ArticlePDF Available

Do Science Parks Generate Regional Economic Growth? An Empirical Analysis of their Effects on Job Growth and Venture Capital

  • Technology Policy Institute

Abstract and Figures

Agglomerations, or "clusters" of industries, and especially of high-technology industries, can be major sources of economic growth. Policy makers therefore often search for ways to catalyze such clusters. A popular approach is to establish a science or research park in the hopes that it will attract companies and fuel regional economic growth. In this paper I assemble a county-level panel dataset to explore the effects of science parks on job growth and on venture capital. Non-parametric and econometric analysis reveals no positive effect of science parks on regional development overall. In other words, while success stories do exist, the analysis suggests that successes are the exception rather than the rule. Thus, policies intended to promote cluster development by subsidizing scien+C30ce or research parks are unlikely to be effective.
Content may be subject to copyright.
Do Science Parks Generate Regional Economic Growth?
An Empirical Analysis of their Effects on Job Growth and
Venture Capital
Scott Wallsten
Working Paper 04-04
March 2004
Scott Wallsten is a fellow at the AEI-Brookings Joint Center and a resident scholar at the American
Enterprise Institute. I thank Tim Bresnahan, Alfonso Gambardella, Bob Hahn, AnnaLee Saxenian, and
Andy Toole for valuable comments, and Shenyi Wu for excellent research assistance. All mistakes,
however, are entirely my own.
In order to promote public understanding of the impact of regulations on
consumers, business, and government, the American Enterprise Institute and the
Brookings Institution established the AEI-Brookings Joint Center for Regulatory
Studies. The Joint Center’s primary purpose is to hold lawmakers and regulators
more accountable by providing thoughtful, objective analysis of relevant laws and
regulations. Over the past three decades, AEI and Brookings have generated an
impressive body of research on regulation. The Joint Center builds on this solid
foundation, evaluating the economic impact of laws and regulations and offering
constructive suggestions for reforms to enhance productivity and welfare. The
views expressed in Joint Center publications are those of the authors and do not
necessarily reflect the views of the Joint Center.
Executive Director
Stanford University
University of Maryland
Covington & Burling
Massachusetts Institute
of Technology
Stanford University
Stanford University
University of Michigan
Milken Institute
Massachusetts Institute
of Technology
Harvard University
University of Chicago
Harvard University
All AEI-Brookings Joint Center publications can be found at
© 2004 by the author. All rights reserved.
Executive Summary
Agglomerations, or “clusters,” of industriesand especially of high-technology
industriescan be major sources of economic growth. Policy makers therefore often search for
ways to catalyze such clusters. A popular approach is to establish a science or research park in
the hopes that it will attract companies and fuel regional economic growth. In this paper I
assemble a county-level panel dataset to explore the effects of science parks on job growth and
on venture capital. Non-parametric and econometric analysis reveals no positive effect of
science parks on regional development overall. In other words, while success stories do exist,
the analysis suggests that successes are the exception rather than the rule. Thus, policies intended
to promote cluster development by subsidizing science or research parks are unlikely to be
Do Science Parks Generate Regional Economic Growth?
An Empirical Analysis of their Effects on Job Growth and Venture Capital
Scott Wallsten
1. Introduction
Policy makers around the world are anxious to find tools that will help their regions
attract high-tech jobs and become centers of innovation and new technology. During the late
1990s this meant trying to emulate Silicon Valley. The dot-com crash put an end to those hopes,
which have now morphed into a desire by officials in many regions to become biotech hubs. For
example, Florida announced plans in late 2003 to give more than $500 million in subsidies to the
Scripps Research Institute to build a research facility in Palm Beach County; Virginia officials
urged a county to give the Howard Hughes Medical Institute a break on property taxes to make
the state appear biotech-friendly; and the city of Baltimore is hoping that a planned city-
subsidized plan will remake a run-down neighborhood into a thriving biotech center.1 Local
officials generally hope that with the right ingredients they can create a “cluster” of high-
technology activity that, in a virtuous circle, will attract more people and businesses.
Unfortunately, while we have learned a great deal about firm clustering, the composition
of clusters like Silicon Valley, and the various components of successful high-tech regions, there
is little empirical evidence on the effectiveness of public policies intended to start these clusters
from scratch. In this paper I use U.S. county-level panel data to investigate the effects of
research parks on job growth and on venture capital. I take two empirical approaches to test the
regional effect of the parks. First, I match counties with research parks to “similar” counties
without parks and compare them over time (in the spirit of Goldstein and Luger (1991a),
discussed below). Second, I test econometrically the effect of establishing science parks. I find
no evidence that research parks had any measurable economic impact, suggesting that public
subsidies to these ventures were not wise investments.
1 See Ulferts (2003) on Florida, Laris (2003) on Virginia, and Barbaro (2003) on Baltimore.
2. Industrial clustering
It is well known that many industries concentrate regionally (e.g., Krugman 1991a, 1998,
Porter 2003) and that this clustering is greater than would be expected if geographic distributions
were random (Ellison and Glaeser 1997). Industry agglomeration is not limited to high-
technology sectors. In the early nineteenth century, U.S. manufacturing was concentrated in a
small part of the Northeast and the Midwest. Historically, shoes were produced in Massachusetts
and rubber in Akron, Ohio. Carpet producers are still disproportionately located in Dalton,
Georgia, and jewelry producers are Providence, Rhode Island (Krugman 1991a). Today, high-
tech firms concentrate in areas like Silicon Valley.
Alfred Marshall in 1920 hypothesized three reasons for industrial clustering: benefits of a
pooled labor supply, access to specialized inputs, and information flows between people and
firms.2 These features may generate a positive feedback loop, in which firm concentration brings
additional labor and other inputs, encouraging additional firm concentration, and so on (e.g.,
Arthur 1994, Krugman 1991b). The specialized inputs required for industrial concentration
differ by industry, of course. It is by now conventional wisdom that universities and venture
capital are necessary components of any high-tech agglomeration.
Universities not only draw scientists and engineers to a region, but also generate
knowledge that nearby firms can use. Indeed, there is evidence of knowledge spillovers between
firms and universities. Jaffe (1989) finds that university research positively impacts patenting by
firms in the same state. Anselin, et al. (1997) find evidence of the same spillovers at a smaller
regional level using a more sophisticated spatial analysis. Saxenian (1994), meanwhile,
documents the importance of knowledge transfer between Stanford University and firms in
Silicon Valley. The existence of these knowledge spillovers suggests that universities are
important components in the virtual circle of high-tech agglomerations.
