do ties really bind? the effect of knowledge and commercialization ...
-
Upload
nguyenliem -
Category
Documents
-
view
218 -
download
1
Transcript of do ties really bind? the effect of knowledge and commercialization ...
DO TIES REALLY BIND? THE EFFECT OF KNOWLEDGE ANDCOMMERCIALIZATION NETWORKS ON OPPOSITION
TO STANDARDS
RAM RANGANATHANManagement Department, McCombs School of Business, University of Texas at Austin
LORI ROSENKOPFManagement Department, The Wharton School, University of Pennsylvania
We examine how the multiplicity of interorganizational relationships affects strategicbehavior by studying the influence of two such relationships—knowledge linkages andcommercialization ties—on the voting behavior of firms in a technological standards-setting committee. We find that, while centrally positioned firms in the knowledgenetwork exhibit lower opposition to the standard, centrally positioned firms in thecommercialization network exhibit higher opposition to the standard. Thus, the in-fluence of network position on coordination is contingent upon the type of interorgan-izational tie. Furthermore, when we consider these relationships jointly, knowledgecentrality moderates the opposing effect of commercialization centrality, such that thecommercialization centrality effect increases with decreasing levels of knowledgecentrality. In other words, firms most likely to delay the standard are peripheral in theknowledge network yet central in the commercialization network, which suggests thatthey have the most to lose from changes to current technology.
Over the past two decades, a key thesis that re-verberates across strategy and organization theoryresearch is that interorganizational relationships af-fect both firms’ strategic actions and outcomes. Em-pirical support for this idea demonstrates that afirm’s network position influences its investments(e.g., Stuart & Sorenson, 2007), its alliance forma-tion patterns (e.g., Gulati, 1995a, 1999; Rosenkopf &Padula, 2008; Walker, Kogut, & Shan, 1997), and itschoice of acquisition partners (e.g., Yang, Lin, &Peng, 2011). Similarly, strategy scholars inter-ested in understanding heterogeneity in firmperformance have found that a firm’s network
position influences several different outcomes,including innovation (e.g., Ahuja, 2000), survival(Baum, Calabrese, & Silverman, 2000), market share(Shipilov, 2006), and financial performance (e.g., Za-heer & Bell, 2005). The empirical evidence for theeffect of strategic networks comes from a widerange of industries, including investment bank-ing, computers, chemicals, mobile communica-tions, and biotechnology.
While most prior organizational network studieseither focus exclusively on a single interorganiza-tional relationship or pool different types of tiestogether in an aggregated network, the more com-
Financial support for this research was provided bythe Wharton Center for Leadership and Change Man-agement and the Mack Institute for Innovation Man-agement (both of which are based at the WhartonSchool of the University of Pennsylvania). MatheenMusaddiq, Varun Deedwandiya, Michael Lai, and Bev-erly Tantiansu provided invaluable research assis-tance. We appreciate the comments from Ranjay Gulatiand three anonymous reviewers, which helped to en-hance the quality of this paper substantially. We wouldalso like to acknowledge the many helpful suggestionsfrom the Management Department faculty, doctoralstudents at Wharton, and faculty at the University of
Texas, as well as from seminar participants at the 2011ASQ–OMT, 2011 Academy of Management, and 2011Strategic Management Society (SMS) conferences. Thispaper won a 2011 SMS “Best Conference PhD Paper”Prize and was a finalist for the 2011 SMS “Best Con-ference Paper” Prize. It is also part of a dissertationthat received a Dissertation Grant Award from theSMS’s Strategy Research Foundation, and was a final-ist for the Academy of Management’s Wiley Blackwell“Best Dissertation” Award in the Business Policy & Strat-egy division and for the “Best Dissertation” Award in theAcademy of Management’s Technology and InnovationManagement division.
515
� Academy of Management Journal2014, Vol. 57, No. 2, 515–540.http://dx.doi.org/10.5465/amj.2011.1064
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download, or email articles for individual use only.
plex reality is that firms are simultaneously embed-ded in multiple types of interorganizational net-works (Gulati, 1998; Shipilov & Li, 2012). Oneconcern with suppressing multiplicity is whetherresults obtained are generalizable across the differ-ent network types in which a firm is embedded(Inkpen & Tsang, 2005). Additionally, however, itoverlooks the interplay between a firm’s positionsin different interorganizational networks. In otherwords, a firm’s positions in multiple networks maysimultaneously influence its behavior (Gulati,1999), and, depending upon the behavioral impli-cations of specific types of ties, there may be bothbright and dark sides to firms’ embeddedness ininterorganizational relationships (Gulati & West-phal, 1999). Therefore, teasing out such interplaymay be particularly insightful if divergent behav-iors or outcomes can be predicted when the effectof each relationship is considered independently.1
In this paper, we examine multiplicity in firms’strategic alliance ties—interorganizational relation-ships that encompass a variety of formal agree-ments between firms, including joint research anddevelopment (R&D), technology exchange, licens-ing, and marketing arrangements (Gulati, 1995a).Our specific focus in studying multiplicity is on thedifferent types of alliance ties firms have acrosstheir sets of alliance relationships, and the diver-gent influences of these different types on firms’strategic choices.2 Within the large body of allianceresearch, significant attention has been devoted tothe potential of alliances to function as dual path-ways that can enable both new knowledge creationand exploitation of existing complementary know-how (Koza & Lewin, 1998; Lavie & Rosenkopf,2006; Rothaermel & Deeds, 2004). For instance, al-liances are often conceptualized as boundary-span-ning mechanisms that help organizations incorpo-rate distant knowledge and, ultimately, shape
technological evolution (e.g., Rosenkopf & Almeida,2003; Rosenkopf & Nerkar, 2001). In contrast, alli-ances are also featured as mechanisms that help firmsexploit existing complementary resources to success-fully adapt to the challenges posed by new technol-ogies (Rothaermel, 2001). While numerous networkstudies have examined networks comprising bothtypes of alliances, few acknowledge the exploration–exploitation duality in the underlying ties in con-structing these networks (see Gupta, Smith, & Shal-ley, 2006). As Table 1 shows, studies of alliancenetworks either combine both types into a single net-work or limit their scope to a single type, thus ne-glecting multiplicity within the alliance context.
Given this gap in interorganizational research,our goal is to highlight the tensions inherent inalliance network multiplicity by examining how afirm’s positions in different types of alliance net-works influence its strategic behavior differently.In investigating multiplicity, we dichotomizefirms’ participation in alliance networks based onthe functional objective of their alliance ties—whether the goal of specific ties is to spur explora-tion of new knowledge or to enable exploitationvia commercialization of existing capabilities andknow-how. Our context is a voluntary technologystandards-setting organization (SSO)—an industry-wide committee through which engineers from dif-ferent firms attempt to forge a shared set of rules forfuture technological development (Dokko, Nigam,& Rosenkopf, 2012; Rosenkopf, Metiu, & George,2001; Simcoe, 2012). SSOs have increasingly be-come a preferred arrangement for coordinatingtechnological change and innovation across largenumbers of firms (Chiao, Lerner, & Tirole, 2007;Farrell & Simcoe, 2012; Lavie, Lechner, & Singh,2007). Within this context, we examine the influ-ence of firms’ network positions on their votingbehavior to support or oppose the formation ofstandards. Intuitively, we expect the influence ofmultiple strategic networks on firms’ conduct to beespecially salient in the standards-setting context,as negotiations to arrive at a consensus standard areconducted multilaterally, where prior interorgani-zational linkages are likely to have an importantbearing. Our specific choices of context (SSO in atechnological change setting) and strategic behav-ior (voting for/against the standard) are particularlysuitable for illuminating the contrasting influencesof firms’ positions in knowledge networks focusedon longer-term exploration versus positions incommercialization networks focused on exploita-tion of current technologies.
1 One of the central tenets of the behavioral theory ofthe firm is that firms satisfice across multiple goals andacross different organizational coalitions (Cyert & March,1963). Since a firm’s position in each network signifies astrategic resource that has emerged from a path-depen-dent pattern of investments and commitments (Gulati,1999), it may make decisions that involve trade-offsacross these different network resources.
2 Note that the potential overlap of the different typesof ties a firm has with the same partner firm also consti-tutes a kind of multiplicity or multiplexity. Althoughthese types of ties are included in our analyses, we do notfocus on them either conceptually or empirically.
516 AprilAcademy of Management Journal
TA
BL
E1
All
ian
ceL
iter
atu
rean
dF
un
ctio
nal
Mu
ltip
lici
ty
Res
earc
hA
rtic
leS
etti
ng
All
ian
ceF
un
ctio
ns
Sam
ple
d
Con
cep
tual
Han
dli
ng
ofF
un
ctio
nal
Mu
ltip
lici
tyE
mp
iric
alH
and
lin
gof
Fu
nct
ion
alM
ult
ipli
city
Ah
uja
(200
0)C
hem
ical
sA
llal
lian
ces
Non
eN
one
Bae
&G
argi
ulo
(200
4)T
elec
omm
un
icat
ion
sA
llh
oriz
onta
lal
lian
ces
Non
eN
one
Bau
met
al.
(200
0)B
iote
chn
olog
y(C
anad
ian
)A
llal
lian
ces
Non
eS
pli
ttin
gco
un
tsby
fun
ctio
nG
ula
ti(1
995b
)B
iop
har
mac
euti
cals
,n
ewm
ater
ials
,au
tom
obil
esA
llal
lian
ces
Non
eC
ontr
olfo
rR
&D
alli
ance
s
Gu
lati
(199
5a,
1999
),G
ula
ti&
Gar
giu
lo(1
999)
New
mat
eria
ls,
ind
ust
rial
auto
mat
ion
,au
tom
otiv
ep
rod
uct
sA
llal
lian
ces
Non
eN
one
Gu
lati
&H
iggi
ns
(200
3)B
iote
chn
olog
yD
own
stre
amal
lian
ces
wit
hp
rom
inen
tp
har
mac
euti
cal
firm
sN
one
Non
e
Gu
lati
&S
ingh
(199
8)B
iop
har
mac
euti
cals
,n
ewm
ater
ials
,au
tom
obil
esA
llal
lian
ces
Non
eC
ontr
olfo
rR
&D
alli
ance
s
Kok
a&
Pre
scot
t(2
002,
2008
)S
teel
All
alli
ance
sN
one
Con
trol
for
rati
oof
mu
ltip
lex
par
tner
sto
tota
lp
artn
ers
Lee
(200
7)C
omp
ute
ran
dn
etw
orki
ng
All
hor
izon
tal
alli
ance
sN
one
Non
eL
ee,
Lee
,&
Pen
nin
gs(2
001)
Kor
ean
tech
nol
ogy
star
t-u
ps
Tec
hn
olog
yan
dm
arke
tin
gal
lian
ces
Non
eN
one
Lin
,Y
ang,
&A
rya
(200
9)C
omp
ute
r,st
eel,
ph
arm
aceu
tica
ls,
pet
role
um
,an
dn
atu
ral
gas
All
alli
ance
sN
one
Non
e
Yan
g,L
in,
&L
in(2
010)
Com
pu
ter
All
alli
ance
s(w
ith
inin
du
stry
)N
one
Non
eM
adh
avan
,K
oka,
&P
resc
ott
(199
8,20
04)
Ste
elA
llh
oriz
onta
lal
lian
ces
Non
eN
one
Ow
en-S
mit
h&
Pow
ell
(200
4)B
iote
chn
olog
yA
llal
lian
ces
Non
eC
ontr
olfo
rR
&D
ties
Ph
elp
s(2
010)
Tel
ecom
mu
nic
atio
ns
(equ
ipm
ent
man
ufa
ctu
rin
g)T
ech
nol
ogy
dev
.&
exch
ange
Non
eN
one
Pol
idor
o,A
hu
ja&
Mit
chel
l(2
011)
Ch
emic
als
Tec
hn
olog
yjo
int
ven
ture
sN
one
Non
eP
owel
l,K
opu
t,&
Sm
ith
-Doe
rr(1
996)
Bio
tech
nol
ogy
All
alli
ance
sR
&D
/Oth
ers
R&
D/N
on-R
&D
mea
sure
sR
osen
kop
f&
Alm
eid
a(2
003)
Sem
icon
du
ctor
sA
llal
lian
ces
Non
eN
one
Ros
enko
pf
etal
.(2
001)
Tel
ecom
mu
nic
atio
ns
(wir
eles
s)R
&D
alli
ance
sN
one
Non
eR
osen
kop
f&
Pad
ula
(200
8)T
elec
omm
un
icat
ion
s(w
irel
ess)
All
alli
ance
sN
one
Non
eR
owle
y,B
ehre
ns,
&K
rack
har
dt
(200
0)S
emic
ond
uct
ors
and
stee
lA
llal
lian
ces
Non
eS
tron
g(R
&D
)vs
.w
eak
(lic
ensi
ng)
Sin
gh(1
997)
Hos
pit
also
ftw
are
syst
ems
All
alli
ance
sT
ech
nol
ogy/
Oth
ers
Tec
hn
olog
y/O
ther
mea
sure
sS
tuar
t(2
000)
Sem
icon
du
ctor
sA
llh
oriz
onta
lal
lian
ces
Non
eC
ontr
olfo
rte
chn
olog
yal
lian
ceS
tuar
t,H
oan
g,&
Hyb
els
(199
9)B
iote
chn
olog
yA
llal
lian
ces
Non
eN
one
Wal
ker
etal
.(1
997)
Bio
tech
nol
ogy
All
alli
ance
sby
star
t-u
ps
Non
eN
one
Wh
itti
ngt
on,
Ow
en-S
mit
h,
&P
owel
l(2
009)
Bio
tech
nol
ogy
All
alli
ance
sN
one
Non
e
Yan
g,L
in,
&P
eng
(201
1)C
omp
ute
rin
du
stry
All
alli
ance
sE
xplo
rati
on/O
ther
sE
xplo
rati
onin
dex
We posit that, while centrality in a knowledgenetwork represents a firm’s control and likely ac-ceptance of future technological change in the formof proposed standards, centrality in a commercial-ization network represents a firm’s ability to appro-priate value from using alliance resources in thecurrent state of technology. Accordingly, in ournovel dataset of firms’ voting records over a 14-yearstandard-setting period in the computer industry,we find that firms peripheral in the knowledgenetwork yet central in current commercializationnetworks have the most to lose from technologicalchanges, which leads them to strategically delaythe standard. By demonstrating that different typesof alliance network ties have opposing effects onfirms’ strategic choices within these collaborativeinnovation communities, our study provides im-portant and novel insights to the relationship be-tween alliance networks and innovation. Departingfrom the majority of alliance studies that focus onthe positive effects of interorganizational relationson technological innovation and the benefits ofcomplementary assets, our findings show that spe-cific types of interorganizational ties that are aimedat exploiting current technology can also be a pow-erful force in impeding community-driven techno-logical change. Furthermore, the joint effects ofthese different types of network positions on firms’voting behavior suggest that firms are consideringtheir strategic options, at least in part, based ontheir whole multiplex portfolio of interorganization-al ties. This underscores the importance of explic-itly considering tie multiplexity when studyingfirm behavior and outcomes.
