paper - Office of the University Economist

advertisement
UNIVERSITIES AND THE TECHNOLOGY
INTENSITY OF THE LOCAL WORKFORCE:
EVIDENCE FROM METROPOLITAN AND
MICROPOLITAN AREAS
A report from the Productivity and Prosperity Project (P3), an
initiative supported by the Office of the University Economist
November 2009
Kent Hill, Ph.D.
Research Professor, Department of Economics,
and Center for Competitiveness and Prosperity Research
Dennis Hoffman, Ph.D.
Professor of Economics, University Economist,
and Director, L. William Seidman Research Institute
Timothy D. Hogan, Ph.D.
Professor Emeritus, Department of Economics,
and Research Associate, Center for Competitiveness and Prosperity Research
Eva Madly, M.S. Economics
Research Economist, L. William Seidman Research Institute
Center for Competitiveness and Prosperity Research
L. William Seidman Research Institute
W. P. Carey School of Business
Arizona State University
Box 874011
Tempe, Arizona 85287-4011
(480) 965-5362
FAX: (480) 965-5458
EMAIL: Dennis.Hoffman@asu.edu
www.wpcarey.asu.edu/seid
Over the past three decades, interest has increased in providing an answer to the question of how
research universities shape the economies of the regions in which they are located. The BayhDole Act in 1980 allowed universities, nonprofits, and small businesses to control the inventions
and other intellectual property arising from federally funded research. Since then, universities
themselves have encouraged the commercialization and local application of their research as a
way of generating new revenues from the licensing of university-held patents. State and local
governments now view universities not only as providers of higher education but as potentially
valuable economic development assets.
On a more academic level, economic geographers and growth theorists have also shown interest
in the effects that research universities have on local economies. For geographers, the synergy
between the university and the business community in the production and transfer of knowledge
offers a prime example of the type of economic activity that benefits from the concentration of
resources in localized areas. For growth theorists, the many examples of the localization of the
economic rewards from knowledge developed at universities provides evidence of the selfreinforcing nature of growth from ideas resulting from the interaction between academia and
industry.
The Office of the University Economist at Arizona State University (ASU) has been working
with the Commission on Innovation, Competitiveness, and Economic Prosperity (CICEP) of the
Association of Public and Land-Grant Universities to conceptualize and quantify how
universities contribute to their local economy. (For more information on this effort, see
http://economist.asu.edu/.)
A particular emphasis is the impact of research universities on the economic prosperity of
regions. The initial phase of this line of research, discussed in this paper, focuses on the
influence of research universities on the concentration of high-technology activities in the
private-sector economy of a region. This paper specifically looks at the link between the
presence of research universities and employment in technology-intensive industries in more
than 900 metropolitan and micropolitan areas across the United States.
Numerous models of the relationship between high-technology employment and the presence of
research universities are examined in this paper. High-tech employment is alternatively defined
as actual employment, per capita employment, and the share of total employment. High
technology is defined three ways: total high tech, high-tech services, and scientific research and
development. The presence of research universities also is examined in multiple ways based on
the three classifications of research universities defined by the Carnegie Foundation.
In this initial phase of research, statistically significant positive relationships were found between
variables measuring the presence of research universities and the technological intensity in U.S.
metropolitan/micropolitan areas. Universities with “very high research activity” are more
strongly related to employment in technology-intensive industries than universities with lesser
amounts of research activity.
1
BACKGROUND
The contribution of technology innovations induced by research and development (R&D) to
economic growth is well documented. As noted by Lane (2009), citing Jorgenson (2008) et al., a
substantial portion of productivity gains in recent years in the United States may be traced to
science investments. Yet, as Lane noted in citing Japanese and Swedish examples, rates of return
on investments may not be observed in all circumstances. One focus in this line of research is to
determine accurate rate-of-return measurements for investments made by universities in
developing research faculties and degree programs designed to foster the enhancement of
knowledge and skills in the workforce.
An important issue for policymakers is to measure the benefits to regions from universities that
produce quality undergraduate and graduate students and that maintain faculties who are active
researchers. Many regions of the United States are contemplating transitions from an emphasis
on old-line manufacturing employment bases to alternatives requiring more knowledge, training,
and skills. Sunbelt states that have reaped steady benefits from the north-to-south migration in
recent decades may also be seeking more stable economic base industries by diversifying heavily
growth-reliant economies with knowledge-based activities. The quality of the research
infrastructure maintained by the region can assist in this transition and universities can provide a
key component of that infrastructure.
Regions that nurture research and innovation conceptually should serve as a magnet for
inventors, provide a business climate conducive to the development of a technology-based
workforce, and ultimately attain superior economic growth due to its productivity-enhancing
technological innovations. By this line of reasoning, universities serve as catalysts for research
and innovation, and regions with research universities should reap the benefits predicted
throughout the literature.
Empirically, however, the evidence is weak regarding exactly how universities provide benefits
to surrounding communities and the types of investments that are required of local communities
to yield these benefits. Issues also remain regarding the distribution of university benefits: to
university graduates, local residents, local businesses, and others.
Resolving these questions will help citizens understand the return from their investments of tax
dollars, tuition dollars, or philanthropic dollars to local universities. A starting point is to
compare regions with research universities to similar regions that do not have a research
university.
The Case for the Relationship Between a Research University and the Local Economy
The low cost of and technological improvements to telecommunications have led some to predict
that physical proximity will become a less important attribute to economic development. High
population mobility also may contribute to a spatial diversification of economic activity. Thus, it
is easy to be skeptical of the claim that spatial proximity to a university can have great value for
a commercial firm. Economic geographers, however, remain impressed by and continue to
document the power of agglomeration economies in some industries, including those that create
knowledge.
2
As far as knowledge itself is concerned, if it is codified (written down and transmitted through
formulas or text), the cost of transferring knowledge is negligible. Proximity to an original
source of codified knowledge is unimportant. Some knowledge, however, is highly complex and
difficult to write down. The transfer of tacit knowledge requires frequent face-to-face interaction
with the knowledge source. Geographic proximity is critical for transferring tacit knowledge.
Darby and Zucker (2003) have argued that “metamorphic” innovations—those associated with
the creation of new industries or the radical technological transformation of an existing
industry—typically are driven by breakthrough discoveries in science and engineering. Examples
include integrated circuits, recombinant DNA, and nanotechnology. These kinds of discoveries
are not well understood initially and cannot be codified. In the beginning, the new knowledge is
largely tacit, and it is difficult for anyone other than the discoverer to see commercial value in
the findings. Transfer and application to industry require bench-level relationships between
industry scientists and the pioneering scientists. If the scientist making the metamorphic
discovery has a university appointment that he wishes to maintain and does not want to commute
long distances, he will serve as a fixed factor that determines the location of firms entering the
market to develop the new technology.
The period during which discovering scientists play a major role in transferring new knowledge
to industry may last only 10-to-15 years. Eventually, scientific findings become codified and can
be learned by graduate students at any major research university. But once an industry has been
established in a given location, agglomeration economies associated with the rise of specialized
suppliers or markets for specialized labor may serve to lock in an industry’s location. In this way,
the initial geographic residences of path-breaking researchers have a long-term effect on industry
location.
Research universities also generate local economic impacts through their graduate programs.
Availability of scientific labor is an important concern for managers of industrial laboratories,
and they may choose to site a lab in an area if local universities can provide a steady supply of
highly qualified science and engineering graduates [Malecki (1987), and Malecki and Bradbury
(1992)]. Because of a variety of local attachments people develop while in school, young
professionals often prefer to remain in the vicinity of their graduate school, especially if that
school is located in a large urban area. As shown by Hill (2006), there is a significant positive
correlation between graduate degrees awarded in a state and the share of the state’s population
with a graduate degree. This correlation is especially apparent after adjusting for weather, which
is known to be an important factor in people’s migration decisions and which also correlates with
the location of major universities.