Venture capital, too, is an important component of a high tech agglomeration. Venture
capitalists, though, may be as important as the capital itself, screening business plans and
providing management advice to funded firms (Gompers and Lerner 1999).
2 Indeed, the study of regional economies and industrial clustering has experienced cyclical popularity over the past
century (McCann and Sheppard 2003).
A large high-tech labor force, a combination of large firms and new start-ups, venture
capital, venture capitalists, infrastructure that supports high tech needs (e.g., a good fiber optic
network), and university connections are all important or perceived as important components of
high tech regions. This observation often leads to the view that these are just “ingredients” that,
once in place, will generate a new Silicon Valley or biotech hub. Some of these ingredients may
appear amenable to quick policy interventions and are attractive to politicians who want to
promote regional economic development and desire immediately visible outcomes. Policy
makers sometimes believe that they might be able to create a nucleus of, and catalyst for, such a
cluster by helping establish a science or research park.
Research Parks
Research parks remain popular. The Association of University Research Parks (AURP)
counted 135 parks in the U.S. as
members in its 1998 directory,
while the International Association
of Science Parks has members in 82
countries outside the U.S.3 By
2003, the AURP had more than 200
members. Figure 1 shows the
growth in the number of science
parks in the U.S. from the first
parks in the 1950s through the end
of the 20th century.4 Felsenstein
(1994) notes that science parks are
3 AURP defines a research park as “a property-based venture which has: Existing or planned land and buildings
designed primarily for private and public research and development facilities, high technology and science based
companies, and support services; a contractual and/or formal ownership or operational relationship with one or more
universities or other institutions of higher education, and science research; a role in promoting research and
development by the university in partnership with industry, assisting in the growth of new ventures, and promoting
economic development; [and] a role in aiding the transfer of technology and business skills between the university
and industry tenants.” See (accessed January 8, 2004). The International
Association of Science Parks lists countries in which it has members and other information: (accessed January 8, 2004).
4 The number of parks that closed, if any, is not clear. The organizations that track parks are also advocacy groups
and are not eager to highlight failures.
Figure 1
Num be r of U.S . S ci e n ce Parks
generally established with two primary objectives. The first objective is “to play an incubator
role, nurturing the development and growth of new, small, high tech firms, facilitating the
transfer of university know-how to tenant companies, encouraging the development of faculty-
based spinoffs and stimulating the development of innovative products and processes.” The
second objective is to be a catalyst for regional economic development—a “growth sector
leading the area . . . into a spiral of propulsive expansion.”
Most science parks receive some form of public subsidy. Goldstein and Luger (1991a)
note, “many parks are public corporations or subsidiaries of public universities. Others are
privately owned but may receive various types of government subsidies including land,
buildings, services and infrastructure, and property tax reductions. Less direct government
subsidies to science/technology parks can be through the provision of specially designed
economic development, education, and job training programs, at the state level, and through
favorable land-use policies which favor expansion, at the local level.” Notably absent from the
literature on science parks, however, is any real discussion of their costs or estimates of public
expenditures on them.
Some science parks have been successful. The Research Triangle Park (RTP) in North
Carolina, for example, has been considered a success for some time (e.g., Braun and McHone
1992, Goldstein and Luger 1991b). RTP currently hosts 38,500 full-time employees and 131
organizations.5 Success stories like this, along with a few others such as the Stanford Research
Park in the heart of Silicon Valley, encourage others to build parks in the hopes of emulating that
success. Even if a science park itself is successful, however, that success may not spill over into
the local economy. While the Research Triangle region now exhibits many features of a high
tech area,6 RTP had not stimulated a regional technology cluster even by the early 1990s, despite
having been established in 1959 (Braun and McHone 1992).
Success stories like RTP seem to be more the exception than the rule. For example, San
Antonio broke ground on its Texas Research Park in the mid-1980s among predictions of hosting
50,000 jobs and generating another 100,000 spinoff jobs within 30 years (Haines-Saine 1985).
While it has not been that long yet, it does not look promising: about 300 jobs so far (Hundley
5 See (accessed
January 8, 2004).
6 See, for example, Hillner (2000) who ranked the Research Triangle among the 46 “locations [in the world] that
matter most in the new digital geography” in the July 2000 issue of Wired.
2003). A research park established in Prince George’s County, Maryland, in the mid-1980s
promised 12,000 jobs on the park and 25,000 related spinoff jobs (PR Newswire 1988). A local
council member recently called that park a “failure,” and the state of Maryland wants a refund on
some of the millions of dollars it invested in the site’s infrastructure (Wilen 2003).
A possible explanation for the few examples of “science park-led local economic
development” (Felsenstein 1994), is that, as Jowitt (1991) observes, research parks are often just
a political quick fix to industrial decline. Indeed, policy makers in regions experiencing
economic downturns (either absolutely or relative to other regions) are likely to face pressure to
generate economic development. A science park may be a politically attractive option since it
can be constructed relatively quickly, generating at least an appearance of economic
development activity. It can further generate an appearance of success when firms move into the
park. Cities and research park organizations routinely count as “success” any firms or
employment in the park, with no regard to whether that economic activity was new to the region
or simply relocated into the park, and no analysis of whether that activity would have been likely
to occur without the park. Moreover, as noted above, the costs of the park (many of which might
be hidden, such as the opportunity cost of the land) are rarely calculated. In other words, cost-
benefit analyses of research parks are likely to count as benefits any economic activity in the
park regardless of whether it is, in fact, a net benefit, and ignore the costs altogether.
In order to generate economic growth, a science park would have to encourage firm
growth that would not have happened without the park or generate spillovers that would
otherwise be absent.7 The first criteria would be difficult and data-intensive to answer; to my
knowledge no study explores it comprehensively. Still, some research explores differences
between firms on and off science parks, and other research explores potential links between
parks and their surrounding communities. A small body of additional research looks for regional
effects of these parks.
There is some evidence that firms located in science parks differ from firms located
outside the parks (but in the same region). Braun and McHone (1992), for example, found that
firms in the Central Florida Research Park were more likely to be branch plants than firms
outside the park. Ferguson (1999) found that firms in Swedish science parks tended to be
younger and smaller than firms outside the park. But differences between firms on and off the
7 And in a global sense, the park has an impact only if firms do not simply relocate from one place to another.
parks are to be expected and by itself this phenomenon is difficult to interpret. Parks may have
selection or investment criteria, for example, ensuring that they host only particular types of
firms. While the selection mechanism or incentives offered may attract particular types of firms,
the differences among firms, per se, does not imply an economic effect of the park.
