Global cities and innovation: a local melting pot for knowledge flows, cosmopolitanism and human capital
Koen Wolvers 10457623 15-8-2018
MSc. Business Administration: International Management Track University of Amsterdam
Final Draft Thesis
Abstract
Given the alleged importance of global cities in the worldwide knowledge production, remarkably little studies have quantitatively investigated the interplay between global cities and innovation. This paper investigates the relationship between the unique characteristics of a global city - a high degree of international connectivity, a cosmopolitan environment and a high degree of human capital - and the innovative output of inventors in the global cities. We hypothesize that global city connectivity will attract R&D activities and make it easier for inventors to connect with distant knowledge areas. Next to this, we argue that the cosmopolitan environment will enhance creativity and recombination opportunities. Moreover, we argue that global cities’ high degree of human capital helps in understanding and processing
knowledge. In turn, we hypothesize that these factors will enhance the innovative performance of inventors in global cities. Analysing USPTO patent data of European and North-American global cities during the period 2001 to 2005, our results however suggest no significant relationship between global city characteristics and the
innovative performance of inventors in global cities. This research contributes to the scientific literature by providing a novel theoretical approach that can be consulted and built upon by future researchers and policymakers. Moreover, we provide interesting suggestions for future researchers.
Key words: global cities; connectivity; external knowledge sourcing; innovation;
Statement of originality
This document is written by Student Koen Wolvers who declares to take
full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original
and that no sources other than those mentioned in the text and its
references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents.
Table of content
1. Introduction……… 5
2. Literature review……… 8
2.1 External knowledge sourcing………. 8
2.2 Geographic proximity and knowledge spillovers………... 10
2.2.1 Global knowledge networks……… 12
2.2.2 Specialized regions……… 13
2.3 The concept of global cities……….. 14
3. Theoretical framework……….. 19
3.1 Global city connectivity and innovation………... 19
3.2 Global city cosmopolitanism and innovation……… 20
3.3 Global city human capital and innovation………... 22
4. Sample selection, data collection and variables……… 24
4.1 Sample selection……….. 24 4.2 Data collection……….. 25 4.3 Variables……….. 27 4.3.1 Dependent variable………... 27 4.3.2 Independent variables………. 27 4.3.3 Control variables………. 28 5. Results………. 29 5.1 Descriptive statistics………. 29 5.2 Models……… 32
6. Discussion and conclusion………..……….. 33
7. Limitations and future research……… 38
8. References.……… 39
9. Appendix….……… 46
List of figures Figure 1. Conceptual model……… 24
List of tables Table 1. Overview sample by tier……… 25
Table 2. Overview patent distribution………. 27
Table 3. Descriptive statistics……… 31
1. Introduction
In today’s increasingly globalized world, firms more and more face competition both within and across country boundaries. In order to keep up with this international competition, firms are required to be open to external knowledge sources that can provide firms access to new ideas and technologies that are not available internally. This can allow for recombination opportunities, which enhances the creation of new knowledge (Carlo et al., 2009; Laursen et al., 2009). The creation of new knowledge can be crucial for the survival of business as innovation allows organizations to maintain their competitive position in the face of new emerging and disruptive technologies (Christensen, 1997).
Yet not all locations provide the same opportunities for external knowledge sourcing. One dimension along which locations can differ is geographic proximity. It has been long argued that this type of proximity can provide firms with an increased access to external knowledge via knowledge spillovers - a positive type of externality that could arise as a result of actors (e.g. suppliers, customers, universities,
competitors, individuals) being closely located together. Geographic proximity would also enhance the development of formal and informal networks (Audretsch &
Feldamn, 1996; Marshall, 1890; Owen-Smith & Powel, 2004). For instance, locating near other organizations will increase the likelihood for employees of different organizations to meet up by chance (i.e. at a local event), exchange knowledge and built up ties (Marquis, 2003; Putnam, 2000).
Though, the ability to reap benefits from geographic proximity differs per firm. For example, some firms might not have the sufficient knowledge base that is required to process and exploit knowledge spillovers successfully (Boschma, 2005; Boschma & Lambooy, 1999; Perez & Soete, 1988). Others might end up with a negative knowledge spillovers balance in case they ‘leak’ more knowledge into a
region than they absorb (Alcácer & Chung, 2007). Yet others might fall into a ‘technology trap’ and end up with a restricted technological path (Ahuja & Lampert, 2001; Levintal & March, 1993). This becomes the case when actors within a region become too dependent of each other and mostly circulate knowledge among
themselves. This could hurt their innovative capacities as they keep looting knowledge from the same pool with little refreshments from outside the region (Boschma & Frenken, 2011; Pouder & St John, 1996). In order to prevent or overcome the latter, it is important that organizations or regions built global knowledge networks, as these networks can serve as vessels for flows of resources and knowledge in an out of different knowledge areas (Asheim & Coenen, 2006; Bathelt et al., 2004).
In this paper, we will investigate the global city, a specific type of location that has developed unique characteristics overtime largely due to the constant co-location of (international) firms (e.g. Multinationals and Advanced producer services firms) who attracted to the global city by the urge to serve or get services from other firms. For example, MNEs often seek support from APS-firms to support their overseas operations whereas APS-firms seek for clients such as MNEs in the global city. This constant interaction has resulted in an abundance of large international firms within the boundaries of a global city (Dunning and Norman, 1983). In turn, this abundance has implications on the characteristics of a city. First, because these
international firms have far and wide stretching international networks, it is argued that these cities are highly connected to other locations. This would enable global cities to serve a gateway function for local firms, allowing them to connect to a wide and diverse amount of actors across international boundaries (Sassen, 2002; Glückler, 2007; Perri & Scalera, 2017; Belderbos et al., 2017). Moreover, international firms
employ well-educated expatriates from different parts in the world (Martinez & Jarillo, 1989), which contributes to the human capital and cosmopolitan character of global cities (Findlay, 1996; Hannerz, 1996; Beaverstock, 2002; Warf, 2015). As highlighted in the literature, this results in the following distinctive global city characteristics: 1) a high degree of international connectivity, 2) a cosmopolitan environment and 3) a disproportionate share of human capital (Beaverstock et al., 2002; Derudder et al. 2010; Sassen, 2001; Taylor, 2001, Goerzen et al., 2013; Belderbos et al., 2017).
Although some studies have examined the relationship between global city characteristics and the business environment in the city (e.g. Goerzen et al., Belderbos et al., 2017), remarkably little studies investigated the relationship between global city characteristics and the innovative environment in the city. To our knowledge, studies that have examined the relationship between locations and innovation have mostly focussed on industrial clusters or knowledge hotspots (e.g. Boschma, 2005; Malmberg & Maskell, 2006). Though, given the alleged importance of global cities in the
worldwide knowledge production (e.g. Jacobs, 1969; Florida et al., 2017), it is a topic that requires more research. We will examine the relationship between each global city characteristic and the innovative output in global cities. Our research question is the following:
To what extend do the unique characteristics of a global city enhance the innovative performance of inventors?
