The role of geographic proximity for university-industry linkages in brazil: an empirical analysis.
Garcia, Renato ; Araujo, Veneziano ; Mascarini, Suelene 等
1. INTRODUCTION
Innovation and technological change in firms depends upon the
creation and the diffusion of new knowledge. As pointed out by Nelson
(1959), knowledge can be seen as a non-rival asset and it plays a
fundamental role as an input to firms' innovative efforts. Hence,
one of the main sources for the creation and dissemination of new
knowledge, the university, has played a widely recognised role to foster
innovation in firms. In addition university-industry linkages have
become an increasingly important factor for firms' innovative
efforts.
Many authors, such as Jaffe (1989) and Audretsch and Feldman
(1996), have identified the existence of co-localisation between
universities and firms. This is because spatial concentration can
stimulate the maintenance of frequent contacts between academic
researchers and firms' research and development (R&D) staff.
The geographic proximity allows face-to-face contacts and the building
of specific channels of communication between firms and academic
research.
Based on this assumption, the primary aim of this paper is to
present an empirical investigation into the importance of localisation
in the development of university-industry linkages. It has been
suggested that proximity matters for the cooperation between academic
research and firms' R&D activities. To explore this, data from
the Census 2004 of the Brazilian National Council of Scientific and
Technological Development (CNPq) Directory of Research Groups were used.
This dataset collates information regarding the activities of research
groups in Brazil. From this database, it was possible to gather
information on 2 108 interactive research groups that interact with 3
068 firms. The location of research groups and firms allows analysis of
the spatial pattern of university-industry linkages. Through the
examination of these spatial patterns the role of geographic proximity
in fostering the exchange of information and knowledge sharing between
firms' R&D staff and academic researchers in the university can
be determined.
This paper is organised in four main sections including this
introductory section. The following section presents some brief
conceptual remarks on the role of geographic proximity for
university-industry linkages. Section 3 presents the methodology used,
such as the main characteristics of the database and the construction of
the control group for randomisation. The Final section contains the
results of the empirical analysis and some final remarks are presented.
2. BRIEF CONCEPTUAL REMARKS
The geographic concentration of producers and other agents can
benefit from the presence of strong externalities, especially pecuniary
knowledge externalities. The pecuniary knowledge externalities emerge
from unintentional contacts and interactions among local agents with
positive effects for the dissemination of new information, knowledge
sharing and technological learning (Antonelli, 2008). In many cases, the
diffusion of information and knowledge occurs within a complex social
network, in which personal ties and informal contacts among local
workers allow trust building and enable the diffusion of information and
knowledge that foster local innovation.
There is intense debate on the importance of local knowledge
spillovers and the main ways to measure them. Firms' R&D
activities, skilled labour and academic research are among the major
sources of local knowledge spillovers. Hence, academic research plays an
important role for the generation of new knowledge and for disseminating
this knowledge among local agents.
As stated by many authors, such as Klevorick et al. (1995),
universities are a very important source of knowledge for the innovative
efforts of firms, especially in industries in which new academic
research findings are closely linked to industrial innovation.
Nevertheless, in the case of developing countries such as Brazil, this
role of the university must be deeply investigated, since the industrial
structure of these economies has not shown the strong presence of firms
in high-tech industries. In contrast with the role of academic research
in developed economies, universities in developing countries could have
different characteristics and distinct patterns of interactions with
firms (Suzigan et al., 2009).
In addition to the importance of the university and the academic
research for the firms' innovative activities, many authors, such
as Jaffe (1989) and Audretsch and Feldman (1996) observed that
geographic proximity and spatial concentration of firms and universities
can be an important factor for knowledge sharing. In fact, these authors
were trying to measure the importance of local knowledge spillovers,
ever since the existence of spatially-mediated knowledge spillovers was
identified. They also stated that academic research is one of the main
ways in which local knowledge spillovers occurred.
In the same way, Varga (2000) considered the importance of spatial
proximity between universities and firms for innovation, especially in
high technology industries. He stated that geographic proximity of
academic research institutions and industry is an important source of
positive knowledge externalities. The author identified that personal
networks of academic and industry researchers, university spin-off firms
and graduate students are the main important channels for the diffusion
of the new knowledge from the university to local firms. Thus, his study
provides some empirical evidence of the role of agglomeration effects,
by using a modified version of the Griliches' knowledge production
function. His results show the positive effects of the spatial
concentration on the transfer of academic knowledge to firms.
