Van Oort & Bosma Aglomeration Economies, Inventor and Entrepreneurs

Ann Reg Sci (2013) 51:213–244
DOI 10.1007/s00168-012-0547-8
SPECIAL ISSUE PAPER
Agglomeration economies, inventors and entrepreneurs

as engines of European regional economic development
Frank G. van Oort · Niels S. Bosma
Received: 8 March 2012 / Accepted: 21 November 2012 / Published online: 13 December 2012
© Springer-Verlag Berlin Heidelberg 2012
Abstract In economic agglomeration studies, the distinction of various externali-

ties circumstances related to knowledge spillovers remains largely unclear. This paper
introduces human capital, innovation and several types of entrepreneurship as potential
drivers of regional economic performance with an impact of agglomeration economies.
We use measures of specific types of entrepreneurship, discerned at the individual
level, as well as human capital and invention through patenting activity for the period
2001–2006. The empirical application on 111 regions across 14 European countries
investigates their relation with observed regional productivity rates in 2006. Our main
findings indicate that (i) human capital, patenting activity and entrepreneurship are
all linked to regional performance, more so in regions containing large as well as
medium-sized cities; (ii) they act as complements rather than substitutes, facilitating
productivity differently; and (iii) accounting for patenting activity and entrepreneur-
ship captures agglomeration externalities effects previously subscribed only to the
density of resources of regional performance. The particular role of regions with
medium-sized cities next to regions with large cities complies with observed growth
trends as well as recently proposed place-based development approaches that assume
that interactions between institutions and geography are critical for regional economic
performance.
JEL Classification R12 · L26 · O30
F. G. van Oort (B)

Urban and Regional Research Center Utrecht, Utrecht University,
PO Box 80115, 3508TC Utrecht, The Netherlands
e-mail: [email protected]
N. S. Bosma
Utrecht School o Economics, Utrecht University,
PO Box 80215, 3508TC Utrecht, The Netherlands
123
214 F. G. van Oort, N. S. Bosma
1 Introduction
The tendency of firms and workers to agglomerate their activities in space showed
to be persistent throughout the history. Theoretical and empirical studies in spatial
economics have shown that agglomeration economies may be the most important
source of this uneven distribution of economic activities and economic growth across
cities and regions. Looking at the arguments behind this modern urbanization view,
we find two relatively new approaches. One is based on the theory of agglomeration
economies with increasing returns and easy access to knowledge (Krugman 2009),
and a second is based on the idea that (larger) cities are strong because they are
the physical concentration of skilled knowledge workers (Glaeser and Maré 2001).
Both approaches lead to the hypothesis of expected higher labour productivity. This
is confirmed in much empirical research. In Europe, the largest urban areas, more
in particular in the axis of London-Randstad-Paris-Frankfurt-Milan, contribute much
more to their national GDPs than could be expected by looking at their population
sizes (Ciccone 2002). The same conclusion has been made by Ciccone and Hall (1996)
for the USA.
The empirical works on agglomeration economies are characterized by a large
diversity of approaches. In Rosenthal and Strange (2004), there is a brief review of
papers focusing on urbanization economies (being advantages of cities applying to
every firm or consumer). Noticeable is that most of the early (pre-1990’s) works on
agglomeration simply used the cities’ population as a measure of agglomeration. These
studies assume that the population elasticity of productivity is constant. Rosenthal and
Strange (2004) conclude that this literature has found relatively consistent evidence:
doubling the population of a city increases productivity by 3–8 %.
Since Glaeser et al. (1992), it has become more apt to analyse growth variables using
employment in cities, suggesting a relationship between agglomeration and economic
growth, coining the possibility that urban increasing returns are working in a dynamic,
rather than static, context. Sector-specific localization economies, stemming from
input–output relations and transport cost savings of firms, human capital externalities
and knowledge spillovers, are generally offset against the earlier customary measured
general urbanization economies (Henderson 2003). A large literature builds on this
conceptualization of agglomeration economies, reflected in three recent overview and
meta-studies (Melo et al. 2009; Beaudry and Schiffauerova 2009; De Groot et al.
2009). These studies show that the relation agglomeration-growth is ambiguous and
indecisive on either specialization or diversity being facilitated by (sheer) urbanization
as context. The ambiguity is fuelled by measurement issues and heterogeneity in terms
of scale of time and space, aggregation, growth definitions, and the functional form
of the models applied.
Based on an overview of historical and current conceptualizations of knowledge,
knowledge diffusion and innovation in cities, several scholars plea for conceptual
and methodological renewal and rigour in research in order to address this current
impasse in economic agglomeration studies (Van Oort and Lambooy 2012). Only
recently, various conceptualizations of distance and proximity empirically address the
heterogeneity in actors and processes involved and capture the role of cities in this. It is
argued that research should more and explicitly focus on both the transfer mechanisms
123
Agglomeration economies, inventors and entrepreneurs 215
of knowledge diffusion, like spin-offs, research collaborations and social networks,

and on the contexts that facilitate the acquaintance and diffusion of knowledge by
individual firms. New methodological views appear needed here as well, in particular
modelling techniques that link appropriate levels of analysis, from the firm to regional
contexts and agglomeration circumstances (Van Oort et al. 2012).
The first goal of this paper is to take a step towards incorporating appropriate context
and transfer mechanism issues in spatial externalities and productivity–agglomeration
debate. Consistent with recent literature, we argue that mechanisms embedded in entre-
preneurship and innovation processes constitute a relevant additional explanation for
observed regional and national variation in economic agglomeration and performance
(Acs and Armington 2004; Acs and Varga 2005; Aghion et al. 2009; Fritsch and
Mueller 2008; Acs et al. 2010; Bosma et al. 2011). We argue that the nexus around
invention, innovation and high-quality entrepreneurship can be related to urbanization
effects.
The second goal of the paper is to contribute to the recent policy discussion on place-
based or place-neutral development strategies in the European Union. This debate is
highlighted in the context of a series of recent major policy reports: the place-neutral
policies in the 2009 World Bank report (World Bank 2009) and the European place-
based development strategies in Barca (2009) and Barca et al. (2012). Place-neutral
strategies rely on the agglomerative forces of the largest cities and metropolitan regions
to attract talent and growth potential. Place-based development strategists claim that
the polycentric nature of a set of smaller- and medium-sized cities in Europe, each
with their own peculiar characteristics and specializing in the activities to which they
are best suited, creates fruitful urban variety, which enhances optimal economic devel-
opment. This implies that medium-sized city regions have not declined in importance
compared to larger urban ones, which has been indicated in monitoring publications by
the OECD (2009, 2011), but there is until now little empirical support by explanations
based in entrepreneurship, innovation and productivity. A related place-based regional
policy dimension relevant for the European Union concerns objective-1 regions that
have been supported in cohesion policy. Recent research shows that this distinction is
important for productivity and productivity growth (Dogaru et al. 2011; Marrocu et al.
2012). This paper will, therefore, additionally test the relationship between produc-
tivity, innovation and entrepreneurship in distinctive large, medium-sized and small
urban regions, as well as in objective-1 versus non-objective-1 regions simultaneously,
and conclude on the place-based policy implications suggested in the recent discourse.
In order to reach the two research goals, this paper is structured as follows. In
Sect. 2, we introduce a synopsis of both relevant theories on agglomeration, entrepre-
neurship and innovation, as well as of the current place-based development strategies
aimed for by the European Union. We argue that including regional innovation and
entrepreneurship indicators captures additional localized externalities next to urban-
ization economies. In Sect. 3, we present a model that formalizes this in a testable
framework for 111 regions in 14 countries in Europe. In the empirical analysis, we
use data at the individual level from the Global Entrepreneurship Monitor (GEM) in
2001–2006, comprising information of entrepreneurship for over 350,000 working
age adults, and are we able to discern quality aspects of the entrepreneur. We make a
distinction between early-stage entrepreneurs’ growth expectation and their innovative
123
orientation. Patent activity is used to capture potential innovation based on new inven-
tions. These data and the estimation approach applied are discussed in Sect. 4. Section 5
presents the results of 2SLS models in which regional productivity is explained by
agglomeration economies, entrepreneurship and innovation indicators. We also intro-
duce spatial regime analyses that distinguish the relationship between productivity
and explanatory variables in objective-1 and non-objective-1 regions simultaneously,
as well as in large, medium-sized and small urban regions in Europe. This explicitly
tests for the suggested implications of place-based development trajectories that may
be useful for policies on the local level instead of on the aggregate level. We present
conclusions and a policy discussion in Sect. 6.
2 Entrepreneurship and innovation as nexus of regional productivity

and place-based development strategies
2.1 Agglomeration economies in need of conceptual renewal
The recent regional economic development literature has shown a renewed interest in
agglomeration economies from the New Economic Geography (NEG) and the related
empirical literature on economic geography and urban economics. Both sets of theo-
ries share the recently observed trend towards increased urbanization as an outcome.
The new theoretical insights from NEG are in line with the empirical observation that
inter-regional disparities in Europe, especially within countries, have grown since the
1980s. The evidence reviewed in Montfort (2009) leads to the conclusion that in the
last 10–15 years, disparities have diminished among countries and increased within
countries. Theories on agglomeration advantages as an explanation for such observed
spatial concentration of economic activities are increasingly used in economic geog-
raphy (McCann and van Oort 2009). The role of knowledge and human capital as
a determinant of such economic growth has gained greater appeal after its incorpo-
ration in economic growth models. In these models, knowledge spillovers between
economic agents play a crucial role in the growth and innovation process and lead
to external economies of scale in production. New technological knowledge is seen
as tacit, meaning that its accessibility, as well as its growth spillovers, is bounded
by geographical proximity of high-tech firms and entrepreneurs or knowledge insti-
tutions. Also, in endogenous growth theory, the generation of new knowledge and
innovations is explained by increased investment in knowledge, like research and
development (R&D). As knowledge is hard to appropriate, it generates benefits to
other agents through several spillover mechanisms. Understanding the geographical
structures that underlie these spillover benefits is necessary for any evidence-based
innovation or growth policy to stimulate Europe’s transformation towards a cohesive
as well as knowledge economy society. Spatially bounded externalities (spillovers)
are related to location decisions of firms or individuals (entrepreneurs) within their
network. The driving mechanism in agglomeration economies is then that increased
size of (urban) agglomerations leads to increased productivity, which will attract more
people to migrate to larger agglomerations. This in turn will cumulatively cause higher
productivity levels and higher economic growth.
123
The empirical evidence of agglomeration economies is strong, and in an overview

