Does technological diversification spur university patenting?

Abstract

Technological diversity, or the breadth of technological knowledge embedded in patented inventions, refers to the range of different technological or economic fields covered by a patent. This paper explores the role of diversification scope in encouraging the production of new patents in European universities by including the diversification scope as an explanatory variable in a patent production function. We hypothesize that the more diversified the patented technology in the university, the greater the production of new patents in subsequent periods. To test this hypothesis we rely on a cross-sectional sample of patents owned by 141 European universities across Europe in 2001-2004. Our empirical findings support the hypothesis of diversification, which means that the production of new patents can be spurred by promoting or stimulating greater levels of technological diversification. This result is robust to both the use of various measures of diversification and to different econometric specifications.

Full text

1
Does technological diversification spur university patenting?
Manuel Acosta
Daniel Coronado*
M. Ángeles Martínez
UNIVERSITY OF CADIZ
(SPAIN)
Address:
Facultad de Ciencias Económicas y Empresariales
Duque de Nájera, 8
11002-CADIZ
SPAIN
Abstract: Technological diversity, or the breadth of technological knowledge embedded in patented
inventions, refers to the range of different technological or economic fields covered by a patent. This
paper explores the role of diversification scope in encouraging the production of new patents in European
universities by including the diversification scope as an explanatory variable in a patent production
function. We hypothesize that the more diversified the patented technology in the university, the greater
the production of new patents in subsequent periods. To test this hypothesis we rely on a cross-sectional
sample of patents owned by 141 European universities across Europe in 2001-2004. Our empirical
findings support the hypothesis of diversification, which means that the production of new patents can be
spurred by promoting or stimulating greater levels of technological diversification. This result is robust to
both the use of various measures of diversification and to different econometric specifications.
Keywords: University patenting, technological diversification, entropy index, multilevel negative
binomial model, knowledge production function, European universities.
JEL Classification: O31, O32
This is an original manuscript of an article published by Springer in Journal of
Technology Transfer in 2018, available at https://doi.org/10.1007/s10961-015-9414-y
Please, quote as:
Acosta, M., Coronado, D., & Martínez, M.A. (2018). Does technological diversification
spur university patenting? The Journal of Technology Transfer, 43, 96-119.
* Daniel Coronado. daniel.coronado@uca.es

2
1 Introduction
University patenting is one of the roles that universities conduct as part of the “third mission”,
conceptualised as an additional function to teaching and research (Etzkowitz et al., 2000). Other activities
include a variety of technology transfer mechanisms such as spin-offs, licensing, entrepreneurial
education, publications with industry scientists, etc. (Etzkowitz and Zhou, 2006; Rothaermel et al. 2007;
Link et al. 2007). Although patents are a small part of the knowledge produced by universities, the
analysis of university patenting is relevant because of its consequences for both the productive system and
the university.
On the one hand, the positive effects of university patenting on industry stem from at least three
separate but complementary factors. Firstly, there is a direct contribution when a university produces
useful patented technologies with some application to industrial processes and transfers them to private
firms, thereby increasing private innovation and inducing regional economic growth (e.g. Bercovitz and
Feldmann, 2006; Carlsson et al. 2009). Secondly, the production of a patent may affect innovation in
surrounding areas because of the flow of technological knowledge between universities and firms. This
flow of knowledge can occur through a variety of interaction channels between academics and firms
(such as by reading the patent, or via direct conversation or informal meetings with the inventors). This
situation may induce localized spillovers from university to industry, with effects on firms’ innovative
outputs (Anselin et al. 1997, 2000; Feldman and Florida, 1994; Fischer and Varga, 2003; Jaffe, 1989;
Varga, 1998) or new company formation across geographical areas (Acosta et al. 2011; Audretsch and
Lehmann, 2005; Di Gregorio and Shane, 2003; Woodward et al. 2006; Zucker et al. 1998, Bonaccorsi et
al. 2014). Thirdly, academic researchers may use patents to exploit these technological outputs by
creating new companies (spin-offs) that will increase the technological capacity of neighbouring areas
(Chiesa and Piccaluga, 2000; Mustar et al., 2006, Djokovic and Souitaris, 2008; O’Shea et al. 2008; Zahra
et al. 2007).
On the other hand, patenting also has positive consequences for universities. Patents have great
potential as a source of licensing revenues for academic institutions (Siegel et al. 2004, Link et al., 2007).
Universities benefit from patenting when they gain revenue through licensing or by participating in spin-
offs to exploit the results of a patent (Geuna and Nesta, 2006; Sterckx, 2011, include some recent data for
the US and Europe). Moreover, in some contexts, there is evidence that researchers can tap into

3
technological knowledge to develop new ideas for research in related scientific fields; for example,
Stephan et al. (2007) found that patenting US academic researchers publish more than do members of a
non-patenting control group. Carayol (2007) and Breschi et al. (2007) obtained similar results using data
from Europe. Van Looy et al. (2006) concluded that inventors publish significantly more than their
colleagues (non-inventors), which suggests that the two activities - research and patenting - may actually
reinforce each other. A recent paper by Crespi et al. (2011) indicates that in the UK academic patenting
may complement publishing, at least up to a certain point. However, patenting and licensing have caused
some concern about the extent to which the use of proprietary protections for knowledge detracts from
their primary mission as creators of public knowledge. According to Hall and Harhoff (2012), this is one
of the main reasons for the enormous interest in this topic by academic researchers, raising questions
about the effects of patenting on the quantity and quality of research, the changes in its nature towards
other types of investigations focused on commercial ends, or the consequences of university patenting on
the diffusion of knowledge by privatizing some of it. Although some literature also points to a
substitution effect, most of the empirical evidence reveals no such trade-offs and even signals positive
effects (Franzoni, 2009, Van Looy et al. 2011, Franzoni and Scellato, 2011).
This previous literature suggests several arguments that support the importance of analysing patenting
activities in universities. Particularly, providing insight into the causes encouraging or hindering the
capacity of universities to produce patents would help to design policy measures in order to foster the
production of university patents. In this respect, an important consideration in managing technology in
universities is technological diversification, which has an implication on how widely a technology can be
applied (Lerner, 1994; Leten et al. 2007; Toh, 2014). More specifically, technological diversity, or the
breadth of technological knowledge, refers to the antecedent quality of the body of knowledge, which
allows creativity and generates new perspectives and insights into a given problem (Lettl et al. 2009). The
extent to which the production of university technology is diversified or specialised can have
consequences on the academic outputs. The scarce empirical research on this topic suggests, for example,
that the breath of knowledge embedded in patents affects the quality of the patent (Tantiyaswasdikul,
2012), the probability of commercialization (Nerkar and Shane, 2007), or its value (Jaffe, 2000), which in
turn is relevant for technology transfer. Moreover the immediate consequence of widening or narrowing
the scope of technology produced in universities may go beyond the university itself. According to Rosell

4
and Agrawal (2009), university flows of knowledge are narrowing (even only those associated with
patented inventions in certain fields), and this fact could throw into question the traditionally conceived
arrangement between academia and society, affecting policies and economic growth.
In this study, we examine the role of technological diversification in the production of patents by
including the diversification of technology at university level as an explanatory factor in a patent
production function. By doing so, we add an additional aspect to the literature on the factors affecting
university patent production that, to our knowledge, has not been analysed so far. Furthermore, our
approach differs from previous econometric studies with respect to the empirical framework. While
previous works have applied one level of analysis (e.g. university), we put forward a multilevel model in
which the dependent variable is the number of universities, and the independent variables include
university characteristics (the technological scope along with other variables) and control for regional
factors at a second level. For testing the role of technological diversification in the production of patents
we use the university as unit of analysis. Based on a knowledge production function, we estimate several
empirical models using a cross-sectional sample of university-owned patents for 141 European
universities located in 74 European regions.
The paper is organised as follows. Section 2 summarises the relevant literature. Section 3 describes
the data and the patterns of university technological diversification in university-owned patent production
across European universities. Section 4 presents the models, explaining the effects of technological
diversification on the production of new patents. Section 5 provides the results and robustness checks. We
briefly summarise the conclusions and provide some policy implications in Section 6.
2 Literature review
In recent years an increasing number of studies carried out both in Europe and in the US has
contributed to our knowledge on the factors affecting the production of patents in universities. We know
which universities produce more patents, which receive more revenues in licensing, in which fields
patents are being issued, who is patenting in the university and how patent activity relates to personal
characteristics (Stephan et al., 2007; Azoulay et al., 2007). We also know something about certain
institutional and regional factors encouraging academic patent activities and others that are hampering the

