THESIS
MSc International Economics and Business
The Effect of the Presence of Research Institutes on Regional Innovation
August 24, 2014
Student: Jonas Bulthuis (s2040425)(jonas.bulthuis@rug.nl) Supervisor: Prof. Dr. H. Van Ees
Co-assessor: Dr. M.J. Gerritse
ABSTRACT
This thesis explores how knowledge from research institutes contributes to regional innovation. A dataset was compiled covering characteristics of research institutes in the European Union. Elaborating on theories of innovation systems, this thesis argues that the quality of research conducted by research institutes matters for innovation on a regional level and theorizes on the different channels through which knowledge is diffused. The effect of the perceived quality of research on regional innovation is found to be stronger than the effect of the numbers of students and employees on regional innovation. In addition, spillovers to surrounding regions are observed.
TABLE OF CONTENTS
ABSTRACT ... 2
TABLE OF CONTENTS ... 3
INTRODUCTION ... 4
THEORY ... 5
DATA AND METHODS ... 15
RESULTS AND DISCUSSION ... 21
CONCLUSION ... 28
REFERENCES ... 31
APPENDIX A - DISTRIBUTIONS OF VARIABLES ... 35
APPENDIX B - SCATTER DIAGRAMS ... 43
INTRODUCTION
This thesis focuses on the relationship between the geographical presence of research institutes and innovation in the surrounding area. For the purpose of this paper, the term “research institute” is used when referring to an organization that performs research activities and potentially publishes in scientific journals. The term “university” is used for research institutes that aside from conducting research also provide education with the possibility to reward a doctorate (Ph.D.) degree to its graduates. In addition, “innovation” is defined as the results of the process of translating an idea or invention into a good or service that leads to growth of the national economy, increase in employment, and creation of profit (Popadiuka & Wei Choob, 2006; Urabe, 1988).
Research institutes are seen as important players (Charles, 2006; Flores et al., 2009; Fritsch & Schwirten, 1999; McCall, 2010) when it comes to the production of new ideas and diffusion of knowledge. Their presence is expected to affect the levels of innovation and therefore economic growth in the surrounding areas. Traditionally dedicated to conserving, producing, and transmitting knowledge, universities are increasingly seen as drivers of economic and social development (Etzkowitz & Ranga, 2013).
Building on existing theories of the Triple Helix system, National Innovation Systems, and Regional Innovation systems, this thesis contributes new knowledge to the current incomplete understanding about the factors that determine the impact of research institutes on a regional level of innovation (Fritsch & Schwirten, 1999). As Jaffe (1989) pointed out, the mechanisms of transport for any of the observed spillovers from research institutes to the region are not fully understood. Existing literature does not seem to be clear about the characteristics of research institutes that matter for innovation on a regional level. Various papers have explored different channels through which knowledge can flow to and from research institutes. However, it is not clear how strong the impact of these channels is on regional innovation. Therefore, information that is important to regional innovation policymaking can be considered incomplete.
explores the connections between research institutes and regional actors outside the research institutes that are involved with innovation and argues that quality of research is among the most important factors that affect regional innovation.
For the aim of this thesis, the following research question was formulated: How is the
presence of research institutes affecting regional innovation? By examining the
effects of the geographical presence of research institutes, some additional light is shed on the mechanisms that stimulate regional development. Data about filed patents at the U.S. Patent and Trademark Office, regional R&D expenditures in the private sector, and citations of papers originating from regional research institutes allows those interested to draw various conclusions. The effect of the perceived quality of research on regional innovation is stronger than the effect of the number of students and employees on regional innovation. The perceived quality of research originating from research institutes in neighboring regions is also found to have a strong effect on regional innovation. A comparison of regions could expose whether critical success factors are in place to make research institutes effectively contribute to regional innovation. Results from this research can be used to compare characteristics of regions or research institutes. Comparing characteristics of regions and research institutes may reveal the need for new policies that stimulate regional innovation. The rest of this paper is structured as follows. Section 2 elaborates on theories that can be used for understanding regional innovation. Section 3 gives an overview of the data and methods that are deployed for the statistical analyses supporting the arguments in this thesis. Section 4 summarizes the results of the statistical analyses and discusses these. Finally, section 5 concludes the paper with an overview of the limitations of this research, as well as providing suggestions for future research.
THEORY
related to a number of actors. In the model, the three groups of actors increasingly take the role of one another and are thereby stimulating innovation (Etzkowitz & Ranga, 2013). Similar to other theories of innovation systems, the Triple Helix System is conceptualized in terms of components, relationships, and functions (Etzkowitz & Ranga, 2013). At the current state, however, other theories seem to be more specific when it comes to describing these things and therefore are more comprehensive. In addition, a drawback of the Triple Helix system is that, when examining regional innovation, this model might overlook specific features relevant for innovation on the regional level.