Evidence of Local Impacts
Evidence of local economic impacts from research universities comes from a variety of sources:
case studies of local industries born from the ideas of university scientists, analysis of data on
patent citations which make it possible to trace the spatial flow of new knowledge, and use of
econometric techniques to quantify the relationship between the level of economic activity in an
area and the presence of a research university. The evidence shows conclusively that university
research programs can have significant local economic impacts. But these impacts are highly
skewed across universities and, for the average university, are modest in size.
3
Silicon Valley and Route 128
The most highly celebrated cases of local economic development stimulated by university
research are the electronics clusters in Silicon Valley, California, and the Route 128 beltway
around Boston, Massachusetts. Agglomeration economies associated with knowledge spillovers
and thick markets for specialized suppliers have played an important role in reinforcing industry
growth in these areas. But the conventional wisdom is that the initial reason the industry took
root in these particular locations is to be close to researchers and research facilities at Stanford
University and the Massachusetts Institute of Technology. Local firms also were aided by a
readily available supply of electrical and computer engineers graduating from nearby schools.
See Dorfman (1983), Rogers and Larsen (1984), and Saxenian (1996) for a review of the origins
of the electronics industry in Silicon Valley and Route 128.
The Biotech Industry
Biotechnology offers the most recent example of an important new industry built directly on
basic scientific research in which commercial firms are known to have close ties to universitybased scientists. A single scientific moment defines the beginning of the industry—the 1973
discovery by Stanford professor Stanley Cohen and University of California-San Francisco
professor Herbert Boyer of the basic technique for recombinant DNA. Techniques for genetic
engineering would eventually become standardized and widely known. But for 15 years
following the discovery, knowledge of how to identify promising gene sequences and even the
skills of gene transfer were held by a small group of discovering scientists and their co-workers.
Knowledge of the techniques was difficult to transfer because of its complexity and tacitness.
Commercial development required frequent face-to-face contact with discovering scientists.
Since many of these scientists were academics who were unwilling to leave university
appointments, their location often served to determine the location of commercial firms.
Zucker, Darby, and Brewer (1998) were among the first to systematically test for and find a
geographic coincidence between new biotechnology firms and university scientists who made
early contributions to gene sequencing. Zucker, Darby, and Armstrong (1998, 2002) showed that
the most successful biotech firms were those in which university scientists had a financial
interest and remained actively involved throughout the development phase of the product.
Analysis of Patent Citations
The case for linkages between university research and local economic development ultimately
hinges on the argument that knowledge created at universities tends to stay in the local area.
Jaffee and Trajtenberg (2002) used a clever way of finding a “paper trail” with which to track
invisible knowledge flows. U.S. patent records contain highly detailed information on the
identity and location of the inventor and references or citations to previous patents. The citations
make it possible to study knowledge spillovers by tracing the connections between inventors. In
one particular analysis of U.S. patents originating in 1980 and their citations through 1989,
Jaffee, Trajtenberg, and Henderson (1993) found that after excluding self-citations, the frequency
that a university-owned patent is cited in the same metropolitan area as the university is six times
higher than what would be expected given the existing distribution of technical activity.
4
Location of Corporate R&D
A number of authors have used econometric techniques to measure the effect of university
research on the level of local corporate R&D activity. Industry labs directly promote local
economic development by providing high-paying jobs for scientists and technical workers. They
may also generate competitive advantages for local producers who make use of the innovations
coming out of these labs.
Jaffe (1989) used state-level data to relate the number of patents assigned to corporations in a
state to industry R&D expenditures and university research in the state. Jaffee found a large and
statistically significant effect of university research on corporate patent activity. His results
indicate that the most important way university research stimulates corporate patent awards is by
inducing additional corporate R&D spending.
Anselin, Varga, and Acs (1997) employed the same model as Jaffee to represent the relationships
between corporate innovation, corporate R&D, and university research. Counts of innovation are
measured using data from the Small Business Administration on the number of new hightechnology products introduced into the U.S. market in 1982. Anselin, et al. were able to sharpen
the geographic focus of Jaffee’s analysis by using professional employment in private high-tech
research laboratories as a proxy for corporate R&D activity. The employment data were
aggregated to the metropolitan area level. The findings of Anselin, et al. generally confirm the
earlier results of Jaffee. University research has a positive and significant effect on local
corporate innovation, both directly and through its effect on private R&D activity.
Bania, Calkins, and Dalenberg (1992) examined the role of university research in determining
the geographic distribution of industry R&D activity. R&D activity is measured by the number
of Ph.D.s working in industry labs. The geographic unit of analysis is the metropolitan area and
the data are for 1986. Bania, et al. find that university research expenditures have a statistically
significant effect on R&D doctorate employment. The estimated effects are fairly small,
however.
Local Area Income, Employment, and Productivity
Beeson and Montgomery (1993) used Census data for 218 metropolitan areas to measure the
contribution of university variables to alternative measures of labor market activity, including
average individual income in 1980 and employment growth over the periods 1975-1980 and
1980-1989. University characteristics included in the regressions were university R&D funding,
university quality (as measured by the number of highly rated science and engineering
programs), the total number of bachelor’s degrees awarded, and the percentage of degrees
awarded in science and engineering. The authors also included other variables that might affect
the local labor market such as weather, crime rates, taxes, and the size of the metro area
population.
Beeson and Montgomery found that none of the university variables were statistically significant
in explaining average income, at least in random effects models that allow for metro areaspecific error components. The estimated coefficients also were small in value. More significant
relationships were found between university characteristics and metro area employment growth.
For both time periods considered, the authors were able to reject the hypothesis of no
5
relationship between university variables and the rate of employment growth. The estimated
coefficients also were significant in size.
Goldstein and Renault (2004) used annual data on 312 metropolitan areas from 1968 through
1998 to test for the effect of universities on local economic development. Economic development
was measured by an index that indicates whether and in what direction relative earnings in an
area changed over time. The presence of a university was measured by the existence of a top-50
research university, the size of university R&D, the number of degrees awarded, and the number
of patents assigned to universities within the metro area. Goldstein and Renault find that over the
period 1969-1986, universities had no effect on an area’s relative earnings. However, the
presence of universities did become significant in the 1986-1998 period. The authors attributed
this result to the often-cited claim that economic activity in the United States has become more
knowledge-based in recent decades.
In a notable international study, Andersson, Quigley, and Wilhelmsson (2009) took advantage of
a natural experiment in which the government of Sweden undertook a spatial decentralization of
higher education in the country beginning in the late 1980s. Using time series data for 284 local
civil divisions, the authors found systematic evidence that expansion of university research in a
community as measured by the size of the university research staff led to increases in output per
worker in the local area. These productivity effects were highly localized. Approximately 40
percent of the cumulative gain in productivity occurred within 10 kilometers of the institution.
CONDITIONING FACTORS
Two universities with research programs that are similar in scale and quality may have very
different local economic impacts. MIT and Harvard University have had huge, documented
effects on the Boston area economy. However, Johns Hopkins University, which is routinely
among the largest recipients of federal government research funds, has failed to stimulate
significant high-tech production in the Baltimore area [Feldman and Desrochers (2003)]. A high
level of research activity is not by itself a sufficient condition for a university to have large
impacts on jobs and incomes in the local economy. Certain complementary factors need to be
present if a university is to significantly affect the local economy.
Quality of University Research and Graduate Programs
Universities with the greatest local economic impacts are those with high-quality research and
graduate programs. As argued by Zucker, Darby, and Brewer (1998) and Darby and Zucker
(2003), the most compelling reason for new firms to locate near a university is to facilitate tacit
knowledge transfer from faculty who are on the cutting edge of scientific breakthroughs. It is
only these star researchers who have the power to determine firm location.
Another reason why star academic researchers are the ones most likely to be successful in
attracting new industry to an area is what Audretsch and Stephan (1996) referred to as “drawing
power.” University researchers with a national reputation or researchers from an eminent
university serve as a signal of quality which helps to attract resources such as venture capital,
management, and technical workers that are necessary to start up new companies.
6
Finally, Malecki (1987) argued that availability of science and engineering workers is an
important determinant of the location of industrial R&D facilities. But he noted that firms are
particular about the institutions they rely on for new researchers. Especially among large firms,
only the best graduate programs are an attracting factor.
