Knowledge base differentiation in urban systems of innovation and entrepreneurship

Abstract

This research aims to understand knowledge bases in urban systems of innovation and entrepreneurship. Using principal component analysis, it develops a new typology that differentiates urban knowledge bases into management knowledge, biomedical knowledge, engineering knowledge, arts and humanities knowledge, transportation knowledge and agricultural knowledge. The following multivariate analysis shows that management knowledge and engineering knowledge are of major importance in facilitating innovation and high technology entrepreneurship in US cities. Additionally, arts and humanities knowledge is positively associated with innovation but not with entrepreneurship. This research sheds light on public policy to build a vibrant urban system of innovation and entrepreneurship.

Keywords

economic development entrepreneurship innovation knowledge bases technology

Introduction

Innovation and entrepreneurship have been increasingly recognised among the core drivers of urban economic performance (Acs, 2002; Chinitz, 1961; Glaser et al., 2010; Malecki, 1994; Wennekers and Thurik, 1999). New products or new production processes replace less productive, old ones, and drive economic and productivity growths through such ‘creative destruction’ (Schumpeter, 1934). Entrepreneurship or new firm formation not only contributes to job creation (Haltiwanger et al., 2013) but also strengthens regional competitiveness either via effective allocation of market resources (Kirzner, 1997) or innovative activity (Audretsch, 1995). Accordingly, it is becoming the new norm that local and regional policymakers address innovation and entrepreneurship in their economic development strategies.

Building a vibrant urban system of innovation and entrepreneurship is by no means an easy task. In our definition, an urban system of innovation and entrepreneurship is a combination of local resources (broadly construed) and regional environmental factors that encourages and streamlines innovative and entrepreneurial activities in a city. The core function of this system is to commercialise new knowledge or new combinations of existing knowledge by innovators or entrepreneurs. However, knowledge as an uncertain, asymmetric input factor (Arrow, 1962; Audretsch, 1995) is not well understood in urban economic development. Scholars (Asheim and Hansen, 2009; Asheim et al., 2007, 2011; Lever, 2002) have developed some typologies to differentiate knowledge bases in cities, but these approaches, as we argue in the literature review section, have some major drawbacks. Efforts to further understand differentiated urban knowledge bases are still needed.

This research contributes to the literature in two aspects. First, we develop a new typology for urban knowledge bases. Compared with existing ones, our approach allows for identifying comprehensive, mutually exclusive knowledge bases in cities. Second, and based on this new typology, we examine the associations between differentiated knowledge bases and innovation and entrepreneurship in US cities. Our multivariate analysis sheds light on a better understanding of the role of knowledge in urban systems of innovation and entrepreneurship.

This paper is organised into five sections. After this introductory section, we review the literature on innovation, entrepreneurship and knowledge base differentiation in the second section. Meanwhile, a conceptual framework of urban systems of innovation and entrepreneurship is introduced. The third section develops a new typology of knowledge bases in US cities. The fourth section presents an inferential analysis on the relationships between different types of knowledge and innovative and entrepreneurial activities. The fifth section concludes.

Theories

Urban systems of innovation and entrepreneurship: Connecting knowledge with innovation and entrepreneurship in cities

The relationships between knowledge, innovation and entrepreneurship have been well established in the literature. Knowledge and innovation are two closely related terms. Innovation, generally defined by new products or new production processes (Nelson, 1993; Schumpeter, 1934), relies fundamentally on new knowledge or new combinations of existing knowledge. In a production function approach, Jaffe (1989) considers corporate innovative output as a function of private R&D and university R&D, implying that new technological knowledge as a result of R&D activity underpins innovation.

The discussion on the relationship between knowledge and entrepreneurship can be traced to Schumpeter (1934) who considers entrepreneurs as those carrying out innovative activity, but it is not until recently that knowledge and entrepreneurship have been studied in the context of economic growth models. Knowledge lies at the core of endogenous growth theory (Romer, 1986, 1990). Romer argues that knowledge is the primary driver of long-term economic growth. However, not all knowledge is commercially useful (Arrow, 1962; Audretsch, 1995). Even for knowledge with commercial potentials, the absence of entrepreneurial skills creates barriers to the appropriation of the market value of the knowledge (Michelacci, 2003). Acs et al. (2009) and Audretsch (1995) introduce a knowledge spillover theory of entrepreneurship. New firm formation according to this theory is one way by which entrepreneurs commercialise new knowledge developed in incumbent firms, universities or other research entities. Thus, new knowledge represents not only innovative opportunities (in the sense of new products) but also entrepreneurial opportunities (in the sense of new firm formation).

In the context of cities, the theories discussed above suggest that cities or regions with higher levels of knowledge stock, ceteris paribus, should have higher levels of innovative and entrepreneurial activities (Audretsch and Lehmann, 2005; Qian, 2013). While it is possible that knowledge created locally may be commercialised in other regions, the literature on the geographies of innovation and entrepreneurship (Acs et al., 2002; Anselin et al., 1997; Audretsch and Feldman, 1996; Audretsch and Lehmann, 2005; Jaffe et al., 1993; Marshall, 1920; Peri, 2005) generally supports the localisation nature of economic knowledge. This is because geographic proximity and face-to-face interaction facilitate technology transfers and knowledge spillovers.

One powerful perspective of examining the relationship between knowledge, innovation and entrepreneurship in cities or regions is the framework of regional innovation systems (RIS) and its sister framework of regional entrepreneurship systems (RES). A systems approach to innovation addresses all of the important factors that interactively facilitate learning and innovation (Edquist, 1997). In the regional context, Cooke (2004) considers RIS as ‘interacting knowledge generation and exploitation subsystems’ connected with national, international and other regional economies (p. 3). Innovation in the RIS framework reflects a ‘socially and territorially embedded and culturally and institutionally contextualised’ learning process (Asheim and Coenen, 2005: 1175). In the era of globalisation when conventional factors of production can be easily made available across different regions, the competiveness of one RIS over others is indeed embedded in its distinct, sticky knowledge base (Asheim and Coenen, 2005; Maskell et al., 1998). The stickiness of knowledge results from the spatial division of the skilled labour force shaped by history (Asheim and Coenen, 2005).