Another perspective for looking at factors that drive competitiveness and innovation is Porter's diamond. Porter and Stern (2001) highlight the well-known four attributes of Porter’s diamond as important factors affecting innovation. These four attributes are the presence of high-quality and specialized inputs, a context that encourages investment together with intense local rivalry, pressure and insight gleaned from sophisticated local demand, and the presence of related and supported industries. Porter's diamond focuses on the level of nations while within a nation there may still be substantial differences between the prevalence of and accessibility to its attributes. Because of this, many of the resources that form the basis for innovation are localized (McCall, 2010). Some of the aspects currently regarded as important to innovation are completely lacking in this system. For example, one can consider the role of governance and how it affects the system. Furthermore, Porter's diamond concerns a static system.
There exist several theories of innovation systems that address the above-mentioned lacking features of Porter's diamond in order to provide a more comprehensive perspective for viewing innovation. The literature on innovation systems conceptualizes innovation as an evolutionary and social process similar to evolutionary theories of economic and technological change (Edquist, 2004; Kastelle et al., 2009). Theories of innovation systems include National Innovation Systems, Regional Innovation Systems, Sectorial Innovation Systems, local clusters, and Technological Innovation Systems (Kastelle et al., 2009; Perez Vico, 2013). The innovating agents that play a role in these systems tend to be similar, although the different system typologies tend to co-ordinate the activities of the agents in different ways (Kastelle et al., 2009).
An often used approach for viewing innovation systems at the national level is the theory of National Innovation Systems (NIS). An NIS consists of elements, such as firms and knowledge institutions, that interact. The innovation performance relies on the quantity as well as the quality of these interactions, which are affected by market forces, networking and human resource mobility (Organisation for Economic Cooperation and Development [OECD], 2002). Growth in interactions may lead to growth in different dimensions, such as manpower, population, operational reserves, autonomy, transformation, and goal-changing abilities (OECD, 2002). These dimensions may in turn affect the system as a whole causing the system to evolve into an increasingly innovative environment.
Unlike the theory on NISs, this thesis focuses on innovation on a regional level. As Porter and Stern (2001) put it: “location exerts a strong influence.” When firms are located near to one another, their costs of production may drop due to the scale advantages of competing suppliers. When these supply to multiple firms, it lowers transportation costs. Tacit knowledge may is diffused easier when relevant actors are located near to one another, and greater specialization may be allowed (Porter & Stern, 2001). Regional agglomeration is found to be one of the most important factors that affect innovative output (Porter & Stern, 2001). In an agglomeration, different agents pursue related activities.
particular fields that compete but also cooperate (Cooke, 2001; Porter, 2001). This tends to create dynamism, flexibility, learning, and innovation (McCall, 2010). The fact that innovation takes place disproportionately in agglomerations has been validated empirically (Porter & Stern, 2001).
As a response to the notion that global competitiveness is not necessarily driven by factors attributed to national systems that drive innovation, the theorizing on the Regional Innovation System (RIS) emerged (Autio, 1998; Braczyk et al., 1998; Cooke, 2001; McCall, 2010). There is no commonly accepted definition for RISs, but usually, it is similarly to the NIS, "...a set of interacting private and public interests, formal institutions, and other organizations that function according to organizational and institutional arrangements and relationships conducive to the generation, use, and dissemination of knowledge" (Doloreux, 2003). A region can be defined as a geographic area of any size within a nation that is larger than a city (Asheim & Isaksen, 2002). It often concerns an area that can be distinguished from surrounding areas by borders or by specific traits, such as language and culture (McCall, 2010). Using the RIS approach, one can distinguish two subsystems. The first is the knowledge application and exploitation subsystem, which comprises companies, their clients, suppliers, and competitors, as well as their industrial cooperation partners. The second is the knowledge generation and diffusion subsystem, which comprises public research institutions, technology mediating organizations, and educational institutions (Tödtling & Trippl, 2005). Between the two subsystems, exchanges of knowledge, resources and human capital flow and interactions take place. The way in which this happens depends on a regional socioeconomic and cultural setting. In order to provide a clear overview of the elements and how they relate to one another, figure 1 gives a schematic overview of the RIS.
FIGURE 1.
Schematic overview of a regional innovation system (Tödtling & Trippl, 2005).
Given that there are many different actors in the system, which are affected by the earlier discussed governance framework, culture and cooperation (Fritsch & Schwirten, 1999), it is not hard to imagine that a lot of different configurations are possible for the RIS. To illustrate, the organizations within the RIS may have large numbers of employees who are connected in different kinds of relationships. Unfortunately, it remains unclear what the exact nature of the interactions between the actors is. The interactions need further exploration (Doloreux & Parto, 2004) in order to understand the dynamics of the RIS (Tunzelman, 2005; Kastelle et al., 2009). Research and educational organizations play a prominent role in the knowledge generation and diffusion subsystem of the RIS (Tödtling & Trippl, 2005). In order to increase the understanding of the nature of interactions in the RIS and how research institutes contribute to innovation on a regional level, this thesis focuses on the role of research institutes within the RIS. It is widely acknowledged that research institutes are important facilitators of innovation. Research institutes absorb, accumulate and generate knowledge, which they diffuse into the economy (Fritsch & Schwirten, 1999). A nation’s university system provides a bridge between technology and companies (Porter & Stern, 2001). Therefore, its presence may spur innovation, not only globally, but also on a regional level.
