Firm Collaboration and Modes of Innovation in
Norway
Rune Dahl Fitjar, IRIS
Andrés Rodríguez-Pose, LSE
Research questions
› How does industrial (DUI) and scientific (STI) collaboration affect the innovative
capacity of firms?
› Does it matter whether industrial and scientific partners are located nearby or at a
distance?
› Are there any differences between city regions in Western Norway and other city
regions when it comes to
• the use of industrial and scientific partners?
• the geographical location of such partners?
• the effect of industrial and scientific partners on innovation?
› Are there any differences across sectors?
Papers:
Rune Dahl Fitjar and Andrés Rodríguez-Pose (2013): Firm Collaboration and Modes of
Innovation in Norway. Research Policy 42 (1): 128-138.
Rune Dahl Fitjar and Andrés Rodríguez-Pose (2012): Interaction and Innovation across
Different Sectors: Findings from Norwegian City Regions. Paper under review by
Regional Studies.
Sources of innovation
› Scientific/technical knowledge
› Learning by doing
• Linear model of innovation (Bush 1945,
Maclaurin 1953)
• Knowledge spillovers (Audretsch and
Feldman 1996, Sonn and Storper 2008)
• Regional innovation systems (Lundvall
1992, Cooke and Morgan 1998),
industrial districts (Becattini 1987),
learning regions (Morgan 1997),
innovative milieux (Aydalot 1986)
• Key variables: R&D investment, human
capital, links to scientific partners
• Key variables: Informal interaction, social
capital, organisations, institutions,
markets
• Key skills: Know-why, know-what
(Jensen et al. 2007)
• Key skills: Know-how, know-who (Jensen
et al. 2007)
› STI mode of innovation
› DUI mode of innovation
Two types of interaction in DUI mode
› Within supply-chain
› Outside supply-chain
• With suppliers and customers
• With other firms, such as competitors
• Close complementary bonds within
supply chain
• Transfer of knowledge not the main
purpose
• Clear economic purpose, joint aim of
improving products
• Unintended knowledge spillovers may
happen
• Contractual links
• Externalities from diversification (Jacobs),
potential for excessive cognitive distance
(Boschma 2005)
• Externalities from specialisation
(Marshall) or related variety (Frenken et
al. 2007, Boschma and Iammarino 2009)
The geography of STI and DUI
› STI mode
› DUI mode
• Costly search for knowledge requires
purpose-built connections – global
pipelines (Bathelt et al. 2004)
• Based on shared problems and
experiences
• Analytical and codified knowledge travels
well (Asheim and Gertler 2005)
• More frequent in industries with
synthetic or symbolic knowledge base
(Moodysson et al. 2008)
• Geographical distance not necessarily a
problem
• Top research centres often located far
away
• Tacit knowledge
• Local buzz (Storper and Venables 2004),
informal interaction
• ’Being there’ (Gertler 1995)
• Strong value-added of local cooperation
Modes of innovation and knowledge bases
Synthetic knowledge
Characteristics
In-house
Interactions
Industries
Doing, using,
learning
Characteristics
Industries
Science,
technology,
innovation
In-house problem solving
Inductive learning processes
Interaction between departments within
the firm
Aeronautics
Aerospatial industry
Petroleum and energy
Based on frequent, often informal,
interaction
Important role for tacit knowledge
Process innovations and incremental
product innovations
General manufacturing
Construction
Retail trade
Role of R&D constrained to applied
research
Characteristics
Industries
Specialised manufacturing
Biotechnology
Nanotechnology
Analytical knowledge
Creation of in-house R&D labs or
departments
Large firms
Capital intensive sectors
Formal interactions between in-house
R&D units and research centres
Key role for codified knowledge
Radical product innovations
Formalised interaction
Key role for scientific research and
models
Radical and complex product and process
