DOI: 10.1177/2041386614564105 A meta-analysis and path ...

Article

The social network sideof individual innovation:A meta-analysis andpath-analytic integration

Markus Baer and Karoline EvansWashington University in St. Louis, USA

Greg R. OldhamTulane University, USA

Alyssa BoassoThync Inc., USA

AbstractThe current study provides a comprehensive analysis and integration of the literature on the socialnetwork correlates of individual innovation. Reviewing the extant literature, we cluster existingnetwork measures into five general properties—size, strength, brokerage, closure, and diversity.Using meta-analysis, we estimate the population effect sizes between these network propertiesand innovation. Results showed that brokerage had the strongest positive relation to innovation,followed by size, diversity, and strength. Closure, by contrast, had a weak, negative associationwith innovation. In addition, we offer a path-analytic integration of the literature proposing andtesting the direct and indirect effects of the five properties on innovation. We suggest thatnetwork size and strength impact innovation through a web of relations with the more proximalfeatures of brokerage, closure, and diversity. Our path-analytic integration considers the twodominant perspectives on the effects of social networks—brokerage versus closure—simultane-ously allowing us to establish their relative efficacy in predicting innovation. In addition, our modelhighlights that network strength can have both negative and positive effects (via different direct andindirect pathways) and thus inherently involves a tradeoff. We discuss the implications of theseresults for future research and practice.

Paper received 5 June 2014; revised version accepted 15 November 2014.

Corresponding author:

Greg R. Oldham, A. B. Freeman School of Business, Tulane University, 7 McAlister Dr., New Orleans, LA 70118, USA.

Email: [email protected]

Organizational Psychology Review2015, Vol. 5(3) 191–223

ª The Author(s) 2015Reprints and permission:

sagepub.co.uk/journalsPermissions.navDOI: 10.1177/2041386614564105

opr.sagepub.com

OrganizationalPsychologyReview

at DUQUESNE UNIV on August 26, 2015opr.sagepub.comDownloaded from

http://www.sagepub.co.uk/journalsPermissions.nav

http://opr.sagepub.com

http://opr.sagepub.com/

KeywordsCreativity & innovation, diversity & relational demography, social networks

There is widespread scholarly consensus that

innovation constitutes a critical competitive

advantage and is becoming increasingly impor-

tant in allowing organizations to survive and

thrive (Damanpour & Schneider, 2006; Tellis,

Prabhu, & Chandy, 2009). Over the past two

decades, scholars and commentators have

increasingly emphasized that the relationships

individuals cultivate and the positions they

occupy in their social networks hold some of

the keys to unleashing their innovative poten-

tial (e.g., Brass, 1995; Ibarra, 1993; Perry-

Smith & Shalley, 2003). Indeed, as evidence

is accumulating that neither the production of

new ideas nor the execution of these ideas is the

domain of the lone genius (e.g., Baer, 2012;

Uzzi & Spiro, 2005; Wuchty, Jones, & Uzzi,

2007), there has been burgeoning interest in

identifying the features of individuals’ social

networks that are most powerful in boosting

individual innovation in organizations (Phelps,

Heidl, & Wadhwa, 2012).

This growing body of research has focused on

a wide array of network characteristics—ranging

from relational (e.g., tie strength) and nodal (e.g.,

functional background) to positional (e.g.,

betweenness centrality) and structural (e.g.,

density) features—and has produced an impres-

sive number of empirical findings. Properties that

have been featured more prominently in this lit-

erature and that have been found to impact indi-

vidual innovation, include the size of the network

(e.g., Baer, 2010; McFadyen & Cannella, 2004),

the average strength of the relationships com-

prising the network (e.g., McFadyen, Semadeni,

& Cannella, 2009; Rost, 2011) or, alternatively,

the number of weak/strong ties in the network

(e.g., Perry-Smith, 2006; Zhou, Shin, Brass,

Choi, & Zhang, 2009), the extent to which net-

work contacts are interconnected or, alterna-

tively, the extent to which the network provides

opportunities for brokerage (e.g., Burt, 2004;

Obstfeld, 2005), a person’s centrality in the net-

work (e.g., Mehra, Kilduff, & Brass, 2001;

Perry-Smith, 2006), and the network’s diversity

(Baer, 2010; Rodan & Galunic, 2004).

Despite some evidence that the resources

(e.g., task advice, strategic information, buy-

in, social support; Podolny & Baron, 1997)

accessible through social networks play a key

role in sparking creative ideas and fueling

attempts to subsequently execute and capita-

lize on these ideas (e.g., Anderson, 2008;

Cross & Sproull, 2004), a number of important

issues regarding the link between social net-

works and innovation remain unresolved.

First, the proliferation of properties that are

being studied, the different ways in which these

features have been operationalized across stud-

ies, and the fact that work in this area is frag-

mented across multiple academic disciplines

(e.g., management, sociology, political sci-

ence) represents a challenge for the literature

and, as a result, it is not clear which network

properties impact innovation more than others.

Thus, there is a need for these literatures to be

summarized in an integrative fashion allowing

for the comparison of the effects of various net-

work features. A meta-analysis allows us to iden-

tify true population effect sizes (Hunter &

Schmidt, 2004) and compare the various effects

with one another to determine which forces are

most potent.