There are different channels through which research institutes exchange knowledge with regional actors. Known channels of knowledge transfer are publications, patents,1 informal meetings, consulting, joint ventures, research contracts and personal exchange (Flores, 2009). According to Fritsch and Schwirten (1999), the main channels are the following: education of students; carrying out contract research or innovation-related services such as testing, consulting, and training personnel; joint R&D projects of research institutions and private firms; and informal exchanges of know-how. Most of these channels could be utilized for diffusion as well as accumulation of knowledge. For the sake of simplicity, the channels of knowledge transfer are classified into four types: publications, education, collaboration, and personal exchange.
Publications
Research institutes diffuse codified knowledge through different forms of publications. Perhaps the most known form of diffusion of knowledge from research institutes is publications in scientific journals. Scientific journals allow researchers to share their research results with their peers. As it concerns knowledge in codified form, it can be transferred easily, not only regionally but also globally.
This is also the case for press releases. Press releases could be used as a means to spread information about research findings as well. Whether spread through scientific journals or through press releases, both channels information could target a local audience while at other times it may involve a global audience.
The same is true for distribution of codified knowledge through the filing of patents. Results from the research conducted at research institutes may contribute to the invention of new technology for which patents are granted. Often, the research institutes themselves file patents since the protection of intellectual property rights is considered increasingly important to them (Autio, 1998). Information about patents is publicly available and contains knowledge that is derived from research results in codified form. This could be picked up by others, regionally as well as globally. When research is produced in high quantities, it might still lack the quality that is needed for innovation.
H1: The number of times that papers published by a research institute are cited has a positive effect on a regional level of innovation.
Education
Some of the human capital in the RIS may flow to and from local research institutes. For example, research institutes offer personnel training (Fritsch & Schwirten, 1999). Through courses created for companies, knowledge that is useful for firms in the region is spread. Much of the knowledge of research institutes is diffused through the education of students. Graduated students are expected to make a significant
contribution to innovation. However, as students do not always reside in the region
after graduation it is questionable how much they contribute to regional innovation. Therefore, differences in the number of students that remain in the region after graduation and those that migrate to another region might play a role. Nevertheless, even when students remain in the region only during the period of studies, knowledge may be diffused by students through informal and formal meetings. Therefore, the second hypothesis of this thesis is the following.
H2: The number of university students has a positive effect on a regional level of innovation.
Collaboration
consulting, research contracts, and testing of new technology, such as medical equipment and computer systems. All these different settings have the potential to exchange of information that could be used for innovation.
Research institutes often are actively involved in knowledge application and exploitation activities (Autio, 1998). Many universities are equipped with a liaison office with experienced exploitation and commercialization personnel who contribute to innovation (Cooke, 2001). Liaison offices assist with applying for patents and finding partners for collaboration. Having a liaison office that investigates opportunities for commercialization helps universities to become more responsive to the market. The responsiveness of the university system to industrial innovation opportunities is regarded as an important ingredient for innovation (Porter & Stern, 2001).
level. The survey of German research institutes by Fritsch and Schwirten (1999) points out that university’s proximity positively influences its chance of being chosen as a collaborator, and therefore supports the notion that universities are an important player in RISs. It is to be expected that a mix of public research that is considered useful and private R&D spurs regional innovation.
Personal exchange
Crucial parts of the RIS are knowledge, resource, and human capital flows and interactions (Tödtling & Trippl, 2005; Autio, 1998). Research institutes can be regarded as creators of commoditized knowledge, human capital and social capital (Charles, 2006). As discussed earlier, part of the knowledge is transmitted through codified information, such as scientific reports and publications. Other kinds of knowledge need personal interaction in order to be transferred (Fritsch & Schwirten, 1999). An important part of the knowledge that contributes to innovation is tacit (Flores, 2009). This kind of knowledge is spatially bounded (Tödtling & Trippl, 2005; Anselin, 1997) and its exchange requires intensive personal contacts of trust-based character, which are facilitated by geographic proximity (Tödtling & Trippl, 2005). Much of the tacit knowledge is transferred by means of informal contacts (Fritsch & Schwirten, 1999). Therefore, when examining links that affect innovation, the informal contacts should also be taken into account. Faulkner and Senker (1994) conclude from a rather small sample of 60 R&D personnel in British and American enterprises that the cooperation with universities in many cases was based on personal contacts. Researchers, either in formal or informal settings, may act as regional animators (Fritsch & Schwirten, 1999). Any meeting that brings the actors of an RIS in touch with one another could potentially facilitate the exchange of tacit as well as codified knowledge. The number of university employees in a region may affect the innovation performance of the region. The more university employees are part of an RIS, the more potential there is for the exchange of both codified and tacit knowledge to and from research institutes. Therefore, the third hypothesis of this thesis is the following.
H3: The staff size of the regional university system has a positive effect on a regional level of innovation.
TABLE 1.
Overview of hypotheses.H1 The number of times that papers published by a research institute is cited has a positive effect on a regional level of innovation.
H2 The number of university students has a positive effect on regional level of innovation.
H3 The staff size of the regional university system has a positive effect on the a regional level of innovation.