ORGANIZATIONAL CONTEXT—STANDARDS-SETTING COMMITTEES
Shaped by heterogeneous, path-dependent capa-bilities and beliefs, firms make different strategicbets on technologies (Denrell, Fang, & Winter,2003; Nelson & Winter, 1982). The presence of pos-itive network externalities and switching costs intechnology-driven industries such as personalcomputers (PCs) (Katz & Shapiro, 1986) provides aselection mechanism for a “winning” or dominantdesign amongst these technologies (Schilling, 2002;Tushman & Anderson, 1986). Although the emer-gence of a dominant design selects between differ-ent technological platforms (Baldwin & Woodard,2009), it leaves considerable scope for future tech-nical elaboration of standards, at both the compo-
nent and the intercomponent level.3 A firm’s abilityto control this subsequent technological evolutionsuch that its capabilities are sustained or even en-hanced may become a crucial determinant of itsadvantage (Teece, 2007). Technology standards-set-ting committees (i.e., SSOs) are contexts in whichfirms have opportunities to shape such change(Dokko et al., 2012). These industry-wide organiza-tions are venues where firms debate and coordinatethe technological rules that define a commonpath for future technological development (e.g.,Doz, Olk, & Ring, 2000). By shaping choiceswithin these organizations, firms thus have op-portunities to build attributes of their specifictechnologies into the evolving industry-widestandard (Garud, Jain, & Kumaraswamy, 2002;Gomes-Casseres, 1994).
Why Might Firms Contest the Standard?
Although the overall objective of standards com-mittees is to reduce technological uncertainty andavoid costly standards wars (Farrell & Simcoe, 2012),these cooperative arrangements are also character-ized by conflicts (e.g., Browning, Beyer, & Shelter,1995). Firms are likely to have divergent opinionson which particular technological alternative topursue in the standard, or even whether thereshould be a standard at all (Garud et al., 2002). Thisis because decisions made within these committeeshave significant and divergent consequences forthe value of firms’ technological capabilities (seeGomes-Casseres, 1994; Dokko & Rosenkopf, 2010;Rysman & Simcoe, 2008). In particular, as thesecommittees propose standardized rules of interac-tion between different system components (e.g.,Henderson & Clark, 1990), they have the potentialto cause adverse technological, economic, and or-ganizational consequences for participating firms.Technologically, by selecting between different al-ternatives, standards have the potential to differen-tially reward some firms while disadvantaging oth-
3 For example, even after the emergence of “Wintel”—the dominant design in the PC industry—there has beencontinuous change and refinement of the system, withmajor component-level innovations (e.g., solid state, op-tical, and flash memory following tape and disk technol-ogies) as well as architectural innovations (e.g., emer-gence of USB, Firewire, and SCSI (small computersystem interface) as alternative peripheral interface stan-dards to serial and parallel ports).
518 AprilAcademy of Management Journal
ers (e.g., Rysman & Simcoe, 2008).4 Economically, afirm’s technologies may need to be reconfigured tofollow new standardized rules—this may requireadditional investments without assured returns.Organizationally, the changes that a firm needs tomake in processes, structure, and resource alloca-tions to conform to the new standard may faceopposition within the organization (e.g., Chris-tensen & Bower, 1996; Henderson & Clark, 1990;Tripsas & Gavetti, 2000).
As a result, as discussions in a committee prog-ress, some firms are likely to view the emergingstandard as beneficial while others consider it det-rimental. This, in turn, will influence their respec-tive strategic behavior—specifically, firms that aremore likely to be disadvantaged by the standardwill contest its passage, while firms that are morelikely to benefit from it will exhibit support for itsexpeditious acceptance. As decision making in vol-untary standards committees is consensus based,even contestation by a small number of firms islikely to delay the standard-setting process, if notderail it completely. For instance, Simcoe (2012)finds that conflicts from divergent firm interests ledto an eight-month delay for the Internet Engineer-ing Task Force (IETF). Even if a standard doeseventually emerge, delays may allow inertial firmsto adapt to the new technological rules. In dynamicsettings marked by frequent technological change, afirm’s ability to negotiate and extend the life cycleof technologies, even by a few months, may becritical to its competitiveness and survival. In ourarguments, our focus is therefore on understandingthe drivers behind firms’ actions to contest the pas-sage of the standard.
Empirical Setting—INCITS
We focus our empirical inquiry on the Interna-tional Committee for Information Technology Stan-
dards (INCITS), a leading voluntary standardscommittee in the computer industry. Sustainedtechnological change, divided technical leader-ship, and an extensive use of strategic alliances byfirms in this industry make it an ideal setting tostudy the effect of interorganizational ties on oppo-sition to standards (Bresnahan & Greenstein, 1999;Rosenkopf & Schilling, 2007). Member firms inINCITS included both complementers (Adner & Ka-poor, 2010) and competitors (see Hagedoorn, Caray-annis, & Alexander, 2001), with major semiconductorfirms, hard disk manufacturers, cable and controllerfirms, and systems software firms involved in theprocess.5 Financially backed by the InformationTechnology Industry Council (ITI)—a large trade as-sociation representing the majority of firms in theinformation technology sector6—INCITS supportedan open governance structure and allowed for equalcontribution and representation of all organizations,large and small.7,8 Irrespective of size, a member firmcould appoint only one principal voting representa-tive—thus, no individual member firm controlled thestandard-setting process. Further, membership wasopen to all (including the general public) and the lowmembership fee spurred the involvement of several
4 By narrowing the feasible options for future techno-logical development, standards also reduce uncertainty,thereby promoting increased modularity (Sanchez & Ma-honey, 1996). As products become more modular, com-ponent tasks get decoupled, resulting in further special-ization and architectural change that may be competencedestroying for some firms (Baldwin & Clark, 2000; Hen-derson & Clark, 1990). Standards also allow differentcomponents to be produced separately and different vari-ants of the same component to be used interchangeably(Garud & Kumaraswamy, 1995)—again, entailing archi-tectural shifts that may be challenging for firms.
5 The membership was representative of the popula-tion of firms in these sectors. In 2008, INCITS firms inour sample classified under the Standard Industrial Clas-sification (SIC) code 3570 (Computers and office equip-ment) had a combined market share of 95%, those clas-sified under 3571 (Electronic computers) had a combinedmarket share of 98%, and those classified under 3678(Electronic connectors) had a combined market shareof 70%.
6 ITI members employ more than one million people inthe United States, and, in 2000, their revenues exceeded$668 billion worldwide (Source: www.incits.org).
7 Between 1961 and 1997, INCITS was known as theAccredited Standards Committee X3, Information Tech-nology (Source: www.incits.org).
8 Accredited by the American National Standards In-stitute (ANSI), INCITS brings together more than 1,700firms for the creation and maintenance of formal de jureIT standards. It operates more than 50 different technicalcommittees under ANSI rules, which are designed toensure that voluntary standards are developed by the“consensus of directly and materially affected interests”(Source: www.incits.org). The influence of INCITS in theinformation technology sector is evident from 750� stan-dards its subcommittees have published, encompassing arange of technology domains including programminglanguages, computer graphics, cyber security, distributedprocessing, and computer peripheral interfaces.
2014 519Ranganathan and Rosenkopf
small start-up firms as well as independent technol-ogy consultants.
INCITS committees also followed several checksand balances to ensure that the standards reflecteda true consensus. During the standards develop-ment process, firms voted on ballot measures toratify important milestones, which encompassedthe entire range of standards-setting activities, fromthe initiation of a new project to the approval of thefinal specification document.9 Firms were also freeto propose ballot measures for any issues that war-ranted a vote from all participants. INCITS furtherrequired that member firms addressed the contest-ing votes and associated comments on a ballot,even if the required majority to pass the ballot hadbeen achieved.10 In other words, specific objectionsof firms needed to be addressed before the standardcould progress, even though these firms may rep-resent a small minority. Although such mecha-nisms prevented standards from progressing if out-standing concerns existed,11 they also implied thatfirms could delay the standard by voting anythingother than a “yes” on ballots.12,13
Within INCITS, we focus on decision makingwithin three interrelated subcommittees that de-vise standards for computer peripheral interfaces—the T10, T11, and T13 subcommittees.14 While all
three subcommittees run in parallel, they developclosely related interface standards (the T11 andT13 were created out of the T10 committee). Wetrack activity in these committees from 1994 (theyear T10 was formed) up until 2008.
HYPOTHESES
In the hypotheses that follow, we contend thatfirms’ decisions to contest or support the standardare shaped independently and jointly by their po-sitions in the “knowledge network” and the “com-mercialization network” of member firms in thecommittee. Our choice of networks follows thework of several scholars studying technologicalchange and innovation who have identified thesetwo types of relationships as being instrumentalinfluences on technological evolution (e.g., Adner& Kapoor, 2010; Afuah & Bahram, 1995; Henderson& Clark, 1990; Rosenkopf & Almeida, 2003; Rosen-kopf & Nerkar, 2001; Stuart & Podolny, 1996).
Effect of Knowledge Network Position
The main principle behind how a firm’s knowl-edge network position affects its opposition to thestandard is that any firm’s influence on standardsdiscussions depends on the relational context thatthe firm’s knowledge is positioned within (Podolny& Stuart, 1995). Conceptually, each node in aknowledge network of member firms in the stan-dards committee represents a particular firm’s tech-nological knowledge base, and the ties betweenfirms represent some convergence of their knowl-edge bases (e.g., Stuart, 2000).15 A firm that is rela-tionally embedded in such a knowledge network istherefore one with knowledge that has been instru-
9 INCITS identifies a total of eight milestones: (1) ap-proval of the standards project, (2) notification to thepublic, (3) technical development, (4) initial public re-view, (5) management review, (6) executive board ap-proval, (7) ANSI approval, and (8) publication.
10 “[T]he purpose of . . . letter ballot resolution is toresolve any comments submitted with ‘No’ votes in re-sponse to . . . letter ballots, such that those ‘No’ votesbecome ‘Yes’ votes and indicate greater consensus”(Source: www.incits.org).
11 As prior research has pointed out, such standardsmay ultimately face legitimacy issues and run the riskthat they don’t attract a crucial mass of committed firmsto develop products (Garud et al., 2002).
12 We provide greater detail on the different votingoptions in the Methods section of this paper.
13 Firms are also not permitted from discussing futurevotes. Ballots are generally submitted electronically andthe results of the ballots are available only after thevoting process is complete.
14 T10 is responsible for developing standards for con-necting peripheral devices to personal computers, partic-ularly the series of SCSI standards. T11 develops periph-eral standards targeted at higher-performance computingapplications, including the high-performance parallel in-terface and fibre channel sets of standards. Finally, T13
develops a family of standards relating to the ATA/SerialATA (AT Attachment) storage interface used to interfacethe majority of hard disks in PCs. These are interfacestandards committees in the “architectural innovation”sense (Henderson & Clark, 1990), as the specificationsthey draft affect different components of a computer sys-tem, including the microprocessor circuitry and digitallogic to support different peripheral devices, the algo-rithms and protocols to transfer data between these de-vices and the computer, the connectors (e.g., USB cables,converter plugs, ports, and sockets) that physically trans-mit this data, and the peripherals that store or generatethis data (e.g., disks, cameras, portable drives).
15 We measure this network in two distinct ways—with alliance ties and also with cross citations of patents.
520 AprilAcademy of Management Journal
mental in driving the innovation efforts of the otherfirms in the network. As discussions within thecommittee progress and different technological al-ternatives are evaluated by member firms, thegroup as a whole is less likely to find a consensussolution in a knowledge area that is distant fromthe core of the knowledge network (Fleming & So-renson, 2004; Levinthal, 1997). It follows, then, thatfirms that are the most centrally positioned in thisknowledge network are likely to be the closest interms of “knowledge distance” from the emergingstandard. In other words, because a central firm’sposition represents the extent to which a firm’sknowledge has been foundational amongst peerfirms in the committee (Stuart & Podolny, 1996),there is a greater likelihood that the technologicalrules that the committee identifies as part of thestandard will continue to build on this foundationalknowledge.
The idea that the relational structure of knowl-edge exhibits inertial pressures has important im-plications for the ability of firms to derive economicrents based on their technological capabilities. Onthe one extreme, the most central firms will be ableto expeditiously develop innovative productsbased on the standard compared to the other firms(the advantage of possessing foundational knowl-edge). They also benefit from reinforcing effects ofincreased royalties as other firms continue to buildon this knowledge (Stuart, 1998). At the other ex-treme, firms on the periphery of the knowledgenetwork face the challenging prospect that theirknowledge, which, thus far, has been a tangentialinfluence on other firms’ innovation efforts, will befurther marginalized by its exclusion from theemerging standard. During deliberations, thesefirms will not only be more likely to propose diver-gent ideas and viewpoints, but also be more likelyto have these proposals excluded from the stan-dard. This idea is also consistent with prior re-search on institutional change that suggests thatchallenges to a prevailing order are more likely tooriginate from the periphery of a field than the core(e.g., Kraatz & Moore, 2002; Leblibici, Salancik,Gopay, & King, 1991). For such firms, the economicand organizational costs of acceding to an unfavor-able standard will outweigh any shared industry-wide benefits that the standard is likely to usher in.We contend that these considerations will influ-ence firms’ strategic behavior in the committeesuch that:
Hypothesis 1. The more central a focal firm iswithin the knowledge network of memberfirms, the lower its opposition to the standard.