paper by Rosenthal and Strange (2004), it is shown that a doubling in the size of an
agglomeration leads to an increase in productivity between 3 and 8 %. Melo et al.
(2009), using a sample of 34 studies on agglomeration economies for 729 estimated
values of elasticity, find a variation up to 29 % . In another meta-analysis considering
31 studies, De Groot et al. (2009) conclude that the theory provides ‘strong indications
for sectoral, temporal and spatial heterogeneity’. Beaudry and Schiffauerova (2009)
confirm this view in their extensive—qualitative—overview of most recent agglomera-
tion studies. Instaed of refining or standardizing existing localization and urbanization
indicators in empirical research, it may be better to introduce more relevant indicators
that capture knowledge spillover contexts more adequately (Van Oort and Lambooy
2012). We will do this in this paper by focussing on innovation and entrepreneurship
as a novel nexus of productivity in regions.
2.2 Regional innovation and entrepreneurship as externalities contexts
The link between agglomeration and regional performance resulting from innovation-
based knowledge spillovers and entrepreneurship was first addressed by Audretsch
and Feldman (1996). In their search for economic effects of entrepreneurship, thus
far scholars have encountered two important phenomena. First, there is large hetero-
geneity within the broad category of entrepreneurship (Santarelli and Vivarelli 2007;
Stam 2008). In empirical analyses, most often a straightforward measure of new firm
formation is used, but we know from the entrepreneurship literature that some types
of entrepreneurship are more important for economic growth than others (Audretsch
and Keilbach 2005; Acs 2008; Coyne et al. 2010). A second phenomenon is the pro-
nounced uneven spatial distribution of entrepreneurship (in cities), which also seems
to be persistent over time. Until now, studies on the geography of entrepreneurship
have largely focused on (determinants of) spatial variations of (nascent) new firms
(Keeble and Wever 1986; Reynolds et al. 1994; Acs and Storey 2004; Tamásy 2006;
Koster 2007) or self-employment (Parker 2005). As such, these studies neglect the het-
erogeneity in entrepreneurship, like innovation-oriented entrepreneurship and growth-
oriented entrepreneurship versus imitative entrepreneurship. Some recent studies also
confirm the need to integrate urbanization when linking entrepreneurship to growth,
as the impact of entrepreneurship for regional growth may be higher in urban areas
(Fritsch and Schroeter 2011; Bosma et al. 2011).
Entrepreneurship and innovation are often expected to go hand in hand even though
they are not the same. Conceptually, a positive effect of entrepreneurship on regional
economic development is assumed via the commercialization of new combinations of
resources (Schumpeter 1942). The process of ‘creative destruction’ brings about com-
petition and selection among both new and incumbent firms, resulting in more efficient
producers which in the end fuels economic growth (e.g Aghion et al. 2009; Bosma et
al. 2011). Holcombe (1998) argued that there is an entrepreneurial multiplier effect
in the sense that entrepreneurship leads to more entrepreneurial opportunities through
innovation. However, from the empirical literature, we know that only a small part
of the (new) firms owned and managed by entrepreneurs is innovative or ambitious
123
in terms of growth. Referring to Baumol’s categorization of entrepreneurship in pro-

ductive, unproductive and destructive (Baumol 1990; Coyne et al. 2010 argue that
Holcombe’s statement should be considered for both productive entrepreneurship and
non-productive entrepreneurship. This means that creative destruction can also have
negative outcomes when non-productive entrepreneurs behave as predators on the
productive members of the (regional or national) society. Typical examples include
lobbying activities by interest groups of incumbent entrepreneurs and rent-seeking
behaviour (Olson 1982). In essence, non-productive entrepreneurship emerges from
the established sets of firms in their attempts to reduce competition. Thus, focusing on
early-stage entrepreneurship—and particularly promising types of entrepreneurship
in terms of growth and innovation—seems to be a sound way of trying to capture
productive types of entrepreneurship.
An important stream of literature argues that knowledge spillovers stemming from
innovation and entrepreneurship do not only contribute to the competitiveness of firms,
but also to that of regions. Concepts like innovative milieux, technological districts,
regional innovation systems, learning regions, etc., have been introduced to underline
the importance of regions as key drivers of innovation (Camagni 1991; Storper 1992;
Asheim 1996; Cooke 2001). This body of literature refers to success stories like Silicon
Valley to stress that key technological advances take place in only a limited number of
regions worldwide. Knowledge spillovers also are hypothesized to be one of the main
reasons causing localization economies in agglomeration (Henderson 2003).
Many of these studies argue that knowledge spillovers tend to be geographically
bounded and that geographical proximity facilitates knowledge sharing and in turn
innovation. In addition, this body of literature stresses that all firms in the district
benefit from these knowledge spillovers, because they belong to the same social and
cultural environment. Numerous case studies have been carried out to provide empir-
ical evidence for these theoretical statements (e.g. Cooke 2001), while econometric
studies also demonstrated that knowledge spillovers are indeed geographically local-
ized (Jaffe et al. 1993; Audretsch and Feldman 1996). In a similar vein, Acs and
Plummer (2005) argue that entrepreneurs role of ‘filtering’ the available knowledge is
also a primarily regional phenomenon. Because of the relevance of regional linkages
between human capital, innovation and entrepreneurship, it is clear that they should
all be considered in the search to explain regional economic development.
2.3 Building regional indicators from individual-level attributes
The issues above all relate to the question how the drivers of human capital, invention
and entrepreneurship—embodied in our empirical application by knowledge work-
ers, inventors and different types of entrepreneurs—are related. We can subdivide
this question into two: (i) which—if any—of the three drivers is most important for
economic development; and (ii) is there a bottleneck (can regions not develop with-
out one of these). Schumpeter (1934, 1942) seemed to be indecisive about the first
question of knowledge drivers considering his emphasis on Mark I (growth through
creative destruction—mainly by new and small firms and knowledge less appropri-
able) and Mark II (growth through innovation in large research labs—mainly big firms
123
that can appropriate knowledge) regimes, respectively. However, the Mark I and Mark
II regimes need not be exclusive for regions; they can be seen as complements and
the regional industry structure is likely to be conducive to Mark I or Mark II domi-
nance: sectoral patterns are related to the nature of the underlying technological regime
(Breschi et al. 2000).
The individual is the most relevant level for studying the entrepreneurial process
from opportunity recognition to resource mobilization and exploitation, while recog-
nizing the context this individual operates in. Schumpeter (1947) also asserted that
“the inventor produces ideas, the entrepreneurs ‘gets things done”’. This is a clear call
for making distinctions between innovative activity and entrepreneurial activity—and
for making these distinctions at the individual level. We argue that indeed individuals
should be more promising as a unit of analysis than firms, as it is ultimately individ-
uals who eventually perform the economic activities. Moreover, the traditional dis-
tinctions between firms, entrepreneurs and employees are increasingly getting blurred
in the current Western service-based era of individualism, (social) networking and
ICT-developments. The independent entrepreneurs and inventors, forming (tempo-
rary) economic coalitions with other independent professionals or(small) firms, have
escaped the traditional firm growth path perspective—and our attention thus far (Van
den Born 2009). Looking beyond the traditional input factors of labour and capi-
tal, regions can outperform in terms of human capital (knowledge workers), innova-
tion (inventors) and creative destruction through business dynamics (entrepreneurs).
Accordingly, we take into account that these drivers of competitiveness are concep-
tually based on individuals—and hence, we argue that in empirical exercises at the
regional level, measure used should be based on individual-level data.
2.4 Cities, transfer mechanisms and place-based development
A regional perspective means taking into account agglomeration economies, and

urbanization economies and localization economies in particular. In this paper, we
focus on both and discuss how the three drivers of growth discussed above can be seen
in conjunction with agglomeration. A relatively new insight into modelling entrepre-
neurship as a driver of regional economic performance is the idea that urban fea-
tures of a region may have an impact on the magnitude of the measured effect. Not
only may levels of entrepreneurship in cities exceed those at the country level (see
Acs et al. 2011 for initial evidence on world cities), but also the consequences for
economic growth may be higher in cities (see Becker and Henderson 2000). In partic-
ular, Fritsch and Schroeter (2011) find that, for densely populated areas in Germany,
the long-term impact of regional firm-formation rates on employment growth exceeds
the impact found for rural areas. Combined with Duranton and Puga’s (2004) and
Van Oort and Lambooy’s (2012) calls for a more explicit treatment of innovation and
knowledge spillovers in (micro-founded) economic models, these findings suggest
that part of urbanization economies, captured in Ciccone and Hall (1996) and Ciccone
(2002) by estimating the impact of employment density on labour productivity, may
be accounted for by inventive, innovative and/or entrepreneurial activity.
123
This may have important implications for regional development strategies, as place-
based policies may be appropriate in the European case alongside people-based poli-
cies. Barca et al. (2012) summarize the place- and people-based policy debate in the
European context in detail. Based on current economic geographical theories of inno-
vation and density of skills and human capital in cities and globalization, spatially
blind approaches argue that intervention regardless of the context is the best way to
resolve the old dilemma of whether development should be about “places” or about
“people” (Barca et al. 2012, p. 140). It is argued that agglomeration in combination
with encouraging people’s mobility not only allows individuals to live where they
expect to be better off but also increases individual incomes, productivity, knowledge
and aggregate growth (World Bank 2009). Consequently, development intervention
should be space-neutral, and factors should be encouraged to move to where they are
most productive. In reality, this is primarily in large cities. In contrast, the place-based
approach assumes that the interactions between institutions and geography are critical
for development, and many of the clues for development policy lie in these interac-
tions. To understand the likely impacts of a policy, the interactions between institutions
and geography, therefore, require explicit consideration of the specifics of the local
and wider regional context (Barca et al. 2012, p.140).
According to place-based development strategists, economic growth is not uniquely
related to mega-city regions (Barca et al. 2012). Instead, growth can be distributed
across various urban systems in different ways in different countries (OECD 2011). The
place-based approach’s emphasis on interactions between institutions and economic
geography has allowed for the examination of urban development in European cities of
all sizes. Because the roles of very large and small communities have been addressed
extensively, Barca et al. (2012) emphasize the simultaneous role of medium-sized cities
and argue that these are overrepresented in Europe. Many highly productive cities in the
EU indeed are small- to medium-sized cities whose dominant competitive advantage
is that they exhibit high degrees of connectivity rather than urban or home market
scales. A distinguishing dimension in our research period is further the objective-1
versus non-objective-1 regions, as the former are targeted by cohesion policies1 in the
period 2000–2006 that coincide with innovation and entrepreneurship policies. We
will put the hypothesis on urban size in relation to growth to a test in the empirical
model of this paper.
1 The adjective ‘objective-1’ is specifically associated with European Regional Development Funds (ERDF)
in the programming period 2000–2006. ERDF aims to strengthen economic and social cohesion in the
European Union by correcting imbalances between its regions and mainly refers as ‘objective-1’ to the
regions lagging behind in terms of development. The term is region (Europe) and time (2000–2006) specific.
In the new programming period, the old ‘objective 1’ regions together with the Cohesion Fund are understood
as ‘convergence regions’ and consider as well ‘objective 2’ regions (Dühr et al. 2010, pp. 273–286). In our
research period, although objective-1 regions are not the only representatives of cohesion policy regions,
they form an important part of it (compare Dogaru et al. 2011; Lopez-Rodriguez and Faiña 2006). For
regions in Slovenia in our analyses, we used preaccession funds as indicator. Objective-1 regions can be
both urban and rural in character.
123
3 The model
The model takes as point of departure that the production in region i, situated in country
j, is driven by the density of production in the region. Following Ciccone (2002), we
model a regional productivity function where q denotes the output produced on an
acre of land (within the region situated in the country), n captures employment, and
k the amount of physical capital. Furthermore, Q denotes value added at the regional
level and A the acreage of the region in square kilometres. Next to the average level of
human capital (H ) that is also included in Ciccone’s analysis, we introduce measures
of specialization/diversity (SD) which as reciprocal measures sectoral diversity (see
below), entrepreneurship (E) and invention (I ).
q = i j f (n H, n E, n I, k; Q i j , Ai j )
α Q (λ−1)
λ
= i j (n Hi j )β (nS Di j )γ (n E i j )δ (n Ii j )ε k 1−β−γ −δ−ε
ij
(1)
Ai j
Here, i j denotes an index of total productivity in the region. Coefficient α