In order to answer the research question, we analyse all United States Patent and Trademark Office (USPTO) patents that have been granted to inventors resident in European and North-American global cities during the period 2001-2005. The classification of global cities was based on the rankings of the GaWC (2001). Our
results do not show a significant relationship between global city characteristics and innovative output of inventors. Even though we did not find the relationship we expected, our work does have contributions to the scientific literature and
implications for policymakers. From a theoretical perspective, we demonstrate how the characteristics of a global can influence the innovative capacities of inventors in the city. Considering that little research has been conducted on the relationship
between global city characteristics and their innovative climate, our framework can be consulted by researchers who wish to investigate this topic. Furthermore, our research raises new questions that provide interesting avenues for future research. Is our expected relationship more dominant in global cities from emerging countries? And, are global cities in fact the innovative machines many assume them to be? Moreover, policymaker can use this paper to gain information on the relationship between global city characteristics and innovation.
The remainder of this paper is structured as follows. In the next section, the relevant literature on external knowledge sourcing, geographic proximity, knowledge spillovers, global knowledge networks and global cities is discussed. Subsequently, a theoretical framework with the development of hypothesis is stated in section 3. Next, we explain how we collected the data, selected the sample and created the variables in section 4. Hereafter, the results of the statistical analysis are presented in section 5. Furthermore, in section 6, the results are discussed and conclusions are drawn. Finally, in section 7 limitations of our research are discussed and suggestions for future research are given.
2. Literature review
The importance of external knowledge sourcing has long been recognized in the economic literature as an important determinant of innovation. It is argued that, nowadays, firms are seldom able to successfully compete in the global markets by solely relying on their internal innovative efforts now that the developments of new technologies and the changing business environment have accelerated (Carlo et al., 2009; Vega-Jurado et al., 2009; Kang & Kang, 2009; Chesbrough, 2003). In order for firms to compete successfully in this fast changing world, an increasingly important capability has become external knowledge sourcing – the ability to access new ideas and technologies from beyond a firms’ boundaries (Laursen & Salter, 2006;
Rosenkopf & Nerkar, 2001; Katila & Ahuja, 2002). External knowledge sources can provide firms access to new opportunities and enables a firm to integrate
complementary knowledge sets from external agents (Teece, 1986; March, 1991). In turn, this can enhance a firms’ innovative capacity since novel innovations are often a result of combining existing technologies (Van de Bergh, 2008). Internal knowledge sourcing on the other hand often fails to provide solutions to new technological problems (Postrel, 2002).
Empirical evidence shows us that external knowledge sourcing has been a growing phenomenon over the past decades. For example, in most Organization for Economic Co-operation and Development (OECD) countries, since the 1980s, the total amount spent on external R&D has increased gradually relative to the total amount of business expenditures (Howells, 1999; Bonte, 2003). Moreover,
Hagedoorn (2002) shows that the number of inter-firm R&D partnerships has grown significantly from the 1960s to the 1990s. These observations are accompanied by a decreasing strategic importance of internal R&D activities and a decrease in the number of internal R&D departments (Chesbrough, 2003).
Though, internal R&D does still have an important role in the innovation process of a firm. Chesbrough (2003) argues that external knowledge sources and internal R&D activities are in fact complementary. Where external knowledge can serve as an opportunity to learn and enhance technology, internal knowledge would enable firms to identify, acquire and exploit external knowledge. There is empirical evidence that supports Chesbroughs’ idea of complementarities between internal and external knowledge. In a study of contracting behaviors of major independent R&D labs during the period 1940 – 1945, Mowery (1983) and Mowery and Rosenberg (1989) found that the demand for external R&D contracts was higher when firms had the expertise to capably identify their needs for external research. This line of
reasoning has been extended by Cohen and Levinthal (1989, 1990) who argue that internal R&D can both generate innovations as well as increase the absorptive capacity of a firm; identified as the ability of a firm to identify, process and exploit new knowledge. Moreover, studies of Arora and Gambardella (1990, 1994) show that the more firms conduct internal R&D, the larger the number of external linkages aimed at acquiring technology are.
2.2 Geographic proximity and knowledge spillovers
Some locations provide better opportunities for external knowledge sourcing than others. One dimension along which locations can differ is geographic proximity, referring to the spatial distance between actors in bot relative and absolute terms (Boschma, 2005). Back in 1890, Marshall already argued that geographic proximity could enhance knowledge spillovers - a type of positive externality that could arise as a result of actors (e.g. suppliers, customers, universities, competitors) being located
closely together (Audretsch & Feldamn, 1996; Marshall, 1890). Four mechanisms can be identified to understand this phenomenon. First, locating near other organizations increases the likelihood of chance encounters. For instance, participation in local events increases the chance of employees of different organizations to interact and exchange knowledge (Marquis, 2003; Putnam, 2000). Second, proximity makes it easier for firms to monitor each other activities. This allows firms to prioritize their R&D on promising ideas of competitors (Porter & Stern, 2001; Sorenson & Stuart, 2001). Third, a geographic concentration of firms often leads to informal and professional networks that help channel diverse knowledge among local actors (Saxenian, 1996; Owen-Smith & Powel, 2004). Lastly, common interests of firms such as improving the ease and efficiency of absorbing knowledge within an area can lead to joint initiatives such as local institutions or local events that stimulates the development of the region (Malmberg & Maskell, 2006).
Though, geographic proximity is not per se beneficial for every firm, as geographic proximity will not always lead to a positive knowledge spillover balance. This is the case when firms ‘leak’ more knowledge into a region than they absorb (Mariotti et al., 2010). Therefore it would only be beneficial for companies to locate their exploiting activities into knowledge hotspots and their competence-creating activities in less core regions (Cantwell & Mudambi, 2005). Moreover, as touched upon earlier, it is important that firms have a strong internal knowledge base in order to identify, process and exploit knowledge. Therefore it is argued that, in order to successfully transfer knowledge, firms or individuals are required to have a certain base level of knowledge (Boschma & Lambooy, 1999). For that matter, in order to successfully absorb and benefit from a knowledge spillover, cognitive
and institutional) can also influence the ability of a firm to benefit from knowledge spillovers. Like geographic proximity, these forms of proximity can also bring actors within and between organizations together and can even function as a substitute spatial proximity (Boschma, 2005).
Moreover, too much spatial proximity can lead to regional ‘lock-in’ when regions, especially specialized regions, become too much inward looking. Lock-in occurs when actors within a region merely circulate existing knowledge among each other without refreshments from outside the region. This could lead to a restricted technological path and turn hot spots into ‘blind spots’ as actors within the region keep looting from the same knowledge pool, which could hurt the innovative
capacities of these actors (Boschma, 2004, 2005; Boschma & Frenken, 2011; Pouder & St John, 1996). Therefore, it is argued that regions must not solely rely on its own local knowledge base but also need to attract external knowledge inflows (Asheim & Coenen, 2006; Bathelt et al., 2004).
2.2.1 Global knowledge networks
Since the attraction of external knowledge inflows is so important, researchers have emphasized the role of global knowledge networks, they can ‘de-lock’ a region by serving as channels for resource flows in and out of different knowledge areas (Martin & Sunley, 2006; Amin, 2002; Davenport, 2005; Gertler & Levitte, 2005). Lorenzen and Mudambi (2012) distinguish between two types of global linkages that support the global knowledge network. One type, described as “weak-ties”, refers to linkages that are hold by individuals based on friendships or other kinds of personal
by birth; these people can leverage their shared national background as advantage to build trust (Meyerson et al., 1996). The other type is what the literature refers to as “pipelines” (Bathelt et al., 2004), are those links created by organizations through ownerships and strategic alliances.
