-mode network used to compute complexity is
vation investments. This is not to deny the fact that pub- based on the two-digit technological domains and not IPC
lic policy can play a key role in initiating transformative classes because of the high size heterogeneity of IPC
change when fundamental uncertainty is the rule, but classes. In fact, using IPC classes as nodes in the region–
only when embedded in an appropriate territorial context technology network would lead to a noisy complexity esti-
(Boschma, 2017; Montresor & Quatraro, 2017; Uhlbach, mation because some IPC classes are very small (fewer than
Balland, & Scherngell, 2017; Unterlass et al., 2015). Our 20 patent applications a year), while others are very large
proposed framework may be instrumental to make such (more than 3000 patent applications a year). Using the
informed choices. broader and more homogenous classification proposed by
Schmoch (2008) allows one to circumvent this problem.
The EU complexity ranking is consistent with that in the
ACKNOWLEDGEMENTS United States described by Balland and Rigby (2017). As
a robustness check, we also computed complexity using
The authors thank Christopher Esposito and Wolf- the model proposed by Fleming and Sorenson (2001)
Hendrik Uhlbach for their excellent research assistance. and found no major differences.
7. In the entry models, we cannot distinguish between
diversification induced by new or existing firms, by small
DISCLOSURE STATEMENT or big firms, or by local or non-local firms because firm-
level data are not available. See Neffke et al. (2018) for
No potential conflict of interest was reported by the
such a firm-level analysis of regional diversification.
authors.
8. The control variables GDP per capita, population and
technological size have a positive and significant impact on
FUNDING technological entry in most models, while the regional
technological stock and population density tend to have a
Financial support from the Regional Studies Association negative and significant impact.
(RSA) through the Early Career Grant awarded to 9. In the baseline model 1, for instance, the unconditional
Pierre-Alexandre Balland is gratefully acknowledged. The probability of entry is around 16% (as all independent vari-
authors also thank the European Commission for its finan- ables are mean centred, the constant is equal to the uncon-
cial support through the project ‘Technological Diversifica- ditional probability of entry ¼ 0.1632872). An increase by
tion of Europe’s Regions’. Funding support by the JPI 10% in relatedness density (relatedness density ranges from
Urban Europe project ‘Resilient Cities: Industrial Network 0 to 100%) increases the relative probability of entry by
and Institutional Perspectives on Economic Growth and (0.0042477*10)/0.1632872 ¼ 26%. In the most conserva-
Well-Being’ [grant number 438-13-406] is also acknowl- tive model (fixed effect, model 5), we find an increase in
edged by Ron Boschma. the relative probability of entry of about (0.0037696*10)/

REGIONAL STUDIES
1266 Pierre-Alexandre Balland et al.

0.1639320 ¼ 23%. The intercept in a fixed-effect model Boschma, R., Balland, P. A., & Kogler, D. (2015). Relatedness and
cannot be interpreted as the unconditional probability of technological change in cities: The rise and fall of technological
entry by definition. The unconditional probability of knowledge in US metropolitan areas from 1981 to 2010.
Industrial and Corporate Change, 24(1), 223–250. doi:10.1093/
entry for this model can be found as the intercept of
icc/dtu012
model 4. Boschma, R., & Gianelle, C. (2013). Regional branching and smart
10. Therefore, relatedness density is excluded from this specialization policy (Policy Note November 2013). Seville:
model. Institute for Prospective Technological Studies (IPTS).
Breschi, S., Lissoni, F., & Malerba, F. (2003). Knowledge-related-
ness in firm technological diversification. Research Policy, 32(1),
ORCID 69–87. doi:10.1016/S0048-7333(02)00004-5
Camagni, R., & Capello, R. (2013). Regional innovation patterns
Pierre-Alexandre Balland http://orcid.org/0000-0003- and the EU regional policy reform: Toward smart innovation pol-
2787-4376 icies. Growth and Change, 44(2), 355–389. doi:10.1111/grow.
Joan Crespo http://orcid.org/0000-0003-4053-8158 12012
Capello, R., & Kroll, H. (2016). From theory to practice in smart
specialization strategy: Emerging limits and possible future tra-
REFERENCES jectories. European Planning Studies, 24(8), 1393–1406. doi:10.
