Clusters and Entrepreneurship∗

Mercedes Delgado
Michael E. Porter
Scott Stern

December, 2009

Abstract

This paper evaluates the role of regional clusters on entrepreneurship in regional industries. We
focus on the distinct influences of convergence and agglomeration on the rate of growth in the
number of start-up firms and in employment by start-up firms. While reversion to the mean and
diminishing returns to specialization within a location can result in a convergence effect, the
presence of complementary economic activity creates externalities that enhance incentives and
reduce barriers for new business creation. Clusters are a particularly important channel by which
location-based complementarities are realized. Using a novel panel dataset, there is significant
evidence for the impact of clusters on entrepreneurship, after controlling for the impact of
convergence at the region-industry level: industries located in regions with a large presence of
related industries (i.e., strong clusters) experience higher growth in new business formation and
start-up employment. Furthermore, strong clusters contribute to the level of employment in young
start-ups in regional industries.

Keywords: Entrepreneurship, Industry Clusters, Dynamic Economies of Agglomeration.

∗
Acknowledgements: William Kerr, Ben Jones, Christian Ketels, Jim Davis, Christian Asmussen, Roc
Armenter and two referees contributed helpful suggestions.. The research in this paper was conducted
while the authors were Special Sworn Status researchers of the US Census Bureau at the Boston Census
Research Data Center (BRDC). Research results and conclusions expressed are our own and do not
necessarily reflect the views of the Census Bureau. This paper has been screened to insure that no
confidential data are revealed. Author contact information: Mercedes Delgado (Temple University, Fox
School of Business; [email protected]), Michael E. Porter (HBS; [email protected]); and Scott Stern
(Northwestern University, MIT, NBER; [email protected]).

1
 1. Introduction
In this paper, we examine the impact of agglomeration on the growth of
entrepreneurship at the regional level. In particular, we focus on the role of clusters, or
agglomerations of closely related industries, in new business formation. Large variations
in regional employment growth and in the rate of firm creation are a striking feature of
the US economy (Porter, 2003).1 While a significant body of work explores why some
regions experience more rapid growth than others (Porter, 1990, 1998; Saxenian, 1994;
Glaeser, et al, 1992; Barro and Sala-i-Martin, 1995; Venables, 1996; Henderson, 1997;
Fujita, Venables, and Krugman, 1999), there is increasing academic and policy interest in
the particular role played by entrepreneurship. Startups seem to be an important driver of
net regional employment growth (Davis et al, 1996; Haltiwanger, et al, 2009), and there
is large regional variance in startup formation across regions (Armington and Acs, 2002).
A significant debate is underway regarding the role of the regional economic
environment in shaping differences in the rate of regional entrepreneurship and overall
economic performance (Porter, 1990 1998; Saxenian, 1994; Feldman, 2001; Armington
and Acs, 2002; Acs, et al, 2009; Glaeser and Kerr, 2009).
In an effort to explain region-industry growth two countervailing economic forces
must be accounted for: convergence and agglomeration (Delgado, Porter, and Stern,
2007). Convergence arises when, due to diminishing returns, the potential for growth is
declining in the level of economic activity (Barro and Sala-i-Martin, 1992).2
Agglomeration exerts an opposite force on regional evolution. In the presence of
agglomeration economies, the potential for growth is increasing in the level of economic
activity (Glaeser, et al, 1992; Henderson, et al, 1995). From an empirical perspective,
distinguishing the relative importance and differential impact of convergence and
agglomeration has been problematic. For example, if both convergence and
agglomeration effects are present, the impact of the initial level of economic activity on

1
For example, using Bureau of Economic Analysis Economic Areas (EAs) as the unit of analysis, Porter
(2003) documents large cross-EA differences in employment growth during the 1990s, even when
conditioning on the initial level of EA employment, and there are even larger cross-EA differences in the
creation of new firms.
2
While many studies of convergence focus on diminishing returns at the regional level, (Barro and Sala-i-
Martin, 1995), convergence may also arise at more micro levels of analysis, such as the region-industry
level (Henderson, et al, 1995; Dumais, et al, 2002, Bostic, et al, 1997; Cingano and Schivardi, 2004;
Higgins et al, 2006). See Magrini (2004) for a recent review of the convergence literature.

2
growth will reflect a balancing of the two effects, making it infeasible to identify either
effect in isolation (Henderson, et al, 1995).
We move beyond this traditional impasse by identifying the impact of industrial
agglomeration while simultaneously accounting for the impact of convergence. Our key
insight is that while forces that give rise to both convergence and agglomeration operate
within narrow economic units, agglomeration across complementary economic units can
have a separate and distinctive impact (Delgado, Porter, and Stern, 2007). Building on
Porter (1990, 1998, 2001), we focus specifically on the role of clusters –complementary
industries related by technology, skills, demand and other linkages. Agglomeration is a
consequence of the presence of complementary economic activity and clusters are a
particularly important mechanism by which location-based complementarities are
realized.
The main contribution of this paper is to examine a particularly important channel
through which cluster-driven agglomeration might operate: entrepreneurship. The
presence of a cluster of related industries in a location will foster entrepreneurship by
lowering the cost and risk of starting a business, enhancing opportunities for innovation,
and enabling better access to a more diverse range of inputs and complementary products.
(Saxenian 1994, Porter 1998, 2000, Feldman and Francis 2004, Glaeser and Kerr 2009).
The co-location of companies, customers, suppliers, and other institutions increases the
perception of innovation opportunities while amplifying the pressure to innovate (Porter,
2000). Since entrepreneurs are essential agents of change and innovation, a strong cluster
environment should foster entrepreneurial activity.
The empirical analysis exploits the establishment-level Longitudinal Business
Database (LBD) of the Census Bureau and the cluster definitions from the US Cluster
Mapping Project (Porter 2001, 2003). This classification system defines clusters as
collections of industries with high levels of co-location in terms of employment. We
consider several related measures of the cluster environment surrounding a region-
industry, including a measure based on individual clusters, a more encompassing
measures that incorporates linkages among related clusters (i.e., “linked” clusters), and a
third measure that captures the strength of similar clusters in neighboring regions. Using
both databases we measure entrepreneurship and industrial composition at the region-

3
cluster-industry level. We focus on a dataset that spans the years 1990-2005, includes
177 mutually exclusive Economic Areas (EAs) in the contiguous United States, and
incorporates (up to) 588 “traded” industries spanning 41 clusters for each EA.3
Our empirical work focuses on early stage entrepreneurship, which we measure
using two related indicators of start-up activity. We measure the number of new
establishments by new firms in a region within a given traded industry (which we refer to
as the level of start-up establishments), and by the employment in these new firms
(which we refer to as the level of start-up employment). We then compute the growth
rate in start-up establishments and start-up employment in each regional industry, and
estimate the impact of cluster-driven agglomeration while accounting for the impact of
convergence. Our core specifications incorporate detailed controls, including region and
industry fixed effects. In other words, we estimate the impact of cluster composition on
entrepreneurship, relying exclusively on variation arising from the relative size or
strength of each cluster within a given region, and accounting for the overall growth in
start-up activity of a given region and industry.
We find striking evidence for the simultaneous yet distinct influences of
agglomeration and convergence on the growth rate of start-up establishments and start-up
employment. Growth in start-up employment at the region-industry level is declining in
the initial level of start-up employment at the region-industry level, consistent with the
presence of a convergence effect. At the same time, the start-up employment growth rate
is increasing in measures of the strength of the cluster environment. We find similar
findings for the growth rate of start-up establishments. By accounting for convergence
and the potential for competition within each regional industry, we are able to isolate the
positive impact of cluster-related complementarities on entrepreneurship. The positive
impact of clusters is both quantitatively and statistically significant, and robust to a
variety of checks. The results provide support for the hypothesis that strong clusters
facilitate growth in entrepreneurship over time.
While we primarily focus on the formation of start-up firms, for robustness we
also look at total new establishments (including new establishments of existing firms).

3
Traded industries are those that sell products and services across regional and national boundaries. See
Porter (2003) and Section 4 for detailed explanation on traded industries and clusters.

4
We find that stronger clusters also facilitate the entry of new establishments of existing
firms. While more research is needed on the locational and organizational decisions of
multi-establishment (and multinational) firms (Holm, Malmberg, and Solvell, 2003;
Mudambi and McCann, 2005; Alcacer, 2006), these firms may be opening establishments
in new locations to seek complementary clusters and benefit from the competitive
resources of each location (Enright, 2000; Bresnahan and Gambardella, 2004; Manning
2008). The contribution of these firms to generate entrepreneurial opportunities in a
particular regional cluster is an open research question that we examine in related work.
Finally, we also examine the role of clusters in the performance of start-up firms.
A strong cluster may raise the productivity of the participating firms, raising as well the
bar for survival of new businesses (Sorenson and Audia, 2000). At the same time, a
strong cluster in a region could facilitate the growth of start-up firms by providing better
access to the necessary specialized inputs to commercialize their products and services.
Thus, we expect clusters will enhance the performance for the most productive start-ups.
To test this, we study the level of employment in young start-up survivors in a region-
industry. After controlling for the level of start-up activity in the base period, we find that
the cluster environment contributes to improve the level of employment in young start-up
survivors, suggesting that clusters facilitate the survival and potentially the growth of
successful start-ups.
The rest of the paper is organized as follows. We begin by discussing the
relationship between entrepreneurship and cluster-driven agglomeration, and develop the
main hypotheses. Section 3 presents the empirical framework. Section 4 explains the
data, and Section 5 discusses the main findings. A final section concludes.

2. Clusters and Entrepreneurship

Numerous types of mechanisms are associated with entry of new businesses in
agglomerated areas. On the one hand, starting with Marshall (1920), regional studies
have highlighted at least three distinct drivers of agglomeration: knowledge spillovers,
input-output linkages, and labor market pooling. Over time, an extensive literature has
also incorporated additional agglomeration drivers, including local demand, specialized
institutions and the structure of regional business and social networks (Porter 1990, 1998,

5
2000; Malecki, 1990; Saxenian 1994; Markusen 1996). While most empirical studies of
agglomeration focus on variables such as the aggregate rate of employment growth, an
emerging theoretical and empirical literature emphasizes the role of the formation and the
dynamics of new businesses in regional economic growth (Davis et al, 1996; Acs and
Armington, 2006; Haltiwanger, et al, 2009). Relative to business expansions by
incumbent firms, entrepreneurs may be more likely to be able to identify opportunities --
both in the form of new technologies and new markets -- that exploit distinctive sources
of regional comparative advantage. However, there is a large churning of firms in most
countries and sectors, especially amongst young and small businesses (Davis et at, 1996;
Dunne et al, 1988, 2005; Barteisman, et al, 2005). Start-up firms have greater exit rates
than new establishments of existing firms since they lack experience in the industry and
in the location (Dunne et al, 2005; Kerr and Nanda, 2009). Importantly, those start-up
firms that survive tend to have greater growth potential than incumbent firms
(Barteisman, et al, 2005).4 As such, the presence of a strong cluster environment that
reduces barriers to entry and to firm growth and enables regional comparative advantage
will be a central driver of entrepreneurial growth.
The precise mechanisms by which the structure of the regional economic
environment impacts entrepreneurship are numerous and subtle. Chinitz (1961)
hypothesizes that a key requirement for entrepreneurship is the presence of a network of
smaller suppliers, and attributes differences in the rate of entrepreneurship between New
York and Pittsburgh at that time to differences in the structure of suppliers. Building on
these earlier studies, a rich (though mostly qualitative) literature has emerged examining
the relationship between entrepreneurship and regional economic performance. For
example, Saxenian (1994) attributes the success of Silicon Valley to the culture of
entrepreneurship (relative to Route 128) and a more decentralized organization of
production. An extensive literature also highlights the broader relationship between
entrepreneurship and the regional innovation system (e.g., Audretsch, 1995; Feldman
2001; Shane 2001; Armington and Acs, 2002; Acs, et al, 2009). Recently, Glaeser and
Kerr (2009) carefully test for the impact of specific Marshallian economies of

4
Barteisman, et al (2005) find that start-ups are of significantly lower size in the US than in Europe, but
they seem to enjoy lower barriers to firm growth reaching rapidly a higher average size.