Links through personal relationships usually offer more scope for diversity and knowledge recombination within different technologies and industries than organizational pipelines. Links through organizational pipelines are usually strategic, coordinated from the top and managed by profit-oriented organizations (Lorenzen & Mudambi, 2012). These firms tend to be very industry focused, i.e. linking with suppliers, buyers and partners (Christensen, 1997). Therefore, these kinds of links are associated with less technological diversity (Feldman & Audretsch, 1997). On the other hand, personal relationships offer more connections outside of the firms industry, which provides a wider and more diverse range of inputs (Lorenzen & Mudambi, 2012). Particularly networks with linkages to extra-regional technological hotspots can be of great value as these often provide knowledge that is not available locally (Bathelt et al., 2004; Cantwell & Santangelo, 1999). The characteristics, emergence and development of these extra-regional areas are briefly discussed in section 2.2.2.
2.2.2 Specialized regions
Literature on national innovations systems shows us that, due to competition of nations around the world, different specialized regions have emerged over time in various technologies and industries (Cantwell & Jane, 1999; Porter, 1990; Engel, 2015). In these places, a particular type of economic activity establishes a stronghold
as knowledge becomes embedded in individual skills of its people, and in the routines and procedures of its organizations (Malmberg & Maskell, 1997). Much like the development of business clusters, as described by Porter (1990), these specialized regions initially emerge by ‘chance’ or ‘historical events’ and gradually develop a comparative advantage over other locations by attracting more organizations and competencies to the region (Cantwell & Janne, 1999). In the literature, these
specialized regions are often referred to as “knowledge hotspots” (Bathelt et al., 2004; Pouder and St. John, 1996), “international centres of excellence in research and innovation” (Cantwell & Janne, 1999) or “clusters of innovations” (Engel, 2015).
In this paper, we will focus on global cities. Much like specialized regions, these locations have also been able to attract many organizations within its
boundaries. Yet, whereas first is highly specialized in a particular type of industry or business (e.g. Silicon valley in technology or the Hollywood in film productions), the latter has attracted organizations from a wide array of industries or businesses
(Goerzen et al., 2013; Glückler, 2007). In the next section, we discuss the concept of global cities and also introduce the research question.
2.3 The concept of global cities
The concept of global cities comes from urban and geography studies and was originally introduced by Sassen (1991). She describes global cities as the command and control centres of this planet, being able to apply a considerable amount of influence within the globalized world. This influence arises from its disproportional share of worldwide “command points”, i.e. office networks, that has accumulated within its boundaries. Global cities are host to a disproportional share of global advanced producer services (APS) firms (e.g., accounting, advertising, finance,
insurance and law), who have far stretching networks worldwide (Sassen, 2001). These firms are drawn to global cities by the urge of reducing coordination costs (Arzaghi and Henderson, 2008). Some of the most salient examples of global cities are New York, London and Tokyo (A.T. Kearney, 2016).
Global cities should not be confused with “world cities” or “megacities” as they can be intrinsically different. A world city refers to a type of city that we have seen over centuries (e.g. Rome, Athens) and the term megacity merely refers to the size of the city. In the lights of this, it can be said that many of today’s global cities are also world cities or megacities. However, not all world or megacities are global cities (Sassen, 1991). Global cities differentiate themselves by their high degree of connectivity, being at the intersect of various flows such as ideas, goods, services, travellers, brains, investments and knowledge, giving them a central and leading role in global networks (Perri & Scalera, 2017). Next to this, they typically have a
cosmopolitan environment and a disproportionate share of human capital, expressed in skilled workers, innovative companies and high quality institutions (Beaverstock et al., 2002; Derudder et al. 2010; Sassen, 2001; Taylor, 2001, Goerzen et al., 2013; Belderbos et al., 2017).
These distinctive characteristics are largely a result of a constant interplay between firms and locations, in which firms are attracted to the global city by the need to serve or to get services from other firms. For example, MNEs operations require a global supply of business services that enable them to support their activities abroad. Therefore they would locate in a global city, which houses many APS firms who could serve them. Vice versa, APS firms are attracted to the global city by the demand for services coming from MNEs. This clustering pattern explains why global cities are host to a disproportional share of APS -and MNE firms (Dunning and
Norman, 1983). In this process, these MNEs and APS firms are able to weave global cities into their global networks, largely contributing to global cities’ high degree of connectivity. By this means, global city connectivity is often estimated by observing the number of APS firms in the city (Taylor, 2001; Belderbos et al., 2017).
Another measure that is often used to measure global city connectivity is international airline flows (e.g. Bel & Fageda, 2008; Matsumoto, 2004). According to Keeling (1995) this measurements is used since they are 1) readily available, 2) the most visible form of connections between cities, 3) measure face-to-face meetings, 4) are the preferred mode of transportation for expatriates, 5) are an important
component of a city’s aspiration towards world status.
Furthermore, APS companies typically make use of expatriates as part of their coordination and control strategy (Martinez & Jarillo, 1989). In senior management positions, expatriates are an essential shackle between their companies and the international markets. Since these expatriates are from “elsewhere”, they often bring with them their internationally established networks, cultural practises and social relations. This allows them to monitor the developments in other parts of the world. Beyond this, such global labour brings highly specific knowledge, skills and
cosmopolitan practices into the city, which contributes largely to the transformation of cities. While expatriates initially come to the global city through employment by international companies, such workers later become locally available for other firms. The existence of such a skilled labour pool in turn attracts more (international)
business into the city. Next to this, since expatriates bring their cultural practises with them, they can shape the socio-cultural environment of a city, making it more
1996; Beaverstock, 2002). This largely contributes to the cosmopolitan character that global cities have (Warf, 2015).
Global cities are naturally ethnically diverse and racially tolerant and values such as autonomy, freedom, egalitarianism and mutual respect are typically placed highly (Mosterin, 2005; Florida, 2002). Next to this, many of their citizens have access to a wide range of information and are exposed to diverse cultures through museums, food, schools, tourists and ethnic festivals (Warf, 2015). Because of this, these cities offer opportunities for creativity, excitement and novel experiences. These opportunities are often linked to high-tech information industry, research and
development, arts, fashion, media, and the music businesses. Economists and geographers often measure cosmopolitanism in terms of the degree of foreign-born inhabitants (Short, 2004). Scholars in other fields like philosophy, sociology and political science focus on measures such as level of freedom and egalitarianism (Sevincer et al., 2015).
The last global city characteristic that needs to be discussed in this literature review is the high degree of human capital. As touch upon earlier, global cities offer international firms a wide pool of skilled expatriates that have been drawn to the city. However, other factors also stimulate this high degree of human capital. Compared to other cities and regions, global cities are host to many Universities and Colleges (Florida et al., 2017), who can have a big contribution in the regional development of a city (Haapanen & Tervo, 2012). Universities can collaborate with local actors (e.g. local government, industry and citizens) and help them bridge the gap between science and business. They can do this by applying some of their research into practise and by coaching local actors. Since Universities attract and educate students, they contribute to the local pool of talent that is present in global cities (Markkula &
Kune, 2015). Next to this, after students graduate, they are generally more likely to stay in the city than they would in other areas (Florida et al., 2017; Haapanen & Tervo, 2012). Besides students, Universities also attract engineers, business specialists and scientists, thus stimulating the human capital that is in the region (Huffman & Quigley, 2002).