Agglomeration and Research Networks
Agglomeration economies are known to be an important factor in the production of knowledge.
Spatial concentration of research activity promotes the development of markets for specialized
suppliers of materials, testing equipment, and even legal services. Agglomeration also helps to
support informal channels of knowledge transfer. University research will be more productive
and more likely to influence local economic activity if it takes place in an area with an existing
concentration of corporate research activity and high-tech production. Studies of the biotech
industry, for example, have found that university faculty who collaborate with industry in
commercial ventures are more likely to do so with local firms if the industry has a significant
local presence. Otherwise, faculty involvement will be long distance [Audretsch and Stephan
(1996)].
In a case study of the Cleveland area, Fogarty and Sinha (1999) found that technology developed
in local universities did not generate local jobs and incomes but instead was quickly diffused to
Japan, California, and Texas. The authors attribute this to the fact that the Cleveland economy is
heavily oriented toward mature industries and lacks the local research networks necessary to
develop university technologies. Fogarty and Sinha measure the extent of local research
networks by tracing the direct and indirect citations of university patents and calculating the
tendency for subsequent innovations to be localized. Metropolitan areas with the strongest local
R&D networks are San Francisco, New York, Boston, and Los Angeles. Areas with much
weaker networks are Washington-Baltimore, Philadelphia, Chicago, Detroit, and Cleveland.
Varga (2000) provided an exacting test of the importance of agglomeration as a factor
conditioning the size of the effect of university research on local innovative activity. He used an
econometric model to explain variations across metropolitan areas in counts of new product
innovations made by the Small Business Administration for 1982. Agglomeration effects are
identified using the amount of high-tech employment in a metro area. Using interactive
variables, Varga found that university research led to a significant number of local area
innovations only when high-tech employment was at least 160,000 workers.
University Policy and Culture
University policy toward the commercialization of research can have important effects on the
extent to which faculty engage in and develop commercially relevant research. In an attempt to
raise what are generally considered to be disappointing financial returns from resources used to
promote technology transfer, more universities are making use of equity arrangements when
licensing university inventions. When faculty have a financial interest in the performance of the
firm that licenses their research, they are more likely to assist the firm in product development.
Licensing firms believe that university equity positions confer a kind of halo effect that helps
them secure venture capital funding. Data analysis indeed shows that universities that are
permitted to take an equity position in companies that license their research have 70 percent
more start-ups than universities that cannot.
7
The general orientation of a university toward local economic development also seems to
influence the impact its research has on the local economy. Based on an analysis of eight cases of
U.S. universities and their local areas, Paytas, et al. (2004) found that universities with the largest
local economic impacts were those that were broadly engaged with local industry clusters—
providing not only research services but business, marketing, legal services, and workforce
education. The most engaged universities were those with a commitment to local economic
development across many academic units including colleges of business, engineering, law,
medicine, and public policy.
Nature of Local Industry
Paytas, et al. (2004) found through their case study approach that the nature of local industry was
as important as were the characteristics of the university. Established clusters with mature
products and processes were less able to absorb new technologies generated at universities.
Innovation was likely to flow out of the region. Universities were more successful in generating
local impacts when interacting with young, emerging clusters.
In an econometric study of the effect of university research on local economic growth, Bania,
Eberts, and Fogarty (1993) examined the number of start-ups of new manufacturing firms in 25
large metropolitan areas over the period from 1976 through 1978. Bania, et al. explained firm
births using an econometric model with both traditional business climate variables, such as labor
costs and taxes, and variables relating to knowledge infrastructure, including university research
expenditures and the percentage of employed workers who were scientists or engineers. They
found mixed evidence for the effect of university research on new company start-ups. University
research had a positive and significant effect on new business formation in the electronics
industry but was statistically insignificant for five other manufacturing industries. The authors
conjectured that the reason their strongest findings were for electronics was because their period
of study coincided with the emergence of the electronics industry. They offered these results as
evidence supporting the theory that the local impacts of university research are greatest in newly
emerging industries.
EMPIRICAL APPROACH
A vast literature describes the relationships between knowledge/discovery and economic
prosperity. If investments by universities are important to regions, the impact should be
measurable by examining the economic geography of the region, such as the number and
concentration of inventors, technology in the workforce, the growth or in-migration of
technology-based firms, and the pace of prosperity gains. Such university impacts have not been
well measured to date, at least to the extent and with the precision proposed in the research
agenda of ASU’s Office of the University Economist.
Outside academia, papers designed to extol the virtues of universities, produced largely by
marketing and communications organizations, measure the value of research universities in the
same way economists measure the value of a newly constructed prison—by counting direct
payroll jobs and associated locally induced effects using input-output models. In contrast, the
work by ASU is designed to provide tangible results for those seeking accurate measures of the
impact of universities on the economic vitality of regional economies.
8
Numerous empirical questions could be addressed to inform this broad line of inquiry. In this
paper, the focus is to investigate the empirical question of the role played by research universities
in determining the technology intensity of a region’s workforce. Since technology and skills are
associated with higher productivity, increases in the technological skills of the workforce due to
the presence of a research university should result in higher standards of living.
The foremost challenge in the investigation of this empirical question is to control for factors that
affect high-technology employment other than the presence of a research university. The
literature notes that, at a minimum, a region’s amenities, climate, natural attractiveness, culture,
fiscal stability and predictability, and size are factors that can influence economic development.
An area that is perceived favorably on such factors may attract educated and skilled workers and
the companies that employ them regardless of the presence of a local university.
Modeling the Technology Intensity of the Workforce
The empirical model presented in this paper is based in part on the work of Jaffe (1989).
Drawing from the knowledge production framework, a fundamental component in the production
process is the flow of the knowledge created by university research from universities to industry.
The basic hypothesis is that this production process attracts industry to locate in proximity to the
universities to reap the direct effects of the research underway and to benefit from some key byproducts of that research activity. The additional benefits include the availability of faculty talent
in consulting, lecture, and contract research projects and the availability of students, lab
assistants, and research staff employed at the university but who could work at businesses in
close regional proximity to the university. Indeed, there may be advantages and opportunities for
workers to share joint appointments at universities and high-tech businesses. Thus, the model is
consistent with the notion that while knowledge can easily diffuse and benefit industries located
far away from the source of its creation, the by-products of this knowledge creation process will
accrue disproportionately to those businesses in close proximity to the source.
The amount and share of employment in high-technology industries in any particular
metropolitan area may be influenced by many factors that need to be controlled for in the
empirical specification. The control variables include measures of metro size, region of the
country, cost of living, wealth or income, and the educational attainment and skill levels of the
residents. The empirical models are then designed to measure whether the technological intensity
of the workforce is influenced by the geographic proximity to research universities and by the
volume of research dollars expended by all the universities in the metropolitan area.
Data and Variable Definitions
The unit of analysis is the 940 metropolitan and micropolitan statistical areas located in the 50
states and the District of Columbia, as currently defined by the U.S. Office of Management and
Budget. Incomplete data were available for six small micropolitan areas, so 934 observations
were used. All models are cross-sectional in nature. The time period for the analysis is 2003
through 2006. To eliminate any short-term aberrations in the data, the average over the four-year
period was calculated and used for each variable with annual values.
The dependent variables are alternative measures of high-technology employment. Technology
intensity is alternatively defined as total high-tech employment, employment in high-tech
9
services, or employment in the scientific R&D industry group. Each of these three definitions is
alternatively measured as actual employment, per capita employment, and the high-technology
share of total employment.
The North American Industry Classification System (NAICS) was used to identify hightechnology activities. High-technology services includes the following industry groups: 5112–
Software Publishers; 5179–Other Telecommunications; 5182–Data Processing, Hosting, and
Related; 5413–Architectural and Engineering Services; 5415–Computer Systems Design and
Related; and 5417–Scientific R&D. Total high-tech employment includes these services industry
groups and the following high-tech manufacturing industry groups and industries: 3254–
Pharmaceuticals and Medicine; 3341–Computers and Peripherals; 3342–Communications
Equipment; 3343–Audio and Video Equipment; 3344–Semiconductors and Other Electronics;
3345–Navigational, Measuring, Electromedical, and Control Instruments; 3364–Aerospace;
333295–Semiconductor Machinery; 333314–Optical Instruments and Lenses; and 333315–
Photographic and Photocopying Equipment.