DATA AND METHODS
to society. This makes U.S. patent counts a strong indicator of innovation for the following four reasons:
1. "When a foreign inventor files a U.S. patent, it is a sign of the innovation's potential economic value because of the costs involved" (Porter & Stern, 2001). 2. When a patent is filed in the U.S., it indicates that the invention concerned may
have already been successful domestically, and an expansion to the large U.S. market is taking place.
3. The U.S. is at the technological frontier, so investing in a U.S. patent is more likely to occur for inventions that are truly novel.
4. Patent data on the U.S. is available.
U.S. patents granted to actors in an RIS can be seen as results from the processes of innovation. Knowledge that is diffused by research institutes may be incorporated into inventions that are commercialized. The configuration of the RIS may lead to U.S. patents, which in turn may be capitalized by numerous actors in the region in the form of growth of the regional economy, increase in employment, and profit. Part of the patents is granted to the universities and, therefore, U.S. patents as the dependent variable may seem tautological because most of the independent variables relate to the presence of universities. Although a small bias may be present because universities also file patents, the share of university patents tends to be very small in the European Union (Guena & Nesta, 2006). For the far majority of the patents, actors other than those related to the universities seem to benefit from it. Moreover, this is also reflected in the fact that, from the compiled dataset for this thesis, significant spillovers from research institutes of neighboring regions were observed, i.e. actors outside the involved research institutes benefited from it.
The following independent variables were used to test the hypotheses.
• Citations captured the number of times publications from research institutes in a given region had been cited.
• UniStaff captured the combined number of employees for all the universities in a given region.
Apart from the independent variables mentioned above, an additional variable, Grav was used for capturing spillover effects from other regions. Each variable was indexed per region and per year as indicated below here with the subscripts region and year respectively. After an invention has been conceived, filing a patent for it could take several years. Therefore, it was not expected that knowledge diffused to and from research institutes would lead to granted patents within the same year, and lags of several years are incorporated into all the models. There is a small variation in the lag sizes between the different models due to limitations in available data. For the scope of this thesis and due to the limited timeframe, data for the dependent variable,
Patents, was selected of the most recent year of data available (2013) for all models.
knowledge produced at research institutes leads to patents. Therefore, citation scores were collected from an earlier period than the period of which patents counts were collected.
The locations of the research institutes were not readily available. After identifying the majority of research institutes in the sample, the data was processed with geocoding software to find the region where the research institute was located. Then, the number of times papers were cited between 2005 and 2009 in the NUTS 2 region were the papers originated was counted. The geocoding software did not associate all citations with NUTS 2 regions. The large majority of citation data, however, has been incorporated for each of the regions. Data for Internet use, as well as population data for calculating per capita values, was collected from Eurostat. The dataset contained the size of the population in different years per NUTS 2 region and was based upon January 1st of the related year. For the variables UniStaff and UniStudent a dataset compiled in 2010 by the EUMIDA consortium was used. The dataset contained a list of research institutes for which the staff size, the number of students and NUTS 2 region were provided. In order to illustrate the differences in variable dependent effects, this thesis deployed the econometric models listed in table 2.
TABLE 2.
Econometric models. Model specification(1) PatentsRegion2013 /CapitaRegion2009 =
CitationsRegion2005-2009 /CapitaRegion2009 + θControls + ε
(2) PatentsRegion2013 / CapitaRegion2009 =
CitationsRegion2005-2009 /CapitaRegion2009 + GravRegion2013/CapitaRegion2009 + θControls + ε
(3) PatentsRegion2013/CapitaRegion2010 =
UniStudentsRegion2010 /CapitaRegion2010 + θControls + ε
(4) PatentsRegion2013/CapitaRegion2010 =
The innovative output of a region also depended on the size of the population in the region. More precisely, as seen from the perspective of the RIS, it depended on the available stock of human capital (Fritsch & Schwirten, 1999; Autio, 1998; Tödtling & Trippl, 2005). In order to correct for the fact that not all regions were sized equally, variables that concerned absolute numbers were divided by the size of the total population and were expressed in per capita values.
Unlike data on human capital per region, data on population size was more readily available. The population size used to calculate a value for the dependent variable was the population size at the time when the independent variable was indexed because this corrected for possible fluctuations in population size and maintained the effect on innovative output per capita. The variable Capita captured the population size of a given region during the given year.
Besides the size of the regional population, other factors explaining innovative output based on different models in the literature were related to R&D expenditure. The models used for this thesis were derived from a Cobb-Douglas model that was modified by Jaffe (1989) in order to explain the number of patents granted to actors in a given state in the U.S. As independent variables Jaffe (1989) included R&D performed by universities, R&D performed by private companies, and the measure of geographic coincidence of university and industrial research. R&D expenditure in the private sector is used as a control variable when testing each of the hypotheses for this thesis. For this thesis, the R&D performed by universities was replaced by different characteristics of research institutes: citation scores, staff size, and the number of students. A measure for geographic coincidence with research institutes outside a given region is included in model 2 in order to capture spillover effects of neighboring regions. The variables incorporated into the models used for this thesis are similar to the variables in various models for explaining regional innovation (Anselin, 1997; Buesa et al., 2010; Rodríguez-Pose et al., 2006; Regional Innovation Scoreboard, 2014). The variables of the final sample can be defined as follows.
