1. Paper to be presented at the Summer Conference 2009
on
CBS - Copenhagen Business School
Solbjerg Plads 3
DK2000 Frederiksberg
DENMARK, June 17 - 19, 2009
UNIVERSITY-INDUSTRY R&D COLLABORATIONS: A JOINT-PATENTS ANALYSIS
Antonio Messeni Petruzzelli
DIMeG - Politecnico di Bari
a.messeni-petruzzelli@poliba.it
Abstract:
Empirical studies on R&D collaborations between universities and firms have mainly centred their attention
on universities and firms characteristics and strategies that favour the establishment of collaborative agreements.
In this paper, we extend the current research framework investigating the role that specific relational attributes
may play on the relevance of such collaborations. Specifically, we focus on three relevant factors, namely
technological relatedness, national culture similarity, and prior collaborations ties between universities and
firms. We develop testable hypotheses about their impact on the innovative performance of R&D university-
industry collaborations, and test them on a sample of 796 university-industry collaborations, established by 27
universities located in 12 different European countries.
Our results suggest that innovation value has an inverted U-shaped relation with partners technological
relatedness. In addition, universities and firms belonging to similar cultural contexts and having had previous
ties are more able to achieve better innovative outcomes.
JEL - codes: O32, -, -
2. University-Industry R&D Collaborations:
A Joint-Patents Analysis
ABSTRACT
Empirical studies on R&D collaborations between universities and firms have mainly centred their attention
on universities and firms’ characteristics and strategies that favour the establishment of collaborative
agreements. In this paper, we extend the current research framework investigating the role that specific
relational attributes may play on the relevance of such collaborations. Specifically, we focus on three
relevant factors, namely technological relatedness, national culture similarity, and prior collaborations ties
between universities and firms. We develop testable hypotheses about their impact on the innovative
performance of R&D university-industry collaborations, and test them on a sample of 796 university-
industry collaborations, established by 27 universities located in 12 different European countries.
Our results suggest that innovation value has an inverted U-shaped relation with partners’ technological
relatedness. In addition, universities and firms belonging to similar cultural contexts and having had
previous ties are more able to achieve better innovative outcomes.
Key words: university-industry collaborations; innovation value; relational attributes
1. INTRODUCTION
Nowadays, it is well understood that the creation and application of new knowledge are the primary factors
that drive the economic growth. Moreover, it is also commonly accepted that universities are important
sources of new knowledge, especially in the areas of science and technology (e.g. Rosenberg and Nelson,
1994; Nelson and Rosenberg, 1998; Etzkowitz and Leydesdorff, 2000). Thus, researchers have devoted a
great effort to investigate the nature and the importance of the relationships between university and industry,
tying to build a clear picture of which mechanisms may favour universities and firms interaction, thus
promoting knowledge transfer and acquisition. A better comprehension of university-industry links has
assumed a great importance also at policy level, as shown by the several initiative launched by the European
Commission to proactively enhance the transfer of technological knowledge from university to industry and
identify effective and efficient innovation policies.
The aim of this study is to contribute to the analysis of university-industry relationships, focusing on the role
that relational attributes may play on the relevance of such collaborations. In fact, studies on this topic have
mainly centred their attention on the universities and firms’ characteristics and strategies that favour the
establishment of collaborative agreements. In particular, universities’ entrepreneurial orientation, faculty
-1-
3. incentive mechanisms, national policies, government support, type of industry, and involvement in
complementary innovative activities have been described as the main important factors leading universities
and firms to fruitful collaborate (e.g. Debackere and Veugelers, 2005; Veugelers and Cassiman, 2005;
Rothaermel et al., 2007). Nevertheless, few attention has been devoted to understand how relational specific
attributes may affect the value of university-industry R&D collaborations. In an attempt to fill this gap, we
identify three relevant such attributes, namely technological relatedness between partners, national culture
similarity, and previous collaboration ties.
Collecting data from the European Patent Office (EPO), we study university-industry collaborations in terms
of joint patents, and present an econometric analysis examining the impact of the three relational variables
on the value of the collaboration innovative output. 796 collaborations are considered, developed by 27
universities located in 12 countries belonging to the European Union. Results show that the value associated
to university-industry joint innovations presents an inverted U-shaped relation with partners’ technological
relatedness, and it is favoured by national culture similarity and the existence of previous collaboration ties
between the organizations.
The paper is structured as follows. Section 2 reports the theoretical background, analysing the relevance of
knowledge complementarity and collaboration agreements to innovate, and the role of universities as
knowledge sources. Section 3 presents the hypotheses about the influence of technological relatedness,
national culture similarity, and prior collaboration ties on the innovation value of university-industry
collaborations, whereas in Section 4 the research methodology and approach are described. Finally, Section
5 and 6 discuss the main research results and conclusions, respectively.
2. THEORY
2.1. Knowledge Complementarity & Innovation
There is a strong consensus in the literature (e.g. Hamel and Prahalad, 1994) that the development of
innovation is strongly related to the organizations’ capability to collect and manage knowledge, since its use
and combination provide the creativity and the novelty necessary to move outside existing paradigms. In
fact, the innovation process can be conceived as an open process, where complementary and heterogeneous
inputs (i.e. pieces of knowledge) are transformed into outputs (i.e. results of innovations) (Katz and Khan,
1996).
However, organizations are becoming more and more specialized on specific fields of knowledge and, then,
rarely have all the required resources internally. Therefore, to successfully innovate they need to acquire
knowledge from other external sources, such as customers, suppliers, competitors, universities, research
centres, and other institutions (see also Freeman, 1987; Owen-Smith and Powell, 2004).
-2-
4. The importance of complementarity in the organizations’ innovation strategy is also well analysed by
Cassiman and Veugelers (2007), who demonstrate the tight relationship between organizations’ internal
R&D activities and external knowledge acquisition to effectively develop innovations, and access and
capture their benefits. Further light on the complementarity issue can be added quoting the Philips CEO
Gerard Kleisterlee (Economist, 2002), who stated that “we used to start by identifying our core competencies
and then looking for market opportunities. Now we ask what is required to capture an opportunity and then
either try to get those skills via alliances or develop them internally to fit”.