The RIS framework, as argued by Qian et al. (2013), does not sufficiently discuss the role of entrepreneurs. Although entrepreneurship can significantly facilitate innovation (Acs et al., 2009), it also has other notable economic development effects that have little to do with the latter, such as job creation (Haltiwanger et al., 2013) and better allocation of market resources (Kirzner, 1997). Qian et al. (2013) therefore propose the RES framework, addressing the regional economic system’s function in facilitating entrepreneurial activity (versus learning in RIS). They also emphasise the importance of knowledge bases where entrepreneurial opportunities reside. While the outcome of RIS is about product or process innovations, the outcome of RES is about new and particularly high growth firms.

As just discussed, both RIS and RES scholars highlight the knowledge base of regions, which represents the source of innovative opportunities or knowledge-based entrepreneurial opportunities. We therefore think that it is useful to adopt an integrative approach in a new framework called urban systems of innovation and entrepreneurship. It can be defined as a combination of local resources (broadly construed) and regional environmental factors that encourages and streamlines innovative and entrepreneurial activities in a city. Locally specific knowledge as one of the major resources lies at the centre of such a system. Commercialising localised economic knowledge via innovation and entrepreneurship is the core function of urban systems of innovation and entrepreneurship. A vibrant system of innovation and entrepreneurship represents the key competitive advantage of cities in the knowledge economy era.

Knowledge base differentiation in urban systems of innovation and entrepreneurship

The vibrancy of urban systems of innovation and entrepreneurship depends on localised knowledge bases of cities. While there is a consensus over the positive associations between knowledge and innovation and between knowledge and entrepreneurship, it is much less clear what types of knowledge bases really matter in urban systems of innovation and entrepreneurship. In the literature, it is common to see use of a single measure of knowledge in regions or cities, such as patents (Acs et al., 2002) or human capital (Audretsch and Keilbach, 2006). This is not surprising, as both categorising and quantifying knowledge in cities are challenging tasks. Nevertheless, there are two noticeable approaches to categorising knowledge of regions or cities.

First, and following the tradition of Polanyi (1967), knowledge is either codified or tacit. While codified knowledge can be documented in books, journals, newspapers or other outlets, tacit knowledge cannot be easily described due to lack of self-awareness and/or difficulties in communication (Gertler, 2003). As Gertler further puts, the transmission of tacit knowledge generally requires face-to-face interaction, which makes geographical proximity or cities important in knowledge spillovers and innovation. Lever (2002) makes an effort to quantify urban knowledge bases along the line of codified/tacit knowledge. He decomposes the knowledge bases of cities into codified knowledge (measured by the number of students and academic publications), tacit knowledge (measured by the presence of major knowledge companies, airport connectivity and major fairs/exhibitions, and new enterprises) and knowledge infrastructure (measured by the quality of telecommunications). There are two problems with Lever’s work. First, although adding knowledge infrastructure to the Polanyian dichotomy expands the indicators used to measure the knowledge base of cities, it also makes the three components not mutually exclusive. Second, the measures for these components, especially for tacit knowledge, are apparently rough.

The second typology is developed by Asheim and his colleagues (Asheim and Hansen, 2009; Asheim et al., 2007, 2011). They differentiate regional knowledge bases into a science-based analytical knowledge base, an engineering-based synthetic knowledge base and an arts-based symbolic knowledge base. The analytical knowledge base is defined by ‘economic activities where scientific knowledge based on formal models and codification is highly important’ (Asheim et al., 2011: 896). By definition, the analytical knowledge base is more tied with codified knowledge than tacit knowledge, and thus face-to-face interaction and geographical proximity play a less important role (Asheim et al., 2007). The synthetic knowledge base is defined by ‘economic activities where innovation takes place mainly through the application or novel combinations of existing knowledge’ (Asheim et al., 2011: 897). While codified knowledge is still important, the synthetic knowledge base is also built on know-how, tacit knowledge and problem solving, all of which require face-to-face communication and which benefit from geographical proximity (Asheim et al., 2007). Lastly, the symbolic knowledge base is defined by ‘the creation of meaning and desire as well as aesthetic attributes of products, producing designs, images and symbols, and to the economic use of such forms of cultural artifacts’ (Asheim et al., 2011: 897). For this base, tacit knowledge is even more important, involving not only know-how and but also know-who (i.e. finding the right person for a specific task) that makes geographically mediated face-to-face interaction and buzz very important (Asheim et al., 2007). Asheim and Hansen (2009) quantify these three knowledge bases by measuring the location quotient of most relevant occupations for each, an approach similar to Richard Florida’s measure of the creative class (Florida, 2002). Asheim’s typology is a major step in differentiating regional knowledge bases, but it overlooks other types of knowledge bases in cities beyond science, engineering and arts. For instance, business or management knowledge, greatly desired in the process of innovation or entrepreneurship, is left out.

Neither Lever (2002) nor Asheim and his colleagues (Asheim and Hansen, 2009; Asheim et al., 2007, 2011) further examine the relationships between differentiated knowledge bases and innovation and entrepreneurship in regions or cities, except for the correlation analysis by Asheim and Hansen (2009) where positive associations between the three types of knowledge and technology/patents are found. Additional work to identify the types of knowledge bases that are important to innovation and entrepreneurship is needed to better understand urban economic systems.

A new typology of knowledge bases of US cities

This research introduces a new typology of knowledge bases in the context of US Metropolitan Statistical Areas (MSAs). In this study, cities and MSAs are interchangeably used. The US federal government defines MSAs in terms of commute-to-work. Each MSA can be considered as a regionally integrated labour market. Unlike the deductive approach by Asheim and his colleagues (Asheim and Hansen, 2009; Asheim et al., 2007, 2011), we adopt an inductive approach, using principal component analysis towards a unique dataset on occupational knowledge, called Occupational Information Network or O*NET. Compared with previous work (Asheim and Hansen, 2009; Asheim et al., 2007, 2011; Lever, 2002), our approach allows for identifying comprehensive, mutually exclusive knowledge bases in cities.