Effect of Commercialization Network Position
While the knowledge network position reflects afirm’s control and acceptance of future technologi-cal change in the form of the proposed standard, itscommercialization network position captures the pat-tern of its formal agreements to commercialize cur-rent technologies. Firms most central in thesecommercialization networks have maximized the uti-lization of complementary resources through exten-sive partnering with upstream and/or downstreamfirms (Hamel, Doz, & Prahalad, 1989; Koza & Lewin,1998).16 In the existing state of technology, such firmsare positioned advantageously to appropriate rentsusing existing network resources.17
To analyze the influence of firms’ positions insuch a commercialization network on their strate-gic behavior in the standards committee, we firstnote that, in an industry characterized by systemicinnovation and separability of innovation activi-ties, the critical complementary assets that a firmneeds to access in order to successfully commer-cialize its innovation are the other interconnectedtechnologies or components needed to completethe system (Rosenkopf & Schilling, 2007; Teece,1986). For example, in the computer industry, op-erating system software is a complementary assetfor hardware, application software is a complemen-tary asset to the operating system, and peripheralports are complementary to cables and connectors.
16 It is important to note that, although there is likelyto be some degree of overlap between these two kinds ofties—knowledge and commercialization—they are notconceptually the same. In other words, a firm’s search forpartners that is driven by commercialization needs (i.e.,product market penetration) is distinct from the samefirm’s pattern of knowledge flows underlying its own andothers’ innovations.
17 Although prior research has highlighted the“pipes” aspect (Podolny, 2001) of complementary assetnetworks (i.e., networks as resource access relation-ships that help firms adapt to technological change),emerging work suggests that such capabilities may alsoact as “prisms” through which firms evaluate newtechnological options (e.g., Wu, Wan, & Levinthal,2013; Taylor & Helfat, 2009). Clearly, supporting anemerging industry-wide standard is one such strategicchoice for firms that their positions in complementaryasset networks might influence.
2014 521Ranganathan and Rosenkopf
The function of a firm’s commercialization rela-tionships in such settings is primarily to ensurecompatibility of its innovation with the larger sys-tem.18 Second, beyond actually ensuring technicalintegration across the system, these agreementsserve as important signals to consumers that certaintechnologies offer them greater flexibility to mixand match components from different vendors.This becomes particularly decisive in industries inwhich concerns of interoperability and lock-in mayhinder adoption of new technologies (Katz & Sha-piro, 1986). Thus, a firm that is in a central positionin such a network of complementers also com-mands a certain competitive advantage by beingable to, ultimately, offer a broader range of productsto users.
If technology standards are adopted by the com-mittee, then the industry-wide transition to thesestandards and subsequent user adoption may com-pletely devalue a central firm’s prior investmentsin these complementary assets (Taylor & Helfat,2009). The cornerstone of advantage for the cen-trally positioned firm in the commercialization net-work is the lack of a system-wide standard—it isthe absence of the standard that creates the veryneed (and value) for such complementary bound-ary-spanning relationships. In contrast, the veryrationale for a technological standard is to achieveconvergence across the entire industry on how dif-ferent components of the system should interoper-ate. With the adoption of a standard, any particularfirm only needs to design its product to the stan-dards specification to automatically obtain thesame interoperability or complementary benefitsthat a central firm has achieved from investing inits network of commercialization ties. This createsstrong disincentives for firms occupying central po-sitions in the commercialization network to sup-port the quick passage of the standard. Even thoughthe passage of the standard may, ultimately, beunavoidable, central firms may choose to contest
the standard because the nature of the consensus-driven decision-making process makes it vulnerableto these delays (Simcoe, 2012), allowing them to con-tinue to exploit current network positions. With suf-ficient time, these firms may even be able to leapfrogthe standard by introducing the next generation oftechnologies that can then tap into existing relation-ships to preserve their advantage.19 Therefore:
Hypothesis 2. The more central a focal firm iswithin the commercialization network ofmember firms, the higher its opposition to thestandard.
Joint Influence of the Two Network Positions
In the preceding hypotheses, we argued for dis-tinct and opposing effects for a firm’s knowledgenetwork position and for its commercialization net-work position on its strategic behavior in the stan-dards committee. Since member firms are simulta-neously connected via knowledge ties and viacommercialization ties, how do these distinct stra-tegic network resources jointly interact to influencefirm behavior? This interaction is best exemplifiedby considering firms peripheral in the knowledgenetwork but central in the commercialization net-work. For these firms, as discussed in Hypothesis 1,the industry-wide adoption of a standard that theyare in a weak position to control is likely to causetechnological competence erosion. Such firms arefaced with choosing between two difficult alterna-tives should the standard emerge—either reinvestto align their own technologies with what the stan-dard mandates, or adopt a non-conforming strategyand “go it alone” against the industry standard. Forsuch firms, the need to protect interfirm commer-cialization investments in complementary technol-
18 For example, in our sample, firms that made harddisk drives entered into strategic alliances with firms thatmade disk controllers to ensure that their products inter-operated with one another (e.g., the alliance betweenSeagate Technology and Adaptec). Similarly, systemsoftware firms entered into alliances with PC makers tomarket compatible systems (e.g., VMware’s alliance withDell). Some of these alliances were targeted at specificproduct markets (e.g., VMware’s alliance with HP to in-tegrate its software ESX 3i with different models of HPProLiant servers).
19 The relationship between a firm’s existing ties and,consequently, its narrow view of technological changehas been highlighted in prior research—for instance,Christensen and Bower (1996) have found that existingcommitments to downstream customers and upstreamsuppliers restricts the ability of firms to adapt to techno-logical change. We hypothesize on a similar dynamic instandards-setting committees, with the distinction beingthat firms actually have the option to block technologicalchange before it occurs. Similarly, research on networkshas also suggested that certain kinds of institutionalchanges have the potential to rupture the value of exist-ing network relationships, and that firms that are moreembedded in these relationships risk losing their competi-tive advantage when such change occurs (Uzzi, 1997).
522 AprilAcademy of Management Journal
ogies becomes even more critical. Faced with aninevitable decline in the future value of their currenttechnological capabilities, firms peripheral in theknowledge network are likely to attempt to appropri-ate the most value that they can out of their currentcommercialization network. This will be reflected ina higher rate of opposition to the proposed standard,which is magnified by the strength of the firm’s po-sition in the commercialization network.
The same logic applies to firms on the other end ofthe spectrum—those central in the knowledge net-work but peripheral in the commercialization net-work have more to gain and less to lose by supportingthe standard, and, thus, are likely to exhibit a lowerrate of opposition to the standard. Therefore:
Hypothesis 3. Knowledge network centrality willnegatively moderate the effect of commercializa-tion network centrality on a firm’s opposition tothe coordinated standard. Specifically, the moreperipheral the firm’s knowledge network posi-tion, the greater the positive effect of commer-cialization centrality on its rate of opposition.
METHODOLOGY
Data Sources
Table 2 summarizes the variety of different sourcesfor the study’s data, the variables and measures thatwere constructed from these data, and the method ofconstruction.
Measures
Dependent variable. “Firm’s vote on a ballotmeasure”—to capture a firm’s opposition to thestandard, we collected data from the standards sub-committee archival records on all the 241 differentballot measures, across the three subcommittees,and across the 14-year study period. Since we ob-tained this data from the very first year of operationof these subcommittees, left censoring is not anissue. For each ballot, we first identified the yearthe ballot occurred and the specific technical sub-set to which it was related. We also identified theworking groups responsible for developing the par-ticular technical subset.20 We then manually ex-
tracted the record of all the firms’ votes for thisballot, mapping individual votes to unique memberfirms already identified based on the membershiproster. There were four different voting choices thatfirms could make—“yes,” “abstain,” “yes with com-ments/conditions,” and “no.” We assigned ordinalvalues of 0 (yes), 1 (abstain), 2 (yes with comments/conditions), and 3 (no) to these votes.21
Independent variables. Our main independentvariables are the measures of the firms’ positions inthe two strategic networks and the interactions be-tween these positions. We operationalized bothknowledge and commercialization ties with strate-gic alliance data by interrogating Dow Jones & Com-pany’s search tool Factiva for the alliance an-nouncements for each member firm.22 Our searchincluded the entire range of formal agreements;following prior research, we used five-year movingwindows to define these ties (e.g., Gulati & Gar-giulo, 1999; Gulati, 1999), and our first year ofrecord for alliances was 1989 (to appropriatelymatch the first observed year for the dependentvariable, which is 1994). We carefully cleaned thedata to remove duplicate ties (by flagging duplicatealliance announcements between the same memberfirms that appeared close in time to each other) andties that were rumored but did not materialize,resulting in a total of 10,389 uniquealliances with3,365 dyadic alliance ties between member firms.23
20 The standards subcommittees are organized intoseveral working groups. Each working group is respon-sible for a specific technical subset of the overall stan-dard. Member firms are free to join any working group
and contribute to the standard. We provide more detailson the working groups when we discuss the controlvariables, below.
21 Our particular ordering of votes was based on theidea that the underlying construct captured in the votewas the level of firms’ opposition the standard. Thus,although we tested the robustness of our models to thisordering, our interpretation of the standard committee’srules indicated that the level of effort required to addressand resolve objections from a “no” vote was higher thanthat required to address a “yes with comments” vote,which, in turn, was higher than that required to addressan “abstain” vote—as there was really no effort to addressa “yes” vote, this was assigned the base value of 0.
22 As Lavie (2007) and Schilling (2009) have shown,the Securities Data Company’s “SDC Platinum” data onalliances covers only a small subset (less than 50%) ofthe alliance population and is therefore inappropriate toaccurately construct an alliance network. By includingall leading news sources, Factiva provides a more com-prehensive dataset to track alliances.
23 Multilateral alliances (with more than two memberfirms) were elaborated to include all dyadic tie combina-tions amongst the participating firms.
2014 523Ranganathan and Rosenkopf
TA
BL
E2
Var
iabl
es,
Mea
sure
s,an
dD
ata
Sou
rces
Var
iabl
esM
easu
res
Dat
aS
ourc
es
Dep
end
ent
vari
able
Fir
m’s
vote
ona
ball
otC
oded
as0
�ye
s;1
�ab
stai
n;
2�
yes
wit
hco
nd
itio
ns/
com
men
ts;
3�
no
INC
ITS
ball
otd
atab
ase
Ind
epen
den
tva
riab
les
Kn
owle
dge
net
wor
kce
ntr
alit
y(1
)D
egre
ece
ntr
alit
yin
exp
lora
tory
alli
ance
net
wor
kof
mem
ber
firm
s,w
ith
edge
sw
eigh
ted
byn
um
ber
ofd
isti
nct
exp
lora
tory
ties
betw
een
firm
s(f
ive
year
mov
ing
win
dow
)
Fac
tiva
for
alli
ance
ann
oun
cem
ents
(2)
In-d
egre
ece
ntr
alit
yin
pat
ent
cita
tion
sn
etw
ork
ofm
embe
rfi
rms
(fiv
eye
arm
ovin
gw
ind
ow)
NB
ER
pat
ent
dat
a
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
yD
egre
ece
ntr
alit
yin
anco
mm
erci
aliz
atio
nal
lian
cen
etw
ork
ofm
embe
rfi
rms,
wit
hed
ges
wei
ghte
dby
nu
mbe
rof
dis
tin
ctco
mm
erci
aliz
atio
nti
esbe
twee
nfi
rms
(fiv
eye
arm
ovin
gw
ind
ow)
Fac
tiva
for
alli
ance
ann
oun
cem
ents
Con
trol
vari
able
sP
arti
cip
atio
nin
wor
kin
ggr
oup
asso
ciat
edw
ith
ball
otC
um
ula
tive
firm
-rep
rese
nta
tive
s’at
ten
dan
ceat
wor
kin
ggr
oup
mee
tin
gsa
INC
ITS
wor
kin
ggr
oup
doc
um
ents
Pat
ent
stoc
kon
stan
dar
d’s
tech
nol
ogie
sS
tock
offi
rm’s
pat
ents
onte
chn
olog
ies
emer
gin
gfr
omth
est
and
ard
Der
wen
tP
aten
tst
ock
(ove
rall
)S
tock
offi
rm’s
pat
ents
inre
leva
nt
broa
dte
chn
olog
ical
cate
gori
esre
late
dto
stan
dar
da
NB
ER
pat
ent
dat
a
Pat
ent
stoc
kd
iver
sity
Nu
mbe
rof
dis
tin
ctte
chn
olog
ical
clas
ses
refl
ecte
din
firm
’sp
aten
tst
ocka
NB
ER
pat
ent
dat
aC
itat
ion
sto
pat
ent
stoc
kN
um
ber
ofci
tati
ons
tofi
rm’s
pat
ent
stoc
kfr
omp
aten
tsn
otow
ned
bym
embe
rfi
rmsa
NB
ER
pat
ent
dat
a
Kn
owle
dge
insu
lari
tyN
um
ber
ofci
tati
ons
firm
mak
esto
own
pat
ents
aN
BE
Rp
aten
td
ata
Ext
ern
alal
lian
ces
(to
firm
sn
oton
stan
dar
ds
com
mit
tee)
Cou
nt
ofal
lian
ces
ton
on-m
embe
rsa
Fac
tiva
Siz
eF
irm
’sas
sets
and
firm
’sre
ven
ues
in$b
nb
Sta
nd
ard
&P
oor’
sC
omp
ust
atd
atab
ase,
Du
n&
Bra
dst
reet
’sH
oove
r’s
dat
abas
e,C
orp
orat
eT
ech
nol
ogy
Dir
ecto
ry(C
orp
tech
)F
inan
cial
slac
kF
irm
’sD
ebt
and
Fir
m’s
Cas
hin
$bn
bC
omp
ust
atF
inan
cial
per
form
ance
Net
inco
me
in$b
nC
omp
ust
atS
ecto
rsi
zeS
ecto
r’s
reve
nu
es(s
ecto
rd
efin
edby
pri
mar
yfo
ur-
dig
itS
IC)b
Com
pu
stat
,H
oove
r’s,
Cor
pte
ch
aS
quar
ero
otof
mea
sure
toal
levi
ate
skew
nes
s.b
Nat
ura
llo
gof
mea
sure
toal
levi
ate
skew
nes
s.