1080/09654313.2016.1156058
Abernathy, W. J., & Clark, K. B. (1985). Innovation: Mapping the Castaldi, C., Frenken, K., & Los, B. (2015). Related variety, unre-
winds of creative destruction. Research Policy, 14(1), 3–22. lated variety and technological breakthroughs: An analysis of
doi:10.1016/0048-7333(85)90021-6 US state-level patenting. Regional Studies, 49(5), 767–781.
Asheim, B., Boschma, R., & Cooke, P. (2011). Constructing regional doi:10.1080/00343404.2014.940305
advantage: Platform policies based on related variety and differen- Chatterjee, S., & Wernerfelt, B. (1991). The link between resources
tiated knowledge bases. Regional Studies, 45(7), 893–904. doi:10. and type of diversification: Theory and evidence. Strategic
1080/00343404.2010.543126 Management Journal, 12(1), 33–48. doi:10.1002/smj.
Asheim, B., Grillitsch, M., & Trippl, M. (2016). Smart specialization 4250120104
as an innovation-driven strategy for economic diversification: Coffano, M., & Foray, D. (2014). The centrality of entrepreneurial
Examples from Scandinavian regions (Papers in Innovation discovery in building and implementing a smart specialisation
Studies No. 2016/23) (pp. 73–97). Lund: Centre for strategy. Italian Journal of Regional Science, 13, 33–50. doi:10.
Innovation, Research and Competence in the Learning 3280/SCRE2014-001003
Economy (CIRCLE), Lund University. Cortinovis, N., Xiao, J., Boschma, R., & van Oort, F. (2017). Quality
Asheim, B. T., & Gertler, M. S. (2005). The geography of inno- of government and social capital as drivers of regional diversifica-
vation: Regional innovation systems. In J. Fagerberg, D. C. tion in Europe. Journal of Economic Geography, 17, 1179–1208.
Mowery, & R. R. Nelson (Eds.), The Oxford handbook of inno- doi.org/10.1093/jeg/lbx001
vation (pp. 291–317). Oxford: Oxford University Press. Dosi, G. (1982). Technological paradigms and technological trajec-
Atkinson, A. B., & Stiglitz, J. E. (1969). A new view of technological tories: As suggested interpretation of the determinants and direc-
change. Economic Journal, 79, 573–578. doi:10.2307/2230384 tions of technical change. Research Policy, 11(3), 147–162. doi:10.
Balland, P. A. (2016). Relatedness and the geography of innovation. 1016/0048-7333(82)90016-6
In R. Shearmur, C. Carrincazeaux, & D. Doloreux (Eds.), Essletzbichler, J. (2015). Relatedness, industrial branching and tech-
Handbook on the geographies of innovation (pp. 127–141). nological cohesion in US metropolitan areas. Regional Studies, 49,
Northampton: Edward Elgar. 752–766. doi:10.1080/00343404.2013.806793
Balland, P. A. (2017). EconGeo: Computing key indicators of the spatial Fitjar, R. D., & Rodríguez-Pose, A. (2011). When local interaction
distribution of economic activities, R package version 1.3. Retrieved does not suffice: Sources of firm innovation in urban Norway.
from https://github.com/PABalland/EconGeo Environment and Planning A, 43(6), 1248–1267. doi:10.1068/
Balland, P. A., & Rigby, D. (2017). The geography of complex a43516
knowledge. Economic Geography, 93(1), 1–23. doi:10.1080/ Fleming, L., & Sorenson, O. (2001). Technology as a complex adap-
00130095.2016.1205947 tive system: Evidence from patent data. Research Policy, 30, 1019–
Balland, P. A., Rigby, D., & Boschma, R. (2015). The technological 1039. doi:10.1016/S0048-7333(00)00135-9
resilience of U.S. cities. Cambridge Journal of Regions, Economy Foray, D. (2015). Smart specialization: Opportunities and challenges for
and Society, 8(2), 167–184. doi:10.1093/cjres/rsv007 regional innovation policy. London: Routledge/Regional Studies
Barca, F. (2009). An agenda for a reformed Cohesion Policy: A place- Association.
based approach to meeting European Union challenges and expec- Foray, D., David, P. A., & Hall, B. H. (2009). Smart specialization.
tations (Independent report prepared at the request of the The concept (Knowledge Economists Policy Brief No. 9, June).
European Commissioner for Regional Policy, Danuta Hübner). Brussels: European Commission.
Brussels: European Commission. Foray, D., David, P. A., & Hall, B. H. (2011). Smart specialization.