(0 < α ≤ 1)
captures returns to physical capital, human capital, entrepreneurs and inventors.
The coefficients β, γ , δ and ε are positive, and 0 < β + γ + δ + ε ≤ 1. Analogous
to Ciccone (2002, p. 216), that is, assuming that (i) these parameters are equally
distributed among the acres in each region; and (ii) the rental price of the capital is the
same within a country, average regional labour productivity can be derived as:
αλϕ θ
Qi j β Ni j
= j iωj Hi j S D γ E iδj Iiεj (2a)
Ni j Ai j
Or alternatively,

Qi j 1−αλ(1−γ ) 1−αλ(1−δ) 1−αλ(1−ε) ϕ
= j iωj Hi j
1−αλ(1−β)
S Di j Ei j Ii j
Ni j

Ni j Hi j S Di j E i j Ii j θ
× (2b)
Ai j
where j depends on the rental price of the capital in the country (meaning that
national differences can be taken into account by including intercepts at the national
level), and ω is a constant,
αλ − 1 1
θ= and ϕ = (3)
1 − αλ(1 − β − γ − δ − ε) 1 − αλ(1 − β − γ − δ − ε)
Coefficient θ , therefore, measures the effect of the regional density of employ-

ment in conjunction with the four other focal indicators on regional productivity: the
degree of specialization/diversity and the densities of human capital, invention and
entrepreneurship. We will denote this combined effect as the effect of agglomeration
123
economies in the remainder of this paper. In case (αλ = 1), which, for example, holds
in the special case of constant returns to scale (α = 1) in combination with the absence
of externalities from the density of production in the region (λ = 1), the degree of
specialization/diversity and the density of employment, human capital, invention and
entrepreneurship are irrelevant and productivity is explained by human capital, entre-
preneurial activity and innovation independent from the density. In case αλ > 1, θ is
positive and its value will be higher if production is more reliant on physical capital as
the assumptions made imply that an increase in total factor productivity—caused by
increases in density of employment, human capital, invention or entrepreneurship—
will be followed by an inflow of physical capital.
Taking logarithms of (2a), we have the following:
log Q i j − log Ni j = log j + θ (log Ni j − log Ai j )

+ αλϕ β log(Hi j ) + γ log(S Di j ) + δ log(E i j ) + ε log(Ii j )
+ ω log i j , (4)
which we can rewrite in the form of an estimation equation as follows:
log Q i j − log Ni j = α j + θ (log Ni j − log Ai j )

+ηHi j + ρ S Di j + κ E i j + ν Ii j + μi i j ; (5)
where H, SD, E and I denote, respectively, human capital, specialization/diversity