Breschi and Lissoni (2001) also pointed out the importance of the
"knowledge externalities bounded in space", finding that firms
that are operating near important knowledge sources tend to be more
innovative than rival firms located elsewhere. They also emphasize the
importance of increasing empirical research on local knowledge
spillovers towards a better understanding of their nature and primary
characteristics, and argue that "the concept of local knowledge
spillover is no more than 'a black box', whose content remains
ambiguous" (Breschi and Lissoni, 2009; p. 976). The authors,
Breschi and Lissoni (2009) illustrate the role of these knowledge flows
by verifying the contribution of the mobility of inventors and the
network of researchers to the diffusion of knowledge across firms and
within regions. However, their results show that the effect of spatial
proximity on knowledge diffusion is not so strong, since the main
channel for knowledge diffusion is the co-inventors network, which is
not necessarily spatially concentrated. Hence, Breschi and Lissoni
(2009) pointed out that social and cognitive proximity among agents
could be as important as geographic concentration.
Other studies, such as D'Este and Iammarino (2010), indicate
that, in general, the smaller the spatial distance between universities
and firms, the greater the interactions among them. The main reason,
according to the authors, is the reduction in costs involved with the
exchange of information and knowledge sharing over smaller spatial
distances. In addition, D'Este and Iammarino (2010) show that
geographic proximity and research quality are positively associated to
university-industry linkages, even though there are strong differences
across knowledge areas. Similarly, Laursen et al. (2010) highlighted the
importance of academic research quality and found evidence that
geographic proximity tends to be particularly important when the
cooperation with universities presents very good academic performance.
Therefore, there is a general assumption that geographic proximity
can play an important role in fostering university-industry linkages,
since it allows the building of specific channels of communication and
local networks, which facilitate the dissemination of information and
knowledge sharing.
3. METHODOLOGY
Characteristics of the Database
In order to evaluate the role of geographic proximity in
university-industry linkages, data from the Census 2004 of the Brazilian
National Council of Scientific and Technological Development (CNPq)
Directory of Research Groups was used. A research group is defined by
CNPq as "a group of researchers, students and technical support
staff that is organised around the development of scientific research
lines following a hierarchical rule based on expertise and
technical-scientific competence" (CNPq Directory of Research
Groups, 2004).
This is the broader database of research activities in Brazil and
gathers information on the activities of Brazilian research groups, both
in universities and public research organisations (PRO). The database is
fed by the research group leader, who provides information regarding:
human resources, such as researchers, students and technical staff; main
research lines; knowledge specificities; academic production, measured
by scientific publications, patents, and artistic production; industries
linked with the research groups activities; and patterns of interaction
between the research group and firms.
These data are the main source of information about the patterns of
interaction between universities and firms in Brazil. For this reason,
the database allows the examination of the main features of the
relationships between academic research and the firm's R&D
activities. However, there are two methodological limitations that
should be pointed out: filling out the form that feeds the database is
voluntary and the data are collected by self-declaration. Therefore,
there is a high possibility that the interactions between research
groups and firms are underestimated in the database, since not all the
research group leaders complete the form with the correct information.
Nevertheless, the number of research groups that are filling out
the form for the database is growing. In addition, there are an
increasing number of studies that are using the Directory of Research
Groups to examine the Brazilian university-industry linkages, such as
Rapini et al. (2009), Suzigan et al. (2009) and Fernandes et al. (2010).
In the 2004 Census, there were 19 470 research groups, encompassing 77
649 researchers from 375 different institutions in all the Brazilian
regions. For the purposes of this paper, only the research groups that
declared linkages with firms were selected. Table 1 shows the lists of
variables selected for analysis.
Data on the location of the research groups are available in the
Directory, and data on the location of the firms were collected from the
Brazilian Treasury Department. Therefore it was possible to use
information on the location of both the firm and the research group.
Concerning the localisation of interactions between research group and
firm, the location was included in three different aggregation levels:
states, meso-region and micro-region. These levels of regional
aggregation were created by the Brazilian Office of Statistics (IBGE),
which defines micro-regions as a cluster of neighbour cities, and
meso-regions as a wider geographic space that normally involves three or
four micro-regions. In Brazil, there are 558 micro-regions and 137
meso-regions in its 27 states. In general terms, micro-regions can be
associated with the EU NUTS3 regions, and meso-regions with EU NUTS2.
With regard to the interaction with firms, the research group
leader may register up to three different types of interactions from a
list of 14 options. Since the leader can select more than one type of
linkage with the same firm, interactions may be counted more than once.
Therefore, for the empirical analysis, the repetitions that included
different types of relations were removed, to ensure that each pair
"research group-firm" was counted only once. Each register of
the database contains one research group and one interactive firm.