5
production of patents. For example, prior empirical research on the determinants of university patenting
has largely evaluated regulatory changes such as the US Bayh-Dole Act (e.g., Henderson et al., 1998,
Mowery et al. 2001) or IPR regulatory policies in Europe (Geuna and Rossi 2011). Other research has
focussed on university-specific factors (for example, the size, prestige of the university, universities’
policies, etc.) and whether the economic environment can affect the production of patents at university
level (Coupe, 2003; Payne and Siow, 2003; Saragossi and de la Potterie, 2003; Azagra-Caro et al. 2003;
Dai et al., 2005; Geuna and Nesta, 2006; Baldini, 2006; Acosta el al. 2009; Rizzo and Ramaciotti, 2014).
However, the question of whether the breadth of the technologies embedded in university patents can
contribute to or hinder the production of new patents remains unanswered. In the following paragraphs
we advance some positive and negative consequences of technological diversification.
Universities, as with other organisations, can develop knowledge in a variety of technological
domains; they can make choices about the knowledge fields in which they want to develop their research
and technological activities. By deciding upon the development of knowledge in one technological
domain (or discipline), they reduce their options to develop expertise in other domains, and hence
decisions regarding the knowledge depth and breadth become strategic questions with consequences in
the long run (Moorthy and Polley, 2010). Therefore, there are two alternatives. On the one hand,
universities may follow a technological diversification strategy expanding the production of technology
base embedded in patents into a wide range of fields. On the other hand, universities may promote
specialisation by reinforcing and stimulating the production of technology in a small number of areas.
Both options could produce different outcomes. While the final decision about university research might
remain centred on the individual researcher, some studies show that these decisions are influenced by the
university environment, societal forces, and policy (Dai et al., 2005).
In order to provide clues about which hypothesis (diversification or concentration) would be more
compatible with the increase of patent performance at universities, there are two questions to address.
First, we discuss some advantages and drawbacks of one or another strategy. Second, we summarise some
empirical background.
With respect to the first question, while the role of technological diversification has been studied at
company level, there is not much research at university scope. Some literature has analysed the
advantages of diversifying the production of technology. For example, Wade and Gravill (2003) argue

6
that it reduces the risks of innovation and creates synergy in cross-disciplinary technology integration
(Gambardella and Torrisi, 1998), thereby enhancing the sources of competitive advantage. Henderson and
Cockburn (1996) point out the benefits of diversity in the research agendas of pharmaceutical firms. They
suggest that having a range of research approaches and expertise within firms permits the cross-
fertilization of ideas through knowledge spillovers between units and therefore greater innovative output.
Granstrand (1998) argues that a technological diversification strategy can help firms to enhance
innovation efficiency because diversification can stimulate firms to generate innovative ideas through the
combination and recombination of various technologies. By maintaining a broad technology portfolio,
firms can explore and exploit new opportunities emerging from scientific and technological
breakthroughs (Lin et al. 2006). Some drawbacks are also inherent in technological diversification.
Coordination costs may increase exponentially with higher levels of technological diversification because
more diversified firms are more likely to encounter difficulties in combining mature, or exploitative,
technologies with explorative trajectories (Leten et al. 2007). Uncertainty in conducting sometimes
unrelated activities (Brown, 1992) is another of the main negative consequences. Some of these
advantages and drawbacks at company level can be observed at university level. For example, in
academic circles, a relationship between the diversity of knowledge and information accessed via network
connections during the various stages of the research process should positively influence both an actor’s
creativity and the knowledge creation quality since knowledge diversification offers opportunities to
exploit and combine inputs from different sources (Chen and Liu, 2012). The costs of co-ordination are
probably higher for a technological diversification strategy compared to a concentration option. For
example, diversification in university research sometimes stems from the incorporation into a project of
researchers from different universities, working in related fields. Many (though not all) European funding
opportunities require some sort of collaboration with partners in other European countries; for example,
most of the funding in the framework of the European strategy Horizon 2020 require organisations from
at least three different countries. When several universities are involved in a project, some face-to-face
contact between researchers is necessary for co-ordination and exchange of ideas, increasing the
displacement costs. Another risk is uncertainty; there is no guarantee that technological diversification
will increase the likelihood of generating technological outputs in the form of patents. Sometimes,
attempts to cope with several fields of research result in fewer outputs because of lack of expertise in

7
some of them, or perhaps there is the same quantity of outputs, but of worse quality than in a
specialisation scenario.
With regard to the empirical background, the consequences of technological diversification has been
widely analysed at firm level (e.g. Breschi et al. 2003; Lin et al. 2006; García-Vega, 2006; Quintana-
García and Benavides-Velasco, 2008; Chiu et al. 2010). Overall, this literature confirmed the positive
impact of technological diversification on firms’ technological performance. However, the question of
whether the scope of technological knowledge embedded in university patents affects the production of
new patents remains quite unexplored. Only a few number of papers have dealt with a related topic about
the breadth (knowledge diversification) created in universities with several outcomes, such as the quality
of patents and commercialisation. The research by Rosell and Agrawal (2009) examined the technological
breadth of university inventions, which are more likely to be cited by a more concentrated set of
subsequent patent owners. Their results suggest that the rapid rise in patenting during the 1980s in the US
was associated with a narrowing of knowledge flows both to and from universities, but only in specific
technology fields. Tantiyaswasdikul (2012) investigated the effect of the breadth of patent protection –
including both of the number of IPC classes and the number of claims - to account for breadth or
diversification on the value of patents measured with patent citation. Using a sample of 1120 Japanese
university patents the author found a significant effect and concluded that patent diversification does
influence the value of university patents. Nerkar and Shane (2007) examined the effect of the attributes of
inventions on their commercialisation, particularly technology scope. Using patents assigned to the
Massachusetts Institute of Technology between 1980 and 1996 and licensed to a private firm, their work
supports the hypothesis that the greater the scope of the invention’s patent, the greater the likelihood of its
commercialisation. These two papers show the positive effects of technological diversification of
university patents on its quality and commercialisation, but not on the subsequent performance.
Diversification seems to bring benefits for technology transfer as well. Some research has shown that
broader scope of patented inventions provides a wider range of alternative inventions that can be blocked
by the patent. The size of the region of technology space may exclude other patentees from operating, and
then a broader patent is more valuable to the patentee (Jaffe, 2000). This idea suggests that a
diversification patent generates greater appropriateness and increases the value of the patented invention
(Merges and Nelson, 1990; Lerner, 1995; Shane, 2001). To summarise, the empirical research on

8
technological diversification in other institutions (firms) shows some benefits on technological
performance, and the limited research of this topic at university level also points in the same direction: a
positive relationship between diversification and some indicators such as the patent quality and patent
commercialisation. This similarity with respect to the empirical research between firms and universities is
not surprising, given that universities have become entrepreneurial in their inner dynamic as well as
through external connections with business firms in research contracts and transfer of knowledge and
technology (Etzkowitz, 2003). Moreover, although the intra-organisational processes of these academic
units are relatively nascent and fragmented (Hunter et al. 2011), some similarities between the production
of knowledge in firms and universities are clea,r to the extent that universities are divided into research
groups that operate as firm-like entities; they lack only a direct profit motive to make them companies
(Etzkowitz, 2003). On these grounds, we expect a positive causal relationship between technological
diversification and the production of university patents in subsequent periods.
3 Patent specialisation across European universities
3.1 University-owned patents across European universities
Following the terminology in Lissoni et al. (2008), ‘academic patenting’ comprises both university-
owned patents, in which the property is retained by the universities that employ the academic inventors,
and university-invented patents, where at least one of the inventors is from a university but the owner of
the patent is a firm. This latter category of patents consists of inventions arising from collaboration with
industry, with three actors involved in negotiations over the intellectual property: the academic inventor,
his/her business partner, and the university administration (Lissoni et al, 2013)
1
. In this paper we consider
university-owned patents. Lissoni (2012) suggests that, by looking solely at university-owned patents, we
may end up losing sight of a substantial amount of inventing activity that is registered under a company’s
or an individual’s heading. However, although our choice might be seen as a partial view (we exclude
invented patents), the fact is that university-owned patents fit better to the objective of this paper for the
1
Geuna and Nesta (2006) concluded that the number of university-invented patents is higher than the number of
patents owned by universities. Lissoni et al. (2008) provide figures to show that university-owned patents in France,
Italy and Sweden constitute no more than 10% of all academic patents (most patent ownership lies with firms),
although this can be as high as 69% in the US. Recently, some European countries have introduced legislative
changes in an attempt to retain the rights of invention where universities produce the information included in patents;
see Geuna and Rossi (2011) for a discussion.