For each alliance, we coded the announcementdate and the unique member firm identifier foreach member firm that was in the alliance. As Fac-tiva does not automatically distinguish betweenknowledge and commercialization alliances, wefollowed the procedure used by Lavie and Rosen-kopf (2006); that is, manually reading the an-nouncement of each alliance to determine its cate-gory.24 Specifically, if the alliance involved a newknowledge-generating agreement (e.g., R&D, tech-nology co-development), then we categorized it as aknowledge network tie, whereas, if the alliance in-volved an agreement based on existing technologies,including interoperability testing and certification,joint marketing, original equipment manufacturer/value-added reseller, licensing, or production, thenwe coded it as a commercialization tie. Although bothtypes of ties may involve a technology component,what distinguishes knowledge ties from commercial-ization ties is a focus on developing new and rela-tively uncertain technologies oriented towards thefuture.25 The following is an example of an an-nouncement that we coded as a commercializationtie, despite the presence of a technology component,as it clearly involved combining current complemen-tary solutions rather than generating new ones (Katila& Ahuja, 2002):
Adaptec, Inc. (NASDAQ: ADPT), a global leader instorage solutions, today announced that SeagateTechnology (NYSE: STX) has combined UnifiedSerial Controllers from Adaptec with its ownCheetah15K 146 GB Serial Attached SCSI (SAS)disk drives to deliver a comprehensive SAS Evalu-ation Kit to Seagate system builders and resellers.
Similarly, the following is an example of an alli-ance agreement that we coded as a knowledge tie,as the focus is clearly on joint development ofnext-generation products and technologies:
International Business Machines Corp. has forged amajor alliance with rival electronics giants SiemensAG of Germany and Toshiba Corp. of Japan to de-
velop the next generation of computer memorychips. The companies said yesterday they will co-operate in the development of 256-megabit chipsthat will have 16 times more capacity than the chipscommonly in use. . . . The 256-megabit chips likelywould be used in future generations of small, pow-erful personal computers and workstations. The ad-vanced semiconductor should be ready by the endof the decade, the companies said.
In all, 38% of our member alliance ties werecoded as knowledge ties and 62% were coded ascommercialization agreements.
“Knowledge network centrality” and “commer-cialization network centrality”—for each subcom-mittee and year, we first identified knowledge net-work ties between two firms if we coded at leastone knowledge alliance between them in the rele-vant five-year window. We then calculated a degreecentrality measure for each firm using the numberof such ties as edge-weights to account for thestrength of these relationships (Miura, 2012).26 Wesimilarly derived commercialization network cen-trality, using commercialization ties instead ofknowledge ties.
In addition to knowledge generation alliances,knowledge flows between firms are also revealed inpatent citations, which document the technologicalantecedents of inventions (Benner & Tushman,2002). Following several studies that have usedpatent citations to measure such knowledge flows(e.g., Mowery, Sampat, & Ziedonis, 2002; Rosen-kopf & Almeida, 2003; Stuart, 1998), we also com-puted an alternate measure of knowledge networkcentrality using U.S. National Bureau of EconomicResearch (NBER) citation data (Hall, Jaffe, & Tra-jtenberg, 2005). From the full patent dataset, weextracted patents that member firms owned (match-ing on the assignee names) and then filtered thesefurther using technological categories relevant tothe industry.27 We then used a five-year moving
24 A research assistant with considerable technicalknowledge about computer hardware and data storagecarried out the coding. It was then repeated by one of theco-authors for a random subsample of 10% of the mem-ber firm alliances. Inter-rater agreement was 88.5%, withCronbach’s alpha of .81 and Cohen’s kappa of .75.
25 We encountered 64 “hybrid” member firm alliancesthat had some elements of knowledge generation andcommercialization—our results are robust to the inclu-sion of these alliances in the regression models.
26 Since alliances are bilateral arrangements, we treatedthe networks as undirected.
27 These included the HJT (Hall, Jaffe, and Trajtenberg)category 2 (Computers and Communications), with sub-categories Communications (21), Computer hardwareand software (22), Computer peripherals (23), Informa-tion storage (24), Electronic business methods and soft-ware (25), and the HJT category 4 (Electrical and Elec-tronic), with subcategories Electrical devices (41),Measuring and testing (43), Power systems (45), semicon-ductor Devices (46), and Miscellaneous electrical/elec-tronic (49).
2014 525Ranganathan and Rosenkopf
window of cross citations between these patents(Katila & Ahuja, 2002; Phelps, 2010; Stuart, 1998)to define the citation ties and calculated an in-degreecentrality measure for every member firm.28,29
Control variables. Fixed effects are used to con-trol for unobserved heterogeneity across both firmsand ballots. We also use a number of variables tocontrol for observable time-varying factors that mayinfluence firms’ strategic behavior with respect tospecific ballots, subcommittees, or years. For bal-lot-specific effects, “Participation in working groupassociated with ballot” controls for the stake that afirm has in a particular ballot. Since ballots areheterogeneous with regard to the technical issuesbeing debated, some firms may have higher stakesin particular ballots. To construct this variable, wefirst assembled all the standards-related documentfiles from the committees’ archival records. Basedon the document titles of these files, we identifiedthe subset of 2,185 documents that constitutedworking group meeting minutes, which mapped onto 63 distinct working groups. We automated theparsing of each document to estimate the extent towhich the firm was involved in each working groupmeeting. We used two different measures—count-ing the overall number of occurrences of the firm’sname (and possible variants) in these documentsand counting the number of firm representativesthat attended the meeting (by parsing only the at-tendance section of the minutes documents).30
Since the standards development process was se-quential and path dependent, and because someworking groups had multiple ballots over time, weused a cumulative stock of this measure.
We included “Patents on standard’s technolo-gies” to account for the differences in standard-specific technological competencies across firms.Since some firms’ technological competencies maymap more closely to the emerging standard, thesefirms may also have greater stakes in its ballotmeasures. To calculate this measure, we compileda set of technological keywords specific to the stan-dards subcommittees, deriving these from the pub-lished standards and from reading through sub-
committee charters.31 Then, using the DerwentWorld Patents Index32 (e.g., Pavitt, 1985), we com-piled the set of patents for which the descriptioncontained one or more of these keywords. Finally,we matched the names of the patent assignees withthe member firm sample and calculated a patentstock for each member firm.33
In each year, we accounted for the breadth,depth, importance, and insularity of a firm’s tech-nological investments and its financial position.These variables are “Patent stock (overall),” “Patentstock diversity,” “Citations to patent stock,” and“Knowledge insularity”—and their construction isdescribed in greater detail in Table 2. As firms withmore liberal opportunities may be less resistant tothe standard than firms that occupy competitiveniches (e.g., Ahuja, 2000; Stuart, 1998), we alsoinclude “Overall sector size (assets),” using a sumof the assets of all listed firms in the firm’s SICcode. We include “Size,” “Financial slack,” and“Financial performance” to control for effects ofperformance and financial resources. Finally, wealso control for “External alliances (to firms not onstandards committee),” as these might influencefirms’ voting patterns (Gomes-Casseres, 2003; Lei-ponen, 2008).
Estimation
Level of analysis. Our analysis is at the “firm-vote level”—in other words, we pooled all firms’votes across all ballot measures that the firms votedon. We chose the firm-vote level rather than thefirm-year level because the technical agenda under-lying ballot measures varies over time and oversubcommittees—thus, each measure solicits firms’votes on a different subset of technical issues withrespect to the standard.34
28 We used the patent application year for the timewindow since this reflects the establishment of the tie.
29 Since patent citations are unidirectional, we treatedthe networks as directed.
30 In the results shown, we used the latter measure—the results are robust to the use of the first measureas well.
31 For example, keywords for T13 subcommittee in-cluded “Advanced Technology Attachment,” “ATAPI,”and “Serial ATA.”
32 The Derwent database allows searching patents bytechnological keywords—a facility not available with theNBER patent data.
33 We also included a control for the number of “Pro-posals”—this variable dropped out in our regressions.
34 Were we to aggregate the votes up to the firm-yearlevel, we would be making the unrealistic assumptionthat identical issues are deliberated across all ballots.Also, notwithstanding the fact that our ultimate interestis in firms’ overall strategic behavior, rather than behav-ior on specific technical issues, modeling at the firm-vote
526 AprilAcademy of Management Journal
Model. We model the firm’s vote on a ballot as:
Yijk � f�Xij� � Zijk1 � � Zij
2� � Zi3� � ui � vj � wk� � �ijk
where “i” indexes the firm, “j” indexes the subcom-mittee, and “k” indexes the ballot measure. Y is afirm’s vote on a ballot measure in a subcommittee andX is a vector of firm covariates specific to the sub-committee (the positions in knowledge and com-mercialization networks, as well as the interactionbetween them). Z1, Z2, and Z3 are vectors of firmcovariates specific to both the subcommittee andballot (1), specific only to the subcommittee (2),and independent of subcommittee and ballot (3). ui,vj, and wk are unobserved firm, subcommittee, andballot effects, respectively. ijk is a random distur-bance term.
Our estimation uses an ordered logistic regressionmodel (the “ologit” command in Stata, the data anal-ysis and statistical software program), which assumesthe ijk to be independent and identically distributedrandom variables that follow an extreme value distri-bution (i.e., a logistic distribution function).
Unobserved heterogeneity. As is the case withmost non-experimental studies, one concern is thatour results could be biased because of unobservedheterogeneity. We recognize and attempt to controlfor two kinds of unobserved heterogeneity. The firstkind arises because some firms may be more predis-posed than others towards opposing or accepting thestandard because of reasons that are not observable ormeasurable.35 To alleviate this concern, we include“firm fixed effects” (a dummy variable for each firm)in all our main models.36 The second kind of unob-served heterogeneity arises because some ballot mea-sures may result in greater opposition from firms thanothers, due to factors that are independent of ourhypothesized predictors or included controls. To al-leviate this concern, we also include “ballot fixedeffects” (a dummy variable for each ballot measure).37
Finally, we also lag the independent variables andcontrols by one year to mitigate the risk of simulta-neity or reverse causality, and include “year fixedeffects.”38
RESULTS
Table 3 lists the sample statistics and correla-tions (all independent variables are mean cen-tered). The high correlations between the two alli-ance centrality measures are not surprising, as wewould expect firms active in one alliance sphere toalso be active in the other.39 The correlation be-tween the “Knowledge network centrality (patentcitations)” measure and the “Commercializationnetwork centrality” measure is, however, moder-ate. Therefore, one additional advantage of this al-ternate measure is that we can alleviate collinearityconcerns if we obtain consistent results with bothsets of measures.
Table 4 displays the results of the main regres-sions used to test the hypotheses. In Models 1–5,we use the alliance-based measure for “Knowledgenetwork centrality,” and, in Models 6 and 7, we usethe citation-based measure. Model 1 is a controls-only model. Models 2–4 show the results from thestepwise addition of independent variables to thecontrols-only model. All models include firm fixedeffects, and we add ballot fixed effects in Model 5and Model 7. The improvements in log-likelihoodacross successive models are significant (likelihoodratio test), and show that the stepwise addition ofvariables improves model fit.
A negative coefficient for a variable in these re-gressions indicates support for the standard, whilea positive coefficient indicates opposition. Hypoth-esis 1 stated that the more central a focal firm iswithin the knowledge network, the lower will be itsopposition to the standard. The coefficient for thevariable “Knowledge network centrality” is nega-tive and strongly significant in Models 2–6 and
level does allow us the flexibility to control for thesedifferences across ballots (as we have described in thepreceding section on control variables).
35 For instance, some firms may adopt a more openstrategy of knowledge sharing and cross-firm collabora-tion—such strategies may be correlated with central net-work positions and a higher support for open standards.
36 Firm fixed effects models also automatically in-clude time invariant effects specific to the sector orindustry.
37 By including ballot fixed effects, subcommitteefixed effects are automatically included (to control forunobserved heterogeneity in voting behavior across T10,
T11, and T13). We confirmed this in our regressionswhen the dummy variables for the subcommittees droppedout of the regressions.
38 As network scholars have suggested, one way toalleviate concerns around the endogeneity of networkmeasures is to use time-varying data that allow the use oflag structures and the incorporation of fixed effects in theregression models (Stuart & Sorenson, 2007).
39 Variance inflation factors calculated after a pooledlinear regression were below the threshold of 10 (e.g.,Davidson, Jiraporn, Kim, & Nemec, 2004).
2014 527Ranganathan and Rosenkopf
TA
BL
E3
Des
crip
tive
Sta
tist
ics
and
Cor
rela
tion
s
Var
iabl
esM
SD
12
34
56
78
910
1112
1314
1516
1718
19
1F
irm
’svo
teon
ball
ot0.
440.
811
2K
now
led
gen
etw
ork
cen
tral
ity
(R&
Dal
lian
ces)
00.
530.
091
3C
omm
erci
aliz
atio
nn
etw
ork
cen
tral
ity
00.
760.
110.