6
agglomeration on new firm entry; while their analysis does not specifically evaluate the
impact of clusters, they provide complementary evidence that the presence of small
suppliers and workers in relevant occupations is associated with a higher level of new
business creation.
At the same time, a small but growing (and mostly independent) literature within
regional and international business studies examine how the location decision and the
benefits of agglomeration depend on attributes of the firm as well as on attributes of the
industry and the location (Saxenian 1994; Henderson 2003; Rosenthal and Strange 2003;
Alcacer 2006; McCann and Mudambi 2005; among others). There are key interactions
between the internal organization of the firm (start-ups, multi-location, multinational,
small or large, young or old, corporate organization, etc.) and the agglomeration benefits
from a geographical location. One dimension that has received special attention is the
role of small firms in extracting and generating economies of agglomeration. For
example, Henderson (2003) finds that the extent of localization economies is larger for
single-unit plants (vs. multi-unit plants), which tend to be more dependent on the external
environment. Related work suggests that the presence of smaller (and younger) firms
spurs additional new business creation and regional employment growth (Glaeser et al.
1992, Rosenthal and Strange, 2009; Glaeser, Kerr and Ponzetto 2009, and Feberman,
2007, among others).
In contrast, some regional and cluster studies highlight the particular importance
of so-called “anchor” firms (including multi-establishment and multinationals) that
induce spin-offs and attract firms from related industries (Agrawal and Cockburn 2002,
Enright 2000, Feldman, Francis, and Bercovitz 2005, Klepper 2007, Greenstone, et al,
2008; Manning, 2008). Further research is needed to understand the location decision of
multi-location firms and their contribution to the entrepreneurial activities in a particular
region.
The main goal of this paper is to test whether the presence of related economic
activity in a region facilitates the growth of start-up establishments and start-up
employment in regional industries. Drawing on the cluster theory we suggest that to
capture important agglomeration forces the focus should be the presence of
complementary and related industries instead of industry diversity per se (Porter 1990,

7
1998; Feldman and Audretsch, 1999; Delgado, Porter and Stern, 2007).5 Thus, we use
Porter’s (1998, 2003) empirical cluster framework to explore agglomeration across set of
industries related by technology, skills, demand, or other linkages, abstracting from the
individual underlying mechanisms that induce agglomeration benefits.
Our first hypothesis focuses on the relationship between the growth of start-up
activity and the initial level of start-up activity within a region-industry. This relationship
will be subject to a convergence effect, which can be interpreted in terms of mean-
reversion or diminishing marginal returns to entrepreneurial opportunities. Mean
reversion simply implies that a region-industry that has a relatively high level of start-up
activity at t0 (compared to the average start-up activity in the industry in other regions
with similar size and economic composition) is more likely to experience a lower
(stochastically determined) growth rate of start-up activity between t0 and t1 (Barro and
Sala-i-Martin, 1991; Quah, 1996; Henderson, et al, 1995). At the same time, it is
possible that the returns to entrepreneurial activity are diminishing in the level of
entrepreneurial activity as the result of input scarcity. For example, if the price of
specialized (labor or capital) inputs is increasing in the intensity of competition among
start-up firms, there will be diminishing returns to entrepreneurship as a result of
congestion costs (Sorenson and Audia, 2000). As a result, a high level of
entrepreneurship in a particular region-industry at a point in time may result in
diminished near-term opportunities for entrepreneurship in that same region-industry.
Thus, our first hypothesis is that there will be convergence in entrepreneurship at the
narrowest unit of analysis: the region-industry growth rate of start-up activity will be
declining in the level of region-industry start-up activity.
Our remaining hypotheses focus on the impact of related economic activity on the
growth rate of start-up activity. Conditional on the traditional convergence effect, the
relationship between related economic activity and entrepreneurship depends on how the
presence of particular types of economic activity impacts entrepreneurial incentives. On

5
The literature on agglomeration economies highlights numerous channels that may facilitate positive
externalities (see the review by Rosenthal and Strange, 2004), including increasing returns to industry
specialization and to diversity at the regional level (Glaeser et al, 1992; Henderson, et al, 1995). The
cluster theory challenges this conceptualization of industry specialization and regional diversity by
focusing on the role of clusters of related and complementary industries.

8
the one hand, the returns to entrepreneurship are lower in the face of intensive
competition, and so the incentives for start-up entry in a particular location will be lower
in the presence of a higher level of local price-based competition (Porter, 1980;
Bresnahan and Reiss, 1991). At the same time, the presence of complementary economic
activity – specialized suppliers, a local customer base, producers of complementary
products and services – increases the pool of inputs available and enhances the range and
diversity of profitable entry opportunities and so improves entrepreneurial incentives.
The empirical relationship between entrepreneurship and particular types of pre-existing
economic activity will therefore depend on whether these activities are substitutes or
complements (Bulow, et al, 1985).
It is useful to distinguish, then, between the level of specialization of a region in a
particular industry and the strength of the cluster environment around that industry. On
the one hand, the relationship between industry specialization and entrepreneurship
growth is ambiguous. While industry specialization in a particular location may enhance
opportunities for learning, innovation, and entrepreneurial spawning (Audretsch, 1995;
Gompers, et al, 2005; Glaeser and Kerr, 2009), a large presence of established firms
(relative to the size of the national industry) intensifies local competition, dampening
incentives for entrepreneurial entry. Our second hypothesis is, then, that the ultimate
empirical relationship between industry specialization and the growth rate of
entrepreneurship is ambiguous, and will depend on the precise nature of competition
(cost-based or innovation-based) and the pattern of strategic interaction between entrant
and established firms.6
In contrast, a strong cluster environment surrounding a particular region-industry
enhances the incentives and potential for entrepreneurship. The firms within a
geographically concentrated cluster share common technologies, skills, knowledge,
inputs, consumers, and institutions, facilitating agglomeration across complementary and
related industries (Porter 1990, 1998, 2003; Feldman and Audretsch, 1999; Delgado,
Porter and Stern, 2007). A strong cluster environment enhances growth at the region-
industry level by raising the returns to business expansion, capital investment, and

6
In the empirical analysis we do not test how the nature of competition affects start-up activity. Instead,
we test whether positive externalities (e.g., complementarities with the established firms) or congestion
forces (e.g., price-based competition effects) prevail at the region-industry level.

9
innovation, thereby increasing job creation and productivity (see e.g., Porter 1990, 1998,
2003; Saxenian 1994; Swann 1998; Feldman and Audretsch, 1999; Delgado, Porter and
Stern, 2007; Bresnahan and Gambardella, 2004; Bonte, 2004; Delgado, 2005; Cortright,
2006). More specifically, clusters of related and complementary industries facilitate new
business formation and the growth of successful start-ups by lowering the costs and risks
of entry (e.g., by providing low-cost access to specialized capital inputs, offering an
environment in which the costs of failure may be lower), enhancing opportunities for
innovation-based entry (as a stronger cluster environment will allow local entrepreneurs
to develop and commercialize new technologies more rapidly) and allowing start-up
firms to leverage local resources to expand new businesses more rapidly. Finally, strong
clusters are often associated with the presence of innovation-oriented demanding local
consumers, thus providing increased opportunities for entrepreneurial entry into emerging
and differentiated market segments. As a result, entrepreneurship is a particularly
important channel for cluster-driven agglomeration, and may therefore be crucial for the
role of clusters in enhancing regional economic performance (Porter 1998; Saxenian
1994; Swann 1998; Feldman 2001; Feldman and Francis 2004; Feser, Renski, and
Goldstein, 2008; Glaeser and Kerr, 2009; Wennberg and Lindqvist, 2008). Thus, our
third hypothesis is that, after controlling for the convergence effect, the growth rate of
entrepreneurship will be increasing in the strength of the cluster environment in the
region.
It is useful to also consider the impact of clusters in neighboring regions. On the
one hand, strong clusters in neighboring regions enhance the opportunities and lower the
costs of entrepreneurship (e.g., by providing access to suppliers and customers, by
allowing firms to leverage local technology and institutions, etc). Indeed, Delgado,
Porter and Stern (2007) find that clusters and industries that are co-located in nearby
regions benefit from inter-regional spillovers in employment growth. At the same time,
the presence of a strong cluster in a neighboring region is a source of locational
competition, particularly for capital investment and entrepreneurship. Hence
entrepreneurs may move to a neighboring region to open a business when that
neighboring region has a strong cluster environment, reducing the potential for
entrepreneurship growth for locations with weak cluster environments adjacent to strong

10
cluster environments. Therefore, our hypothesis is that the impact of the strength of
neighboring regions’ cluster environment on the growth rate of entrepreneurship is
ambiguous, and will depend on the relative salience of inter-regional spillovers versus
locational competition.
While this paper focuses on new business formation, we recognize that the link
between entrepreneurship and regional growth depends also on the survival and growth
of these new enterprises. Thus, we tentatively examine the role of clusters in the survival
of start-ups in regional industries. As mentioned earlier, a strong cluster in a region could
reduce the barriers to the growth of start-up firms by providing better access to the
necessary inputs to develop and commercialize their products and services. At the same
time, a strong cluster may raise the productivity of the participating firms, raising as well
the productivity bar for survival of new businesses. Thus, our hypothesis is that a strong
cluster environment will enhance the potential for growth for the most productive start-
ups, while also raising the quality and productivity bar for survival.

3. Econometric Model
To test our hypotheses, we need an empirical framework that allows us to
evaluate the distinct impacts of agglomeration and convergence forces on start-up
activity. We measure start-up activity in two related ways: the number of establishments
by new firms (with payroll) in a region within a given traded industry (i.e. start-up
establishments), and the employment in these new firms (i.e., start-up employment). We
are particularly interested in separating out the role played by industrial clusters in start-
up activity, while controlling for the economic activity within a region-industry, as well
as broader factors such as the overall growth of a region or industry. To do so, we adapt
the conditional convergence framework (Barro and Sala-i-Martin, 1991; Henderson, et al,
1995) and evaluate how the growth in start-up activity at the region-industry level is
impacted by the level of start-up activity, industry specialization, the strength of the
cluster environment, and region and industry fixed effects. Our core econometric
specification is therefore:

11
 ⎛ Start-up Activityi,c,r,2002−05 ⎞
ln ⎜
⎜ Start-up Activity ⎟⎟ = α 0 + δ ln(Start-up Activityi,c,r,1991−94 ) + β1 ln( Industry Speci,c,r,1990 ) +
⎝ i,c,r,1991− 94 ⎠

β 2 ln(Cluster Spec) i,c,r,1990

outside i
+ β3 ln(Linked Clusters Specoutside
c,r,1990 ) +
c

β 4 ln(Cluster Spec in Neighborsc,r,1990 ) +α i + α r + ε i,c,r,t . (1)