Second, along with the variety of variables

that have been studied, there are also conflict-

ing theoretical lenses through which the link

between networks and innovation has been

examined. Specifically, there is a longstanding

debate about the relative potency for innovation

of network structures that offer opportunities

for brokerage between otherwise disconnected

parts of the network (e.g., Burt, 1992) and

192 Organizational Psychology Review 5(3)



closed structures, in which most people in the

network are connected to one another (e.g.,

Coleman, 1988). Although many scholars

assume that closure and brokerage are opposite

ends of the same continuum, both are associ-

ated with distinct theoretical arguments and

distinct operationalizations. In the present study,

we consider both perspectives simultaneously

contrasting their relative efficacy in predicting

innovation thereby highlighting the unique role

each plays.

Third, few if any attempts to examine the

linkages between individuals’ social networks

and innovation have considered potential

mediating mechanisms. However, there is good

theoretical and empirical evidence to suggest

that certain network properties may operate

through other elements of the network to exert

their influence on innovation. To address this

issue, in addition to meta-analytically review-

ing the extant literature, we offer a path-

analytic integration of this body of work speci-

fying the indirect paths through which certain

network features may impact other network

characteristics to ultimately shape innovation.

Given that estimates of the population effect

sizes between the various network features

examined here were not readily available, we

conducted a second meta-analysis to derive

these estimates—the first of its kind according

to our knowledge.

Finally, although a number of quantitative

reviews of the innovation literature at various

levels of analysis have been offered in recent

history (e.g., Damanpour, 1991; Hammond,

Neff, Farr, Schwall, & Zhao, 2011; Hulsheger,

Anderson, & Salgado, 2009), none have con-

sidered the role individuals’ social networks

may play in affecting individual innovation in

organizations. By focusing specifically on the

social network correlates of innovation, we

hope to provide a complementary perspective

to these earlier meta-analytic reviews spurring

a more vibrant dialogue between work on

social networks and innovation at the individ-

ual level of analysis.

Theoretical backgroundand hypotheses

Innovation in organizations

We define innovation as the process of gen-

erating new ideas for organizational products,

procedures, processes, and service offerings,

of testing, developing, and refining these ideas

for future use, and of implementing them (e.g.,

Amabile, 1996; Kanter, 1988; Oldham &

Cummings, 1996; van de Ven, 1986; West,

2002). Thus, innovation encompasses creative

idea generation and the subsequent execution,

adoption, or implementation of these ideas.

Although we acknowledge the possibility that

factors impacting the inception of a new idea

may differ from those that shape whether an

idea is subsequently executed (e.g., Axtell

et al., 2000; Clegg, Unsworth, Epitropaki, &

Parker, 2002; Frese, Teng, & Wijnen, 1999),

the literature on the social network correlates

of innovation has rarely examined this dis-

tinction—neither theoretically nor empiri-

cally. Given the state of the literature and the

relatively moderate number of studies that

have examined the link between innovation

and social network properties to begin with,

we elected to follow previous meta-analytic

efforts on the topic of innovation (e.g., Hulsheger

et al., 2009) not distinguishing between idea gen-

eration and implementation but instead to focus

our attention on examining overall innovation.

Scholars have used a wide array of indicators

to measure innovation, ranging from more sub-

jective indicators, such as self/others’ evalua-

tions of a person’s innovation (e.g., Moran,

2005; Rodan & Galunic, 2004) to more objective

indicators, such as the number of patents, pub-

lications, and forward citations obtained or the

prizes awarded for certain innovative achieve-

ments (e.g., Fleming, Mingo, & Chen, 2007;

Liao, 2011). Although there is some evidence to

suggest that predictor–criterion relations may

differ systematically across different measure-

ment approaches (e.g., Hammond et al., 2011;

Baer et al. 193



Hulsheger et al., 2009), the strength of our

analysis lies in the fact that we combine

studies from different fields of inquiry and

provide estimates of the strength of the network–

innovation relations across different ways of

operationalizing innovation.

Linking network properties directlyto innovation

The extant literature on social networks has

considered a wide variety of network proper-

ties and ways of operationalizing the various

features. In fact, there are numerous measures

that seemingly capture the same underlying

conceptual idea. For instance, constraint, effi-

ciency, and betweenness centrality all capture

the extent to which individuals serve as infor-

mation conduits between different parts of the

network. In addition, there has been a pro-

liferation of operationalizations with some

authors creating their own idiosyncratic network

indicators. To clarify this picture, we follow

previous research (e.g., Burt, 1992) and focus

our attention on those concepts and measures

that have been most frequently and consistently

used in the literature: network size, strength,

brokerage, closure, and diversity.

Network size, reflecting the number both of

direct and secondary ties, acts as a channel to

access different sources of information and

knowledge (Freeman, 1979; Gabbay & Leen-

ders, 2001; McFadyen & Cannella, 2004).

In fact, Anderson (2008) showed that managers

with larger networks were able to gather more

task-relevant and diverse information. Expo-

sure to more, and more diverse information and

perspectives, in turn, may give an individual an

edge in identifying existing opportunities for

improvement and in having access to the raw

material needed to develop fresh ideas for how

to exploit them (Campbell, 1960; Mumford &

Gustafson, 1988). Consistent with this asser-

tion, McFadyen and Cannella (2004) showed

that network size had a positive correlation with

a biomedical researcher’s production of

impactful (journal-impact weighted) publica-

tions. In addition, having more connections

enhances the probability of idea implementa-

tion by increasing the access an individual has

to contacts that may be both willing to engage

in and to support a potentially risky venture.

Thus, we predict a positive relation between

network size and innovation.

Hypothesis 1. Network size is positively

related to innovation.