Data Description: Pattern of University-Industry Linkages in Brazil
In the 2004 database, 2 108 research groups from 217 different
institutions indicated some type of linkages with firms. This shows that
of all the research groups in the database (19 470), 10.8 percent
declared to have some type of interaction. Taking the universities and
public research organisations, 57.8 percent of all the institutions had
research groups that presented some kind of linkages with firms (217 out
of the 375 in the database). The declared interactions between research
groups and firms totaled 8 817. After removing the duplications, 4 476
interactions were listed with information about the location of the pair
"research group-firm".
As for the distribution of the interactions with Brazilian
universities and public research organisations, the top 10 institutions
with the largest share of interactions accounted for 41 percent of the
total linkages. The share of interactions by public research
organisations was 12.2 percent. Observing the distribution of
interaction among the 2 108 research groups, it was not possible to
identify any substantial concentration. The single exception concerned
one research group which declared maintaining linkages with 199
different firms, or 4.1 percent of the total.
Suzigan et al. (2009) have presented a comprehensive map of the
university-industry linkages in Brazil. The main assumption of the
authors is that the Brazilian national system of innovation is
incomplete and immature, and this characteristic matters to the
development of university-industry linkages. According to the authors,
interactions between universities and firms in Brazil are characterised
both by the transmission of typical codified knowledge, for example by
rendering of services and training, and by the creation of bidirectional
flows of information and knowledge, through collaborative R&D
projects that involve researchers from the R&D facilities at the
firm and researchers from the university.
In terms of firms that interact with research groups, it is not
possible to identify a greater concentration. The three firms most often
mentioned by the research groups were: Petrobras, the stated-owned oil
Brazilian company and one of the Brazilian firms that presents huge
innovative efforts; Embrapa, an agricultural research institute that is
quite important in Brazil and interacts both with universities and with
firms; and Cemig, a power utility company, the interactions of which are
stimulated by public policy. Among the top 12 firms that had linkages
with a university in Brazil, the main industries were energy, including
oil and gas; pulp and paper; and mining. It is important to mention that
there are some public policy measures in Brazil that stimulate the
interaction of firms with universities. Some of these measures apply for
all industries, while others are specific to certain industries. This is
the case for the energy industry in Brazil, since public policy forces
firms to spend a share of their revenue in R&D. This often
culminates in the establishment of joint projects with universities or
public research organisations.
Other information available in the database concerns the
distribution of interactions over the different knowledge areas. Many
authors, such as Metcalfe (2003), pointed out that some knowledge areas,
such as engineering, pharmacology, agronomy, computing and medicine,
because of the nature of their scientific activities, tend to be closer
to the problems of society and of firms. For this reason they often
bridge the gap between academic activities and applied research within
firms. Schartinger et al. (2002) also points out that different
knowledge areas present distinct patterns of interactions, which is
evident in the different types of mechanisms involved in linkages
between firms and universities.
In the case of university-industry linkages in Brazil, the
importance of Engineering and Agricultural Sciences is verified, as
these are the areas with the largest number of interactions with firms
(Table 2).
These data show the important role played by these two knowledge
areas in fostering firms' innovation in Brazil, which can be seen
by the huge number of interactions with firms. Some of these
interactions are characterised by more routinised activities and
codified knowledge, such as laboratorial tests and essays or the supply
of specialised inputs. However, there are a large number of interactions
that are based on joint research projects between university researchers
and firms' R&D staff, which involve a higher level of tacit
knowledge sharing.
Thus, to evaluate the intensity of knowledge flows between firms
and universities, the main types of relationship between the academic
research groups and firms must be examined. When filling out the form,
the research group leader is required to specify the type of interaction
made with each firm. The leader can thus select up to three types of
interactions out of the 14 types presented.
Table 3 shows all the different types of interaction that the
research group leaders pointed out in the 2004 Census. Some of these
interactions involve unilateral knowledge flows, moving from the
research group toward the firm or vice versa, as is the case of
technology transfer and product development.
On the other hand, there are other types of interaction such as
joint research projects, which have a typically bilateral nature, in
which there are stronger tacit, specific and complex knowledge sharing.
The most common type of interaction is the "Short-term R&D
cooperative projects," which corresponds to 2 422 interactions or
27.5 percent. The second most common is "Technology transfer to the
firm", with 1 472 interactions, and the third is "Long-term
R&D cooperative projects," 1 206 interactions. These results
show that a considerable share of interactions take place via joint
research projects, which require the transfer of complex information and
knowledge sharing in both directions..
4. RESULTS ON THE LOCAL DIMENSION
The use of data from the Directory of Research Groups allowed
examination of the importance that geographic proximity has for the
university-industry linkages in Brazil. An empirical test was built in
order to study the differences in distance and the propensity to
interact in different knowledge areas.