The dependent variable is the growth in start-up activity of traded industry i in cluster c at
region (EA) r, where the base period is the mean level of start-up activity during the
years 1991-1994, and the end period is the mean level of start-up activity during the year
2002-2005 (Section 4 includes a detailed explanation of this measure and its
construction). The explanatory variables include the level of start-up activity in the
region-industry, industry specialization and measures of the strength of related economic
activity: cluster specialization, the strength of linked (related) clusters, and the strength
of similar clusters in neighboring regions. These measures capture the relative scale and
strength of different types of economic activity potentially impinging on start-up activity
at the region-industry level. Our main hypotheses are that the growth rate in start-up
activity is subject to a convergence effect (δ < 0), is increasing in the strength of clusters
and linked clusters (β2 > 0 and β3 > 0), and has an ambiguous relationship with industry
specialization (β1) and the strength of clusters in neighboring regions (β4).
Our main econometric specification also accounts for other differences across
regions and industries that affect the start-up growth rate through the inclusion of
industry (αi) and EA fixed effects (αr). Our analysis thus controls for unobserved factors
(such as idiosyncratic demand shocks, regional policies, etc) that might be correlated both
with our measures of cluster specialization and the start-up growth rate in a particular
region-industry. Thus, our core identification structure estimates the impact of cluster
composition on entrepreneurship, relying exclusively on variation arising from the
relative size or strength of that cluster within a given region, accounting for the overall
growth of a given region and industry. Finally, to account for correlation across

12
industries within each regional cluster, the standard errors are clustered by region-
cluster.7

4. Data
To estimate equation (1), our dataset includes measures of start-up activity at the
region-industry level (at two points in time), as well measures of industry and cluster
specialization during a baseline period. We combine data from the Longitudinal
Business Database (LBD) of the Census Bureau with cluster definitions drawn from the
US Cluster Mapping Project (Porter, 2001, 2003). Before turning to the precise variable
definitions, it is useful to provide an overview of these two data sources.
The LBD provides annual observations of the universe of US establishments with
payroll from 1976 onward. For each establishment, the LBD includes the date of entry,
physical location, industry code, and number of employees of that establishment.
Importantly, the LBD offers both an establishment-level identifier and a firm-level
identifier, so it is possible to distinguish between entrepreneurship – the initial entry of a
new firm in its first establishment – and business expansions by existing firms through
the opening of new establishments.8 Our approach aggregates this data to the region-
industry level and the region-cluster level, using four-digit SIC codes as the primary
industry unit and economic areas (EAs) as the geographic unit.9
Our approach combines the LBD with a classification system for cluster
definitions drawn from the US Cluster Mapping Project (Porter, 2001, 2003). While the

7
Additionally, since nearby regions tend to specialize in the same type of clusters, there might be spatial
dependence of the performance and unobserved attributes of a region and its neighbors. For instance
spatial dependence in performance exists if the growth of neighboring industries and clusters influences
own-industry growth. Similarly, unobserved attributes of the neighboring regions, such as human capital
composition, may induce spatial dependence in the error terms. We take into account this potential spatial
dependence directly by including the cluster specialization of adjacent regions in our main specifications.
8
For detailed information on the LBD data see Jarmin and Miranda (2002). Other papers that examine
start-up formation based on this detailed Census Bureau data include Armington and Acs (2002), Glaeser
and Kerr (2009) and Kerr and Nanda (2009).
9
There are 179 BEA-defined EAs covering the entirety of the United States. To minimize concerns about
differences in transportation costs and the definition of neighboring regions, we exclude the Alaska and
Hawaii EAs. The boundaries of EAs are drawn to reflect meaningful economic regions, ensure
comprehensive regional coverage and have been highly stable over time (Johnson and Kort, 2004). EA
include both rural and urban areas, facilitating the mapping of clusters that span urban and proximate rural
areas (Porter, et al., 2004).

13
measurement of complementary economic activity in a consistent and unbiased manner is
a considerable challenge,10 the US Cluster Mapping Project (USCMP) develops a
methodology for grouping four-digit (and some three-digit) SIC codes into cluster and
linked cluster groupings.11 The methodology first distinguishes between three “types” of
industries with very different patterns of spatial competition and locational drivers:
traded, local, and natural resource-dependent. To focus our analysis on those industries
most closely linked to our underlying hypotheses, we focus exclusively on the traded
industries, where the relationship between start-up activity and cluster-driven
agglomeration is likely to be most salient. These traded industries consist of 588
(mostly) four-digit SIC codes that are associated with service and manufacturing
industries that sell products and services across regional and national boundaries.12
Porter (2001, 2003) assigns each traded industry into one of 41 mutually exclusive traded
clusters (referred to as “narrow clusters”), where the set of industries included in each
cluster primarily reflects pairwise correlations of industry employment across locations
(Appendix A provides a comprehensive list of the 41 traded clusters, and key
attributes).13 Examples of clusters include automotive, apparel, biopharmaceuticals, and

10
A small literature considers alternative classification schemes. Ellison and Glaeser (1997) study the
coagglomeration of manufacturing industries, creating an index reflecting “excess” concentration. Feldman
and Audretsch (1999) group those manufacturing industries that have a common science and technological
base, using the Yale Survey of R&D Managers. Other studies define linkages between industry activities
in terms of their technological and/or market proximity (Scherer, 1982; Jaffe, Trajtemberg and Henderson,
1993; Bloom, Schankerman and Van Reenen, 2005). Finally, Ellison, Glaeser and Kerr (2007) test various
mechanisms inducing co-agglomeration of industries, and conclude that input-output linkages are the most
relevant factor followed by labor pooling. This reasoning is consistent with the methodology developed in
Porter (2001, 2003). See also Feser and Bergman (2000) and Forni and Paba (2002).
11
For consistency with Porter’s cluster definitions, we focus on industries that are included in the publicly
available County Business Patterns data (i.e., private sector non-agricultural production, non-household,
and non-railroad employment). In order to use industry data back to 1990, the analysis employs SIC
system rather than the more refined NAICS systems, which was introduced in 1997 (and modified in 2002).
By construction, recent NAICS-based data can be translated (with some noise) into the older SIC system.
12
Traded industries account for over 87% of domestic US patents and 30% of total US employment
(Porter, 2003). In contrast, local industries do not agglomerate (are not localized) and focus on local
demand.
13
While the co-location of industries in a region does not guarantee interaction or spillovers (Boschma
2005, Torre 2008), consistent co-location across many regions strongly suggests that such interactions are
present. However, it is possible that in a few cases industries with high co-location of employment across
regions may have little economic relationship. Thus, in the USCMP two adjustments are made to the
cluster definitions to eliminate spurious correlations. First, the four-digit SIC industry definitions and list
of products and services are used to reveal the presence of logical externalities. Second, the National Input-
Output accounts are used to look for meaningful cross-industry flows (see Porter, 2003 pp. 563). While
industries that have meaningful economic interactions tend to co-agglomerate in space (see Ellison and

14
information technology. Within a cluster such as information technology, 9 individual 4-
digit SIC code industries are incorporated, including electronic computers (SIC 3571) and
software (SIC 7372). These cluster definitions form our key measures of complementary
economic activity.

Variable Definitions and Sample Description

Entrepreneurship. Our two main measures of entrepreneurial activity are start-up
employment and start-up establishments by new firms within a given EA-industry. By
focusing on start-up firms (those opening their first establishment), our measure offers a
reasonably proxy for the level of entrepreneurial activity in a given industry and location
at a given point in time. Specifically, Start-up Employment is defined as total level of
employment in new firms during their first year of operation (with payroll); and Start-up
Establishment is the count of these new firms. Consistent with prior work (Armington
and Acs, 2002; Glaeser and Kerr 2009), we computer 4-year averages for these annual
start-up activities.14 Using a multi-year span (and including a Census-year in the base
and terminal period) both allows for a more informative signal of the true level of
entrepreneurial activity and also significantly reduces the number of EA-industries in
which we observe zero entrants during a given period.15
One of the main goals of this paper is to evaluate how the cluster environment
impacts the growth rate of entrepreneurship. While this focus allows us to evaluate the
role of the cluster environment on regional dynamics, the most straightforward approach
to evaluating growth – taking ln(Start-up Activityi,r,2002-05 / Start-up Activityi,r,1991-94)--
must account for the fact that there are many EA-industries in which there is a zero level

Glaeser 1997), we recognize that there are non-geographical dimensions of proximity that could also
facilitate the interactions between industries and their firms. For example, Boschma (2005) and Torre
(2008) suggest that institutional, organizational, and temporary geographical proximity may be as
important as geographical proximity in facilitating knowledge transfers among firms. Our cluster
definitions can only indirectly capture these dimensions to the extent that they complement geographical
proximity.
14
When we aggregate the data at the region-industry level, we exclude establishments with missing
industry information, but include these observations to compute region-level and US-level totals.
Additionally, we drop establishments with zero employment and with very low wages (below half of the
minimum wage) or very high wages (above $2 million USD). Our key findings are robust to including
these wage outlier observations.
15
In the LBD data the inflow of new establishments may be recorded with some delay, with Census-years
being most accurate in terms of recording all new establishments.

15
of employment (i.e., non existing regional industry) or, relatedly, a zero level of start-up
activity in the study period. In either of these cases, we are required to either exclude
those observations with a zero value in the starting or ending period, or impose a positive
lower bound on the level of startup activity.
To convey our main results in the most concise way, we focus the bulk of our
analysis on a sample where we include EA-industries that have a non-zero level of
employment during 1990, and then focus on the growth rate in start-up activity among
those EA-industries where there is a pre-existing level of economic activity (thus making
a growth rate analysis meaningful). The resulting sample consists of 53,213 EA-
industries. While excluding EA-industries with zero employment is meaningful, our core
findings are robust to alternative treatments of this data issue that we discuss below.
To include in the analysis EA-industries where we observe zero start-up activity
in either the baseline or terminal periods, we set a minimum level of start-up activity of 1
employee and 0.01 establishments.16 We also demonstrate that our results are robust to
the (un-scaled) subsample which conditions on a positive level of EA-industry start-up
activity in both the baseline and terminal period (sample of 11,981 EA-industries).17
Interestingly, despite the fact that our regional clusters include both regions and
industries units that are quite narrow, a very high share of EA-clusters experience at least
a minimal level of entrepreneurial activity – for example, 85% of all EA-clusters have at
least one start-up establishment during the 1990-2005 period. This suggests that the large
number of EA-industries with zero start-ups is not due to the lack of start-up activity
within-clusters, but likely due to the small scale of some industries and EAs.
Finally, we further account for the large number of zeros by examining the impact
of the industry and cluster environment on the level of entrepreneurial activity, using all
potential EA-industry pairs (i.e., 588 industries by 177 EAs). These analyses include
probit specifications that directly evaluate whether “missing” EA-industry start-up

16
In other words, we scale the start-up activity indicators by adding the minimum annual start-up
employment and start-up establishments in the sample, which is a standard procedure to scale variables (see
e.g., Glaeser and Kerr, 2009). Since the average start-up establishment in a EA-industry is less than one,
we scale the number of start-up establishments by adding the smallest number in the sample (0.01).
17
While more than 60% of the 53,213 EA-industries experience some start-up activity over the 1990-2005
period, only 11,981 experience start-up activity in both the base and terminal periods. This high skewness
of start-up activity in regional industries has been documented in other studies that use narrow regional
and/or industry units (see e.g., Glaeser and Kerr, 2009; Rosenthal and Strange, 2003).