Network strength reflects the average strength

of the relationships comprising a network. The

strength of a tie increases as a function of the

number and reciprocity of interactions, the affec-

tive intensity of a relationship, or the amount of

time the relationship has existed (Granovetter,

1973). While tie strength characterizes the quality

of the dyadic relationship between two parties,

network strength captures the average strength

of all relationships in an actor’s network

(McFadyen & Cannella, 2004). Therefore, net-

work strength increases as the number of ties in

the network that are characterized by frequent

interactions, emotional closeness, or long dura-

tion increases (Perry-Smith & Shalley, 2003).

Prior examinations of the importance of

network strength for innovation have yielded

mixed results, with support for its ability to

enhance (e.g., Moran, 2005) and restrict (e.g.,

Perry-Smith, 2006) innovation. According to

one line of argumentation, network strength has

the potential to boost innovation. While large

networks provide enhanced exposure and

access to more and different information and

knowledge sets, network strength allows for the

transfer of these resources such that individuals

can comprehend and make use of them (e.g.,

Kijkuit & van den Ende, 2010). Repeated inter-

actions help parties become more efficient in

exchanging complex, noncodified information

and allow for the creation of an overlapping

knowledge base upon which to generate and

interpret ideas (e.g., Bouty, 2000; Hansen,

1999; Reagans & McEvily, 2003). In addition,

strength has been suggested and found to be




instrumental in identifying and formulating

problems and in providing validation for indi-

viduals’ ideas by bolstering their confidence

in them, thereby creating the conditions neces-

sary for people to share their ideas persuasively

(e.g., Cross & Sproull, 2004). Consistent with

this line of reasoning, Fleming et al. (2007)

found that repeated collaboration on patents

positively correlated with both the novelty

(number of new patent subclasses assigned) and

future usefulness (number of forward citations)

of an inventor’s patents.

In addition, strong networks, particularly

those in which strength is derived from mem-

bers being emotionally close to one another,

cultivate trust which creates a nonthreatening

arena in which to propose new ideas without the

fear of attack, ridicule, or rejection (Krackhardt,

1992; Levin & Cross, 2004). Innovation typi-

cally involves taking risks, with individuals

exhibiting higher creativity when groups are

high in trust (e.g., Clegg et al., 2002), psycho-

logical safety (e.g., Baer & Frese, 2003), and

social support (e.g., Madjar, 2008). Further,

members who interact frequently and over a

longer period of time are more likely to feel a

sense of obligation to continue the relationship

and to reciprocate any past favors or exchanges,

thereby motivating assistance in new ventures

in order to maintain a balanced relationship

(e.g., Granovetter, 1983; Krackhardt, 1992).

Alternatively, network strength may be det-

rimental to innovation because it decreases the

availability of novel, nonredundant information

in the network. Weak ties are more likely to

connect different groups and to be a source of

diverse, nonredundant information (Grano-

vetter, 1973; Powell & Smith-Doerr, 1994). For

example, Perry-Smith (2006) observed a posi-

tive correlation between the number of weak

ties in an employee’s network and supervisor-

rated creativity. Similarly, individuals whose

connections are predominantly strong are more

likely to build redundant knowledge stocks over

time by transferring knowledge over repeated

interactions (McFadyen & Cannella, 2004). In

fact, McFadyen and Cannella (2004) suggested

and found that increasing network strength can

be detrimental for innovation leading to a

reduction in the number of publications.

In strong networks, expectations and exchange

rules become ingrained in the relationships,

leading individuals to become more cautious

about deviating from established practices in

order to prevent backlash or rejection from their

interaction partners (Krackhardt, 1999). Given

that innovation requires deviation from the

norm, strong networks may therefore suffocate

individuals’ motivation to generate and execute

risky ideas (McPherson, Smith-Lovin, & Cook,

2001). This line of reasoning would suggest

that network strength may undermine innova-

tion. Based on the previous arguments, we

therefore propose competing hypotheses sug-

gesting that network strength may either posi-

tively or negatively relate to innovation.

Hypothesis 2a. Network strength is posi-

tively related to innovation.

Hypothesis 2b. Network strength is nega-


Individuals who bridge structural holes

(i.e., the gaps in the structure that exist

between otherwise disconnected parts of the

network) have certain innovation advantages

by connecting separate groups and controlling

the flow of communication across them. Specif-

ically, brokerage promotes innovation through

two mechanisms: obtaining access to nonredun-

dant information and knowledge sets and con-

trolling the presentation and use of resources

across contacts (Burt, 2005; Kilduff & Tsai,

2003). First, by connecting separate clusters

of individuals, brokers are more likely to be

exposed first to new opportunities and to have

privileged access to nonredundant information

and knowledge sets than if they only had con-

nections within a single, well-connected cluster

(Burt, 1992). Having access to unique, nonover-

lapping sources of information provides a wider

variety of information from which to draw on for

Baer et al. 195



ideas (Harrison & Klein, 2007). Indeed, Mehra

et al. (2001) found that connecting otherwise dis-

connected work colleagues was a positive pre-

dictor of innovation.

Second, because brokers bridge otherwise

disconnected parts of the network, this affords

them a certain amount of control, giving indi-

viduals the ability to decide when and with

whom to share information across groups

(Hargadon & Sutton, 1997). In addition, because

brokers reside between disconnected groups,

brokerage behavior is less prone to monitoring

and control than if groups were able to interact

directly (Burt, 2005). In this way, brokers have

autonomy to gather and disseminate information

between connections, helping them to promote

the acceptance and implementation of their new

ideas by selectively sharing resources (Burt,

1992; Padgett & Ansell, 1993). Burt (2004)

found empirical support for this contention by

showing that managers who bridged structural

holes not only had ideas that were rated as more

valuable, they were also more likely to have their

ideas discussed and engaged by senior manage-

ment and less likely to have them dismissed.