Employment was obtained from Metro Business Patterns, produced by the U.S. Census Bureau.
Undisclosed employment was estimated by multiplying the number of establishments in each of
the several employment ranges presented by the Census Bureau by the national average
employment per establishment in each range and summing the results.
The number and categorization of research universities in each metropolitan/micropolitan area
were obtained from the National Center for Education Statistics IPEDS (Integrated
Postsecondary Data System) Data Center. This dataset includes the classification of each
university produced by the Carnegie Foundation for the Advancement of Teaching in 2005 and
the city in which each university is located. The latter was used to identify the metropolitan or
micropolitan area in which each university resides. Research universities (defined as universities
that awarded at least 20 doctorates in 2003-04) are split into three categories based on the
Carnegie Foundation definition—those with very high research activity, those with high research
activity, and other doctoral/research universities. Universities that do not grant doctorates were
not included in this analysis, though any R&D spending by such institutions are included in the
academic R&D variable discussed below.
Based on the Carnegie Foundation data, three mutually exclusive independent variables were
constructed to measure university presence in a metropolitan/micropolitan area and in its
surrounding area, defined as within a radius of 60 miles of the metro/micro area: the total
number of universities with very high research activity; the total number of universities with high
research activity; and the total number of other research universities. A single composite variable
also was constructed by weighting the universities with very high research activity by a value of
three, the high-level universities by two, and the other research universities by a value of one.
Total university R&D spending also was included as a direct measure of the level of research
activity at universities by metro/micro area. Academic R&D expenditures for each university
were obtained from the National Science Foundation. The R&D spending of all universities
located in each metropolitan/micropolitan area was aggregated to obtain total academic R&D
expenditures at the metro/micro level.
10
The models also incorporate a set of variables meant to control for other factors that may affect
the technology intensity of a region—population, total employment, personal income (total or
per capita), a set of binary regional indicators, cost of living, and educational attainment. The full
list of variables is presented in Table 1. Descriptive statistics for each variable are displayed in
Table 2.
EMPIRICAL RESULTS
Models of the Labor Force Shares of Technology Workers
Tables 3 through 12 report the results of 10 variants of the regression analysis in which the
dependent variable is expressed as a share of total employment. In each of these 10 tables, three
alternative dependent variables were specified: total high-technology employment, employment
in high-technology service (nonmanufacturing) industries, and employment in scientific R&D
industries. In each specification, the control variables include population, regional dummy
variables, and measures of cost of living, educational attainment, and income per person.
The two independent variables of key interest are university presence (within 60 miles of each
metro/micro area) and the amount of academic R&D spending (within each metro/micro area).
Five of the tables differ only in the way that university presence is measured.
Table 3 contains results based on the presence of universities with very high research activity.
Results suggest that having one or more of these very-high-intensity research universities within
60 miles of the metro/micro area has a statistically significant impact on the high-technology
share of the area’s total employment. The impact of the presence of one or more very-highintensity research universities on total high-tech employment (column 1 of Table 3) is an
additional 0.35 percent in the share of technology workers in the workforce. For high-technology
service employment, the impact is 0.18 percent; for R&D employment, it is 0.097 percent.
Similarly, an additional $100 million in academic research spending is associated with an
additional 0.36 percent in the share of all high-technology employment, 0.27 percent more in the
share of high-technology service employment, and a 0.12 percent addition to the share of R&D
employment.
While these shares may appear small, high-technology employment on average was just 2.55
percent of total employment in the metro/micro areas, as seen in Table 2. Thus, the coefficient of
the very-high-intensity research university variable (0.35 percent) represents a 13.7 percent
increase in the share of all high-technology employment. Similarly, the effect is an 11.5 percent
increase in the share of high-technology service employment and a 31.6 percent increase in the
share of R&D employment. Each of these increases was measured after controlling for various
area-specific attributes.
Using the same approach, an additional $100 million in university R&D spending results in
percentage increases of 14.1, 17.3, and 39.1 in the respective employment shares of all hightechnology workers, high-technology service workers, and R&D workers. Given that the average
annual level of university R&D spending across the entire sample of metro/micro areas was
about $28 million, an increment of $100 million is a large change, but the difference between
having a research university and not having a research university located in the area is also very
11
big. Since all of these are linear models, the implication of the choice of a smaller increment is
easy to calculate—for example, an additional $10 million would imply a 1.4 percent increase in
overall technological intensity.
Table 4 also focuses on universities with very high research activity, but expresses the dependent
variables as shares of the overall population. The results are very similar to those based on
employment shares, though the t-scores on the university presence variable and the academic
R&D spending variable are somewhat higher, as is the adjusted R-squared. The presence of one
or more very-high-intensity research universities is associated with what appears to be a small
increment to the overall share of per capita technology workers, but it translates into a 14.9
percent increase in the share. Similarly, the share of high-technology service workers per capita
is boosted by 12.8 percent and the share of R&D workers per capita surges 39 percent. The small
statistically significant coefficients on the university research spending variable represent an 18.1
percent, 22.8 percent, and 46.8 percent increase in the shares of total, services, and R&D workers
per capita as a result of an additional $100 million in university R&D expenditures.
Tables 5 and 6 examine the effects of the presence of universities with high research activity and
other doctoral/research universities respectively. These results should be compared to those in
Table 3, as the dependent variables all were measured as the share of total employment.
Generally, the impacts of these research universities are smaller than those of universities with
very high research activity. The magnitude of difference is small between institutions with very
high and high research activity, and also is small with other research universities based on total
high-technology employment. However, the research universities with less than high levels of
research activity have an insignificant effect on the high-tech services and R&D employment
shares. The impact of academic research spending remains high in these models since this
variable measures total research dollars at all area universities—including those in the very high
research classification.
In Table 7, all three variants of university presence were included as explanatory variables. The
results generally suggest that universities with very high research activity have greater impacts
on the technology employment shares than do the other research universities, though the
estimates are not as statistically significant. Table 8 employs a single composite variable for
university presence: the doctoral, high, and very high classifications receive values of 1, 2, and 3
respectively, with a value of 0 assigned to areas without any research universities. This weighted
variable is strongly significant but considerably smaller in size than the variable estimates in
Table 3.
The sets of models presented in Tables 9 through 12 all define university presence based on
very-high-intensity research universities but are based on different subsamples of the
metro/micro areas (less than 5 million residents, less than 1 million residents, less than 500,000
residents, and a population of less than 250,000 respectively). The results indicate that the
presence of very-high-intensity research universities bolsters the high-technology employment
share, though the strength of the effect wanes as larger metro areas are stripped from the sample.
There is no support for the relationship in areas with populations less than 250,000. The impact
of university research spending on the technology workforce shares is very strong except in areas
of less than 250,000 residents and is especially strong in the less than 1 million and less than
12
500,000 specifications, with the percentage increase in the high-technology employment share
two-to-three times higher than observed in the overall sample.
Models Based on the Total Number of High-Technology Workers in Each Region
The results shown in Tables 13 through 21 are designed to reveal the robustness of results
observed in Tables 3 through 12. They are based on specifications in which the dependent
variables are expressed as actual high-technology employment rather than the share of total
employment. Otherwise, the model specification in Table 13 is nearly identical to that in Table 3,
with Table 14 comparable to Table 5, etc. The set of control variables differs in two ways: a total
employment variable is added and aggregate regional income is used to measure the wealth of
the region rather than per capita income.