9
following reason. Our research question on how diversification of knowledge embedded in university
patents can produce additional, subsequent technological knowledge is an issue that, if corroborated,
would require university competence to put in practice measures affecting the diversification or
specialisation of patented technology. This control competence over technological research is easier for
university-owned patents than for university-invented patents because the former relies on university’s
resources while the latter are totally or partially funded by the firms that end up owning the intellectual
property. Some empirical research supports this idea; for example, university-owned patents have been
found to be more responsive to public funding and university-invented patents more responsive to private
funding (Azagra-Caro et al. 2006; Azagra-Caro, 2014).
In order to construct our sample of university-owned patents we used the following criterion. A patent
was assigned to a university when the name of the university appeared in the list of applicants
2
. When
there was more than one university applicant, each was assigned a patent. To avoid national distortions
arising from differing patent application requirements in different countries, we included only European
Patent Office patents in our sample. We examined the institutions named in the European Indicators,
Cyberspace and the Science–Technology–Economy–System (EICSTES) Project and in the Worldwide
Web of Universities (http://univ.cc, a web site with links to 7,884 universities in 190 countries).
Our search resulted in 3,330 European university-owned patents (obtained from the Derwent
Innovation Index) related to 391 universities located in Europe-15 (before the 2004 enlargement) for the
period from 2001 to 2004
3
. Of the 391 universities, 31 have no university-owned patents; 192 universities
have between one and five such patents; 89 have between six and 10; 37 have 11 to 20; 18 have 21 to 30;
and 24 universities own more than 30 patents. On average, there are 8.5 patents per university for the
whole period.
Table 1 presents the top 10 universities with the highest number of owned patents. Three British
universities lead the ranking, accounting for 17.5% of all patents. That Table also shows that owned
patents appear to be concentrated in a few universities; only ten universities (2.5% of the sample) account
2
There are two issues that we had to solve. First, sometimes the college appears as applicant instead of the university.
Second, it is common that different names and/or abbreviations for the university appear as applicant of the patent. To
avoid losing information we elaborated a complete list of universities (and colleges) considering different
abbreviations as well. (This list is available upon request).
3
Although the first search was performed using the Derwent Innovation Index, we have recently checked and
compared the entire information with another database (PATSTAT), with little change with respect to our original
data.

10
for 25.6% of patents. University-owned patents can be analysed by looking at the distribution across
technological fields. For this purpose, we have classified all patents in the sample according to the
technological subclasses of the International Patent Classification (IPC) at four digit-level. The IPC offers
a means of obtaining an internationally-uniform classification of patent documents of which the primary
objective is the establishment of an effective search tool for the retrieval of patent documents.
Researching the IPC codes is a way to find out the various technologies embodied in a patent because
each classified ‘thing’ in a patent is made according to its nature, function, purpose or application (Guide,
2012)
4
. The last column in Table 1 shows the number of different technological subclasses for all the
patents owned by each university and Table 3 exhibits the ten most dynamic technological subclasses of
the IPC in which European universities have patented in 2001-2004. As this table depicts, more than a
half of all university-owned patents include at least one of those IPC codes.
Table 1. Top Universities by number of university- owned patents ( 2001-2004)
University-owned patents
No.
IPC (*)
University
Country
No.
%
Cum. %
University of London
GB
243
7.30
7.30
114
University of Oxford
GB
168
5.05
12.34
99
University of Cambridge
GB
89
2.67
15.02
81
Ghent University
BE
61
1.83
16.85
48
University of Southampton
GB
50
1.50
18.35
42
Catholic University of Leuven
BE
49
1.47
19.82
51
Universitat Politecnica de Valencia
ES
49
1.47
21.29
39
University of Bristol
GB
48
1.44
22.73
46
Delft University of Technology
NL
48
1.44
24.17
59
Leiden University
NL
47
1.41
25.59
27
University of Manchester
GB
46
1.38
26.97
44
Eberhard-Karls-Universitat Tubingen
DE
45
1.35
28.32
32
University of Glasgow
GB
45
1.35
29.67
27
Catholic University of Louvain
BE
43
1.29
30.96
45
Utrecht University
NL
40
1.20
32.16
20
Others
2259
67.84
100
All Universities in the sample (391)
3330
100
100
(*) Number of different technological fields (IPC codes at four digit-level).
Source: Derwent Innovation Index.
4
Ava il ab le a t http :/ /w ww. wi po .int /c la ss if ic at ion s/ ip c/ en /g ene ra l/

11
Table 2. Ten IPC codes with the greater number of university onwed-patents
IPC Codes
Patents including the IPC code(*)
No.
%
Cum. %
A61K
1051
10.93
10.93
C12N
769
7.99
18.92
G01N
760
7.90
26.82
A61P
686
7.13
33.95
C07K
535
5.56
39.51
C12Q
509
5.29
44.80
C07H
217
2.26
47.06
B01J
199
2.07
49.13
C12P
198
2.06
51.19
H01L
194
2.02
53.20
A61K: Preparations for medical purposes; C12N: Micro-organisms or enzymes;
compositions thereof; G01N: Investigating or analysing materials by
determining their chemical or physical properties; A61P: Specific therapeutic
activity of chemical compounds or medicinal preparations; C07K: Peptides;
C12Q: Measuring or testing processes involving enzymes or micro-organisms;
C07H: Sugars; derivatives thereof; nucleosides; nucleotides; nucleic acids;
B01J: Chemical or physical processes, e. g. catalysis, colloid chemistry; their
relevant apparatus; C12P: Fermentation or enzyme-using processes to
synthesise a desired chemical compound or composition or to separate optical
isomers from a racemic mixture; H01L: Semiconductor devices; electric solid
state devices not otherwise provided for.
(*) When a same code appears for example in two patents, these data include both patents.
Source: Derwent Innovation Index
3.2 Relationship between technological diversification and patent production
To analyse how universities diversify their patented technology, we suggest an entropy index based on
Jacquemin and Berry (1979), which measures the degree of technological diversification across
technological fields:
,.
where Pij is the proportion of patents in field j in university i, and ln(1/Pij) is the weight given to each
field. This measure considers both the number of technological fields and the relative importance of these
fields compared with all the patents owned by the university. Zero indicates the absence of diversity and
ln (number of sectors) indicates the maximum extent of diversity.
In order to calculate the entropy index for each university we have followed two criteria. First,
according to Lerner (1994) and other subsequent research (e.g. Nerkar and Shane, 2007; Huang and
Murray, 2009; Czarnitzki et al., 2011), the technological diversification in university patenting is
measured using the spread of a university’s patent portfolio over technological subclasses of the
International Patent Classification at four-digit level. Second, to avoid problems over the sensitivity of the
results to changes in scale (that is, the number of technology fields used to calculate the entropy indexes),
entro =P
ij
i=1
n
∑ln 1
P
ij
"
#
$
$
%
&
'
'

12
we have considered another way of classifying patents, using an industrial classification instead the IPC
subclasses. For this purpose, we classified all the university-patented technology (by four-digit IPC
codes) into 44 industrial sectors, applying an industrial concordance table developed by Schmoch et al.
(2003). Using this equivalence table, we assigned the original IPC data (all the IPC at the four-digit level)
for each university-owned patent to the Classification of Economic Activities in the European
Community at the two-digit level based on the industrial sector where the patent originated. Once all the
technologies were classified in industrial sectors, we used this information to obtain the entropy indexes
for all universities in the sample for the period 1998–2000.
After applying these two criteria explained above we obtained introipc (entropy index obtained from
IPC codes) and entro (entropy index calculated from industrial sectors). Other alternatives consist of
weighting the entropy index introducing a weight factor (p/p–1) similar to that proposed by Hall et al.
(2001). This diversity index (entroipcw) accounts for the number of technology classes in which a
university was involved and controls for the size of the patent portfolio by weighting for (p/p-1), which is
correct for the small sample bias
5
. Finally, we use the log of the number of different IPC classes for all
the technologies patented by each university. This is a rough measure (useful for comparing with the
previous indexes) that takes into account only the technological scope of the patent (or number of
technologies embedded in the patent). Fig. 1 shows the relationship between the production of university
patents in the period 2001–2004 and the degree of technological diversification in the previous period
(1998–2000) using the indexes described above. These graphs provide some intuition, showing that the
data are consistent with the view that diversification is positively related to the production of patents. The
graphs also suggest that diversification is important but, above a certain point, there is clearly a positive
relationship with the production of patents in the subsequent period 2001–2004 when the values of the
entropy indexes are larger than 1.5.
5
For example, when the number of patents owned by the university is large, let us say 1000, the weight factor is near
1 (p/p-1 = 1000/999) and entroipc is equal to entroipcw; however, if the number of patents is small, e.g. 5, the weight
factor is 1.25 (p/p-1 = 5/4), which means that entroipcw is larger than entroipc for universities with a small number of
patents.

