COLLABORATION NETWORKS AND FIRM PERFORMANCE: THE EFFECT OF FIRM AND PARTNER POSITIONING ON INNOVATIO

1
COLLABORATION NETWORKS AND FIRM PERFORMANCETHE
EFFECT OF FIRM AND PARTNER POSITIONING ON
INNOVATIONHans Frankort(Organization
Strategy)METEOR / April 4, 2007
2
Outline

• Introduction
• Firm and partner positioning
• Relevance
• Hypotheses
• Methods
• Results
• Conclusions

3
Introduction

• PhD project, September 2005-now
• Alliance networks, knowledge flows and
  innovation
• (Advisors John Hagedoorn / Wilko Letterie)
• Today one of the thesis papers work in progress
• Goal testing an extended concept of network
  positioning in an RD alliance network

4
Firm positioning Connecting direct ties
(B)
(A)

• Focus on direct contacts (e.g., Burt, 1992
  Coleman, 1988)
• Timely information
• Access to others knowledge
• Building reputation

5
Partner positioning Indirect ties
(A)
(B)
Singular focus on direct contacts possibly
insufficient (e.g., Baum et al., 2006, on hybrid
partner positions Burt, 2007, on secondhand
brokerage)
6
Relevance in RD networks (1)

• Extant work considers several
• Firm-level (e.g., Ahuja, 2000 Hausman et al.,
  1984 Sampson, 2005),
• Relational (e.g., Gomes-Casseres et al., 2006
  Phelps, 2005 Sampson, 2007), and
• Partner-level (e.g., Baum et al., 2000 Stuart,
  2000) antecedents to firm-level innovation.
• In terms of networks
• Dominant focus on direct ties and numbers of
  indirect ties (e.g., Ahuja, 2000 Gulati, 1995
  Soh, 2003)
• Partner-level network positioning in terms of
  brokerage or closure lacking so far

7
Hypotheses (1) Firm positioning
(A)
(B)

• Hypothesis 1. In an interfirm RD network, a
  firms subsequent innovative performance
  increases with the amount of structural holes it
  spans.

8
Hypotheses (2) Partner positioning

• Competing hypotheses for partner network
  positioning
• Hypothesis 2a. In an interfirm RD network, a
  firms subsequent innovative performance
  increases with the amount of structural holes
  spanned by its partners.
• Hypothesis 2b. In an interfirm RD network, a
  firms subsequent innovative performance
  decreases with the amount of structural holes
  spanned (i.e., increases with the amount of
  redundant ties maintained) by its partners.

9
Data

• RD alliance network in IT (technological
  subcategories 21-24 and 26 Hall et al., 2002)
• Alliance data from CATI (1970-1999 Hagedoorn,
  2002)
• Patent data from NBER / USPTO (1975-2002 Hall et
  al., 2002)
• Data for other variables from SPs Compustat,
  Datastream, etcetera (1975-1999)
• Basic database matching done for Gomes-Casseres
  et al., 2006
• Result 152 unique firm entities in 455
  horizontal RD alliances

10
Dependent variable Innovative performance

• Measured as the firm-level count of patents
  granted, based on years of application (cf.
  Ahuja, 2000 Rothaermel Hess, 2007 Sampson,
  2007)
• Modeled using a conditional fixed-effects
  negative binomial panel specification (Hausman et
  al., 1984)
• Innovative performance takes a one-year lead to
  all independent variables

11
Explanatory control variables

• I applied a 3-year moving window to construct the
  network measures and all relational and
  partner-level controls 1988-1990, 1989-1991,
  1990-1992, etcetera
• I checked the results against a specification
  using a dynamic window results identical
  (consistent with extant work by e.g. Gulati
  (1995), Gulati Gargiulo (1999), Bae Gargiulo
  (2004))

12
Explanatory variable (1) Firm structural holes

• Firm is investment in contact j pijt (1/nit)
  (nijt)
• Constraint on firm i (Burt, 1992 54-55)
• Firm structural holes of i fshit (9/8 cit)

13
Explanatory variable (1) Firm structural holes
Firm structural holes low Firm structural
holes high
14
Explanatory variable (2) Partner structural
holes

• Partner structural holes to firm i (Burt, 2007)
• where Ait is the set of is alters in t, and Ait
  is the number of is alters in t

15
Explanatory variable (2) Partner structural
holes
Partners structural holes Partner structural
holes low high
16
Results (1) Models (n 752)
p lt .05 p lt .01 p lt .001 year fixed
effects included
17
Results (2)

• Hence, the optimal
• network structure in
• terms of innovative
• output to firm A

18
Conclusions Main findings

• A network broker benefits most if its direct
  contacts are well-embedded in their respective
  networks, net of (1) firm positioning and (2) a
  set of firm-level, relational, and partner-level
  effects
• This reinforces work that assumes structural
  holes allow a firm to tap into distinct chunks
  of social structure
• Uncertainty-reducing role of partners
  embeddedness to the focal firm

19
Conclusions Limitations

• The usual suspects
• Generalizability to other sectors and across
  networks, i.e. those that are not high-tech, is
  limited
• Performance limited to innovation other purposes
  may be served by firms alliance networks

20
Conclusions Future research

• Replication in different sectors, using
  alternative performance measures
• Direct measurement of distinct knowledge in dense
  groups
• Consider interactions of structural and
  relational / node-specific contingencies
• Questions?