16
activity is related to the overall cluster environment, and count models that explicitly
accounts for the skewed (count data) distribution of start-up activity in our dataset (see
Table 7).
There are large differences across EAs and clusters in the level of start-up
activity. At the cluster level, the average (1990-2005) annual start-up establishments (as
% of establishments in the cluster) varies from a maximum of 4.4% in Business Services
to a minimum of 0.59% in Power Generation and Transmission (See Table A1). At the
EA level, the annual start-up establishment rate during 1990-2005 (as % of traded
establishments in the EA) is 2.60% on average (with a standard deviation of 0.79). The
top-EAs by the rate of start-up establishments include Las Vegas-Paradise-Pahrump (NV)
and Austin-Round Rock (TX), and the bottom-two EAs are Grand Forks (ND-MN) and
Mason City (IA) (See Table A3).
While we primarily focus on start-up establishments, for robustness we also look
at total new establishments (including both start-up firms and new establishments of
existing firms). We expect the strength of the cluster environment in a location will
foster the entry of both types of new establishments. To test this, we compute two related
entry indicators. Entry employment is defined as the level of employment in all new
establishments within a given EA-industry; and entry establishment is the count of these
new establishments. Δ Entry Employment is then defined as ln(Entry Employmenti,r,02-05 /
Entry Employmenti,r,91-94 ), and Δ Entry Establishments is ln(Entry Establishmenti,r,02-05 /
Entry Establishmenti,r,91-94).
Finally, to test for the role of clusters in the performance of start-up firms, we
examine the level of employment in young (up to five years old) start-ups in regional
industries. Specifically, Employment in Start-up Survivorsi,r,2004-05 is defined as the
average annual level of employment (over 2004-2005) in start-up firms borne during
2001-2003.
Industry Specialization. Our main empirical task is to examine the impact of the
industry and different aspects of the cluster environment on the growth rate in start-up
activity. As such, we require measures of industry and cluster specialization, as well as
the strength of related and neighboring clusters. We draw on a body of prior work which
uses location quotients (LQ) as a primary measure of regional specialization (Glaeser, et

17
al. 1992, Feldman and Audretsch 1999, Porter 2003, among others). Specifically, the
employment-based industry specialization in the base year (1990) is measured by the
share of regional employment in the industry as compared to the share of US total
employi,r employ r
employment in the national industry: INDUSTRY SPECEmploy,i,r,90 = ,
employi,US employ US

where r and i indicate the region (EA) and the industry, respectively. This indicator
captures to what extent the industry is “over-represented” (in terms of employment) in
the EA. Note that the specialization indicators are based on employment in the start-up
employment models, and based on establishments in the start-up establishment models.
In the data, the employment-based industry specialization of regions has a mean of 2.01
and a standard deviation of 6.42 (the establishment-based industry specialization has a
mean of 1.78 and standard deviation of 3.20; Table 1). As mentioned earlier, we include
region and industry fixed effects, and so the independent variation in our main empirical
specifications is driven exclusively by variation in employment in the region-industry.
Cluster Specialization. We utilize an analogous procedure to develop a measure
for cluster specialization. For a particular EA-industry the specialization of the EA in
cluster c is measured by the share of regional employment in the cluster (outside the
industry) as compared to the share of US total employment in the national cluster (outside
employoutside i
employr
the industry): CLUSTER SPECEmploy,icr,90 = c,r
. The average
employoutside
c,US
i
employ US

(employment-based) cluster specialization is 1.21 (and the standard deviation is 2; Table

1). Since this measure of specialization is relative to the overall size of the region, a
region may exhibit specialization within a particular cluster even though that region only
maintains a small share of the overall national employment of that cluster. While it is not
surprising that leading regions in the automotive cluster include Detroit-Warren-Flint
(MI) and Cleveland-Akron-Elyria (OH), there are also pockets of automotive cluster
strength in smaller regions, such as Lexington-Fayette-Frankfort-Richmond (KY) and
Louisville-Elizabethtown-Scottsburg (KY-IN) (Figure A1). It is useful to note that, with
the inclusion of region and industry fixed effects and a measure of industry
specialization, the independent variation that is utilized in the regression comes
exclusively from the employment within a given cluster (outside the industry).

18
Table 2 illustrates key attributes of the top regional clusters based on cluster
specialization in 1990.19 These top clusters tend to have a higher level of start-up
activity, a larger average size of establishments as well as a greater mix of older
incumbents than other regional clusters. Interestengly, on average over 20% of all the
establishments in a regional cluster belong to firms that have establishments in more than
one geographical market (EA), and this figure increases to over 30% in the top regional
clusters. This fact suggests that clusters are not isolated geographical units and they may
establish linkages with other locations, in part facilitated by the presence of these
regionally diversified firms.20
Strength of Linked and Neighboring Clusters. We additionally develop measures
of the strength of “linked” clusters and the presence of clusters in neighboring regions.
The measure of linked clusters is developed using the set of “broad” cluster definitions
defined in Porter (2003). Specifically, while the narrow cluster definitions used for the
earlier measure come from a classification scheme in which each industry is assigned to a
unique cluster, Porter (2003) also develops a broad cluster definition in which each
industry may be associated (measured by locational correlation of employment) with
multiple clusters. To develop a measure based on linkages to cluster c, we include those
broad clusters that have at least 1 of cluster c’s narrow industries in common. For
example, in the case of automotive, the linked clusters include production technology,
metal manufacturing, and heavy machinery; among others (see Table A2).21 Having
identified the set of clusters linked to a cluster (C*), we then measure the degree of
overlap between each pair of clusters (c, j) using the average proportion of narrow
industries that are shared in each direction:
19
We define the set of top EA-clusters by selecting the top-10 EAs by Custer Specialization for every cluster.
The high cluster specialization criterion is complemented with a minimum threshold for the share of
national cluster employment (above values that correspond to the 20th percentile) to limit the concern about
small regions with very low employment and very large location quotients.
20
While systematic quantitative analysis is still missing, the view is that in response to global competition
clusters are becoming more specialized on certain activities within the value chain and increasing their
participation in national and global value chains (see e.g., Porter, 1998b, Bresnahan and Gambardella,
2004; Ketels and Memedovic, 2008). For example, multi-plants and multinational firms can use their
network of subsidiaries to coordinate across clusters (Dunning 1998; Rugman and Verbeke, 2003;
Manning, 2008).
21
Clusters with linkages with many other traded clusters include analytical instruments and
communications equipment, among others; while clusters with few connections to other clusters include
tobacco and footwear.

19
 ⎛ shared industriesc,j shared industries j,c ⎞ 22
ωc,j =Avg ⎜ , ⎟. The strength of a region in clusters
⎜ total industries total industries j ⎟⎠
⎝ c

linked to cluster c is then defined by a weighted sum of the location quotients associated
with each linked cluster:
C*

∑ (ω c, j * em ploy j, r )
em ploy r
j∈ C *c
LIN K E D C LU S T E R S S P E C E m ploy c,r = C*
/ .
em ploy U S
∑ (ω c, j * em ploy j, U S )
j∈ C *c

For instance, based on this weighting which emphasizes the degree of overlap between
clusters, our measure of the strength of linked clusters for industries within the
automotive cluster will weigh the presence of the metal manufacturing cluster more
heavily than the presence of the furniture cluster (Table A2).
We also develop a measure of the presence of like clusters in neighboring regions.
In part, we include this measure based on the simple empirical observation that
specialization in a particular cluster tends to be spatially correlated across neighboring
regions – the historical strength of the automotive cluster near Detroit is likely reinforced
by cluster specialization in automotives in neighboring EAs such as Grand Rapids-
Muskegon-Holland (MI), Toledo-Fremont (OH) and Fort Wayne-Huntington-Auburn
(IN). To explore the role of neighboring clusters in start-up growth in a region-industry,
we compute the (average) specialization of adjacent regions in the cluster (including the
focal industry). In other words, the strength of neighboring clusters is measured by the
average LQ of the adjacent regional clusters.
Finally, as mentioned earlier, we include a complete set of region and industry
fixed effects in our main specifications, and so control for unobserved factors that may be
correlated with these measures of industry and cluster specialization, including regional
or industry-level demand shocks, regional policies, or national regulatory changes that
affect all firms within certain industries.

22
For example, automotive has 5 narrow industries (out of 15) in common with production technology, and
production technology shares 7 narrow industries (out of 23) with automotives; the degree of overlap
between these two clusters is then ωc , j = .32 .

20
5. Results
We now turn to our key findings. The sample consists of a cross-section of
region-industries in 177 EAs and 588 four-digit SIC code industries, grouped into 41
traded clusters. After conditioning on those EA-industries where we observe a positive
level of employment in the base period, our core sample consists of 53,213 observations.
Our analysis begins in Table 3 where we compute the average start-up growth
rates for region-industries based on their initial levels of start-up activity and cluster
specialization. Specifically, we divide all region-industries into four categories based on
whether they are above or below the median level of start-up employment (for their
industry), and above or below the median level of cluster specialization (for their
industry). There are striking differences in the start-up employment growth rate across
these conditions. Whereas region-industries with a relatively high level of start-up
employment and a low level of cluster specialization experience, on average, a 33%
decline in the rate of start-up activity between 1991-1994 and 2002-2005, those region-
industries with low levels of start-up employment and a high level of cluster
specialization register a 36% growth rate on average during this period. There is a
statistically and quantitatively significant increase in the average start-up employment
growth rate when moving from high initial level of start-up employment to a low initial
level of start-up employment, consistent with a mean-reversion process in start-up
activity. However, regardless of the initial level of start-up employment, there is a
statistically and quantitatively significant increase in the growth rate of start-up activity
when one moves from a region-industry with a low level of cluster specialization to one
with high level of cluster specialization.23 In other words, those regional industries that
are located in a relatively strong cluster experience much higher growth rates in
entrepreneurial activity.
While the sharp contrasts in Table 3 are intriguing, it is of course possible that
alternative factors, such as industry specialization, or industry and region effects, are
driving these striking results. We therefore turn in Table 4 to a more systematic
regression analysis. The dependent variable is the start-up growth rate between the

23
We find the same conclusions when looking at start-up establishment growth; and using the no-zeros
subsample.

21
baseline period (1991-1994) and a terminal period (2002-2005). In (4-1), we include
only the level of start-up employment, the level of industry specialization and the level of
cluster specialization. The results provide evidence for the two main findings of this
paper. First, there is a large convergence effect – a doubling of the initial level of start-up
employment is associated with a 31% decline in the expected growth rate of start-up
activity. At the same time, the presence of complementary economic activity in the form
of clusters also has an important influence on the growth rate of entrepreneurship. Both
industry and cluster specialization are associated with higher growth rates of start-up
activity.
These results are reinforced in (4-2), where we incorporate both the strength of
linked clusters and the strength of the cluster in neighboring regions, and control for the
total employment in the region. Both cluster specialization and the presence of linked
clusters have a positive influence on the start-up growth rate, while strength of clusters in
neighboring regions is actually associated with a lower growth rate of start-up activity.
This latter finding is consistent with the hypothesis that, while a strong local cluster
environment enhances incentives for entrepreneurship, neighboring clusters may also
attract entrepreneurs and so provide a substitute for growth within a particular EA.
Interestingly, the employment size of the region contributes to the growth of start-up
activity of its regional industries, countervailing the convergence forces that take place at
the region-industry level.
In (4-3) and (4-4), we implement the core specifications, in which we include
region and industry fixed effects, thus controlling for unobserved shocks such as an
increase in demand or a change in the policy environment towards a particular industry.
The main results concerning mean reversion and the impact of cluster specialization are
robust. The only meaningful change in the estimates concerns the impact of neighboring
clusters; not surprisingly, given that the sign on the coefficient on Cluster Spec in
Neighbors is ambiguous from the perspective of theory, the estimated coefficient depends
on whether we control for industry and region heterogeneity.
Finally, in (4-5), we conduct a robustness check in which we drop all region-
industries in which we observe zero start-up employment during either the base period or
the terminal period (this allows us to avoid the scaling adjustments to the dependent