Table 13 contains models examining the effect of the presence of one or more very high research
universities within a 60 mile radius on the number of technology workers in the metro/micro
area’s workforce. For total technology workers, the results support the notion that proximity to
very high research universities and research dollars bolster the number of technology workers,
although the impact of additional university research dollars is not statistically significant in the
service technology or R&D technology specifications. While the numbers of additional
technology workers associated with the presence of research universities appears small, it
represents a rather substantial share of the workforce, on average. For total technology workers,
results suggest that the presence of one or more very high research universities is associated with
a 19.9 percent increase in the number of total technology workers, a 19.1 percent increase in the
number of service technology workers, and a 34.8 percent increase in the number of R&D
technology workers. In comparison, with the “share” specification results in Table 3, the
magnitude of these impacts is slightly greater. The impact of an additional $100 million in
university research expenditures increases total technology workers by about 17.9 percent, while
the impacts of university research spending on service and R&D workers in the “levels”
specification is small and statistically insignificant.
Tables 14 and 15 present the results of similar models for the other categories of research
universities, and Tables 16 and 17 offer results for models in which all types of research
universities are included simultaneously. As with the results of the “share” models, the impact of
universities with less than very high research activity is not as significant on the size of hightechnology employment as are the effects of very-high-intensity research universities. The
impact of university research spending in these specifications remains mixed, with overall
technology jobs affected but not so with high-tech service and R&D employment.
In contrast with the “share” models, the overall significance of research universities on the
number of technology workers appears to be influenced by the population size of the
metro/micro areas used to examine the impacts. Based on the subsamples of less than 1 million
and less than 500,000 residents (Tables 19 and 20), support for the basic hypothesis that
proximity to research universities matter for the technology intensity of the workforce is very
strong. In the 1 million and under results, the impact on the technology workforce due to the
presence of one or more research universities is 14.4 percent, 17.1 percent, and 46.5 percent on
total, services, and R&D technology workers respectively. The impacts of an additional $100
million in research expenditures are greater, with the number of technology workers expanding
13
51.1 percent, 66.4 percent, and 138.2 percent respectively. In the 500,000 and under subsample,
the impacts on the technology workforce due to the presence of one or more research universities
is 24.7 percent, 21.3 percent, and 63.1 percent on total, services, and R&D technology workers
respectively. The impacts of an additional $100 million in university research spending are even
bigger, with the numbers of technology workers expanding 77.8 percent, 102.9 percent, and
190.6 percent respectively.
With the potential for simultaneity between the dependent variables measuring technological
intensity of the workforce and the personal income variable in the models, versions of the models
excluding the income variable were also estimated. The signs, statistical significance, and
magnitudes of the estimated coefficients for the research university variables were little changed
in these alternate specifications.
CONCLUSION
The link between the presence of research universities and the high-technology intensity of a
region’s economy is timely given that regions with technology/knowledge-based economies
recently have been more successful in economic development. However, this is only one of the
dimensions of interest in examining how research universities shape the economies of the
regions in which they are located.
This paper presents results of the initial analysis of the relationship between research universities
and technology intensity as measured by high-tech employment. The models specified in terms
of the high-tech share of total employment (and population) demonstrate a statistically
significant positive relationship for the variable measuring research university presence. The
relationship is stronger among universities with very high research activity (based on the
Carnegie Foundation classification) than for the two other categories of research universities.
Though small in absolute size, the estimated coefficients indicate substantially greater high-tech
employment shares in metro/micro areas with one or more very-high-intensity research
universities. The results of the models specified in terms of actual high-tech employment
generally support the results of the “share” models.
These results are consistent with previous research that found: (1) links between universities and
the location of high-tech plants (Glasmeier 1991); and (2) connections between university
activities and technology intensity measured in terms of occupation-based measures, such as
employment shares of scientists/engineers/etc. (Beeson and Montgomery 1993).
Models based on different subsamples of the metro areas by size indicate that the strongest
relationship between research universities and technology intensity occurs in larger metro areas;
no relationship is seen in less populous areas (250,000 or fewer residents). These results appear
to be contrary to the findings of Goldstein and Drucker (2006) that the effects of universities are
particularly important in small- and medium-sized metro areas, but Goldstein and Drucker
measured growth in average earnings rather than technology intensity. Future work on the
influence of research universities looking at different characteristics of regional
economies/regional economic performance will address this issue more directly.
14
The analysis presented in this paper should be viewed as an initial exploratory look at the link
between research universities and the technological intensity of regions. Future research on the
general topic of the influence of research universities on the growth and prosperity of regions
will include more sophisticated models of the connections between research universities and the
technological intensity of a region.
Another direction for further work is revealed by the results of the existing simple models that
have included two variables related to research universities, one that measures the presence of
one or more research universities in a region and one that measures R&D spending by all
universities in the metro/micro area. The fact that the results generally demonstrate statistically
significant positive relationships between both of these measures and technology intensity
implies that research universities have effects on regional economies in addition to those
associated with their research spending. Goldstein, Meier, and Luger (1995) identified eight
research university functions, and other researchers have already begun to look at the importance
of factors other than the amount of R&D spending. (See Drucker and Goldstein (2007) for a
summary of this research.)
15
TABLE 1
DATA AND VARIABLE DEFINITIONS
Data
Academic Research
and Development
Expenditures
Cost of Living and
Bachelor’s Degree
Employment
Metropolitan and
Micropolitan Areas