Regional Distribution of Firms and University
A preliminary analysis of university-industry linkages in Brazil
shows that there is a strong regional concentration of research groups
and firms in the Brazilian Southern regions. Taking interactions as the
unit of analysis, it was clear that out of all the research groups that
declared maintaining linkages with firms, more than half of the
interactions occurred with research groups from three Southern Brazilian
states: Sao Paulo, Minas Gerais and Rio Grande do Sul. If other states
are added in the same region (Rio de Janeiro, Parana and Santa
Catarina), it turns out that they account for 78 percent of the
interactions (Table 4).
This result is strongly linked with the regional distribution of
income in Brazil, since these states account for a high share of the
Brazilian gross domestic product (GDP) (70.3%). Other indicators of
innovation, such as deposit of patents, innovative firms or scientific
publications, are even more concentrated in these states. Conversely,
the share of the total interaction of the other states is fairly low.
The other 21 states together, only account for 21.9 percent of the total
interactions of research groups. This confirms, for the Brazilian
experience, the statement from Audretsch and Feldman (1996) that
innovation is more concentrated in space than economic activity.
This spatial concentration is also seen in micro-regions. The top
20 micro-regions encompass 75.1 percent of interactions (i.e. 3 361
interactions). Figure 1 clearly illustrates this concentration,
highlighting the high regional concentration of the university-industry
linkages in Brazil, especially in the Southern regions.
[FIGURE 1 OMITTED]
From the standpoint of the firms that interact with research
groups, the circumstances are similar to those shown in Table 4.
Likewise, most of the interactive research groups could be found in the
same six states, with small changes in the positions. This means that
the Southern concentration remains unchanged. In the micro-regions
level, the concentration is only narrowly lower than for research
groups; the share of the top 20 most interactive micro-regions is 63.8
percent (2 857) of all the interactions.
The Role of Geographic Proximity for University-Industry Linkages
In order to analyse the role of geographic proximity in fostering
university-industry linkages, data from the Directory of Research Groups
were ordered by the location of the interactive research groups and the
firms.
Upon examination of the interactions between research groups and
firms, it is evident that a huge share of the interactions (71.3%) take
place within the same state. However, as stated by authors such as
Breschi and Lissoni (2001), states are not the correct unit of analysis
to examine the role of geographic proximity and its benefits. This is
because the state unit is too large to foster the dissemination of
information and knowledge sharing. Thus, it is not possible to assume
that academic researchers and firms' R&D staff in a given state
are more likely to have face-to-face contacts and to build local
networks. Therefore, at the state level, diffusion of information and
knowledge sharing do not occur in relation to geographic proximity. In
the same way, Beaudry and Schiffauerova (2009) reinforce this problem,
since the results of several empirical tests applied to larger areas
turned out to be far less significant than those of lower geographic
areas.
In order to overcome this problem, three different aggregation
levels were used (states, meso-regions and micro-regions) to analyse the
role of geographic proximity in university-industry linkages of less
aggregated geographic levels. In addition, in the smaller spatial unit,
the micro-regions, an analysis based on geographic distance was added.
An important measure of the role of geographic proximity is the
number of interactions within the same region, i.e., the linkages
between firms and research groups of the same region. As indicated by
authors such as Zucker and Darby (1996) and Audretsch and Feldman
(2004), the spatial proximity of the main sources of knowledge
facilitate the access to differentiated knowledge flows and reduce the
time involved in firms' learning processes. The examination of the
co-located linkages shows that 49.1 percent of total interactions are of
research groups and firms located in the same meso-region and 44.1
percent in the same micro-region (Table 5).
This result shows the importance of geographic proximity for the
interaction between academic research and the firms' innovative
efforts. The main reason for the co-location of university and industry
is that the knowledge flows generated by academic activities remain
local and local firms can benefit from the geographic proximity.
Furthermore, the co-location facilitates the establishment of joint
projects between academic research and firms' R&D. Hence, the
high number of interactions that occur in the same micro-region show the
role of geographic concentration of both universities and firms, in
which the local pool of capabilities, the local networks of
professionals and face-to-face contacts are important factors to
stimulate the diffusion of information and knowledge sharing between
universities and firms.
It is worth observing that the results above, which show that
university-industry linkages in Brazil are highly co-located, may have
been heavily influenced by the spatial distribution of economic
activity, since in Brazil the main economic and innovative indicators
are strongly concentrated in Southern states. Hence, the geographic
concentration of the university-industry linkages could be a result of a
previous geographic concentration of economic activity. Many authors,
such as Jaffe et al. (1993), stated that preexisting location factors
can influence the measurement of local knowledge spillovers, and the
same statement can be applied to university-industry linkages.