22
variable and the convergence effect measure that we discussed in the Data Section). The
main results are not only robust to a subsample that focuses on those region-industries
with a positive level of start-up activity in both the base and terminal period, but the
estimated coefficients on each of the parameters associated with the cluster environment
increase in a meaningful way.
To illustrate the size of the effects, consider a one-standard deviation shift in each
of the measures of industrial and cluster specialization using the coefficient estimates
(and sample) from (4-5).25 A one standard-deviation increase in industry specialization
(4.26) is associated with a 11% increase in the annual start-up employment growth rate,
while a one standard-deviation shift in cluster specialization (1.91) is associated with a
1.6% increase in the expected annual start-up employment growth rate.26 In other words,
after controlling for the impact of convergence, there is a quantitatively important impact
of related economic activity on start-up activity.
The core findings persist when we shift attention towards measures of
entrepreneurship and specialization based on the number of establishments (rather than
the total employment within those new establishments). In Table 5, both the dependent
and independent variables are now based on counts of establishments, and the structure of
the specifications mirrors the logic of Table 4. We begin in (5-1) by including the level
of start-up establishment activity, industry specialization and cluster specialization; we
add the linked cluster and neighboring cluster measures and control for the total
establishments in the region in (5-2); include region and industry fixed effects in (5-3)
and (5-4); and condition on a sample that excludes all region-industries with zero start-up
establishments in either the baseline or terminal period in (5-5). Our core results
concerning the convergence effect and the impact of cluster specialization are robust.
Interestingly, the only significant difference in the results concerns the estimated effect of
industry specialization. Whereas the coefficient on industry specialization was positive
in the start-up employment models (Table 4), the coefficient on industry specialization in

25
We focus our analysis of the magnitudes on this latter specification since this subsample is not subject to
the re-scaling of the dependent variable that we implement in (4-1)-to-(4-4).
26
Alternatively, an increase in the level of industry specialization (cluster specialization) from the 25th to
the 75th percentile value is associated with a 3.4% (0.7%) increase in the annual start-up employment
growth rate.

23
the first two columns of Table 5 is negative (and significant).27 While the coefficient
becomes positive and significant when we include region and industry (or only industry)
fixed effects, the heterogeneity of this parameter across specifications is consistent with
the fact that the effect of industry specialization on entrepreneurial growth is ambiguous.
In contrast, the coefficient on the impact of clusters (both using the narrow cluster
definition as well as the impact of “linked” clusters) is positive across all specifications.
Finally, in Table 6, we consider an alternative measure of “new” economic
activity by examining the growth in employment in “new” establishments (even if the
firm may already exist in other locations) and counts of new establishments (including
new establishments by already existing firms). While this measure combines “pure”
entrepreneurship with more traditional types of business expansion, the opening up of
new establishments (and employing a significant number of workers in those
establishments) is a crucial channel by which entrepreneurial firms grow over time and
contribute to aggregate economic performance. Each of the specifications in Table 6
includes region and industry fixed effects, and only vary in the number of measures of
related economic activity that are included (i.e., whether the linked cluster and
neighboring cluster variables are included) and whether the dependent and explanatory
variables are measured based on employment (models 6-1 and 6-2) or based on counts of
new establishments (models 6-3 and 6-4). The results are robust across all of the
specifications – there is a mean-reversion effect in the data consistent with convergence,
and a positive impact of cluster strength and scale on the growth rate of new
establishments and of the employment in new establishments. The core findings hold
when we examine only new establishments of existing firms (i.e., “new subsidiaries”.
Interestingly, these new establishments often belong to firms that operate in clusters in
other locations (EAs) -- for example in 1990, more than 85% of these new subsidiaries
belong to multi-location firms. While more research is needed on the locational and
organizational decisions of multi-establishment firms, the findings suggests that firms
may be opening establishments in new locations to seek complementary clusters and
benefit from the competitive resources of each location.

27
The coefficient becomes positive if we include the (log of) the number of establishments in the EA-
industry; suggesting that the scale of the EA-industry matters for start-up establishment growth.

24
 In Table 7 we examine the level of entrepreneurial activity using all potential
region-industry pairs (i.e., 588 industries by 177 EAs) to further account for the large
number of EA-industries with zero start-up activity. First, we implement a probit
specification that evaluates whether the existence of start-up activity during 2002-2005 is
related to the industry and the overall cluster environment while controlling for the
existence of any start-up experience in an earlier period (Positive Start-up Activity during
1990-1996) and including region and industry fixed effects (7-1).29 We report the
marginal effects to facilitate the interpretation of the coefficients, and find that
agglomeration forces that take place in the industry and in the set of related industries
within the regional cluster and in neighboring clusters increase the probability of
experiencing start-up activity in the future.30
In model 7-2 we use the same explanatory variables to examine the total count of
start-up establishments during 2002-2005 using a fixed effects Negative Binomial model.
The estimated incidence-rate ratios suggest that the strength of the cluster has more than a
33% boost on the subsequent count of start-up establishments; the local industry has also
a large effect (29% boost), and the linked clusters and neighboring clusters matter but to a
lower extent.31 Finally, these findings are confirmed when we implement a panel
specification to examine the annual count of establishments in 1997 and 2002 Census
years using a Negative Binomial model with year, region and industry fixed effects (7-3
and 7-4).32 Drawing on the dynamic count data model developed by Blundell et al.,
(1995), we include two alternative indicators of the pre-existing start-up activity in a
region-industry, the existence of any start-up activity from 1990 to 1996 (7-3) and the
average annual start-up activity during that period (7-4) to control for the unobservable
heterogeneity of region-industries. Overall our findings suggest that the cluster
environment facilitates the formation of new businesses in regional traded industries.

29
In order to include the EA-industries with zero employment (i.e., zero industry specialization), we
replace industry specialization with the minimum positive value of this variable.
30
We also define the industry and cluster environment based on employment (versus count of
establishments) and the same findings hold. Furthermore, the findings are robust to dropping the
observations with zero employment in the base year (i.e., using our core sample of 53213 observations).
31
We also examine the count of start-up employment and find robust evidence for the contribution of the
industry and the cluster, while the effect of linked clusters and neighboring clusters gets noisier.
32
As mentioned earlier, Census-years are preferred because they offer a better coverage of all new firms.

25
While the focus of this paper is the formation of new businesses, we also examine the
role of clusters on the (medium-term) performance of start-up firms. In particular, in
Table 8 we study the level of employment in young start-up survivors in a region-
industry using our core econometric specification. We find that the cluster environment
contributes to improve the level of employment in young start-up survivors, suggesting
that clusters facilitate the survival and potentially the growth of successful start-ups. In
related work, we will examine more carefully the role of clusters in the dynamics of new
businesses and the attributes of successful start-ups.
Finally, it is useful to emphasize that our core findings on start-up activity growth
(Tables 4-5) are robust to a variety of sensitivity checks. We have used a less nuanced
measure of complementary activity, in which the elements of each “cluster” are defined
as all traded four-digit SIC code industries in each two-digit SIC code.33 We have
included in the model a dummy indicator equal to one for region-industries with positive
start-up activity in the base period to capture unobserved factors that may influence the
entrepreneurial opportunities in a local industry. Regarding the sample, we have varied
both the length and precise dates of both the base and terminal periods to test for
sensitivity to other periods. Specifically, we consider the 1990-2001 period and define
the base and terminal periods based on 5-year averages (versus 4-year averages). By
using this shorter period, we also reduce some of the noise of translating the older SIC
system into the new 2002 NAICS. Furthermore, we have considered a larger (but noisier)
sample by including those establishments with very noisy wages (too low or too high).34
Finally, we have examined whether the impact of clusters is particularly salient in certain
types of regions (e.g., large versus small) and clusters (manufacturing-oriented versus
service-oriented).35

33
In other words, we estimate models (4-3) and (5-3) but using an indicator of the specialization of the
region in other two-digit SIC traded industries (instead of the cluster specialization variable), and find that
the presence in the region of other (two-digit SIC) related industries contributes to start-up activity growth.
34
The new sample consists of 54,001 EA-industries with positive employment in the base period.
35
Interestingly, we find that in larger regions the convergence effect is smaller and the cluster effects are
larger. We also explore how the convergence and cluster effects vary across different types of clusters
(manufacturing-oriented vs. service-oriented) and find that service-oriented clusters have the lowest
convergence and largest cluster-driven agglomeration benefits in entrepreneurship growth.

26
6. Conclusion and Extensions
This paper finds striking evidence for the simultaneous yet distinct influences of
agglomeration and convergence on the growth of the number of new firms and
employment by new firms in regional (traded) industries. The growth in start-up
activities at the region-industry level is declining in the initial level of start-up activity at
the region-industry level due to convergence forces. After controlling for convergence,
however there is strong evidence that the presence of complementary economic activity --
most notably, the presence of strong local clusters -- accelerates the growth in start-up
activities. We find that industries located within a strong cluster or that can access strong
linked clusters are associated with higher start-up employment growth rates and higher
entry of new firms.
These findings offer an important and novel contribution to the ongoing debate
about the impact of related economic activity on entrepreneurship and economic
performance. Most notably, building on the cluster framework developed by Porter
(1990, 1998), this paper moves beyond the traditional debate in which the presence of
related economic activity simultaneously indicated the presence of complementarities as
well as competition for inputs and customers, clouding the interpretation of any particular
empirical finding. Instead, by first accounting for convergence and the potential for
competition within each industry in a region, we are able to isolate the separate and
quantitatively important impact of cluster-related complementarities on entrepreneurship.
In other words, while at a (narrow) industry level firms compete for a given pool of
resources, the cluster environment that surrounds an industry will increase the pool of
competitive resources and reduce the barriers of entry for new firms. Strong regional
clusters enhance the range and diversity of entrepreneurial start-up opportunities while
also reducing the costs of starting a new business. Cluster-driven entrepreneurship is,
moreover, a dynamic process, as the new business creation at one point in time spurs
further start-up activity on an ongoing basis.
While our analysis has focused at new business formation, we also examine
young start-up survivors. We find that clusters contribute to the level of employment in
young start-ups in regional industries, suggesting that a strong cluster environment
enhances the potential for growth for the most productive start-ups. The impact of the

27
cluster environment on the dynamics of new businesses is an open research question that
we are further investigating.
Our findings support the idea that clusters of related and complementary
industries facilitate the growth in the formation of new businesses and the performance of
start-up survivors, even after controlling for region and industry heterogeneity. There is
large heterogeneity in the types of entrepreneurship that we could further explore. Start-
up firms will differ in size, innovative-orientation, and their growth potential. Similarly,
we could also exploit the attributes of the firms that participate in clusters (size, age,
regional diversification). Further understanding the demography of entrepreneurship and
clusters will help design more effective policies to promote entrepreneurship.
In the paper we focus exclusively on traded industries since they are exposed to
greater competition and account for most innovations. In future work we will examine
how the effect of cluster-driven agglomeration forces on entrepreneurship varies for local
industries versus traded industries.
Finally, we find that clusters also matter for the formation of new subsidiaries of
existing firms. These new subsidiaries often belong to firms that operate in clusters in
other locations (i.e., “multi-cluster” firms), suggesting that firms may seek
complementary clusters, benefiting from the competitive resources of each location.
Similarly, the International Business literature emphasizes that in a global economy, local
clusters integrate in national and global value chains often through the complex network
of subsidiaries of MNCs (Dunning 1998; Porter 1998b; Enright, 2000; Manning, 2008).
We conjecture based on our analysis of new establishments of existing firms that strong
clusters may also facilitate the entry of foreign subsidiaries to a location. A related
important questions raised by international business studies is what attributes of US
clusters are more attractive for the location of (domestic and foreign) multi-establishment
firms, and how the participation in multiple clusters affect the strategy, internal
organization and performance of these companies. Furthermore, the contribution of these
firms to generate entrepreneurial opportunities in a particular regional cluster is an open
research question that we examine in related work.

