Building Dynamic Capabilities:
Innovation Driven by Individual, Firm, and Network
Level Effects
Paper forthcoming in Organization Science
Frank T. Rothaermel
Andrew M. Hess
Georgia Institute of Technology
© Rothaermel & Hess, 2007
1
The Biotechnology Revolution(s)
1
Global Pharmaceutical Industry
• R&D expenditures have grown from $6.8 billion in
1990 to $21.3 billion in 2000 (17% of sales)
• Development cost for new drugs have increased from
$231 million to $802 million over the same period
• Average sales per patented product have fallen from
$457 million in 1990 to $337 million in 2001
Constant 1999 dollars.
© Rothaermel & Hess, 2007
3
R&D Investments and New Drug Approvals
© Rothaermel & Hess, 2007
4
2
Value of Expiring Pharmaceutical Patents
8
7
6
5
B illio n
$ $
4
3
2
1
0
1998
1999
2000
2001
2002
2003*
2004*
Source: Warburg Dillon and Read
© Rothaermel & Hess, 2007
5
Research Questions
• Where is the locus of innovation capabilities?
– Is it within the individual, firm, or network level of
analysis
– Is this a multilevel story of capability development
involving interactions across levels of analysis?
• If so, are the different innovation mechanisms
complements or substitutes?
© Rothaermel & Hess, 2007
6
3
Empirical research on capability development
Most research has focused on one level of analysis:
– Network Level:
Powell, Koput, Smith-Doerr (1996), Rothaermel (2001),
Higgins and Rodriguez (2006)
– Firm Level:
Cohen and Levinthal (1989, 1990)
Tushman and Anderson (1986)
– Individual Level: Zucker and Darby studies
Such a focus makes two implicit assumptions:
– Homogeneity within non-focal levels of analysis
– Independence between focal and non-focal levels of analysis
© Rothaermel & Hess, 2007
7
Interactions Across Levels
• Complements vs. Substitutes
– Competing hypotheses are advanced to test the
interdependence across levels
• Two activities are complements (substitutes) if the
marginal benefit of each activity increases (decreases)
in the presence of the other activity:
– Complements: the interactions across levels are positive
– Substitutes: the interactions across levels are negative
© Rothaermel & Hess, 2007
8
4
Theoretical Model
Firm-Level:
R&D Capability (H2)
Network-Level:
H5: Substitutes
H4: Complements
Individual-Level:
Intellectual Human Capital (H1a)
Star Scientists (H1b)
Biotech Alliances (H3a)
Biotech Acquisitions (H3b)
© Rothaermel & Hess, 2007
9
Methodology: Overview
•
Developed a detailed & comprehensive panel dataset
(1980-2004) documenting:
• 900 biotech acquisitions
• 4,000 biotech alliances
• 13,200 biotech patents
• 110,000 non-biotech patents
• 135,000 research scientists
• 480,000 journal publications of biotechnology research
• 9.2 million journal citations
• Last but not least:
– These data are complemented by qualitative fieldwork through interviews
and direct observation before, during, and after completion of the study.
© Rothaermel & Hess, 2007
10
5
Dependent Variable
• Innovation Output
Biotechnology patent applications granted:
• Externally validated measure of technological novelty
• Critical to success in pharmaceutical industry and
correlated with key performance measures
– Citation-weighted patents
– New product development
© Rothaermel & Hess, 2007
11
Independent Variables
• Intellectual Human Capital (IHC):
– Searched ISI Scientific Citation Index for journal
articles published between 1980 and 2004:
• An organization’s name corresponding to a
pharmaceutical firm
• A keyword related to scientific research
• Longer time period than study period
– To address “rising star” effect
– To address right truncation
© Rothaermel & Hess, 2007
12
6
Pharmaceutical Firm Publications in Biotechnology
80,000
3
60,000
Publications
70,000
50,000
•
Resulted in a population of over 480,000 articles and
135,000 authors.