Goerzen et al. (2013) investigated the relationship between global city characteristics and liability of foreignness (LOF) faced by MNE subsidiaries,
described as the additional costs of doing business abroad (Zaheer, 1995). They argue that global city connectivity reduces LOF because it should ease the transfer of capital, people, goods and information to and from the subsidiary (Friedmann, 1986). This connectivity also has an effect on the cosmopolitan environment in the city since it exposes the local population to international stimuli (Sassen, 2002). This in turn would lower the hostility towards foreign firms since cosmopolitanism is associated with tolerance towards outsiders (Warf, 2015).
Although there are studies that have investigated the relationship between global city characteristics and the local business climate (e.g. Goerzen et al., 2013), surprisingly little literature examined the relationship between global city
characteristics and the local innovative climate in global cities. Studies that link location to innovation mostly focussed on industrial cluster or knowledge hotspots (e.g. Malmberg & Maskell, 2006; Boschma, 2005). Though many of the same mechanisms might apply to global cities, it is a topic that deserves more research, especially given their alleged importance the global knowledge economy (e.g. Jacobs, 1969; Florida et al, 2017). By examining the relationship between each global city characteristic and innovative output in global cities, we aim to provide an answer to
the following research question: to what extend do the unique characteristics of a global city enhance the innovative performance of MNEs and other inventors?
3. Theoretical framework
3.1 Global city connectivity and innovative performance
Inventors often innovate by recombining existing knowledge with new knowledge. Therefore the quantity and quality of produced innovations depends on the diversity of the knowledge stock that is available to the inventor (Kogut & Zander, 1992). As local knowledge tends to be fairly homogeneous within a region, it often required to turn to external knowledge sources in order to gain new ideas, insight and expertise (Rosenkopf & Almeida, 2003). Firms that only source their knowledge locally might end up in a restricted technological path as they keep on drawing from the same knowledge pool. This could prevent them from recognizing opportunities in other markets (Lamboy & Boschma, 2001). Next to this, local search can result in an overreliance on local innovation systems, which can lead to learning traps and eventually cause a reduction in a firms’ tendency to try-out new ideas (Ahuja & Lampert, 2001; Levintal & March, 1993). Therefore, distant ties can be important to increase the diversity of knowledge sources (Giuliani, 2013).
By this means the role of global knowledge networks is emphasized, as they can serve as channels for resource flows in and out of different knowledge areas (Lorenzen & Mudambi, 2012), thereby facilitating the access to distant and more diverse technologies. Particularly networks with linkages to extra-regional
technological hotspots can be of great value as these often provide knowledge that is not available locally (Bathelt et al., 2004; Cantwell & Santangelo, 1999).
Compared to other locations, global cities provide more opportunities for creating a global knowledge network. These cities are characterized by a high degree of international connectivity, thus having many links with other locations (Goerzen et al., 2013). This position enables them to serve a global bridging function, allowing for the development of international ties to distant knowledge areas (Sassen, 2002;
Glückler, 2007). Because of this, local inventors are more easily able to connect with distant knowledge sources (Perri & Scalera, 2017). Furthermore, Belderbos et al., (2014) shows that these opportunities in turn attract a vast number of R&D activities towards the city. This can allow local inventors to benefit from local knowledge spillovers if the cognitive distance is too large. Namely, interactive learning processes require at least cognitive proximity besides spatial proximity (Boschma, 2005).
Both of these connectivity functions can thus contribute to a deep and diversified pool of knowledge within the boundaries of global cities, either through increased access to distant knowledge areas or knowledge spillovers. Therefore it is hypothesized that global city connectivity has a positive effect on the innovative output of its inventors. Namely, this access to a deep and diversified pool of knowledge would provide inventors with opportunities for recombining existing knowledge.
H1: Global city connectivity increases the innovative output of its inventors
3.2 Global city cosmopolitanism and innovative performance
Global cities are typically more ethnically diverse and tolerant towards different cultures than most other places (Warf, 2015). Because many of its residents come from different backgrounds, various different cultures are exposed in the city through media, people, food, ethnic festivals, information, music and many other factors
(Warf, 2015; Sevincer, 2017). This diversity tends to come with cosmopolitan cognitive skills; namely, it enables people to understand the perspectives of one another, to reconsider opinions and reduces the urge to force peoples own view on others (Warf, 2015; Marcus et al., 1995). For this reason, global cities can be called cosmopolitan cities.
This environment of openness and tolerance is to attract the so-called “creative class”, as according to Florida (2002). This class is comprised of skilled professionals who’s economic function it is to create new ideas, the class ranges from scientists, engineers and University professors to poets and architects. Next to this, they place high value on tolerance towards ethnic differences and alternative lifestyles,
recreational opportunities and amenities. This class is ought to be responsible for technological innovations and the formation of new industries (Florida, 2002).
Although this claim of the creative class by Florida (2002) is debated (e.g. Scott, 2006; Asheim & Hansen, 2009), but not directly relevant for this section, it has been shown that urban diversity stimulates economic growth through religion
(Noland, 2005), immigrants (Ottaviano & Peri, 2005) or ethnic groups (Quigley, 1998). Thus, it can be argued that cosmopolitanism (i.e. tolerance, diversity and openness) stimulates economic growth.
Beyond the creative class, Florida et al. (2017) argue that this cosmopolitan environment can stimulate creativity in general, which would help overcome roadblocks in the creative process. He states: “It is common for the creative mind to return to ordinary life in the moments when it cannot solve an important problem. In the urban environment, there are many more diverse, but related, influences that might trigger a solution via what is commonly understood as a eureka moment”. If this is the case, this should enhance the innovative output of the global city as the
entire process of human creativity is considered a crucial element for innovation (Amabile, 1988)
Alesina and La Ferrara (2005) argue that cultural diversity within the
workforce may lead to creativity and innovation since it involves a recombination of abilities and knowledge. Similar conclusions are drawn by Hunt el at. (2014) who find that ethnic diversity contributes to the innovative output of companies by increasing the levels of creativity. Nieburh (2006), who investigated the relationship between diversity in the workforce and innovative output, shows that diversity has the strongest impact on innovative output among employees with high levels of
education. We argue that this outcome is especially relevant for global cities, as these cities have a large internationally skilled labour force drawn to the city by MNEs APS firms and (Findlay, 1996; Hannerz, 1996; Beaverstock, 2002).
We argue that the cosmopolitan environment of global cities stimulates creativity and allows for recombination opportunities. This would in turn enhance the innovative output of its inventors.
H2: Global city cosmopolitanism increases the innovative output of its inventors
3.3 Global city human capital and innovative performance
A large and diversified pool of accessible knowledge is essentially useless when nobody has the ability to understand and use the knowledge effectively. In order to understand, process and create new knowledge, it is of foremost importance that people have the right skills and tools to do so (Markkula & Kuhne, 2015). Global cities have drawn knowledgeable people through labour migration, but have also created human capital “on site” via education, training, on the job learning and broad processes of socialization (Storper & Scott, 2009). This has created a strong
concentration of talent and strong institutions (e.g. Universities) in global cities (Doel & Hubbart, 2002).