The O*NET data is a survey dataset from the US Department of Labor. The survey results include importance and level scores on 33 knowledge variables or types for each occupation defined in the Standard Occupational Classification (SOC) System of the US Bureau of Labor Statistics (BLS). Detailed information of this survey can be found at its official website (http://www.onetcenter.org/overview.html). In the questionnaire (https://onet.rti.org/pdf/OE_Knowledge_Questionnaire.pdf), two questions are separately asked for each knowledge variable: ‘How important is [X] knowledge to the performance of the occupation?’ and ‘What level of [X] knowledge is needed to perform the occupation?’. To help respondents understand different levels, the questionnaire gives concrete examples for each type of knowledge. In the case of ‘Economics and Accounting’ knowledge, for instance, the level-2 example is ‘Answer billing questions from credit card customers’; the level-4 example is ‘Develop financial investment programs for individual clients’; and the level-6 example is ‘Keep a major corporation’s financial records’. Based on the choices under each question, the importance of a particular type of knowledge for each occupation has a score on a scale of 1–5, and the level on a scale of 0–7. To obtain one single value for one knowledge type of each occupation, we use the product between the importance score and the level score, following Feser (2003) and Gabe et al. (2012). Next, we calculate the weighted average of a knowledge type as an indicator of an MSA’s score on this knowledge, using the share of an occupation’s employment in the MSA as the weight. It can be mathematically written as:

K_{i, m} = \frac{\sum_{j = 1}^{N} ({k_{i, j}}^{*} E_{j, m})}{\sum_{j = 1}^{N} E_{j, m}}

(1)

Where K_i,m the value of the i-th type of knowledge (among the 33 knowledge variables) in MSA m; k_i,j is the value of the i-th type of knowledge for occupation j (again, the product between the importance score and the level score from O*NET); E_j,m is the employment in occupation j in MSA m; and N is the number of occupations in the O*NET dataset. BLS provides occupational employment data by MSAs. We calculate knowledge values for all MSAs.¹

Using O*NET data, Gabe et al. (2012) have measured the 33 types of knowledge in US and Canadian cities in a similar way. Further, they conduct a cluster analysis to categorise the cities based on their knowledge values. Moving towards a different direction, we categorise not cities but the knowledge bases of cities using principle component analysis. By doing this, we are able to group those highly correlated knowledge variables into one knowledge base and eventually provide a typology that includes mutually exclusive knowledge bases after factor rotation.

Table 1 shows the results of principal component analysis using the 2005–2006 averages of MSA knowledge scores (based on O*NET 15.0) and MSA occupational employment. All 358 lower-state MSAs are considered. The 2005–2006 time period is used to accommodate the data availability issue in the following regression analysis, as discussed in the next section. It results in a six-factor solution by retaining all factors that have an eigenvalue of at least one. The six knowledge bases, which account for 90 percent of the total variances of the 33 knowledge variables, are named as management (26% of total variance), biomedical (22%), engineering (16%), arts and humanities (15%), transportation (6%) and agricultural (5%). Interestingly but not surprisingly, three of these six knowledge bases – biomedical, engineering and arts and humanities – correspond to the three knowledge bases defined by Asheim and his colleagues (Asheim and Hansen, 2009; Asheim et al., 2007, 2011), i.e. the analytical knowledge base, the synthetic knowledge base and the symbolic knowledge base. Additionally, we are able to identify three additional independent knowledge bases – management, transportation and agricultural – and therefore measure differentiated urban knowledge bases more comprehensively. The management knowledge base, which is missing in Asheim’s framework, accounts for 26 percent of the total variance of 33 knowledge variables, the highest among the six knowledge bases. It is therefore an important part of urban knowledge and deserves particular attention. The newly identified agricultural knowledge base may seem surprising at first glance, but US MSAs are defined along county boundaries and the peripheral area of MSAs is likely to be rural.

Table 1.

Principal component analysis – factor loadings after rotation.

Knowledge Variable	Factor 1: Management	Factor 2: Biomedical	Factor 3: Engineering	Factor 4: Arts and humanities	Factor 5: Transportation	Factor 6: Agricultural
Economics and Accounting	0.910
Clerical	0.854
Personnel and Human Resources	0.851
Administration and Management	0.850
Customer and Personal Service	0.799
Telecommunications	0.794
Law and Government	0.793
Sales and Marketing	0.758
Communications and Media	0.743
Computers and Electronics	0.742
English Language	0.721
Medicine and Dentistry		0.962
Therapy and Counselling		0.866
Psychology		0.861
Biology		0.851
Philosophy and Theology		0.764
Sociology and Anthropology		0.756
Foreign Language		0.680
Chemistry		0.669
Physics			0.942
Engineering and Technology			0.912
Design			0.896
Mathematics			0.707
Mechanical			0.687
Building and Construction			0.553
Production and Processing			0.547
History and Archeology				0.932
Fine Arts				0.914
Geography				0.880
Education and Training				0.686
Transportation					0.860
Public Safety and Security					0.788
Food Production						0.783^#
(Share of total variances)	(26%)	(22%)	(16%)	(15%)	(6%)	(5%)

Notes: Principal component analysis was done based on all 358 MSAs in lower US states (2005 MSA definition by the US Office of Management and Budget).

Based on highest factor loadings, six factors are separately named as follows: 1 – management knowledge; 2 – biomedical knowledge; 3 – engineering knowledge; 4 – arts and humanities knowledge; 5 – transportation knowledge; 6 – agricultural knowledge.

Factor 6’s highest loadings were negative; both factor loadings and predicted factor value are therefore reversed, so that a higher factor value represents a higher level of agricultural knowledge.