PatentsRegion2013 / CapitaRegion2009
PatentsRegion2013 / CapitaRegion2010
This is the number of U.S. patents granted for a given NUTS 2 region in 2013 divided by the region's population on January 1, 2010.
CitationsRegion2009 / CapitaRegion2009
This is the number of citations from research institute publications during the period 2005-2009 in the given NUTS 2 region per capita during 2009 divided by the region's population size on January 1, 2009.
GravRegion2013 / CapitaRegion2009
This is the geographic coincidence index, which can also be referred to as the gravity effect of surrounding regions. This gravity variable is calculated by summing the citation scores for surrounding regions, corrected for the distance to each surrounding region. The correction was performed by dividing the citation score for each surrounding region by the square of the distance in kilometers to it. Mathematically this can be expressed as
GravRegion1 = ΣregionX Citations RegionX / (dRegion1RegionX)² ,
with dRegion1RegionX as the distance between region Region1 and region RegionX.
Similar to other variables, the gravity variable is divided by the region's population size on January 1, 2009.
UniStaffRegion2010 / CapitaRegion2010
This is the number of university employees in a given NUTS 2 region during 2010 divided by the region's population size on January 1, 2010.
UniStudentsRegion2010 / CapitaRegion2010
This is the number of university students in the given NUTS 2 region during 2010 divided by the region's population size on January 1, 2010.
PrivateR&DRegion2010 / CapitaRegion2010
PrivateR&DRegion2009 / CapitaRegion2009
This is the number of full-time equivalents of R&D personnel and researchers in the business enterprise sector for the given NUTS 2 region during 2009 divided by the region's population size on January 1, 2009.
The hypotheses for this thesis were tested by means of an ordinary least squares (OLS) regression analyses. There are several reasons why the chosen variables meet the criteria for using OLS. Any correlation found between independent variables that were jointly used in the models was well below the level that makes these variables suspect to multicollinearity. Moreover, all regressions passed the t-tests for multicollinearity with low VIF-values. An overview of correlation matrices can be found in the "Results and Discussion" section. In order to deal with the skewness of the variables, the logarithm function was applied to them, similar to the work of Anselin (1997). The dependent variables passed the Jarque Bera test for normality. The differences in distributions between the variables with and without the application of the logarithm function are depicted in appendix A. Three different versions of the Breusch-Pagan and Cook-Weisberg tests for heteroskedasticity showed that none of the models suffered from any unacceptable amount of heteroskedasticity. Appendix B allows for visual inspection through scatterplots of the dependent variables w.r.t. the independent variables. The geographic distribution of the variables are depicted in Appendix C.
RESULTS AND DISCUSSION
Table 3 shows the descriptive statistics for the 272 observed NUTS 2 regions of which data could be compiled for this thesis. The dataset contains values per NUTS 2 region for each variable. The table shows that the number of observations differs among the variables because some of the NUTS 2 regions data was lacking. This resulted in different sample sizes for the regression analyses depending on the variables included.
TABLE 3.
Descriptive statistics.Tables 4 and 5 show the correlations between the independent variables that were indexed during the year 2009 (models 1 and 2) and the independent variables that were indexed during the year 2010 (models 3 and 4), respectively.
TABLE 4.
Correlation matrix of models with dependent variables indexed at 2009.
TABLE 5.
Correlation matrix of models with dependent variables indexed at 2009.
Variable LOG( PatentsRegion2013/C apitaRegion2010) LOG( PrivateR&DRegion2 010/CapitaRegion2010) LOG( UniStaffRegion2010/ CapitaRegion2010) LOG( UniStudentsRegion2 010/CapitaRegion2010) LOG( PatentsRegion2013/ CapitaRegion2010) 1.0000 LOG( PrivateR&DRegion2 010/CapitaRegion2010) 0.7284*** 1.0000 LOG( UniStaffRegion2010/ CapitaRegion2010) 0.1670** 0.2459*** 1.0000 LOG( UniStudentsRegion2 010 CapitaRegion2010) -0.1139 0.1214 0.7017*** 1.0000 *** p<0.01, ** p<0.05, * p<0.1
To test the hypotheses for this thesis, several regressions were performed. Tables 6 and 7 give the results of the regressions with the independent variables that were indexed during the year 2009 (models 1 and 2) and the results of the regressions with the independent variables that were indexed during the year 2010 (models 3 and 4), respectively. Not surprisingly, a significant positive correlation was found between the number of university students and the number of university employees. In order to avoid issues related to multicollinearity, the hypotheses 2 and 3 were tested with separate models (models 3 and 4). The regressions are numbered from 1 to 6. Regressions 1 and 4 were performed to test the effect of the control variable
TABLE 6.
Independent Variables Indexed at 2009.
(1) (2) (3)
VARIABLES Controls Model 1 Model 2
LOG(CitationsRegion2009 / CapitaRegion2009) 0.328*** 0.327*** (0.0743) (0.0735) LOG(GravRegion2013 / CapitaRegion2009) 0.308*** (0.0568) LOG(PrivateR&DRegion2009 / CapitaRegion2009) 1.312*** 1.044*** 0.950*** (0.0802) (0.0930) (0.0911) Constant -2.687*** -3.131*** -0.592 (0.504) (0.532) (0.774) Observations 205 178 178 R-squared 0.543 0.560 0.615
TABLE 7.