Most research finds little, if any, real effects of research parks. Felsenstein (1994) finds
little evidence that firms in science parks engage in more research, have stronger linkages to
universities, or witness greater transfers between other local firms than do firms not located in
science parks. Indeed, he concludes that the evidence suggests “parks may function as ‘islands’
of innovation or as collections of firms with no real links between them.” Braun and McHone
(1992) note a lack of linkages between firms in science parks and local economic actors outside
the park. Spillovers to the larger region are probably less likely without such linkages.
While the results discussed above suggest it is unlikely that science parks have positive
regional effects, there is scant evidence on this matter. Goldstein and Luger (1991b) provide the
only evidence to date on this score. They matched U.S. counties with science parks to “similar”
counties without science parks and compared total county employment growth rates before and
after the park was established. Of the 45 parks in their sample 32 were in counties that grew
faster than the matched counties, and 26 of the 45 grew more than 20 percent faster than the
matched counties. Still, it is difficult to interpret these results since the analysis—despite the
matched sample—does not control for other factors that could influence economic growth, and
the authors provide little information on how they chose their control counties.
I build on this work by updating and greatly expanding the dataset to include far more
measures of a high-tech economy over a longer period of time. In addition, I conduct more
rigorous econometric tests that attempt to control for reasons parks may have been established in
the first place. I discuss the data, methods, and analysis below.
3. Data
The county-level data I analyze in this paper come from a variety of sources and cover
1988-1997, though the time period covered varies by data source. Venture Capital data comes
from VentureXpert, a database compiled by Venture Economics.8 Again, these data are provided
8 See Gompers and Lerner (Gompers and Lerner 1999) for a description of the Venture Economics data.
at the firm level, along with address and deal amount. The VC data includes information from
1983 through 1999. University data comes from the National Science Foundation’s CASPAR
database. This database provides information on all U.S. universities, including an address. The
firm and university addresses allow me to aggregate this data into counties.9
Employment data and firm counts by industry come from the U.S. Census Bureau County
Business Patterns. While I have this industry data back to 1986, the 1987 changes to the
Standard Industrial Classification (SIC) mean that time series analyses using these data should
begin in 1988. Population estimates, government and military employment, and per capita
income are available from the Department of Commerce Bureau of Economic Analysis Regional
Economic Information System (REIS). I have these data from 1983-1997. These sources all
provide data at the county level.
The policy variable of interest in this paper is whether a county established a research or
science park. This information comes from the Association of University Research Parks, which
compiled this information in its 1998 directory. A park may affect a region in many ways, but I
look at three in particular: high-tech employment, firms, and venture capital. I discuss these in
more detail below.
High-tech Firms and Employment
Because one objective of this paper is to look for impacts on regional technology
development, I need a measure of employment in technology-related industries in addition to
total employment. County Business Patterns provides data down to the four-digit SIC level,
allowing me to construct such a measure. As many authors have noted, however, at least two
problems arise in defining “technology industries.” The first is simply that the term itself is
ambiguous. Almost all industries use and even develop advanced technology to some extent,
making any definition of “high-tech” at least somewhat arbitrary. The second is that the
Standard Industrial Classifications, even at the four-digit level, are fairly crude and do not
accurately classify firms—especially large, multiproduct firms. Nonetheless, following other
authors, it is possible to construct a crude definition of technology firms.
9 I employed a computerized geographic information system (GIS) to aggregate firm and university data into
counties. The GIS reads in the observation’s zip code and matches it to the county containing that zip code.
Occasionally zip codes overlap counties; in these cases the GIS uses the centroid of the zip code to identify a county.
I begin with DeVol’s (1999) definition, which “includes industries that spend an above-
average amount of revenue on research
and development and that employ an
above-average number of technology-
using occupations—such as scientists,
engineers, mathematicians, and
programmers.” Table 1 lists the
industries that comprise “high-tech” in
this paper.10 While a reasonable
definition, it includes many industries
that deal primarily with military
research and manufacturing (e.g.,
guided missiles).
Military spending has
historically had a large impact on
technology development—both in the
technological direction and geographic location of R&D. However, changes in employment
related to military R&D are driven largely by exogenous factors—the end of the Cold War, for
example, brought about dramatic reductions in all areas of military spending, while the war on
terrorism and in Iraq are now increasing military spending. While the effects of changes in
military-related employment are interesting to study, including those changes in an aggregate
measure of employment clouds the picture of regional changes in technology employment.
I thus calculate an alternate measure of high-tech employment excluding three industries:
aircraft and parts (SIC 372), guided missiles and space vehicles (SIC 376), and search, detection,
navigation, and guidance equipment (SIC 381). This variable should measure non-military high-
tech employment.
10 DeVol’s (1999) definition also includes SIC 781, motion pictures and allied services. I exclude this industry
from my analysis because it seems to have been included in the DeVol study largely because of its concentration in
Los Angeles rather than its role as a “high-tech industry.”
Table 1
High Tech Industries
SIC Defintion
283 Pharmaceuticals
357 Computer and Office Equipment
366 Communications Equipment
367 Electronic Components and Accessories
372 Aircraft and Parts
376 Guided Missiles and Space Vehicles
381 Search, Detection, Navigation, and Guidance
382 Laboratory Apparatus
384 Surgical, Medical, and Dental Instruments
481 Telephone Communications
737 Computer Programming & Data Processing
871 Engineering, Architectural, and Surveying
873 Research, Development, and Testing
Industries in italics (372, 276, 381) excluded from some
calculations to remove effects of military spending.
Venture capital
As mentioned above, the venture capital data comes from Venture Economics, which
collects and disseminates data for the National Venture Capital Association. Because “venture
capital” is, in general, so difficult to define, aggregated estimates of VC funding can differ
depending on the source of the data. Venture Economics has one of the broadest definitions of
“venture capital,” including not only venture funds but also “other private equity funds.”11 The
result is that Venture Economics data yields the largest estimates of aggregate VC funding. For
example, Venture Economics reported total venture funding in 1998 of $19.2 billion,
PriceWaterhouseCoopers reported $14.2 billion, and VentureOne reported $12.5 billion.12
Fortunately, while levels differ by source, changes over time and differences across regions do
not. As such, an analysis that makes use of the variance across time and regions, as this paper
does, should not be greatly affected by data source.13 Nonetheless, I partially address this issue
by removing from the VentureEconomics data all “non-high technology” (i.e., those listed as
“consumer-related” or “other”). This deletion brings the VC data closer to other estimates (the
1998 total drops to $17.2 billion) and is more in line with the technology focus of this paper.