The benefits stemming from occupying a

brokerage position, however, may not be

enduring. For example, work by Buskens and

van de Rijt (2008) showed that in situations in

which all actors strive to take advantage of

structural holes, the benefits of brokerage

quickly evaporate. In addition, serving as a

broker may also have the potential to under-

mine a focal individual’s ability to innovate. For

instance, Krackhardt (1999) suggested that

bridging a structural hole subjects a broker to

competing sets of group norms that may in fact

be more constraining than those faced by an

actor in a single connected clique. Moreover,

given the privileged access to diverse informa-

tion sets, relaying this information to others with

potentially very different ways of seeing the

world may result in overload thereby adversely

impacting a broker’s ability to capitalize on his

or her position (Cross & Parker, 2004). Finally,

because a broker’s behavior is less prone to

monitoring, brokers could engage in incidental

or strategic filtering, distortion, or hoarding of

information (Balkundi, Kilduff, Barsness, &

Michael, 2007; Burt, 1992). Recent research

suggests that such behavior, if detected by alters,

causes a spiral of distrust and information and

knowledge hiding on both sides, ultimately

undermining the creativity of the broker (Cerne,

Nerstad, Dysvik, & Skerlavaj, 2014). To the

extent, however, that not all individuals in a

network are striving towards structural holes and

the focal actor is able to deal with disparate

social spheres without taking advantage of his or

her position, brokerage should be positively

associated with individual innovation.

Hypothesis 3: Brokerage is positively


Closure (i.e., the extent to which the contacts

in one’s network are also connected) has the

potential to either enhance or reduce an individ-

ual’s innovation. Dense networks create trust

because overlapping ties make it less likely that

people take advantage of each other—violators

will be easily identified and will be forced either

to amend their behaviors or to leave the network

(Coleman, 1990). In addition, network closure,

by encouraging information exchange among

densely connected people, may enhance innova-

tion by facilitating the creation of a shared base

of knowledge upon which constituents can

rely to interpret and contribute to a new idea

(Milliken, Bartel, & Kurtzberg, 2003; Uzzi,

1997). Thus, when network structures are dense,

norms of cooperation are more likely to emerge

and be internalized, trust is more likely to be

established, and actions can be coordinated more

seamlessly—all of which should make it more

likely that vital information is shared, that tacit

knowledge can be transferred, that conflicting

viewpoints can be integrated, and that resources

needed to act upon one’s ideas can be more eas-

ily secured and mobilized (e.g., Coleman, 1988;

Granovetter, 1985; McFadyen et al., 2009;

Podolny & Baron, 1997; Reagans & McEvily,

2003). Obstfeld (2005) provided support for this




overall logic. He observed a positive associa-

tion between network density and individuals’

involvement in the firm’s innovative activities.

Through additional interviews, he confirmed

that closure allowed for faster coordination

in prototyping processes and for the mobiliza-

tion of support for change.

Alternately, closure may be detrimental to

innovation as the frequent, reciprocal interac-

tions that typify dense networks often result in

redundant sharing of information and reinforce-

ment of existing group norms and processes.

Ties within closed networks tend to connect

similar and overlapping social circles, resulting

in increased likelihood that information dis-

tributed between contacts will be redundant or

already known (Granovetter, 1973). Addition-

ally, closed networks create strong social norms,

which may inhibit individuals’ willingness to

take the risk of suggesting or implementing new

ideas (Coleman, 1990). Closure may create social

obligations that restrict the ability for individuals

to act freely (Mizruchi & Stearns, 2001)—the

overlapping ties in dense networks make it easier

to detect deviant behavior, improving the like-

lihood that actors will adhere to established

expectations to avoid rejection or sanctions

(Coleman, 1988). Individuals in dense networks,

therefore, tend to conform, which can deter the

sharing or introduction of new ideas (Ahuja, 2000;

Obstfeld, 2005). In fact, Fleming et al. (2007)

found that network closure exhibited a negative

relation with the number of new patent combina-

tions, an indicator of invention novelty. However,

patents arising from cohesive networks in their

sample were more likely to generate future use,

through citations by other inventors, reinforcing

that there are positive benefits that may flow from

closure. Therefore, we propose competing

hypotheses suggesting that closure may either

positively or negatively relate to innovation.

Hypothesis 4a: Closure is positively related

to innovation.

Hypothesis 4b: Closure is negatively related

to innovation.

The diversity of the network is a potentially

invaluable resource for innovation. Diversity

typically refers to differences across an individ-

ual’s network contacts in terms of both task-

oriented (e.g., educational background, tenure)

and relations-oriented (e.g., gender, age) fea-

tures (Jackson, May, & Whitney, 1995; Joshi

& Roh, 2009). Diversity is key to innovation

because these task- and relations-oriented dif-

ferences are assumed to reflect more deep-

level differences in information and knowledge

sets as well as other resources that are necessary

for the generation of fresh ideas and for their

execution (Adler & Kwon, 2002; Rodan &

Galunic, 2004). Specifically, access to hetero-

geneous sources of information and knowledge

may heighten an individual’s sensitivity to

existing environmental needs, improving the

likelihood of recognizing opportunities for

innovation (Perry-Smith & Shalley, 2003;

Shane, 2000). Next, being exposed to people

from different backgrounds and walks of life

enhances the probability not only that individu-

als may come across ideas that could solve

existing problems in other parts of the organiza-

tion but also that they combine different ideas

to generate new solutions altogether (Fleming

et al., 2007; Rodan & Galunic, 2004).