13
Fig. 1. Relationship between the number of university-owned patents in 2001-2004 and four measures of
technological diversification 1998-2000 (*)
Note: entroipc is the entropy index computed from the number of IPC class at four-digit level for each university; entro is the entropy
index obtained using the number of economic sectors; entroipcw is the entroipc index weighted by (n/n-1) where n is the number of all
technological classes in each university; lnipc is the logarithm of the number of different IPC clases in each university.
4 Variables, data and method
4.1 Variables and data
4.1.1 Dependent variable
The dependent variable, ownedpat, is the number of European patents owned by university ‘u’ located
in region ‘r’ in period ‘t’. As explained above, the data for this variable were obtained through a search
using the Derwent Citation Index for the period 2001-2004.
4.1.2 Variable of main interest
Our main independent variable is the degree of diversification of the patented technologies produced
in universities in a period before the patent production. We have used the four measures of diversification
described above that were obtained at university level for the period 1998-2000: entroipc, entro,
entroipcw and lnipc.

14
4.1.3 Control variables
• University size (lnpub). The empirical literature seems to support the view that large universities
are more likely to produce technology outputs. For example, Schartinger et al. (2002), in a study of 309
Austrian university departments, found that department size is a significant determinant of industry–
university knowledge and technology transfer. Likewise, Carlsson and Fridh (2002) investigated
technology transfer in the US using data from 170 universities, hospitals and research institutes. One of
their more important findings was the significant relation between institution size and the total number of
patents. Friedman and Silberman (2003) also found that university size was a relevant factor in explaining
the number of invention disclosures in 83 US universities. These references indicate the relevance of size
to the production of more patents or other kinds of technology outputs, but including size in the model is
also important for controlling by the fact that larger universities can count on more resources (such as
researchers or facilities) and consequently have more opportunities to produce patents. For capturing the
size of the university, we employed the number of scientific publications (lnpub). These data refer to the
scientific articles containing at least one author affiliated with a European university for 1998–2000. The
information was retrieved from the Science Citation Index Expanded (SCI) and includes papers from all
scientific fields and engineering, which are the more related to the production of technology (fields in
social sciences and humanities were excluded).
6
• Collaborations with firms or universities in producing patents (collabo). Collaboration provides
benefits that include access to a wide variety of resources, to new foundations or instruments needed to
solve complex problems, and to the creation of new knowledge or technologies (Belkhodja and Landry,
2007). To estimate the existence of spillovers, we need to assess the level of knowledge exchange or
knowledge co-operation with other universities or firms. To capture this variable, an approximate
indicator is the ratio between the number of university patents in collaboration with firms and other
universities and the total number of patents (collabo) for each university in the period 1998–2000.
• University R&D expenditure (rdexpend). This is one of the main variables suggested in the
knowledge production function, and on which the empirical literature has focused most intensely. For
example, Coupé (2003), Payne and Siow (2003), Azagra-Caro et al. (2003), Azagra-Caro et al. (2006),
6
As is well known, the SCI is a bibliographical database produced by the Information Sciences Institute (ISI), which
is in turn a part of Thomson Reuters’ Web of Science. The main advantage of ISI citation indexes is that they provide
a complete list of all authors and their affiliations. .

15
and Acosta et al. (2009) found a significant effect of university R&D expenditure on the production of
university patents, although with different values for the coefficient depending on the context in which
the knowledge production function was applied. This variable was measured using the logarithm of the
annual average for 1998–2000 of Higher Education Expenditure on R&D (HERD) in millions of PPS—
purchasing power standard, at 2000 prices. The source of HERD is Eurostat. We include regional instead
of university data for two reasons. First, there are no available indicators of university R&D funds for a
large sample of European universities, but we can take regional HERD, which is available at the regional
level. This is the only method of control for university R&D expenditure. Second, our variable of interest
is ‘diversification’; we do not aim to obtain accurate estimates of the effects of R&D on patent
production, although it is necessary to control for this factor. Methodologically, there should be no
problem because we take into account the clustered nature of data in the specification (universities are
grouped in regions and as regions for our analysis we chose the territorial units from Eurostat in each
country at the NUTS 2 level of aggregation).
• Technological environment or demand side for university-patented technology (rdgdp). Regions
with a strong industrial sector, supported by high-technology activities or with well-trained human
resources, will demand more intensive university technology than regions with a much lower level of
knowledge in technology and innovation. The inclusion in the model of a control variable capturing
regional factors is important because of the spillovers theory reversed (Casper, 2013). This hypothesis
suggests that the university’s regional environment can significantly affect the university’s success in
producing and commercialising science and technology. However, the empirical literature does not offer a
clear-cut conclusion about the effect of regional demand on university technology. For example, Baldini
et al. (2006) analysed a set of 637 patent applications filed at the Italian Patent and Trademark Office, the
European Patent Office, and the US Patent and Trademark Office between 1965 and 2002, with at least
one applicant belonging to the official list of higher education institutions. They found greater university
patenting activity in the north of Italy, where there is a higher level of industrial development. Using a
longitudinal dataset of all public Italian universities for 2005-2009, Rizzo and Ramaciotti (2014) found
that being embedded in an innovative region is a factor that exerts a positive influence on the university
rate of application for patents. Acosta et al. (2009) found no evidence of the effects of industrial potential
on the production of university patents in Europe. In our model this variable was captured by the ratio

16
between firm R&D expenditure and regional gross domestic product. We obtain an annual average for
1998–2000. The source used to construct rdgdp was Eurostat.
Note that our list of explanatory variables is comprised of data at two levels: our dependent variable
(the number of university-owned patents, the main variables of interest (diversification) and the control
variables capturing size and collaboration are obtained at university level. However, the control variables
on R&D expenditure and the demand side for patented technologies are obtained at regional level. There
are two reasons for this choice: first, according to the main empirical literature, R&D is a fundamental
variable in explaining the production of patent, but this variable is unavailable at university level.
Secondly, controlling for the university demand of the surrounding economic environment requires
considering a variable that nests our dependent variable (universities are located in regions). The
inclusion of these variables at regional level avoids a misspecification problem, and the combination of
several nested structures can be dealt with the appropriate econometric specification. Unfortunately data
about some the previous independent variables are not available for all the universities considered in the
descriptive analysis (391). For some of them was impossible to obtain information about their
characteristics and, on the other hand, information on several variables such as university R&D
expenditure exists only for a specific number of European countries. Consequently, the number of
observations (universities) to estimate the model was 141. Table 3 briefly summarises the variables and
presents the descriptive statistics, while Table 4 shows the correlations among all variables involved in
the analysis. Note that the linear correlations between the diversification indexes and the number of
patents range between 0.32 and 0.56.

17
Table 3 Variables. Summary and Descriptive Statistics
Variable
Description
Mean
Std. Dev.
Min.
Max
ownedpat
Dependent variable. Nº of university-owned patents (2001-2004).
8.764
10.508
0
49
Level 1: University data
entroipc
Entropy index obtained using technological IPC classes. (1998-
2000).
1.540
0.873
0
3.802
entro
Entropy index obtained using technological economic sectors.
(1998-2000).
0.872
0.710
0
2.723
entroipcw
Entropy index weighted by the university patent activity (1998-
2000).
1.763
0.858
0
3.862
lnipc
Log of the number of different IPC class for each university. (1998-
2000)
1.592
0.928
0
3.892
lnpub
Log of number of JCR scientific papers published by each university
(1998-2000).
6.730
1.761
0
8.856
collabo
Number of patents in 1998-2000 developed in collaboration with
firms divided by the number of patents 1998-2000.
0.171
0.314
0
1.000
Level 2: Regional data
lnrduniv
Log of R&D University Expenses (average 1998-2000)
5.534
0.867
3.392
7.550
rdgdp
R&D expenses divided by the regional gross domestic product
(average 1998-2000)
0.012
0.009
0.001
0.043
Table 4 Correlations
Variable
ownedpat
entroipc
entro
entroipc
lnipc
lnpub
collabo
lnrduniv
lnrdfirms
ownedpat
1.0000
entroipc
0.5192
1.0000
entro
0.3232
0.8055
1.0000
entroipcw
0.4660
0.9893
0.8035
1.0000
lnipc
0.5581
0.9959
0.7808
0.9783
1.0000
lnpub
0.4142
0.3294
0.1228
0.3
0.3425
1.0000
collabo
-0.0219
0.0262
-0.0017
0.0298
0.0227
0.0571
1.0000
lnrduniv
0.2351
0.0809
0.0127
0.076
0.096
0.1294
0.1587
1.0000
rdgdp
0.1297
0.1729
0.1175
0.1628
0.1814
0.0648
0.0942
0.4533
1.0000
4.2 Econometric model
This section establishes an econometric model to test the importance of diversification (along with
other variables) in the production of subsequent university-owned patents. The standard starting point in
the literature is a ‘knowledge production function’ (Griliches, 1979) in which knowledge is measured
with a proxy variable (e.g., patents) and the inputs include university R&D funds and other variables.
Several authors have used this empirical tool to analyse the production of university patents at various
levels (e.g. Coupé, 2003; Foltz et al. 2000, 2003; Payne and Siow, 2003; Azagra-Caro et al. 2006; Baldini
et al. 2006; Acosta et al. 2009; Gurmu et al. 2010).