innovation
Data
› Tailor-made survey of firms with more than 10 employees in
Norway
› Targeting the managers of those firms
› Conducted by telephone
› In the five largest urban agglomerations in Norway
› In the spring of 2010
› Examining
• Innovation during the last three years
• The use of external partners in innovation processes
• The location of external partners used
Innovation in Norwegian city regions
Bergen has the lowest
share of innovative
firms in all categories
Product
Process
Oslo has the highest
share of innovative
firms in all categories
Total
Radical
Total
Radical
N
Oslo
59.6 %
34.0 %
50.4 %
20.4 %
403
Bergen
46.4 %
25.1 %
42.4 %
16.5 %
401
Stavanger
54.0 %
33.8 %
46.8 %
18.8 %
400
Trondheim
52.3 %
29.0 %
48.7 %
19.7
% is 300
Stavanger
mostly very
(% yes)
close to the average
Kristiansand
58.0 %
30.0 %
47.0 %
20.0 %
100
Total
53.4 %
30.5 %
46.9 %
18.8 %
1604
Innovation
inThere
different
sectors
are
most
Manufacturing
firms
have
thenon-innovative
highest
firms
in construction
level of product
innovations
% of firms
Manufacturing
Construction Trade
Product
innovation Manufacturing also has the
highest share
No innovation
35.8 of firms 71.7
introducing
(2.8)process innovations
(2.8)
Incremental
21.3
16.3
innovation
(2.4)
(2.3)
Radical
43.9
12.0
innovation
(2.9)
(2.0)
Process
innovation
No innovation
37.5
62.4
(2.8)
(3.0)
Incremental
39.2
20.9
innovation
(2.8)
(2.5)
Radical
23.3
16.7
innovation
(2.4)
(2.3)
N
296
258
39.1
(2.9)
24.6
(2.6)
36.2
(2.9)
65.2
(2.9)
24.6
(2.6)
10.1
(1.8)
276
Transport and communications
has the lowest level of radical
innovation
Hotels and
restaurants
Transport and Other
communication services
However, other services
has a higher
55.8
49.2 share of40.7
radical process
innovation
(4.4)
(4.5)
(2.4)
Hotels and
25.6restaurants 24.2
24.5
have the(3.9)
fewest radical (3.9)
(2.1)
process innovations
18.6
2.6
34.7
(3.4)
(4.0)
(2.3)
62.8
(4.3)
32.6
(4.1)
4.7
(1.9)
129
63.7
(4.3)
20.2
(3.6)
16.1
(3.3)
124
43.5
(2.4)
29.6
(2.2)
26.9
(2.1)
432
Percent of companies using partner type
60
Suppliers and customers
are the most common
partners at all scales
Very few firms collaborate
with scientific partners
abroad
50
40
30
20
10
0
Internal
Suppliers
Customers
Regional
Competitors Consultants
National
International
Universities Research inst.
Percent using partner type, by city region
Collaboration with suppliers and customers
is equally common in all regions
Stavanger has by far
the lowest level of
collaboration with
universities
90
80
70
Stavanger and
Bergen also
collaborate the
least with research
institutes
60
50
40
30
20
10
0
Internal
Suppliers
Oslo
Customers Competitors Consultants Universities Research
inst.
Bergen
Stavanger
Trondheim
Kristiansand
DUI and STI collaboration across different sectors
Trade and manufacturing
firms collaborate the least
with competitors
90%
80%
70%
Construction firms
collaborate the least with
universities and research
institutes, while other service
firms collaborate the most
60%
50%
40%
30%
20%
10%
0%
Internal
Manufacturing
Suppliers
Construction
Customers Competitors Consultancies Universities
Trade
Hotels & rest.
Transp. & comm.
Research
institutes
Other services
Innovation and collaboration
with partner
Collaboration with
customers has a
significant
effect on
Collaboration
New
to market withProcess
product
innovation
competitors
has a
negative effect on -0.02
0.20
product innovation(0.12)
(0.13)
New to industry
Within congl
0.39**
(0.12)
Suppliers
0.39**
(0.14)
0.33*
(0.16)
0.76***
(0.14)
0.38*
(0.19)
Customers
0.36**
(0.13)
0.54***
(0.15)
0.03
(0.13)
-0.03
(0.17)
Competitors
-0.39***
(0.12)
-0.55***
(0.13)
-0.14
(0.12)
-0.09
(0.15)
Consultancies
0.15
(0.12)
0.18
(0.13)
0.16
(0.12)
0.03
(0.15)
Universities
0.30*
(0.16)
0.53***
(0.15)
0.21
(0.15)
0.13
(0.18)
Research inst
0.26
(0.16)
0.20
(0.16)
0.26
(0.16)
0.79***
(0.18)
Logistic regression models, N = 1604.