Finally, network diversity makes it more

likely that individuals know others who can

validate their ideas, who can lend their cred-

ibility and support, or who can provide other

resources (e.g., time, energy) needed to push an

idea through to completion. Thus, having access

to individuals who are not only fundamentally

different from the focal individual but also from

each other makes it more likely that individuals

are exposed to the variety of perspectives and

resources critical to the identification of new

opportunities, to the development of new ideas

to solve existing problems, and to the imple-

mentation of one’s ideas. Consistent with this

logic, Perry-Smith (2006) observed a positive

correlation between task-oriented diversity and

supervisor-rated creativity. In addition, Rodan

and Galunic (2004) showed that the

Baer et al. 197



heterogeneity of knowledge accessible through

one’s network indeed related positively and sig-

nificantly to managerial innovation. Based on

the previous arguments and results, we therefore

predict that network diversity will be positively


Hypothesis 5: Network diversity is posi-


Linking network properties directlyand indirectly to innovation:A path-analytic model

In the following, we develop a path-analytic

model that specifies the direct and indirect

effects through which network size, strength,

brokerage, closure, and diversity impact inno-

vation. Our path model complements our pre-

vious theoretical analysis of the direct effects

on innovation in an important way. Specifi-

cally, by modeling both the direct and indirect

effects of the five social network properties

simultaneously we are able to paint a more

comprehensive picture of the effects of social

networks on innovation and are able to evalu-

ate the relative strength of each effect. The

path model is depicted in Figure 1.

In developing our path model, we con-

ceptualize brokerage and closure as separate

forces impacting innovation. Although closure

has often been considered as the opposite of

brokerage (e.g., Obstfeld, 2005), a close

reading of the literature suggests that both

properties are associated with distinct theore-

tical logics and different operationalizations.

While the beneficial effects of brokerage on

innovation and other outcomes are often

attributed to the fact that brokers have privi-

leged access to more fresh ideas and opportu-

nities and can control the flow of information

(Burt, 1992, 2004), the effects of closure are

typically attributed to the fact that dense network

structures promote the development of trust and

solidarity and allow for the mobilization of

resources and coordinated action (Coleman,

1988; Obstfeld, 2005). In addition, the benefits

associated with brokerage tend to be attributed to

individuals operating at the intersection between

different social spheres whereas the benefits

associated with closure seem to stem largely

from within the primary social sphere in which

the focal actor is operating. Thus, both theore-

tical logics are not mutually exclusive and can

operate simultaneously (Reagans & Zuckerman,

2001). Moreover, each property is typically

captured via different measures—brokerage is

often operationalized as constraint, efficiency, or

centrality (Burt, 1992; Freeman, 1977), while

closure is typically captured via network density.

Hence, we decided to treat these two network

properties as separate entities. This allowed us to

not only consider the different ways in which

each may shape innovation but also to quanti-

tatively evaluate the relative potency of each

property in impacting innovation.

Network size

Closure

Network diversity Innovation

+

Network strength

Brokerage

+

−

−

+

+

+/−

+

Figure 1. Theoretical model of network effects on innovation.




Network size and strength are hypothesized

to exhibit direct relations with brokerage—

network size is expected to be positively related

to brokerage, whereas we expect network

strength and brokerage to be negatively linked.

The positive association between size and

brokerage is rooted both in theoretical and

empirical reasons. From a theoretical stand-

point, as networks increase in size, the prob-

ability increases that individuals from very

different backgrounds and with very different

interests join the network. Individuals may even

cultivate relationships for specific and different

purposes (e.g., to obtain advice or to mobilize

support; Burt, 1992), resulting in the network

becoming more fragmented and allowing indi-

viduals to serve as brokers between these oth-

erwise disconnected social worlds (Kilduff &

Tsai, 2003). From an empirical standpoint, a

number of measures typically used to capture

brokerage (e.g., constraint or efficiency) factor

network size into the measure itself so that a

positive association between the two concepts is

likely empirically (e.g., Rodan & Galunic, 2004).

Indeed, a study by Fleming and Waguespack

(2007) revealed a positive correlation between

a measure of network size (the number of coau-

thors with whom an individual had published)

and brokerage. Burt (2000), in an examination

of multiple managerial samples, also found a

positive correlation between network size and

brokerage—operationalized via constraint.

Network strength, on the other hand is

expected to be negatively related to brokerage.

Granovetter (1973) made the persuasive argu-

ment that weak ties are more likely than strong

ties to serve as bridges between otherwise dis-

connected parts of the network. In contrast,

strong ties are likely to be associated with dense

collections of redundant ties (Perry-Smith,

2006). This suggests that network strength

should be negatively linked to brokerage. Burt

(1992) challenged this argument suggesting

that it is not the quality of the relationship

spanning a chasm that matters, but that a struc-

tural hole exists in the first place—whether the

hole is spanned by a strong or weak tie is sec-

ondary. Although we concur with Burt in that

brokerage is a more proximal causal agent than

network strength, there is nevertheless good

reason to believe that increasing network

strength makes it less likely that individuals

occupy brokerage positions. Empirical evidence

supports this logic. For example, in a study of

government employees, Lee and Kim (2011)

found a negative correlation between network

strength and brokerage—measured via network

efficiency. Together, then, we hypothesize a pos-

itive relation between network size and broker-

age and a negative relation between network

strength and brokerage.


related to brokerage.