Thus, internal knowledge, mainly resulting by R&D activities, is not the only kind of knowledge managed
by organizations, which can also acquire new knowledge from the external environment by activating
collaborative R&D agreements with upstream and downstream sources of knowledge (such as suppliers, and
customers), and with other firms and scientific organizations (such as universities and research centres).
2.2. R&D Collaboration
Collaborative relationships are defined to include the direct and voluntary participation of two or more actors
in designing and/or producing a product or process (Polenske, 2004). The importance of collaboration in the
development of R&D activities has been extensively investigated by several scholars and literature streams.
In the Transaction Costs Economics (TCE), collaborative relationships are seen as hybrid forms of
organization between hierarchical transactions and arms length transactions in the market place (e.g.
Williamson, 1975; Pisano, 1990). Following this perspective, collaboration allows organizations to acquire
new competencies and to reduce the uncertainty and opportunistic behaviours associated to the development
and creation of new knowledge. In fact, organizations must constantly seek out new opportunities for
upgrading and renewing their capabilities. Nevertheless, acquiring capabilities entails uncertainty regarding
the value of the capability and the extent to which it can benefit the firm. Consequently, organizations may
benefit from having a network of knowledgeable collaborations that provides a reliable source of
information about options for enhancing competitive capabilities and minimizes opportunism, being the
partners involved in mutual knowledge exchanges (Nooteboom, 1999; Hagedoorn, 2002; Freel, 2003).
The importance of R&D collaboration to reduce opportunism has been also discussed by the Organizational
Theory, which analyses how inter-organizational ties are effective means to favour the diffusion and transfer
of complex knowledge, since they contribute to create a mutual trust, embeddedness, and social cohesion
between partners, necessary to overcome opportunistic problems and enhance innovation rise (e.g.
Granovetter, 1973; Reagans and McEviliy, 2003; Burt, 2004).
The Strategic Management literature has dealt with R&D collaborations, underlining how they can be used
by organisations as channels to reach and acquire external competencies, necessary to innovate and achieve
a sustainable competitive advantage. In fact, R&D alliances are often aimed at expanding an organisation’s
set of distinctive capabilities through inter-organisational learning, so to shape or respond to competitive
dynamics in a market (e.g. Mowery et al., 1998; Colombo, 2003; Goerzen and Beamish, 2005).
-3-
5. Finally, the Industrial Organisation literature has investigated the R&D collaboration issue, focusing on the
appropriability hazards. Specifically, knowledge presents the features of a public good, since the use by one
organisation of the information and new knowledge produced by R&D activities does not reduce the amount
of information available to other organizations. Furthermore, R&D activities are generally characterised by
an externality problem, since organisations involved in these activities cannot fully appropriate and exploit
the benefits for the occurrence of involuntary knowledge spillovers (Spence, 1984; d’Aspremont and
Jacquemin, 1988; Alcacer and Chung, 2007). Therefore, the establishment of collaborative R&D agreements
between organisations can contribute to control knowledge spillovers and, then, to internalize the positive
effects arising from R&D investments (e.g. Cassiman, 2000).
2.3. Universities as Sources of Knowledge Complementarity
In the previous sections the complementarity character of knowledge and the importance to establish R&D
collaborations as means to acquire such complementarity has been highlighted. Therefore, it is now
interesting to understand which organisations can represent effective sources of knowledge
complementarity.
It is commonly recognized that universities are important sources of new knowledge, especially in the area
of science and technology (see also Agrawal, 2001). In particular, several studies have shown the relevance
of universities as explorative organizations, stressing how they can act as bridges, allowing other
organisations to reach dispersed and heterogeneous information and pieces of knowledge (e.g. Saxenian,
1994; Varga, 2000; Adams, 2005; Audretsch et al., 2005). The knowledge gatekeeper character of university
is strictly related to its research activity, which gives the opportunity to i) access to a wide range of
industries, ii) learn the different knowledge from many industries, and ii) link knowledge across industries
and sectors.
Such gatekeeper character can make universities as ad hoc partners for firms to acquire heterogeneous and
complementary knowledge. In fact, universities have the ability to recombine and integrate such external
knowledge (Henderson and Cockburn, 1994) and act as knowledge brokers that span multiple markets and
technology domains and bring knowledge from where it is known to where it is not.
Recent studies have revealed an increasing attention towards university-industry R&D collaborations, as
channels through which knowledge can be transferred and acquired (e.g. Rothaermel and Thursby, 2005),
mainly focusing on firms and universities’ characteristics favouring such collaborations.
With this regard, Veugelers and Cassiman (2005) have empirically demonstrated that firms’ size, type of
industry, government support, and the involvement in complementary innovative activities positively affect
the likelihood to establish R&D collaborations with universities (see also Bercovitz and Feldman, 2007).
Regarding universities, the entrepreneurial orientation and the existence and productivity of technology
transfer offices (TTOs) are generally seen as the most important factors affecting the universities’ capability
to collaborate and develop joint innovations with the industrial environment (Rothaermel et al., 2007).
-4-
6. Nevertheless, few attention has been devoted to investigate and understand the role played by relational
attributes in explaining university-industry collaborations and their influence on the collaborations’ value.
Specifically, we are interested at analysing how technological relatedness, national culture similarity, and
prior collaboration ties may contribute to clarify why certain university-industry collaborations are more
valuable than others.
3. HYPOTHESES
In the present section, we develop a set of theoretical arguments that lead to the development of specific
hypotheses regarding how the three relational variables affect the innovation value of university-industry
R&D collaborations.
3.2.Technological Relatedness and Innovation Value
The notion of technological relatedness is based on shared technological experiences and knowledge bases
between organizations. It refers not to the technologies themselves, in terms of tools and devices used to
create new products and services, but to the knowledge actors possess about these technologies (Jaffe, 1986;
Mowery et al., 1996; Knoben and Oerlemans, 2006).
The importance of technological proximity is strictly related to the notion of absorptive capacity. In fact, as
shown by Cohen and Levinthal (1990), in order to successfully collaborate, the prior (technological)
knowledge of an organization must be similar to the new knowledge on the basic level, but fairly diverse on
the specialized level. Basic knowledge refers to the general understanding of the techniques upon which a
scientific discipline is based, whereas specialized knowledge refers to the specific knowledge used by the
actors in its everyday functioning. With this regard, Lane and Lubatkin (1998) show that organizations with
greater technological relatedness in basic technologies have greater relative absorptive capacity, and hence
are more likely to learn from each other.