4. Empirical tests
Investigating the effects of relatively small policy interventions is difficult with
aggregated data—even at the county level. I take two different approaches for exploring the
data. While problems exist with each approach, the similarity of the results lends some
robustness to the final conclusions.
First, I match “treatment” counties—those that built research parks—with similar
counties without the policy intervention and compare changes in high-tech employment and
venture capital over time. While matching counties in this way is imprecise at best, it provides a
11 See for a description of Venture Economics’ data
12See,, and for the Venture Economics,
PriceWaterhouseCoopers, and VentureOne data, respectively.
13 Unfortunately, I cannot test this claim rigorously. Venture Capital data is costly, and using it labor-intensive,
making it extremely difficult to directly compare all the sources.
first rough cut of the effects of these policy prescriptions. Second, I conduct a more rigorous
regression analysis.
The first approach to exploring the effects of science parks is to match counties that built
parks to similar counties that did not, much in the spirit of Goldstein and Luger (1991b). My
analysis differs from theirs, however, in several ways. First, they looked only at changes in total
county employment, while I look at high-tech employment and also venture capital. Second, I
define “similar” in a precise way, while how they chose matches is not clearly defined. Finally,
a decade has passed since they completed their work; it is time for an update.
For each county that built a research park, I attempted to identify similar counties that did
not.14 I considered a county to be a control if its population, high tech employment, and venture
capital were all within 30 percent of the levels for the treatment county in the year that it built the
park. This definition yielded matches for 41 counties that opened parks from 1986 onwards
(though because of the SIC changes I compare employment changes only for the 26 counties that
opened a park in 1988 or later). Several counties yielded only one match, while one county,
Gallatin, Montana, with its Advanced Technology Park, yielded 147 matches.
Appendix tables 1 and 2 show high-tech employment and venture capital, respectively, in
the treatment and control counties in the year the park was established and five years later. The
tables reveal little difference between the groups. The number of high-tech jobs increased, on
average, from 5814 to 6283 in the 27 treatment counties, and from 5184 to 6723 in the control
counties. Venture capital increased from about $5.7 million to $6.8 million on average in the 41
treatment counties, and from about $5.0 million to $10.9 million in the control counties. Nine of
the treatment counties ended up with more venture capital than their controls, and in eight cases
both the treatment and control counties attracted no venture capital.
Figure 2 and Figure 3 show means from five years before and to five years after park
establishment for high tech employment and venture capital, respectively. The nature of the data
means that each data point in the graphs cannot, unfortunately, be calculated over the same
sample size. My employment data begins in 1988, meaning that I have no pre-park data when
the park was established that year (and only one year of pre-park data when the park was
established in 1989, and so on). Likewise, my data stops in 1997, meaning that I have no
information for time t+5 for treatment counties and their controls after 1992 (and no information
14 Because my data starts in 1986, the analysis excludes any county that built a park before that year.
for time t+4 after 1993, and so on).
Despite their statistical shortcomings, the
figures strongly suggest no substantial
differences between the treatment
counties and their controls.
While the tables and figures are
compelling, they are not, by themselves,
entirely convincing. In addition to the
data issues discussed above, matching
counties based on a few variables cannot
control for the many other ways in which
counties differ. To look at this question
more rigorously I estimate equation (1)
below, first only with the counties that
established parks to investigate the before-after effects of the park, and then including the control
(1) Yit =
0 +
t +
j +
1*(Science park dummyit) +
2Z +
I estimate Equation (1) three
times, each time using a different
dependent variable. I first define Yit
as high-tech non-military employment
in county i in year t, next as venture
capital in the county-year, and finally
as the number of small high-tech
firms in the county-year. The science
park dummy equals one if there is a
science park in that county-year, and
Figure 2
Science Park Treatment-Control Com parison
High-Tech Employment
t-5 t-4 t-3 t-2 t-1 0 t+1 t+2 t+3 t+4 t+5
year relative to park establishment
high-tech employment
Science park counties Control counties
Figure 3
Science Park Treatment-Control Comparison
Venture Capital
t-5 t-4 t-3 t-2 t-1 0 t+1 t+2 t+3 t+4 t+5
Year relative to park establishment
Venture Capital ($000)
Park Counties Control Counties
zero otherwise. Z is a vector of independent variables that are potentially important components
of regional development. Continuous independent variables that vary across county and time
include university R&D spending, personal income, non high-tech employment, the number of
large high-tech firms not primarily engaged in military activities, and the number of large high-
tech firms engaged primarily in military related activities, where “large” is defined as a firm with
more than 500 employees. Finally, I control for year (
t) and county (
j) fixed effects. I explain
the rationale for each variable’s inclusion below.
Considerable evidence suggests that universities are important components of a region’s
economy. For example, Jaffe (1989) finds spillovers from university research at the state level,
while Saxenian (1994) notes Stanford’s key role in Silicon Valley. I include university R&D
spending in the county both to control for the presence of a research university and also to test
their effects in this context. Personal income proxies for wealth and cost of living. Non high-
tech employment controls for general (non high-tech) economic conditions and size of the labor
I include counts of large high-tech firms since they are important determinants of
regional technology development but are less likely than small firms to move to a region because
of a science park.15 Indeed, newer research parks (those that opened in the time period in this
sample) aim to attract primarily small high-tech firms. Several authors have noted the
importance of the initial conditions in determining the growth trajectory of a region (e.g., Arthur
1994, Krugman 1995). The county fixed effects help control for such initial conditions as well
as county specific, but otherwise unobserved, features that are likely to affect technological
development. Year fixed effects control for time trends, which could otherwise contribute to
spurious correlations.
15 With some exceptions, of course. RTP now hosts several large firms, including Glaxo Wellcome and IBM.
However, RTP is an exception as one of the few very large and successful parks.
Tables 2 and 3 highlight the results of this analysis. Table 2 shows the results when
estimating the equation using only the counties that established science parks in order to get a
sense of the before-after effects.
Within these counties, the analysis
reveals that the number of large
non-military firms is positively and
significantly correlated with high-
tech employment, negatively and
significantly correlated with high-
tech venture capital, and not
statistically correlated with the
number of small high-tech firms.
The number of large military-
focused firms is weakly positively
correlated with the number of small
high-tech firms. The number of
government employees in the
county is positively correlated with
the number of high-tech jobs,
negatively correlated with venture
capital, and positively correlated with the number of small high-tech firms.
The coefficient of interest is on the research park dummy variable. Recall that in this
case the dummy variable equals zero before the county established the park and one afterwards.
The coefficient, while negative, is not statistically significant, suggesting that establishing the
science park had no measurable impact on the number of high-tech jobs, venture capital, or the
number of small firms in the county.