18
Our framework is marked by two differences. First, we introduce the role of diversification as an
explanatory factor, and, second, we take into account non-observable regional influences that can affect
the production of university patents. The following paragraphs explain the empirical model, the variables
and provide some details about the estimation procedure.
Given the nature of the dependent variable, a suitable framework for dealing with this type of data
would be any of the family of the count models. The Poisson regression model is usually the starting
point for the analysis of count data. However, the assumed equality of the conditional mean and variance
functions is typically taken to be the major shortcoming of the Poisson regression model and thus
observed data rarely, if ever, display this feature (Greene, 2012, p. 807). The most common is the
negative binomial (NB) model, which is used to model overdispersion (the extra variability compared
with the mean).
7
Additionally, we extend the standard NB model to incorporate normally distributed
random effects at different hierarchical levels. More specifically, our base empirical specification is a
count model in which the conditional mean explaining the production of university-owned patents
depends on technological diversification along with a set of explanatory variables (Table 3) plus regional
unobserved heterogeneity:
[1]
Where diversif is captured by one of the four diversification indexes put forward in Section 3.2 (see
Table 3 for a summary) and is a random coefficient of unobserved heterogeneity over
regions. This term should capture other characteristics of the regional innovation system (not explicitly
included as explanatory factors in the model) that could affect the production of patents, such as
incentives and other policy measures and interface centres.
All independent variables refer to a previous period, t–µ, in which the patents are produced, because it
takes time for the factors included as independent variables to generate effects on the dependent variable.
Because universities are grouped into regions, the model has two hierarchical levels. The first level of
university effects considers the characteristics of universities affecting the production of patents. The
second level clusters all universities in the same region. This hierarchical structure suggests the extension
7
Other alternatives have been proposed in the literature, for example when the fraction of zeros is too high to be
compatible with standard underlying count data, zero inflated models or hurdle models can be a good option. But in
our analysis it is not appropriate to estimate this model as our sample contains 12.8% of zeros.
E(ownedpaturt | diversifurt−
µ
, lnpuburt−
µ
, collabourt−
µ
ln rdunivrt−
µ
,rdgdprt−
µ
,
ζ
r
(2 ) )=
exp(
α
+
γ
diversifurt−
µ
+
φ
1ln puburt−
µ
+
φ
2collabourt−
µ
+
β
1ln rdunivrt−
µ
+
β
2rdgdprt−
µ
+
ζ
r
(2 ) )
),0(| )2()2(
jz
NX
r»

19
of the simple count regression to a grouped-data model with unobserved effects capturing regional
heterogeneity. The adoption of a grouped-data model to analyse factors at two hierarchical nested levels
has several advantages in avoiding potential endogeneity problems in simple cross-sectional regression
models caused by the omission of variables and intragroup correlations. First, the grouped-data model
takes into account unobserved factors affecting the production of university patents on a regional scale.
Second, the geographical area where the patent is produced may have some influence. For example,
patents produced in the same region may be influenced by identical incentives or transfer office
characteristics (that is, we assume a high probability of intragroup correlation between universities in the
same region). Consequently, the decision to use hierarchical analysis has the main objective of avoiding
bias caused by both the omission of relevant variables and unobserved regional heterogeneity. Failure to
consider clustering of data could result in serious bias when we estimate the effects of the aggregate
explanatory variables on the individual-specific response variables (see, for example, Moulton, 1990;
Antweiler, 2001; Wooldridge, 2003). Furthermore, this procedure makes it possible to estimate
confidence intervals correctly for the estimated coefficient of diversification after controlling for regional
unobserved heterogeneity.
The log-likelihood of the multilevel count model (Poisson or NB) has no closed form, and so must be
approximated. All the multilevel Poisson models in the next section are estimated by adaptive Gaussian
quadrature, using the seven-quadrature points in the Stata statistical package, version 13. The technical
details of the procedure can be found in Raudenbush and Bryk (2002), Pinheiro and Chao (2006) and
Rabe-Hesketh and Skrondal (2012).
5 Results
5.1 Baseline results
Following our empirical specifications and the procedure proposed above, we estimated four negative
binomial regressions with random cluster effects. These models are obtained from a process whereby the
response is generated by counting the number of patents for each level-1 unit (university) and where,
conditional on the fitted explanatory variables and higher-level terms (regions), the mean count for each
university level-1 unit has a Poisson-gamma mixture distribution. This specification may capture some

20
additional overdispersion stemming from unexplained heterogeneity between universities. Table 5
presents the results. The dependent variable ownedpat (number of university-owned patents in 2001–
2004) is explained using four diversification indexes along with other variables that, according to the
standard literature on this topic, affect the production of patents (see Table 3 for a short description of the
variables and the descriptive statistics). Each model in Table 5 corresponds to one of the indexes
capturing the diversification of patented technology. Model I includes the entropy index computed using
IPC codes (entroipc); Model II uses the entropy index calculated from the number of industrial sectors;
Model III is presented with the entropy index weighted by the size of the university portfolio (entroipcw);
and Model IV considers the number of different IPC codes. The coefficient of overdispersion (lnalpha) is
significant in all models, pointing to the gain of the NB model with respect to the Poisson regression for
capturing the variability of the dependent variable. Analysing first the random part of the models, we find
the regional variance is not significant and ranges from 0.03 to 0.11. On the other hand, the LR statistic
that compares the two level random effect negative binomial models with standard NB regressions (last
row in Table 5) is only significant at 10% in one model. This test suggests that there are no unobservable
regional factors affecting the production of university-owned patents that had not been controlled for in
all models, and consequently there is little difference with respect to standard NB regressions.
Focusing now on our main variables of interest, the indexes of diversification, their coefficients
are positive and highly significant in all models. Consequently, the more diversified the number of
patented technologies in a university in a period, the greater the number of new patents produced by the
university in the subsequent period. In order to provide some direction about the magnitude of the effect
of diversification, it can be noted that the estimated coefficients can be interpreted as semielasticities, or
elasticities when the variable enters logarithmically in the exponential conditional mean (Cameron and
Trivedi, 1998, pp. 80-81). The estimates suggest that 10% increase in technological diversification would
lead to a rise in the number of patents ranging between 4 and 4.8%. For example, using the logarithm of
the number of IPC codes as a diversification index (Model IV), a university which owns 9 patents (the
average number of patents in the sample) over 8 different IPC codes at four-level digits (the average
number of different IPC codes in the sample), doubling the number of technologies in which the
university investigates in terms of number of IPC codes (or an increase of 100%) would lead to an
increase of approximately 4 more subsequent patents (an increase of 48%). This result suggests

21
diminishing returns, which means that the increase in the number of subsequent patents is less than the
enlargement of the level of diversification. This is not surprising since part of the research in new
knowledge may be fruitless in terms of future patents. The other coefficients for the indexes based on
entropy have a similar interpretation (Models I, II and III), although it should be taken into account that
the entropy indexes consider both the number of technological fields and the internal distribution of
patents in each field (the increase in entropy can stem from both expanding the number of technologies
and/or balancing the number of patents in each technology).
The other variables in the model present the following results. The variable collabo is not significant
in any model; that is, collaboration with firms to produce new patented technologies in a period is not a
significant variable when the objective is obtaining a greater number of patents in a subsequent period.
Regarding the size of the university, the coefficient of the number of publications (lnpub) is highly
significant. The elasticity of the R&D university expenditure is fairly stable around 0.25, showing
decreasing returns to scale as the literature on university patents suggests. We find no significant effect
from the demand side (captured by the ratio between firms R&D expenditure and GDP).
Table 5 Negative Binomial Models with Nested Random Effects
MODEL I
MODEL II
MODEL III
MODEL IV
Coeff.
Std. E.
Coeff.
Std. E.
Coeff.
Std. E.
Coeff.
Std. E.
C
-2.420
***
0.607
-2.478
***
0.685
-2.565
***
0.996
-2.357
**
0.597
Level 1: University.
entroipc
0.483
***
0.087
entro
0.403
***
0.112
entroipcw
0.449
***
0.091
lnipc
0.477
***
0.080
lnpub
0.333
***
0.055
0.401
***
0.056
0.350
***
0.056
0.322
***
0.055
collabo
-0.233
0.260
-0.197
0.278
-0.247
0.269
-0.218
0.254
Level 2: Region
lnrduniv
0.255
**
0.110
0.237
*
0.122
0.248
**
0.111
0.249
**
0.107
lnrdfirms
-8.696
10.098
-3.799
11.218
-7.527
10.345
-9.118
9.882
lnalpha
-0.604
***
0.224
-0.570
***
0.221
-0.548
***
0.221
-0.647
***
0.224
Lev 2 (region) Variance RE
0.032
0.078
0.112
0.098
0.033
0.084
0.031
0.073
Log-Likelihood
-409.537
-416.741
-412.091
-407.702
No. of Level 1 units
(universities)
141
141
141
141
No. of Level 2 units (regions)
74
74
74
74
LR test ML vs. nbinomial
regression:
0.19
1.65
*
0.18
0.21
***, **, * denote coefficients statistically different from zero at the 1%, 5% and 10% levels, respectively. Dependent variable (pat)
is the number of university-owned patents produced in 2001-2004. All independent variables refer to the period 1998-2000.