Table 2 shows the top-10 ranked cities by the factor score of each knowledge base. These cities can also be found in the Appendix maps (Figure A1), showing spatial variations of the knowledge bases across all MSAs. In terms of the management knowledge base, the highest ranked city is the national capital region, i.e. the Washington MSA. This is not surprising, as the presence of the US federal government and its decentralisation has created numerous business opportunities for local entrepreneurs. According to Feldman et al. (2005), this region has built mature information technology and biotechnology clusters, and entrepreneurs play a leading role in this process. In the Washington area, the share of management occupations accounts for 6.4% of total employment in 2006, compared with 4.4% of the national average (own calculation based on BLS data). Interestingly, among other top-ranked MSAs, San Jose, CA, Austin, TX and Boulder, CO are also highly entrepreneurial and technology-intensive regions. San Jose is home to Silicon Valley, and Boulder is perhaps the most entrepreneurial college town in the United States. These results suggest a complementary relationship between high technology and management.

Table 2.

Top 10 cities among 358 MSAs by six knowledge factors.

Rank	Management	Biomedical	Engineering
1	Washington-Arlington-Alexandria, DC-VA-MD-WV	Alexandria, LA	Warner Robins, GA
2	Tallahassee, FL	Durham, NC	Huntsville, AL
3	Trenton-Ewing, NJ	Gainesville, FL	San Jose-Sunnyvale-Santa Clara, CA
4	Carson City, NV	Morgantown, WV	Palm Bay-Melbourne-Titusville, FL
5	Boulder, CO	Iowa City, IA	Kennewick-Richland-Pasco, WA
6	Austin-Round Rock, TX	Huntington-Ashland, WV-KY-OH	Elkhart-Goshen, IN
7	Boise City-Nampa, ID	Duluth, MN-WI	Victoria, TX
8	Pocatello, ID	Cumberland, MD-WV	Boulder, CO
9	San Jose-Sunnyvale-Santa Clara, CA	Kankakee-Bradley, IL	Bremerton-Silverdale, WA
10	Idaho Falls, ID	Columbia, MO	Lake Charles, LA
Rank	Arts and humanities	Transportation	Agricultural
1	Yuba City, CA	Houma-Bayou Cane-Thibodaux, LA	Madera, CA
2	Vineland-Millville-Bridgeton, NJ	Farmington, NM	Myrtle Beach-Conway-North Myrtle Beach, SC
3	Merced, CA	Laredo, TX	Flagstaff, AZ
4	McAllen-Edinburg-Mission, TX	Vineland-Millville-Bridgeton, NJ	Yuma, AZ
5	El Centro, CA	Odessa, TX	El Centro, CA
6	Ithaca, NY	Pine Bluff, AR	Salinas, CA
7	Monroe, MI	Albany, GA	Hanford-Corcoran, CA
8	Kingston, NY	Joplin, MO	Naples-Marco Island, FL
9	Dover, DE	Lake Charles, LA	Sebastian-Vero Beach, FL
10	Killeen-Temple-Fort Hood, TX	Beaumont-Port Arthur, TX	Ocean City, NJ

Among the five cities with the highest level of biomedical knowledge, four of them (No. 2 – Durham, NC, No. 3 – Gainesville, FL, No. 4 – Morgantown, WV and No. 5 – Iowa City, IA) are notable college towns with major medical schools and hospitals. Take Iowa City as an example. The University of Iowa, Iowa’s flagship public university, has nationally recognised academic programs in medicine, pharmacy, dentistry, nursing and public health. Additionally, the University of Iowa Hospitals and Clinics in 2013 employed over 10,000 physician and non-physician employees (as a reference, the metropolitan population was about 160,000 in the same year), according to its official website. Iowa City is also home to a US Department of Veterans Affairs hospital. These major biomedical-based employers make the Iowa City MSA among the highest-ranked in this category.

The engineering knowledge base highlights drastically different types of cities, but all highly relevant to engineering. Among them, Warner Robins, GA (known as the ‘Carpet Capital of the World’) and Elkhart, IN (known as the ‘Recreational Vehicle Capital of the World’) are both manufacturing-based. Their high value in the engineering knowledge base is driven by the disproportionally high presence of production occupations. By contrast, San Jose (No. 3) and Boulder (No. 8) as discussed above are among the major high technology centres in the United States. Their occupational structure is strongly weighted towards computer, mathematical and engineering occupations. Huntsville, AL (No. 2) relies primarily on federally supported rocket propulsion research anchors including the Marshall Space Flight Center and the US Army Aviation and Missile Command (AMCOM).

Top cities in the arts and humanities knowledge base are generally small cities with strong employment in education, training and library occupations. Three small, southern cities, Houma, LA, Farmington, NM and Odessa, TX, are the top three in the transportation knowledge base. Houma is a coastal city with strong offshore oil exploitation and fishing industries. The economic base of secondly ranked Farmington, NM is mining of energy resources. Though not transportation centres, the economic activities of these two cities are highly reliant on transportation. Laredo is located at the US-Mexico border and is a major transportation centre for international trade between these two countries. Lastly, top-ranked MSAs in the agricultural knowledge base generally have a large workforce in farming, fishing and forestry occupations. For instance, 15% of employment of No. 1 Madera, CA in 2006 was in these occupations, compared with 0.3% nationwide (own calculation based on BLS data).

Knowledge bases, innovation and entrepreneurship: empirical evidence from US cities

After identifying the six knowledge bases of US cities, we now examine the differentiated effects of these knowledge bases on innovation and entrepreneurship. The purpose is to reveal what types of knowledge bases really matter in urban systems of innovation and entrepreneurship. Neither Lever (2002) nor Asheim and his colleagues (Asheim and Hansen, 2009; Asheim et al., 2007, 2011) have clearly answered this question after building their knowledge base typology.

Model specification

We run cross-sectional regressions in which innovation and entrepreneurship are separately used as the dependent variable and the primary explanatory variables are the six knowledge bases discussed in the previous section. Although a panel data approach may provide more reliable results, it is not used for two reasons: (1) the MSA definition and the occupational classification have been continuously changing, making it difficult to build a temporally consistent dataset, and (2) the results of principle component analysis on all 33 knowledge variables are not always temporally stable. In this study, we examine metropolitan innovative and entrepreneurial activities in 2006, the year before the 2007–2009 recession. While it would also be interesting to explore a more recent time after the recession, data on new firm formation by industries used to measure high technology entrepreneurship were not available to the authors when this study was performed. To obtain more reliable measures while mitigating possible reverse causalities, all the independent variables are based on the average of the 2005 and 2006 data, except for population growth that stretches for a longer time period (2000–2006).