Bathelt, H., Malmberg, A., & Maskell, P. (2004). Clusters and From academic idea to political instrument, the surprising career of
knowledge: Local buzz, global pipelines and the process of a concept and the difficulties involved in its implementation
knowledge creation. Progress in Human Geography, 28, 31–56. (MTEI Working Paper, November). Lausanne: MTEI.
doi:10.1191/0309132504ph469oa Foray, D., Goddard, J., Morgan, K., Goenaga Beldarrain, X.,
Boschma, R. (2014). Constructing regional advantage and smart Landabaso, M., Nauwelaers, C., & Ortega-Argilés, R. (2012).
specialisation: Comparison of two European policy concepts. Guide to research and innovation strategies for smart specialisation
Italian Journal of Regional Science, 13(1), 51–68. doi:10.3280/ (RIS 3), S3 smart specialisation platform. Seville: Institute for
SCRE2014-001004 Prospective Technological Studies (IPTS)-Joint Research
Boschma, R. (2017). Relatedness as driver of regional diversification: Centre (JRC) of the European Commission.
A research agenda. Regional Studies, 51(3), 351–364. doi:10. Frenken, K. (2016). A complexity-theoretic perspective on
1080/00343404.2016.1254767 innovation policy (Papers in Evolutionary Economic Geography

REGIONAL STUDIES
 Smart specialization policy in the EU: relatedness, knowledge complexity and regional diversification 1267

No. 16.19). Utrecht: Department of Economic Geography, McCann, P., & Ortega-Argilés, R. (2013). Transforming European
University of Utrecht. regional policy: A results-driven agenda and smart specialization.
Frenken, K., & Boschma, R. A. (2007). A theoretical framework for Oxford Review of Economic Policy, 29(2), 405–431. doi:10.1093/
evolutionary economic geography: Industrial dynamics and urban oxrep/grt021
growth as a branching process. Journal of Economic Geography, McCann, P., & Ortega-Argilés, R. (2015). Smart specialization,
7(5), 635–649. doi:10.1093/jeg/lbm018 regional growth and applications to European Union Cohesion
Frenken, K., Van Oort, F. G., & Verburg, T. (2007). Related variety, Policy. Regional Studies, 49(8), 1291–1302. doi:10.1080/
unrelated variety and regional economic growth. Regional Studies, 00343404.2013.799769
41(5), 685–697. doi:10.1080/00343400601120296 McCann, P., van Oort, F., & Goddard, J. (Eds.). (2017). The empiri-
Grillitsch, M. (2016). Institutions, smart specialisation dynamics and cal and institutional dimensions of smart specialisation. Abingdon:
policy. Environment and Planning C: Government and Policy, 34, Routledge.
22–37. doi:10.1177/0263774X15614694 Miguelez, E., & R. Moreno. (2017). Relatedness, external linkages
Grillitsch, M., & Nilsson, M. (2015). Innovation in peripheral and regional innovation in Europe. Regional Studies. doi:10.
regions: Do collaborations compensate for a lack of local knowl- 1080/00343404.2017.1360478
edge spillovers? Annals of Regional Science, 54(1), 299–321. Montresor, F., & Quatraro, F. (2017). Regional branching and Key
doi:10.1007/s00168-014-0655-8 enabling technologies: Evidence from European patent data.
Harhoff, D., Scherer, F., & Vopel, K. (2003). Citations, family size, Economic Geography, 93(4), 367–396. doi:10.1080/00130095.
opposition and the value of patent rights – Evidence from 2017.1326810
Germany. Research Policy, 32, 1343–1363. doi:10.1016/S0048- Moodysson, J., M. Trippl, & E. Zukauskaite. (2016). Policy learning
7333(02)00124-5 and smart specialization: Balancing policy change and continuity
Hausmann, R., Hidalgo, C. A., Bustos, S., Coscia, M., Simoes, A., & for new regional industrial paths. Science and Public Policy, 44,
Yildirim, M. A. (2014). The atlas of economic complexity: Mapping 382–391. doi:10.1093/scipol/scw071
paths to prosperity. Cambridge, MA: MIT Press. Morgan, K. (2015). Smart specialisation: Opportunities and chal-
Heimeriks, G., & Balland, P. A. (2016). How smart is specialisation? An lenges for regional innovation policy. Regional Studies, 49(3),
analysis of specialisation patterns in knowledge production. Science 480–482. doi:10.1080/00343404.2015.1007572
and Public Policy, 43(4), 562–574. doi:10.1093/scipol/scv061 Morgan, K. J. (2017). Nurturing novelty: Regional innovation policy
Hidalgo, C., & Hausmann, R. (2009). The building blocks of econ- in the age of smart specialisation. Environment and Planning C,
omic complexity. Proceedings of the National Academy of Sciences, 35(4), 569–583.