28
7. References

Acs, Z.J., and C. Armington, 2006, Entrepreneurship, Geography and American

Economic Growth, New York, NY: Cambridge University Press.
Acs, Z.J., D. Audretsch, P. Braunerhhjem, and B. Carlsson, 2009, “The Knowledge
Spillover Theory of Entrepreneurship,” Small Business Economics 32 (1), pp. 15-
30.
Acs, Z.J., E. Glaeser, R. Litan, L. Fleming, S. Goetz, W. Kerr, S. Klepper, S. Rosenthal,
Ol. Sorenson, and W. Strange, 2008, “Entrepreneurship and Urban Success:
Toward a Policy Consensus,” Kaufman Foundation Report.
Agrawal, A and I. Cockburn, 2003, “The Anchor Tenant Hypothesis: Exploring the Role
of Large, Local, R&D-intensive Firms in Regional Innovation Systems,”
International Journal of Industrial Organization 21, pp. 1227-1253.
Alcacer, J., 2006, “Location Choices across the Value Chain: How Activity and
Capability Influence Collocation,” Management Science 52 (10), pp. 1457-1471.
Armington, C., and Z.J. Acs, 2002, “The Determinants of Regional Variation in New
Firm Formation,” Regional Studies 36(1), pp. 33-45.
Audretsch, D.B., 1995, Innovation and Industry Evolution, MIT Press, Cambridge, MA.
Barro, R.J. and X. Sala-i-Martin, 1991, “Convergence across States and Regions,”
Brookings Papers on Economic Activity 1, pp. 107-182.
Barro, R.J. and X. Sala-i-Martin, 1992, “Convergence,” Journal of Political Economy
100(2), pp. 223-51.
Barro, R.J. and X. Sala-i-Martin, 1995, Economic Growth, Cambridge, MA: MIT Press.
Barteisman, E., S. Scarpetta, and F. Schivardi, 2005, “Comparative Analysis of Firm
Demographics and Survival: Evidence from Micro-Level Sources in OECD
Countries,” Industrial and Corporate Change 14(3), pp. 365-391.
Bloom, N., M. Schankerman and J. Van Reenen, 2005, “Identifying Technology
Spillovers and Product Market Rivalry,” CEP Discussion Paper, 675.
Blundell, R., R. Griffith, and J. Van Reenen, 1995, “Dynamic Count Data Models of
Technological Innovation,” The Economic Journal, 105 (No. 429), pp. 333-344.
Bonte, W, 2004, “Innovation and Employment Growth in Industrial Clusters: Evidence
from Aeronautical Firms in Germany,” International Journal of the Economics
and Business 11 (3), pp. 259-278.
Boschma, R., 2005, “Proximity and Innovation: A Critical Assessment,” Regional
Studies, 39 (1), pp. 61-74.
Bostic, R.W., J.S. Gans, and S. Stern, 1997, “Urban Productivity and Factor Growth in
the Late Nineteenth Century,” Journal of Urban Economics 41(1), pp 38-55.
Bresnahan, T.F., A. Gambardella (eds.), 2004, Building High-Tech Clusters. Silicon
Valley and Beyond. Cambridge University Press, New York.
Bresnahan, T.F. and P.C. Reiss, 1991, “Entry and Competition in Concentrated Markets,”
The Journal of Political Economy 99 (5), pp. 977-1009.

29
Bulow, J. L., J.D. Geanakoplos, and P.D. Klemperer, 1985, “Multimarket Oligopoly:
Strategic Substitutes and Complements,” Journal of Political Economy 91(31),
pp.488-511.
Chinitz, B., 1961, “Contrasts in Agglomeration: New York and Pittsburgh,” American
Economic Review 51(2), pp. 279-289.
Ciccone, A. and R. Hall, 1996, “Productivity and the Density of Economic Activity,”
American Economic Review 86(1), pp. 54-70.
Cingano F. and F. Schivardi, 2004, “Identifying the Sources of Local Productivity
Growth,” Journal of the European Economic Association 2(4), pp. 720–742
Cortright, J., 2006, “Making Sense of Clusters: Regional Competitiveness and Economic
Development,” The Brookings Institution Metropolitan Policy Program,
http://www.brookings.edu/reports/2006/03cities_cortright.aspx.
Davis, S., J. Haltiwanger and S. Schuh, 1996, Job Creation and Destruction, MIT Press.
Delgado M, M.E. Porter and S. Stern, 2007, “Clusters, Convergence and Economic
Performance,” manuscript.
Delgado, M., 2005, “Do Clusters Really Matter for Companies’ Competitive Strategies?
Evidence at the Country Level,” manuscript.
Dumais, G., G. Ellison, E.L. Glaeser, 2002, “Geographic Concentration as a Dynamic
Process,” Rev. Econ. and Stat. 84 (2), pp. 193-204.
Dunne, T, S.D. Klimek, M.J. Roberts, 2005, “Exit from regional manufacturing markets:
The role of entrant experience,” Int. J. Ind. Organ. 23, pp. 399-421.
Dunne, T., M. Roberts, and L. Samuelson, 1989, "Patterns of Firm Entry and Exit in U.S.
Manufacturing Industries," RAND Journal of Economics 19, pp. 495-515.
Dunning, J. H., 1998, “Location and the Multinational Enterprise: A Neglected Factor?”
Journal of International Business Studies, 29 (1), 45-66.
Duranton, G. and Puga, D., 2001, “Nursery Cities: Urban diversity, process innovation,
and the life cycle of products,” American Economic Review 91(5), pp. 1454-1477.
Ellison, G. and E. Glaeser, 1997, “Geographic Concentration in U.S. Manufacturing
Industries: A Dartboard Approach,” Journal of Political Economy 105, pp. 889-
927.
Ellison, G. and E. Glaeser, W. Kerr, 2007, “What Causes Industry Agglomeration?
Evidence from Coagglomeration Patterns,” NBER Working Paper, No. 13068.
Enright, M., 2000, “Regional clusters and multinational enterprises: independence,
dependence or interdependence?,” International Studies of Management and
Organization, 30 (2), pp. 114-138.
Feldman, M.P., 1994, “The University and High-Tech Start-ups: The Case of Johns
Hopkins University and Baltimore,” Economic Development Quarterly 8, pp. 67-
77.
Feldman, M.P., 2001, “The Entrepreneurial Event Revisited: An Examination of New
Firm Formation in the Regional Context,” Industrial and Corporate Change 10,
pp. 861-891.

30
Feldman, M.P. and D. Audretsch, 1999, “Innovation in Cities: Science-based Diversity,
Specialization and Localized Competition,” European Economic Review 43, pp.
409-429.
Feldman, M.P., J. Francis, and J. Bercovitz, 2005, “Creating a Cluster While Building a
Firm: Entrepreneurs and the Formation of Industrial Clusters.” Regional Studies
39 (1), pp. 129-141.
Feser, E.J. and E.M. Bergman, 2000, “National Industry Cluster Templates: A
Framework for Applied Regional Cluster Analysis,” Regional Studies 34(1), pp.
1-19.
Forni, M and S. Paba, 2002, “Spillovers and the Growth of Local Industries,” Journal of
Industrial Economics 50, pp. 151-171.
Frenken, K., F.G. van Oort & T. Verburg, 2007, “Related Variety, Unrelated Variety and
Regional Economic Growth,” Regional Studies 41, 685-697.
Feser, E., H. Renski, H. Goldstein, 2008, “Clusters and Economic Development
Outcomes,” Economic Development Quarterly 22 (4), pp. 324-344.
Fujita, M., P. Krugman, and A.J. Venables, 1999, The Spatial Economy, MIT Press,
Cambridge, Massachusetts.
Glaeser E.L., et al., 1992, “Growth in Cities,” Journal of Political Economy 100(6),
1126-52.
Glaeser, E.L., and W.R. Kerr, 2009, "Local Industrial Conditions and Entrepreneurship:
How Much of the Spatial Distribution Can We Explain?" Journal of Economics
and Management Strategy, 18(3).
Glaeser, E.L, W.R. Kerr, and G.A. M. Ponzetto, 2009, “Clusters of Entrepreneurship,”
Journal of Urban Economics (forthcoming).
Gompers, P., J. Lerner, and D. Scharfstein, 2005, “Entrepreneurial Spawning: Public
Corporations and the Formation of New Ventures, 1986-1999,” Journal of
Finance 60 (2), pp. 577-614.
Greenstein S., C. Forman and A. Goldfarb, 2008, “Understanding the Inputs into
Innovation: Do Cities Substitute for Internal Firm Resources?” Journal of
Economics and Management Strategy 17(2).
Greenstone M., R. Hornbeck, and E. Moretti, 2008, “Identifying Agglomeration
Spillovers: Evidence from Million Dollar Plants,” NBER Working Paper No.
13833.
Haltiwanger, J., R.S. Jarmin, and J. Miranda, 2009, “Who Creates Jobs? Small vs. Large
vs. Young,” Working Paper.
Henderson, J.V., 2003, “Marshall's Scale Economies,” Journal of Urban Economics 53,
pp. 1-28.
Henderson, J.V., A. Kuncoro and M. Turner, 1995, “Industrial Development in Cities,”
Journal of Political Economy 103, pp. 1067-1090.
Higgins, MJ., D. Levy, and AT Young, 2006, “Growth and Convergence across the
United States: Evidence from County‐Level Data,” Review of Economics and
Statistics 88(4), pp. 671‐681.

31
Jaffe, A., M. Trajtenberg, R. Henderson. 1993. Geographic Localization of Knowledge
Spillovers as Evidenced by Patent Citations. Quarterly J. Econ. 108 577-598.
Jarmin, R, and J Miranda, 2002, “The Longitudinal Business Database,” CES Working
Paper.
Johnson K.P. and J.R. Kort, 2004, “2004 Redefinition of the BEA Economic Areas,” at
http://www.bea.gov/scb/pdf/2004/11November/1104Econ-Areas.pdf.
Ketels, C. and O. Memedovic, “From Clusters to Cluster‐Based Economic
Development,” International Journal of Technological Learning, Innovation, and
Development 1, pp. 375-392.
Kerr, W.R. and R. Nanda, 2009, “Democratizing Entry: Banking Deregulations,
Financing Constraints, and Entrepreneurship, Journal of Financial Economics 94, pp.
124–149.Klepper, S., 2007, “The Geography of Organizational Knowledge,” manuscript.
Krugman, P., 1991, “Increasing Returns and Economic Geography,” Journal of Political
Economy 99(3), pp. 483–499.
Malecki, E.J., 1990, “New Firm Formation in the USA: Corporate Structure, Venture
Capital, and Local Environment,” Entrepreneurship and Regional Development 2,
pp. 247-265.
Manning, S., 2008, “Customizing Clusters: On the Role of Western Multinational
Corporations in the Formation of Science and Engineering Clusters in Emerging
Economies,” Economic Development Quarterly, 22(4), pp. 316-323.
Magrini, S., 2004, Regional (Di)Convergence, in J. V. Henderson and J. F. Thisse (eds.)
Handbook of Regional and Urban Economics, 4, Amsterdam: Elsevier
North‐Holland, pp. 2243‐2292.
Markusen, A., 1996, “Sticky Places in Slippery Space: A Typology of Industrial
Districts,” Economic Geography 72(3), pp. 293-313.
McCann, P. and Mudambi, R. 2005, “Analytical Differences in the Economics of
Geography: The Case of the Multinational Firm,” Environment and Planning A,
37(10), pp. 1857-1876.
Paci R. and S Usai, 2000, “The Role of Specialization and Diversity Externalities in the
Agglomeration of Innovative Activity,” Rivista Italiana degli Economisti 2(2), pp.
237-268.
Porter, M. E., 1980, Competitive Strategy: Techniques for Analyzing Industries and
Competitors, New York: Free Press.Porter, M.E., 1990, The Competitive Advantage of
Nations, Free Press, New York.
Porter, M.E., 1998a, Clusters and Competition: New Agendas for Companies,
Governments, and Institutions. In M.E. Porter (ed.). On Competition. Harvard
Business School Press, Boston, pp. 197-299.
Porter, M.E., 1998b, ‘Competing Across Locations: Enhancing Competitive Advantage
Through a Global Strategy, in Porter, M. (Ed.): On Competition, Harvard
Business School, Boston.
Porter, M.E, 2000, “Location, Competition, and Economic Development: Local Clusters
in a Global Economy.” Economic Development Quarterly 14 (1), pp. 15-34.