In the innovation model, the dependent variable – innovation – is measured by the number of patents normalised by population. Patents have been widely used as an indicator of regional innovation (Acs et al., 2002; Plummer and Acs, 2014; Qian, 2013), despite not being a perfect measure in that lots of patents are not commercialised and lots of innovations are not associated with patents (Parkers and Griliches, 1980). Patent data and population data are available from the US Patent and Trademark Office (USPTO) and the US Census Bureau’s Population Estimates respectively. Besides the six knowledge base variables measured by their factor scores, the regression model also controls for four variables that may impact innovation, as demonstrated by the literature. The first one is human capital, measured by the share of adults (age 25+) with a bachelor’s degree or above. As shown in many studies (e.g. Faggian and McCann, 2009; Florida et al., 2008; Qian, 2013), human capital is a strong predictor of regional innovative activity. The second control variable is population size (in its logarithm form) as a measure of agglomeration or urbanisation economies. Large cities have advantages in innovations because of high levels of knowledge spillovers especially across different sectors (Duranton and Puga, 2001; Jacobs, 1969). The third one is the share of foreign-born population in the metro, addressing the role of cultural diversity and immigrants in innovation (Florida, 2002; Niebuhr, 2010; Qian, 2013; Qian and Stough, 2011). The data source for the first three control variables is the US Census Bureau, either its American Community Survey (ACS) or its Population Estimates. The last control variable in the innovation model is a dummy variable indicating the presence of one or more research universities in the metro (1 = yes; 0 = no). Research universities, as noted in many studies (Anselin, et al., 1997; Audretsch et al., 2013; Fritsch and Slavtchev, 2007; Jaffe, 1989), are sources of innovation. We use the list of US universities with very high research activity identified by the Carnegie Foundation in 2010.

In the case of entrepreneurship, we use two separate dependent variables – entrepreneurship and high technology entrepreneurship. High technology entrepreneurship is much more likely to be determined by knowledge. Entrepreneurship is measured by the number of new single-unit establishments normalised by employment; and high technology entrepreneurship is measured by the number of new single-unit high technology establishments normalised by employment. We use the high technology industries defined by BLS economist Daniel Hecker (2005). New single-unit establishment data are obtained from the Business Information Tracking Series of the US Census Bureau,² and employment data are available from the US County Business Patterns (CBP). In addition to the six knowledge base variables, we also include the four control variables – human capital, population size, foreign born and research university – that are used in the innovation model. The effects of these four variables on entrepreneurship in the knowledge economy context are also well documented in the literature. Specifically, human capital contributes to the absorptive capacity of entrepreneurs, which further promotes regional entrepreneurial activity (Qian and Acs, 2013; Qian et al., 2013). Population size or agglomeration economies facilitate knowledge spillovers that, according to the knowledge spillover theory of entrepreneurship (Acs et al., 2009; Audretsch, 1995), encourage the creation of new firms. Cultural diversity or foreign born signifies diverse perspectives that make market opportunities more likely to be commercially exploited by entrepreneurs (Audretsch et al., 2010; Cheng and Li, 2012; Qian, 2013; Qian and Haynes, 2014). New knowledge generated via universities’ R&D may be discovered and commercialised by local entrepreneurs (Bercovitz and Feldman, 2006), highlighting the role of research universities in regional entrepreneurial activity. Further, we control for three additional variables that are less relevant to knowledge-based entrepreneurship but may be directly associated with new firm formation: population growth, establishment size and unemployment. Population growth is measured by the percentage change of metropolitan population between 2000 and 2006. According to Reynolds et al. (1994), it reflects increased demand that creates additional market opportunities for entrepreneurs. Data source for this variable is the US Census Bureau’s Population Estimates. Establishment size is an indicator of the level of market competition. The smaller the average establishment size, the more small businesses there are in the market. The presence of a large number of small businesses represents an environment that is friendly to small and new businesses. Therefore, establishment size should be negatively associated with new firm formation (Acs and Armington, 2006). Establishment size is calculated by dividing the number of paid employees by the number of establishments in the metropolitan area. Data are available from CBP. Lastly, unemployment rate may be positively associated with entrepreneurship, while unemployed individuals may be forced to start their own businesses (Reynolds et al., 1994). Unemployment rate data are from ACS.

Table 3 summarises all of the variables used in regression analysis. Table 4 is a correlation matrix among these variables. The correlation matrix table helps to identify variables that have high explanatory power in the later regression results.

Table 3.

Descriptive statistics.

Variable	Obs.	Mean	Std. Dev.	Min	Max
Innovation	358	0.252	0.438	0.000	4.549
Entrepreneurship	358	5.781	2.265	2.802	16.597
H.T. Entrepreneurship	358	0.413	0.274	0.000	2.030
Management	358	0.000	1.000	−4.231	2.990
Biomedical	358	0.000	1.000	−2.464	2.960
Engineering	358	0.000	1.000	−2.852	4.632
Arts and Humanities	358	0.000	1.000	−2.767	2.672
Transportation	358	0.000	1.000	−2.437	5.521
Agricultural	358	0.000	1.000	−3.561	3.684
Human Capital	356	24.779	7.878	11.100	56.850
Population Size	358	12.647	1.063	10.926	16.750
Population Growth	358	6.167	6.928	−22.118	38.433
Foreign Born	356	7.428	6.818	0.851	36.792
Establishment Size	358	15.448	2.596	6.457	24.752
Unemployment	356	6.707	1.734	2.300	13.800
Research University	358	0.221	0.415	0.000	1.000

Table 4.

Correlation matrix.