The average scientist published 3.8 papers that were
cited an average of 66.4 times
The average firm employed 214 publishing research
scientists per year
•
•
40,000
30,000
20,000
10,000
Ph
arm
Pf
iz
ac
W ia & Ba er
ar U ye
ne p r
r-L jo
am hn
Rh
Sh ber
on
ion t
e-P
ou Sa ogi
len nk
Ho
As c R yo
ec
tra o
hs
r
t M Zen er
ec
ari
on A a
Ya Ro HP
ma us
n se
Br N Aji ouc l
ist ip no hi
ol po mo
-M n to
ye Ro
rs ch
Sq e
ui
A
M lli Ser bb
its ed on
ub -S o
ish ign
To H
a
yo oe M i Ka l
Bo ch oc sei
se st R hid
ki o a
(T us
o se
Hy yob l
br o)
it
Ja
ns K ech
s
Ni en Paken
pp ha
on rm
In Ze
te o
Te rfer n
Ca Ky ch on
rte ow nic
r-W a are
M
al l ed
ac ex
e
Al Inc
tan
Ar a
An mou
aly r
tab
0
© Rothaermel & Hess, 2007
13
Independent Variables
Star Scientists:
• Number of publications & times cited
• Defined stars based on 3 standard deviations
above the mean in publications and citations
• Sample Statistics:
– Number of Stars @ st. dev > 3:
• By publication: 2,392 stars
• By citation:
1,570 stars
• Both:
851 stars
© Rothaermel & Hess, 2007
< 0.65% of total pop.
is responsible for
15.2% of total pubs &
27.3% of total cites
14
7
Distribution of Innovative Output
Star scientists
• publish 25x more articles
• are cited 45x more
45,000
Count of Publication Authors
40,000
35,000
30,000
Publication Count
25,000
20,000
15,000
10,000
5,000
76
71
66
61
56
51
46
41
36
31
26
21
16
11
6
1
0
Numbe r of Publications/Pate nts pe r Individual
15
© Rothaermel & Hess, 2007
4,000
40,000
3,500
35,000
3,000
Publication Count
30,000
2,500
Patent Count
25,000
2,000
20,000
1,500
15,000
1,000
10,000
Count of Patent Inventors
45,000
500
5,000
76
71
66
61
56
51
46
41
36
31
26
21
16
11
0
6
0
1
Count of Publication Authors
Distribution of Innovative Output
Numbe r of Publications/Pate nts pe r Individual
© Rothaermel & Hess, 2007
16
8
20,000
900
18,000
800
16,000
700
14,000
500
10,000
400
8,000
6,000
Non-star Authors
4,000
Star Authors
StarPubs
600
12,000
300
200
100
2,000
0
0
19
73
19
75
19
77
19
79
19
81
19
83
19
85
19
87
19
89
19
91
19
93
19
95
19
97
19
99
20
01
20
03
Non-star
pubs
The Role of IHC – Publication Count
© Rothaermel & Hess, 2007
17
Independent Variables and Controls
Other IV’s:
• R&D Capability
– R&D expenditures
• Biotech alliances and acquisitions
Controls:
–
–
–
–
–
–
–
–
–
Lagged biotech patents
Non-biotech patents
Time to Cohen-Boyer patent citation
Diversified pharmaceutical firm
Horizontal merger
Firm size (total assets)*
Firm performance (net income & revenues)
Country Effects: U.S., European, Asian (Japanese) Firm
Year Effects
* All financial data are in constant U.S. dollars
© Rothaermel & Hess, 2007
18
9
Results
Model 1: Controls Only
BPA
IRR
Factor Change
Firm Merged
p < 0.001
1.20
20%
Total Assets
p < 0.001
0.62
- 38%
Total Revenues
p < 0.001
1.22
22%
Time to Cohen-Boyer
Patent Citation
p < 0.001
0.55
- 45%
Non-Biotech Patents
p < 0.001
1.14
14%
Lagged Biotech Patents
p < 0.001
1.18
18%
19
© Rothaermel & Hess, 2007
Results – Direct Effect Hypotheses
Intellectual Human Capital
BPA
IRR
Factor Change
p < .001
1.15
15%
© Rothaermel & Hess, 2007
20
10
Results – Direct Effect Hypotheses
BPA
IRR
Factor Change
Intellectual Human Capital
p < .001
1.15
15%
R&D Expenditures
p < .05
1.32
32%
R&D Expenditures Squared
p < .001
0.92
-8%
21
© Rothaermel & Hess, 2007
Results – Direct Effect Hypotheses