Universities can contribute to the regional innovation system by actively coaching local actors in their innovative efforts. By doing this, they are able to apply their research into practice, creating an active science-society dialogue. Next to this, University researchers keep a region plugged into the global knowledge network because of their outward orientation (Owen-Smith & Powel, 2004). Perhaps just as important, Universities attract students to cities, and prepare them for taking an active role in society after graduating (Markkula & Kune, 2015). This contributes to the local knowledge pool since students often tend to stay in the cities, which is especially the case with global cities (Florida et al., 2017)
Beyond this, Universities tend to attract innovative business to the city. Jaffe et al. (1993) finds that innovative activities are heavily concentrated around
University regions. Locating closely to Universities can increase the innovative output of the inventor, as shown by Malecki (2005). Locating near Universities can result in local knowledge spillovers that can be exploited by inventors.
Since talent and institutions are highly concentrated within global cities, and since they can have an important contribution to the innovative climate of a region, it is expected that global city human capital increases the innovative output of inventors.
4. Sample selection, data collection and variables 4.1 Sample selection
The overall sample population is initially based on all cities in the continents North America and Europe. We choose to focus on only these two continents because it would not be realistic to collect the required data of all global cities in the world, considering the time span of this thesis. For example, collecting reliable
demographics data of some Asian or African cities would be a time consuming process. Therefore, we narrowed it down to a sizeable sample of global cities from which the required data can be obtained. Two regions where then possible: the European Union and the United States. Both of these regions host many global cities and have the required data available. The specific time period that was chosen is 2001 to 2005 because this period fits well with the available data and is relatively recent.
Figure 1. Conceptual model Global city Connectivity
Global city cosmopolitanism
Global city human capital
H1
H2
Innovative output global city inventors
Since not all (global) cities have the same amount of global city traits, we chose to select only the most global cities in North America and Europe, based on the global city index as listed by the Globalization and World Cities research group and network (GaWC) (2001). This resulted in an initial research sample of 625 cities (125 cities times 5 years). This index also classifies cities as Alpha, Beta, Gamma or high-sufficiency tier, based on the amount of global city traits the city has. Alpha cities in this case have the most global traits, sufficiency cities the least. Since
high-sufficiency cities do not meet the requirements of global city status, such as a
disproportionate share of worldwide “command points” (Sassen, 2002), 185 cases were dropped (37 cities times 5 years). Due to data unavailability, Moscow and Istanbul cases were also dropped. The above filtering resulted in a research sample of 430 cases (86 cities times 5 years). Table 1 shows the
sample cities by tier. Out of this sample, 34 cities were located in North America, corresponding to 7 Alpha, 12 Beta and 15 Gamma tier cities. The most Alpha cities are located around the east- and west coast of the United States. Europe contributes to the sample with 52 cities, corresponding to 15 Alpha, 23 Beta and 14 Gamma tier cities. The most Alpha cities are located in Western Europe.
4.2 Data collection
Following other studies about innovative activity (e.g. Walker et al., 1997; Ahuja, 2000, Perri & Scalera, 2017), the patent inventor database from the United States Patent Trade Office (USPTO) is used as main source of data. The choice for the
Table 1. Overview sample by tier
Tier City
Alpha New York Miami Dublin Chicago Brussels Amsterdam Los Angeles Paris Milan Washington Zurich Stockholm San Francisco Frankfurt Vienna Luxembourg Lisbon Barcelona London Toronto Madrid Warsaw
Beta Boston Minneapolis Stuttgart Atlanta Denver Munich Dallas Seattle Manchester Houston St. Louis Dusseldorf Philadelphia San Diego Lyon Cologne Edinburgh Copenhagen Hamburg Budapest Prague
Athens Oslo Rotterdam
Berlin Rome Sofia
Helsinki Antwerp Zagreb Bratislava Montreal Tallinn Vancouver Kiev
Gamma Cleveland Cincinnati Raleigh Detroit Kansas City Orlando San Jose Tampa Columbus Phoenix Charlotte Pittsburgh Austin Baltimore Krakow Bologna Edmonton Marseille Lausanne Nuremberg Belfast Turin Gothenburg Leipzig
Malmo Leeds Bilbao
USPTO Patent database ensures us that innovations for which a patent has been granted have been thoroughly evaluated and are fairly novel. This makes it an adequate choice for measuring inventive efforts of inventors (Archibugi and Coco, 2005). Namely, patent data is regarded as a well understood and widely used measurement for innovative activity, adopted in various studies (e.g. Ahuja, 2000; Juda, 2007). Patent data provide information on the technical characteristics of inventions as well as ownership of inventions. Next to this, the geographic span of patent data is wide, covering all countries with patent systems. Though, the use of patent data as measurement for innovative activities does have some flaws. For example, patent data does not capture inventions that have not been patented. Also, patenting procedures or systems can be different among countries, which has to be accounted for in the measurement (de Rassefosse & de la Potterie, 2018).
The specific dataset that is used is called “Disambiguation and co-authorship networks of the U.S. patent inventor database (1975 - 2010)” provides inventor-level data and is distributed by the Harvard Dataverse Network (Li et al., 2014). Out of this dataset, we first of all only selected the patents that were applied for after the year 2000 and before the year 2006. Since it could take some time before patents get granted, we consider the application as the moment at which an invention was produced, even if it has been granted at a later time. For this reason also, to ensure that none of the ‘to be granted patents’ are excluded, the latest year of measurement is 2005. After this filter, we selected all the unique patents and assigned each to a city based on the residence location of the lead inventor that is on the patent. Thus, in case that there are multiple inventors on the patent, only the residence location of the lead inventor is accounted for. By doing this, we ensure that the same patent is not counted more than once or assigned to different cities. We also accounted for cities that are
listed under different names like Köln/Cologne. For example, whenever a patent is initially listed under Köln, the unique patent is assigned to Cologne.
The distribution pattern of patents among the sample cities is mostly in line with the different tiers. The highest patent concentrations are generally in Alpha tier cities and the lowest in Gamma tier cities. Exceptions to this pattern are Beta or Gamma tier cities where at least ten percent of the workforce consists of technology workers and cities located in Eastern Europe.
4.3 Variables and measures 4.3.1 Dependent variable
Innovative activity in global cities. The innovative activity of a city was
captured by counting the number of successful patent applications after the year 2000 and before 2006. This would mean that a patent is assigned to a city if the lead inventor on the patent is a resident of the city. As mentioned before, the USPTO inventor patent database is used to count the number of patents per city.
4.3.2 Independent variables
Global city connectivity. Global city connectivity is measured as the total
number APS firms residing in a city in each year. Counting the total number of global APS firms is often used as a proxy for city connectivity (e.g. Belderbos et al., 2017; Taylor, 2001). APS firms are known to have a large global network with connection to many locations (Taylor, 2001; Belderbos, 2017). Next to this, they often employ a lot of expatriates who also contribute to the global cities connectivity by bringing
Table 2. Overview Patent Distribution
Tier City Patents
Average Patents Alpha - 277,4
Highest Alpha New York 2001 1342 Lowest Alpha Luxembourg 2005 4
Average Patents Beta - 170,4
Highest Beta San Diego 2001 1306
Lowest Beta Talinn 2003 1
Average Patents Gamma - 93,7
Highest Gamma Cincinnati 2003 489
their internationally established network with them (Beaverstock, 2002). The data was retrieved from the Globalization and World Cities Research Network (GaWC) (2000), specifically ‘dataset 11’. This dataset only provides the total APS firms in the year 2000. Because it is expect that these total have not undergone significant changes in the other sample years, the totals are kept constant in this research.