(often coined localization/urbanization in the literature), entrepreneurship and inven-
tion. In Ciccone’s computations, the inclusion of regional dummies for different spatial
scales did not affect the estimated effect of urbanization economies. These are consis-
tently estimated at a rate ranging between 4.0 and 4.5 %. In our empirical exercises,
we only deal with regional and national level; hence, the dummies α j reflect countries.
4 Data and approach
4.1 Data
The selection of countries and regions included in the empirical study was restricted
particularly by data availability on (high) quality types of entrepreneurial activity.
These data are based on representative samples among the population between 18 and
64 years and available for 142 regions across 17 European countries. By eliminating
regions in Croatia due to data availability issues for other indicators, as well as the
regions with sample sizes lower than 700, we arrive at a sample of 127 regions across 16
European countries (see Bosma 2011).2 A second restriction was the availability of the
specialization/diversity index. Of the 127 regions for which we had entrepreneurship
2 The initial spatial scale was adopted along a regional classification developed by ESRI and consisted of
125 regions corresponding to Nuts1 levels for Belgium, France, Germany, Greece, Ireland, the Netherlands,
and the United Kingdom. Nuts2 levels are applied to Croatia, Denmark, Finland, Hungary, Norway, Portugal,
Slovenia, and Sweden, and a combination of Nuts1 and Nuts2 to Italy, Spain, and Switzerland. In a second
123
data, 111 could be assigned the Theil localization-diversity measure over their location
quotients (see below).
For analysing Eq. (5), we require regional-level data on value added, employment,
the acreage of each region, education, the degree of specialization/diversity, inven-
tion and types of entrepreneurship that allow to discern innovativeness and growth
ambitions. Data on value added and employment are available at the NUTS3 level and
are drawn from the Cambridge Econometrics database on European Regions. Data on
education are obtained from Eurostat’s regional database, which distinguishes three
major categories of education. We included the shares of the numbers of people who
have tertiary education in 2001 (the denominator is the population aged between 25
and 64). These indicators are available for the NUTS3 regions for all countries except
Belgium, Norway and Sweden.3 The square kilometres for the acreage of the NUTS3
regions are also drawn from Eurostat.
There is a burgeoning literature looking for micro-foundations and causes of
agglomeration economies (Rosenthal and Strange 2001; Feser 2002). Whether due
to firm size or a large initial number of local firms, a high level of local factor employ-
ment may allow the development of external economies within the group of local
firms in a sector. These are termed localization economies. The strength of these local
externalities is assumed to vary, so that these are stronger in some sectors and weaker
in others (Duranton and Puga 2001). The associated economies of scale comprise
factors that reduce the average cost of producing outputs in that locality. On the other
hand, urbanization economies reflect external economies passed to enterprises as a
result from savings from the large-scale operation of the agglomeration or city as a
whole, and which are, therefore, independent from industry structure. Relatively, more
populous localities, or places more easily accessible to metropolitan areas, are also
more likely to house universities, industry research laboratories, trade associations and
other knowledge generating institutions. It is the dense presence of these institutions,
which are not solely economic in character, but are social, political and cultural in
nature, that support the production and absorption of know how, stimulating innova-
tive behaviour and differential rates of interregional growth (Frenken et al. 2007). The
diverse industry mix in an urbanized locality, therefore, improves the opportunities to
interact, copy and modify practices and innovative behaviour in the same or related
industries.
The degree of sectoral specialization and diversity is an important variable in our
models, as it tests our hypothesis on agglomeration. The degree of regional specializa-
tion is measured by the Theil index over the location quotients of 59 products including
agriculture, manufacturing and services, for the year 2000 (see Thissen et al. 2011;
Dogaru et al. 2011 for a more detailed introduction of this measure). This unique data
set has been collected by the Netherlands Environmental Assessment Agency (PBL)
and is based on regionalized production and trade data for 256 European NUTS2
Footnote 2 continued
step, some dense sub-regions were identified within the previously identified larger regions; if the sample
size allowed, these dense regions are abstracted and treated separately from the larger region of which they
form part. This resulted in an augmented sample of 142 regions.
3 This is no problem for our data set, as we deal with Nuts2 regions in these countries in our analysis.
123
regions, 14 sectors, and 59 product categories (compare Combes and Overman 2004).
Location quotients measure the relative specialization of a region in a certain sector
as the percentage of production accounted for by a sector in a region relative to the
percentage of production accounted for by that sector in Europe as a whole. This quo-
tient measures whether a sector is over- or underrepresented in a region compared with
its average representation in a larger area and, therefore, is to comprise localization
or specialization economies of agglomeration. The Theil coefficient then measures
deviations from the European average distribution of production specializations in all
sectors. A high score represents a large degree of sectoral specialization in a region,
and a low score represents sectoral diversity. Dogaru et al. (2011) show that on the level
of European NUTS2-regions, the largest national economies of Germany, France and
the United Kingdom regions have high levels of sectoral diversity (all regions contain
most of the existing sectors, including services). Eastern European regions are rela-
tively specialized, as are Scandinavian, Greek and Irish regions. These regions miss
concentrations of certain activities, for example, specific types of services, manufac-
turing, distribution or agricultural activities. A group of medium-sized economies,
like The Netherlands, Austria, Belgium, Denmark, Italy, Portugal and Spain, show
moderate levels of specialization.
Invention is captured by the number of applicants who have filed patents between
2001 and 2006. In endogenous growth theory, the generation of new knowledge and
innovations is explained by increased investment in R&D (Acs 2002). This was also the
perspective of Schumpeter’s book of 1942, where he saw the entrepreneurial function
of management replacing the risk-taking and innovative individual hero-entrepreneur.
In this view, it is possible to create a knowledge production function (KPF) with invest-
ments in R&D as input and knowledge, in the form of patents as output. Consensus
has grown among economists and economic geographers that knowledge production
and knowledge spillovers are to an important extent geographically localized (Breschi
and Lissoni 2009). To test for knowledge spillovers, most scholars apply a knowledge
production function approach to explain the regional production of patents or inno-
vations as a result of public and private R&D inputs and a local spillover index. In
more than one case, and for different spatial levels, scholars have been able to indi-
cate that such spillovers turn out to be statistically significant, that is, exert a signifi-
cant and positive effect on knowledge output as measured by patents or innovations.
Geographical proximity is often claimed to be beneficial for successful knowledge
exchange. Most often, this is explained by the importance of face-to-face contacts
for the exchange of tacit knowledge. In many studies, this localized interaction is,
however, only assumed implicitly rather than examined in an explicit manner. It is
for this reason that Van Oort and Lambooy (2012) remark that it is not yet clear how
the lines of knowledge-generation and application can exactly be connected to pro-
ducers and consumers in urban contexts. It is particularly necessary to investigate the
relative importance of markets (prices) and formal and informal networks as carriers
of knowledge. We, therefore, focus on regionally aggregated inventor-level patents,
as that population captures spillover opportunities most optimally. Information on
patent activity was derived from the REGPAT database, abstracted from the Euro-
pean Patent Office and cleaned by the KITeS centre based at Bocconi University
(Thoma et al. 2010). Applicants include individual applicants, company-based
123
applicants and inventors. The dates of the applications refer to application year (not
priority year) as we are interested in the moment that the invention—as an indicator
of potential innovation—was introduced to the region.
Our entrepreneurship indicators are derived from the Global Entrepreneurship Mon-
itor (GEM; see Reynolds et al. 2005; Bosma et al. 2012). The indicators are based
on telephone surveys among the adult population. A key GEM indicator is the total
early-stage entrepreneurial activity (TEA) rate. This measure is defined as the preva-
lence rate (in the 18–64 population) of individuals who are involved in either nascent
entrepreneurship or as an owner-manager in a new firm in existence for up to 42
months. Nascent entrepreneurs are identified as individuals who are, at the moment
of the GEM survey, setting up a business. Moreover, they have indicated (i) that they
have ‘done something to help start a new business, such as looking for equipment or
a location, organizing a start-up team, working on a business plan, beginning to save
money, or any other activity that would help launch a business’; and (ii) that they will
be the single owner or a co-owner of the firm in gestation. Also, they have not paid
any salaries, wages or payments in kind (including to themselves) for more than 3
months; if they have, they are considered to be an owner-manager of a (new) firm.
While the TEA rate is an overall measure of early-stage entrepreneurial activity,
identifying different types of TEA is also possible. We draw distinctions between
growth expectation (three categories) and innovation orientation. The four types of
early-stage entrepreneurial activity are as follows:
1. Early-stage entrepreneurial activity with no growth ambitions TEAGR_NO): Indi-
viduals in early-stage entrepreneurial activity who expect to have no or one
employee in the next 5 years
2. Early-stage entrepreneurial activity with some growth expectations (TEAGR_SM):
Individuals in early-stage entrepreneurial activity who expect to have 2 or more
employees in the next 5 years
3. Early-stage entrepreneurial activity with high-growth expectations (TEAGR_HI):
Individuals in early-stage entrepreneurial activity who expect to have 10 or more
employees in the next 5 years
4. Early-stage entrepreneurial activity with innovative ambitions TEAINNOV: Indi-
viduals in early-stage entrepreneurial activity who expect (i) at least some customers
to consider the product or service new and unfamiliar and (ii) not many businesses
offering the same products or services.
In order to get reasonably precise measures at the regional level, GEM data between
2001 and 2006 have been merged; hence, the entrepreneurship data in our regression
refers to the period 2001–2006. Regional variation in entry rates is often very per-
sistent (Fritsch and Mueller 2007; Brenner and Fornahl 2008; Andersson and Koster
2011), as is economic performance (Martin and Sunley 2006). To see the relevance
of identifying the types of entrepreneurial activity, we provide maps relating to three
of the types of entrepreneurship in Figs. 1, 2 and 3. The figures point at large dif-
ferences in regional patterns. The average non-growth regional entrepreneurship rate
(TEAGR_NO) pictured in Fig. 1 is 2.8 % and ranges from 1.2 % in the western part of
France to 6.0 % in Western Transdanubia (Hungary). The rate of high-growth-oriented
TEA in Fig. 2 ranges from 0.6 % in the French Parisien Bassin to 2.6 % in the Hamburg
123
Fig. 1 Total early-stage entrepreneurial activity (TEA) with low-growth orientation (zero or one employee
in the next 5 years), percentage of population between 18 and 64 years, 2001–2006
area. We should note that, since the indicators are estimates rather than count data,
there are confidence intervals attached to these estimates. Therefore, when examining
the maps, one should especially focus on general patterns and not so much on the
outcome for one particular region.
Although we can still figure out national borders in these European maps, regional
variations within countries are also large. Focusing on the main differences between
lower ambitious types of entrepreneurship (Fig. 1) and higher growth ambitious types
of entrepreneurship (Fig. 2), we see some notable differences. In general, the growth-
oriented entrepreneurship rates appear to be somewhat higher in or around strongly
populated regions. Compared to other European regions, in many Spanish areas, there
123
Fig. 2 Total early-stage entrepreneurial activity (TEA) with high-growth orientation (ten or one more
employees in the next 5 years), percentage of population between 18 and 64 years, 2001–2006
are fairly many early-stage entrepreneurs with low-growth expectations, but the rate
of ambitious ones with respect to hiring employees is relatively low. The same goes
for Northern Portugal, Greece and parts of France. Sweden is an example of a country
showing low overall entrepreneurship rates, but performing better on growth-oriented
entrepreneurship. This is even stronger for the northern part of Italy, where there is
relatively little participation in TEA with low-growth orientation, but the scores on
growth ambitious entrepreneurship are clearly higher. In this respect, the western part
of Slovenia connects to Northern Italy. Within France, only the Paris and Mediterranean
areas have relatively many growth ambitious early-stage entrepreneurs, while this rate
is low in all other regions. Regions performing relatively bad in all types of independent
123
Fig. 3 Total early-stage entrepreneurial activity (TEA) with innovative orientation (regardless of sector),
percentage of population between 18 and 64 years, 2001–2006
entrepreneurship are situated in the east of France, and to a lesser extent, some Swedish
regions and the whole of Belgium.4
The innovation-oriented early-stage entrepreneurship rates (Fig. 3) show an even
larger regional variation. In the UK, the London area and the Eastern region (including
Cambridge) outperform other regions with respect to innovation-oriented early-stage
entrepreneurship rates. The Mediterranean area (including Nice/Sophia-Antipolis)
seems to be rather innovative, as compared to the rest of France; Emilia-Romagna
4 We would like to point out that the entrepreneurship measure in our analysis disregards entrepreneurial
activity conducted by employees. In Bosma et al. (2012), in a first international comparison of entrepreneurial
employee activity, countries like Sweden, Belgium and France appear to exhibit relatively high levels of
entrepreneurial employee activity.
123
Table 1 Descriptive statistics, regional level, N = 111
Mean SD Correlation table
1. 2. 3. 4. 5. 6. 7. 8.
1. Labour 3.88 0.39

productivity
2006 (ln)
2. Employment −2.57 1.44 0.31*
density
2001–2006 (ln)
3. Acreage of the 27.0 30.5 0.04 −0.68*
region (*1000)
4. Specialization 0.073 0.049 0.02 −0.45* 0.25*
(Theil), 2000
5. Patent activity, 0.37 0.34 0.56* 0.26* −0.08 −0.14
total, (*1000),
2001–2006
6. Entrepreneurship: 2.78 0.84 −0.30* 0.10 −0.16 −0.04 −0.03
low-growth
expectation,
2001–2006
7. Entrepreneurship: 0.73 0.44 0.31* 0.51* −0.38* −0.17 0.29* 0.37*
high-growth
expectation,
2001–2006
8. Entrepreneurship: 0.91 0.46 0.44* 0.26* −0.05 0.02 0.25* 0.30* 0.61*
innovation,
2001–2006
9. Share tertiary −2.25 0.46 0.58* 0.33* −0.08 017 0.44* 0.04 0.31* 0.38*
education, 2003
(ln)
* p < 0.01
and Sardinia show far higher innovative entrepreneurship rates than Sicily. Sweden
and Finland show high levels of innovation-oriented entrepreneurial activity, while
especially in the latter country, regional variation is also large.
Table 1 shows the means, standard deviations and correlations of all the variables
concerned in the estimation of Eqs. (2a), (2b). The average value of our dependent
variable, regional labour productivity, equals 3.88. Further inspection reveals that
there are substantial differences between countries (more so than within countries).
The standard deviation between countries equals 0.45, whereas the standard devia-
tion within countries equals 0.13. This result confirms that country borders need to
be appreciated in our empirical models explaining labour productivity. Still, regional
differences are important as the top performing regions are situated in different coun-
tries: highest labour productivity rates are found in Oslo and surroundings, the west-
ern parts of Switzerland, Ile de France (Paris), Stockholm area and Copenhagen area.
Many urbanized areas feature in the top 15 regions in terms of labour productivity.
The bottom 15 consists of regions in Hungary, Portugal, Slovenia and Extremadura
(Spain).
123
Table 1 further indicates that high regional education levels, patenting activity,
high-growth-oriented entrepreneurship and innovation-oriented entrepreneurship are
all positively correlated with regional levels of labour productivity, while low growth-
oriented entrepreneurship correlates negatively with labour productivity. Also, high-
growth-oriented entrepreneurship correlates positively with employment density. The
correlation between high-growth-oriented entrepreneurship and innovation-oriented
entrepreneurship equals 0.61, which could potentially lead to multicollinearity prob-
lems in case they are both included in a model. We have, therefore, decided to not
include two entrepreneurship measures in the same regression. Employment density
is strongly correlated with its instrumental variable, the acreage of the region. The
instrument is not correlated with labour productivity, and therefore, this assumption
for proper instruments is not violated.5 We also calculated the correlations with over-
all regional TEA rates. Interestingly, it turned out these overall rates did not correlate
significantly with any of the three different types of TEA listed in Table 1. This finding
confirms the relevance of identifying different types; their regional patterns turn out
to differ from each other.
We finally observe that although innovation-oriented entrepreneurial activity and
innovation through patent activity are positively linked, the relation is rather weak. In
addition, there seems to be no clear pattern for these measures in relation to employ-
ment density. We should note that these findings may turn out to be different if we
would use a different spatial classification. For instance, if the Cambridge and Oxford
region could be abstracted from the South East (England) region that we employ as
observation in our analysis, employment density and patenting activity would probably
both be high in this particular area. Both measures would be lower for the remainder
of the South East region.
4.2 Approach
Having described our data and revealed some initial statistical associations, we now
turn to the steps in our regression analysis. The first step consists of an analysis rem-
iniscent to that of Ciccone (2002), although not all variables of that analysis can be
reproduced exactly and the period of analysis is much more recent. The main reason
for building so closely on this previous work is to see whether the spatial applica-
tion we adopt in this paper makes a difference for the analysis: we change to such a
spatial scale of analysis that allows us to include measures of (types of) entrepreneur-
ship and the measure of specialization/diversity. Next, we add the index measuring
specialization/diversity in order to appreciate agglomeration economies as a com-
bination of urbanization economies and localization economies. In a third step, we
add our measures of entrepreneurship and invention and see (i) whether accounting
for entrepreneurship and patenting activity provides for an additional explanation of
regional differentials in productivity levels; and (ii) what this means for the estimated
effect of ‘urbanization economies’ as explained in the previous section. Finally, we
5 There is also no significant correlation between the error terms resulting from our regression and the
acreage of the region.
123
Table 2 2SLS estimation results with different spatial settings and estimation techniques
Model 1 Model 2a Model 2b Model 2c
Employment density (ln), Average 2001–2006 0.053*** 0.039** 0.034* 0.043**