This has to do with the technical and market competencies organisations own and have acquired when
dealing with specific technologies and markets. If these are not sufficient, search and imitation cost will
increase too much. In this vein, Perez and Vein (1988) stress a negative relationships between the current
knowledge base of an organization and the costs firms have to sustain to acquire the required knowledge of a
new technology. In fact, the authors argue that for each new technology exists a minimum level of
knowledge under which firms are incapable of bridging their knowledge gap.
However, when partners’ technological bases are too similar, it can be detrimental for learning and
innovation (Noteboom, 2000). In fact, it may result in a technological lock-in, in the sense that similar
knowledge bases limit the rising of new technologies or new market possibilities (Knoben and Oerlemans,
2006).
-5-
7. Divergences in technological specializations can be an important condition to establish R&D collaborations,
since it can allow partners to reach new and distinctive resources and capabilities (Colombo, 2003). In fact,
the exposure to partners’ different cognitive and technological frames may yield novel insights, as firms
benefit from “external economies of cognitive scope” (Nooteboom, 1999; Wuyts et al., 2005).
For instance, Sakakibara (1997) analyses the motivations of Japanese firms in participating in government-
sponsored R&D consortia and shows that firms perceive obtaining complementary knowledge and sharing
specialized skills as the most important objectives of such projects. Similarly, Brockhoff et al. (1991) find
that the possibility of capturing synergistic gains from the exchange of complementary technical knowledge
is the most important reason for collaborative R&D in Germany.
This reasoning leads to state that there may be an optimal amount of technology overlap between partners
that affects both the potential benefits (higher when partners are technologically distant) and the ability to
collaborate (higher when partners are close). Following Nooteboom (2000), it is possible to argue that too
little technological distance may imply a lack of sources of novelty, whereas too much technological
distance implies problems of communication and mutual understanding.
Thus, a non-monotonic relation between the technological relatedness and the value of the innovation
developed through university-industry collaborations may be expected.
Therefore, following this analysis, we argue that:
Hypothesis 1. Technological relatedness between universities and firms collaborating in R&D activities has
a curvilinear effect (inverted U) on the value of joint innovations.
3.2. National Culture and Innovation Value
Culture can be defined as the “complex whole which includes knowledge, beliefs, art, moral, laws, customs,
and any other capabilities and habits acquired by a man as a member of a society” (Taylor, 1871, p. 38).
Therefore, it is reasonably to assume that people belonging to the same community have a common culture
and system of opinions. Consequently, people of a same culture share the same tacit background and
ideology, adopt similar ways of thinking, behaving, deciding, and do not need to communicate a lot to
explain their opinion to other members of their culture, since the whole community grounds on the same
social awareness pre-existing and accumulated knowledge base.
In order to investigate the influence of cultural proximity on the knowledge transfer processes and
innovations development, we adopt a macro-level approach, focusing on the differences between continents,
nations, or regions’ culture, assuming that organisations located within the same geographical areas share the
same culture (Hofstede, 1980; Gerler, 1995).
In the business literature, several empirical studies have highlighted the importance of cultural proximity at
the macro-level, showing that this similarity can contribute to explain knowledge flows and partnerships
between organisations (e.g. Kogut and Singh, 1988; Folta and Ferrier, 2000; Hargadoorn, 2002; Van
Everdingen and Waarts, 2003). This depends on the tight relation between culture and institutions (Zuckin
-6-
8. and Di Maggio, 1990). In fact, organisations located in countries sharing similar cultures, are also
characterised by similar institutional frameworks, such as legislative conditions, labour relations, and
business practices, that can reduce transaction costs and, then, favour the likelihood of collaborations in
R&D activities, for instance providing analogous norms and laws for protecting intellectual property rights
(Capello, 1999; Kirat and Lung, 1999; Knoben and Oerlemans and, 2006).
These findings are also supported by some theoretical studies, suggesting that a similar culture encourages
coordination and facilitates transfer and feedbacks of information, and leads to a high rate of trust among
members, thus allowing communication and learning to proceed relatively smoothly (e.g. Maskell and
Malmberg, 1999; Knoben and Oerlemans, 2006).
The specificity of culture is seen as an important factor also for explaining university-industry collaborations
(Juniper, 2000). Specifically, studies on knowledge transfer between universities and firms in the Alsatian
region show the existence of few partnerships between French firms and German universities, due to the
cultural distance between the organisations (Heraud and Nanopoulous, 1994; Levy and Woessner, 2001). In
fact, when universities and companies collaborate in research activities institutional differences may
generate a great complexity in terms of coordination and arrangements, that can be mitigated by the
similarity between the cultural frameworks of the organizations’ countries.
Thus, we hypothesize that:
Hypothesis 2. Similar national culture between universities and firms collaborating in R&D activities has a
positive effect on the value of joint innovations.
3.3. Prior Collaborations Ties and Innovation Value
Strategic alliances and collaborations between organizations are now considered as a ubiquitous
phenomenon, that has received a great deal of attention from a number of perspectives.
Recently, scholars have focused on various path-dependent and sociological factors affecting the
performance of such collaborations, especially referring to innovation processes. With this regard, authors
have shown that higher level of familiarity, trust, and mutual understanding make existing relationships
efficient to establish and easy to maintain. Thus, prior collaboration ties have a clear and persistent influence
on the choice of future partners (Gulati, 1995; Hagedoorn et al., 2003; Goerzen, 2007; Kim and Song, 2007).
Moreover, it has been empirically demonstrated that this embeddedness has a positive effect on the transfer
of knowledge between actors, since it favours economies of time, integrative agreements, Pareto
improvements in allocative efficiency, and complex adaptation (Uzzi, 1997).
The underlying mechanisms of repeated collaborations are related to the establishment of inter-personal ties
that tend to increase over time, giving a greater understanding of each others’ needs and capabilities (Gulati,
1995). The existence of prior ties contributes to rise trust between management teams, which is transferred at
the level of inter-organizational trust (Zucker, 1986), and increases the transaction efficiency, in terms of
lower transaction costs (Zollo et al., 2002; Dyer and Chu, 2003; Goerzen, 2007; Kim and Song, 2007).