Hypothesis 7. Network strength is nega-

tively related to brokerage.

Network size and strength are also hypothe-

sized to exhibit relations with closure—net-

work size is expected to be negatively related

to closure whereas we expect network strength

and closure to be positively linked. Assuming

a finite level of time and energy, increasing

network size makes it less likely for individu-

als to be able to interact with all of their con-

tacts, or a subset of them, at the same time,

thereby reducing the possibility of closure to

occur (Burt, 2000). In addition, increasing net-

work size may serve strategic purposes, such

that individuals may divide up their time

between different individuals to obtain differ-

ent resources and to achieve different goals.

In essence, this logic suggests that network

size coincides with reduced multiplexity of

relationships (Louch, 2000), resulting in sets

of ties that each serves very different functions

(e.g., task advice, strategic information, buy-

in, social support; Podolny & Baron, 1997).

As a result, closure between contacts is

unlikely to occur to the same extent as when

networks are smaller. Empirical evidence pro-

vides support for this logic (e.g., Moran, 2005;

Baer et al. 199



Mors, 2010; Reagans & McEvily, 2003). For

example, Carter and Feld (2004) observed a neg-

ative correlation between network size and den-

sity in a sample of college students.

Network strength is also expected to relate

to density—but positively. According to Louch

(2000), as the frequency of contact—one indi-

cator of the strength of a relationship—between

an individual and his/her contacts increases, the

greater the chance that the contacts themselves

form a bond. Assuming limited time resources,

people are constrained in the number of strong

ties they can maintain at a given point in time

causing people with multiple strong ties to spend

time with their contacts at the same time (e.g.,

Parks, Stan, & Eggert, 1983). In addition, ties

that are well-established and have a long his-

tory—another indicator of the strength of a rela-

tionship—tend to settle more often into

transitive patterns than ties with a brief history

(e.g., Hallinan & Hutchins, 1980; transitivity

posits that if a chooses b as a friend and b

chooses c as a friend, then a will choose c as a

friend; Holland & Leinhardt, 1970). Empirical

evidence supports these arguments (e.g., Louch,

2000; Reagans & McEvily, 2003). For example,

McFadyen et al. (2009) reported a positive corre-

lation between average tie strength and density.

Together, then, we predict a negative relation

between network size and closure and a positive

relation between network strength and closure.

Hypothesis 8. Network size is negatively

related to closure.

Hypothesis 9. Network strength is positively

related to closure.

We also expect that network diversity

increases as a function of the size of the net-

work. Intuitively, as the number of members

in one’s network increases, so does the likeli-

hood that they represent different functional

and social circles. However, given that time and

energy are finite resources, there is a limit to the

number of ties one can cultivate and maintain. In

fact, adding similar (in terms of functional and

social category diversity) contacts to one’s net-

work increases operating costs without offering

many additional benefits thereby producing neg-

ative returns (e.g., McFadyen & Cannella, 2004;

Zhou et al., 2009). Therefore, strategic networ-

kers seek to acquire additional ties that provide

the greatest returns on their expenditures, and

increasing size may not provide meaningful

returns if each additional contact is embedded

in the same social circles (Aral & van Alstyne,

2011; McPherson et al., 2001). By reaching out

to new connections possessing distinct qualities,

individuals can maximize the likelihood that

these contacts provide access to diverse and

unique pools of information and knowledge.

Indeed, Ibarra (1995) demonstrated that purpo-

sive network development strategies inject

diversity into the network, even those utilizing

homophily to expand their network. As a result,

there should be a positive association between

the size and diversity of a network.


related to diversity.

Next, we consider the possibility that broker-

age, closure, and diversity mediate the effects

of network size on innovation. The idea that

the composition and structure of individuals’

networks and the positions they occupy within

their social spheres may play an integral role in

realizing the benefits associated with increas-

ing network size is not new. Burt (1992), for

instance, suggested that although the number

of structural holes is likely to be a function

of an increase in network size, it is not the

number of people one is connected to but the

number of disconnected actors that determines

whether individuals come to enjoy the benefits

of privileged and enhanced access to diverse

ideas and of having opportunities to exert con-

trol. In other words, size alone does not create

an advantage; the advantage comes from the

configuration of relationships, the positions

that people occupy within the structure, and

the types of contacts they are connected to.

Thus, network size is only likely to positively




impact innovation to the extent that it provides

individuals access to contacts who are differ-

ent from each other and who are otherwise dis-

connected thereby creating opportunities for

brokerage and reducing closure. Serving as a

broker between otherwise disconnected and

different parts of one’s network and operating

under conditions of reduced closure while

being able to tap into diverse social spheres,

should directly enhance innovation. Together,

then, these arguments suggest that brokerage,

closure, and diversity fully mediate the posi-

tive effect of network size on innovation.

Hypothesis 11. Brokerage, closure, and

diversity fully mediate the effects of

network size on innovation.

Brokerage and closure are also expected to

mediate the effects of network strength on

innovation. As with network size, there is some

evidence that the benefits attributed to network

strength—strong networks serve as a mechan-

ism for the transfer of knowledge, particularly

tacit knowledge, and provide access to help

and support (Hansen, 1999; Uzzi, 1997)—are

a function of the structure of the network

(Reagans & McEvily, 2003). Tie strength tends

to produce closure—strong ties are more likely

to be embedded in dense third-party relation-

ships (Granovetter, 1973; Hansen, 1999)—and

it is the emergence of cooperative norms in

such dense structures and the internalization

of these norms that sets the stage for the

successful generation and implementation of

creative ideas. Specifically, closure and the

cooperative norms that it promotes make it

more likely that individuals invest their time

and energy in assisting others, that the level

of trust emerges that is needed to access

resources crucial to the innovative process, and

that actions can be efficiently and effectively

coordinated within the network (e.g.,

Krackhardt, 1992; Podolny & Baron, 1997;

Reagans & McEvily, 2003). Thus, to the extent

that closure enhances innovation, we expect

that the positive effect of network strength on

innovation will be mediated by closure.