Independent Variables Indexed at 2010.
(4) (5) (6)
VARIABLES Controls Model 3 Model 4
LOG(UniStudentsRegion2010 / CapitaRegion2010) 0.200 (0.285) LOG(UniStaffRegion2010 / CapitaRegion2010) 0.305 (0.211) LOG(PrivateR&DRegion2010 / CapitaRegion2010) 1.235*** 1.211*** 1.179*** (0.101) (0.130) (0.128) Constant -3.445*** -2.937** -2.138 (0.648) (1.341) (1.473) Observations 139 88 87 R-squared 0.531 0.500 0.511
One of the most outstanding results from the regression analyses was that citation scores of the research institutes corresponded with the regional level of innovation. Model 1 (regression 2) and model 2 (regression 3) tested the effect of the number of citations. The results showed that the number of times papers published by regional research institutes were cited had a positive effect on a regional level of innovation. This is a clear indicator that knowledge spillovers from regional universities contributed to regional innovation. The regional level of innovation was also positively affected by the number of times papers published by research institutes in surrounding regions were cited. This is a clear indicator that spillover effects reach beyond the borders of NUTS 2 regions and that neighboring regions benefit from the presence research institutes.
The correlation coefficients for Citations and Grav with Patents were highly significant and showed almost equal explanatory power, as can be witnessed from the correlation matrix in table 4. Results from the regression analyses in table 6 show that a 1% increase in citation scores on average leads to a 0.33% increase in regionally granted U.S. patents (models 1 and 2 / regressions 2 and 3). Similarly, a 1% increase in citation scores in surrounding regions on average leads to a 0.3% increase in regionally granted U.S. patents (model 2 / regression 3). By definition of Grav, the spillover effect is inversely proportional to the distance. With a p-value of less than 0.01, the found coefficients for both of these models were highly significant. Based on the results of the analyses, hypothesis 1 of this thesis is supported by both models 2 and 3.
was found. Table 5 shows a low, but positive and significant correlation coefficient of
UniStaff w.r.t. Patents.
The result from regression 6 also shows a positive relationship that indicates that a 1% increase in the number of university employees leads to a 0.3% increase in the regional level of innovation. However, the result from this regression was not significant. Therefore, hypotheses 3 of this thesis is rejected. For any possible effect of university students on regional innovation, even less evidence was found, therewith also rejecting hypothesis 2. The effect of university students may be small due to student migration to other regions after graduating. In this case, knowledge obtained from the university would be utilized elsewhere and not increase innovation in the university region. The currently used dataset does not reveal sufficient information about the effects of university students as well as staff.
CONCLUSION
After an extensive process of collecting and analyzing data, the compiled dataset uncovered interesting findings regarding regional innovation systems and the role of research institutes in the European Union. The collected data revealed regional knowledge spillovers from research institutes. These knowledge spillovers can occur through publications, education, collaboration, and personal exchanges as supported by the results of the statistical analyses. The different models provided insight into relationships between diverse factors affecting regional innovation systems. One of the main findings from this research is that the quality of research that is conducted in a region has a strong positive effect on innovative performance at the regional level. This supports the notion that location matters in relation to research institutes from which knowledge can be transferred in codified form over greater distances. Results from this research indicate that knowledge from research institutes is utilized in the surrounding area.
picture became more complete when considering research institutes and their characteristics.
The ability to compile the data limited the research. While collecting data for NUTS 2 regions of the European Union, it was difficult to find complete data sets for all variables covering all regions. When connecting the different variables into one dataset, the number of observations for which all values were known became smaller. The differences in the number of observations between models reflected this.
In addition, data on patents and citations had to be connected to the NUTS 2 regions of the European Union. This involved a geocoding process that was not completely accurate. As a consequence, there were gaps in the variables accounting for the number of patents per region and the number of citations per region. If these gaps in the data were closed, the resulting improved version of the dataset would yield more precise results.
Another limitation was the variability of the number of patents filed among the regions. For many regions, the number of patents is small on the order of magnitude of 1 because many regions have one or zero patents. The regression excluded regions with zero patents due to missing data as a result of the LOG function. Examining the effects on a more aggregated level, such as the level of NUTS 1 regions, would mitigate this limitation. Moreover, larger regions had more data available from the statistical agencies, and as such the results would have fewer gaps in data. This would provide the additional advantage of further analyzing spillovers from research institutes that cover a larger area than the NUTS 2 regions. The explanatory power of the variables UniStudents and UniStaff does not appear to be strong enough for the current sample sizes (88 observations when accounting for students and 87 observations when accounting for staff). Larger sample sizes could have revealed more evidence in support or opposition of the hypotheses.
Innovative output was measured by the number of granted patents and may have biased results. With patents, the grantee and the inventor do not necessarily reside within the same region. Therefore, it is possible the data inaccurately reflects innovation conceived in a different region.
these differences by examining the university type. It may also be worthwhile to differentiate between different types of university staff (e.g. research vs. support staff) in order to find the effect of different staff groups on regional innovation.