-7-
9. Given the specific nature of academic knowledge, R&D collaborations between universities and firms are
generally affected by high uncertainty, information asymmetries, transaction costs, and appropriability
hazards) (Hall et al., 2001; Veugelers and Cassiman, 2005), which can hamper the development of
innovations. Therefore, repeated collaborations may mitigate these problems for two main reasons. First, the
reputation effect (in terms of character, skill, reliability, competence, and other attributes) is essential to
exchange and it is an important platform to mitigate problems of information asymmetry and causal
ambiguity. Second, trust indicates a willingness to have openness to trade partners for value creation through
exchange and combination. Referring to the governance structure of R&D collaborations, trust offers a
sociological element of exchange giving more flexibility in operation and reducing coordination costs by
providing the ability to smooth conflicts (Murray, 2004; Lin, 2006).
Consequently, we suggest that:
Hypothesis 3. Prior collaboration ties between universities and firms collaborating in R&D activities have a
positive effect on the value of joint innovations.
4. METHODOLOGY
4.1. Research Setting
To empirically test our hypotheses we analyse the university-industry R&D collaborations, in terms of joint
patents, carried out by different universities belonging to the European Union (EU). In particular, we
consider the industry R&D relationships created by the three most innovative universities for each EU
country, identified on the basis of the overall number of patents registered at the EPO. The choice to
consider only the three most innovative universities is leaded by two main reasons. First, to investigate how
these organizations, generally considered as a benchmark in research activities at both the national and
international level, manage relationships to fully capture the benefits arising from industry collaborations.
Second, since we use patents as proxy for innovations, only the most innovative universities present a
sufficient set of relationships with the industrial environment for testing our hypotheses.
The use of patents as a proxy to evaluate innovations has been largely adopted in the literature, as shown by
several empirical works evaluating organizations’ innovative performance and the diffusion and transfer of
knowledge (e.g. Jaffe et al., 1993; Flor and Oltra, 2004; Singh, 2005; Fritsch and Slavtchev, 2007;
Nooteboom et al., 2007). Several factors can explain their intensive use (Ratanawaraha and Polenske, 2007).
First, patent data are readily available in most countries, thus permitting cross-country comparisons. Second,
the extensiveness of patent data enables researchers to conduct both cross-sectional and longitudinal
analysis. Third, patent data contain detailed useful information, such as the technological fields, the
assignees, the inventors, and some other market features. Finally, patents provide a measure of innovation
that is externally validated through the patent examination process (see also Griliches, 1990; Schilling and
-8-
10. Phelps, 2007), thus giving a certain degree of confidence to the relevance and result of the R&D
collaborations.
4.2. Sample
First, we identified all the universities, both public and private, located in each of the 27 countries of the EU,
thus defining a list of 812 universities. Then, we identified the three most innovative universities in each
country on the basis of the overall number of patents registered at the EPO between 1998 and 2003. From
this analysis, 81 universities have been classified. Finally, for each of the 81 universities, we analysed
patents jointly registered with firms. Thus, 29 universities have been selected, located in 12 different
countries and establishing 796 R&D university-industry collaborations.
To assess the value of the collaborations’ innovative output, we considered the patents registered between
1998 and 2003, since a moving window of five years is the appropriate time frame for assessing
technological impact (Stuart and Podolny, 1996; Henderson and Cockburn, 1996). In fact, studies about
R&D depreciation (e.g. Griliches, 1985) suggest that knowledge capital depreciates sharply, losing most of
its value within five years.
4.3. Dependent Variable
The analysis and assessment of patent value is a very debated and controversial topic, occupying a number
of pages on scientific journals. In the literature, several empirical strategies have been used to approximate
the patent’ value. Despite the strong heterogeneities across studies, in terms of indicators adopted, data
sources, time spans, and research methodologies, some similarities emerge. The most important one is that
the patent’s value is closely associated with the number of forward citations.
The use of forward citations has been introduced by the pioneer work of Trajtenberg (1990) and fully
developed by Jaffe et al., (1993) and validated as measure of patent’s value in numerous subsequent studies
(e.g. Hirschey and Richardson, 2001; Harhoff and Reitzig, 2002; Gittleman and Kogut, 2003; Harhoff et al.,
2003; Hall et al., 2005; Bonaccorsi and Thoma, 2007; Giuri et al., 2007; Singh, 2008).
Thereby, we measure the value (InnValue) associated to each innovation as the number of citations received
by each patent.
4.4. Independent Variables
Technological relatedness. The technological relatedness (TechRel) is evaluated by means of the degree of
overlapping between the organizations’ technological bases, in terms of technological fields in which they
patent. In particular, in this research the technological similarity is evaluated following the measure
proposed by Jaffe (1986), who uses the patent technological class information to construct a measure of the
closeness between two actors in the technology space. In this case the technology space is represented by the
-9-
11. 129 patent classes (three-digit) assigned by the International Patent Classification (IPC). Hence, the
technological relatedness is evaluated as:
f i f j'
Tech Re li , j = (1)
( f f )( f f )
i i
'
j j
'
where the vectors fi and fj (apex indicates the transposed vector) are constituted by all the patents registered
by the university (i) and the company (j) at the EPO from the previous five years up to date of the
collaboration, respectively, and allocated to the patent class n (n=1,…,129). Thus, the firms’ patent portfolio
is compared to the patent portfolio of each university has developed a patent with it. TechReli,j, which
represents the uncentered correlation between the two vectors, assumes value one, if i and j’s patent
activities perfectly coincide (fi = fj). On the contrary if they do not overlap at all, i.e. the two vectors are
orthogonal, it assumes value 0.