Degree of cosmopolitanism. Geographers and economists often conceptualize
cosmopolitanism as the degree to which a city is ethnically diverse, and it is typically operationalized as the proportion of the inhabitants who are foreign born (Short, 2004). This information was obtained from ‘citypopulation.de’, a website that links to various cities-and state websites that contain the demographics of a city.
Degree of human capital. As a proxy for the degree of human capital, we
counted the number of top 1000 Universities in a city or just outside of the city. This proxy is used because Universities have an important role in educating and training local actors. Data was retrieved from Timeshighereducation.com.
4.3.3 Control variables
To ensure robust results, several control variables are included in the analysis. First, Population was added as control variable since it is likely that cities with larger populations naturally produce more patents. Data on the city populations was
obtained from ‘citypopulation.de’, a website that links to various cities-and state websites that contain the demographics of a city. Second, as a proxy for innovative performance in previous years, cumulative patent output was added. This control variable was added since it is likely that a cities past innovative performance influences future innovative performance. This variable counts all the successful patent applications from before 2001. Data is drawn from the USPTO inventor patent database. Third, technological hotspot is added as a dummy to the control variables.
In technological hotspots, technology workers agglomerate, which lead to a larger patent output naturally (Florida et al., 2017). In this analysis, a technological hotspot is identified as a city where more than ten percent of its workforce is active in the technological industry. For Europe, this information was obtained from Eurostat (2017), “Statistics on European Cities”. For American cities, this information was obtained from the American Community Survey. Canadian cities have also been checked, but do not contain technological hotspots. Fourth, local wage level in terms of average after-tax salary is added. This data was retrieved from Numbeo.com. Fifth, we controlled for the region a city is located in. Differences in for example regional infrastructure or proximity to other urban areas could possibly influence the dependent variable. Sixth, we control for output in different city tiers. Lastly,
previous years are added as dummy variable since it is likely that for example 2001
output strongly relates to 2002 output.
5. Results
5.1 Descriptive statistics
The descriptive statistics of the dependent, independent and control variables are presented in Table 2. We performed a test for multicollinearity by assessing the correlation between the predicting variables. Variables can be multicollinear if they correlate very highly (above 0,7) (Field, 2009). We note that some variables appear to be multicollinear. Namely, APS firms with population, and top 1000 Universities with APS firms. To access whether these variables are indeed multicollinear, we ran a variance inflation factor (VIF) test. This test shows us that multicollinearity is not an issue as the highest VIF score is 7.488, which is below the critical limit of 10
(Graham, 2003). These tests indicate that multicollinearity is not an issue in this thesis.
The descriptive statistics shows us that, on average, the global cities in our sample contained 93 APS firms, 2 to 3 Universities, have population of around 1 million people, a foreign born ratio of around 0.23, and an average salary after tax of around 2700 USD. We can also note that most of the cities are located in the North American region, followed by Western Europe. Central Europe contained the least global cities.
31 ia bl e M ea n St d. D ev ia ti on 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 nt ed Pa te nt s 18 7, 27 21 25 8, 01 25 6 1 ul at iv e pa te nt s 32 06 ,4 30 2 38 81 ,2 52 09 0, 88 8 ** 1 la ti on pe r 10 00 00 10 ,8 07 8 12 ,9 22 49 0, 50 9* * 0, 56 6* * 1 no lig ic al H ot sp ot (Y es ) 0, 20 93 0, 40 72 8 0, 44 3* * 0, 43 4* * -0 ,1 28 ** 1 ry pe r 10 00 U SD 2, 66 86 1, 05 38 5 0, 43 1* * 0, 39 4* * 0, 02 2 0, 32 7* * 1 fir m s 92 ,9 39 8 67 ,7 32 84 0, 55 0* * 0, 55 1* * 0, 73 4* * -0 ,1 16 * 0, 27 9* * 1 ig n bo rn ra ti o 0, 23 10 0, 15 55 3 0, 14 0* 0, 10 1 0, 20 4* * -0 ,2 00 ** 0, 25 6* * 0, 28 7* * 1 10 00 U ni ve rs it ie s 2, 44 19 2, 61 60 9 0, 43 3* * 0, 50 5* * 0, 70 0* * -0 ,0 32 0, 17 4* * 0, 75 9* * 0, 12 1* 1 20 01 0, 20 00 0, 40 04 7 0, 06 1 0, 00 0 -0 ,0 06 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 1 20 02 0, 20 00 0, 40 04 7 0, 04 2 0, 00 0 -0 ,0 03 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,2 50 ** 1 20 03 0, 20 00 0, 40 04 7 0, 00 2 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,2 50 ** -0 ,2 50 ** 1 20 04 0, 20 00 0, 40 04 7 -0 ,0 50 0, 00 0 0, 00 3 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,2 50 ** -0 ,2 50 ** -0 ,2 50 ** 1 20 05 0, 20 00 0, 40 04 7 -0 ,0 54 0, 00 0 0, 00 7 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,2 50 ** -0 ,2 50 ** -0 ,2 50 ** -0 ,2 50 ** 1 ra lE ur op e 0, 04 65 0, 21 08 4 -0 ,0 96 * -0 ,0 89 -0 ,0 87 -0 ,1 14 * 0, 40 1* * 0, 00 6 0, 26 1* * -0 ,0 37 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 1 er n Eu ro pe 0, 10 47 0, 30 64 6 -0 ,2 23 ** -0 ,2 47 ** 0, 02 9 -0 ,1 76 ** -0 ,5 69 ** -0 ,1 35 ** -0 ,1 92 ** -0 ,1 01 * 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,0 76 1 th -A m er ic a 0, 37 21 0, 48 39 3 0, 45 3* * 0, 48 0* * 0, 06 2 0, 49 1* * 0, 43 8* * 0, 06 4 0, 00 2 0, 07 2 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,1 70 ** -0 ,2 63 ** 1 th er n Eu ro pe 0, 06 98 0, 25 50 5 -0 ,1 12 * -0 ,1 58 ** -0 ,1 03 * -0 ,1 41 ** 0, 06 4 -0 ,0 56 -0 ,0 52 -0 ,0 64 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,0 60 -0 ,0 94 -0 ,2 11 ** 1 he rn Eu ro pe 0, 10 47 0, 30 64 6 -0 ,1 40 ** -0 ,1 44 ** 0, 04 7 -0 ,1 76 ** -0 ,3 59 ** 0, 02 3 -0 ,1 25 * 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,0 76 -0 ,1 17 * -0 ,2 63 ** -0 ,0 94 1 te rn Eu ro pe 0, 30 23 0, 45 98 0 -0 ,1 29 ** -0 ,1 16 * -0 ,0 19 -0 ,1 52 ** -0 ,0 62 0, 03 7 0, 07 9 0, 04 4 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,1 45 ** -0 ,2 25 ** -0 ,5 07 ** -0 ,1 80 ** -0 ,2 25 ** 1 ha ti er ci ti es 0, 25 58 0, 43 68 3 0, 20 5* * 0, 22 1* * 0, 37 4* * -0 ,2 36 ** 0, 16 7* * 0, 75 6* * 0, 31 4* * 0, 46 2* * 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 12 4* -0 ,1 13 * -0 ,0 65 -0 ,0 56 0, 14 8* * 0, 02 0 1 ti er ci ti es 0, 43 02 0, 49 56 9 -0 ,0 57 -0 ,0 60 -0 ,0 48 0, 13 0* * -0 ,0 64 -0 ,1 68 ** 0, 04 1 -0 ,0 66 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,0 80 0, 24 0* * -0 ,0 37 0, 03 9 -0 ,1 44 ** -0 ,0 10 -0 ,5 09 ** 1 m a ti er ci ti es 0, 31 40 0, 46 46 4 -0 ,1 32 ** -0 ,1 44 ** -0 ,3 00 ** 0, 08 3 -0 ,0 88 -0 ,5 52 ** -0 ,3 31 ** -0 ,3 64 ** 0, 00 0 0, 00 0 0, 00 0 0, 00 0 0, 00 0 -0 ,0 30 -0 ,1 49 ** 0, 10 1* 0, 01 1 0, 01 4 -0 ,0 09 -0 ,3 97 ** -0 ,5 88 ** 1 or re la ti on is si gn ifi ca nt e 0. 01 le ve l( 2-ta ile d) . rr el at io n is si gn ifi ca nt at 0. 05 le ve l( 2-ta ile d) . e 3. D es cr ip ti ve st at is ti cs ;m ea ns ,s ta nd ar d de vi at io n an d co rr el at io ns
5.2 Models
Models 1 to 4 are presented in table 3. All models have granted patents as dependent variable. Since the dependent variable can only take on whole values, we treat it a count variable. Next to this, we note that the variance of the data is high compared to the mean. Therefore, we have conducted a negative binomial regression on all models. This choice ensures us that these factors are accounted for (Hilbe & Joseph, 2011).