Therefore, it is necessary to control pre-existing location factors, in
order to verify if, despite the geographic concentration of economic
activity, university-industry linkages remain geographically
concentrated. Without this kind of control, the co-localisation between
research groups and firms could be related solely to the fact that
research groups are close to firms, but not to the presence of
externalities associated with the geographic proximity.
For this reason, the analysis of the localisation patterns of
university-industry linkages requires the use of some methodological
tools that can separate the importance of geographic proximity and the
pre-existing concentration of agents. To do this, in the same way as
Jaffe et al. (1993), a random control sample was built, in which
previous location factors could be removed. The comparison between the
database and the control group allows the analysis of the importance of
geographic proximity for university-industry linkages, in spite of other
location factors.
The construction of the control group is quite simple. It is
presumed that the choice made by the firm to interact with the
university is related to its specific needs in a certain knowledge area.
Thus, the firm will search a research group that masters this knowledge
and is able to help with the solution to its problem. A total of 77
sub-areas of knowledge were identified based on the Brazilian
classification. Assuming the decision to interact is made by the firm,
the control group was built by taking each firm of the original database
and associating it with a new research group randomly selected among the
groups of the same sub-area of knowledge. Thus, any chance that
geographic proximity could influence the firm's decision to
interact with the university was removed.
Through the randomisation, in the control group, the geographic
proximity has no influence on the decision of the firm to interact, and
firms can choose any research group in any part of the country, despite
the existence of benefits related to the closeness to the university.
Therefore, each interaction between a research group and a firm would
have the same probability of occurring with any research group in that
knowledge area, taking into account the unequal distribution of research
groups in that field. If after randomisation the proportion of
co-located interactions differs with statistical significance, it is
possible to conclude that there is a non-random process underlying the
location of university-industry linkages, and geographic proximity may
play an important role for the interactions between research groups and
firms.
One potential problem regarding the randomisation procedure is that
some of the 77 knowledge sub-areas only had a small number of research
groups. For example, in the database, 34 sub-areas presented less than
25 of the interactions in Brazil. In those cases, the random set of
research groups is small and the probability that the original pair
firm-research group is selected to the control group increases. This
problem can introduce a bias toward co-localisation. However, this
problem is not so important since interactions in those 34 sub-areas
were responsible for a only 6 percent of the database. It would be
possible to also use the type of interaction in the randomisation, since
there could be a relationship between the distance and the type of
interaction. However, this would impose a new restriction for building a
control, since firms could interact with research groups in more than a
one way. For example, a research group could engage in a joint research
project with one firm and provide training to another one; or another
group that has an agreement to transfer technology to a company acts via
a consultancy contract with other firms.
In a preliminary examination of the results, the importance of the
co-localisation could be seen by the share of interactions that took
place in the same region. Table 6 presents the interactions between
firms and research groups of the same state, the same meso-regions and
the same micro-regions, both in the database and in the control group by
knowledge areas.
A simple comparison between them shows that local interactions were
more than three times higher in the database than in the control groups
in every spatial aggregation level and for every knowledge area. This
clearly shows that university-industry linkages are quite localised,
which suggests the important role played by geographic proximity. This
result is similar to what Jaffe et al. (1993) found in the analysis of
patent citations in the U.S., using the same method of randomisation. At
the micro-region level, where the benefits of the geographic proximity
are more powerful, the co-localised interactions are up to five times
higher, which strengthens the conclusion that geographic proximity
matters.
Furthermore, by looking at the distinct knowledge areas, it is
possible to see different co-localisation patterns. In the case of
Agricultural Sciences, interactions are less co-localised than the
average, even though localisation remains quite important since a huge
share of interactions occurs in the same region, and almost 30 percent
of all interactions take place in the same micro-region, ten times more
than in the control group. On the other hand, Human and Social Sciences
tend to be more spatially concentrated, since 64.7 percent of all
interactions take place in the same micro-region.
Moreover, it is possible to analyse the difference regarding
distance, in kilometers (km), between the location of research groups
and firms. This was done both for the whole database and for the control
group for each knowledge area. For both groups (database and control
group), the distance from the centroid of the research group's to
the firm's micro-regions was added. In the case of interactions
within the same micro-region, a null distance was assumed. Table 7 shows
the main results.