Table 2
Science parks and high-tech employment
(includes only counties that established parks)
dependent variable number of
high-tech jobs
number of
small high
tech firms
mean of dependent var: 9226 5338 307
Science park? -110.118 -967.370 -11.945
(0.26) (0.55) (1.41)
University R&D spending 0.005 0.009 0.000
(0.77) (0.32) (0.74)
non-tech employment 0.001 0.164 0.0008
(0.19) (5.23)** (2.64)**
personal income 0.788 1.260 0.060
(11.06)** (4.33)** (42.65)**
num large non-military firms 818.167 -1,589.503 2.614
(6.19)** (2.94)** (1.01)
num large military firms -662.339 -2,009.042 16.476
(1.57) (1.17) (1.99)*
government employment 0.223 -0.789 0.008
(5.38)** (4.65)** (10.03)**
Constant -11,368.250 6,586.608 -649.597
(4.78)** (0.68) (13.92)**
Observations 672 672 672
R-squared 0.38 0.20 0.90
Absolute value of t statistics in parentheses
+ significant at 10%; * significant at 5%; ** significant at 1%
Year and county fixed effects included in all regressions
Table 3 incorporates the control
counties, which allows us to compare
how the science park counties performed
relative to the other counties once they
established their parks. In this case the
table shows that university R&D
spending has a positive and statistically
significant correlation with the number of
high-tech jobs and the number of small
high-tech firms, consistent with research
that has demonstrated local spillovers of
university research.
The science park dummy variable
here turns out to be statistically
significant and negatively correlated with
the number of high-tech jobs and venture
capital. It is positively correlated with
the number of small firms, but is not statistically significant. In other words, the results here
suggest that counties that established science parks actually did worse—on average and all else
equal—than the counties that did not establish the parks. In sum, the econometric results match
the nonparametric results: there is scant evidence that establishing a research park aided regional
5. Conclusion
Industrial clusters of economic activity are real and can be major sources of economic
growth. While this has been true for a long time, since the 1980s clusters of high-technology
activity have gotten the most attention, culminating in the late 1990s with a focus on Silicon
Valley and today on biotech hubs like San Diego. Politicians are attracted to the idea of
clustering because they are attracted by the idea that with the right “ingredients” their region can
Table 3
Science parks and high-tech employment
(includes counties with and without parks)
Dependent variable
number of
number of
small high
tech firms
mean of dependent variable 1788 862 74
Science park? -882.951 -4,081.143 2.139
(7.63)** (7.84)** (0.78)
University R&D spending 0.008 0.010 0.001
(4.93)** (1.32) (15.93)**
non-tech employment 0.011 0.125 0.001
(5.41)** (14.03)** (15.85)**
personal income 0.648 1.116 0.055
(36.98)** (14.18)** (132.76)**
num large non-military firms 847.653 -29.152 5.462
(30.31)** (0.23) (8.28)**
num large military firms -311.038 -94.156 7.105
(4.44)** (0.30) (4.30)**
government employment 0.091 -0.647 0.005
(8.84)** (13.99)** (20.14)**
Constant -2,027.551 -296.348 -169.404
(15.73)** (0.51) (55.72)**
Observations 8181 8181 8181
R-squared 0.400.120.86
Absolute value of t statistics in parentheses
+ significant at 10%; * significant at 5%; ** significant at 1%
Year and county fixed effects included in all regressions
become another such cluster. One common technique for trying to catalyze such growth is to
establish or subsidize a research park.
This paper uses a host of county-level panel data to test whether such plans tend to be
effective. This paper is not, of course, the final answer. Investigating all the effects of science
parks requires data disaggregated below the county level. A more detailed analysis would
determine whether firms and other organizations moved into the park (which might be an
indicator of park success), whether they simply moved from one location within the region to
another (which would mean no net regional impact), and—importantly—calculate the costs of
the park.
Nonetheless, this analysis suggests that establishing a research park tends to have no net
impact on job growth, the number of firms, or on the amount of venture capital attracted to the
county. That is, while there are successful research parks, they seem to be the exception rather
than the rule. These results are consistent with Porter’s (2003) conclusion that it is difficult to
start new regional clusters from scratch. While high tech clusters can be major sources of
economic growth, and industrial clustering is common, the results in this paper suggest that
research parks are not, in general, likely to help generate one, and that subsidies spent on them
are likely to be ineffective.
Anselin, Luc, Attila Varga, and Zoltan Acs. "Local Geographic Spillovers between University
Research and High Technology Innovations." Journal of Urban Economics 42 (1997):
Arthur, Brian. Increasing Returns and Path Dependence in the Economy. Ann Arbor: The
University of Michigan Press, 1994.
Barbaro, Michael. "Baltimore Makes a Bold Bid to Transform Neighborhood." The Washington
Post 2003, A01.
Braun, Bradley, and W. Warren McHone. "Science Parks as Economic Development Policy: A
Case Study Approach." Economic Development Quarterly 6, no. 2 (1992): 135-47.
DeVol, Ross. America’S High-Tech Economy: Growth, Development, and Risks for
Metropolitan Areas. July ed. Santa Monica, CA: The Milken Institute, 1999.
Ellison, Glenn, and Edward Glaeser. "Geographic Concentration in U.S. Manufacturing
Industries: A Dartboard Approach." Journal of Political Economy CV (1997): 889-927.
Felsenstein, Daniel. "University-Related Science Parks-‘Seedbeds’ or Enclaves of Innovation?"
Technovation 14, no. 2 (1994): 93-110.
Ferguson, Richard. "What's in a Location? Science Parks and the Support of New Technology-
Based Firms." The Swedish University of Agricultural Sciences, 1999.
Goldstein, Harvey, and Michael Luger. Science/Technology Parks and Regional Development:
Prospects for the United States. Edited by Ulrich Hilpert, Regional Innovation and
Decentralization: High Tech Industry and Government Policy. London and New York:
Routledge, 1991a.
———. Technology in the Garden. Chapel Hill: University of North Carolina Press, 1991b.
Gompers, Paul, and Josh Lerner. The Venture Capital Cycle. Cambridge, MA: MIT Press, 1999.
Haines-Saine, Renee. "San Antonio Readies Biotech Boomtown Pitch." United Press
International, June 9 1985.
Hundley, Kris. "Boom or Bust?" St. Petersburg Times, November 24 2003.
Jaffe, Adam. "Real Effects of Academic Research." The American Economic Review LXXIX
(1989): 957-70.
Jowitt, Anthony. Science Parks, Academic Research and Economic Regeneration: Bradford and
Massachusetts in Comparison. Edited by Ulrich Hilpert, Regional Innovation and
Decentralization: High Tech Industry and Government Policy. London and New York:
Routledge, 1991.