32
Porter, M.E., 2001, “Regions and the New Economics of Competition,” in A.J. Scott
(ed.), Global City-Regions. Trends, Theory, and Policy, New York: Oxford
University Press, pp. 145-151.
Porter, M.E., 2003, “The Economic Performance of Regions,” Regional Studies 37, pp.
549-578.
Porter, M.E. and S. Stern, 2003, “The Impact of Location on Global Innovation: Findings
from the National Innovative Capacity Index,” in P. K. Cornelius (ed.), The
Global Competitiveness Report 2002-2003, New York: Oxford University Press,
pp. 227-253.
Porter, M.E., et al., 2004, “Competitiveness in Rural U.S. Regions: Learning and
Research Agenda,” Economic Development Administration, U.S. Department of
Commerce.
Rosenthal, S.S. and W.C. Strange, 2003, “Geography, Industrial Organization, and
Agglomeration,” Review of Economics and Statistics 85 (2), pp. 377-393.
Rosenthal, S.S. and W.C. Strange, 2004, “Evidence on the Nature and Sources of
Agglomeration Economies,” in J.V. Henderson and J.F. Thisse (eds.), Handbook
of Regional and Urban Economics, vol. 4, Amsterdam: Elsevier North-Holland.
Rosenthal, S.S., and W.C. Strange, 2009, “Small Establishments/Big Effects:
Agglomeration, Industrial Organization and Entrepreneurship, forthcoming in
Edward Glaeser (ed.), 2009, Economics of Agglomeration.
Rugman, A. M. and A. Verbeke, 2003, “Multinational Enterprises and Clusters: An
Organizing Framework,” Management International Review, 43(3), pp. 151–169.
Shane, S., 2001, “Technological Opportunities and New Firm Creation,” Management
Science 47, pp. 205–220.
Scherer, F. M., 1982, “Inter-industry Technology Flows and Productivity Growth,” The
Review of Economics and Statistics 64(4), pp. 627-634.
Sorenson, O., P.G. Audia. 2000. The Social Structure of Entrepreneurial Activity:
Geographic Concentration of Footwear Production in the United States, 1940-
1989. Amer. J. Sociology 106 (2) 424-62.
Swann, G.M. P., 1998, “Clusters in the US Computing Industry,” in Swann, G.M. P., M.
Prevezer and D. Stout (eds.), 1998, The Dynamics of Industrial Clustering:
International Comparisons in Computing and Biotechnology, Oxford: Oxford
University Press, pp. 76-105.
Torre, A., 2008, “On the Role Played by Temporary Geographical Proximity in
Knowledge Transmission,” Regional Studies, 42 (6), pp. 869-889.
Venables, A.J., 1996, “Equilibrium Locations of Vertically Linked Industries,”
International Economic Review 37(2), pp. 341-360.
Wennberg, K. and G. Lindqvist, 2008, “The Effect of Clusters on the Survival and
Performance of New Firms,” Small Business Economics.

33
 Table 1: Region-Industry Descriptive Statistics
EA-Industries with positive employment in 1990*
“No-zeros”
Positive start-up
activity in 1991-
94 & 2002-05
N=53213 N=11981
Variables Definition
ESTABLISHMENTS90 EA-industry establishments 16.16 56.00
(71.23) (142.01)
STARTUP ESTABLISHMENT91-94 EA-industry (4-year average) annual .47 1.87
startup establishments (91-to-94) (2.59) (5.21)
ENTRY ESTABLISHMENT91-94 EA-industry (4-year average) annual entry 0.82 2.54
establishments (91-to-94) (5.28) (9.39)
EMPLOYMENT90 EA-industry employment 554.67 1544.09
(2464.7) (4778.32)
STARTUP EMPLOYMENT91-94 EA-industry ( 4-year average) annual 6.37 20.99
startup employment (86.42) (179.05)

ENTRY EMPLOYMENT91-94 EA-industry (4-year average) annual entry 15.39 43.75

employment (146.70) (262.82)

EMPLOYMENT IN STARTUP EA-industry (2-year average) annual

SURVIVORS04-05 employment in start-up firms borne during 13.61 51.55
2001-2003 (120.10) (241.85)
∆ STARTUP ESTABLISHMENT Growth rate in startup establishments .10 .04
⎛ startup estabi,r,02−05 ⎞ (1.78) (.91)
log ⎜⎜ ⎟⎟
⎝ startup estabi,r,91−94 ⎠
∆ STARTUP EMPLOYMENT Growth rate in startup employment .16 .47
(1.02) (1.73)
∆ ENTRY ESTABLISHMENT Growth rate in entry establishments .07 .02
(1.86) (.94)
∆ ENTRY EMPLOYMENT Growth rate in entry employment .14 .36
(1.25) (1.80)
INDUSTRY SPECEstab, 90 Industry establishment-based Location 1.78 1.40
estabi,r estab r (3.20) (2.57)
Quotient. LQi,r =
estabi,US estab US
CLUSTER SPECEstab, 90 Cluster establishment-based LQ 1.06 1.13
(outside the industry) (1.14) (1.25)
LINKED CLUSTERS SPECEstab, 90 Linked clusters’ establishment-based LQ .97 .98
(weighted by cluster overlap) (.46) (.35)
CLUSTER SPEC in Neighboring clusters’ average 1.03 1.02
NEIGHBOREstab, 90 establishment-based LQ (.80) (.81)

INDUSTRY SPECEmploy, 90 Industry employment-based LQ 2.01 1.47

(6.42) (4.26)
CLUSTER SPECEmploy, 90 Cluster employment-based LQ 1.17 1.18
(outside the industry) (1.95) (1.91)
LINKED CLUSTERS SPECEmploy, 90 Linked clusters’ employment-based LQ 1.04 1.01
(weighted by cluster overlap) (.92) (.70)
CLUSTER SPEC in Neighboring clusters’ average 1.13 1.06
NEIGHBOREmploy, 90 employment-based LQ (1.19) (1.06)

Note: The core sample includes EA-industries with positive employment in 1990, resulting in 53213 observations.
For this sample, the start-up and entry indicators are scaled by adding 1 employee and 0.01 establishments
(before taking logs). The “No-zeros” subsample includes EA-industries with positive startup activity in both
the base and terminal periods (1991-1994 and 2002-2005). For the Entry variables (i.e., those based on all the
new establishments) the “No-zeros” sample consists of 16016 observations.

34
 Table 2: Attributes of EA-Clusters in 1990 (Mean and Std. Deviation)
Start-up Multi-EA Avg. Size Young Old
Estab91-94 Estab Estab Estab Estab
(Annual, (rate) Age<5 Age>10
level) (rate) (rate)
Top-10 8.8 .313 247.761 .249 .520
EA-Clusters1990 (29.16) (.248) (394.20) (.162) (.210)
(by Cluster)
N=410
Other 3.11 .193 55.661 .304 .426
EA-Clusters1990 (12.97) (.217) (93.301) (.214) (.240)
N=6220
t-test of 5.95 10.684 28.273 -5.095 7.717
diff. of means
Note: We define the set of top EA-clusters by selecting the top-10 EAs by Custer Specialization for every
cluster (E.g., for automotive the top one local cluster would be in Detroit EA). The high cluster specialization
criterion is complemented with a minimum threshold for the share of national cluster employment (above the
20th percentile values) to limit the concern about small EAs with very low employment and very high cluster
specialization. Multi-EA Establishments measures the presence in a local cluster of firms that are active in
more than one geographical market (EA) for a given cluster (e.g., Microsoft has establishments in the IT
cluster in many EAs). Specifically, this variable is the rate of establishments of multi-EA firms in a given
EA-cluster. Avg. Size Establishment is the average size of the establishments in a given EA-cluster. Young
(Old) Establishments is the rate of establishments with Age<5 (Age>10) in a given EA-cluster.

Table 3: EA-industry Average Growth Rate in Start-up Employment

(by Level of Start-up Employment and Cluster Specialization, N=53213)

STARTUP EMPLOYMENT1991-94
Low High

Low ∆STARTUP ∆STARTUP

CLUSTER SPECEmploy, 90 EMPLOYMENT= .25 EMPLOYMENT= -.33
(Outside the industry)
N= 20,507 N=6,265

High ∆STARTUP ∆STARTUP

EMPLOYMENT= .36 EMPLOYMENT= -.09

N= 17,474 N=8,967
Notes: Low versus High is based on the median of the variable for each industry.
All the averages are significantly different from each other at 1%.

35
 Table 4: EA-Industry Growth in Start-up Employment (N=53213)
STARTUP EMPLOYMENT GROWTH
No zeros
N=11,981
1 2 3 4 5
Ln STARTUP EMPLOY91-94 -.299 -.358 -.680 -.684 -.827
(.012) (.011) (.009) (.009) (.011)
Ln INDUSTRY SPECEmploy, 90 .030 .045 .112 .107 .283
(.003) (.003) (.003) (.003) (.014)
Ln CLUSTER SPECEmploy, 90 .031 .017 .025 .013 .093
(Outside the industry) (.002) (.003) (.003 ) (.003) (.019)
Ln LINKED CLUSTERS SPECEmploy, 90 .013 .061 .110
(.007) (.007) (.030)
Ln CLUSTER SPEC in -.018 .031 .076
NEIGHBORSEmploy, 90 (.007) (.007) (.027)
Ln REGIONAL EMPLOYMENT .157
(.005)
EA FEs No No Yes Yes Yes
INDUSTRY FEs No No Yes Yes Yes
R-Squared .084 .115 .267 .269 .400
Notes: Bold, Bold-Italic numbers refer to coefficients significant at 1% and 5% levels. Robust standard
errors clustered by EA-Cluster. The explanatory variables are in logs.

Table 5: EA-Industry Growth in Start-up Establishments (N=53213)

STARTUP ESTABLISHMENT GROWTH
No zeros
N=11,981
1 2 3 4 5
Ln STARTUP ESTABLISHMENTS91-94 -.375 -.406 -.863 -.865 -.654
(.007) (.006) (.005) (.005) (.012)
Ln INDUSTRY SPECEstab, 90 -.107 -.066 .570 .557 .295
(.008) (.009) (.009) (.010) (.012)
Ln CLUSTER SPECEstab, 90 .086 .047 .026 .007 .067
(Outside the industry) (.006) (.007) (.006) (.006) (.017)
Ln LINKED CLUSTERS SPECEstab, 90 .092 .152 .100
(.020) (.018) (.032)
Ln CLUSTER SPEC in NEIGHBORSEstab, 90 .018 .050 .010
(.018) (.014) (.025)
Ln REGIONAL ESTABLISHMENTS .213
(.010)
EA FEs No No Yes Yes Yes
INDUSTRY FEs No No Yes Yes Yes
R-Squared .176 .193 .440 .442 .311
Notes: Bold, Bold-Italic numbers refer to coefficients significant at 1% and 5% levels. Robust standard
errors clustered by EA-Cluster. The explanatory variables are in logs.