(0.008) (0.020) (0.019) (0.019)
Specialization-diversity, 2000 0.87
(0.56)
Share tertiary education (ln), 2001 −0.02 0.29*** 0.19** 0.18**
(0.03) (0.09) (0.08) (0.08)
Constant 3.92*** 4.63*** 4.38*** 4.32***
(0.05) (0.10) (0.15) (0.15)
No. of countries 5 17 14 14
No. of regions 766 127 111 111
Regional classification NUTS3 NUTS1/3 NUTS1/3 NUTS1/3
Treatment country effects
Fixed effects Yes Yes Yes Yes
Hausman test X 2 71.44 1.29 2.01 5.02
( p < 0.01) ( p = 0.52) ( p = 0.37) ( p = 0.17)
R-squared overall 0.189 0.313 0.336 0.313
Within countries 0.164 0.457 0.343 0.375
Between countries 0.473 0.274 0.342 0.312
Dependent variable: regional levels of labour productivity, 2006, in logarithm
Acreage of the region enters the regressions as instrumental variable for employment density
* p < 0.10; ** p < 0.05; *** p < 0.01
introduce spatial regimes in our 2SLS models allowing for coefficients to structurally
differ over objective-1 versus non-objective-1 regions (based on Dogaru et al. 2011
and Lopez-Rodriguez and Faiña 2006) and over large (at least 3 million inhabitants),
medium-sized (between 1.5 and 3 million inhabitants) and small (lower than 1.5 mil-
lion inhabitants) regions.6 The next section is structured based on these three steps in
our analysis.
5 Results
5.1 The initial model
The first model in Table 2 is largely in vein of the model introduced by Ciccone (2002)
with country dummies and employment density instrumented with the acreage
of the region. For 1989, Ciccone reported a coefficient for employment den-
sity equal to 0.046 with a robust standard error of 0.005. Whether NUTS0
6 Although this distinction differs somewhat from that in OECD (2012), these cutoff points yield a distribu-
tion for the European regional classification adopted in the paper that is comparable to the OECD-distribution
on a global scale.
123
(country), NUTS1 or NUTS2 regions as fixed effects were included did not
make any significant difference in the estimated size of this coefficient. Our
analysis of the effects on labour productivity in 2006 in model 1 yields very
similar results, although we arrive at a somewhat higher coefficient (equal to
0.053). Augmenting the number of countries and adopting the regional level for
which we have data on patenting and entrepreneurship leads to a decrease in
the number of observations. For these 127 regions over 17 countries, the esti-
mated effect in model 2 decreases somewhat to 3.9 % (model 2a) and is signif-
icant at the 5 % level. In this different spatial setting, the coefficient measuring
the effect of education is now positive and significant. The standard errors are
somewhat larger owing to the lower number of observations. Model 2b shows the
results of the very same model, however, limited to the observations for which
the specialization/diversity measure is also available. Model 2c then shows the
estimates including the specialization/diversity measure, leading to an estimated
coefficient to employment density of 4.3 %, very similar to the Ciccone (2002)
findings.7
Hausman tests indicate that random effects estimation may not be used with
model 1b but they may be used with models 2. Estimating with random effects
appeared to better explain variations between countries, but not within countries.
As we are primarily interested in capturing regional effects, we proceeded with fixed
effects estimations—also to be consistent with Ciccone (2002).8 In addition, tests
revealed that there are no indications of the necessity of spatial autocorrelation model
corrections.9
5.2 Adding agglomeration economies
As noted, model 2c reveals that the specialization/diversity measure that is hypothe-

sized to capture sector-specific agglomeration economies has no effect on productivity
when looking at their direct effect in isolation. Employment density as the second indi-
cator of agglomeration is positively significant in all models in Table 2. On the basis
of these results, one would conclude that specialization and/or diversity effects (as
measured in the Theil index) do not matter for productivity and have no added value
on the top of urbanization per se (as measured by employment density). This con-
firms the signalled ambiguity of empirical studies on specialization and diversity in
relation to productivity and growth: no dominant force of either one of them is found.
When including innovation and entrepreneurship variables in model 3, the relation of
agglomeration indicators with productivity changes. Again, the employment density
7 Even though the coefficient for the specialization/diversity measure is not significant, its addition to model
2b leads to a better model fit ( p < 0.05).
8 Auxiliary regressions (not reported) indicate very similar outcomes for Tables 2 and 3 when applying
random effects estimation.
9 Using a binary weight matrix with information on neighbouring regions, the Getis–Ord statistics did not
point at spatial lags for the dependent and independent variables, nor for the residuals. In contrast to the
more often used Moran’s I statistic, the Getis–Ord statistic focuses on clustering of high and low values
(Getis and Ord 1992), which would be the relevant issue of concern in our analysis.
123
indicator is always positively significant attached to productivity. But now also, in

models 3a, 3f and 3g that all include the patenting variable, the level of (aggregate)
specialization is positively related to productivity and has an additional effect over the
density variable. The model fit improves considerably. The indicators for innovation
(patenting) and entrepreneurship are positively and significant attached to productivity
as well.
5.3 Adding entrepreneurship and invention as knowledge transfer mechanisms
Introducing variables that capture knowledge transfer mechanisms much more closely
instead of aggregate agglomeration variables in isolation changes the picture and leads
to more interpretable results. The effect of agglomeration economies is captured by a
complex interplay of the variables density, specialization, entrepreneurship and inno-
vation. Initial models only focusing on some of these dimensions in isolation miss
important effects that are revealed in our analysis. This confirms our plea for concep-
tualizing and empirically measuring agglomeration variables that capture spillovers
and externalities as close as possible.
The first model in Table 3 repeats the results of model 2c for the sake of comparison.
Model 3a adds invention (measured by patenting activity), and we observe that this
coefficient is positive and highly significant ( p < 0.01). From the results in models
3b–3d, entrepreneurial activity also appears to be a positive contributor and significant
at p < 0.10. This holds for all four distinguished types of entrepreneurship, while
we particularly expected the impact to be positive for high-growth and for innovation-
oriented entrepreneurial activity. Our expectations are, however, confirmed in the sense
that the regional variations in labour productivity are best explained by patent activity
as well as growth and innovation-oriented entrepreneurship. A possible explanation
of the positive and weakly significant effect of low-growth-oriented entrepreneurial
activity may be that we pick up some of the effects not captured by employment and
education. Self-employment is not very often captured in employment statistics used
to estimate growth models. Put differently, the effect found for low-ambition self-
employment may be very similar to the traditional effect of employment. The model
fit statistics suggest that focusing on low-ambition self-employment as a measure of
entrepreneurship particularly leads to poorer explanation of productivity differentials
between countries.
Interestingly, we find distinctive results for patenting activity and the measure of
some growth-oriented entrepreneurship. These types exhibit the strongest interplay
with employment density controlling for national effects—since the estimated ‘urban-
ization economies’ effect drops from 4.3 % (and statistically significant at p < 0.05)
in model 2c to 2.9 and 3.4 % (statistically insignificant at p < 0.10) in models 3a and
3d, respectively. Thus, acknowledging the effect of employment density and human
capital alone may not give a sufficient picture of economic advantages to urbanization;
regions also require people with new ideas (innovation) and/or entrepreneurs who can
create employment opportunities—the existence of social capital may reinforce the
occurrence of such entrepreneurs, see for example Bosma et al. (2004). Accounting for
the three drivers of economic growth considered in our study (human capital, inventors
123
234
123
Table 3 2SLS estimation results including measures of types of entrepreneurship
Model 2c (repeated) Model 3a Model 3b Model 3c Model 3d Model 3e Model 3f Model 3g
Employment density 2001–2006 0.043** 0.029 0.044** 0.038* 0.034 0.043** 0.025 0.030
(0.019) (0.019) (0.019) (0.021) (0.021) (0.019) (0.020) (0.019)
Specialization-diversity index, 2000 0.87 1.27** 0.80 0.72 0.75 0.66 1.11** 1.04*
(0.56) (0.53) (0.56) (0.56) (0.57) (0.58) (0.53) (0.54)
Share tertiary education, 2001 0.18** 0.20*** 0.17** 0.16** 0.16** 0.16** 0.18** 0.18**
(0.08) (0.08) (0.08) (0.08) (0.08) (0.08) (0.07) (0.07)
Patenting activity, 2001–2006 0.17*** 0.16*** 0.17**
(0.04) (0.04) (0.04)
Entrepreneurship, 2001–2006
No growth aspiration TEA 0.031*
(0.016)
Some growth aspiration TEA 0.043** 0.037**
(0.018) (0.017)
High-growth aspiration TEA 0.085*
(0.044)
Innovation TEA 0.066* 0.065*
(0.037) (0.035)
F. G. van Oort, N. S. Bosma
Table 3 continued
Model 2c (repeated) Model 3a Model 3b Model 3c Model 3d Model 3e Model 3f Model 3g
Constant 4.32*** 4.24*** 4.22*** 4.17*** 4.20*** 4.24*** 4.11*** 4.17***