Personal Income
Population
Research Universities
Regions
Variable Name
Bachelor’s Degree
Cost of Living
Doctoral Research
Universities
Far West
Great Lakes
High Research
Universities
High-Tech
Employment
High-Tech
Employment Share
High-Tech Services
Employment
High-Tech Services
Employment Share
Mideast
New England
Per Capita High-Tech
Employment
Per Capita High-Tech
Services Employment
Per Capita Personal
Income
Definition and Source
The average from 2003 through 2006 as reported by the National Science
Foundation, Division of Science Resources Statistics
(http://www.nsf.gov/statistics/showpub.cfm?TopID=8&SubID=1)
From Sperling’s Best Places (http://www.bestplaces.net/)
All employment variables are expressed as the average from 2003 through
2006, as reported by the U. S. Department of Commerce, Census Bureau, in
Metro Business Patterns (http://www.census.gov/econ/cbp/index.html)
As defined by the U.S. Office of Management and Budget
(http://www.whitehouse.gov/omb/assets/omb/bulletins/fy2009/09-01.pdf)
The average from 2003 through 2006 as reported by the U. S. Department of
Commerce, Bureau of Economic Analysis (BEA;
http://www.bea.gov/regional/index.htm#state)
The average from 2003 through 2006 as reported by the U. S. Department of
Commerce, Census Bureau (obtained from the BEA website)
As defined in 2005 by the Carnegie Foundation for the Advancement of
Teaching; obtained from the National Center for Education Statistics
(http://nces.ed.gov/ipeds/datacenter/)
As defined by the U. S. Department of Commerce, Bureau of Economic
Analysis
Definition
Percentage of the population age 25 or older who have earned a bachelor's
degree
Cost of living index
Number of research universities other than those with very high or high
research activity, in and within 60 miles of the metro area
Alaska, California, Hawaii, Nevada, Oregon, and Washington
Illinois, Indiana, Michigan, Ohio, and Wisconsin
Number of research universities with high research activity, in and within 60
miles of the metro area
Total of high-technology services employment and high-tech manufacturing
Total high-technology employment as a percentage of total employment
High-technology services employment
Total high-technology services employment as a percentage of total
employment
Delaware, District of Columbia, Maryland, New Jersey, New York, and
Pennsylvania
Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, and
Vermont
Total high-technology employment divided by population
Total high-technology services employment divided by population
Personal income divided by population
16
TABLE 1 (CONTINUED)
DATA AND VARIABLE DEFINITIONS
Variable Name
Definition
Per Capita R&D
Employment
Personal Income
Plains
Population
R&D Employment
Scientific research and development employment divided by population
R&D Employment
Share
R&D Expenditures
Scientific research and development employment as a percentage of total
employment
Academic research and development spending expressed in 2006 dollars
Rocky Mountain
Southeast
Colorado, Idaho, Montana, Utah, and Wyoming
Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North
Carolina, South Carolina, Tennessee, Virginia, and West Virginia
Arizona, New Mexico, Oklahoma, and Texas
All employment reported in Metro Business Patterns; excludes the public sector,
farms, self-employed, and others
Number of research universities with very high research activity, in and within
60 miles of the metro area
Total number of research universities, in and within 60 miles of the metro area,
with very high research universities receiving a weight of three and high
research universities receiving a weight of two
Southwest
Total Employment
Very High Research
Universities
Weighted Research
Universities
Earnings by place of work; dividends, interest and rent; and transfer payments
Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota
Number of residents in the metro area
Scientific research and development employment
17
TABLE 2
DESCRIPTIVE STATISTICS
Variable
Bachelor’s Degree
Cost of Living
‘Doctoral’ Research Universities
Far West
Great Lakes
‘High’ Research Universities
High-Tech Employment
High-Tech Employment Share
High-Tech Services Employment
High-Tech Services Employment Share
Mideast
New England
Per Capita High-Tech Employment
Per Capita High-Tech Services Employment
Per Capita Personal Income
Per Capita R&D Employment
Personal Income
Plains
Population
R&D Employment
R&D Employment Share
R&D Expenditures
Rocky Mountain
Southeast
Southwest
Total Employment
‘Very High’ Research Universities
Weighted Research Universities
Number
of
Observations
934
934
934
934
934
934
912
911
911
911
934
934
912
911
934
694
934
934
934
694
694
934
934
934
934
912
934
934
18
Mean
0.1211
86.6
0.4454
0.0953
0.1734
0.5974
5,968
0.0255
4,099
0.0156
0.0824
0.0300
0.0094
0.0057
29,227
0.0012
1.07e+07
0.1263
294,580
919.4
0.0031
28.516
0.0557
0.3148
0.1221
119,488
0.5824
3.3878
Standard
Deviation
0.0455
19.1
0.9749
0.2938
0.3788
1.0356
26,124
0.0300
18,851
0.0198
0.2752
0.1706
0.0120
0.0078
5,831
0.0027
4.33e+07
0.3324
1.01E+6
4,466
0.0068
101.0
0.2294
0.4647
0.3275
427,154
1.0076
5.1363
Minimum
0.0356
64
0
0
0
0
0
0.00014
1.59
0.00014
0
0
0
0.00006
12,5011
5.38e-06
219,649
0
11,501
1.59
0.00002
0
0
0
0
2,027
0
0
Maximum
0.3602
276
9
1
1
11
346,036
0.27566
275,872
0.25634
1
1
0.11352
0.08213
76,308
0.02786
8.83e+08
1
1.87e+07
61,765
0.08716
843.9
1
1
1
7.60e+06
8
55
TABLE 3
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
8.1e-11
3.4e-10
0.0035***
0.0018**
3.6e-05***
2.7e-05***
0.0095
0.0131***
-0.0032
0.0052
-0.0072
8.8e-04
0.0018
0.0097**
-4.4e-04
0.0102**
0.0069
0.0129**
0.0080
0.0123**
-1.8e-04*
-1.2e-04**
0.2410***
0.1569***
6.5e-07**
7.2e-07***
-0.0111
-0.0242***
Adjusted R-squared
Number of Observations
0.281
911
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
19
0.340
911
R&D
Employment
Share
-4.5e-10
9.7e-04***
1.2e-05***
0.0053***
0.0028
0.0014
0.0026
0.0029
0.0035*
0.0036*
-2.6e-05
0.0229**
1.4e-07*
-0.0055*
0.158
694
TABLE 4
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY WORKERS PER CAPITA SHARES
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
Per Capita High- Per Capita HighTech
Tech Services
Per Capita R&D
Employment
Employment
Employment
1.5e-10
2.1e-10
-2.0e-10*
0.0014***
7.3e-04**
4.5e-04***
1.7e-05***
1.3e-05***
5.4e-06***
0.0022
0.0045***
0.0020***
-0.0011
0.0024
0.0014*
-0.0019
9.1e-04
8.2e-04
4.2e-04
0.0038**
0.0013*
-9.2e-04
0.0035*
0.0013*
0.0022
0.0050***
0.0016*
0.0014
0.0038**
0.0015*
-7.6e-05**
-4.9e-05**
-1.1e-05
0.0900***
0.0617***
0.0088**
4.7e-07***
4.0e-07***
8.1e-08***
-0.0103**
-0.0132***
-0.0032***
0.355
912
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
20
0.439
911
0.204
694
TABLE 5
IMPACT OF HIGH CLASSIFIED RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES
Independent Variables
Population
‘High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
2.7e-10
2.7e-10
0.0024*
0.0019**
4.2e-05***
3.0e-05***
0.0055
0.0111**
-0.0079
0.0029
-0.0112
-5.8e-04
-0.0028
0.0077*
-0.0057
0.008*
0.0024
0.0111**
0.0036
0.0107**
-1.6e-04*
-1.1e-04*
0.2329***
0.1523***
6.7e-07**
7.1e-07***
-0.0083
-0.023***
Adjusted R-squared
Number of Observations
0.277
911
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
21
0.342
911
R&D
Employment
Share
-4.4e-10
8.3e-04**
1.4e-05***
0.004**
0.0015
4.1e-04
0.0014
0.0016
0.0024
0.0025
-1.9e-05
0.0200*
1.4e-07*
-0.0047*
0.155
694
TABLE 6
IMPACT OF DOCTORAL CLASSIFIED UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES
Independent Variables
Population
‘Doctoral’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
6.4e-11
7.0e-10
0.0027**
1.9e-04
4.1e-05***
2.9e-05***
0.0078
0.0112**
-0.0064
0.0025
-0.0111
-0.0023
-0.0016
0.0066