Of course, there is a drawback of having a

network populated with predominantly strong

ties. Network strength can limit brokerage

(Fleming et al., 2007)—strong ties are less

likely to serve as bridges between otherwise

disconnected parts of the network. This, in turn,

limits exposure to diverse information and

restricts the ability to exert control—both of

which should stifle innovation. Taken together

then, the aforementioned arguments suggest

that closure might mediate the positive effect of

strength on innovation while brokerage may

mediate its negative effect.

Hypothesis 12. Brokerage and closure fully

mediate the effects of network strength on

innovation.

Methods

Data collection

We used multiple search techniques to identify

relevant empirical studies examining the lin-

kages between the various social network

properties and innovation. First, we performed

a search of computerized databases, including

PsycARTICLES, Psychology and Behavioral

Sciences Collection, PsycINFO, SocINDEX,

Communication Abstracts, Business Source

Complete, ABI/Inform, ProQuest Digital Dis-

sertations, and Web of Science. We used broad

keywords related to the primary concepts, such

as ‘‘creativity,’’ ‘‘innovation,’’ and ‘‘knowledge

creation,’’ in conjunction with ‘‘social/knowledge

networks,’’ ‘‘network structure,’’ and ‘‘social

capital’’ as well as terms specifically related to the

individual network properties (e.g., ‘‘network

density,’’ ‘‘network centrality’’). The electronic

search was supplemented with a manual search of

forward references to selected theoretical and

empirical articles on social networks and inno-

vation (e.g., Burt, 2004; Perry-Smith & Shalley,

2003). In addition, we manually searched the

Baer et al. 201



reference lists and forward citations of all studies

that met our inclusion criteria. Third, we searched

national conference programs (e.g., Academy of

Management, Strategic Management Society)

using the same search terms as noted before.

Finally, we contacted selected researchers in

the area to obtain current or unpublished studies

that may fit our criteria for inclusion. This

search yielded 6,128 relevant citations.

From this initial set of articles, we applied

additional criteria for inclusion. First, we only

included studies at the individual level of

analysis eliminating studies, for example, that

examined the link between social network

properties of teams and innovation (e.g.,

Leenders, van Engelen, & Kratzer, 2003;

Reagans & Zuckerman, 2001) or that focused

on interdepartmental or interorganizational

networks and outcomes related to innovation

(e.g., Hansen, 1999; Tsai & Ghoshal, 1998).

Second, to be included, a study had to report on

at least one relation between a network property

and innovation. Third, we included only studies

with original data. In cases where multiple

studies utilized the same data set (e.g., Rodan,

2002, 2010; Rodan & Galunic, 2004), we

eliminated any redundant studies including

only the study that had been published first or

constituted the most comprehensive analysis of

the underlying data set. We included non-

overlapping data from dissertations for four of

the samples (Baer, 2010, 2012; Obstfeld, 2005;

Perry-Smith, 2006; Tortoriello & Krackhardt,

2010). Fourth, we only included studies that

relied on actual network data to derive indica-

tors of the various network properties and

excluded studies that used network-type mea-

sures without having collected egocentric or

complete network data. For example, although

Teigland and Wasko (2003) examined the link

between frequency of communication (a mea-

sure of average strength) and creativity, the fre-

quency measure was not based on previously

identified network contacts but rather referred

to people in general in the organization (e.g.,

‘‘coworkers in my location’’). Consequently,

studies such as these were excluded from our

sample. Finally, included studies had to report

sample sizes and either a correlation coefficient

or enough information to calculate a correlation

coefficient. We contacted authors of studies

that met all criteria, except the reporting of cor-

relations (e.g., Burt, 2004), obtaining useable

data for an additional nine studies.

All authors then independently coded these

for inclusion in our final data set based on

whether the network and innovation measures

in each of the studies were consistent with our

definition of these concepts. At this point, we

excluded studies that used idiosyncratic indi-

cators of network properties that, in many

cases, combined a number of network measures

into one overall amalgam (e.g., Casanueva &

Gallego, 2010; Sorenson, Rivkin, & Fleming,

2006) making it difficult to discern which net-

work feature ultimately is driving the observed

effect. This procedure achieved satisfactory rates

of agreement for both the network properties

(92% agreement) and the innovation measures

(83% agreement), and perfect agreement was

reached through subsequent discussion. The

final data set contained 45 studies (see Table 1

for a summary of the included studies).

Using a coding scheme based on the defi-

nitions of network properties provided by

Wasserman and Faust (1994) and the definition

of innovation provided by van den Ven (1986)

and others, all authors classified all possible

effect sizes in each of the 45 studies (studies

could contain multiple effect sizes). Some

studies included panel data or lagged variables

(e.g., McFadyen & Cannella, 2004; McFadyen

et al., 2009); in these cases, we only coded

those effect sizes for which independent and

dependent variables referred to the same time

period to be consistent with the other cross-

sectional studies in our sample.

Operationalization of variables

Network size reflects a count of the number

of ties present in an individual’s network.