841
4K
now
led
gen
et.
cen
tral
ity
(R&
Dal
lian
ces)
�C
omm
erci
aliz
atio
nn
etw
ork
cen
tral
ity
04.
630.
060.
860.
791
5K
now
led
gen
etw
ork
cen
tral
ity
(pat
ent
cita
tion
s)0
0.27
00.
470.
490.
291
6K
now
led
gen
et.
cen
tral
ity
(pat
ent
cita
tion
s)�
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
y0
0.64
0.09
0.86
0.97
0.84
0.55
1
7P
arti
cip
atio
nin
wor
kin
ggr
oup
asso
ciat
edw
ith
ball
ot0
3.79
0.18
0.2
0.23
0.2
0.26
0.25
18
Pat
ents
onst
and
ard
’ste
chn
olog
ies
01.
990.
080.
450.
530.
340.
50.
530.
211
9P
aten
tst
ock
(ove
rall
)0
25.3
10.
060.
610.
610.
460.
710.
660.
150.
411
10P
aten
tst
ock
div
ersi
ty0
2.76
00.
460.
450.
260.
820.
480.
110.
420.
821
11C
itat
ion
sto
pat
ent
stoc
k0
85.1
40.
020.
540.
630.
440.
790.
690.
150.
580.
840.
721
12K
now
led
gein
sula
rity
035
.90
0.05
0.6
0.7
0.5
0.72
0.75
0.16
0.56
0.88
0.69
0.93
113
Ext
ern
alal
lian
ces
(to
firm
sn
oton
stan
dar
ds
com
mit
tee)
02.
400.
110.
710.
750.
660.
370.
740.
160.
30.
60.
380.
530.
611
14S
ize
(ass
ets)
02.
20�
.01
0.43
0.49
0.27
0.7
0.49
0.05
0.43
0.61
0.7
0.67
0.62
0.34
115
Siz
e(r
even
ues
)0
2.24
�.0
20.
430.
490.
280.
740.
50.
070.
430.
610.
720.
690.
640.
350.
961
16F
inan
cial
slac
k(c
ash
)0
3.23
0.01
0.51
0.56
0.41
0.5
0.58
0.07
0.56
0.45
0.48
0.6
0.57
0.36
0.68
0.67
117
Fin
anci
alp
erfo
rman
ce(n
etin
com
e)0
3.48
0.02
0.42
0.42
0.37
0.2
0.4
0.08
0.18
0.29
0.19
0.26
0.31
0.42
0.22
0.23
0.27
118
Fin
anci
alsl
ack
(deb
t)0
0.92
�.0
30.
30.
360.
250.
520.
420.
050.
480.
510.
510.
670.
580.
20.
750.
720.
670.
151
19S
ecto
rsi
ze0
1.16
0.03
0.31
0.31
0.19
0.37
0.32
0.08
0.33
0.43
0.48
0.43
0.38
0.25
0.47
0.43
0.45
0.14
0.44
1
TA
BL
E4
Ord
ered
Log
isti
cR
egre
ssio
nR
esu
lts
Var
iabl
es
Mod
els
12
34
56
7
Ind
epen
den
tva
riab
les
Kn
owle
dge
net
wor
kce
ntr
alit
y(R
&D
alli
ance
s)�
0.41
***
(0.1
1)�
0.71
***
(0.1
4)�
0.42
**(0
.18)
�0.
53**
*(0
.19)
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
y0.
35**
*(0
.10)
0.47
***
(0.1
1)0.
62**
*(0
.12)
0.85
***
(0.2
6)0.
85**
*(0
.27)
Kn
owle
dge
net
wor
kce
ntr
alit
y(R
&D
)�
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
y�
0.04
**(0
.02)
�0.
04**
(0.0
2)
Kn
owle
dge
net
wor
kce
ntr
alit
y(P
aten
tci
tati
ons)
�0.
86**
(0.3
7)�
0.76
*(0
.45)
Kn
owle
dge
net
wor
kce
ntr
alit
y(P
aten
tci
tati
ons)
�C
omm
erci
aliz
atio
nn
etw
ork
cen
tral
ity
�0.
97**
*(0
.28)
�0.
86**
*(0
.30)
Con
trol
sP
arti
cip
atio
nin
wor
kin
ggr
oup
asso
ciat
edw
ith
ball
ot0.
07**
*(0
.01)
0.07
***
(0.0
1)0.
07**
*(0
.01)
0.07
***
(0.0
1)0.
13**
*(0
.01)
0.08
***
(0.0
1)0.
13**
*(0
.01)
Pat
ents
onst
and
ard
’ste
chn
olog
ies
0.05
*(0
.03)
0.07
**(0
.03)
0.07
**(0
.03)
0.07
**(0
.03)
0.07
**(0
.03)
0.05
(0.0
3)0.
05(0
.03)
Pat
ent
stoc
k(o
vera
ll)
0.01
(0.0
1)0.
01(0
.01)
0.02
*(0
.01)
0.02
*(0
.01)
0.02
**(0
.01)
0.01
(0.0
1)0.
02*
(0.0
1)P
aten
tst
ock
div
ersi
ty�
0.11
(0.0
7)�
0.10
(0.0
7)�
0.08
(0.0
7)�
0.09
(0.0
7)�
0.07
(0.0
7)�
0.10
(0.0
7)�
0.08
(0.0
7)C
itat
ion
sto
pat
ent
stoc
k0.
01**
*(0
.00)
0.01
***
(0.0
0)0.
01**
(0.0
0)0.
01**
*(0
.00)
0.01
***
(0.0
0)0.
01**
*(0
.00)
0.01
***
(0.0
0)K
now
led
gein
sula
rity
�0.
01**
(0.0
1)�
0.02
***
(0.0
1)�
0.02
***
(0.0
1)�
0.02
***
(0.0
1)�
0.03
***
(0.0
1)�
0.02
**(0
.01)
�0.
02**
*(0
.01)
Ext
ern
alal
lian
ces
(to
firm
sn
oton
stan
dar
ds
com
mit
tee)
�0.
10**
*(0
.03)
�0.
08**
(0.0
3)�
0.09
***
(0.0
3)�
0.09
***
(0.0
3)�
0.12
***
(0.0
3)�
0.11
***
(0.0
3)�
0.13
***
(0.0
3)S
ize
(ass
ets)
0.25
***
(0.0
8)0.
23**
*(0
.08)
0.22
***
(0.0
8)0.
22**
*(0
.08)
0.21
**(0
.09)
0.24
***
(0.0
8)0.
24**
*(0
.09)
Siz
e(r
even
ues
)�
0.07
(0.0
9)�
0.08
(0.0
9)�
0.11
(0.0
9)�
0.12
(0.0
9)�
0.17
*(0
.10)
�0.
16(0
.10)
�0.
20**
(0.1
0)F
inan
cial
slac
k(c
ash
)�
0.01
(0.0
2)0.
01(0
.02)
0.01
(0.0
2)0.
02(0
.02)
0.01
(0.0
2)�
0.01
(0.0
2)�
0.01
(0.0
2)F
inan
cial
per
form
ance
(net
inco
me)
�0.
00(0
.01)
�0.
00(0
.01)
�0.
01(0
.01)
�0.
00(0
.01)
�0.
00(0
.01)
�0.
00(0
.01)
�0.
00(0
.01)
Fin
anci
alsl
ack
(deb
t)�
0.12
(0.1
0)�
0.13
(0.1
0)�
0.12
(0.1
0)�
0.14
(0.1
0)�
0.12
(0.1
0)�
0.12
(0.1
0)�
0.12
(0.1
0)S
ecto
rsi
ze0.
04(0
.14)
0.08
(0.1
4)0.
10(0
.14)
0.11
(0.1
4)0.
06(0
.15)
0.17
(0.1
4)0.
09(0
.15)
cut
1C
onst
ant
2.72
***
(0.6
1)2.
67**
*(0
.61)
2.55
***
(0.6
1)2.
42**
*(0
.62)
4.14
***
(0.8
1)2.
32**
*(0
.62)
4.06
***
(0.8
2)cu
t2
Con
stan
t4.
08**
*(0
.61)
4.03
***
(0.6
1)3.
91**
*(0
.62)
3.78
***
(0.6
2)5.
64**
*(0
.81)
3.69
***
(0.6
2)5.
55**
*(0
.82)
cut
3C
onst
ant
5.03
***
(0.6
2)4.
98**
*(0
.62)
4.86
***
(0.6
2)4.
74**
*(0
.62)
6.67
***
(0.8
1)4.
64**
*(0
.62)
6.58
***
(0.8
2)O
bser
vati
ons
9,12
09,
120
9,12
09,
120
9,12
09,
120
9,12
0F
irm
s13
513
513
513
513
513
513
5B
allo
ts24
124
124
124
124
124
124
1F
irm
fixe
def
fect
sY
esY
esY
esY
esY
esY
esY
esB
allo
tfi
xed
effe
cts
No
No
No
No
Yes
No
Yes
Yea
ref
fect
sY
esY
esY
esY
esY
esY
esY
esIn
du
stry
effe
cts
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Log
like
lih
ood
�71
39�
7132
�71
27�
7123
�66
80�
7128
�66
91C
hi-
squ
are
1504
1518
1529
1536
2421
1527
2400
Pse
ud
o-R
20.
0953
0.09
620.
0969
0.09
730.
153
0.09
670.
152
Not
es.
Dep
end
ent
vari
able
isfi
rm’s
vote
ona
ball
otm
easu
re,
cod
edas
0�
yes,
1�
abst
ain
,2
�ye
sw
ith
com
men
ts/c
ond
itio
ns,
3�
no.
All
ind
epen
den
tva
riab
les
are
lagg
edby
one
year
.F
irm
vote
sar
ep
oole
dac
ross
all
year
s.M
odel
s1–
5u
seal
lian
ced
ata
toco
nst
ruct
the
know
led
gen
etw
ork;
mod
els
6an
d7
use
pat
ent
cita
tion
dat
a.S
tan
dar
der
rors
inp
aren
thes
es.
*p
�.1
0**
p�
.05
***
p�
.01
moderately significant in Model 7, indicating sup-port for Hypothesis 1. A one-unit increase in“Knowledge network centrality” (approximatelytwo standard deviations above the mean) decreasesthe odds of not voting “yes” by 0.41 times. Hypoth-esis 2 stated that the more central a focal firm iswithin the commercialization network, the higherits opposition to the standard. The coefficient forthe variable “Commercialization network central-ity” is positive and strongly significant in all themodels, indicating support for Hypothesis 2. Aone-unit increase in “Commercialization networkcentrality” (approximately 1.3 standard deviationsabove the mean) increases the odds of not voting“yes” by 1.85 times. Finally, Hypothesis 3 exploresthe joint effect of the two network positions—itstated that “Knowledge network centrality” wouldhave a moderating effect on “Commercializationnetwork centrality,” such that the more technolog-ically peripheral a focal firm is, the greater theeffect of its commercialization network position on
its opposition to the standard. The coefficient forthe interaction term “Knowledge network cen-trality � Commercialization network centrality”is negative and significant, indicating support forHypothesis 3. Figure 1 further examines this re-lationship graphically. For the same “Commer-cialization network centrality” score, the likeli-hood of a firm not voting “yes” on the ballot islower, and the higher its “Knowledge networkcentrality” score.
To examine the implications of the interplay be-tween the two network positions in more detail,note that our arguments were motivated by thelogic that the effect of network multiplicity onfirms’ strategic behavior would be more prominentwhen the effects of the firm’s knowledge networkposition align with those of its commercializationnetwork position. In other words, when the firm iscentral in one network but peripheral in the other,we would expect a clear and unambiguous effect on
FIGURE 1Graph of Dependent Variable (Y-Axis) vs. Commercialization Network Centrality (X-Axis) for Different
Levels of Knowledge Network Centrality
A
B
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Pro
babi
lity
(Y
>0)
Commercialization network centrality
Low Knowledge network centrality
Moderate Knowledge network centrality
High Knowledge network centrality
-1.52 -1.14 -0.76 -0.38 0 0.38 0.76 1.14 1.52
Notes. Low knowledge network centrality � mean �1 SD; moderate knowledge network centrality � mean; high knowledge networkcentrality � mean �1 SD. Y-axis measures the probability that the dependent variable is greater than 0 (i.e., vote is not an unconditional“yes”). Coefficient estimates are based on Model 5 in Table 4.
530 AprilAcademy of Management Journal
its voting pattern. Table 5 demonstrates empiricalevidence for this logic.
We subdivided the firms into four quadrantsbased on their two network positions (Low–Lowcentrality, Low–High centrality, High–Low central-ity, and High–High centrality)40 and included anindicator variable for each quadrant in the regres-sion. Models 2, 3, and 5 demonstrate support forour logic—the coefficient for “Low knowledge cen-trality AND High commercialization centrality” ispositive and significant, while the coefficient for“High knowledge centrality AND Low commercial-ization centrality” is negative and significant. Fur-ther bolstering our arguments, we do not see anyeffects for firms in the Low–Low quadrant or theHigh–High quadrant—since the effects of the twonetwork positions are in opposition for these firms,an overall effect is therefore not discernible.
Figure 1 further illustrates this dynamic—thelikelihood that a firm does not vote “yes” is approx-imately three times higher if a firm is central in thecommercialization network and peripheral in theknowledge network (point “A” on the graph) than ifit is peripheral in the commercialization networkand central in the knowledge network (point “B”on the graph).41
Alternative Explanations
One additional concern regarding unobservedheterogeneity could be that a particular firm’s vot-ing choice on a particular ballot may be affected bythe relationship between the set of technologicalissues being decided upon in the ballot and thefirm’s technological orientation. For instance, if theballot’s technological agenda is distant from or notrelevant to the firm’s technological area, then thefirm may be less inclined to vote against it. Unfor-tunately, to directly assess this concern empiricallyrequires not only advanced technical skills in thecomputer and data storage area, but also a dispro-portionate effort of reading through several thou-sand patents and several thousand working groupmeeting minutes to devise a measure. Moreover,even with true expertise in this area, the measure
would still be subject to interpretation. However,our models included two proxy measures—“Partic-ipation in working group associated with ballot”and “Patents on standard’s technologies”—thatcapture at least some portion of this heterogeneity,and reflect particular firms’ stakes in particular bal-lots and subcommittees. Table 6 displays the re-sults of a post hoc analysis, which includes theinteractions of these measures with our indepen-dent variables.