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1	Innovation	1.000
2	Entrepreneurship	0.045	1.000
3	H.T. Entrepreneurship	0.414	0.658	1.000
4	Management	0.265	0.231	0.499	1.000
5	Biomedical	−0.072	−0.227	−0.208	0.007	1.000
6	Engineering	0.273	−0.057	0.256	−0.002	0.005	1.000
7	Arts and Humanities	0.171	−0.046	0.047	0.011	−0.008	0.004	1.000
8	Transportation	−0.310	0.102	−0.155	0.001	−0.003	0.002	−0.002	1.000
9	Agricultural	−0.034	0.525	0.244	0.001	0.001	−0.001	0.001	0.001	1.000
10	Human Capital	0.535	0.104	0.564	0.578	−0.009	0.171	0.202	−0.469	0.043	1.000
11	Population Size	0.168	−0.007	0.336	0.369	0.100	0.302	0.300	−0.128	−0.165	0.342	1.000
12	Population Growth	0.006	0.599	0.474	0.102	−0.325	0.092	0.006	0.118	0.262	0.053	0.181	1.000
13	Foreign Born	0.234	0.309	0.410	0.098	−0.143	0.109	0.301	−0.094	0.262	0.187	0.463	0.402	1.000
14	Establishment Size	0.049	−0.798	−0.432	−0.195	0.065	0.216	0.033	−0.127	−0.556	−0.009	0.239	−0.215	−0.112	1.000
15	Unemployment	−0.193	−0.138	−0.359	−0.312	0.160	−0.093	0.143	0.132	0.037	−0.445	−0.033	−0.172	0.139	0.032	1.000
16	Research University	0.331	−0.043	0.331	0.383	0.055	0.221	0.200	−0.303	−0.027	0.599	0.510	0.002	0.277	0.133	−0.143

Note: The American Community Survey (ACS), the data source of human capital, foreign born and unemployment variables, does not provide 2005–2006 data for two MSAs: Carson City, NV and Lewiston, ID-WA. Therefore, the correlation analysis and regression analysis exclude these two MSAs and are based on all other 356 variables.

Regression results

Table 5 shows the OLS regression results. For the innovation model, five of the six knowledge bases show a significant association with regional innovation at least at the 0.10 level. Among them, management knowledge, engineering knowledge and arts and humanities knowledge are positively associated with patenting activity. The positive effects of the management knowledge base and the arts knowledge base on innovation are of particular interest, as these relationships have been less studied in the literature. A strong management knowledge base may increase the awareness and streamline the process of applying for patents that have potential economic benefits. This relationship, however, should be read with caution, as it is only significant at the 0.10 level. The arts knowledge base may contribute to innovation in an incremental way via new designs of existing products. Transportation knowledge and agricultural knowledge are negatively associated with innovation, suggesting that a deep presence of either of these two may defer new knowledge generation and/or commercialisation.

Table 5.

OLS regression results.

	Dependent variable:Innovation	Dependent variable:Entrepreneurship	Dependent variable:High-tech entrepreneurship
Management	0.049* [0.028]	−0.062[0.074]	0.026** [0.011]
Biomedical	−0.003[0.019]	−0.139*** [0.053]	−0.020** [0.008]
Engineering	0.115*** [0.021]	0.044[0.054]	0.057*** [0.008]
Arts and Humanities	0.042** [0.021]	−0.170*** [0.056]	−0.022*** [0.008]
Transportation	−0.059** [0.024]	0.027[0.062]	−0.010[0.009]
Agricultural	−0.071*** [0.021]	0.078[0.063]	−0.030*** [0.009]
Human Capital	0.022*** [0.004]	0.023** [0.012]	0.012*** [0.002]
Population Size	−0.114*** [0.026]	0.242*** [0.071]	0.030*** [0.011]
Population Growth		0.123*** [0.009]	0.009*** [0.001]
Foreign Born	0.016*** [0.004]	0.008[0.010]	0.006*** [0.001]
Establishment Size		−0.624*** [0.025]	−0.048*** [0.004]
Unemployment		−0.009[0.034]	−0.012** [0.005]
Research University	0.000[0.061]	−0.188[0.159]	−0.014[0.024]
Number of observations	356	356	356
Adjusted R²	0.372	0.841	0.732

Notes: *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level; standard errors in brackets.

To test robustness of the results, we have also tried to use only the level scores of the O*NET knowledge variables. Both factor analysis results and regressions results are consistent with the results presented in this article.

For the entrepreneurship model, no knowledge base significantly improves general entrepreneurship. Four of the six knowledge bases are insignificantly associated with the dependent variable. The negative and significant coefficients of the biomedical knowledge base and the arts and humanities knowledge base suggest that they are not friendly to general startup activity. The lack of explanatory power by the knowledge bases nevertheless is supported by the evidence provided in Qian et al. (2013) that most new firms in the United States are not knowledge- or technology-based.

For the high technology entrepreneurship model, both the management knowledge base and the engineering knowledge base are positively associated with high technology startup activity, and these relationships are significant at least at the 0.05 level. This finding is consistent with Qian and Acs’s argument (2013) that entrepreneurs need both market knowledge and technological knowledge for successful commercialisation of new technology in new firms. Biomedical knowledge, generally associated with high technology, is negatively associated with high technology startups, providing more insights into the nature of this knowledge type that does not encourage new firm formation. Strict federal regulations in approving new drugs further discourage startups in the biomedical field. Perhaps not surprisingly, a negative, significant association with high technology entrepreneurship is also found for the arts and humanities knowledge base and the agricultural knowledge base.