106, 10570–10575. doi:10.1073/pnas.0900943106 Neffke, F., Hartog, M., Boschma, R., & Henning, M. (2018).
Hidalgo, C., Klinger, B., Barabassi, A., & Hausmann, R. (2007). The Agents of structural change. The role of firms and entrepreneurs
product space conditions the development of nations. Science, in regional diversification. Economic Geography, 94, 23–48.
317, 482–487. doi:10.1126/science.1144581 Neffke, F., Henning, M., & Boschma, R. (2011). How do regions
Iacobucci, D. (2014). Designing and implementing a smart specialis- diversify over time? Industry relatedness and the development
ation strategy at regional level: Some open questions. Italian of new growth paths in regions. Economic Geography, 87(3),
Journal of Regional Science, 13, 107–126. doi:10.3280/ 237–265. doi:10.1111/j.1944-8287.2011.01121.x
SCRE2014-001006 Petralia, S., Balland, P. A., & Morrison, A. (2017). Climbing the lad-
Iacobucci, D., & Guzzini, E. (2016). Relatedness and connectivity in der of technological development. Research Policy, 46(5), 956–
technological domains: Missing links in S3 design and 969. doi:10.1016/j.respol.2017.03.012
implementation. European Planning Studies, 24(8), 1511–1526. Pinheiro, F. L., Alshamsi, A., Hartmann, D., Boschma, R., &
doi:10.1080/09654313.2016.1170108 Hidalgo, C. (2018). Shooting low or high: Do countries benefit
Karo, E., & Kattel, R. (2015). Economic development and from entering unrelated activities? (Papers in Evolutionary
evolving state capacities in Central and Eastern Europe: can Economic Geography No. 18.07). Utrecht: Utrecht University.
‘smart specialization’ make a difference? Journal of Economic Radosevic, S., & Ciampi Stankova, K. (2015). External dimensions of
Policy Reform, 18(2), 172–187. doi:10.1080/17487870.2015. smart specialization. Opportunities and challenges for trans-regional
1009068 and transnational collaboration in the EU-13 (Joint Research
Kogut, B., & Zander, U. (1992). Knowledge of the firm, combinative Centre S3 Working Paper Series No. JRC96030).
capabilities, and the replication of technology. Organization Luxembourg: Publications Office of the European Union.
Science, 3, 383–397. doi:10.1287/orsc.3.3.383 Rigby, D. (2015). Technological relatedness and knowledge space:
Kogut, B., & Zander, U. (1993). Knowledge of the firm and the evol- Entry and exit of US cities from patent classes. Regional Studies,
utionary theory of the multinational corporation. Journal of 49(11), 1922–1937. doi:10.1080/00343404.2013.854878
International Business Studies, 24, 625–645. doi:10.1057/ Rigby, D. L., & Essletzbichler, J. (1997). Evolution, process variety,
palgrave.jibs.8490248 and regional trajectories of technological change in U.S.
Kroll, H. (2015). Efforts to implement smart specialization in practice manufacturing. Economic Geography, 73(3), 269–284. doi:10.
– Leading unlike horses to the water. European Planning Studies, 2307/144484
23(10), 2079–2098. doi:10.1080/09654313.2014.1003036 Rodríguez-Pose, A., di Cataldo, M., & A. Rainoldi. (2014). The role
Lawson, C., & Lorenz, E. (1999). Collective learning, tacit knowl- of government institutions for smart specialisation and regional
edge and regional innovative capacity. Regional Studies, 33, development (S3 Policy Brief Series No. 4). Luxembourg:
305–317. doi:10.1080/713693555 Publications Office of the European Union. doi:10.2791/71842
Maskell, P., & Malmberg, A. (1999). Localised learning and indus- Rodríguez-Pose, A., & Wilkie, C. (2015). Institutions and the entre-
trial competitiveness. Cambridge Journal of Economics, 23(2), 167– preneurial discovery process for smart specialization (Papers in
186. doi:10.1093/cje/23.2.167 Evolutionary Economic Geography No. 15.23). Utrecht:
Matti, C., Consoli, D., & Uyarra, E. (2017). Multilevel policy mixes Utrecht University.
and industry emergence. The case of wind energy in Spain. Santoalha, A. (2016). New indicators of smart specialisation. A related
Environment and Planning C, 35(4), 661–683. diversification approach applied to European regions (TIK Working
Mazzucato, M. (2013). The entrepreneurial state, debunking public vs. Papers on Innovation Studies No. 20161220). Oslo: University
Private sector myths. London: Anthern. of Oslo.