Controls: Sector, region, education, age, board memberships, ownership, size
0.10
(0.15)
* p < 0.05, ** p < 0.01, *** p < 0.001
Collaboration with
universities has a
Product
significant effect on
product innovation
Collaboration with
types suppliers significantly
affect all forms of
innovation
Fitted probabilities of innovation
Product
with a supplierNew
or to
New to Collaborating
market Process
Collaborating
withthe
a competitor
customer
increases
likelihood
industry
likelihood
8 %ofreduces
productthe
innovation
byby
9 %points
points
compared
any
0.30 to not having
0.10
partners
0.30
0.11
No partners
0.34
0.15
Within congl
0.43
0.17
Suppliers
0.43
0.19
0.48
0.14
Customers
0.43
0.23
0.31
0.10
Competitors
0.26
0.09
0.27
0.10
Consultancies
0.38
0.17
0.34
0.11
Universities
0.41
0.23
0.35
0.12
Research inst
0.40
0.17
0.36
0.20
Now, let’s run the model for firms in Bergen
Collaboration with
consultancies is
consistently negative
for firms in Bergen
New to market
Process
New to industry
Within congl
0.39
(0.25)
0.08
(0.29)
0.04
(0.25)
0.22
(0.34)
Suppliers
0.29
(0.29)
0.57
(0.36)
0.66**
(0.29)
1.12***
(0.43)
Customers
0.31
(0.30)
0.86**
(0.38)
Competitors
-0.43*
(0.25)
-0.61**
(0.30)
Consultancies
-0.09
(0.24)
Universities
Research inst
The coefficients
for universities
0.09
-0.76*
and research
(0.29)institutes are twice
(0.39) as
high for firms
0.03 in Bergen
0.28
(0.24)
(0.33)
-0.20
(0.28)
-0.21
(0.24)
-1.12***
(0.34)
0.60*
(0.32)
0.92***
(0.33)
0.25
(0.31)
0.12
(0.41)
0.52
(0.36)
0.71*
(0.37)
0.23
(0.35)
1.15***
(0.42)
Logistic regression models, N = 401.
Controls: Sector, education, age, board memberships, ownership, size
* p < 0.10, ** p < 0.05, *** p < 0.01
Product
And the same model for
Within congl
0.31
(0.24)
0.03
(0.25)
-0.13
(0.24)
0.03
(0.30)
Suppliers
0.81***
(0.29)
0.38
(0.31)
0.50*
(0.29)
-0.09
(0.37)
Customers
0.66**
(0.27)
0.78***
(0.30)
Competitors
-0.83***
(0.25)
-0.61**
(0.26)
-0.32
(0.24)
-0.21
(0.31)
Consultancies
-0.06
(0.24)
-0.02
(0.24)
0.05
(0.23)
0.38
(0.30)
Universities
0.02
(0.34)
0.43
(0.32)
-0.36
(0.33)
-0.36
(0.39)
Research inst
0.38
(0.33)
0.25
(0.32)
0.36
(0.32)
1.11***
(0.35)
University0.51*
collaboration has0.43
no effect on
innovation
for firms in Stavanger
(0.27)
(0.36)
Logistic regression models, N = 400.
Controls: Sector, education, age, board memberships, ownership, size
* p < 0.05, ** p < 0.01, *** p < 0.001
Product
The effects of
inter-firm
firms in Stavanger
collaboration are
stronger than in
New to market
Process
the full model New to industry
Do the effects of collaboration differ across sectors?
Generalised ordinal
regression model
Suppliersincremental
Suppliersradical
Customersincremental
Customersradical
Competitorsincrem.
Competitorsradical
Consultanciesincrem.