(0.15) (0.14) (0.15) (0.15) (0.15) (0.14) (0.14) (0.13)
No. of countries 14 14 14 14 14 14 14 14
No. of regions 111 111 111 111 111 111 111 111
Regional classification NUTS1/3 NUTS1/3 NUTS1/3 NUTS1/3 NUTS1/3 Nuts1/3 NUTS1/3 NUTS1/3
R-squared overall 0.313 0.407 0.224 0.319 0.320 0.361 0.431 0.454
Within countries 0.375 0.453 0.400 0.405 0.393 0.400 0.476 0.489
Between countries 0.312 0.436 0.167 0.319 0.320 0.372 0.451 0.454
Fixed effects (country level) used in all models. Standard errors reported between parentheses. Regions with a GEM sample size lower than 700 have been excluded
(15 regions). Employment density instrumented with acreage of the region
* p < 0.10; ** p < 0.05; *** p < 0.01
Agglomeration economies, inventors and entrepreneurs
123
235
and entrepreneurs), the importance of the density of employment in conjunction with

these drivers for regional productivity appears to be limited at best. Adding entrepre-
neurship and patenting activity to model 2c does lead to a significant improvement for
most of the models: a likelihood ratio test supports the relevance of the inclusion of
early-stage entrepreneurial activity at p < 0.05 for models 3a–3c and at p < 0.1 for
models 3d and 3e. In models 3f and 3g, we allow for separate effects of innovation
(from patent activity) and entrepreneurial activity. Results from both models suggest
complementarities rather than substitutes: sizes and significances are only slightly
lower in comparison with those in models 3a, 3c and 3e, and the effects remain pos-
itive and significant, whereas the model fit improves that of 3a ( p < 0.05 for model
3f and p < 0.10 for model 3g).
5.4 Spatial regime analyses
Table 4 presents 2SLS estimates of a model of productivity where the sample of 111
regions is split in two regimes (regions that obtained objective-1 funds in the research
period versus those that did not) and three regimes (large-medium-sized and small
urban regions), respectively. In the models in Table 4, all entrepreneurship variables
are introduced.10 Regime analysis estimates the two equations simultaneously and per-
forms a spatial Chow–Wald test to determine the significance of the regime. Because
much of region-specific heterogeneity is now captured by the respective regimes, no
spatial fixed effects are introduced in the model. The spatial Chow–Wald test for the
productivity equation shows that the two regimes concerning objective-1 funds signifi-
cantly differ from each other, while those on size do not. Coefficients that significantly
differ over regimes (tested by t tests) are boxed in Table 4). Non-objective-1 regions
contain a significant relation of high growth aspiration entrepreneurship with produc-
tivity, a larger impact of the absence of entrepreneurs without growth aspirations, and
a larger impact of education. Interestingly, although the coefficients over the urban
size regimes do not significantly differ from each other in direction, they do differ in
magnitude. In large urban regions, entrepreneurs with a mediate level of growth aspi-
ration are more (and significantly) attached to productivity, while in medium-sized
urban regions, innovative entrepreneurs are significantly contributing to productivity.
The explained variance is in both regime analyses considerable higher than in the
non-regime analyses, suggesting that the regimes capture an additional part of the
regional variation in productivity. The share of higher education remains important
in all regimes except for the small urban regions, and patent activity is an important
contributor to productivity in the largest as well as medium-sized urban regions. The
regime analyses suggest that spatial heterogeneity is a crucial issue in explaining pro-
ductivity patterns. Besides large urban regions, medium-sized regions are indicative
for explaining productivity by (innovative) entrepreneurial activity. This suggests that
a focus on particular types of regions (large, medium-sized, or objective-1 regions)
can also direct policymakers in place-based development strategies.
10 The correlations between these variables are not very high (and never higher than 0.61).
123
Table 4 2SLS estimation results by objective-1 and urban size regimes
Model 4 Model 5
Objective-1 Non-obj.-1 Small Medium Large
Employment density 0.016 0.097 −0.060 0.052 −0.007

2001–2006 (0.04) (0.06) (0.08) (0.06) (0.06)
Specialization- 0.887 −0.225 −1.180 0.890 0.230
diversity Index, 2000 (1.02) (0.95) (1.61) (0.73) (1.60)
Share tertiary 0.221*** 0.325*** 0.269 0.289** 0.510***
education, 2001 (0.10) (0.14) (0.18) (0.16) (0.17)
Patenting activity, 0.100 0.432 0.135 0.339*** 0.465**
2001–2006 (0.09) (0.326) (0.14) (0.12) (0.26)
Entrepreneurship, 2001–2006
No growth aspiration TEA 0.032 −0.218*** −0.189*** −0.183*** −0.214***
(0.53) (0.040) (0.059) (0.07) (0.050)
Some growth aspiration TEA −0.542 0.113* 0.028 0.093 0.144***
(0.44) (0.07) (0.09) (0.08) (0.06)
High-growth aspiration TEA 0.049 0.247** 0.249 0.016 −0.078
(0.11) (0.13) (0.17) (0.17) (0.17)
Innovation TEA 0.449 0.107 0.218 0.241** 0.123
(0.84) (0.15) (0.19) (0.13) (0.127)
Constant 4.467*** 4.206*** 4.49*** 4.604*** 4.98***
(0.26) (0.15) (0.38) (0.36) (0.55)
No. of countries 14 14
No. of regions 111 111
Regional classification NUTS1/3 NUTS1/3
R-squared overall 0.786 0.726
Spatial Chow–Wald 41.414 17.979
test (prob.) (0.000) (0.457)
Fixed effects (country level) used in all models. Standard errors reported between parentheses. Regions with
a GEM sample size lower than 700 have been excluded (15 regions). Employment density instrumented
with acreage of the region. Coefficients that significantly differ over regimes are bold
* p < 0.10; ** p < 0.05; *** p < 0.01
6 Conclusions and discussion
In this paper, we posited that for the discussion on the merits of agglomeration
economies, it is vital to acknowledge the underlying knowledge spillover circum-
stances of innovation and entrepreneurship. We considered externalities related
to innovation and entrepreneurship as an explanation of regional differences of
economic performance in a specific spatial and economic context. We concen-
trated on the occurrences of specific types of entrepreneurship in the early-stage
phase, when the venture is in the exploration phase (nascent entrepreneurship) or
123
in the early years after the start-up. The types of early-stage entrepreneurship—
innovation-oriented entrepreneurial activity and measures related to growth expec-
tation entrepreneurial activity—are determined at the individual level, and their
regional prevalence rates exhibit significant variation. Similarly, we looked at indi-
viduals filing patents at the EPO for retrieving measures of patent activity, pos-
sibly leading to innovation. A first inspection of the data showed that although
positive relationships exist between levels of growth- and innovation-oriented entre-
preneurship and patent activity, the regional patterns are essentially different. Hence,
innovative activity through patenting, entrepreneurship that exploits innovations and
entrepreneurship that is growth-oriented in general measure different aspects of
the creative destruction mechanism. Besides our interest in the empirical relation-
ship between productivity, entrepreneurship and innovation in European regions, we
were interested in opportunities for place-based development (as opposed to people-
based or place-neutral development), as recently advocated by the European Union.
Given our regional data that were built-up from the individual level (of entrepre-
neurship and patenting data) and recent delineations on urban size and cohesive-
ness, we are able to test the proposition that medium-sized urban regions, besides
larger or small communities, or so-called objective-1 hold good cards for regional
development.
In our empirical investigation, we found confirmation for the importance of both
types of early-stage entrepreneurship and patenting activity in explaining regional vari-
ation in labour productivity. Moreover, we found in our analysis of European regions
that the impact of growth-oriented entrepreneurship and patenting activity is comple-
mentary to an important part of the urbanization economies effects, as found by for
instance Ciccone (2002), who also examined European regions. Thus, urbanization
economies can partly be explained by the effect of differentials in regional levels of
growth-oriented entrepreneurship and patent activity. We did not find this particular
effect for innovation-oriented entrepreneurship or entrepreneurs who do not expect to
grow.
An interesting finding was also that regions with high levels of low-growth-oriented
entrepreneurship (that is, early-stage entrepreneurs expecting to generate at most one
job apart from their own over the next 5 years) were also associated with higher levels
of labour productivity. Regions with a large number of such early-stage entrepre-
neurs, overall constituting over 50 % of all early-stage entrepreneurs, may be more
productive because there are more people who are responsible for their own income
and, therefore, willing to work hard. In addition, these findings may reflect produc-
tivity gains stemming from the trend of increasing ‘independent professionals’ in
service sectors. Many of these entrepreneurs are focused on increasing their earn-
ings, basically under the condition that they will not grow in terms of employment.
As a result, many of these independent professionals subcontract and, as a regional
aggregate, their impact may be sizable in terms of regional productivity measures.
Another explanation may be that the effect of low-growth entrepreneurship adds to
the effect of labour, since in most statistics, the number of employed exclude the
self-employed. In this perspective, the positive effect found with low-growth-oriented
entrepreneurship may be interpreted similarly to the contribution of the traditional
factor of labour.
123
Our finding in specially conducted regime analyses relates to the second aim of
this paper, which is to contribute empirically to the recent discussion on place-based
or place-neutral development strategies in the European Union (Barca et al. 2012). A
conceptual development discussion burgeons between, on the one hand, spatially blind
approaches that argue that intervention regardless of context (“people-based policy”)
is the best way to go forward, and on the other hand, place-based approaches assume
that interactions between institutions and geography are critical for development. This
idea has recently been translated into either a focus on the largest urban concentra-
tions (“people-based policies”) or on an urban network setting combining clusters of
especially medium-sized cities (“place-based policies”). Our framework combining
productivity, entrepreneurship and invention shows that although the spatial regimes of
urban regional size are not significantly different from each other, there are still indica-
tive differences in the qualitative entrepreneurship–productivity relation. Innovative
entrepreneurship is positively attached to productivity only in medium-sized urban
regions, while mediate growth-oriented entrepreneurship is more attached to produc-
tivity in large urban regions. Also, the distinction in regions based on objective-1
funding is important, with better performance opportunities (still) to be found outside
cohesion regions. This regional heterogeneity suggests that microeconomic processes
work out differently in different regions, thereby supporting European place-based pol-
icy strategies alongside place-neutral (people-based) policy strategies. Furthermore,
growth- and innovation-oriented entrepreneurships are particularly individually driven
phenomena and could be constrained by, for example, national levels of employment
protection as Bosma et al. (2009) found. Their analysis also suggested that urbaniza-
tion has an indirect impact on entrepreneurship via individual-level variables such as
age, education and the degree to which people see entrepreneurial examples (a form
of entrepreneurship-specific social capital). These circular relationships underline the
long-lasting trend towards more urbanization (United Nations 2007).
Finally, our approach using information at the individual level and the firm level
to create measures at the regional level opens other avenues for further multilevel
research. To some extent, we accounted for the interaction between human capital
and growth-oriented entrepreneurship when we investigated the impact on urbaniza-
tion economies, but this could be modelled more explicitly. Such models can then
also control for endogeneity issues arising from the omission of such interactions in
so-called intangible regional assets (Artis et al. 2012). For instance, the data would
also allow examining the impact of highly educated entrepreneurs on regional perfor-
mance viz. a viz. the impact by lower educated entrepreneurs. Similarly, one could
include more quality aspects in the invention measures by also considering patent cita-
tions. Our results do point at separate positive effects of innovation and entrepreneur-
ship, however, contingent on characteristics of the regional environment. Therefore,
Schumpeter’s (1947) claim that “the inventor produces ideas, the entrepreneur ‘gets
things done”’ could be aligned with the existence of spatially bounded spillover effects.
7 Appendix
See Table 5.
123
Table 5 Regions included in the empirical analysis
Belgium (NUTS1) Greece (NUTS2-3) cont’d Slovenia (NUTS2/3) cont’d