National Culture Distance. This variable aims at capturing the differences and similarities between national
cultural frameworks at the macro-level, in terms of norms and values of conduct. To achieve this goal, we
adopt the Kogut and Singh (1988) modified index of Hofstede that measures the cultural distance (CultDist)
between universities and companies collaborating in R&D activities (see also Clodt et al., 2006). In
particular, this index analyses four distinct dimensions: i) power distance (as the extent to which the less
powerful members of organisations and institutions accept and expect that the power is distributed
unequally), ii) individualism (as the degree to which individuals are not integrated into groups), iii)
masculinity (as the distribution of roles between the genders), iv) and uncertainty avoidance (as the society’s
tolerance for uncertainty and ambiguity). Through the analysis of these four key issues, a positive continue
index (CDij) is identified, which measures the institutional distance between actors i and j as:
CDij = ∑ { I di − I dj ) 2 / Vd }/ 4
4
( (2)
d =1
where Idj stands for the index for the d-th considered dimension and j-th actor, Vd is the variance of the index
of the d-th dimension.
Prior collaboration ties. To evaluate the existence of prior ties between universities and firms jointly
developing a patent, we account for previous research experiences between the partners. In particular, we
measure this variable as a binary one (PriorTies), assuming value one if, before the partnership under
analysis, the two actors have established previous R&D collaborations, in terms of other patents jointly
assigned. Otherwise, the variable assumes value zero. To identify such prior collaborations, we use a five-
year moving window following previous studies suggesting that the lifespan for alliances is usually no more
than five years (Kogut, 1988; Gulati, 1995; Kim and Song, 2007).
4.5. Control Variables
We include several variables to control for alternative factors that can explain the value of innovations
jointly developed by universities and firms.
- 10 -
12. We introduce dummy variables to control for industry fixed effects, since university-industry relationships
can be strongly affected by specific sector capabilities and competences (see also Pfeffer and Novak, 1978;
Pavitt, 1984). In particular, 14 main different industrial sectors are identified according to the standard
industrial classification (SIC): pharmaceuticals; engineering services; chemicals; industrial and commercial
machinery; electric services, measuring, analysing, and controlling systems; fabricated metal products;
transportation equipments; textile mills products; rubber and miscellaneous plastic products; food and
kindred products; business services; agriculture; fishing.
Then, we control for the firms absorptive capacity (Cohen and Levinthal, 1990) measured by means of firms
size (FirmSize), in terms of natural logarithm of number of employees, and natural logarithm of the overall
number of patents successfully filled from the previous five years up to date of the collaboration with
university (FirmPatents), which can be used also to take into account the technological capital owned by the
sampled companies (e.g. Phene et al., 2006; Nooteboom et al., 2007; Rothaermel and Boeker, 2008).
Regarding universities, we control for their entrepreneurial orientation and the existence of TTOs (see also
Debackere and Veugelers, 2005; Rothaermel et al., 2007). The entrepreneurial orientation has been widely
discussed in relation with the aptitude of universities to create new firms, such as spin-offs and incubators.
Thus, we introduce two binary variables, Spin-Off and Incubator, assuming value one if the universities have
created spin-offs or firms incubators, respectively. To control for the existence of TTOs, another dummy
variable (TTO) is introduced, which takes value one if the university has at least one technology transfer
office.
Other potential explanations to successful university-industry collaborations can be represented by
university’s reputation (UnivReputation) and university’s capability to be involved in scientific projects with
the industrial environment (UnivProjects). The former is measured following the Academic Ranking of
World Universities, compiled by the Shanghai Jiao Tong University’s Institute of Higher Education. The
report includes major institutes of higher education ranked according to a formula that takes into account
different criteria, such as teaching quality, staff quality, and research productivity, quality and efficiency. We
code UnivReputation as a dummy variable assuming value one if the sample universities are ranked in the
first ten positions.
UnivProjects is measured by means of the number of market-oriented and industrial R&D projects
developed by the sample universities during the observation period (1998-2003). Data are collected through
the EUREKA database, which provides several financial and technical information about European
university-industry joint projects aimed at creating innovative products, processes and services.
We control also for the university’s patenting propensity, as the natural logarithm of the overall number of
patents successfully filled by universities from the previous five years up to the date of the industry
collaboration (UnivPatents), and for their size, in terms of natural logarithm of number of full-time
researchers (UnivSize). In addition, we take into account the university’s country fixed effects. In particular,
country dummies are included to control for universities located in Belgium, Germany, Netherland, UK, that
- 11 -
13. count for about 80% of the overall number of university-industry relationships (see Table 1), and other
countries (Austria, Czech Republic, Denmark, France, Ireland, Italy, Poland, and Spain). Exogenous shocks
characterising the year of the relationship are also controlled.
Finally, we evaluate the effects of geographical distance between partners. Following a broad literature on
the effect of geography on learning and innovation rise (e.g. Audretsch and Stephan, 1996; Lublinski, 2003;
Siegel et al., 2003; Alcacer, 2006), we measure geographical distance (GeoDist) as a continue positive
variable, evaluated by the spatial distance (expressed in kilometres) between the location sites of universities
and companies jointly registered patents. To avoid problems related to companies’ multiple locations,
especially referring to multinationals, information about the site where the patents have been developed are
obtained analysing inventors’ addresses. Given the skewed distribution of the variable, also this variable has
been transformed using a log transformation.
In Table 1, all the model variables are described.
Table 1. Definition of variables.
Dependent variable
InnValue Number of citations received by each university-firm joint patent
Independent variables
Degree of overlapping between the technology profile of univeristy and firm jointly developing a
TechRel patent. The technology profile is represented by all patents registered by the university and the firm
from the previous five years to the date of the collaboration, and assigned to the 129 IPC (three-digit).
TechRel2 Squared term of the previous variable.
CultDist Degree of overlapping between the national cultures of unveristy and firm jointly developing a patent.
Dummy variable assuming value 1 if university and firm jointly developing a patent have registered
PriorTies
another patent in the previous five years.
Control variables
FirmSize Number of full time employers of each firm jointly developing a patent with university (Source:..).
Number of patents that each firm firm jointly developing a patent with university has registered from
FirmPatent
the previous five years up to the date of the collaboration.
UnivSize Number of full time researchers of each university.
Number of patents that each university has registered from the previous five years up to the date of
UnivPatent
industry collaboration.
Incubator Dummy variable assuming value 1 if university has at least one incubator.
Spin-off Dummy variable assuming value 1 if university has at least one spin-off.
TTO Dummy variable assuming value 1 if university has a technology transfer office.
Dummy variable assuming value 1 if university is ranked in the first ten positions of the Academic
UnivReputation
Ranking of World Universities.