-0.0048
0.0066
0.0037
0.0098*
0.0038
0.0087*
-1.8e-04*
-1.1e-04*
0.2430***
0.1540***
6.4e-07**
7.6e-07***
-0.0075
-0.0224***
Adjusted R-squared
Number of Observations
0.277
911
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
22
0.334
911
R&D
Employment
Share
-3.2e-10
2.9e-04
1.3e-05***
0.0043**
0.0014
-2.3e-04
0.001
0.0012
0.002
0.0019
-2.2e-05
0.0229**
1.5e-07*
-0.0044*
0.144
694
TABLE 7
IMPACT OF ALL CLASSIFICATIONS OF UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES
Independent Variables
Population
‘Very High’ Research Universities
‘High’ Research Universities
‘Doctoral’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
-3.2e-10
3.1e-10
0.0024
0.0016*
0.0011
0.0016*
0.0012
-0.0011
3.8e-05***
2.8e-05***
0.0094
0.0120**
-0.0036
0.0044
-0.0070
0.0010
0.0018
0.0092*
-5.4e-04
0.0098*
0.0072
0.0125**
0.0084
0.0124**
-1.8e-04*
-1.1e-04*
0.2422***
0.1520***
6.0e-07*
7.2e-07***
-0.0103
-0.0245***
Adjusted R-squared
Number of Observations
0.281
911
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
23
0.343
911
R&D
Employment
Share
-4.8e-10
8.4e-04*
5.5e-04
-2.9e-04
1.2e-05***
0.0049**
0.0026
0.0015
0.0024
0.0029
0.0034*
0.0037*
-2.2e-05
0.0206*
1.4e-07*
-0.0056*
0.160
694
TABLE 8
WEIGHTED IMPACT OF ALL CLASSIFICATIONS OF UNIVERSITIES AND
RESEARCH ON TECHNOLOGY EMPLOYMENT SHARES
Independent Variables
Population
Weighted Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
-2.7e-10
1.4e-10
7.6e-04***
4.1e-04**
3.9e-05***
2.8e-05***
0.0089
0.0129***
-0.0039
0.0049
-0.0071
0.0010
0.0015
0.0096**
-8.4e-04
0.0101**
0.0069
0.0131**
0.0082
0.0125**
-1.8e-04*
-1.2e-04**
0.2403***
0.1566***
6.1e-07**
7.0e-07***
-0.0104
-0.0239***
Adjusted R-squared
Number of Observations
0.283
911
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
24
0.341
911
R&D
Employment
Share
-5.4e-10*
2.1e-04***
1.3e-05***
0.0051**
0.0026
0.0014
0.0025
0.0028
0.0035*
0.0036*
-2.5e-05
0.0223**
1.3e-07*
-0.0053*
0.160
694
TABLE 9
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES:
METROS LESS THAN 5 MILLION IN POPULATION
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
4.0e-09*
3.5e-09**
0.0034**
0.0017*
3.6e-05***
2.8e-05***
0.0089
0.0120**
-0.0033
0.0048
-0.0067
0.0011
0.0012
0.0089*
-0.0013
0.0094*
0.0072
0.0131**
0.0070
0.0114**
-1.5e-04*
-1.0e-04*
0.2275***
0.1448***
5.1e-07*
6.0e-07***
-0.0082
-0.0212***
Adjusted R-squared
Number of Observations
0.277
903
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
25
0.336
903
R&D
Employment
Share
-5.7e-10
9.7e-04***
1.3e-05***
0.0048**
0.0027
0.0013
0.0024
0.0028
0.0035*
0.0036*
-2.4e-05
0.0202*
1.3e-07*
-0.0049*
0.153
686
TABLE 10
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES:
METROS LESS THAN 1 MILLION IN POPULATION
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
1.6e-08**
1.2e-08**
0.0028*
0.0012
4.7e-05***
3.9e-05***
0.0076
0.0111**
-0.0053
0.0033
-0.0077
3.9e-04
-6.5e-04
0.0075*
-0.0043
0.0075
0.0076
0.0132**
0.0040
0.0096*
-2.0e-04**
-1.2e-04**
0.2065***
0.1269***
4.5e-07
5.9e-07***
-1.7e-04
-0.0167**
Adjusted R-squared
Number of Observations
0.243
861
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
26
0.295
861
R&D
Employment
Share
-5.2e-10
8.2e-04**
2.0e-05***
0.0044**
0.0019
8.8e-04
0.0019
0.0020
0.0036*
0.0030
-3.2e-05
0.0124
1.5e-07*
-0.0035
0.159
644
TABLE 11
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES:
METROS LESS THAN 500,000 IN POPULATION
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
2.2e-08*
2.3e-08***
0.0031*
0.0012
5.6e-05***
4.7e-05***
0.0058
0.0100*
-0.0069
0.0026
-0.0095
4.5e-04
-0.0026
0.0064
-0.0062
0.0072
0.0051
0.0128**
0.0021
0.0083
-1.9e-04*
-1.0e-04*
0.1924***
0.1084***
4.6e-07
6.3e-07***
0.0015
-0.0176**
Adjusted R-squared
Number of Observations
0.224
815
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
27
0.277
815
R&D
Employment
Share
1.8e-09
7.5e-04*
2.2e-05***
0.0041*
0.0019
9.2e-04
0.0018
0.0020
0.0036
0.0028
-3.0e-05
0.0093
1.5e-07*
-0.0035
0.146
598
TABLE 12
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT SHARES:
METROS LESS THAN 250,000 IN POPULATION
Independent Variables
Population
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Per Capita Personal Income
Constant
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
Share
Share
3.6e-08
4.2e-08***
0.0021
2.9e-04
3.1e-05
2.0e-05
0.0075
0.0105*
-0.0073
0.0022
-0.0090
6.8e-04
-0.0041
0.0050
-0.0072
0.0060
0.0024
0.0107*
0.0013
0.0076
-1.7e-04*
-8.6e-05
0.1692***
0.0945***
3.9e-07
5.3e-07**
0.0047
-0.0147*
Adjusted R-squared
Number of Observations
0.137
742
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
28
0.176
742
R&D
Employment
Share
4.9e-09
2.9e-04
8.4e-06
0.0035
3.7e-04
-1.0e-04
-3.9e-05
5.4e-04
0.0020
0.0015
-2.4e-05
0.0080
8.5e-08
-4.2e-04
0.039
525
TABLE 13
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0607***
-0.0491***
0.1616***
0.1030***
1,188**
783**
10.67**
1.29
2,254
4,633**
818
2,858
1,409
2,195
2,188
3,928*
5,279*
5,442**
2,657
3,295
6,647**
6,105***
43.63
7.73
1,925
1.8e+04*
3.6e-04***
5.1e-04***
-6,693*
-6,282**
0.881
912
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
29
0.855
911
R&D
Employment
-0.0126***
0.0145***
320***
0.52
1,593**
945
519
992*
1,308*
674
1,527**
-0.07
4,762
2.2e-04***
-1,454*
0.783
694
TABLE 14
IMPACT OF HIGH CLASSIFIED RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT
Independent Variables
Population
Total Employment
‘High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
-0.0599***
0.1599***
77.9
12.30***
779
-1,039
-795
30
2,748
348
4,113*
55.37*
1,716
3.6e-04***
-5,081
Dependent Variables
High-Tech
Services
Employment
-0.0484***
0.1017***
335.4
2.56
3,688*
1,729
1,052
2,715
4,033*
2,045
4,788**
14.77
1.8e+04*
5.1e-04***
-5,443*
0.880
912
0.854
911
High-Tech
Employment
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
30
R&D
Employment
-0.0123***
0.0139***
179.8*
1.06
1,136*
452
66
483
734
158
1,009*
2.66
4,417
2.2e-04***
-1,108
0.781
694
TABLE 15
IMPACT OF DOCTORAL CLASSIFIED UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT
Independent Variables
Population
Total Employment
‘Doctoral’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0602***
-0.0486***
0.1611***
0.1017***
709*
-179
12.37***
2.31
1,478
3,478*
-451
1,462
-140
500
783
2,265
3,542
3,509*
1,292
1,467
4,917*
4,144*
47.62*
17.58
2,340
1.8e+04*
3.5e-04***
5.2e-04***
-5,414
-5,078*
0.881
912
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
31
0.854
911
R&D
Employment
-0.0124***
0.0140***
-13
0.91
1,105*
348
-163
313
523
-56
754
2.91
5,099*
2.2e-04***
-951
0.779
694
TABLE 16
IMPACT OF ALL CLASSIFICATIONS OF UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
‘High’ Research Universities
‘Doctoral’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0612***
-0.0490***
0.1627***
0.1022***
1,301**
1,073**
-516
126
291
-715*
10.22**
0.87
2,636
4,295*
1,078
2,741
1,369
2,146
2,354
3,745*
5,410*
5,321**
2,809
2,971
6,625**
5,996***
40.47
12.51
3,274
1.8e+04*
3.6e-04***
5.2e-04***
-6,604*
-6,393**
0.881
912
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
32
0.855
911
R&D
Employment
-0.0125***
0.0142***
401***
74
-231*
0.44
1,461**
893
511
919
1,243*
564
1,497**
1.64
4,312
2.2e-04***
-1,481*
0.784
694
TABLE 17
WEIGHTED IMPACT OF ALL CLASSIFICATIONS OF UNIVERSITIES AND
RESEARCH ON TECHNOLOGY EMPLOYMENT
Independent Variables
Population
Total Employment
Weighted Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0601***
-0.0487***
0.1606***
0.1024***
177*
118*
11.81***
2.04
1,643
4,235*
47
2,356
712
1,745
1,423
3,433*
4,378*
4,857**
1,930
2,825
5,856**
5,594**
47.41*
10.16
1,223
1.8e+04*
3.5e-04***
5.1e-04***
-6,147*
-5,927**
0.881
912
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
33
0.854
911
R&D
Employment
-.0125***
.0143***
54**
.81
1,432**