36
 Table 6: EA-Industry Growth in Entry (All new establishments, N=53213)
ENTRY EMPLOYMENT ENTRY ESTABLISHMENT
GROWTH GROWTH
1 2 3 4
Ln ENTRY91-94 -.708 -.712 -.872 -.875
( .007) (.007) (.005) (.005)
Ln INDUSTRY SPEC90 .145 .139 .621 .604
(.003) (.003) (.010) (.010)
Ln CLUSTER SPEC90 .032 .019 .035 .011
(Outside the industry) (.003) (.004) (.006) (.007)
Ln LINKED CLUSTERS SPEC90 .083 .185
(.009) (.019)
Ln CLUSTER SPEC in NEIGHBORS90 .031 .068
(.007) (.015)
EA FEs Yes Yes Yes Yes
INDUSTRY FEs Yes Yes Yes Yes
R-Squared .325 .328 .454 .455
Notes: Bold and Bold-Italic numbers refer to coefficients significant at 1% and 5% levels. Robust standard errors
clustered by EA-Cluster. The explanatory variables are based on employment (establishments) in the entry employment
(establishment) growth models.

Table 7: EA-Industry Level of Start-up Activity (Using the Full Sample)

Probit Negative Binomial
(Marginal Effects) (Incidence-Rate Ratios)
POSITIVE START-UP ANNUAL
START-UP ESTABLISHMENTS START-UP
ACTIVITY (during 2002-05) ESTABLISHMENTS
(during 2002-05) (1997, 2002)
1 2 3 4
POSITIVE START-UP .054 1.349 1.289
ACTIVITY1990-96 (.003) (.016) (.019)
AVG ANNUAL START-UP 1.007
ESTABLISHMENTS1990-96 (.000)
Ln INDUSTRY SPEC Est 90 .011 1.286 1.360 1.362
(.000) (.005) (.006) (.006)
Ln CLUSTER SPEC Est 90 .008 1.338 1.385 1.367
(Outside the industry) (.002) (.011) (.012) (.012)
Ln LINKED CLUSTERS SPEC Est 90 .004 1.117 1.097 1.100
(.003) (.017) (.019) (.019)
Ln CLUSTER SPEC in .025 1.107 1.044 1.046
NEIGHBORSEst 90 (.003) (.014) (.015) (.015)
EA FEs Yes Yes Yes Yes
INDUSTRY FEs Yes Yes Yes Yes
YEAR FEs Yes Yes
R-Squared .450
Log-likelihood -29470.9 -58403.9 -63412.1 -63307.1
Obs. 103368 103014 206028 206028
Note: In (6-1) the dependent variable is a dummy equal to one for EA-industries with positive start-up activity during
2002-2005; the coefficients are marginal effects from the probit model. In (6-2) we examine the count of start-up
establishments during 2002-2005 using a Negative Binomial model, and the coefficients are the incidence-rate ratios.
In (6-3) and (6-4) we examine the annual count of start-up establishments in two Census years (1997, 2002) using a
Negative Binomial model. All models omit data for which all observations corresponding to a given fixed effect have
zero observed outcome. Note that to avoid convergence problems, in the Negative Binomial models we combine the
10% smallest EAs by employment size in a single region fixed effect, including in the models a total of 160 EA
dummies (vs. 177). Similarly, we include 529 industry fixed effects in the model (vs. 588) by combining the 10%
smallest national industries. Alternatively, we drop the 10% smallest EAs and industries and the same findings hold.

37
 Table 8: EA-Industry Level of Employment in Start-up Survivors (N=53213)
(Ln) EMPLOYMENT
in START-UP SURVIVORS2004-2005
No-zeros
11,981
1 2 3
Ln STARTUP EMPLOY91-94 .338 .334 .172
(.011) (.011) (.011)
Ln INDUSTRY SPECEmploy, 90 .100 .097 .227
(.003) (.003) (.014)
Ln CLUSTER SPECEmploy, 90 .020 .012 .075
(Outside the industry) (.003) (.003) (.019)
Ln LINKED CLUSTERS SPECEmploy, 90 .053 .095
(.008) (.028)
Ln CLUSTER SPEC in NEIGHBORSEmploy, 90 .016 .039
(.007) (.026)
EA FEs Yes Yes Yes
INDUSTRY FEs Yes Yes Yes
R-Squared .283 .284 .375
Notes: Bold, Bold-Italic, and Italic numbers refer to coefficients significant at 1%, 5% and 10% levels. Robust
standard errors clustered by EA-Cluster. The dependent variable is the (log of) average annual employment (over
2004-2005) of start-up establishments borne in 2001-2003. We add 1 to the dependent variable before taking logs. The
explanatory variables are in logs

38
 Appendix A: Traded Clusters Attributes

Table A1: Entry by Traded Clusters (Annual averages over 1990-2005)

Name (41 traded clusters) Type # Narrow Startup New Startup New Startup
Industries Estab subsidiary Employ subsidiary Estab
Estab (1,000) Employ (%
(1,000) Estab)
total svc.
Aerospace Engines High-tech 2 0 7.6 5.1 0.229 1.820 1.4
Aerospace Vehicles & Defense High-tech 6 0 25.2 10.6 0.861 5.449 2.0
Analytical Instruments High-tech 10 0 164.2 58.3 5.404 7.041 1.7
Biopharmaceuticals High-tech 4 0 44.0 25.0 2.476 2.220 2.1
Chemical Products High-tech 21 0 90.4 95.7 2.070 3.915 1.2
Communications Equipment High-tech 9 0 60.1 40.8 2.781 4.282 1.8
Information Technology High-tech 9 3 410.0 403.5 7.431 14.127 3.2
Medical Devices High-tech 8 0 89.9 39.2 2.005 3.591 1.8
Distribution Services Service 19 19 2452.6 1741.2 19.584 40.042 2.5
Education & Knowledge Creation Service 10 9 1042.6 490.6 16.399 16.107 2.9
Business Services Service 21 21 10539.3 4457.4 90.254 116.404 4.4
Entertainment Service 13 9 1597.9 207.9 23.572 12.545 4.0
Financial Services Service 21 21 2241.0 8692.7 35.963 137.821 1.9
Heavy Construction Services Service 19 6 2811.4 521.4 22.056 16.493 2.5
Hospitality & Tourism Service 22 19 1805.8 924.0 29.968 38.447 2.6
Oil & Gas Products & Services Service 12 6 377.1 277.3 4.312 8.032 2.4
Power Generation & Transmission Service 6 1 16.6 98.5 0.632 3.472 0.6
Transportation & Logistics Service 17 16 1186.2 1544.3 14.752 38.298 2.3
Agricultural Products Other 20 6 527.6 53.4 3.504 2.055 3.1
Apparel Other 27 0 454.2 46.5 8.727 5.323 3.9
Automotive Other 15 0 304.8 87.9 8.836 10.849 1.8
Building Fixtures, Equipment & Other 25 2 578.0 68.5 5.855 3.188 2.6
Services
Construction Materials Other 11 0 125.3 33.5 1.733 1.762 2.3
Fishing & Fishing Products Other 3 0 57.0 8.2 0.496 0.732 3.0
Footwear Other 5 0 15.6 3.7 0.460 0.380 2.6
Forest Products Other 8 0 114.3 38.7 2.529 3.477 2.4
Furniture Other 10 0 165.5 28.9 2.394 2.538 2.4
Heavy Machinery Other 10 2 188.0 106.3 2.793 3.439 1.4
Jewelry & Precious Metals Other 7 1 261.5 15.7 1.114 0.477 2.7
Leather & Related Products Other 13 0 114.3 12.7 1.342 0.765 2.7
Lighting & Electrical Equipment Other 10 0 66.7 24.7 2.009 1.939 1.4
Metal Manufacturing Other 44 0 386.6 122.8 10.683 7.935 1.6
Motor Driven Products Other 12 0 46.3 21.8 1.691 2.698 1.7
Plastics Other 9 0 202.3 112.9 4.831 7.834 1.6
Prefabricated Enclosures Other 12 0 55.0 20.8 1.670 1.946 2.0
Processed Food Other 43 2 325.6 237.7 8.956 11.205 1.3
Production Technology Other 23 0 192.5 67.2 4.038 4.315 1.6
Publishing & Printing Other 26 3 887.1 330.1 9.717 11.995 2.4
Sporting, Recreational & Other 3 0
Children's Goods 95.9 10.5 1.299 1.345 3.5
Textiles Other 20 0 120.2 27.9 3.773 3.472 2.5
Tobacco Other 4 0
Notes: Data sources: Longitudinal Business Database and US Cluster Mapping Project (Porter, 2003).
There are 589 traded four-digit SIC code industries (146 of them are service (svc.) industries).
Service clusters are those with more than 35% of employment in service industries.
High-tech clusters are manufacturing clusters with high R&D and patenting.
The last column reports startup establishments (as % of establishments in the cluster).
“New subsidiary” refers to new establishments of existing firms.

39
 Table A2: Automotive (narrow) Cluster: Breadth of Industries and Linked
Clusters
Clusters Linked to Automotive (Shared industries )

Manufacturing

Motor Driven
Technology
Production

Machinery

Aerospace
4-digit SIC Industries

Furniture
Products

Engines
Heavy
Metal
2396 Automotive and apparel trimmings
3052 Rubber and plastics hose and belting X
3061 Mechanical rubber goods X
3210 Flat glass
3230 Products of purchased glass
3322 Malleable iron foundries X
3465 Automotive stampings X
3519 Internal combustion engines, n.e.c. X X
3544 Special dies, tools, jigs and fixtures X X
3549 Metalworking machinery, n.e.c. X X
3592 Carburetors, pistons, rings, valves X
3711 Motor vehicles and car bodies X
3714 Motor vehicle parts and accessories X
3799 Transportation equipment, n.e.c.
3824 Fluid meters and counting devices X
Cluster Overlap ( ωc , j ) with linked clusters .32 .25 .08 .08 .03 .03
Porter's (2003) cluster definitions. The Automotive cluster has more than 30% overlap with the Production Technology
cluster (by average of the percent of narrow industries shared in each direction)

Figure A1: Location of strong regional Automotive clusters (1997) (Top 20% of EAs
by employment Location Quotient)
EAs with high cluster specialization
EAs with high cluster specialization and high share of US cluster employment (top 10% of EAs)
EAs with high share of US cluster employment but without high cluster specialization (these
regional clusters are not defined as strong clusters)

Source: Delgado, Porter and Stern (2007). Calculations based on US Cluster Mapping Project.

40
Table A3: Annual Start-up Activities in EAs (average 1990-2005): Top and Worse
Performers

Top-5 EAs by start-up employment Top-5 EAs by start-up establishments

(% traded employment) (% traded establishments)
Miami-Fort Lauderdale-Miami Beach, FL Las Vegas-Paradise-Pahrump, NV
Houston-Baytown-Huntsville, TX Austin-Round Rock, TX
San Antonio, TX Sarasota-Bradenton-Venice, FL
Pensacola-Ferry Pass-Brent, FL Bend-Prineville, OR
Los Angeles-Long Beach-Riverside, CA Miami-Fort Lauderdale-Miami Beach, FL

Bottom-5 EAs by start-up employment Bottom -5 EAs by start-up establishments

(% traded employment) (% traded establishments)
Mason City, IA Grand Forks, ND-MN
Waterloo-Cedar Falls, IA Mason City, IA
Kearney, NE Aberdeen, SD
Cedar Rapids, IA Waterloo-Cedar Falls, IA
Joplin, MO Scotts Bluff, NE
Source: Authors’ calculations based on LBD data. Start-up activities in the region are based on traded
industries.

41