BPA
IRR
Factor Change
Intellectual Human Capital
p < .001
1.15
15%
R&D Expenditures
p < .05
1.32
32%
R&D Expenditures Squared
p < .001
0.92
-8%
NS
-
-
Biotech Alliances
© Rothaermel & Hess, 2007
22
11
Results – Direct Effect Hypotheses
BPA
IRR
Factor Change
Intellectual Human Capital
p < .001
1.15
15%
R&D Expenditures
p < .05
1.32
32%
R&D Expenditures Squared
p < .001
0.92
-8%
NS
-
-
p < .05
1.04
4%
Biotech Alliances
Biotech Acquisitions
23
© Rothaermel & Hess, 2007
Results – Direct Effect Hypotheses
•
The effect of stars
BPA
IRR
Factor Change
Star Scientists
p < .01
1.08
8%
R&D Expenditures
p < .05
1.32
32%
R&D Expenditures Squared
p < .001
0.92
-8%
NS
-
-
p < .01
1.06
6%
Biotech Alliances
Biotech Acquisitions
© Rothaermel & Hess, 2007
24
12
Results – Direct Effect Hypotheses
•
•
•
The effect of stars disappears while controlling for non-stars
Unobserved heterogeneity
Non-stars fully mediate any star effect
BPA
IRR
Factor Change
NS
-
-
Non-Star Scientists
p < .05
1.10
10%
R&D Expenditures
p < .05
1.32
32%
R&D Expenditures Squared
p < .001
0.92
-8%
NS
-
-
p < .05
1.05
5%
Star Scientists
Biotech Alliances
Biotech Acquisitions
25
© Rothaermel & Hess, 2007
Results – Interaction Effect Hypotheses*
Individual x Firm Level
BPA
IRR
Factor Change
IHC x R&D Expenditures
p < .05
0.91
- 9%
Star Scientists x R&D Exp.
p < .05
0.92
- 8%
* only significant interactions are shown
© Rothaermel & Hess, 2007
26
13
Results – Interaction Effect Hypotheses*
Individual x Firm Level
BPA
IRR
Factor Change
IHC x R&D Expenditures
p < .05
0.91
- 9%
Star Scientists x R&D Exp.
p < .05
0.92
- 8%
Individual x Network Level
BPA
IRR
Factor Change
IHC x Bio Alliances
p < .001
0.94
- 6%
Non-Stars x Bio Alliances
p < .001
0.91
- 9%
* only significant interactions are shown
27
© Rothaermel & Hess, 2007
Results – Interaction Effect Hypotheses*
Individual x Firm Level
BPA
IRR
Factor Change
IHC x R&D Expenditures
p < .05
0.91
- 9%
Star Scientists x R&D Exp.
p < .05
0.92
- 8%
Individual x Network Level
BPA
IRR
Factor Change
IHC x Bio Alliances
p < .001
0.94
- 6%
Non-Stars x Bio Alliances
p < .001
0.91
- 9%
Firm x Network Level
BPA
IRR
Factor Change
R&D Exp. x Alliances
p < .01
1.08
8%
* only significant interactions are shown
© Rothaermel & Hess, 2007
28
14
Conclusions – Direct Effects
• Locus of innovation capabilities resides across different
levels
– In the intersection between individual, firm, and networklevel effects
• Significant amount of the variance in biotech patenting
is explained by individual-level factors
– Mediation of star effect on innovation by non-stars
• Stars close cognitive gap, while non-stars close operational gap (Lavie,
2006)
© Rothaermel & Hess, 2007
29
Conclusions – Direct Effects
• Firms are able to build, buy and access innovation
capabilities through
– Recruitment of IHC and star scientists,
– R&D spending,
– Acquisitions of new technology firms,
• But: Firms must already possess necessary R&D
capabilities to be a means by which firms can leverage
different innovation mechanisms
© Rothaermel & Hess, 2007
30
15
Conclusions – Interaction Effects
When attempting to innovate:
• Individual-level effects appear to be substitutes to firm
or network-level antecedents
• In contrast, firm and network-level effects appear to be
complements
© Rothaermel & Hess, 2007
31
16