Krugman, Paul. Geography and Trade. Cambridge: MIT Press, 1991a.
———. "Increasing Returns and Economic Geography." Journal of Political Economy XCIX
(1991b): 483-99.
———. "Space: The Final Frontier." Journal of Economic Perspectives XII (1998): 161-74.
Laris, Michael. "Biomedical Institute's Tax Break Uncertain." The Washington Post, December 1
2003, B01.
McCann, Philip, and Stephen Sheppard. "The Rise, Fall and Rise Again of Industrial Location
Theory." Regional Studies 37, no. 6&7 (2003): 649-63.
Porter, Michael. "The Economic Performance of Regions." Regional Studies 37, no. 6&7 (2003):
PR Newswire. "Um, Government and Business Leaders Officially Open the Center." 1988.
Saxenian, AnnaLee. Regional Advantage: Culture and Competition in Silicon Valley and Route
128. Cambridge: Harvard University Press, 1994.
Ulferts, Alisa. "Huge Incentives Lure Research Group to Florida." St. Petersburg Times, October
10 2003, 1A.
Wilen, John. "P.G. To Rezone Mie's Failed Technology Park." Baltimore Business Journal, June
16 2003.
... Thompson, Purdy and Ventresca (2017) found that development was furthered by endogenous factors (language and interaction) rather than public policy. Common interventions often seek to install the features of successful systems or identify likely 'winners' among new ventures (Harrison & Leicht, 2010), but shortcuts such as establishing a science park, absent other regional resources, do not yield vibrant entrepreneurial ecosystems and regional prosperity (Wallsten, 2004). Moreover, many efforts that yield high new venture founding rates fail to stimulate the formation of the high-growth potential firms responsible for regional prosperity (Brown & Mason, 2017;Shane, 2009), suggesting that the necessary elements to foster growth were missing. ...
Full-text available
In this research, we develop the concept of entrepreneurship platforms, a new perspective on entrepreneurial ecosystems. We examine their growth from pre-platform regions to emergent, adolescent, and mature entrepreneurship platforms. At each stage, we propose policy interventions to help foster platform development based on the prevailing network effects within and across groups of entrepreneurs and investors. Our conceptualization contributes to what is known about entrepreneurial ecosystems and has implications for both theory and practice.
... However, at the same time, the set of factors varies substantially from case to case and cannot be summarised into a simple formula, applicable everywhere and anytime (Weiner, 2016). In contrast, we can give a fair number of examples where policies aiming to boost growth based on science and technology have failed spectacularly: from underperforming USSR science towns (Josephson, 1997) to failed science parks in the US (Luger & Goldstein, 1991;Kefalides, 1991;Wallsten, 2004), Greece (Bakouros, Mardas, & Varsakelis, 2002), India (Phan, Siegel, & Wright, 2005), Poland (Najwyz˙sza Izba Kontroli, 2013), and the UK (Massey, Quintas, & Wield, 1992). ...
... Recognition of the heterogeneity within types of STPs and their close relations clearly shows that in analyzing outcomes and impact it is insufficient to compare firms participating in these support mechanisms with those that do not using a simple dummy variable. For example, with regard to a county-level panel dataset, Wallsten (2004) explored the provision or not of STPs at a county level in relation to job growth and the take-up of venture capital. This study found no positive effect of STPs on regional development. ...
This book is the first collection of scholarly writings on science and technology parks (STPs) that has an international perspective. It explores concrete ways to systematically collect information on public and private organizations related to their support of and activities in STPs, including incubation to start-up and scale-up, and collaborations with centers of knowledge creation. Rather than perpetuate the qualitative assessment of successful practices, the focus of this book is to present quantitative and qualitative evidence of the impact of STPs on regional development and to raise awareness on the importance of systematic data collection and analysis. Only through a systematic collection of data on fiscal identification numbers of companies, universities, and university spin-offs will it be possible to conduct current and especially future analyses on the impact of STPs on entrepreneurship, effectiveness of technology transfer, and regional economic development. To this extent, the synergistic views of academics, representatives from STPs, and policy experts are crucial.
... Recognition of the heterogeneity within types of STPs and their close relations clearly shows that in analyzing outcomes and impact it is insufficient to compare firms participating in these support mechanisms with those that do not using a simple dummy variable. For example, with regard to a county-level panel dataset, Wallsten (2004) explored the provision or not of STPs at a county level in relation to job growth and the take-up of venture capital. This study found no positive effect of STPs on regional development. ...
This chapter has two objectives. First, it discusses how the importance of Science and Technology Parks (STPs) as tools for place-based policy and for the development of innovation eco-systems is not matched by a systematic collection of data and unbiased impact evaluation. Second, it summarizes the main findings, common challenges and recommendations deriving from the “Workshop on Science and Technology Parks in Europe – Steps towards a systematic and harmonized data collection” (Seville, Spain, 10–11 October 2018). The contributions to the workshop are grouped into three thematics: (1) the relevance of STPs and clusters for innovation eco-systems; (2) heterogeneity and evolution of STPs and (3) identification and measurement issues.
... Recognition of the heterogeneity within types of STPs and their close relations clearly shows that in analyzing outcomes and impact it is insufficient to compare firms participating in these support mechanisms with those that do not using a simple dummy variable. For example, with regard to a county-level panel dataset, Wallsten (2004) explored the provision or not of STPs at a county level in relation to job growth and the take-up of venture capital. This study found no positive effect of STPs on regional development. ...
This chapter presents a conceptual framework and two cases illustrating the coordination or orchestration role played in their local innovation ecosystem by two distinctive science and technology parks in two place-based innovation ecosystem in Europe, Lindholmen Science Park in Gothenburg, Sweden, and Technology Park Ljubljana in the capital city of Slovenia—which are also presented individually in this publication. This chapter is extensively based on the research conducted for the Joint Research Centre (JRC) Series on Place-based Innovation Ecosystems since 2017 with various place-based innovation ecosystems mapped out in Europe and beyond.
... Recognition of the heterogeneity within types of STPs and their close relations clearly shows that in analyzing outcomes and impact it is insufficient to compare firms participating in these support mechanisms with those that do not using a simple dummy variable. For example, with regard to a county-level panel dataset, Wallsten (2004) explored the provision or not of STPs at a county level in relation to job growth and the take-up of venture capital. This study found no positive effect of STPs on regional development. ...