Among the significant results from control variables, human capital is positively associated with innovation, entrepreneurship and high technology entrepreneurship, consistent with the literature (Faggian and McCann, 2009; Florida et al., 2008; Qian, 2013; Qian and Acs, 2013). Population size is positively associated with entrepreneurship measures but negatively associated with innovation measured via patents. These results may imply that the large size of cities benefits urban systems of innovation and entrepreneurship through providing more market opportunities instead of knowledge spillovers. The effect of population growth is positive and significant both in the entrepreneurship model and in the high technology entrepreneurship model, which is in line with the finding by Reynolds et al. (1994). Foreign born is positively associated with knowledge-based innovative activity and high technology entrepreneurship, supporting existing literature on the role of cultural diversity or immigrants (Audretsch et al., 2010; Cheng and Li, 2012; Florida, 2002; Niebuhr, 2010; Qian, 2013; Saxenian, 2002). Establishment size as hypothesised exerts a negative effect on both entrepreneurship and high technology entrepreneurship. Unemployment only presents a significant relationship with high technology entrepreneurship, and the coefficient is negative. Therefore, a high employment rate is more an indicator of economic stagnation, which discourages knowledge-based startup activity. Lastly, the presence of one or more research universities does not contribute to regional innovative and entrepreneurial activity after human capital is controlled for.

Summary and policy discussions

This research aims to provide a better understanding of knowledge bases in urban systems of innovation and entrepreneurship. Using principal component analysis on the O*NET data, it develops a new typology that differentiates urban knowledge bases into management knowledge, biomedical knowledge, engineering knowledge, arts and humanities knowledge, transportation knowledge and agricultural knowledge. The multivariate analysis results show that management knowledge and engineering knowledge are of major importance in facilitating innovation and high technology entrepreneurship in US cities. Additionally, the arts and humanities knowledge contributes to innovation but not entrepreneurship.

Our research sheds light on how to build a vibrant urban system of innovation and entrepreneurship. It is important to strengthen the management and engineering knowledge bases in the labour market, which are conducive to innovation and technology-driven startups. Additionally, results from the control variables in our regression models show that human capital and foreign-born population are important as well. Cities need to create environments that can attract and retain well-educated population and immigrants.

The success of US high technology centres (e.g. Silicon Valley, Route 128 in Boston and the Research Triangle in North Carolina) has created an illusion that it is critical to build a system of innovation and entrepreneurship around one or more major research universities. Our analysis shows that there are no significant associations between the presence of major universities and innovative and entrepreneurial activities after controlling for human capital. The core question should rather be whether a city could attract and retain talented college graduates, who do not have to be locally ‘produced’.

This research has two limitations. First, it would require some survey data similar to O*NET in order to generalise this analysis to cities in other countries. Second, for the reasons discussed in the fourth section, we have only conducted cross-sectional analysis when investigating the effects of differentiated knowledge bases on innovation and entrepreneurship. Therefore, our multivariate results reflect associations but not necessarily causalities.

Future research may take the direction of studying knowledge diversity in urban systems of innovation and entrepreneurship, an idea originated from Jacobs (1969), instead of examining individual effects of differentiated knowledge bases as done in this study. Recently, the literature on evolutionary economic geography has significantly advanced this line of research by addressing the importance of ‘related variety’ (Asheim et al., 2011; Boschma and Iammarino, 2009; Frenken et al., 2007). While existing work has mostly been focused on industrial diversity, O*NET data can be used to measure the knowledge diversity of US cities.

Footnotes

Appendix

Acknowledgements

The author would like to thank Hyejin Jung for her research assistance. An earlier version of this research was presented at the annual meeting of the Association of Collegiate Schools of Planning (ACSP) on 30 October 2014 in Philadelphia.

Funding

This research is funded by the Early Career Grant Scheme of the Regional Studies Association and the Faculty Scholarship Initiative of Cleveland State University.

Notes

References

Acs

(2002) Innovation and the Growth of Cities. Cheltenham: Edward Elgar Publishing.

Acs

Armington

(2006) Entrepreneurship, Geography, and American Economic Growth. New York: Cambridge University Press.

Acs

Anselin

Varga

(2002) Patents and innovation counts as measures of regional production of new knowledge. Research Policy 31(7): 1069–1085.

Acs

Audretsch

Braunerhjelm

. (2009) The knowledge spillover theory of entrepreneurship. Small Business Economics 32: 15–30.

Anselin

Varga

Acs

(1997) Local geographic spillovers between university research and high technology innovations. Journal of Urban Economics 42(3): 422–448.

Arrow

(1962) The economic implications of learning by doing. Review of Economic Studies 29(3): 155–173.

Asheim

Hansen

(2009) Knowledge bases, talents, and contexts: On the usefulness of the creative class approach in Sweden. Economic Geography 85(4): 425–442.

Asheim

Coenen

Vang

(2007) Face-to-face, buzz, and knowledge bases: Sociospatial implications for learning, innovation, and innovation policy. Environment and Planning C 25(5): 655–670.

Asheim

Coenen

(2005) Knowledge bases and regional innovation systems: Comparing Nordic clusters. Research Policy 34(8): 1173–1190.

10.

Asheim

Boschma

Cooke

(2011) Constructing regional advantage: Platform policies based on related variety and differentiated knowledge bases. Regional Studies 45(7): 893–904.

11.

Audretsch

(1995) Innovation and Industry Evolution. Cambridge, MA: MIT Press.

12.

Audretsch

Feldman

(1996) R&D spillovers and the geography of innovation and production. American Economic Review 86(3): 630–640.

13.

Audretsch

Keilbach

(2006) Entrepreneurship, economic growth and restructuring. In: Casson

Yeung

Basu

al.

(eds) The Oxford Handbook of Entrepreneurship. New York: Oxford University Press, pp. 281–310.

14.

Audretsch

Lehmann

(2005) Does the knowledge spillover theory of entrepreneurship hold for regions? Research Policy 34(8): 1191–1202.

15.

Audretsch

Dohse

Niebuhr

(2010) Cultural diversity and entrepreneurship: A regional analysis for Germany. The Annals of Regional Science 45(1): 55–85.

16.

Audretsch

Leyden

Link

(2013) Regional appropriation of university-based knowledge and technology for economic development. Economic Development Quarterly 27(1): 56–61.

17.

Bercovitz

Feldman

(2006) Entrepreneurial universities and technology transfer: A conceptual framework for understanding knowledge-based economic development. The Journal of Technology Transfer 31: 175–188.

18.

Boschma

Iammarino

(2009) Related variety, trade linkages, and regional growth in Italy. Economic Geography 85(3): 289–311.