Model 1 only contains the control variables, next to the dependent variable. The results show that this model is significant (LR test = 1624.558, Chi-square = 373.816, P < 0.000). The control variables Cumulative patents, tier and technological hotspot have a significant (P < 0.05) impact on the dependent variable. The other control variables Population, Average salary, Region and Year do not have a significant impact at p < 0.05 on the dependent variable.
In order to test Hypothesis 1, Model 2 includes the independent variable APS-firms. The results show that this model is significant as a whole (LR test = 1519.807, Chi-square = 348.586, P < 0.000), but a worse fit than model 1. The control variables Cumulative patents and Technological hotspot have a significant impact on the dependent variable (P < 0.05). The other control variables Population, Average salary, Tier, Region and Year do not have a significant impact at P< 0.05 on the dependent variable. The relationship between the granted patents and the number of APS firms is statistically insignificant (P > 0.05) in this model. Therefore, we reject hypothesis 1.
With the purpose of testing Hypothesis 2, Model 3 includes the independent variable foreign-born ratio. The results show us that this model is significant (LR test = 1519.772, Chi-square = 348.656, P < 0.000), but a worse fit than model 1 and a
similar fit as model 2. The control variable Cumulative patents has a significant impact on the dependent variable (P < 0.05). The other control variables Population, Average salary, Tier, Technological hotspot, Region and Year do not have a
significant impact on the dependent variable. The Independent variable APS-firms does also not have a significant impact on the dependent variable at P < 0.05. The relationship between foreign-born ratio and granted patents is statistically
insignificant (P > 0.05) in this model. Therefore, we reject hypothesis 2.
To test hypothesis 3, Model 4 includes the independent variable Top 1000 Universities. The results show us that this model is significant (LR test = 1519.597, Chi-square = 349.004, P < 0.000), but a worse fit than model 1 and a similar fit as model 2 and 3. The control variable Cumulative patents has a significant impact on the dependent variable (P < 0.05). The other control variables Population, Average salary, Tier, Technological hotspot, Region and Year do not have a significant impact on the dependent variable. The independent variables APS-firms and Foreign born ratio do not have a significant (P < 0.05) impact on the dependent variable. The
Table 4. Regression results Number of APS firms, Percentage of foreign born population and top 1000 Universities
Dependent variable: granted patents Model 1 Model 2 Model 3 Model 4
Beta and significance B Sig. B Sig. B Sig. B Sig. Constant 5,712 0,0000 5,463 0,000
Independent variables
Number ofAPS Firm 0,005 0,141 0,005 0,147 0,004 0,270
Percentage foreign born population -0,134 0,791 -0,089 0,863
Top 1000 Universities 0,023 0,556
Control variables
Cumulative Patents in 1000s 0,163 0,0000 0,153 0,000 0,153 0,000 0,149 0,000
Population per 100000 0,001 0,8806 -0,010 0,356 -0,010 0,379 -0,011 0,344
Technological hotspot (Yes) 0,577 0,001 0,602 0,008 0,323 0,151 0,320 0,156
Average Salary in 1000 USD 0,029 0,216 0,094 0,508 0,085 0,557 0,091 0,531
Tier Included Included Included Included Included Included Included Included
Region Included Included Included Included Included Included Included Included
Year Included Included Included Included Included Included Included Included
Model fit
N 265 265 265 265
Log likelihood 1624,558 1519,807 1519,772 1519,597
Likelihood ratio Chi-sqaure 373,816 348,586 348,656 349,004
P-value 0,000 0,000 0,000 0,000
Note: Values for tier, region and year are in the appendix
relationship between the granted patents and top 1000 Universities is statistically insignificant (P>0.05) in this model. Therefore, we reject hypothesis 3.
6. Discussion and conclusion
In this paper, we have argued that the characteristics of a global city would enhance the innovative performance of its inventors. We started with the hypothesis that global city connectivity would increase the innovative output of its inventors. Namely, this connectivity would make it easier for local inventors to connect with distant
knowledge areas (Sassen, 2002; Glückler, 2007). Next to this, it was argued that these opportunities would attract MNE subsidiaries and R&D activities to the city. Local inventors would then be able to benefit from these R&D activities by exploiting knowledge spillovers (Belderbos, 2014; Boschma, 2005). In turn, this would provide inventors with a deeper and more diverse knowledge pool that would enhance innovations. Our results however, based on the all United States Patent and
Trademark Office (USPTO) patents that have been granted to inventors resident in European and North-American global cities during the period 2001-2005, do not show a significant relationship between global city connectivity and innovative output. More specifically, we do not find a significant relationship between our two proxy variables, granted patents and APS firms.