We summarize and synthesize the frameworks of analysis and quantitative and qualitative evidence from several countries of the impact of science and technology parks (STPs) both on regional development and on tenant firms and territorial development. We observe that the chapters recognize the importance of taking into account the multi-dimensional environmental context within which STPs operate, emphasize the importance of understanding interactions and networks among the various stakeholders involved with STPs, and identify the multiplicity of types of STPs regarding their ownership and goals, business models and strategies, resources and outcomes over time. We outline a high level schematic view of these relationships that form a basis for identification of data sources and policy development.
... However, at the same time, the set of factors varies substantially from case to case and cannot be summarised into a simple formula, applicable everywhere and anytime (Weiner, 2016). In contrast, we can give a fair number of examples where policies aiming to boost growth based on science and technology have failed spectacularly: from underperforming USSR science towns (Josephson, 1997) to failed science parks in the US (Luger & Goldstein, 1991;Kefalides, 1991;Wallsten, 2004), Greece (Bakouros, Mardas, & Varsakelis, 2002), India (Phan, Siegel, & Wright, 2005), Poland (Najwyz˙sza Izba Kontroli, 2013), and the UK (Massey, Quintas, & Wield, 1992). ...
Full-text available
Science is increasingly defined by multidimensional collaborative networks. Despite the unprecedented growth of scientific collaboration around the globe—the collaborative turn—geography still matters for the cognitive enterprise. The spatial location and distance between scholars and research organisations affect their likelihood to collaborate and to achieve results that expand the knowledge frontier. At the same time, the developmental prospects of cities, regions, and countries result from, and depend on, their place in the global network of research collaboration. This book explores how geography conditions scientific collaboration and how collaboration affects the spatiality of science. Addressing these questions requires the reconstruction of historical developments that led to the collaborative turn in science, the examination of mutual relations between science and places, and the analysis of spatial patterns of research collaboration at various levels: from individual to global. By combining a vast array of approaches, concepts, and methodologies, the volume offers a comprehensive theoretical framework for the geography of scientific collaboration. The momentous role of collaboration for contemporary science implies that understanding the processes and patterns of research collaboration is now essential, not only for scholars interested in science studies, but also for policymakers and managers of research organisations. Those more practically oriented readers may be particularly attracted by the case studies of scientific collaboration policies from the European Union, the United States, and China, as well as the catalogue of tools for scientific collaboration policy.
Full-text available
Das Modellvorhaben „Land(auf)Schwung“ des Bundesministeriums für Ernährung und Landwirtschaft hat zum Ziel, neue Wege in der Entwicklung ländlicher Räume zu erproben. 13 periphere, ländliche Kreise wurden von 2015 bis 2019 mit jeweils ca. 2,5 Mio. Euro gefördert, um Ansätze zur Sicherung der Daseinsvorsorge sowie zur Stärkung der regionalen Wertschöpfung zu entwickelnund umzusetzen. Benachteiligte ländliche Regionen sind oftmals geprägt von einer begrenzten Innovationskraft aufgrund fehlender Wissensinfrastrukturen, einer Branchenstruktur mit niedrigemInnovationsniveau, weniger Startups und einer Dominanz von Klein- und Kleinstbetrieben. Die regionale Wirtschaft bietet häufig wenige Arbeitsplätze für Hochqualifizierte an und hat gleichzeitigaufgrund geringerer Produktivität Schwierigkeiten, Fachkräfte anzuziehen. Diese Herausforderun-gen wurden in zahlreichen Projekten zur Stärkung der regionalen Wertschöpfung angenommen. So wurden u. a. Kleinstbetriebe im Lebensmittelbereich mit Investitionshilfen bei der Entwicklungneuer Produkte unterstützt, regionale Vermarktungsinitiativen wurden auf- und ausgebaut, um denAbsatz regionaler Produkte zu steigern, und Technologie- und Gründerzentren wurden eingerichtet. Die Begleitforschung nutzte diese entstandene Projektlandschaft als Datengrundlage für dieAbleitung übergeordneter Erkenntnisse und Handlungsempfehlungen. Es wurden vertiefende Fallstudien (insgesamt 83 Interviews) zu regionalen Produkten durchgeführt sowie die Einrichtung vonzwei Gründerzentren untersucht und eine standardisierte Onlinebefragung von insgesamt 166 Klein(st)betrieben, die in regionalen Vermarktungsinitiativen organisiert sind, ausgewertet. Die Untersuchungen weisen auf ein reges Innovationsverhalten der Klein- und Kleinstbetriebe hin. Durch investive Förderung von an die Nischenstrategie angepassten Fertigungsanlagen sowie Unterstützung bei der Netzwerkbildung können sie im Wachstumsprozess unterstützt werden. RegionaleVermarktungsinitiativen fördern Kooperationen zwischen den beteiligten Klein(st)betrieben und können so Innovationen anstoßen und letztendlich das Betriebswachstum positiv beeinflussen. Da Vermarktungsinitiativen teilweise mit hohem Aufwand insbesondere für eine gemeinsame Logistikverbunden sind, sollten zunächst die regionalen Voraussetzungen für einen wirtschaftlichen Betriebüberprüft werden. Technologie- und Gründerzentren können zwar die Eintrittsbarrieren in die Selbstständigkeit senken, sollten aber immer in eine kohärente regionale Strategie der Wirtschafts-förderung eingebettet sein, um ihr Potenzial entfalten zu können.
Full-text available
This article addresses the issue of technology transfer in the context of institutional conditions of this process with particular focus on intermediary institutions, such as technology parks. The study presents the role of institutions in the effective process of technology transfer. The research conducted by the Polish Agency for Enterprise Development and the Association of Organizers of Innovation and Entrepreneurship Centres in Poland was used.
This paper examines the role of science parks as ‘seedbeds’ of innovation. Making the distinction between the spatial and the behavioural conceptions of the seedbed metaphor, the paper surveys the evidence related to the limited interaction effects between science park firms on the one hand and their neighbouring park firms, local universities and off-park firms on the other. This suggests that science parks might be functioning as ‘enclaves’ of innovation rather than seedbeds.This hypothesis is empirically tested on the basis of a survey of over 160 high-technology firms in Israel located both on and off-park. Specifically, the following questions are addressed: (1) are seedbed effects important inputs to a firm's innovation level? and (2) to what extent are these effects contingent on the physical proximity and clustering afforded by science park location? The results indicate that, first, seedbed effects, as indicated by level of interaction with a local university and the entrepreneur's educational background, are not necessarily related to the firm's innovative level; second, science park location is shown to have only a weak and indirect relationship with innovation level. It is proposed that the role of the science park is thus innovation-entrenching rather than innovation-inducing. The attraction of science park location could therefore be due to perceived status and prestige conferred rather than benefits in terms of technology transfer and information flow.