22
5.2 Robustness check
As pointed out above, the main coefficients do not change substantially when the technological
diversity is captured using different indexes, which means that the results are robust to four alternative
measures of diversification. In this Section we provide additional robustness checks in several ways. First,
by carrying out a sensitivity analysis to determine any possible change in the coefficients of either
diversification indexes as a consequence of omitting variables from the main models; secondly, by
estimating different count data specifications (Poisson models with robust standard errors); and thirdly,
by choosing Tobit models as an alternative specification to the count models.
With respect to the sensitivity analysis, we first assess potential multicollinearity computing the
variance inflation factor (VIF) in the estimated NB models. The mean VIFs range between 1.14 and 1.19
with a maximum of 1.3. These levels are considerably lower than the usual threshold of 10, and
consequently the results are not biased by multicollinearity. Additionally, we removed from the models
the variable capturing the size of the university (lnpub, number of publications), which is the variable
with the higher coefficient of correlation with the entropy indexes. Dropping this variable does not
change the main result: the diversification variables remain significant with only small variations in the
value of the coefficients.
The NB model is one way to model data in the presence of significant overdispersion, yet this
commonly chosen specification is not the only one. As pointed out by Cameron and Trivedi (2009, p.561)
we can simply use Poisson models with robust standard errors. We have followed this procedure to obtain
four additional Poisson models (one model for each diversification index along with all other variables
considered in the NB models) with robust standard errors and random effects to capture regional
unobserved heterogeneity. The results do not change the significance of the four indexes of
diversification. Although the values of the coefficients are slightly smaller than those obtained for our
baseline models, all the coefficients of the diversification variables are statistically significant at the 1%
level.
8
Regarding the alternative specification, Table 6 presents four Tobit models including exactly the same
variables as in the NB models and the same number of observations. These results show that all the
8
The estimated models for the sensibility analysis and the additional Poisson models are not presented, but are
available upon request.

23
coefficients of the diversification indexes remain highly significant, and the other significant coefficients
are the same as in the previous NB models. Note however that the Tobit model should only be viewed as
an approximation for these data. The dependent variable is a count, not a continuous measurement, and
thus a count model is always a preferable modelling framework (Greene, 2012, p. 815).
Table 6 Tobit models
MODEL V
MODEL VI
MODEL VII
MODEL VIII
Coeff.
Std. E.(1)
Coeff.
Std. E.(1)
Coeff.
Std.
E.(1)
Coeff.
Std.
E.(1)
C
-27.691
***
6.126
-28.676
***
6.871
-29.356
***
6.653
-26.794
***
5.820
Level 1: University.
entroipc
5.529
***
1.286
entro
4.555
***
1.203
entroipcw
4.997
***
1.249
lnipc
5.660
***
1.275
lnpub
1.967
***
0.410
2.651
***
0.485
2.166
***
0.425
1.834
***
0.398
collabo
-1.899
1.684
-1.891
1.591
-1.971
1.644
-1.789
1.706
lnrduniv
2.588
***
0.964
2.668
**
1.061
2.554
**
0.999
2.519
***
0.932
lnrdfirms
-21.642
111.610
16.097
115.381
-5.073
114.898
-31.654
109.889
Log-Likelihood
-456.55
-464.37
-459.82
-453.62
F
9.125
***
8.849
***
8.601
***
9.464
***
Pseudo R2
0.070
0.054
0.063
0.076
No. Obs.
141
141
141
141
***, **, * denote coefficients statistically different from zero at the 1%, 5% and 10% levels, respectively.
Dependent variable (ownedpat) is the number of university-owned patents in 2001-2004. All independent variables refer to the
period 1998-2000.
(1) Robust standard errors allowing for intra-group correlation (regions were used as a grouped variable for the universities).
6 Conclusions
In this paper we have analysed the relationship between technological diversification and the
production of new university patents. Our framework combines the literature on university patent
production with other theoretical and empirical literature about the consequences of technological
diversification. As noted in a previous section of the paper, several authors have shown the relevance of
some particular university characteristics to the encouragement of patents production. This study extends
these findings by including the role of technological diversification as an additional explanatory variable
in a modified patent production function that controls for some university characteristics and unobserved
regional factors. By analysing the effects of technological diversification to the framework on the factors
affecting university patent production, this paper contributes a new aspect to our knowledge that has not
been analysed so far at university level. Besides this, our research relates to several papers that have

24
found a significant impact of technological diversification on variables such as the quality of university
patents and their commercialisation.
Using a sample of university-owned patents distributed across 141 European universities, our
findings provide empirical evidence which suggests that greater diversification of the technologies
embedded in patents is relevant to spur the production of new patents. The estimations are robust as to the
use of a variety of technological fields to compute several indexes of diversification as well as to
alternative specifications. Other factors included in the models show evidence that variables usually
analysed in the university patenting literature, such as university R&D expenditure and the size of the
university (particularly the size of the research carried out in the university measured by the number of
published papers), are relevant in explaining the production of university patents.
Our empirical findings should nonetheless be interpreted with caution because according to the
broad classification of patents between university-owned patents and university-invented patents, we have
considered only the first group (patents in which the name of the university appears as applicant) across
Europe. Despite the fact that the use of university-owned patents is justified for the purpose of this paper,
we have considered only one side of the story. Although most European countries have been moving
away from inventor ownership of patent rights towards different systems of institutional ownership
(Geuna and Rossi, 2011), there are still many university-invented patents where at least one of the
inventors is a researcher affiliated to a university, but the applicant for the patent is a firm.
Bearing in mind the stated limitation, this study contributes by providing a new element for
discussing policy steps regarding how universities should diversify their technology in order to strengthen
the production of new university patents. Although a lengthy discussion of policy measures is beyond the
scope of this paper, we suggest some ideas aimed at technological diversification which could be
explored. For example, supporting projects with outcomes in multiple sectors, particularly with the
involvement of the private sector; and intensifying network collaboration within and between universities
and other public research institutes so that they share knowledge, resources and capabilities, thus
enhancing technological diversification and the subsequent generation of new patented knowledge.
Another course of action to be considered could consist of providing institutional encouragement for
researchers towards a wider range of fields or activities in order to promote technology transfer in fields
with greater industrial demand for university knowledge (e.g. introducing differences in the way in which

25
patents are assessed in the professors’ CVs; release from teaching time to develop complex patents with a
wide range of technological applications, etc.).
Finally, there are other ways in which the role of technological diversification at universities can
be explored. A natural extension of this research would be to analyse the effects of technological
diversification of European university-invented patents and to compare them with our results based on
university-owned patents. This contrast would provide a complete picture at European level on the role of
technological diversification on the production of new patents. Other suggestions consist of using other
objective variables such as the quality of the outputs or their commercialisation. The researcher and the
research group as one unit of analysis to study the production of technology could also give an interesting
perspective, given the recent debate about how patenting (or more generally, a greater involvement in
technological commercialisation) may divert or delay the research agenda at individual or research-group
level.
Acknowledgements
The authors highly appreciate the helpful comments of an anonymous Reviewer, which significantly
contributed to improving the clarity and quality of the paper. We are also grateful for the financial
assistance provided by Junta de Andalucía.
References
Acosta, M., Coronado, D., León, D. & Martínez, M.A. (2009). The production of university technological
knowledge in European regions: evidence from patent data. Regional Studies, 43(9), 1167–1181.
Acosta, M., Coronado, D. & Flores, E. (2011). University spillovers and new business location in high-
technology sectors: Spanish evidence. Small Business Economics, 36 (3), 365–376.
Anselin, L., Varga, A. & Acs, Z.J. (1997). Local geographic spillovers between university research and
high technology innovations. Journal of Urban Economics, 42 (3), 422–448.
Anselin, L., Varga, A. & Acs, Z.J. (2000). Geographical spillovers and university research: a spatial
econometric perspective. Growth and Change, 31(4), 501–515.
Antweiler, W. (2001). Nested random effects estimation in unbalanced panel data. Journal of
Econometrics, 101(2), 295–313.
Audretsch, D.B. & Lehmann, E.E. (2005). Does the knowledge spillover theory of entrepreneurship hold
for regions?, Research Policy, 34 (8), 1191–1202.
Azagra-Caro, J.M., Fernández-de-Lucio, I. & Gutiérrez-Gracia, A. (2003). University patents: output and
input indicators...of what?, Research Evaluation, 12(1), 5–16.
Azagra-Caro, J.M., Yegros-Yegros, A. & Archontakis, F. (2006). What do university patent routes
indicate at regional level?. Scientometrics, 66 (1), 219–230.
Azagra-Caro, J. M., Carayol, N., & Llerena, P. (2006). Patent production at a European Research
University: Exploratory evidence at the Laboratory Level. Journal of Technology Transfer, 31(3),
257–268.