UnivProjects Number of EUREKA projects developed by university during the observation period.
Natural logaritm of the physical distance expressed in kilometres between the location sites
GeoDist
(headquarter of local affiliates) of university and firm jointly developing a patent.
Industry dummies
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Pharma
pharmaceutical industry (SIC codes 2833, 2834, 2835, 2836).
- 12 -
14. Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
EngServices
engineering services industry (SIC codes 8711, 8712, 8713, 8748).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Chem
chemicals industry (SIC codes 281-, 282-, 285-, 286-, 287-, 288-, 289-).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
IndusMachinery industrial and commercial machinery industry (SIC codes 3531, 3552, 3556, 3559, 3565, 3568, 3569,
3682, 3585, 3589, 3599).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
ElectricServices
electric services industry (SIC codes 4931, 4939).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
MeasurSystems
measuring, analysing, and controlling systems industry (SIC codes 3823).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Metal
fabricated metal products industry (SIC codes 3443, 3449, 3479, 3498, 3499).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Transp
transportation equipments industry (SIC codes 3715, 3732, 3743, 3799).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Textile
textile mills products industry (SIC codes 2211, 2221, 2241, 2273, 2295, 2299).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Rubber
rubber and miscellaneous plastics products industry (SIC codes 3011, 3021, 3052, 3053, 3061, 3069).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Food
food and kindred products industry (SIC codes 2011, 2013, 2032, 2038, 2041, 2043, 2087, 2099)
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
BusinessServices
business services industry (SIC codes 7335, 7336, 7363, 7389).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Agric
agriculture industry (SIC codes 01-, 02-, 07-).
Dummy variable assuming value 1if firms jointly developing a patent with universities operate in the
Fish
fishing industry (SIC codes 0919, 0921)
University country
dummies
BE Dummy variable assuming value 1if university is located in Belgium.
DE Dummy variable assuming value 1if university is located in Germany.
NL Dummy variable assuming value 1if university is located in Netherland.
UK Dummy variable assuming value 1if university is located in United Kingdom.
Dummy variable assuming value 1if university is located in Austria, Czech Republic, Denmark,
Others
France, Ireland, Italy, Poland, and Spain.
Year dummies
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
1998
1998.
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
1999
1999.
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
2000
2000.
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
2001
2001.
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
2002
2002.
Dummy variable assuming value 1 if the patent jointly developed by university and firms is dated
2003
2003.
4.6. Estimation Model
- 13 -
15. The dependent variables of this study are represented by a nonnegative, integer count variable. Verified by a
statistical test of overdispersion (Gourieroux et al., 1984), the negative binomial estimation provides a
significant better fit for the data than the more restrictive Poisson model. Negative binomial regression
accounts for an omitted variable bias, while simultaneously estimating heterogeneity (Hausman et al., 1984;
Cameron and Trivedi, 1986). Thus, the following model is adopted:
P ( nit / ε ) = e − λit exp( ε ) λinit / nit −1!
where n is a nonnegative integer count variable, representing the value associated to each university-industry
relationship (patent). Therefore, P( nit / ε ) indicates the probability that each relationship (patent) has
received n citations in year t.
The application of a negative binomial estimation, jointly with a rich set of detailed control variables, allows
us to effectively address any potential endogeneity (Hamilton and Nickerson, 2003; Rothaermel and Hess,
2007).
5. RESULTS
In Table 2 basic descriptive statistics and pairwise correlations are reported. All the correlations between the
independent variables fall below the 0.70 threshold, thus indicating acceptable discriminant validity (Cohen
et al., 2003).
Table 2. Descriptive statistics and correlation matrix (N=796).
Panel (A): independent variables
Variables Mean S.D. Min Max 1 2 3 4
1. InnValue .477 1.304 0 12 1.000
2. CultDist .583 1.054 0 4.435 -.099 1.000
3. PriorTies .797 .417 0 1 -.033 -.070 1.000
4. TechRel .556 .308 0 .991 .071 -.026 .124 1.000
Panel (B); main control variables
Variables Mean S.D. Min Max 1 2 3 4 5 6 7 8 9 10 11
1. InnValue .477 1.304 0 12 1.000
2. FirmSize 7.134 2.706 1.792 13.541 .025 1.000
3. FirmPatent 3.709 3.081 0 11.278 -.032 .512 1.000
4. UnivSize 7.861 .832 5.561 8.854 -.006 .030 .184 1.000
5. UnivPatent 5.484 .898 3.178 6.942 .040 -.038 .124 .172 1.000
6. Spin-off .987 .111 0 1 .041 .015 -.025 -.042 .137 1.000
7. Incubaor .739 .438 0 1 .019 .150 .150 .363 .053 .113 1.000
- 14 -
17. 11. Belgium 105 .132 .339 .048
12. UK 286 .359 .480 .027
The results of the negative binomial regression are reported in Table 3. Model 1 loans only the control
variables, whereas in Models 2-5 the impact of technological relatedness, national culture distance, and prior
collaboration ties on innovation value is investigated. Regarding firm industry, university country, and
collaboration year fixed effects, the omitted industry is pharmaceutical, the omitted country is others, and the
omitted year is 1998.
Table 3. Negative binomial estimates of joint innovations’ value.
Dependent variable
ScienValue
Model 1 Model 2 Model 3 Model 4 Model 5
Independent variables
TechRel 1.796** 1.230*
TechRel2 -1.439* -1.134*
CultDist -.312*** -.332***
PriorTies -.232* -.251*
Control variables
FirmSize .003 .015 -.001 .005 .005
FirmPatent .007 -.002 .015 .001 .004
UnivSize -.821*** -.711*** -.528** -.761*** -.370
UnivPatent -.673 -.423 -.668 -.547 -.482
Incubator -1.575*** -1.303*** -1.448*** -1.443*** -1.129***
Spin-Off -1.313*** -1.346*** -1.285*** -1.473*** -1.388***
TTO 1.799*** 1.528** 1.401** 1.676** 1.127**
UnivReputation 5.524*** 5.622*** 5.364*** 5.792*** 5.619***
UnivProjects 0.163*** .142** .146** .149*** .122**
GeoDist .041** .049** .084*** .034* .089***
Industry dummies included included included Included included
University country dummies included included included Included included
Year dummies included included included Included included
Log likelihood -236.199 -234.303 -232.438 -235.160 -229.794
(*, **,***) ρ < 0.10 (0.05, 0.01).