This unexpected result might have been caused by an inadequacy in our measurement of connectivity. In the literature review, we showed that many
researchers (e.g. Belderbos, 2017; Taylor, 2001) have made use of the number of APS firms as measurement for connectivity. Namely, it was argued that these firms are able to weave cities in to their global networks. Though, some researchers (e.g. Derudder & Witlox, 2008; Keeling, 1995) argue that international airline flows provide a better indicator of connectivity. Unlike APS firms, this measurement tool
takes account for face-to-face meetings, which are particularly valuable for
exchanging tacit knowledge (Castellani, 2013). In our view, a combination of both connectivity measurements would provide us with a more comprehensive variable. However, we were not able to obtain the required air traffic data for most cities. Furthermore, in our research sample, we only included European and North American global cities because we believed that, given the amount of time we had, it would not be realistic to obtain the required data from all global cities the world. This could have influenced our results, as different mechanisms can be active in other continents. For example, Perri and Scalera (2017) show that Chinese global cities exhibit a greater ability to serve as gateway then global cities in advanced economies. Moreover, it can be argued that the impact and need of accessing foreign knowledge is larger for inventors in emerging economies than for inventors in advanced
economies. Namely, given the relative technological backwardness of emerging countries, it can be assumed that the domestically accessible pool of knowledge is smaller in these countries than in countries with advanced economies. Thus, the impact that global city connectivity has on innovative output of inventors might be greater in global cities that are located in emerging countries.
An alternative explanation is that we have overstated the impact of global city connectivity. Iammarino and McCann (2015) argue that MNEs role in global
innovation networks is not as large or important as some researchers believe it to be. Goerzen et al. (2013) have shown that MNEs have a tendency to locate their
competence-exploiting activities into global cities and their competence-creating activities into less core location (Cantwell & Mudambi, 2005). In practice this means that corporate headquarters are largely located in global cities whereas research activities are located in lower tier locations. As a consequence, political and financial
power of MNEs is clustered in global cities and strategic functions such as the
generation of new innovations are located in lower tier areas (Iammarino & McCann, 2015).
This would have implications for our argument. First, it would mean that not as many R&D activities are attracted to the global city as we expect. This would in turn impact the amount of exploitable knowledge spillovers in the global city and thus lowers recombination and innovation opportunities. Second, in this situation it can be assumed that the MNE linkages with other locations are now more skewed towards other power functions and less skewed towards innovation functions than in our theoretical model. This would in turn lower the accessible and exploitable knowledge from distant knowledge areas, which will damage recombination and innovation opportunities.
We also hypothesized that the cosmopolitan character of global cities would increase the innovative output of its inventors. Namely, the cosmopolitan character of a global city exposes its residents to a large variety of inputs that stimulates creativity and provides them with recombination opportunities (Florida et al, 2017; Alesina & Ferrara, 2005). The results however do not show a significant relationship between global city cosmopolitanism and innovative output. In other words, there is no significant relationship between dependent variable granted patents and independent variable foreign-born ratio.
Even though economists often use foreign-born ratio as a measurement for cosmopolitanism, it might not completely capture cosmopolitanism. As outlined in the literature review, it also encompasses other more intangible values such as autonomy, freedom or mutual respect (Mosterin, 2005). For that matter, it is hard to capture in one variable. Scholars in other fields such as philosophy or sociology
attempt to do so with measure such as level of freedom and egalitarianism (Sevincer et al., 2017). Also, besides the earlier hypothesized positive effects of
cosmopolitanism on innovation, the negative effects might be underexposed. For example, Ottaviano and Peri (2006) emphasize the role of integration as a prerequisite for realising potential benefits from diversity.
In our final hypothesis, we expected a positive relationship between global city human capital and innovative output of global city inventors. We expected this relationship because talent and supportive institutions (e.g. universities) are highly concentrated in and around global cities. Our results however do not confirm this relationship. Specifically, we do not find a significant relationship between our dependent variable granted patents and out independent variable number of top 1000 Universities.
This outcome could have been influenced by our measurement method of patent output, in which we assign a patent to a city based on the residence location of the inventor. Though, this might be not completely accurate as a University that is located in the city might influence inventors outside of the city as well. For example, students or University staff (which can be inventors) might study or work at the University in the city, but are not an actual resident of the city. Thus, once these inventors file for a patent, the patent will not be assigned to the city, but rather to the inventors’ residence location outside of the city.
Even though our hypotheses have been rejected, this research still has important contributions for the scientific literature and policymakers. From a theoretical point of view, we show how global city characteristics can influence the innovative abilities of inventors. Considering that most of the global city literature focuses on the relationship between global city characteristics and the business
environment, future researchers on the innovative climate in global cities can consult our framework. Furthermore, despite the limitations of our research, our results do raise questions about alleged importance of global cities in the global knowledge production – are global cities indeed innovative machines many assume them to be? More quantitative research is required. Moreover, we provide promising research avenues that require more research. As hinted at before, we suspect that the impact of global city connectivity is particularly big for the innovative climate in global cities from emerging economies. Finally, policy makers in global cities can use this paper to gain insights on the interplay between global city characteristics and innovation.
7. Limitations and future research
Our research has several limitations that need to be addressed and could provide promising opportunities for future research. First, while our research sample focuses on all global cities in Europe and North America, it could be that different
mechanisms apply to global cities located in other continents, which could influence the results. For example, Perri and Scalera (2017) have already shown that global cities in China are better able to serve a gateway function than global cities in advanced economies. Moreover, considering the technological backwardness of emerging countries, linkages to foreign knowledge areas might be especially valuable in these cities. Thus, testing our theoretical framework with a specific focus on global cities located in China or other emerging countries could be an interesting avenue for future research. Second, the variable that we use to measure connectivity might not be the most adequate measurement, as it does not take account for face-to-face meetings that are particularly valuable for exchanging tacit knowledge (Castellani, 2013). For that matter, we suggest that future researchers create a variable that tracks both the
movement of people in-and out of the city (e.g. air traffic) as global networks (APS-firms). We did not take account for face-to-face meetings in our connectivity variable as were not able to obtain the required data. Third, we tried to capture
cosmopolitanism by measuring the percentage of foreign-born residents in the city, an approach that is often used by economist (Short, 2004). Though, this approach might not completely capture cosmopolitanism, as it does not take account for more abstract cosmopolitan values such as autonomy, freedom or mutual respect. Future research could include these values by incorporating various indexes in their variable such as the human freedom index (Heratige.org, 2018) or tolerance index (Das et al., 2008). Fourth, we assigned patents to a city by accounting for the residence location of the lead inventor on the patent. However, it may very well be that an inventor is
influenced by the characteristics of the city (e.g. he/she works in the city), but it not a resident in the city. Future researchers could take this into account by assigning a patent to the location of the organisation where the inventor is active.
8. References
Ahuja, G. (2000). Collaboration networks, structural holes, and innovation: A longitudinal study. Administrative science quarterly, 45(3), 425-455. Alesina, A., & Ferrara, E. L. (2005). Ethnic diversity and economic performance.
Journal of economic literature, 43(3), 762-800.
Amabile, T. M. (1988). A model of creativity and innovation in organizations. Research in organizational behavior, 10(1), 123-167.
Archibugi, D., & Coco, A. (2005). Measuring technological capabilities at the country level: A survey and a menu for choice. Research policy, 34(2), 175-194. Arora, A., & Gambardella, A. (1990). Complementarity and external linkages: the
strategies of the large firms in biotechnology. The journal of industrial economics, 361-379.
Arzaghi, M., & Henderson, J. V. (2008). Networking off madison avenue. The Review of Economic Studies, 75(4), 1011-1038.
Asheim, B. T., & Coenen, L. (2006). Contextualising regional innovation systems in a globalising learning economy: On knowledge bases and institutional