In Model 1, we interact “Participation in workinggroup associated with ballot” with both networkcentrality variables, and, in Model 2, we interact“Patents on standard’s technologies” with thesevariables. Importantly, our main effects are robustto this inclusion. Further, the results suggest thatfirms indicating a high interest (stake) in the ballotvia either of these two measures will decrease theirexpected levels of opposition when they are morecentral in the knowledge network. This supportsour theorizing, in Hypothesis 1, that centrality inthe knowledge network reflects both control andacceptance of the technological change imposed bythe standard. In contrast, the level of commercial-ization network centrality is either not significantor slightly positive on the level of expected oppo-sition. This supports our theorizing in Hypothesis 2that argues for the effect of a commercializationposition that is independent of technological con-siderations of the standard.
Robustness Checks
We also carried out an extensive set of robustnesschecks to test the sensitivity of our assumptionsregarding model specification, variable construc-tion, and choice of measures. Several of these re-sults are reported in the appendix (Table A1), andwe summarize them here. In Model A1, we use analternate ordinal ranking for the dependent vari-able (0 � yes, 1 � yes with comments/conditions,2 � abstain, and 3 � no). In Models A2–A4, we usean ordinary least squares model, a logistic specifi-cation (without rank ordering the dependent vari-able; i.e., 0 � yes, 1 � everything else), and anordered probit specification, respectively. InModel A5, we use Bonacich’s eigenvector central-ity (e.g., Gulati & Gargiulo, 1999) instead of degreecentrality. We also carried out tests without theinclusion of firms disconnected from thetwo net-works, with the inclusion of 64 hybrid ties and bylimiting the study period to 2007, 2006, and 2005,
40 The mean of each centrality measure was used todemarcate the boundaries of the quadrants.
41“Central” is represented in the graph (Figure 1) bythe firm for which the network centrality score is mean�1 standard deviation. “Peripheral” is represented in thegraph by the firm for which the network centrality scoreis mean �1 standard deviation.
2014 531Ranganathan and Rosenkopf
TA
BL
E5
An
alyz
ing
the
Inte
ract
ion
betw
een
Kn
owle
dge
Net
wor
kC
entr
alit
y(A
llia
nce
s)an
dC
omm
erci
aliz
atio
nN
etw
ork
Cen
tral
ity:
Ord
ered
Log
isti
cR
egre
ssio
nR
esu
lts
Var
iabl
es
Mod
els
12
34
5
Ind
epen
den
tva
riab
les
Low
know
led
gece
ntr
alit
yA
ND
Low
com
mer
cial
izat
ion
cen
tral
ity
�0.
00(0
.09)
Low
know
led
gece
ntr
alit
yA
ND
Hig
hco
mm
erci
aliz
atio
nce
ntr
alit
y0.
18*
(0.1
0)0.
21*
(0.1
2)H
igh
know
led
gece
ntr
alit
yA
ND
Low
com
mer
cial
izat
ion
cen
tral
ity
�0.
25**
(0.1
1)�
0.20
*(0
.11)
Hig
hkn
owle
dge
cen
tral
ity
AN
DH
igh
com
mer
cial
izat
ion
cen
tral
ity
0.07
(0.1
2)0.
14(0
.14)
Con
trol
sP
arti
cip
atio
nin
wor
kin
ggr
oup
asso
ciat
edw
ith
ball
ot0.
13**
*(0
.01)
0.13
***
(0.0
1)0.
13**
*(0
.01)
0.13
***
(0.0
1)0.
13**
*(0
.01)
Pat
ents
onst
and
ard
’ste
chn
olog
ies
0.05
*(0
.03)
0.05
*(0
.03)
0.04
(0.0
3)0.
05*
(0.0
3)0.
05(0
.03)
Pat
ent
stoc
k(o
vera
ll)
0.02
*(0
.01)
0.02
*(0
.01)
0.02
(0.0
1)0.
02*
(0.0
1)0.
02(0
.01)
Pat
ent
stoc
kd
iver
sity
�0.
09(0
.07)
�0.
09(0
.07)
�0.
08(0
.07)
�0.
10(0
.07)
�0.
08(0
.07)
Cit
atio
ns
top
aten
tst
ock
0.01
***
(0.0
0)0.
01**
*(0
.00)
0.01
***
(0.0
0)0.
01**
*(0
.00)
0.01
***
(0.0
0)K
now
led
gein
sula
rity
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
Ext
ern
alal
lian
ces
(to
firm
sn
oton
stan
dar
ds
com
mit
tee)
�0.
11**
*(0
.03)
�0.
12**
*(0
.03)
�0.
11**
*(0
.03)
�0.
11**
*(0
.03)
�0.
12**
*(0
.03)
Siz
e(a
sset
s)0.
25**
*(0
.09)
0.23
***
(0.0
9)0.
23**
*(0
.09)
0.25
***
(0.0
9)0.
21**
(0.0
9)S
ize
(rev
enu
es)
�0.
11(0
.10)
�0.
11(0
.10)
�0.
11(0
.10)
�0.
12(0
.10)
�0.
12(0
.10)
Fin
anci
alsl
ack
(cas
h)
�0.
02(0
.02)
�0.
02(0
.02)
�0.
02(0
.02)
�0.
02(0
.02)
�0.
02(0
.02)
Fin
anci
alp
erfo
rman
ce(n
etin
com
e)�
0.00
(0.0
1)�
0.00
(0.0
1)�
0.00
(0.0
1)�
0.00
(0.0
1)�
0.00
(0.0
1)F
inan
cial
slac
k(d
ebt)
�0.
12(0
.10)
�0.
11(0
.10)
�0.
12(0
.10)
�0.
12(0
.10)
�0.
11(0
.10)
Sec
tor
size
�0.
02(0
.15)
�0.
01(0
.15)
0.04
(0.1
5)�
0.01
(0.1
5)0.
06(0
.15)
cut1
Con
stan
t4.
50**
*(0
.81)
4.48
***
(0.8
1)4.
47**
*(0
.81)
4.49
***
(0.8
1)4.
43**
*(0
.81)
cut2
Con
stan
t5.
99**
*(0
.82)
5.97
***
(0.8
1)5.
96**
*(0
.81)
5.98
***
(0.8
1)5.
93**
*(0
.81)
cut3
Con
stan
t7.
02**
*(0
.82)
7.00
***
(0.8
1)7.
00**
*(0
.81)
7.01
***
(0.8
1)6.
96**
*(0
.81)
Obs
erva
tion
s9,
120
9,12
09,
120
9,12
09,
120
Fir
ms
135
135
135
135
135
Bal
lots
241
241
241
241
241
Fir
mfi
xed
effe
cts
Yes
Yes
Yes
Yes
Yes
Bal
lot
fixe
def
fect
sY
esY
esY
esY
esY
esY
ear
effe
cts
Yes
Yes
Yes
Yes
Yes
Ind
ust
ryef
fect
sY
esY
esY
esY
esY
esL
ogli
keli
hoo
d�
6699
�66
97�
6696
�66
99�
6694
Ch
i-sq
uar
e23
8523
8823
9023
8523
94
Not
es.
Dep
end
ent
vari
able
isfi
rm’s
vote
ona
ball
otm
easu
re,
cod
edas
0�
yes,
1�
abst
ain
,2
�ye
sw
ith
com
men
ts/c
ond
itio
ns,
3�
no.
All
ind
epen
den
tva
riab
les
are
lagg
edby
one
year
.F
irm
-vot
esar
ep
oole
dac
ross
all
year
s.A
llm
odel
sin
clu
de
firm
fixe
def
fect
s,ba
llot
fixe
def
fect
s,an
dye
aref
fect
s.S
tan
dar
der
rors
inp
aren
thes
es.
*p�
.10
**p
�.0
5**
*p�
.01
respectively, to assess sensitivity to U.S. NBER pat-ent data right-censoring42 (results not shown). Ourresults were robust across these models.43
To empirically underscore the importance of in-cluding network multiplicity, we also ran the re-gressions with a composite measure of centrality(Model A6) where we constructed a single alliancenetwork by pooling both types of ties. Not surpris-ingly, the coefficient of the composite alliance net-work centrality measure was insignificant, as we
42 NBER patent data is available only until 2006.43 Additionally, we reduced collinearity between the two
independent variables by orthogonalizing them using theGram–Schmidt procedure (e.g., Sine, Mitsuhashi, & Kirsch,2006) (the “orthog” command in Stata software). This removesthe common variance between the two variables and trans-forms them into uncorrelated measures. The effects for the
orthogonalized centrality measures are consistent in compa-rable regression models (results available from authors).
TABLE 6Analyzing the Interaction between Firm’s Network Positions and Its Stake in a Ballot: Ordered Logistic Regression
Results
Variables
Models
1 2
Independent variablesKnowledge network centrality (R&D alliances) �0.65*** (0.16) �0.51*** (0.18)Commercialization network centrality 0.50*** (0.12) 0.40*** (0.13)Participation in working group � Knowledge network centrality �0.04** (0.02)Participation in working group � Commercialization network centrality 0.01 (0.01)Patents on standard’s technologies � Knowledge network centrality �0.12*** (0.04)Patents on standard’s technologies � Commercialization network centrality 0.06* (0.03)
ControlsParticipation in working group associated with ballot 0.14*** (0.01) 0.13*** (0.01)Patents on standard’s technologies 0.07** (0.03) 0.07** (0.03)Patent stock (overall) 0.02** (0.01) 0.03*** (0.01)Patent stock diversity �0.07 (0.07) �0.07 (0.07)Citations to patent stock 0.01*** (0.00) 0.01** (0.00)Knowledge insularity �0.03*** (0.01) �0.03*** (0.01)External alliances (to firms not on standards committee) �0.11*** (0.03) �0.10*** (0.03)Size (assets) 0.21** (0.09) 0.19** (0.09)Size (revenues) �0.17* (0.10) �0.16* (0.10)Financial slack (cash) 0.01 (0.02) 0.02 (0.02)Financial performance (net income) �0.00 (0.01) �0.00 (0.01)Financial slack (debt) �0.11 (0.10) �0.14 (0.10)Sector size 0.05 (0.15) 0.02 (0.15)cut1 Constant 4.17*** (0.81) 4.07*** (0.82)cut2 Constant 5.66*** (0.81) 5.57*** (0.82)cut3 Constant 6.70*** (0.81) 6.60*** (0.82)Observations 9,120 9,120Firms 135 135Ballots 241 241Firm fixed effects Yes YesBallot fixed effects Yes YesYear effects Yes YesIndustry effects Yes YesLog likelihood �6679 �6679Chi-square 2425 2425
Notes. Dependent variable is firm’s vote on a ballot measure, coded as 0 � yes, 1 � abstain, 2 � yes with comments/conditions, 3 �no. All independent variables are lagged by one year. Firm votes are pooled across all years in the sample. All models include firm fixedeffects, ballot fixed effects, and year effects.
Standard errors in parentheses.* p � .10
** p � .05*** p � .01
2014 533Ranganathan and Rosenkopf
would expect the individual effects from the un-derlying knowledge and commercialization net-works to offset each other for a large part of thesample. We also conducted a principal componentanalysis of the two alliance network centrality mea-sures, and included the component that accountedfor 90% of the common variation (Model A7). Again,the insignificant coefficient of this component furtherhighlights the necessity of accounting for multiplicityin the alliance construct.
DISCUSSION AND CONCLUSIONS
This study makes several important theoreticaland empirical contributions to research on rela-tional pluralism. The simultaneous considerationof multiple networks highlights the reality that thetypes of ties measured can shape network effectsdifferentially, providing a stark contrast to mostnetwork studies that focus on a single type of tie,ignoring the pluralism embodied by firms (Shipilov& Li, 2012). That is, while prior studies typicallyfocus on the implications of a firm’s position ineither a technological network (e.g., Stuart &Podolny, 1996) or its embeddedness in a relationalnetwork (e.g., Ahuja, 2000; Uzzi, 1997), here weshow that firms’ strategic decisions in the stan-dards-setting arena are shaped by the opposing andinteracting effects of their positions in both knowl-edge and commercialization networks.
Importantly, in our study, the effects of these twonetwork types are neither symmetric nor merelyadditive, underscoring the importance of the func-tion of the network ties themselves. Regarding sym-metry, while knowledge network centrality is associ-ated with favoring new standards, commercializationnetwork centrality is associated with opposing them,thus cautioning us against generalizing effects of cen-tral network positions on behavior without regard tothe specific network (and outcome) under consider-ation. Indeed, our results demonstrate that, whileknowledge networks and central positions withinthese networks promote technological innovationin the form of compatibility standards, commer-cialization networks and central positions withinthis network can decelerate the pace of innova-tion. Regarding additivity, while each networkgenerates significant main effects, their interac-tion demonstrates how the overlap of the twonetworks suggests a firm’s overall disposition to-wards a ballot proposal. Specifically, firms withlow knowledge centrality but high commercial-ization centrality are most likely to cast opposingvotes, because the rents these firms appropriate
are largely a function of their current allianceagreements that are rooted in the current stan-dards. Given the insights derived from examiningmultiple types of networks simultaneously, itwill be critical for future research to extend thisapproach by incorporating other forms of plural-ism, such as the interpersonal networks consti-tuted via mobility or common institutional affil-iations such as graduate degree programs.