Finally, the scope of this thesis is limited to certain years and regions within the European Union. Broadening it to include more regions and time periods would provide a clearer picture not only of innovation stimulus but also of how it has changed over the years.
REFERENCES
Acs, Z. J., Anselin, L., & Varga, A. 2002. Patents and innovation counts as measures of regional production of new knowledge. Research Policy, 31: 1069-1085.
Anselin, L. 1997. Local geographic spillovers between university research and high technology innovations. Journal of Urban Economics, 42: 422-448.
Asheim, B. T., & Isaksen, A. 2002. Regional innovation systems: the integration of local ‘sticky’ and global ‘ubiquitous’ knowledge. Journal of Technology Transer, 27: 77-86.
Autio, E. 1998. Evaluation of RTD in regional systems of innovation. European Planning Studies, 6(2): 131-140.
Bergek, A., Jaconsson, S., Carlsson, B., Lindmark, S., & Rickne, A. 2008. Analyzing the functional dynamics of technological innovation systems: a scheme of analysis. Research Policy, 37(3): 407-429.
Braczyk, H. J., Cooke, P., & Heidenreich, M. 1998. Regional innovation systems: The role of governances in a globalized world: 2-25. London: UCL Press.
Buesa, B., Heijs, J. & Bauert, T. 2010. The determinants of regional innovation in Europe: A combined factorial and regression knowledge production function approach. Research Policy, 39: 722–735.
Charles, D. 2006. Universities as key knowledge infrastructures in RISs. Innovation, 19(1): 117-130.
Cooke, P. 2001. RISs, clusters, and the knowledge economy. Industrial & Corporate Change, 10(4): 945-974.
Cowana, R. & Zinovyevac, N. 2013.University effects on regional innovation. Research Policy, 42: 788-800.
Doloreux, D. & Parto, S. 2005. RISs: Current discourse and unresolved issues. Technology in Society, 27: 133–153.
Etzkowitz, H., & Ranga, M. 2013. A triple helix system for knowledge-based regional development: from "spheres" to "spaces". Industry and Higher Education, 27(4): 237-262.
EUMIDA European University Data Collection 2010. EUMIDA European
University Data Collection, http://datahub.io/nl/dataset/eumida, viewed on 6 July 2014.
European Commission. 2014. Regional innovation score board 2014.
http://ec.europa.eu/enterprise/policies/innovation/files/ius/ius-2014_en.pdf, viewed on 6 July.
Eurostat. 2014. Population on 1 January by age and sex - NUTS 2 regions,
http://epp.eurostat.ec.europa.eu/portal/page/portal/statistics/themes, viewed on 6 July 2014.
Eurostat. 2014. Total R&D personnel and researchers by sectors of performance, sex and NUTS 2 regions,
http://epp.eurostat.ec.europa.eu/portal/page/portal/statistics/themes, viewed on 6 July 2014.
Faulkner, W. & Senker, J. 1994. Making sense of diversity: public-private sector research linkage in three technologies. Research Policy, 23: 673-695.
Flores, M., Boe, C., Huber, C., Plu, A., Schoc, R. & Poulyd, M. 2009.Universities as key enablers to develop new collaborative environments for innovation: successful experiences from Switzerland and India. International Journal of Production Research, 47(17): 4935–4953.
Fritsch, M. & Schwirten, C. 1999. Enterprise-university co-operation and the role of public research institutions in RISs. Industry and Innovation, 6(1): 69-83.
Geuna, A., & Nesta, L. J. J. 2006. University patenting and its effects on academic resarch: the emerging European evidens. Research Policy, 35: 790-807.
Gunasekara, C. 2006. Reframing the role of universities in the development of RISs. Journal of Technology Transfer, 31: 101–113.
Jaffe, A. B. 1989. Real effects of academic research. The American Economic Review, 79(5): 957-970.
Kastelle, T., Potts, J., & Dodgson, M. 2009. The evolution of Innovation Systems. Paper presented at DRUID Summer Conference 2009, Copenhagen.
Lee, T. L., & von Tunzelmann, N. 2005. A dynamic analytic approach to national innovation systems: the IC industry in Taiwan. Research Policy, 34: 425-440.
McCall, T. 2010. Regional innovation systems. Working paper, University of Tasmania.
Jaffe, A. B., Trajtenberg, M. & Henderson, R. 1993. Geographic localization of knowledge spillovers as evidenced by patent citations. Quarterly Journal of Economics, 108: 577-598.
Jalles, J. T. 2010. How to measure innovation? New evidence of the technology– growth linkage. Research in Economics, 64: 81–96.
Janeiro, P., Proença, I., & Da Conceição Gonçalves, V. 2013. Open innovation: Factors explaining universities as service firm innovation sources. Journal of Business Research, 66: 2017-2023.
Organisation for Economic Cooperation and Development. 2002. Dynamising National Innovation Systems. Organisation for Economic Cooperation and Development, http://www.oecd-ilibrary.org, viewed on 15 July 2014.
Perez Vico, E. 2013. The impact of academia on the dynamics of innovation systems - capturing and explaining utilities from academic R&D. Thesis for the degree of doctor of philosofy, Chalmers University of Technology, Gothenburg.