The average geographic distance between all the Brazilian
micro-regions is very high, around 1 500 km. After creating the control
group, in which the geographic distribution of all interactive research
groups is considered, it was possible to observe a substantial decrease
in distance to 936 km. In addition, in the examination of each knowledge
area, the control group distance varies from 845 km in Human Sciences to
1 111 in Natural and Earth Sciences. In the same way as the share of
co-localised interactions presented above, the distance of the
interactions between research groups and firms of the control group is
more than three times as high in comparison with the whole database. The
difference between these averages is more than 600 km, which confirms
that geographic proximity matters for the university-industry linkages,
even when pre-existing location factors are controlled for. The average
interaction distance measured in the whole database was 294 km while in
the control group the average was 936 km. A Mann-Whitney U test
(Wilcoxon rank-sum) revealed that these values were distinct at a 0.1
percent statistically significant level--it is important to note that it
was not possible to apply a T-test because both Shapiro-Wilk and
Skewness-Kurtosis tests rejected the normality both of the database and
the control group. This led to the conclusion that geographic proximity
plays a very important role in the university-industry linkages, as
shown by the descriptive analysis.
Another important factor that may influence the distance of
university-industry linkages is the knowledge area. Many authors, such
as Metcalfe (2003), point out the different roles played by knowledge
areas in the firms' innovative efforts, which implies different
patterns of interactions with firms. In addition, Schartinger et al.
(2002) points out that the firm's industry affects the
collaboration with university. Since industries are differently
distributed in space, they directly impact the average geographic
distance of university-industry linkages. For example, industries more
concentrated in large urban centres are normally closer to universities.
Table 7 presents the main patterns of interaction, in particular
the last column indicates the difference for each knowledge area
compared to the average of the whole database. There are significant
differences among groups. Interactions between firms and research groups
in Agricultural Sciences are 28 km farther than the average. On the
other hand, interactions in Human and Social Sciences are almost 100 km
closer than the average. It can be verified, therefore, that in some
knowledge areas, such as Human and Social Sciences, interactions are
more co-localised than others, such as Agricultural Sciences.
These differences could be associated with the spatial distribution
of economic activities. The pattern of interaction in Agricultural
Sciences, which is more dispersed in space, could reflect the dispersion
of agricultural activities. Conversely, in Humanities, linkages to
university are more local, probably linked to urban areas.
5. FINAL REMARKS
Universities are frequently presented in the literature as an
important source of information and knowledge to firms' innovative
efforts. Hence, geographic proximity between universities and firms can
foster university-industry linkages, since face-to-face contacts and
local knowledge networks are important ways of interaction.
Based on this assumption, this paper tries to shed further light on
the role of geographic proximity for university-industry linkages. Data
from the Brazilian Directory of Research Groups were used, as it has a
comprehensive database on the research groups' activities in
Brazil. The analysis shows a strong regional concentration of
interactions in the Southern states of the country, and the top 20
micro-regions encompass slightly higher than three quarters of all the
interactions. In terms of geographic proximity, a large share of
co-localised interactions can be observed, since 44 percent of all the
interactions occur between universities and firms in the same
micro-region.
However, the geographic proximity, measured by the co-localized
interactions, could be a result of the pre-existence of location
patterns, since there is a huge geographic concentration of
university-industry linkages in Brazil. In order to control these
location pre-existent patterns, an empirical test was used with a
control group, in the same way as Jaffe et al. (1993). The results show
that, after controlling the pre-existing geographic concentrations of
interactive academic research groups, geographic proximity remains an
important factor for the university-industry linkages. In addition,
geographic proximity plays a different role in each knowledge area in
Brazil. For some knowledge areas, such as Human and Social Sciences,
interactions occur more frequently in closer geographic areas; in
others, such as Agricultural Sciences, linkages to university are more
distant.
The main contribution of this paper is that there isn't a
random geographic distribution of the interactions between research
groups and firms, which allows us to conclude that university-industry
linkages are strongly localized. This conclusion, based on the Brazilian
experience, is in agreement with previous work, such as Jaffe (1989) and
Audretsch and Feldman (1996), which identified the existence of
co-localisation between university and firms. In addition, results show
that there is a remarkable difference among the distinct knowledge
areas.
The findings from this study also have important policy
implications. The results emphasise the importance of universities in
fostering firms' innovative efforts at the local level. In this
way, policy measures should be designed in order to strengthen academic
research at regional level, since local firms can benefit themselves
from the local universities' activities. In this way, local
research groups play a very important role not only by the creation of
local capabilities, but also in the support of local firms'
innovative efforts.
Even though the results show the importance of geographic
proximity, there are some open questions that require further
investigation. One being; what is the impact of the quality of the
university on the role of geographic proximity? This question requires
further investigation because the university's performance can
affect the pattern of interactions between firms and universities
(D'Este and Iammarino 2010; Laursen et al., 2010). Finally the type
of collaboration between university and firms also requires further
investigation (Schartinger et al., 2002).
ACKNOWLEDGEMENTS: The authors would like to thank Fapesp and CNPq
for their financial support, and Emerson Gomes dos Santos e Ariana
Costa, who provided useful comments. Usual disclaimers apply.