Regarding control variables, firms’ characteristics have no impact on the innovation value, whereas
universities’ attributes seem to significantly affect it. Specifically, Table 3 shows that the presence of TTO in
academic organizations has a significant and positive impact on the scientific value, whereas the existence of
- 16 -
18. incubators and spin-offs has a negative influence. Moreover, the development of more valuable innovations
is favoured by the universities’ involvement in applied R&D projects and by their reputation.
Also geographical distance between partners matters, as showing by the positive and significant coefficients.
Probably, it is due to the spatial stickiness of knowledge. Thus, technological knowledge coming from
partners located in distant areas are generally characterised by different paradigms, providing a potential for
non-overlapping knowledge bases and favouring the creation of more radical and scientific valuable
innovations.
Firms developing rubber and miscellaneous plastic products are more able to achieve greater innovation
performance than pharmaceutical companies. Differently, the electric services sector is characterised by
lower values than the pharmaceutical one.
Universities located in Belgium and Netherland seem to scientifically perform better than academic
organizations located in Austria, Czech Republic, Denmark, France, Ireland, Italy, Poland, and Spain.
Finally, no statistical differences occur between dummy years.
Considering the independent variables, data reveal that technological relatedness has an inverted U-shaped
relationship with the innovation value, thus confirming Hypothesis 1. In fact, it emerges that it is necessary a
minimum threshold of technological similarity to favour mutual understanding, but an excessive value may
be harmful for discovering the novelty necessary to improve the scientific relevance of innovations.
Similarity between national cultures has a positive and significant impact on the innovation value, as shown
by β coefficients of cultural distance in Models 3 (-.312) and 5 (-.332), thus supporting Hypothesis 2.
Finally, also Hypothesis 3 is confirmed, since the existence of prior collaborations between universities and
firms positively affects the value of innovation. Thereby, it emerges that universities and firms that have
been previously involved in R&D collaborations have a greater likelihood to develop more valuable
innovations.
6. DISCUSSION & CONCLUSIONS
The present study wants to shed further light on the university-industry R&D collaborations, exploring how
relational attributes may influence the value of the innovations jointly developed. Previous works have
mainly investigated the role played by specific universities and firms’ attributes, such as universities’
entrepreneurial orientation, national policies, government support, types of industry, and the involvement in
complementary innovative activities, devoting few attention to the dyadic properties, rising from the
interaction between path-dependent partners characteristics. In particular, we have focused our study on
three key aspects: i) technological relatedness, ii) national cultural similarity, and iii) prior collaboration ties,
in order to show their impact on the collaboration innovative output.
- 17 -
19. Our results suggest that technological relatedness between universities and firms presents an inverted U-
shaped relation with the value of innovation. This finding reveals that to increase the relevance of
innovations a certain threshold of similar technological competencies is required. Nevertheless, too much
similarity may be detrimental since the development of valuable innovations requires dissimilar and
complementary bodies of knowledge, generally available in different technological partners.
In addition, confirming our second hypothesis, national cultural similarity between partners seems to be a
fundamental condition to improve the innovation value. In fact, the similarity between countries’ rules, laws,
norms, and values can provide a common ground on which technological strategies can be based, thus
favouring goals alignment and the achievement of innovative results.
Finally, also prior ties positively contribute to enhance the value associated to joint innovations. In fact,
previous collaborations may promote the creation of an initial base of inter-partner trust, so developing such
relational routines useful to proceed further to the joint development and ownership of technologies.
The present study contributes to the existing literature on university-industry relationships, stressing the
relevance of specific relational attributes and how these may predict the development of successful joint
innovations. With this regard, our findings seem to suggest that policy makers should promote and support
the establishment of university-industry collaborations, considering also partners’ specific relational
features. Thereby, founds and aids destined to sustain collaborative R&D projects between academic
organizations and companies should be allocated not only evaluating the specific project and partners’
characteristics, but also taking into account how these characteristics interact. In fact, we have shown that
the relation between organizations’ technological bases, cultural frameworks, and the degree of past mutual
experiences may significantly impact on the value of the resulting innovations.
Of course the paper presents some limitations. First, the use of joint patents is not able to capture all the
university-industry collaborations. However, since we are interested in analysing successful collaborations,
joint patents can describe with a certain degree of confidence the success of such partnerships in terms of
innovations development (see also Kim and Song, 2007). Second, joint patents between universities and
companies are often registered only with the name of the researcher(s) and firms engaged in the innovations
development. Nevertheless, we have not considered these cases, since our focus is represented by the
interactions between universities and firms at the institutional level. To include also collaborations of single
professors and researchers with the industrial environment, other aspects, more devoted to capture the social
dynamics occurring between the academic and industrial environment, should be analysed.
Third, the study focuses only on the impact that three specific relational attributes, dealing with
technological competencies, culture, and embeddedness, exert on the value of resulting innovations. Future
studies could complement the present work investigating how these attributes may differently affect the
innovation value, according to both its scientific or economic relevance and the more explorative or
exploitative collaboration aim.
- 18 -
20. Finally, future studies could validate and improve the robustness of our results extending the research
setting, in order to include industry R&D collaborations also established by non-European universities.
REFERENCES
Adams, S.B. (2005): Stanford and Silicon Valley: lessons on becoming a high-tech region, in: California Management
Review, 48, p. 29-51.
Agrawal, A. 2001. University-to-industry knowledge transfer: literature review and unanswered questions.
International Journal of Management Reviews, 3, 285-302.
Alcacer, J., Chung, W. (2007) Location strategies and knowledge spillovers, Management Science, Vol. 53, pp. 760-
776.
Audretsch, D.B; Lehmann, E.E.; Warning, S. (2005): University spillovers and new firms location, in: Research
Policy, 34, p. 1113-1122.
Bercovitz, J.E.L., Feldman, M.P. 2007. Fishing upstream: Firm innovation strategy and university research alliances.
Research Policy, 36, 930-948.
Burt, R.S. (2004) Structural holes and good ideas, The American Journal of Sociology, Vol. 110, pp. 349-399.