Brussel Macedonia & Thrace Koroska, Savinjska,
Vlaams Gewest Thessaly & Epiros Spodnjeposavska
Region Wallone Central + Ionian Islands Pomurska and Podravska
+ Antwerp (NUTS2) Peloponnesus/Aegean/Crete
+ Ghent (NUTS2) Spain (NUTS2)
Hungary (NUTS2) Galicia
Denmark (NUTS2) Central Transdanubia Asturias
Copenhagen area Western Transdanubia Pais vasco
Sealand and Bornholm Southern Transdanubia Navarra
Funen Northern Hungary Aragon
Jutland Northern Great Plain Madrid
+ Aarhus area (NUTS3) Southern Great Plain Castilla y León
+ Budapest area (NUTS3) Castilla La Mancha
Finland (NUTS2) Extremadura
Ita-Suomi Ireland (NUTS2) Catalunya
Etela-Suomi Dublin + Barcelona (NUTS3)
+ Helsinki Area (NUTS3) Border, Midl., Western Comm Valenciana
Lansi-Suomi Southern and Eastern Baleares
Pohjois-Suomi Andalucia
Italy (NUTS1-2) + Valencia (NUTS3)
France (NUTS1) Nord-Ovest + Sevilla (NUTS3)
Ile de France Lombardia + Malaga (NUTS3)
Parisien Bassin Nord-Est Murcia
East Centro Canarias
West Campania
South-West Sud Sweden (NUTS2)
Center-East Sicilia Stockholm area
Meditéranee Östra Mellansverige
Netherlands (NUTS1) Sydsverige
Germany (NUTS1) Noord-Nederland Norra Mellansverige
Baden-Württemberg Oost-Nederland Mellersta Norrland
+ Stuttgart (ROR) West-Nederland Övre Norrland
Bayern + Utrecht (NUTS3) Småland med öarna
+ München (ROR) + Amsterdam (NUTS3) Västsverige
Berlin + The Hague (NUTS3)
Brandenburg + Rotterdam (NUTS3) Switzerland (∼NUTS1/2)
Hamburg Zuid-Nederland North-East
Hessen North-West
+ Rhein-Main (ROR) Norway (∼NUTS2) South
Mecklenburg-Vorpommern North Norway West (French speaking)
123
Table 5 continued
Niedersachsen Middle Norway