26
Azagra-Caro, J. M. (2014). Determinants of national patent ownership by public research organisations
and universities. The Journal of Technology Transfer, 39(6), 898-914.
Azoulay, P., Ding, W., & Stuart, T. (2007). The determinants of faculty patenting behavior:
Demographics or opportunities?. Journal of Economic Behavior & Organization, 63(4), 599-623.
Baldini, N., Grimaldi, R. & Sobrero, M. (2006). Institutional changes and the commercialization of
academic knowledge: a study of Italian universities’ patenting activities between 1965 and 2002.
Research Policy, 35(1), 518–532.
Belkhodja, O., & Landry, R. (2007). The Triple-Helix collaboration: Why do researchers collaborate with
industry and the government? What are the factors that influence the perceived barriers?.
Scientometrics, 70(2), 301-332.
Bercovitz, J. & Feldmann, M. (2006). Entrepreneurial universities and technology transfer: a conceptual
framework for understanding knowledge-based economic development. Journal of Technology
Transfer, 31(1), 175–188.
Bonaccorsi, A., Colombo, M. G., Guerini, M., & Rossi-Lamastra, C. (2014). The impact of local and
external university knowledge on the creation of knowledge-intensive firms: Evidence from the
Italian case. Small Business Economics, 43(2), 261-28.
Breschi, S., Lissoni, F. & Malerba, M. (2003). Knowledge-relatedness in firm technological
diversification. Research Policy, 32(1), 69–87.
Breschi, S., Lissoni, F. & Montobbio, F. (2007). The scientific productivity of academic inventors: new
evidence from Italian data. Economics of Innovation and New Technology, 16(2), 101–118.
Brown, R. (1992). Managing the “S” curve of innovation. Journal of Consumer Marketing, 9(1), 61–72.
Cameron, A. & Trivedi, P., 1998. Regression Analysis of Count Data. Cambridge University Press,
Cambridge.
Cameron, A. C., & Trivedi, P. K. (2009). Microeconomics using stata. Lakeway Drive, TX: Stata Press
Books.
Carayol, N. (2007). Academic incentives, research organization and patenting at a large French
university. Economics of Innovation and New Technology, 16(2), 119–138.
Carlsson, B. & Fridh, A.C. (2002). Technology transfer in United States universities. Journal of
Evolutionary Economics, 12(1-2), 199–232.
Carlsson, B., Acs, Z.J. Audretsch, D.B. & Braunerhjelm, P. (2009). Knowledge creation,
entrepreneurship, and economic growth: a historical review. Industrial and Corporate Change,
18(6), 1193–1229.
Casper, S. (2013). The spill-over theory reversed: The impact of regional economies on the
commercialization of university science. Research Policy, 42(8), 1313-1324.
Chen, C. S., & Liu, C. H. (2012). Impact of network position and knowledge diversity on knowledge
creation: The empirical setting of research communities. Canadian Journal of Administrative
Sciences, 29(4), 297-309.
Chiesa, V. & Piccaluga, A. (2000). Exploitation and diffusion of public research: the case of academic
spin-off companies in Italy. R&D Management, 30(4), 329–340.
Chiu, Y.-C., Lai, H.-C., Liaw, Y.-C. & Lee, T.-Y. (2010). Technological scope: diversified or specialized.
Scientometrics, 82(1), 37–58.
Coupé, T. (2003). Science is golden: academic R&D and university patents. Journal of Technology
Transfer, 28(1), 31–46.
Crespi, G., D’Este, P., Fontana, R. & Geuna, A. (2011). The impact of academic patenting on university
research and its transfer. Research Policy, 40(1), 55–68.
Czarnitzki, D., Hussinger, K., & Schneider, C. (2011). Commercializing academic research: the quality of
faculty patenting. Industrial and Corporate Change, 20(5), 1403-1437.
Dai, Y., Popp, D., & Bretschneider, S. (2005). Institutions and intellectual property: The influence of
institutional forces on university patenting. Journal of Policy Analysis and Management, 24(3), 579-
598.
Di Gregorio, D. & Shane, S. (2003). Why do some universities generate more start-ups than others?.
Research Policy, 32(2), 209–227.
Djokovic, D. & V. Souitaris (2008). Spinouts from academic institutions: a literature review with
suggestions for further research. Journal of Technology Transfer, 33(3), 225–247.
Etzkowitz, H. (2003). Research groups as “quasi-firms”: the invention of the entrepreneurial university.
Research Policy, 32(1), 109–121.

27
Etzkowitz, H., Webster, A., Gebhardt, C., & Terra, B. R. C. (2000). The future of the university and the
university of the future: evolution of ivory tower to entrepreneurial paradigm. Research policy,
29(2), 313-330.
Etzkowitz, H. & Zhou, C. (2006). Triple Helix twins: innovation and sustainability. Science and Public
Policy, 33(1), 77–83.
Feldman, M.P. & Florida, R. (1994). The geographic sources of innovation: technological infrastructure
and product innovation in the United States. Annals of the Association of American Geographers,
84(2), 210–229.
Fischer, M. & Varga, A. (2003). Spatial knowledge spillovers and university research: evidence from
Austria. Annals of Regional Science, 37(2), 303–322.
Foltz, J.D., Barham, B. & Kim, K. (2000). Universities and agricultural biotechnology patent production.
Agribusiness, 16(1), 82–95.
Foltz, J.D., Kim, K. & Barham, B. (2003). A dynamic analysis of university agricultural biotechnology
patent production. American Journal of Agricultural Economics, 85(1), 187–197.
Franzoni, C. (2009). Do scientists get fundamental research ideas by solving practical problems?.
Industrial and Corporate Change, 18(4), 671-699.
Franzoni, C., & Scellato, G. (2011). Academic patenting and the consequences for scientific research.
Australian Economic Review, 44(1), 95-101.
Friedman, J. & Silberman, J. (2003). University technology transfer: do incentives, management, and
location matter?. Journal of Technology Transfer, 28(1), 17–30.
Gambardella, A. & Torrisi, S. (1998). Does technological convergence imply convergence in markets?
Evidence from the electronics industry. Research Policy, 27(5), 445–463.
García-Vega, M. (2006). Does technological diversification promote innovation?: an empirical analysis
for European firms. Research Policy, 35(2), 230–246.
Geuna, A. & Nesta, L.J.J. (2006). University patenting and its effects on academic research: the emerging
European evidence. Research Policy, 35(6), 790–807.
Geuna, A. & Rossi, F. (2011). Changes to university IPR regulations in Europe and the impact on
academic patenting. Research Policy, 40(8), 1068–1076.
Granstrand, O. (1998). Towards a theory of the technology-based firm. Research Policy, 27(5), 465–489.
Granstrand, O. & Oskarsson, C. (1994). Technology diversification in “MUL-TECH” corporations. IEEE
Transactions on Engineering Management, 41(4), 355–364.
Greene, W. H. (2012): Econometric Analysis (7th edition). Boston: Prentice Hall.
Griliches, Z. (1979). Issues in assessing the contribution of research and development to productivity
growth. Bell Journal of Economics, 10(1), 92–116.
Gurmu, S., Black, G. C., & Stephan, P. E. (2010). The knowledge production function for university
patenting. Economic Inquiry, 48(1), 192-213.
Hall, B. H., & Harhoff, D. (2012). Recent research on the economics of patents. Working Paper No.
w17773. National Bureau of Economic Research.
Hall, B.H., Jaffe, A.B. & Trajtenberg, M. (2001). The NBER Patent Citation Data File: Lessons, Insights
and Methodological Tools, NBER Working Paper 8498.
Henderson, R. & Cockburn, I. (1996). Scale, scope, and spillovers: the determinants of research
productivity in drug discovery. RAND Journal of Economics, 27(1), 32–59.
Henderson, R., Jaffe, A. B., & Trajtenberg, M. (1998). Universities as a source of commercial
technology: a detailed analysis of university patenting, 1965–1988. Review of Economics and
Statistics, 80(1), 119-127.
Huang, K. G., & Murray, F. E. (2009). Does patent strategy shape the long-run supply of public
knowledge? Evidence from human genetics. Academy of Management Journal, 52(6), 1193-1221.
Hunter, E.M., Perry, S.J. & Currall, S.C. (2011). Inside multi-disciplinary science and engineering
research centers: the impact of organizational climate on invention disclosures and patents. Research
Policy, 40(9), 1226–1239.
Jacquemin, A.P. & Berry, C.H. (1979). Entropy measure of diversification and corporate growth. The
Journal of Industrial Economics, 27(4), 359–369.
Jaffe, A.B. (1989). Real effects of academic research. American Economic Review, 79(5), 957–970.
Jaffe, A. B. (2000). The US patent system in transition: policy innovation and the innovation process.
Research policy, 29(4), 531-557.
Lerner, J. (1994). The importance of patent scope: an empirical analysis. The RAND Journal of
Economics, 319-333.
Lerner, J. (1995). Patenting in the Shadow of Competitors. Journal of law and economics, 38, 463-495.