Consultanciesradical
Universities
Research institutes
0.11
(0.25)
0.16
(0.30)
Trade
Hotels &
restaurants
Collaboration with
0.22
a
(0.41)
Collaboration
with
1.04**positive effect in
1.52**
competitors
has
a
(0.48)manufacturing
(0.64)
-0.30 and
effect in trade
-1.03 negative
(0.34)
-0.41
other
services
(0.67) and other services (0.82)
0.27
0.37
customers
has
(0.32)
(0.28)
0.66**
(0.30)
-0.01
(0.38)
0.01
(0.27)
-0.29
(0.34)
0.39
(0.28)
-0.94*
(0.51)
Transport &
comms.
Other services
0.83*
(0.48)
0.26
(0.20)
0.09
(0.59)
0.57**
(0.23)
-0.02
(0.46)
0.94***
(0.26)
0.41
(0.55)
(0.20)
0.62
STItopartners are particularly
They also tend
(0.64)
important
construction
have a positive
effect in-0.91*
-0.37
in manufacturing
(0.32)
(0.51)
-0.32
-0.52***
-1.41***
(0.37)
(0.44)
-0.36
(0.36)
-0.35
(0.30)
1.30**
(0.58)
0.40
(0.30)
0.88***
(0.32)
1.26**
(0.53)
0.89**
(0.37)
-0.14
(0.49)
0.47
(0.54)
0.15
(0.25)
0.16
(0.30)
0.99**
(0.47)
-0.14
(0.42)
0.54
(0.63)
-0.53
(0.74)
0.08
(0.26)
0.30
(0.25)
0.71
(0.44)
2.28***
(0.70)
0.56
(0.57)
-0.03
(0.20)
* p < 0.05, ** p < 0.01, *** p < 0.001
Within conglomerate
Manufacturing Construction
Geographical dimension of DUI and STI
DUI non-supply-chain
Regional
28.4
(1.1)
Non-regional
19.0
(1.0)
DUI supply-chain
67.0
(1.2)
61.5
(1.2)
STI
48.0
(1.2)
29.1
(1.1)
It is much more
common to collaborate
with STI partners within
the region
Firms collaborate almost
as frequently with nonregional suppliers and
customers as with
regional ones
Percent using partner type, by city region and
Firms in Bergen collaborate the least with
partner location
The difference between Stavanger and Bergen is
both DUI and STI partners outside the region
80%
Firms in Oslo collaborate the least with
regional partners of all types
mainly that Stavanger firms collaborate more with
non-regional consultancies. For university and
research institutes, the numbers for Stavanger and
Bergen are almost identical
70%
60%
50%
40%
30%
20%
10%
0%
More firms in Stavanger collaborate with regional
consultancies (43 %) than in any other region,
while Stavanger has the lowest level of
collaboration with its regional university (13 %) of
DUI non-supply- DUI non-supply- DUI supply-chain DUI supply-chain
STI regional
all regions except Oslo (11 %). In Trondheim, 30 %
chain regional chain non-regional
regional
non-regional
of firms collaborate with the regional university.
Oslo
Bergen
Stavanger
Trondheim
Kristiansand
STI non-regional
Product
In the STI mode, regional
cooperation also has a
positive effect on product
innovation
travels better?
New to market Process
New to
For regionalindustry
collaboration, STI-
DUI non-supp
regional
-0.20
(0.13)
-0.51***
(0.15)
DUI non-supp
non-regional
-0.30*
(0.15)
-0.13
(0.16)
DUI supply-ch
regional
0.12
(0.12)
DUI supply-ch
non-regional
typeFor
hasnon-regional
a stronger effect
-0.13
-0.08 DUI-type
collaboration,
(0.13)
(0.17) effect than
has a stronger
STI-type-0.01
-0.07
(0.15)
(0.18)
0.17
(0.13)
0.13
(0.12)
-0.03
(0.15)
0.73***
(0.12)
0.72***
(0.14)
0.50***
(0.12)
0.42**
(0.16)
Scientific
regional
0.23*
(0.12)
0.40**
(0.13)
0.20
(0.12)
0.14
(0.15)
Scientific nonregional
0.37**
(0.14)
0.33*
(0.14)
0.33*
(0.13)
0.35*
(0.16)
Logistic regression models, N = 1602.