While our reported results on knowledge andcommercialization networks clarify that usingmerely a single-mode network to predict conductwould be substantially underspecified, future re-search must further assess both differentiationwithin typical tie classifications as well as com-monality across distinct tie classifications. Mostresearch on the effects of alliance network posi-tion have focused either on R&D ties alone (e.g.,Rosenkopf et al., 2001; Sampson, 2007) or apooled combination of both R&D and commer-cialization ties (e.g., Gulati & Gargiulo, 1999;Koka & Prescott, 2002). Given this common prac-tice of pooling knowledge and commercializationties under the generic heading of alliance ties,two of our findings are particularly important.First, the pooling of these ties in our settingyields no discernible effect on standards votingas these two forces counterbalance each other.Second, the knowledge alliance position is moreeffectively proxied by patent citation positionrather than commercialization alliance position,despite the fact that knowledge generation alli-ance ties are jointly volitional and consideredsymmetric while patent citation ties are asym-metric, more emergent, and less social in nature.Accordingly, future research should address therobustness of tie classifications that our field hastaken for granted. Just as all alliances are not thesame, neither are all director interlocks (e.g., in-side versus outside director ties) and nor are allmobility ties (e.g., to client versus to competitor).
Of course, our findings may be particular to thetype of industry we study and its specific dependentvariables regarding conduct. Obviously, our choice ofa systemic industry in which a great deal of techno-logical competition occurs in the standards forumrather than purely in the market may limit the scopeof our findings. Our dependent variable—oppositionto a standard—is purely a function of this context, yeta clear indicator of a firm’s strategic behavior as itattempts to create competitive roadblocks. Whilemost dependent variables for network position stud-ies have focused on performance outcomes such asinnovation or financial measures, our study focuses
534 AprilAcademy of Management Journal
on a standards-specific behavioral measure (voting)and finds that a single network predictor is insuffi-cient to explain variation in this measure. Futureresearch should seek to extend these ideas both inother industries and for other context-specific depen-dent variables.
This paper also makes an important distinctionspecific to multifirm settings such as standardscommittees: Although firms that are central in analliance network are posited to possess relationalcapital and interfirm trust (e.g., Kale, Dyer, & Singh,2002; Koka & Prescott, 2002; Gulati & Singh, 1998;Gulati, 1995b), the value of accumulated social cap-ital may be limited in a standards-setting commit-tee, particularly when a firm cannot control mem-bership, define the technical agenda for discussion,or cast more than one organizational vote. Withengineers as participants, formal criteria for evalu-ating proposals are primarily based on technicalknowledge and excellence (Rosenkopf et al., 2001).Although the standard itself may be strategic tofirms, the process of developing it has a “bottom-up” flavor (see Rosenkopf et al., 2001). Tactics ofpolitics and influence, including lobbying outside theconfines of formal meetings, may backfire. Thus, afirm’s alliance network position may not readilytranslate into advantageous technical committee de-cisions. Future research may seek to uncover addi-tional arenas in which the typical finding that cen-trality leads to positive effects is refuted.
More generally, this setting is one in which thevalue-creating potential of alliance networks isconstrained by the institutional relationshipsforced by standards committee membership andparticipation. Accordingly, the opportunities to ex-tend this research are considerable. Future researchmight examine, for example, the effects of diverg-ing votes or changing committee participationamongst longstanding alliance partners. In sum-mary, our focus on multiple network positions andtheir effects on strategic behavior in technologicalstandards committees has allowed us to demon-strate how typical findings on the typical associa-tion between centrality and outcome variables canbe dramatically altered by the choice of networktie, industry context, and strategic behaviors understudy. Acknowledging and comparing thesechoices across studies of ties, contexts, and behav-iors will allow for the development of stronger,mid-range theories about how network structureaffects conduct for firms embedded in a host ofinterorganizational relationships.
REFERENCES
Adner, R., & Kapoor, R. 2010. Value creation in innovationecosystems: How the structure of technological inter-dependence affects firm performance in new technol-ogy generations. Strategic Management Journal, 31:306–333.
Afuah, A., & Bahram, N. 1995. The hypercube of innova-tion. Research Policy, 24: 51–76.
Ahuja, G. 2000. Collaboration networks, structural holes,and innovation: A longitudinal study. Administra-tive Science Quarterly, 45: 425–455.
Bae, J., & Gargiulo, M. 2004. Partner substitutability, al-liance network structure, and firm profitability inthe telecommunications industry. Academy of Man-agement Journal, 47: 843–859.
Baldwin, C. Y., & Clark, K. B. 2000. Design rules: Thepower of modularity, vol. 1. Cambridge, MA: MITPress.
Baldwin, C. Y., & Woodard, C. J. 2008. The architecture ofplatforms: A unified view. In A. Gawer (Ed.) Plat-forms, markets, and innovation: 19–25. Chelten-ham, U.K.: Edward Elgar.
Baum, J. A. C., Calabrese, T., & Silverman, B. S. 2000.Don’t go it alone: Alliance network composition andstartups performance in Canadian biotechnology.Strategic Management Journal, 21: 267–294.
Benner, M., & Tushman, M. 2002. Process managementand technological innovation: A longitudinal studyof the photography and paint industries. Adminis-trative Science Quarterly, 47: 676–706.
Bresnahan, T. F., & Greenstein, S. 1999. Technologicalcompetition and the structure of the computer in-dustry. Journal of Industrial Economics, 47: 1–40.
Browning, L. D., Beyer, J. M., & Shelter, J. C. 1995. Buildingcooperation in a competitive industry: SEMATECHand the semiconductor industry. Academy of Man-agement Journal, 38: 113–151.
Chiao, B., Lerner, J., & Tirole, J. 2007. The rules of stan-dard setting organizations: An empirical analysis.RAND Journal of Economics, 38: 905–930.
Christensen, C. M., & Bower, J. L. 1996. Customer power,strategic investment, and the failure of leading firms.Strategic Management Journal, 17: 197–218.
Cyert, R., & March, J. 1963. A behavioral theory of thefirm. Englewood Cliffs, NJ: Prentice Hall.
Davidson, W. N., Jiraporn, P., Kim, Y. S., & Nemec, C.2004. Earnings management following duality-creat-ing successions: Ethnostatistics, impression manage-ment, and agency theory. Academy of ManagementJournal, 47: 267–275.
Denrell, J., Fang, C., & Winter, S. 2003. The economics of
2014 535Ranganathan and Rosenkopf
strategic opportunity. Strategic Management Jour-nal, 24: 977–990.
Dokko, G., Nigam, A., & Rosenkopf, L. 2012. Keepingsteady as she goes: A negotiated order perspective ontechnological evolution. Organization Studies, 33:681–703.
Dokko, G., & Rosenkopf, L. 2010. Social capital for hire?Mobility of technical professionals and firm influ-ence in wireless standard committees. OrganizationScience, 21: 677–695.
Doz, Y. L., Olk, P. M., & Ring, P. S. 2000. Formationprocesses of R&D consortia: Which path to take?Where does it lead? Strategic Management Journal,21: 239–266.
Farrell, J., & Simcoe, T. 2012. Choosing the rules forconsensus standardization. RAND Journal of Eco-nomics, 43: 235–252.
Fleming, L., & Sorenson, O. 2004. Science as a map intechnological search. Strategic Management Jour-nal Special Issue, 25: 909–928.
Garud, R., Jain, S., & Kumaraswamy, A. 2002. Institu-tional entrepreneurship in the sponsorship of com-mon technological standards: The case of Sun Mi-crosystems and Java. Academy of ManagementJournal, 45: 196–214.
Garud, R., & Kumaraswamy, A. 1995. Technological andorganizational designs to achieve economies of sub-stitution. Strategic Management Journal, 16: 93–110.
Gomes-Casseres, B. 1994. Group versus group: How alli-ance networks compete. Harvard Business Review,72: 62–66.
Gomes-Casseres, B. 2003. Competitive advantage in alli-ance constellations. Strategic Organization, 1: 327–335.
Gulati, R. 1995a. Social structure and alliance formationpatterns: A longitudinal analysis. AdministrativeScience Quarterly, 40: 619–652.
Gulati, R. 1995b. Does familiarity breed trust? The impli-cations of repeated ties for contractual choices.Academy of Management Journal, 35: 85–112.
Gulati, R. 1998. Alliances and networks. Strategic Man-agement Journal, 19: 293–317.
Gulati, R. 1999. Network location and learning: The in-fluence of network resources and firm capabilitieson alliance formation. Strategic Management Jour-nal, 20: 397–420.
Gulati, R., & Gargiulo, M. 1999. Where do interorganiza-tional networks come from? American Journal ofSociology, 104: 1439–1493.
Gulati, R., Nohria, N., & Zaheer, A. 2000. Strategic net-
works. Strategic Management Journal, 21 (SpecialIssue: (Strategic Networks): 203–215.
Gulati, R., & Singh, H. 1998. The architecture of cooper-ation: Managing coordination costs and appropria-tion concerns in strategic alliances. AdministrativeScience Quarterly, 43: 781–814.
Gulati, R., & Westphal, J. D. 1999. Cooperative or control-ling? The effects of CEO–board relations and thecontent of interlocks on the formation of joint ven-tures. Administrative Science Quarterly, 44: 473–506.
Gupta, A. K., Smith, K. G., & Shalley, C. E. 2006. Theinterplay between exploration and exploitation.Academy of Management Journal, 4: 693–706.
Hagedoorn, J., Carayannis, E., & Alexander, J. 2001.Strange bedfellows in the personal computer indus-try: Technology alliances between IBM and Apple.Research Policy, 30: 837–849.
Hall, B. H., Jaffe, A. B., & Trajtenberg, M. 2005. Marketvalue and patent citations. Rand Journal of Econom-ics, 36: 16–38.
Hamel, G., Doz, L., & Prahalad, C. K. 1989. Collaboratewith your competitors—And win. Harvard BusinessReview, 67: 133–139.
Henderson, R., & Clark, K. 1990. Architectural innova-tion: The reconfiguration of existing product tech-nologies and the failure of established firms. Admin-istrative Science Quarterly, 35: 9–30.
Inkpen, A. C., & Tsang, E. W. K. 2005. Social capital,networks, and knowledge transfer. Academy ofManagement Review, 30: 146–165.
International Committee for Information TechnologyStandards (INCITS). 2011. Available from http://www.incits.org.
Kale, P., Dyer, J. H., & Singh, H. 2002. Alliance capability,stock market response, and long-term alliance suc-cess: The role of the alliance function. StrategicManagement Journal, 23: 747–767.
Katila, R., & Ahuja, G. 2002. Something old, somethingnew: A longitudinal study of search behavior andnew product introduction. Academy of Manage-ment Journal, 45: 1183–1194.
Katz, M. L., & Shapiro, C. 1986. Technology adoption inthe presence of network externalities. Journal of Po-litical Economy, 94: 822–841.
Koka, B. R., & Prescott, J. E. 2002. Strategic alliances associal capital: A multidimensional view. StrategicManagement Journal, 23: 795–816.
Koka, B. R., & Prescott, J. E. 2008. Designing alliancenetworks: The influence of network position, envi-ronmental change, and strategy on firm performance.Strategic Management Journal, 29: 639–661.
536 AprilAcademy of Management Journal
Koza, M. P., & Lewin, A. Y. 1998. The coevolution of strategicalliances. Organization Science, 9: 255–264.
Kraatz, M. S., & Moore, J. H. 2002. Executive migrationand institutional change. Academy of ManagementJournal, 45: 120–143.
Lavie, D. 2007. Alliance portfolios and firm performance:A study of value creation and appropriation in theU.S. software industry. Strategic ManagementJournal, 28: 1187–1212.
Lavie, D., Lechner, C., & Singh, H. 2007. The performanceimplications of timing of entry and involvement inmultipartner alliances. Academy of ManagementJournal, 50: 578–604.
Lavie, D., & Rosenkopf, L. 2006. Balancing explorationand exploitation in alliance formation. Academy ofManagement Journal, 49: 797–818.
Leblibici, H., Salancik, G. R., Gopay, A., & King, T. 1991.Institutional change and the transformation of interor-ganizational fields: An organizational history of theU.S. radio broadcasting industry. AdministrativeScience Quarterly, 36: 333–363.
Lee, C., Lee, K., & Pennings, J. M. 2001. Internal capabil-ities, external networks, and performance: A study oftechnology-based ventures. Strategic ManagementJournal, 22: 615–640.
Lee, G. K. 2007. The significance of network resources inthe race to enter emerging product markets: Theconvergence of telephony communications and com-puter networking, 1989–2001. Strategic Manage-ment Journal, 28: 17–37.
Leiponen, A. E. 2008. Competing through cooperation:The organization of standard setting in wireless.Telecommunications – Management Science, 54:1904–1919.
Levinthal, D. 1997. Adaptation on rugged landscapes.Management Science, 43: 934–950.
Lin, Z., Yang, H., & Arya, B. 2009. Alliance partners andfirm performance: Resource complementarity andstatus association. Strategic Management Journal,30: 921–940.
Madhavan, R., Gnyawali, D. R., & He, J. 2004. Two’scompany, three’s a crowd? Triads in cooperative-competitive networks. Academy of ManagementJournal, 47: 918–927.
Madhavan, R., Koka, B. R., & Prescott, J. E. 1998. Net-works in transition: How industry events (re)shapeinterfirm relationships. Strategic ManagementJournal, 19: 439–459.
Miura, H. 2012. Stata graph library for network analysis.STATA Journal, 12: 94–129.
Mowery, D. C., Sampat, B. N., & Ziedonis, A. A., 2002.Learning to patent: Institutional experience, learn-
ing, and the characteristics of U.S. university patentsafter the Bayh–Dole Act, 1981–1992. ManagementScience, 48: 73–89.
Nelson, R., & Winter, S. 1982. An evolutionary theory ofeconomic change. Harvard, MA: Belknap Press.
Owen-Smith, J., & Powell, W. W. 2004. Knowledge net-works in the Boston biotechnology community. Or-ganization Science, 15: 5–21.
Pavitt, K. 1985. Patent statistics as indicators of innova-tive activities: Possibilities and problems. Sciento-metrics, 7: 77–99.