28
Leten, B., Belderbos, R., & Van Looy, B. (2007). Technological diversification, coherence, and
performance of firms. Journal of Product Innovation Management, 24(6), 567-579.
Lettl, C., Rost, K., & Von Wartburg, I. (2009). Why are some independent inventors ‘heroes’ and others
‘hobbyists’? The moderating role of technological diversity and specialization. Research Policy,
38(2), 243-254.
Lin, B.-W., Chen, C.-J. & Wu, H.-L. (2006). Patent portfolio diversity, technology strategy, and firm
value. IEEE Transactions on Engineering Management, 53(1), 17–26.
Link, A. N., Siegel, D. S., & Bozeman, B. (2007). An empirical analysis of the propensity of academics to
engage in informal university technology transfer. Industrial and Corporate Change, 16(4), 641-
655.
Lissoni, F., Llerena, P., McKelvey, M. & Sanditov, B. (2008). Academic patenting in Europe: new
evidence from the KEINS database. Research Evaluation, 16(2), 87–102.
Lissoni, F. (2012). Academic patenting in Europe: An overview of recent research and new perspectives.
World Patent Information, 34(3), 197-205.
Lissoni, F., Pezzoni, M., Poti, B., & Romagnosi, S. (2013). University autonomy, the professor privilege
and academic patenting: Italy, 1996–2007. Industry and Innovation, 20(5), 399-421.
Merges, R. P., & Nelson, R. R. (1990). On the complex economics of patent scope. Columbia Law
Review, 90(4), 839-916.
Moorthy, S., & Polley, D. E. (2010). Technological knowledge breadth and depth: performance impacts.
Journal of Knowledge Management, 14(3), 359-377.
Moulton, B.R. (1990). An illustration of a pitfall in estimating the effects of aggregate variables on micro
units. The Review of Economics and Statistics, 72(2), 334–338.
Mowery, D. C., Nelson, R. R., Sampat, B. N., & Ziedonis, A. A. (2001). The growth of patenting and
licensing by US universities: an assessment of the effects of the Bayh–Dole act of 1980. Research
policy, 30(1), 99-119.
Mustar, P., Renault, M., Colombo, M. G., Piva, E., Fontes, M., Lockett, A., Wright M., Claryssed, B., &
Moray, N. (2006). Conceptualising the heterogeneity of research-based spin-offs: A multi-
dimensional taxonomy. Research Policy, 35(2), 289-308.
Nerkar, A., & Shane, S. (2007). Determinants of invention commercialization: An empirical examination
of academically sourced inventions. Strategic Management Journal, 28(11), 1155-1166.
O’Shea, R.P., Chugh, H. & Allen, T.J. (2008). Determinants and consequences of university spinoff
activity: a conceptual framework. Journal of Technology Transfer, 33(6), 653–667.
Payne, A. & Siow, A. (2003). Does federal research funding increase university research output?.
Advances in Economic Analysis & Policy, 3(1), Article 1.
Pinheiro, J.C. & Chao, E.C. (2006). Efficient Laplacian and adaptive Gaussian quadrature algorithms for
multilevel generalized linear mixed models. Journal of Computational and Graphical Statistics, 15,
58–81.
Quintana-García, C. & Benavides-Velasco, C.A. (2008). Innovative competence, exploration and
exploitation: the influence of technological diversification. Research Policy, 37(3), 492–507.
Rabe-Hesketh, S. & Skrondal, A. (2012). Multilevel and Longitudinal Modeling Using Stata (Third
Edition). College Station, TX: Stata Press.
Raudenbush, S.W. & Bryk, A.S. (2002). Hierarchical Linear Models: Applications and Data Analysis
Methods. Thousand Oaks, CA: Sage.
Rizzo, U., & Ramaciotti, L. (2014). The determinants of academic patenting by Italian universities.
Technology Analysis & Strategic Management, 26(4), 469-483.
Rosell, C., & Agrawal, A. (2009). Have university knowledge flows narrowed?: Evidence from patent
data. Research Policy, 38(1), 1-13.
Rothaermel, F.T., Agung, S.D. & Jiang, L. (2007). University entrepreneurship: a taxonomy of the
literature. Industrial and Corporate Change, 16(4), 691–791.
Saragossi, S., & de la Potterie, B. V. P. (2003). What patent data reveal about universities: the case of
Belgium. The Journal of Technology Transfer, 28(1), 47-51.
Schartinger, D., Rammer, C., Fischer, M.M. & Fröhlich, J. (2002). Knowledge interactions between
universities and industry in Austria: sectoral patterns and determinants. Research Policy, 31(3),
303–328.
Schmoch, U., Laville, F., Patel, P. & Frietsch, R. (2003). Linking Technology Areas to Industrial Sectors,
Final Report to the European Commission, DG Research.
Shane, S. (2001). Technology regimes and new firm formation. Management science, 47(9), 1173-1190.

29
Siegel, D. S., Waldman, D. A., Atwater, L. E., & Link, A. N. (2004). Toward a model of the effective
transfer of scientific knowledge from academicians to practitioners: qualitative evidence from the
commercialization of university technologies. Journal of Engineering and Technology Management,
21(1), 115-142.
Stephan, P.E., Gurmu, S., Sumell, A.J. & Black, G. (2007). Who’s patenting in the university? Evidence
from the survey of doctorate recipients. Economics of Innovation and New Technology, 16(2), 71–
99.
Sterckx, S. (2011). Patenting and licensing of university research: promoting innovation or undermining
academic values?. Science and Engineering Ethics, 17(1), 45–64.
Tantiyaswasdikul, K. (2012). The impact of the breadth of patent protection and the Japanese university
patents. International Journal of Innovation, Management and Technology, 3(6), 754-758.
Toh, P. K. (2014). Chicken, or the egg, or both? The interrelationship between a firm's inventor
specialization and scope of technologies. Strategic Management Journal, 35(5), 723-738.
Van Looy, B., Callaert, J. & Debackere, K. (2006). Publication and patent behavior of academic
researchers: conflicting, reinforcing or merely co-existing?. Research Policy, 35(4), 596–608.
Van Looy, B., Landoni, P., Callaert, J., van Pottelsberghe, B., Sapsalis, E. & Debackere, K. (2011).
Entrepreneurial effectiveness of European universities: an empirical assessment of antecedents and
trade-offs. Research Policy, 40(4), 553–564.
Varga, A. (1998). University Research and Regional Innovation: A Spatial Econometric Analysis of
Academic Technology Transfers. Boston, MA: Kluwer.
Wade, M.R. & Gravill, J.I. (2003). Diversification and performance of Japanese IT subsidiaries: a
resource-based view. Information & Management, 40(4), 305–316.
Woodward, D., Figueiredo, O. & Guimaraes, P. (2006). Beyond the Silicon Valley: university R&D and
high-technology location. Journal of Urban Economics, 60(1), 15–32.
Wooldridge, J.M. (2003). Cluster-sample methods in applied econometrics. American Economic Review,
93(2), 133–138.
Zahra, S.A., Van de Velde, E. & Larrañeta, B. (2007). Knowledge conversion capability and the
performance of corporate and university spin-offs. Industrial and Corporate Change, 16(4), 569–
608.
Zucker, L.G., Darby, M.R. & Brewer, M.B. (1998). Intellectual human capital and the birth of U.S.
biotechnology enterprises. American Economic Review, 88(1), 290–306.