REGIONAL STUDIES
1268 Pierre-Alexandre Balland et al.

Schmoch, U. (2008). Concept of a technology classification for country Trajtenberg, M. (1990). A penny for your quotes: Patent citations and
comparisons (Final report to the World Intellectual Property the value of innovations. RAND Journal of Economics, 21, 172–
Organisation – WIPO). 187. doi:10.2307/2555502
Simon, H. A. (1962). The architecture of complexity. Proceedings of Uhlbach, W. H., Balland, P. A., & Scherngell, T. (2017). R&D policy
the American Philosophical Society, 106, 467–482. and technological trajectories of regions: Evidence from the EU frame-
Sorenson, O., Rivkin, J., & Fleming, L. (2006). Complexity, net- work programmes (Papers in Evolutionary Economic Geography
works and knowledge flow. Research Policy, 35(7), 994–1017. No. 17.22). Utrecht: Utrecht University.
doi:10.1016/j.respol.2006.05.002 Unterlass, F., Reinstaller, A., Friesenbichler, K. S., Charos, A.,
Sörvik, J., Midtkandal, I., Marzocchi, C., & E. Uyarra. (2016). How Hranyai, K., Reschenhofer, P., … Weingärtner, S. (2015). The
outward looking is smart specialization? Results from a survey on relationship between export and technological specialisation profiles
inter-regional collaboration in smart specialisation strategies (S3 across EU countries and regions and the identification of development
Policy Brief Series No. 16/2016, EUR 27795 EN). potentials (Research Report). Vienna: Austrian Institute of
Luxembourg: Publications Office of the European Union. Economic Research (WIFO).
doi:10.2791/943671 Valdaliso, J. M., Magro, E., Navarro, M., Aranguren, M. J., &
Steijn, M. P. A. (2017). Improvement on the association strength: imple- Wilson, J. R. (2014). Path dependence in policies supporting
menting probability measures based on combinations without rep- smart specialisation strategies: Insights from the Basque case.
etition (Working Paper). Utrecht: Utrecht University. European Journal of Innovation Management, 17(4), 390–408.
Storper, M. (1997). The regional world. Territorial development in a doi:10.1108/EJIM-12-2013-0136
global economy. New York: Guilford. Van Eck, N. J., & Waltman, L. (2009). How to normalize cooccur-
Tanner, A. N. (2014). Regional branching reconsidered: Emergence rence data? An analysis of some well-known similarity measures.
of the fuel cell industry in European regions. Economic Geography, Journal of the American Society for Information Science and
90(4), 403–427. doi:10.1111/ecge.12055 Technology, 8, 1635–1651.
Tavassoli, S., & Carbonara, N. (2014). The role of knowledge variety Webber, M., Sheppard, E., & Rigby, D. (1992). Forms of technical
and intensity for regional innovation. Small Business Economics, change. Environment and Planning A, 24, 1679–1709. doi:10.
43, 493–509. doi:10.1007/s11187-014-9547-7 1068/a241679
Thissen, M., van Oort, F., Diodato, D., & Ruijs, A. (2013). Regional Weitzman, M. L. (1998). Recombinant growth. Quarterly
competitiveness and smart specialization in Europe. Place-based Journal of Economics, 113(2), 331–360. doi:10.1162/003355
development in international economic networks. Cheltenham: 398555595
Edward Elgar. Xiao, J., R. Boschma, & M. Andersson. (2018). Resilience in the
Tödtling, F., & Trippl, M. (2005). One size fits all? Towards a dif- European Union: The effect of the 2008 crisis on the ability of
ferentiated regional innovation policy approach. Research Policy, regions in Europe to develop new industrial specializations.
34(8), 1203–1219. doi:10.1016/j.respol.2005.01.018 Industrial and Corporate Change, 27, 15–47. doi:10.1093/icc/dtx023

REGIONAL STUDIES