Cassiman, B. 2000. Research joint ventures and optimal R&D policy with asymmetric information. International
Journal of Industrial Organization. International Journal of Industrial Organization, 18, 283-314.
Cassiman, B., Veugelers, R. (2007) In search of complementarity and innovation strategy: internal R&D and external
knowledge acquisition. Management Science, 52, 68-82.
Cohen, W.M., Levinthal, D.A. (1990) Absorptive capacity: a new perspective on learning and innovation,
Administrative Science Quarterly, Vol. 35, pp. 128-152.
Colombo, M.G. (2003) Alliance form: a test of the contractual and competence perspective, Strategic Management
Journal, Vol. 24, pp. 1209-1229.
Debackere, K. and R. Veugelers (2005), ‘The role of academic technology transfer organizations in improving
industry science links,’ Research Policy, 34(3), 321–342.
Etzkowitza, H., Leydesdorff, L., 2000. The dynamics of innovation: from National Systems and “Mode 2” to a Triple
Helix of university–industry–government relations. Research Policy 29, 109-123.
Freel, M.S. (2003) Sectoral patterns of small firm innovation, networking and proximity, Research Policy, Vol. 32, pp.
751-770.
Freeman, C. (1987) Technology Policy and Economic Performance: a lesson from Japan (London, Pinter).
Goerzen, A. 2007. Alliance networks and firm performance: the impact of repeated partnerships. Strategic
Management Journal, 28, 487-509.
Goerzen, A., Beamish, P. 2005 The effect of alliance network diversity on multinational enterprise performance.
Strategic Management Journal, 26, 333-354.
Goureoux, C., Monfort, A., Trognon, A. 1984. Pseudo maximum likelihood methods: theory, Econometrica, 52, 681-
700.
Granovetter, M. (1973) The strength of weak ties, American Journal of Sociology, Vol. 78, No. 6, pp. 1360–1380.
- 19 -
21. Hagedoorn, J. 2002. Inter-firm R&D partnerships: an overview of major trends and patterns since 1960. Research
Policy, Vol. 31, pp.477-492.
Hamel, G., Prahalad, C.K. (1994) Competing For The Future (Cambridge, MA, Harvard Business School Press).
Hausman, J., Hall, B., Griliches, Z. 1984. Econometric models for count data with an application to the patents-R&D
relationship. Econometrica, 52, 909-938.
Henderson, R. and I. Cockburn, 1996, Scale, Scope and Spillovers: The determinants of Research Productivity in Drug
Discovery, RAND Journal of Economics, 27, 32-59.
Jaffe, A.B. (1986) Technological opportunity and spillovers of R&D: evidence from firms’ patents, profits, and
market values, The American Economic Review, Vol. 76, pp. 984-1001.
Katz, D., Kahn, R. (1996) The Social Psychology of Organizations (New York, Wiley).
Kim, C., Song, J. 2007. Creating new technology through alliances: An empirical investigation of joint patents.
Technovation, 27, 461-470.
Knoben, J., Oelremans, L.A.G. (2006) Proximity and inter-organizational collaboration: a literature review,
International Journal of Management Review, Vol. 8, pp. 71-89.
Kogut, B., Singh, H. 1988. The effect of national culture on the choice of entry mode. Journal of International
Business Studies, 19, 411-432.
Lane, P.J., Lubatkin, M. (1998) Relative absorptive capacity and interorganizational learning, Strategic
Management Journal, Vol. 19, pp. 461-477.
Mowery, D.C., Oxley, J.E., Silverman, B.S. (1998) Technological overlap and interfirm cooperation: implications for
the resource-based view of the firm, Research Policy, Vol. 27, pp. 507-523.
Nooteboom, B. (1999) Innovation and inter-firm linkages: new implications for policy, Research Policy, Vol. .28, pp.
793-805.
Nooteboom, B. (2000) Learning and Innovation in Organizations and Economies (Oxford, Oxford University Press).
Nooteboom, B., Van Haverbeke, W., Duysters, G., Gilsing, V., van den Oordc, A. 2007. Optimal cognitive distance
and absorptive capacity. Research Policy, 36, 1016-1034.
Owen-Smith, J., Powell, W.W. (2004) Knowledge networks as channels and conduits: the effects of spillovers in
Boston biotechnology community, Organization Science, Vol. 15, pp. 2-21.
Pisano, G., 1990. The R&D boundaries of the firm. An empirical analysis. Administrative Science Quarterly, 35, 153-
176.
Polenske, K.R. (2004) Competition, collaboration and cooperation: an uneasy triangle in networks of firms and
regions, Regional Studies, Vol. 38, pp. 1029-1043.
Reagans, R., McEvily, B. (2003) Network structure and knowledge transfer: the effects of cohesion and range,
Administrative Science Quarterly, Vol. 48, pp. 240-267.
Rothaermel, F.T., Agung, S.D., Jiang, L. 2007. University entrepreneurship: a taxonomy of the literature. Industrial
and Corporate Change, 16, 691-791.
Saxenian, A. (1994): Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge,
MA: Harvard University Press.
Stuart, T.E., Podolny, J.M. (1996) Local search and the evolution of technological capabilities, Strategic Management
Journal, Vol. 17, pp. 21-38.
- 20 -
22. Uzzi, B. (1997) Social structure and competition in interfirm networks: the paradox of embeddedness, Administrative
Science Quarterly, Vol. 42, pp. 37-69.
Varga, A. (2000): Local academic knowledge spillovers and the concentration of economic activity, in: Journal of
Regional Science, 40, p. 289-309.
Veugelers, R., Cassiman, B. 2005. R&D cooperation between firms and universities. Some empirical evidence from
Belgian manufacturing. International Journal of Industrial Organization, 23, 355-379.
Williamson, O.E. 1975 Markets and hierarchies, analysis and anti-trust implications: A study in the economics of
internal organization. New York: Free Press,.
Wuyts, S., Colombo, M.G., Dutta, S., Noteboom, B. (2005) Empirical tests of optimal cognitive distance, Journal of
Economic Behavioral & Organization, Vol. 58, pp. 277-302.
Zukin S. and Di Maggio P. (Eds) (1990) The Social Organization of the Economy. Cambridge University Press,
Cambridge.
- 21 -