Nordrhein-Westfalen West Norway United Kingdom (NUTS1)
+ Duisburg-Essen (ROR) South Norway Scotland
+ Düsseldorf (ROR) Oslo and surroundings North East
+ Köln (ROR) North West
Rheinland-Pfalz Portugal (NUTS1) Yorkshire Humberside
Sachsen Norte (incl Porto) East Midlands
Sachsen-Anhalt Centro West Midlands
Schleswig-Holstein Lisboa e Vale de Tejo East Anglia
Thüringen Greater London
Slovenia (NUTS2-3) South East
Greece (NUTS2-3) Kraska South West
Athens Dolenjska, Osrednjeslovenska, Wales
Macedonia & Thrace Zasavska Northern Ireland
This list only includes the regions with GEM sample sizes higher than 700 observations. Regions in Norway,
Slovenia and Switzerland are not included in model 2b and beyond
+ Urban area has been abstracted from larger surrounding areas (these abstracted areas are not visualized
in Figs. 1, 2 and 3)
ROR: ‘Raumordnungsregionen’. This classification for German regions indicates labour market areas; its
spatial scale lies between the European NUTS2 and NUTS3 classification
References
Acs ZJ (2002) Innovation and the growth of cities. Edward Elgar, Cheltenham
Acs ZJ (2008) Foundations of high impact entrepreneurship. Jena economic research papers. #, 2008–060.
Max Planck Institute of Economics, Jena
Acs ZJ, Armington C (2004) Employment growth and entrepreneurial activity in cities. Reg Stud
38:911–927
Acs ZJ, Varga A (2005) Entrepreneurship, agglomeration and technological change. Small Bus Econ
24(3):323–334
Acs ZJ, Storey DJ (2004) Introduction: entrepreneurship and economic development. Reg Stud
38(8):871–877
Acs ZJ, Plummer LA (2005) Penetrating the “knowledge filter” in regional economies. Ann Reg Sci
39(3):439–456
Acs ZJ, Audretsch DB, Braunerhjelm P, Carlsson B (2010) The missing link: knowledge diffusion and
entrepreneurship in endogenous growth. Small Bus Econ 34(2):105–125
Acs ZJ, Bosma NS, Sternberg R (2011) Entrepreneurship in world cities. In: Minniti M (ed) The dynamics
of entrepreneurial activity. Oxford University Press, Oxford, pp 125–151
Aghion P, Blundell R, Grifith R, Howitt P, Prantl S (2009) The effects of entry on incumbent innovation
and productivity. Rev Econ Stat 91(9):20–32
Andersson M, Koster S (2011) Sources of persistence in regional start-up rates-evidence from Sweden.
J Econ Geogr 11(1):179–201
Artis MJ, Miguelez E, Moreno R (2012) Agglomeration economies and regional intangible assets: an
empirical investigation. J Econ Geogr, first published, online 4 Nov 2011. doi:10.1093/jeg/lbr031
Asheim BT (1996) Industrial districts as ‘learning regions’: a condition for prosperity. Eur Plan Stud
4(4):379–400
Audretsch DB, Feldman MP (1996) R&D spillovers and the geography of innovation and production. Am
Econ Rev 86(3):630–640
Audretsch DB, Keilbach MP (2005) Entrepreneurship capital and regional growth. Ann Reg Sci 39:457–469
123
Barca F (2009) An agenda for a reformed cohesion policy: a place-based approach to meeting European
Union challenges and expectations. Report for the European Commission, Brussels
Barca F, McCann P, Rodriguez-Pose A (2012) The case for regional development intervention: place-based
versus place-neutral approaches. J Reg Sci 52:134–152
Baumol WJ (1990) Entrepreneurship: productive, unproductive, and destructive. J Polit Econ 98(5):893–921
Beaudry C, Schiffauerova A (2009) Who’s right, Marshall or Jacobs? The localization versus urbanization
debate. Res Policy 38:318–337
Becker R, Henderson JV (2000) Intraindustry specialization and urban development. In: Huriot JM,
Thisse JF (eds) Economics of cities: theoretical perspectives. Cambridge University Press, Cambridge,
pp 138–166
Bosker M (2007) Growth, agglomeration and convergence: a space-time analysis for European regions.
Spatial Econ Anal 2:91–108
Bosma NS, Van Praag M, Thurik R, De Wit G (2004) The value of human and social capital investments
for the business performance of startups. Small Bus Econ 23(3):227–236
Bosma N, Schutjens V, Stam E (2009) Entrepreneurship in European regions: implications for public policy.
In: Leitao J, Baptista R (eds) Public policies for fostering entrepreneurship. Springer, New York, pp 59–89
Bosma NS (2011) Entrepreneurship, urbanisation economies and productivity of European regions. In:
Fritsch MF (ed) Handbook of research on entrepreneurship and regional development, Cheltenham
(UK). Edward Elgar, Northampton, pp 107–132
Bosma NS, Wennekers S, Amorós JE (2012) Global entrepreneurship monitor 2011 extended global report:
entrepreneurs and entrepreneurial employees across the globe, Babson Park, MA, US: Babson College,
Santiago, Chile: Universidad del Desarollo, Kuala Lumpur, Malaysia: Universiti Tun Abdul Razak and
London, UK: Global Entrepreneurship Research Association
Bosma NS, Stam E, Schutjens VAJM (2011) Creative destruction and regional productivity growth: evidence
from Dutch manufacturing and services. Small Bus Econ 36(4):401–418
Brenner T, Fornahl D (2008) Regional path-dependence in start-up activity. Papers in evolutionary economic
geography nr. 08.12. Utrecht University, Utrecht
Breschi S, Lissoni F (2009) Mobility of skilled workers and co-invention networks: an anatomy of localized
knowledge flows. J Econ Geogr 9:439–468
Breschi S, Malerba F, Orsenigo L (2000) Technologicl regimes and Schumpetarian patterns of innovation.
Econ J 110(463):388–410
Camagni R (ed) (1991) Innovation networks. Spatial perspectives. Bellhaven Press, London
Ciccone A (2002) Agglomeration effects in Europe. Eur Econ Rev 46(2):213–227
Ciccone A, Hall RE (1996) Productivity and the density of economic activity. Am Econ Rev 86(1):54–70
Combes PP, Overman H (2004) The spatial distribution of economic activities in the European Union.
In: Henderson JV, Thisse J (eds) Handbook of regional and urban economics. Elsevier, Amsterdam,
pp 2120–2167
Cooke P (2001) Regional innovation systems, clusters, and the knowledge economy. Ind Corp Change
10(4):945–974
Coyne CJ, Sobel RS, Dove JA (2010) The non-productive entrepreneurial process. Rev Austrian Econ.
doi:10.1007/s11138-010-0124-2
De Groot HLF, Poot J, Smit MJ (2009) Agglomeration externalities, innovation and regional growth:
theoretical perspectives and meta-analysis. In: Nijkamp P, Capello R (eds) Handbook of regional growth
and development theories. Edward Elgar, pp 256–281
Dogaru T, van Oort F, Thissen M (2011) Agglomeration economies in European regions: perspectives for
objective-1 regions. Tijdschrift voor Econ Soc Geogr 102:486–494
Dühr S, Colomb C, Nadin V (2010) European spatial panning and territorial cooperation. Routledge, London
Duranton G, Puga D (2001) Nursery cities: urban diversity, process innovation, and the life cycle of products.
Am Econ Rev 91:1454–1477
Duranton G, Puga D (2004) Micro-foundations of urban agglomeration economies. In: Henderson JV, Thisse
J-F (eds) Handbook of urban and regional economics, vol. 4. North Holland, New York, pp 2063–2118
Feser EJ (2002) Tracing the sources of local external economies. Urban Stud 39:2485–2506
Frenken K, van Oort FG, Verburg T (2007) Related variety, unrelated variety and regional economic growth.
Reg Stud 41:685–697
Fritsch M, Mueller P (2007) The persistence of regional new business formation-activity over time—
assessing the potential of policy promotion programs. J Evol Econ 17:299–315
123
Fritsch M, Mueller P (2008) The effect of new business formation on regional development over time: the
case of Germany. Small Bus Econ 30(1):15–29
Fritsch M, Schroeter A (2011) Why does the effect of new business formation differ across regions? Small
Bus Econ 36(4):383–400
Getis A, Ord JK (1992) The analysis of spatial association by use of distance statistics. Geogr Anal
24(3):189–206
Glaeser EL, Kallal HD, Scheinkman JA, Schleifer A (1992) Growth in cities. J Polit Econ 100:1126–1152
Glaeser EL, Maré DC (2001) Cities and Skills. J Labor Econ 19(2):316–342
Henderson JV (2003) Marshall’s scale economies. J Urban Econ 53:1–28
Holcombe RG (1998) Entrepreneurship and economic growth. Q J Austrian Econ 1(2):45–62
Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localization and knowledge spillovers as evi-
denced by patent citations. Q J Econ 108:577–598
Keeble D, Wever E (1986) New firms and regional development in Europe. Croom Helm, London
Koster S (2007) The entrepreneurial and replication function of new firm formation. Tijdschr Econ Soc
Geogr 98(5):667–674
Krugman P (2009) The increasing returns revolution in trade and geography. Am Econ Rev 99:561–571
Lopez-Rodriguez J, Faiña A (2006) Objective 1 regions versus non-Objective 1 regions: what does the Theil
Index tell us? Appl Econ Lett 13:815–820
Marrocu E, Paci R, Usai S (2012) Productivity growth in the old and new Europe: the role of agglomeration
externalities. J Reg Sci. doi:10.111/jors.12000
Martin R, Sunley P (2006) Path dependence and regional economic evolution. J Econ Geogr 6:395–437
McCann P, van Oort FG (2009) Theories of agglomeration and regional economic growth: a historical
review. In: Capello R, Nijkamp P (eds) Handbook of regional growth and development theories. Edward
Elgar, Cheltenham, pp 19–32
Melo PC, Graham DJ, Noland RB (2009) A meta-analysis of estimates of urban agglomeration economies.
Reg Sci Urban Econ 39:332–342. doi:10.1016/j.regsci.urbeco.2008-12.002
Montfort P (2009) Regional convergence, growth and interpersonal inequalities across the EU. DG Regional
Policy, European Commission
OECD (2009) Regions matter: economic recovery, innovation and sustainable growth. OECD, Paris
OECD (2011) Regions at a glance 2011. OECD, Paris
OECD (2012) Redefining “urban”. A new way to measure metropolitan areas, OECD, Paris
Olson M (1982) The rise and decline of nations: economic growth, stagflation, and social rigidities. Yale
University Press, Connecticut
Ottaviano GP (2011) ‘New’ new economic geography: firm heterogeneity and agglomeration economies.
J Econ Geogr 11:231–240
Parker SC (2005) Explaining regional variations in entrepreneurship as multiple equilibria. J Reg Sci
45:829–850
Reynolds P, Storey DJ, Westhead P (1994) Crossnational comparisons of the variation in new firm formation
rates: an editorial overview. Reg Stud 28(4):343–346
Reynolds PD, Bosma NS, Autio E, Hunt S, De Bono N, Servais I, Lopez-Garcia P, Chin N (2005) Global
entrepreneurship monitor: data collection design and implementation 1998–2003. Small Bus Econ
24(3):205–231
Rosenthal SS, Strange WC (2004) Evidence on the nature and sources of agglomeration economies. In:
Henderson JV, Thisse J-F (eds) Handbook of urban and regional economics, vol 4. Elsevier, Amsterdam,
pp 2119–2172
Rosenthal SS, Strange WC (2001) The determinants of agglomeration. J Urban Econ 59:191–229
Santarelli E, Vivarelli M (2007) Entrepreneurship and the process of firms’ entry, survival and growth. Ind
Corp Change 16(3):455–488
Schumpeter JA (1947) The creative responses in economic history. J Econ Hist 7(2):149–159
Schumpeter JA (1934) The theory of economic development. Harvard University Press, Cambridge
Schumpeter JA (1942) Capitalism, socialism and democracy. Harper and Row, New York
Stam E (2008) Entrepreneurship and innovation policy. In: Nooteboom B, Stam E (eds) Micro-foundations
for innovation policy. Amsterdam University Press, Amsterdam
Storper M (1992) The limits to globalization. Technology districts and international trade. economic Geogr
68(1):60–93
Tamásy C (2006) Determinants of regional entrepreneurship dynamics in contemporary Germany: A Con-
ceptual and Empirical Analysis. Reg Stud 40(4):365–384
123
Thissen MJPM, Ruijs A, van Oort F, Diodato D (2011) De concurrentiepositie van Nederlandse regio’s.
Regionaal-economicsche samenhang in Europa, Netherlands Environmental Assessment Agency (PBL),
The Hague
Thoma G, Torrisi S, Gambardella A, Guellec D, Hall BH, Harhoff D (2010) Harmonizing and combining
large datasets—an application to firm-level patent and accounting data. NBER working paper 15851,
National Bureau of Economic Research, Cambridge
United Nations (2007) World population prospects: the 2006 revision and world urbanization prospects: the
2007 revision. Population division of the Department of economic and social affairs, UN Washington
Van den Born JA (2009) The drivers of career success of the job-hopping professional in the new net-
worked economy; The challenges of being an entrepreneur and an employee. Ph.D Dissertation Utrecht
University, Utrecht/Born to Grow
Van Oort F, Lambooy J (2012, forthcoming) Cities, knowledge and innovation. In: Fischer M, Nijkamp P
(eds) Handbook of regional science. Springer, Berlin
Van Oort FG, Burger M, Knoben J, Raspe O (2012) Multilevel approaches and the firm-agglomeration
ambiguity in economic growth studies. J Econ Surv 26:468–491
World Bank (2009) World Development Report: Reshaping economic geography. World Bank,
Washington, DC
123
Copyright of Annals of Regional Science is the property of Springer Science & Business
Media B.V. and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

Van Oort & Bosma Aglomeration Economies, Inventor and Entrepreneurs

Uploaded by

Copyright:

Available Formats

Van Oort & Bosma Aglomeration Economies, Inventor and Entrepreneurs

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Van Oort & Bosma Aglomeration Economies, Inventor and Entrepreneurs

Uploaded by

Copyright:

Available Formats

Ann Reg Sci (2013) 51:213–244

SPECIAL ISSUE PAPER

Agglomeration economies, inventors and entrepreneurs

Frank G. van Oort · Niels S. Bosma

Abstract In economic agglomeration studies, the distinction of various externali-

JEL Classification R12 · L26 · O30

F. G. van Oort (B)

of knowledge diffusion, like spin-offs, research collaborations and social networks,

2 Entrepreneurship and innovation as nexus of regional productivity

2.1 Agglomeration economies in need of conceptual renewal

The empirical evidence of agglomeration economies is strong, and in an overview

2.2 Regional innovation and entrepreneurship as externalities contexts

in terms of growth. Referring to Baumol’s categorization of entrepreneurship in pro-

2.3 Building regional indicators from individual-level attributes

2.4 Cities, transfer mechanisms and place-based development

A regional perspective means taking into account agglomeration economies, and

Here, i j denotes an index of total productivity in the region. Coefficient α

Coefficient θ , therefore, measures the effect of the regional density of employ-

log Q i j − log Ni j = log j + θ (log Ni j − log Ai j )

which we can rewrite in the form of an estimation equation as follows:

log Q i j − log Ni j = α j + θ (log Ni j − log Ai j )

where H, SD, E and I denote, respectively, human capital, specialization/diversity

4 Data and approach

Table 1 Descriptive statistics, regional level, N = 111

Mean SD Correlation table

1. Labour 3.88 0.39

Model 1 Model 2a Model 2b Model 2c

Employment density (ln), Average 2001–2006 0.053*** 0.039** 0.034* 0.043**

5.1 The initial model

5.2 Adding agglomeration economies

As noted, model 2c reveals that the specialization/diversity measure that is hypothe-

indicator is always positively significant attached to productivity. But now also, in

5.3 Adding entrepreneurship and invention as knowledge transfer mechanisms

Model 2c (repeated) Model 3a Model 3b Model 3c Model 3d Model 3e Model 3f Model 3g

Model 2c (repeated) Model 3a Model 3b Model 3c Model 3d Model 3e Model 3f Model 3g

Constant 4.32*** 4.24*** 4.22*** 4.17*** 4.20*** 4.24*** 4.11*** 4.17***

and entrepreneurs), the importance of the density of employment in conjunction with

5.4 Spatial regime analyses

Table 4 2SLS estimation results by objective-1 and urban size regimes

Employment density 0.016 0.097 −0.060 0.052 −0.007

6 Conclusions and discussion

Table 5 Regions included in the empirical analysis

Belgium (NUTS1) Greece (NUTS2-3) cont’d Slovenia (NUTS2/3) cont’d

Niedersachsen Middle Norway

You might also like

Employment density (ln), Average 2001–2006 0.053* 0.039 0.034* 0.043**

Constant 4.32* 4.24* 4.22* 4.17* 4.20* 4.24* 4.11* 4.17*