769
383
831
1,128*
529
1,375**
0.67
4,534
2.2e-04***
-1,339
0.782
694
TABLE 18
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT:
METROS LESS THAN 5 MILLION IN POPULATION
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0441***
-0.0284***
0.0216*
0.0034
129.9
-35.9
4.50
1.61
1,674
2,500**
1,466
2,410**
1,510
2,078**
1,968
2,769***
3,149*
3,102***
2,876
3,359***
3,290*
3,081***
-8.27
-18.28*
6,459
6,609
0.0016***
0.0012***
-2,270
-2,067*
0.877
904
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
34
0.917
903
R&D
Employment
-0.0074***
-0.003
127.5
1.06
694
595
286
520
694
421
693
-2.95
2,543
3.0e-04***
-580
0.684
686
TABLE 19
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT:
METROS LESS THAN 1 MILLION IN POPULATION
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0123***
-0.0082***
0.0309***
0.0264***
249*
189**
8.84***
7.30***
249
711
-123
460
160
491
355
875*
439
878*
1,384*
1,405***
389
898*
-13.43*
-7.23
6,213*
4,647**
5.0e-04***
2.4e-04***
-369
-1,017*
0.698
862
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
35
0.691
861
R&D
Employment
-0.0042***
0.0051*
119***
3.5***
419*
289
258
371*
376
485*
392*
-0.96
-377
1.1e-04***
-309
0.428
644
TABLE 20
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH
ON TECHNOLOGY EMPLOYMENT:
METROS LESS THAN 500,000 IN POPULATION
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
Dependent Variables
High-Tech
High-Tech
Services
Employment
Employment
-0.0245***
-0.0186***
0.0427***
0.0239***
276**
153*
8.68***
7.36***
11
427
-381
213
-286
287
101
595
101
657
909
1,201***
349
686
-12.02*
-7.72*
3,605
1,872
6.9e-04***
5.8e-04***
293
-289
0.570
816
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
36
0.550
815
R&D
Employment
-0.0071***
0.0086***
108**
3.30***
282
244
222
345*
366*
477**
398*
-1.32
-893
1.7e-04***
-164
0.350
598
TABLE 21
IMPACT OF VERY HIGH RESEARCH UNIVERSITIES AND RESEARCH ON
TECHNOLOGY EMPLOYMENT:
METROS LESS THAN 250,000 THOUSAND IN POPULATION
Independent Variables
Population
Total Employment
‘Very High’ Research Universities
R&D Expenditures
Mideast
Great Lakes
Plains
Southeast
Southwest
Rocky Mountain
Far West
Cost of Living
Bachelor’s Degree
Personal Income
Constant
Adjusted R-squared
Number of Observations
High-Tech
Employment
-0.0103***
0.0212***
10.61
2.35***
90
-461*
-372
-323
-357
-53
-127
-6.07*
2,891*
4.5e-04***
341
Dependent Variables
High-Tech
Services
Employment
-0.0066***
0.0016
-26.97
1.65***
238
4
57
103
107
338*
187
-4.14**
2,134**
4.4e-04***
-115
R&D
Employment
-0.0014
-8.4e-04
13.01
0.56***
80
-9
-1
-12
-3
60
54
-0.92
122
9.3e-05**
32
0.454
743
0.473
742
0.153
525
* The t-statistic is significant at 95 percent confidence
** The t-statistic is significant at 99 percent confidence
*** The t-statistic is significant at 99.9 percent confidence
37
REFERENCES
R. Andersson, J. Quigley, and M. Wilhelmsson, “Urbanization, Productivity, and Innovation:
Evidence from Investment in Higher Education,” Journal of Urban Economics 66 (2009), 2-15.
L. Anselin, A. Varga, and Z. Acs, “Local Geographic Spillovers between University Research
and High Technology Innovations,” Journal of Urban Economics 42 (1997), 422-48.
D. Audretsch and P. Stephan, “Company-Scientist Locational Links: The Case of
Biotechnology,” American Economic Review 86 (1996), 641-52.
N. Bania, L. Calkins, and D. Dalenberg, “The Effects of Regional Science and Technology
Policy on the Geographic Distribution of Industrial R&D Laboratories,” Journal of Regional
Science 32 (1992), 209-28.
N. Bania, R. Eberts, and M. Fogarty, “Universities and the Startup of New Companies: Can We
Generalize from Route 128 and Silicon Valley?” Review of Economics and Statistics 75 (1993),
761-6.
P. Beeson and E. Montgomery, “The Effects of Colleges and Universities on Local Labor
Markets,” Review of Economics and Statistics 75 (1993), 753-61.
M. Darby and L. Zucker, “Growing by Leaps and Inches: Creative Destruction, Real Cost
Reduction, and Inching Up,” Economic Inquiry 41 (2003), 1-19.
N. Dorfman, “Route 128: The Development of a Regional High Technology Economy,”
Research Policy 12 (1983), 299-316.
J. Drucker and H. Goldstein, “Assessing the Regional Economic Development Impacts of
Universities: A Review of Current Approaches,” International Regional Science Review 30
(2007), 20-46.
M. Feldman and P. Desrochers, “Research Universities and Local Economic Development:
Lessons from the History of the Johns Hopkins University,” Industry and Innovation 10 (2003),
5-24.
M. Fogarty and A. Sinha, “Why Older Regions Can’t Generalize from Route 128 and Silicon
Valley: University-Industry Relationships and Regional Innovation Systems,” in L. Branscomb
et al. (eds.) Industrializing Knowledge: University-Industry Linkages in Japan and the United
States (The MIT Press, 1999), 473-509.
A. Glasmeier, “The High-Tech Potential: Economic Development in Rural America,” Center for
Urban Policy Research, Rutgers University (1991).
H. Goldstein and J. Drucker, “The Economic Development Impacts of Universities on Regions:
Do Size and Distance Matter?” Economic Development Quarterly 20 (2006), 22-43.
38
H. Goldstein, A. Maier, and M. Luger, “The University as an Instrument for Economic and
Business Development: U.S. and European Comparisons,” in D.D. Dill and B. Sporn (eds.)
Emerging Patterns of Social Demand and University Reform: Through a Glass Darkly
(Pergamon, 1999), 105-33.
H. Goldstein and C. Renault, “Contributions of Universities to Regional Economic
Development: A Quasi-Experimental Approach,” Regional Studies 38 (2004), 733-46.
K. Hill, “University Research and Local Economic Development,” Center for Competitiveness
and Prosperity Research, Arizona State University (2006).
A. Jaffe, “Real Effects of Academic Research,” American Economic Review 79 (1989), 957-70.
A. Jaffe, M. Trajtenberg, and R. Henderson, “Geographic Localization of Knowledge Spillovers
as Evidenced by Patent Citations,” Quarterly Journal of Economics 108 (1993), 577-98.
A. Jaffee and M. Trajtenberg. Patents, Citations, and Innovations: A Window on the Knowledge
Economy (The MIT Press, 2002).
D. Jorgenson, M. Ho, K. Stiroh, “A Retrospective Look at the U.S. Productivity Growth
Resurgence,” Journal of Economic Perspectives 22 (2008), 2-24.
J. Lane, “Science Innovation: Assessing the Impact of Science Funding,” Science 324 (2009),
1273-75.
E. Malecki, “The R&D Location Decision of the Firm and “Creative’ Regions—A Survey,”
Technovation 6 (1987), 205-22.
E. Malecki and S. Bradbury, “R&D Facilities and Professional Labour: Labour Force Dynamics
in High Technology,” Regional Studies 26 (1992), 123-35.
J. Paytas, R. Gradeck, and L. Andrews, “Universities and the Development of Industry Clusters,”
Center for Economic Development, Carnegie Mellon University, prepared for the Economic
Development Administration, U.S. Department of Commerce (2004).
E. Rogers and J. Larsen. Silicon Valley Fever: Growth of a High-Technology Culture (Basic
Books, 1984).
A. Saxenian. Regional Advantage: Culture and Competition in Silicon Valley and Route 128
(Harvard University Press, 1996).
A. Varga, “Local Academic Knowledge Transfers and the Concentration of Economic Activity,”
Journal of Regional Science 40 (2000), 289-309.
L. Zucker, M. Darby, and M. Brewer, “Intellectual Human Capital and the Birth of U.S.
Biotechnology Enterprises,” American Economic Review 88 (1998), 290-306.
39
L. Zucker, M. Darby, and J. Armstrong, “Geographically Localized Knowledge: Spillovers or
Markets?” Economic Inquiry 36 (1998), 65-86.
L. Zucker, M. Darby, and J. Armstrong, “Commercializing Knowledge: University Science,
Knowledge Capture, and Firm Performance in Biotechnology,” Management Science 48 (2002),
138-53.
40
Download