Phelps, C. C. 2010. A longitudinal study of the influenceof alliance network structure and composition onfirm exploratory innovation. Academy of Manage-ment Journal, 53: 890–913.
Podolny, J. M. 2001. Networks as the pipes and prisms of themarket. American Journal of Sociology, 107: 33–60.
Podolny, J. M., & Stuart, T. E. 1995. A role-based ecologyof technological change. American Journal of Soci-ology, 100: 1224–1260.
Polidoro, F., Ahuja, G., Jr., & Mitchell, W. 2011. When thesocial structure overshadows competitive incen-tives: The effects of network embeddedness on jointventure dissolution. Academy of ManagementJournal, 54: 203–223.
Powell, W. W., Koput, K. W., & Smith-Doerr, L. 1996.Interorganizational collaboration and the locus ofinnovation: Networks of learning in biotechnology.Administrative Science Quarterly, 41: 116–145.
Rosenkopf, L., & Almeida, P. 2003. Overcoming localsearch through alliances and mobility. ManagementScience, 49: 751–766.
Rosenkopf, L., Metiu, A., & George, V. 2001. From thebottom up? Technical committee activity and alli-ance formation. Administrative Science Quarterly,46: 748–772.
Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:Boundary spanning, exploration, and impact in theoptical disk industry. Strategic Management Jour-nal, 22: 287–306.
Rosenkopf, L., & Padula, G. 2008. Investigating the mi-crostructure of network evolution: Alliance forma-tion in the mobile communications industry. Organ-ization Science, 19: 669–687.
Rosenkopf, L., & Schilling, M. A. 2007. Comparing alli-ance network structure across industries: Observa-tions and explanations. Strategic EntrepreneurshipJournal, 1: 191–209.
Rothaermel, F. T. 2001. Incumbent’s advantage throughexploiting complementary assets via interfirm coop-eration. Strategic Management Journal Special Is-sue, 22: 687–699.
2014 537Ranganathan and Rosenkopf
Rothaermel, F. T., & Deeds, D. L. 2004. Exploration andexploitation alliances in biotechnology: A system ofnew product development. Strategic ManagementJournal, 25: 201–221.
Rowley, T., Behrens, D., & Krackhardt, D. 2000. Redun-dant governance structures: An analysis of structuraland relational embeddedness in the steel and semi-conductor industries. Strategic Management Jour-nal, 21: 369–386.
Rysman, M., & Simcoe, T. 2008. Patents and the perfor-mance of voluntary standard-setting organizations.Management Science, 54: 1920–1934.
Sampson, R. C. 2007. R&D alliances & firm performance:The impact of technological diversity and allianceorganization on innovation. Academy of Manage-ment Journal, 50: 364–386.
Sanchez, R., & Mahoney, J. T. 1996. Modularity, flexibil-ity, and knowledge management in product and or-ganizational design. Strategic Management Jour-nal, 17: 63–76.
Schilling, M. A. 2002. Technology success and failure inWinner-take-all markets: The impact of learning ori-entation, timing, and network externalities. Acad-emy of Management Journal, 45: 387–398.
Schilling, M. A. 2009. Understanding the alliance data.Strategic Management Journal, 30: 233–260.
Shipilov, A. 2006. Network strategies and performance ofCanadian investment banks. Academy of Manage-ment Journal, 49: 590–604.
Shipilov, A., & Li, S. 2012. The missing link: The effect ofcustomers on the formation of relationships amongproducers in the multiplex triads. Organization Sci-ence, 23: 472–491.
Simcoe, T. 2012. Standard setting committees: Consen-sus governance for shared technology platforms.American Economic Review, 102: 305–336.
Sine, W. D., Mitsuhashi, H., & Kirsch, D. A. 2006. Revis-iting Burns and Stalker: Formal structure and newventure performance in emerging economic sectors.Academy of Management Journal, 49: 121–132.
Singh, K. 1997. The impact of technological complexityand interfirm cooperation on business survival.Academy of Management Journal, 40: 339–367.
Stuart, T. E. 1998. Network positions and propensities tocollaborate: An investigation of strategic alliance for-mation in a high-technology industry. Administra-tive Science Quarterly, 43: 668–698.
Stuart, T. E. 2000. Interorganizational alliances and theperformance of firms: A study of growth and inno-vation rates in a high-technology industry. StrategicManagement Journal, 21: 791–811.
Stuart, T. E., Hoang, H., & Hybels, R. 1999.Interorganiza-
tional endorsements and the performance of entre-preneurial ventures. Administrative Science Quar-terly, 44: 315–349.
Stuart, T. E., & Podolny, J. 1996. Local search and the evolu-tion of technological capabilities. Strategic Manage-ment Journal, Summer Special Issue, 17: 21–38.
Stuart, T. E., & Sorenson, O. 2007. Strategic networks andentrepreneurial ventures. Strategic Entrepreneur-ship Journal, 1: 211–227.
Taylor, A., & Helfat, C. 2009. Organizational linkages forsurviving technological change: Complementary as-sets, middle management, and ambidexterity. Organ-ization Science, 20: 718–739.
Teece, D. J. 1986. Profiting from technological innova-tion: Implications for integration, collaboration, li-censing and public policy. Research Policy, 15:785–805.
Teece, D. J. 2007. Explicating dynamic capabilities: Thenature and microfoundations of (sustainable) enter-prise performance. Strategic Management Journal,28: 1319–1350.
Tripsas, M., & Gavetti, G. 2000. Capabilities, cognition,and inertia: Evidence from digital imaging. StrategicManagement Journal, 21: 1147–1161.
Tushman, M. L., & Anderson, P. 1986. Technologicaldiscontinuities and organizational environments.Administrative Science Quarterly, 31: 439–465.
Uzzi, B. 1997. Social structure and competition in inter-firm networks. Administrative Science Quarterly,42: 35–67.
Walker, G., Kogut, B., & Shan, W. 1997. Social capital,structural holes and the formation of an industrynetwork. Organization Science, 8: 109–125.
Whittington, K. B., Owen-Smith, J., & Powell, W. W.2009. Networks, propinquity, and innovation inknowledge-intensive industries. AdministrativeScience Quarterly, 54: 90–122.
Wu, B., Wan, Z., & Levinthal, D. A. 2013. Complementaryassets as pipes and prisms: Innovation incentivesand trajectory choices. Strategic Management Jour-nal, doi: 10.1002/smj.2159.
Yang, H., Lin, Z., & Lin, Y. 2010. A multilevel frameworkof firm boundaries: Firm characteristics, dyadic dif-ferences, and network attributes. Strategic Manage-ment Journal, 31: 237–261.
Yang, H., Lin, Z., & Peng, M. 2011. Behind acquisitions ofalliance partners: Exploratory learning and networkembeddedness. Academy of Management Journal,54: 1069–1080.
Zaheer, A., & Bell, G. 2005. Benefiting from network position:Firm capabilities, structural holes, and performance.Strategic Management Journal, 26: 809–826.
538 AprilAcademy of Management Journal
AP
PE
ND
IX
TA
BL
EA
1
Rob
ust
nes
sC
hec
ks
Var
iabl
es
Mod
elsa
A1
A2
A3
A4
A5
A6
A7
Kn
owle
dge
net
wor
kce
ntr
alit
y(R
&D
alli
ance
s)�
0.54
***
(0.1
9)�
0.15
***
(0.0
6)�
0.49
***
(0.1
9)�
0.30
***
(0.1
1)�
0.87
*(0
.53)
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
y0.
51**
*(0
.12)
0.23
***
(0.0
4)0.
35**
*(0
.12)
0.35
***
(0.0
7)2.
55**
*(0
.70)
Kn
owle
dge
net
wor
kce
ntr
alit
y�
Com
mer
cial
izat
ion
net
wor
kce
ntr
alit
y�
0.03
*(0
.02)
�0.
01**
*(0
.00)
�0.
03*
(0.0
2)�
0.02
**(0
.01)
�4.
08**
(2.0
6)A
llia
nce
cen
tral
ity
(poo
led
alli
ance
s)�
0.02
(0.0
6)A
llia
nce
cen
tral
ity
(pri
nci
pal
com
pon
ent)
�0.
04(0
.04)
Par
tici
pat
ion
inw
orki
ng
grou
pas
soci
ated
wit
hba
llot
0.07
***
(0.0
1)0.
06**
*(0
.00)
0.05
***
(0.0
1)0.
07**
*(0
.01)
0.13
***
(0.0
1)0.
13**
*(0
.01)
0.13
***
(0.0
1)P
aten
tson
stan
dar
d’s
tech
nol
ogie
s0.
06**
(0.0
3)0.
02**
(0.0
1)0.
07**
(0.0
3)0.
04**
(0.0
2)0.
05*
(0.0
3)0.
05*
(0.0
3)0.
06*
(0.0
3)P
aten
tst
ock
(ove
rall
)0.
01(0
.01)
0.01
***
(0.0
0)0.
01(0
.01)
0.01
***
(0.0
1)0.
03**
(0.0
1)0.
02*
(0.0
1)0.
02*
(0.0
1)P
aten
tst
ock
div
ersi
ty�
0.03
(0.0
7)�
0.03
(0.0
2)�
0.05
(0.0
7)�
0.05
(0.0
4)�
0.10
(0.0
7)�
0.09
(0.0
7)�
0.09
(0.0
7)C
itat
ion
sto
pat
ent
stoc
k0.
01**
*(0
.00)
0.00
**(0
.00)
0.01
**(0
.00)
0.00
**(0
.00)
0.01
***
(0.0
0)0.
01**
*(0
.00)
0.01
***
(0.0
0)K
now
led
gein
sula
rity
�0.
02**
*(0
.01)
�0.
01**
*(0
.00)
�0.
02**
*(0
.01)
�0.
02**
*(0
.00)
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
�0.
02**
*(0
.01)
Ext
ern
alal
lian
ces
(to
firm
sn
oton
stan
dar
ds
com
mit
tee)
�0.
07**
(0.0
3)�
0.04
***
(0.0
1)�
0.06
*(0
.03)
�0.
06**
*(0
.02)
�0.
11**
*(0
.03)
�0.
11**
*(0
.03)
�0.
10**
*(0
.03)
Siz
e(a
sset
s)0.
20**
(0.0
9)0.
05*
(0.0
3)0.
28**
*(0
.09)
0.10
**(0
.05)
0.21
**(0
.09)
0.25
***
(0.0
9)0.
25**
*(0
.09)
Siz
e(r
even
ues
)�
0.28
***
(0.1
0)�
0.00
(0.0
3)�
0.23
**(0
.10)
�0.
07(0
.06)
�0.
11(0
.10)
�0.
11(0
.10)
�0.
11(0
.10)
Fin
anci
alsl
ack
(cas
h)
0.01
(0.0
2)0.
01(0
.01)
0.01
(0.0
2)0.
01(0
.01)
�0.
02(0
.02)
�0.
01(0
.02)
�0.
01(0
.02)
Fin
anci
alp
erfo
rman
ce(n
etin
com
e)�
0.01
(0.0
1)�
0.00
(0.0
0)�
0.03
**(0
.01)
�0.
00(0
.00)
�0.
00(0
.01)
�0.
00(0
.01)
�0.
00(0
.01)
Fin
anci
alsl
ack
(deb
t)�
0.13
(0.1
0)�
0.03
(0.0
3)�
0.16
*(0
.10)
�0.
07(0
.06)
�0.
12(0
.10)
�0.
12(0
.10)
�0.
12(0
.10)
Sec
tor
size
0.16
(0.1
5)�
0.02
(0.0
4)0.
16(0
.15)
�0.
00(0
.08)
�0.
01(0
.15)
�0.
01(0
.15)
�0.
01(0
.15)
cut1
Con
stan
t3.
83**
*(0
.81)
——
2.31
***
(0.4
2)4.
27**
*(0
.81)
4.49
***
(0.8
1)4.
48**
*(0
.81)
cut2
Con
stan
t4.
23**
*(0
.81)
——
3.17
***
(0.4
2)5.
76**
*(0
.81)
5.98
***
(0.8
1)5.
97**
*(0
.81)
cut3
Con
stan
t6.
27**
*(0
.81)
——
3.73
***
(0.4
2)6.
80**
*(0
.82)
7.01
***
(0.8
1)7.
00**
*(0
.81)
Con
stan
t0.
09(0
.17)
�2.
58**
*(0
.62)
Obs
erva
tion
s9,
120
9,12
09,
000
9,12
09,
120
9,12
09,
120
Fir
ms
135
135
135
135
135
135
135
Log
like
lih
ood
�67
75—
�46
97�
6669
�66
87�
6699
�66
98C
hi-
squ
are
2232
0.25
b14
2224
4524
0923
8523
86
Not
es.
Sta
nd
ard
erro
rsin
par
enth
eses
.a
Ref
erto
arti
cle
for
des
crip
tion
ofm
odel
s.b
R2.
*p
�.1
0**
p�
.05
***
p�
.01
Ram Ranganathan ([email protected]) is an assistant professor of management in the Mc-Combs School of Business at the University of Texas-Aus-tin. He received his PhD from the Management Department ofthe Wharton School at the University of Pennsylvania. Hisresearch focuses on understanding firm strategy in settings oftechnological change. His work emphasizes the pressuresfirms face in adapting to such change as well as how theyattempt to control future technological evolution by participa-tion in multifirm alliances and standards-setting committees.
Lori Rosenkopf ([email protected]) is theSimon and Mildred Palley professor in the ManagementDepartment of the Wharton School of the University ofPennsylvania. She received her PhD in management oforganizations from Columbia University. She studiesinterorganizational networks and knowledge flow in var-ious high-tech industries by analyzing participation intechnical committees, alliances, the mobility of techno-logical professionals, and patents within technologicalcommunities.
540 AprilAcademy of Management Journal
Copyright of Academy of Management Journal is the property of Academy of Managementand its content may not be copied or emailed to multiple sites or posted to a listserv withoutthe copyright holder's express written permission. However, users may print, download, oremail articles for individual use.