REFERENCES
Antonelli, C. (2008). Pecuniary Knowledge Externalities: the
Convergence of Directed Technological Change and the Emergence of
Innovation Systems. Industrial and Corporate Change, 17(5), pp.
1049-1070.
Audretsch, D. and Feldman, M. (1996). R&D Spillovers and the
Geography of Innovation and Production. American Economic Review, 86(4),
pp. 253-273.
Audretsch, D. and Feldman, M. (2004). Knowledge Spillovers and the
Geography of Innovation. In J.V. Henderson and J. Thisse (Eds) Handbook
of Urban and Regional Economics, Volume 4. North Holland Publishing.
Beaudry, C. and Schiffauerova, A. (2009). Who's Right,
Marshall or Jacobs? The Localization Versus Urbanization Debate.
Research Policy, 38, pp. 318-337.
Breschi, S. and Lissoni, F. (2001). Knowledge Spillovers and Local
Innovation Systems: a Critical Survey. Industrial and Corporate Change,
10, pp. 975-1005.
Breschi, S. and Lissoni, F. (2009). Mobility of Skilled Workers and
Co Invention Networks: an Anatomy of Localized Knowledge Flows. Journal
of Economic Geography, 9, pp. 439-468.
CNPq Directory of Research Groups (2004). Census 2004 of the
Brazilian National Council of Scientific Research. Online version
accessed 2 May 2010, pp.5, http://dgp.cnpq.br/diretorioc/
D'Este, P. and Iammarino, S. (2010). The Spatial Profile of
University Business Research Partnerships. Papers in Regional Science,
89(2), pp. 336-350.
Fernandes, A., Campello de Souza, B., Stamford Silva, A., Suzigan,
W., Chaves, C. and Albuquerque, E. (2010). Academy-Industry Links in
Brazil: Evidence about Channels and Benefits for Firms and Researchers.
Science and Public Policy, 37(7), pp. 485-498.
Jaffe, A.B. (1989). Real Effects of Academic Research. American
Economic Review, 79(5), pp. 957-970.
Jaffe, A.B., Trajtenberg, M. and Henderson, R. (1993). Geographic
Localization of Knowledge Spillovers as Evidenced by Patent Citations.
Quarterly Journal of Economics, 63, pp. 577-598.
Klevorick, A., Levin, R., Nelson, R. and Winter, S. (1995). On the
Sources and Significance of Inter-Industry Differences in Technological
Opportunities. Research Policy, 24, pp. 185-205.
Laursen, K., Reichstein, T. and Salters, A. (2010). Exploring the
Effect of Geographical Proximity and University Quality on
University-Industry Collaboration in the United Kingdom. Regional
Studies, pp. 1-17.
Metcalfe, S. (2003). Equilibrium and Evolutionary Foundations of
Competition and Technology Policy: New Perspectives on the Division of
Labour and the Innovation Process. Revista Brasileira de Inovagao, 2(1).
Nelson, R. (1959). The Simple Economics of Basic Scientific
Research. The Journal of Political Economy, 67, pp. 297-306.
Rapini, M., Albuquerque, E., Chaves, C., Silva, L., Souza, S.,
Righi, H. and Cruz, W. (2009). University-Industry Interactions in an
Immature System of Innovation: Evidence from Minas Gerais, Brazil.
Science and Public Policy, 36, pp. 373-386.
Schartinger, D., Rammera, C., Fischer, M.M. and Frohlich J. (2002).
Knowledge Interactions between Universities and Industry in Austria:
Sectoral Patterns and Determinants. Research Policy, 31, pp. 303-328.
Suzigan, W., Albuquerque, E., Garcia, R. and Rapini, M (2009).
University and Industry Linkages in Brazil: Some Preliminary and
Descriptive Results. Seoul Journal of Economics. 22(4), pp. 591-611.
Varga, A. (2000). Local Academic Knowledge Transfers and the
Concentration of Economic Activity. Journal of Regional Science, 40(2),
pp. 289-309.
Zucker, L. G. and Darby, M. R. (1996). Star Scientists and
Institutional Transformation: Patterns of Invention and Innovation in
the Formation of the Biotechnology Industry. Proceedings of the National
Academy of Science 93 (November): 12709-12716.
Renato Garcia
Polytechnic school, University of Sao Paulo, Brazil, Av. Almeida
Prado, Trav. 2, 128
05508-070. E-mail: renato.garcia@poli.usp.br
Veneziano Araujo
Polytechnic school, University of Sao Paulo, Brazil, Av. Almeida
Prado, Trav. 2, 128 05508-070.
Suelene Mascarini
Polytechnic school, University of Sao Paulo, Brazil.
Table 1. Main Information Collected.