Controls: Sector, region, education, age, board memberships, ownership, size
* p < 0.05, ** p < 0.01, *** p < 0.001
In the DUI mode, only
The effect of collaborating
non-regional
cooperation
What
type
ofisknowledge
with
competitors
has a significant effect on
negative both within and
innovation
outside the region
Fitted probabilities of innovation
Product
with a supplierNew
or to
New to Collaborating
market Process
This
also
holds
radical
customer outside the region industry
morefor
than
product innovation
doubles the likelihood of product
0.09
innovation0.34
for an average firm
No partners
0.40
0.15
DUI non-supp
regional
0.33
0.09
0.30
0.09
DUI non-supp
non-regional
0.30
0.13
0.32
0.09
DUI supply-ch
regional
0.45
0.18
0.39
0.09
DUI supply-ch
non-regional
0.83
0.32
0.57
0.14
Scientific
regional
0.51
0.23
0.42
0.11
Scientific nonregional
0.58
0.21
0.48
0.13
How does this work in Bergen?
Collaboration with nonregional STI partners has a
particularly strong effect
on product innovation in
Bergen
New to market Process
New to
industry
DUI non-supp
regional
-0.23
(0.27)
-0.39
(0.32)
0.02
(0.26)
0.38
(0.35)
DUI non-supp
non-regional
-0.74**
(0.33)
-0.38
(0.36)
-0.11
(0.31)
-0.03
(0.39)
DUI supply-ch
regional
0.24
(0.27)
0.57*
(0.31)
0.32
(0.26)
-0.14
(0.33)
DUI supply-ch
non-regional
0.86***
(0.26)
0.99***
(0.31)
0.48**
(0.25)
0.63*
(0.35)
Scientific
regional
0.16
(0.25)
0.42
(0.27)
-0.04
(0.24)
-0.80**
(0.32)
Scientific nonregional
1.11***
(0.33)
0.96***
(0.33)
0.04
(0.30)
0.40
(0.37)
* p < 0.10, ** p < 0.05, *** p < 0.01
Product
This is mostly the negative effect of
Logistic regression models, N = 401.
consultancies seen in the previous analysis
Controls: Sector, education, age, board memberships, ownership, size
Collaboration with regional
competitors is particularly
harmful in Stavanger
What about Stavanger?
In Stavanger, regional STI
partners tend to have a
positive effect, while nonregional STI partners
matter less
New to market Process
DUI non-supp
regional
-0.89***
(0.26)
-0.81***
(0.29)
-0.41
(0.26)
-0.23
(0.34)
DUI non-supp
non-regional
-0.05
(0.32)
-0.04
(0.32)
0.10
(0.32)
-0.03
(0.38)
DUI supply-ch
regional
0.21
(0.26)
0.15
(0.27)
0.45*
(0.26)
0.00
(0.33)
DUI supply-ch
non-regional
0.90***
(0.25)
0.90***
(0.27)
0.60**
(0.25)
0.61*
(0.33)
Scientific
regional
0.30
(0.24)
0.60**
(0.25)
0.07
(0.24)
0.33
(0.30)
Scientific nonregional
-0.12
(0.27)
-0.22
(0.27)
0.20
(0.27)
0.55*
(0.31)
Logistic regression models, N = 400.
Controls: Sector, education, age, board memberships, ownership, size
New to
industry
* p < 0.10, ** p < 0.05, *** p < 0.01
Product
Conclusion
› External cooperation is an important source of firm innovation
› Both STI and DUI partnerships matter
• Firms in Stavanger and Bergen tend to use fewer STI partners than firms in other regions –
especially universities and research institutes
› However, local interaction has no significant effect on innovation – especially within
DUI mode
› Cooperation with competitors can significantly harm firms’ innovative ability
• Close links to regional competitors has a particularly strong negative effect in Stavanger
› Formal pipeline-type interactions key source of innovation – both in the STI and in
the DUI mode
• Fewer firms in Bergen tend to use non-regional partners in both the STI and DUI mode
• Bergen firms who do use such partners, tend to be much more innovative than their peers
› Excessive cognitive proximity within small and homogeneous regions may be
detrimental to innovation
› Heterogeneity among agents is important
For more info:
Hurrah for 17 May!