Many state and local economic development groups have attempted to create environments that are conducive to the expansion and growth of "high-tech" industry clusters based on the commonly held notion that high-tech clusters offer advantages over other types of industrial activity. There have been numerous attempts to recreate the perceived important elements of the developmental environment that sparked and fostered the growth of the Boston Route 128 and Silicon Valley high-tech clusters. One widely used approach in this regard has been the creation of a focused university-related research park This article uses the results of a recent survey of high-technology firms in Orlando, Florida to compare and contrast the characteristics of firms that have located in a university-related research park with high-tech firms that operate in other parts of the metropolitan area. In general, the survey revealed substantial differences between these two categories of high-tech firms, extending across many dimensions of firm structure and development. Additionally, the survey reveals some similarities in the organizational structure of firms located in Orlando's Central Florida Research Park and firms at North Carolina's Research Triangle Park
This paper re-examines the empirical evidence on the degree of spatial spillover between university research and high technology innovations. The familiar Griliches–Jaffe knowledge production function is estimated at both the state and the metropolitan statistical areas (MSA) level and extended with more precise measures of spatial spillover. Alternatives based on the gravity potential and covering indices are formulated for Jaffe's “geographical coincidence index” and found to provide strong evidence of local spillovers at the state level. At the MSA level, a distinction is made between research and development activities and university research in the MSA and in the surrounding counties. Evidence is found of local spatial externalities between university research and high technology innovative activity, both directly and indirectly via private research and development.
Three principal aspects of venture capital (VC) are empirically explored: fundraising, investing, and exiting those investments. Despite the recent attention to VC, misconceptions abound that the authors attempt to correct. Throughout, the discussions are based on examinations of a large sample of firms, VC funds, and investments. Three themes are elaborated in the volume: (1) The great incentive and information problems venture capitalists must overcome; (2) the interrelatedness of each aspect of the VC process and how it proceeds through cycles; and that (3) the VC industry adjusts slowly to shifts in the supply of capital or the demand for financing. The VC partnership is the intermediary between investors and high-tech start-ups. The fundraising aspect is examined in terms of its structure, means of compensation, and the importance of the structure of the limited partnership form used by most VC funds. The need to provide incentives and shifts in relative negotiating power impact the terms of VC limited partnerships. Covenants and compensation align the incentives of VC funds with those of investors; covenants and restrictions limit conflicts among investors and venture capitalists. Supply and demand and costs of contracting determine contractual provisions. VC contracting may not always be efficient. During periods of high demand and capital flows, partners negotiate compensation premiums. The investing aspect is discussed in terms of why investments are staged, how VC firms oversee firms, and why VC firms syndicate investments. Four factors limit access to capital for firms: uncertainty, asymmetric information, nature of firm assets, and conditions in the financial and product markets. These factors determine a firm's financing choices. Asymmetries may persist longer in high-tech firms, thus increasing the value of delaying investment decisions. Exiting VC investments is examined, in regard to the market conditions that affect the decision to go public, whether reputation affects the decision to go public, why venture capitalists distribute shares, the performance of VC-backed firms, and the future of the VC cycle. Exiting investments affects every aspect of the investment cycle. Venture capitalists add value to the firms in which they invest. The VC cycle is a solution to information and inventive problems. (TNM)
Compares the organization of regional economies, focusing on Silicon Valley's thriving regional network-based system and Route 128's declining independent firm-based system. The history of California's Silicon Valley and Massachusetts' Route 128 as centers of innovation in the electronics indistry is traced since the 1970s to show how their network organization contributed to their ability to adapt to international competition. Both regions faced crises in the 1980s, when the minicomputers produced in Route 128 were replaced by personal computers, and Japanese competitors took over Silicon Valley's market for semiconductor memory. However, while corporations in the Route 128 region operated by internalization, using policies of secrecy and company loyalty to guard innovation, Silicon Valley fully utilized horizontal communication and open labor markets in addition to policies of fierce competition among firms. As a result, and despite mounting competition, Silicon Valley generated triple the number of new jobs between 1975 and 1990, and the market value of its firms increased $25 billion from 1986 to 1990 while Route 128 firms increased only $1 billion for the same time period. From analysis of these regions, it is clear that innovation should be a collective process, most successful when institutional and social boundaries dividing firms are broken down. A thriving regional economy depends not just on the initiative of individual entrepreneurs, but on an embedded network of social, technical, and commercial relationships between firms and external organizations. With increasingly fragmented markets, regional interdependencies rely on consistently renewed formal and informal relationships, as well as public funding for education, research, and training. Local industrial systems built on regional networks tend to be more flexible and technologically dynamic than do hierarchical, independent firm-based systems in which innovation is isolated within the boundaries of corporations. (CJC)
This paper develops a simple model that shows how a country can endogenously become differentiated into an industrialized "core" and an agricultural "periphery. " In order to realize scale economies while minimizing transport costs, manufacturing firms tend to locate in the region with larger demand, but the location of demand itself depends on the distribution of manufacturing. Emergence of a core-periphery pattern depends on transportation costs, economies of scale, and the share of manufacturing in national income. The study of economic geography-of the location of factors of production in space-occupies a relatively small part of standard economic analysis. International trade theory, in particular, conventionally treats nations as dimensionless points (and frequently assumes zero transportation costs between countries as well). Admittedly, models descended from von Thunen (1826) play an important role in urban studies, while Hotelling-type models of locational competition get a reasonable degree of attention in industrial organization. On the whole, however, it seems fair to say that the study of economic geography plays at best a marginal role in economic theory. On the face of it, this neglect is surprising. The facts of economic geography are surely among the most striking features of real-world economies, at least to laymen. For example, one of the most remarkable things about the United States is that in a generally sparsely populated country, much of whose land is fertile, the bulk of the population resides in a few clusters of metropolitan areas; a quarter of the inhabitants are crowded into a not especially inviting section of the East Coast. It has often been noted that nighttime satellite
Pioneering work on an important new approach to economics.
This paper discusses the prevalence of Silicon Valley-style localizations of individual manufacturing industries in the United States. A model in which localized industry-specific spillovers, natural advantages, and random chance contribute to geographic concentration motivates new indices of geographic concentration and coagglomeration. The indices contain controls that facilitate cross-industry and cross-country comparisons. The authors find almost all industries to be more concentrated than a random dart-throwing model predicts but the degree of localization is often slight. They also discuss which industries are concentrated, the geographic scope of localization, coagglomeration patterns, and other topics. Copyright 1997 by the University of Chicago.