Research group level Name
Leader
State
University/PRO
Main knowledge area
Specific knowledge area
Firm level Name
Fiscal Code (CNPJ)
State of the unit that interacts
Type of interaction
ISIC (CNAE)
Source: CNPq Directory of Research Groups, (2004).
Table 2. Number of Interactions in each Knowledge Field.
Knowledge areas Interactions %
Engineering 1 869 41.8
Agrarian Science 946 21.1
Healthy Sciences and Biology 699 15.6
Natural and Earth Sciences 568 12.7
Human and Social Sciences 394 8.8
Total 4 476 100.0
Source: CNPq Directory of Research Groups, (2004).
Table 3. Types of Interaction Between University
and Industry.
Type of Total % Knowledge flow
interaction direction
Firm Research
group
Short-term R&D 2 422 27.5 x x
collaborative
projects
Technology transfer 1 472 16.7 x
Long-term R&D 1 206 13.7 x x
collaborative
projects
Consultancy 680 7.7 x
Training 510 5.8 x
Material supply 385 4.4 x
Non-rotinized 324 3.7 x
engineering
Software development 254 2.9 x
Technology transfer 220 2.5 x
Training 181 2.1 x
Software development 102 1.2 x
Non-rotinized 97 1.1 x
engineering
Material supply 44 0.5 x
Other 513 5.8
NA 407 4.6
Total 8 817 100.0
Source: CNPq Directory of Research Groups (2004).
Table 4. Geographic Distribution of the University-Industry
Linkages in Brazilian States.
Location of the
research groups
Ranking State n. %
1 Sao Paulo 1 227 27.4
2 Minas Gerais 607 13.5
3 Rio Grande 549 12.3
do Sul
4 Rio de Janeiro 424 9.5
5 Parana 385 8.6
6 Santa Catarina 303 6.8
7-27 Other 756 21.9
Total 4 476 100.0
Location of the firms
Ranking State n. %
1 Sao Paulo 1 307 29.2
2 Rio Grande do Sul 484 10.8
3 Minas Gerais 469 10.5
4 Rio de Janeiro 454 10.1
5 Parana 358 8.0
6 Santa Catarina 293 6.5
7-27 Other 212 24.9
Total 4 476 100.0
Source: CNPq Directory of Research Groups (2004).
Table 5. Co-Localisation of Firms and Interactive
Research Groups.
Geographic level No. Share (%)
(N = 4 476)
Same State 3 206 71.6
Same meso-region 2 196 49.1
Same micro-region 1 974 44.1
Source: CNPq Directory of Research Groups, 2004.
Table 6. Comparison: Database and the Control Groups.
Same Same
state (%) meso-region (%)
Group N Interact Control Interact. Control
All interactions 4 476 71.6 21.5 49.1 8.4
Engineering 1 869 72.0 22.7 47.9 9.0
Agricultural 946 69.3 18.7 37.2 2.7
Sciences
Healthy Sciences 699 72.2 21.7 54.8 9.2
& Medicine
Natural and 568 69.4 15.5 53.2 7.0
Earth Sciences
Human and Social 394 77.7 30.5 67.0 19.8
Sciences
Same
micro-region (%)
Group N Interact. Control
All interactions 4 476 44.1 7.6
Engineering 1 869 43.5 8.1
Agricultural 946 28.9 1.9
Sciences
Healthy Sciences 699 50.5 8.0
& Medicine
Natural and 568 49.3 6.5
Earth Sciences
Human and Social 394 64.7 19.0
Sciences
Source: CNPq Directory of Research Groups, (2004).
Table 7. Interactions Distance (in km).
Knowledge areas N Database Control group
Distance Distance
All Interactions 4 476 294 936
Engineering 1 869 284 828
Agricultural Sciences 946 323 1 047
Healthy Sciences and 699 305 985
Biology
Natural and Earth 568 335 1,111
Sciences
Human and Social 394 197 845
Sciences
Knowledge areas Geographic Distance
average difference
distance (+) between
areas (#)
All Interactions 1 502 --
Engineering -- -10.2
Agricultural Sciences -- 28.5
Healthy Sciences and -- 10.4
Biology
Natural and Earth -- 40.9
Sciences
Human and Social -- -97.6
Sciences
Source: CNPq Directory of Research Groups--2004.
(+) This value was measured by the average value of a
matrix with all distances between the 558 Brazilian
micro-regions. It's important to point that the maximum
distance between two micro-regions is 4 504 km.
(#) The Engineering and Nature and Earth Sciences groups
distance differences to the average of all groups presented
no significance in the Wilcoxon rank-sum test with p-value
of 0.6722 and 0.3262, respectively. The Healthy Sciences
and Biology presented a p-value of 0.0236 and the others 0.000.