Porter, M. E. 2000. Location, competition, and economic development: Local clusters in a global economy. Economic Development Quarterly, 14: 15.
Porter, M. E. & Stern, S. 2001. Innovation: Location matters. MIT Sloan Management Review, 42(4): 28.
Rao, B. C. 2010.How to measure innovation. Challenge, 53(1): 109-125.
Van Dierdonck, R., Debackere, K. & Engelen, B. 1990. University-industry relationships: how does the Belgian academic community feel about it? Research Policy, 19: 551-566
Tödtling, F. & Trippl, M. 2005. One size fits all? Towards a differentiated regional innovation policy approach. Research Policy, 34: 1203-1219.
United Nations Economic Commission for Europe. 2014. The United Nations
economic commission for Europe, http://www.immi.gov.au, viewed on 6 July 2014.
Urabe, K. 1988. Innovation and the Japanese management system. In K. Urabe, J. Child, & T. Kagono. (Eds.), Innovation and management international comparisons. Berlin: Walter de Gruyter.
Web of Science 2014. Thomson Reuters, New York,
APPENDIX A -‐ DISTRIBUTIONS OF VARIABLES
PatentsRegion2013 / CapitaRegion2009 LOG(PatentsRegion2013 / CapitaRegion2009) 0 5000 1 .0 e +0 4 1 .5 e +0 4 D e n si ty 0 .0002 .0004 .0006 .0008 patents2013_per_cpta2009US patents in 2013 per capita in 2009
0 .1 .2 .3 D e n si ty -16 -14 -12 -10 -8 ln patents2013_per_cpta2009
PatentsRegion2013 / CapitaRegion2010 LOG(PatentsRegion2013 / CapitaRegion2010) 0 5000 1 .0 e +0 4 1 .5 e +0 4 2 .0 e +0 4 D e n si ty 0 .0002 .0004 .0006 .0008 patents2013_per_cpta2010
US patents in 2013 per capita in 2010
0 .1 .2 .3 D e n si ty -16 -14 -12 -10 -8 ln patents2013_per_cpta2010
PrivateR&DRegion2009 / CapitaRegion2009 LOG(PrivateR&DRegion2009 / CapitaRegion2009) 0 100 200 300 400 D e n si ty 0 .005 .01 .015 RD2009_per_cpta2009
FTE in private R&D in 2009 per capita in 2009
0 .1 .2 .3 .4 D e n si ty -10 -8 -6 -4 ln RD2009_per_cpta2009
PrivateR&DRegion2010 / CapitaRegion2010 LOG(PrivateR&DRegion2010 / CapitaRegion2010) 0 100 200 300 400 D e n si ty 0 .005 .01 .015 RD2010_per_cpta2010
FTE in private R&D in 2010 per capita in 2010
0 .1 .2 .3 .4 D e n si ty -10 -8 -6 -4 ln RD2010_per_cpta2010
CitationsRegion2009 / CapitaRegion2009 LOG(CitationsRegion2009 / CapitaRegion2009) 0 5 10 15 D e n si ty 0 .2 .4 .6 citations2009_per_cpta2009 0 .1 .2 .3 D e n si ty -8 -6 -4 -2 0 ln citations2009_per_cpta2009
GravRegion2013 / CapitaRegion2009 LOG(GravRegion2013 / CapitaRegion2009) 0 1000 2000 3000 4000 5000 D e n si ty 0 .001 .002 .003 grav_citations_per_capita2009 0 .1 .2 .3 D e n si ty -14 -12 -10 -8 -6 ln grav_citations_per_capita2009
UniStaffRegion2010 / CapitaRegion2010 LOG(UniStaffRegion2010 / CapitaRegion2010) 0 100 200 300 400 500 D e n si ty 0 .005 .01 .015 unistaff2010_per_cpta2010 0 .2 .4 .6 .8 D e n si ty -8 -7 -6 -5 -4 lunistaff2010_per_cpta2010
UniStudentsRegion2010 / CapitaRegion2010 LOG(UniStudentsRegion2010 / CapitaRegion2010)
0 10 20 30 40 50 D e n si ty 0 .05 .1 .15 unistudents2010_per_cpta2010 0 .2 .4 .6 .8 D e n si ty -8 -6 -4 -2 lunistudents2010_per_cpta2010
APPENDIX B -‐ SCATTER DIAGRAMS
LOG(PatentsRegion2013 / CapitaRegion2009) w.r.t. LOG(PrivateR&DRegion2009 / CapitaRegion2009 )
LOG(PatentsRegion2013 / CapitaRegion2009) w.r.t. LOG(CitationsRegion2009 / CapitaRegion2009)
LOG(PatentsRegion2013 / CapitaRegion2009) w.r.t. LOG(GravRegion2013 / CapitaRegion2009)
LOG(PatentsRegion2013 / CapitaRegion2010) w.r.t. LOG(UniStaffRegion2010 / CapitaRegion2010)
APPENDIX C -‐ MAP PLOTS
LOG(PatentsRegion2013 / CapitaRegion2009)
LOG(CitationsRegion2009 / CapitaRegion2009)
LOG(PrivateR&DRegion2010 / CapitaRegion2010)