19.

Cheng

(2012) New firm formation facing cultural and racial diversity. Papers in Regional Science 91(4): 759–774.

20.

Chinitz

(1961) Contrasts in agglomeration: New York and Pittsburgh. The American Economic Review 51: 279–289.

21.

Cooke

(2004) Evolution of regional innovation systems – Emergence, theory, challenge for action. In: Cooke

Heidenreich

Braczyk

(eds) Regional Innovation Systems: The Role of Governance in a Globalized World. 2nd edition. London: Routledge, pp. 1–18.

22.

Duranton

Puga

(2001) Nursery cities: Urban diversity, process innovation, and the life cycle of products. American Economic Review 91(5): 1454–1477.

23.

Edquist

(ed.) (1997) Systems of Innovation: Technologies, Institutions and Organizations. London: Pinter.

24.

Faggian

McCann

(2009) Human capital, graduation migration and innovation in British regions. Cambridge Journal of Regions, Economy and Society 33: 317–333.

25.

Feldman

Francis

Bercovitz

(2005) Creating a cluster while building a firm: Entrepreneurs and the formation of industrial clusters. Regional Studies 39(1): 129–141.

26.

Feser

(2003) What regions do rather than make: A proposed set of knowledge-based occupation clusters. Urban Studies 40(10): 1937–1958.

27.

Florida

(2002) The Rise of the Creative Class. New York: Basic Books.

28.

Florida

Mellander

Stolarick

(2008) Inside the black box of regional development – Human capital, the creative class and tolerance. Journal of Economic Geography 8(5): 615–649.

29.

Frenken

Van Oort

Verburg

(2007) Related variety, unrelated variety and regional economic growth. Regional Studies 41(5): 685–697.

30.

Fritsch

Slavtchev

(2007) Universities and innovation in space. Industry and Innovation 14(2): 201–218.

31.

Gabe

Abel

Ross

. (2012) Knowledge in cities. Urban Studies 49(6): 1179–1200.

32.

Gertler

(2003) Tacit knowledge and the economic geography of context, or the undefinable tacitness of being (there). Journal of Economic Geography 3(1): 75–99.

33.

Glaeser

Rosenthal

Strange

(2010) Urban economics and entrepreneurship. Journal of Urban Economics 67(1): 1–14.

34.

Haltiwanger

Jarmin

Miranda

(2013) Who creates jobs? Small versus large versus young. Review of Economics and Statistics 95(2): 347–361.

35.

Hecker

(2005) High-technology employment: A NAICS-based update. Monthly Labor Review 128: 57–72.

36.

Jacobs

(1969) The Economy of Cities. New York: Random House.

37.

Jaffe

(1989) Real effects of academic research. American Economic Review 79(5): 957–970.

38.

Jaffe

Trajtenberg

Henderson

(1993) Geographic localization of knowledge spillovers as evidenced by patent citations. The Quarterly Journal of Economics 108(3): 577–598.

39.

Kirzner

(1997) Entrepreneurial discovery and the competitive market process: An Austrian approach. Journal of Economic Literature 35: 60–85.

40.

Lever

(2002) Correlating the knowledge-base of cities with economic growth. Urban Studies 39(5–6): 859–870.

41.

Malecki

(1994) Entrepreneurship in regional and local development. International Regional Science Review 16: 119–153.

42.

Marshall

(1920) Principles of Economics. London: MacMillan.

43.

Maskell

Eskelinen

Hannibalson

. (1998) Competiveness, Localised Learning and Regional Development: Specialisation and Prosperity in Small Open Economies. London and New York: Routledge.

44.

Michelacci

(2003) Low returns in R&D due to the lack of entrepreneurial skills. The Economic Journal 113: 207–225.

45.

Nelson

(1993) National Innovation Systems: A Comparative Analysis. New York: Oxford University Press.

46.

Niebuhr

(2010) Migration and innovation: Does cultural diversity matter for regional R&D activity? Papers in Regional Science 89(3): 563–585.

47.

Parkers

Griliches

(1980) Patents and R&D at the firm level: A first report. Economic Letters 5: 377–381.

48.

Peri

(2005) Determinants of knowledge flows and their effect on innovation. Review of Economics and Statistics 87(2): 308–322.

49.

Polanyi

(1967) The Tacit Dimension. London: Routledge & K. Paul.

50.

Plummer

Acs

(2014) Localized competition in the knowledge spillover theory of entrepreneurship. Journal of Business Venturing 29(1): 121–136.

51.

Qian

(2013) Diversity versus tolerance: The social drivers of innovation and entrepreneurship in US cities. Urban Studies 50(13): 2718–2735.

52.

Qian

Acs

(2013) An absorptive capacity theory of knowledge spillover entrepreneurship. Small Business Economics 40(2): 185–193.

53.

Qian

Haynes

(2014) Beyond innovation: The Small Business Innovation Research program as entrepreneurship policy. The Journal of Technology Transfer 39: 524–543.

54.

Qian

Stough

(2011) The effect of social diversity on regional innovation: Measures and empirical evidence. International Journal of Foresight and Innovation Policy 7: 142–157.

55.

Qian

Acs

Stough

(2013) Regional systems of entrepreneurship: The nexus of human capital, knowledge and new firm formation. Journal of Economic Geography 13(4): 559–587.

56.

Reynolds

Storey

Westhead

(1994) Cross-national comparisons of the variation in new firm formation rates. Regional Studies 28(4): 443–456.

57.

Romer

(1986) Increasing returns and long-run growth. Journal of Political Economy 94(5): 1002–1037.

58.

Romer

(1990) Endogenous technological change. Journal of Political Economy 98(5): S71–S102.

59.

Saxenian

(2002) Silicon Valley’s new immigrant high-growth entrepreneurs. Economic Development Quarterly 16(1): 20–31.

60.

Schumpeter

(1934) The Theory of Economic Development. Oxford: Oxford University Press.

61.

Wennekers

Thurik

(1999) Linking entrepreneurship and economic growth. Small Business Economics 13(1): 27–56.