Table 1. Summary of papers included in meta-analysis.

Studies Innovation measure Network type Network properties

1. Anderson (2006) Self- & other-rating Ego network Brokerage (betweenness centrality, other),number of weak ties

2. Audretsch,Aldridge, & Sanders(2011)

Startup formation Ego network Number of strong ties

3. Baer (2010, 2012)a Other-rating Ego network Strength, diversity (functional), size, number ofstrong ties

4. Beaudry & Alloui(2012)

Publication count,patent count

Completenetwork

Brokerage (betweenness centrality)

5. Beaudry & Kananian(2013)

Forward citationcount

Completenetwork

Brokerage (betweenness centrality)

6. Bjork, Di Vincenzo,Magnusson, &Mascia (2011)

Other-rating Completenetwork

Size, brokerage (efficiency)

7. Burt (2004) Other-rating Ego network Brokerage (constraint)8. Cannella &

McFadyen (2011)Publication count,

forward citationcount

Completenetwork

Strength, closure (overall network density)

9. Cattani & Ferriani(2008)

Critical awards Completenetwork

Brokerage (constraint)

10. Chen & Liu(2012)

Publication count Completenetwork

Brokerage (efficiency), diversity (functional)

11. Chen & Gable(2013)

Other-rating Ego network Size

12. Chiu (2013) Publication count Completenetwork

Size

13. Chua (2011) Other-rating Ego network Size, strength, closure (overall networkdensity, alter density), diversity (socialcategory)

14. Ding, Levin,Stephan, & Winkler(2010)


Size

15. Fleming, Mingo, &Chen (2007)

Patent count,forward citationcount

Completenetwork

Size, strength, closure (alter density), diversity(functional)

16. Fleming &Waguespack (2007)

Patent count,publication count

Completenetwork

Size, brokerage (constraint)

17. Gonzalez-Brambila, Veloso, &Krackhardt (2013)


Size, strength, brokerage (other), closure(overall network density)

18. Keri (2011) Self-rating Ego network Size, number of strong ties19. Kratzer & Lettl

(2008)Other-rating Complete

networkBrokerage (betweenness centrality)

20. Lee (2009) Patent count Completenetwork

Size, brokerage (other), closure (alter density)

21. Lee (2010) Patent count Completenetwork

Size, brokerage (constraint), diversity(functional)

22. Li, Liao, & Yen(2013)


Completenetwork

Size, brokerage (betweenness centrality)

(continued)

Baer et al. 203



Table 1. (continued)

Studies Innovation measure Network type Network properties

23. Liao (2011) Research awards,forward citationcount

Completenetwork

Strength

24. Liu (2011) Patent count Completenetwork

Brokerage (efficiency, betweenness centrality)

25. Liu, Chiu, & Chiu(2010)

Patent count Completenetwork

Brokerage (efficiency)

26. Liu & Lin (2012) Publication count Completenetwork

Size, brokerage (efficiency)

27. Ma, Huang, &Shenkar (2011)

Self-rating Ego network Strength

28. Maritz (2010) Publication count,self-rating

Ego network Strength

29. McFadyen &Cannella (2004)


Size, strength

30. McFadyen,Semadeni, &Cannella (2009)


Strength, closure (overall network density)

31. Mehra, Kilduff, &Brass (2001)

Other-rating Ego network Size, brokerage (betweenness centrality)

32. Moran (2005) Other-rating Ego network Size, strength, closure (alter density)33. Mors (2010) Other-rating Ego network Size, closure (overall network density)34. Obstfeld (2005)a Self-rating Ego network Size, brokerage (constraint), closure (overall

network density)35. Paruchuri (2010) Forward citation

countComplete

networkSize, diversity (functional)

36. Perry-Smith(2006)a

Other-rating Ego network Strength, brokerage (betweenness centrality),diversity (functional, social category),number of strong & weak ties

37. Rhee & Ji (2011) Other-rating Ego network Number of weak ties38. Rodan & Galunic

(2004)Other-rating Ego network Size, closure (alter density), diversity

(functional), number of strong ties39. Rost (2011) Patent count,

forward citationcount

Completenetwork

Brokerage (other), strength

40. Rotolo & MesseniPetruzzelli (2013)


Size

41. Schultz &Schreyogg (2013)


Ego network Number of strong and weak ties

42. Tortoriello &Krackhardt (2010)a

Patent count Ego network Size, brokerage (constraint, efficiency,betweenness centrality), number of strongties

43. Totterdell,Holman, & Hukin(2008)

Self-rating Ego network Size, brokerage (betweenness centrality,other)

44. Zhou, Shin, Brass,Choi, & Zhang(2009)

Other-rating Ego network Closure (overall network density), number ofstrong & weak ties

45. Zou & Ingram(2013)

Other-rating Ego network Size, brokerage (constraint)

aStudies included supplemental data from the dissertation of one of the authors that may not be in the published piece.




Network size includes both the direct con-

nections in egocentric networks as well as the

combination of direct and indirect connections

in complete networks. We included direct and

indirect ties in complete networks as indirect

ties have been suggested and shown to provide

benefits over and above those provided by

direct ties in terms of access to and the

mobilization of resources relevant to innova-

tion (e.g., Bian, 1997).1,2 In egocentric net-

works, out-degree centrality, or the number of

ties reported by a person, was also considered

a measure of network size (Kilduff & Tsai,

2003).

Network strength captures the average

strength of all relationships in an individual’s

network. The strength of a tie increases as a

function of the ‘‘amount of time, the emotional

intensity, the intimacy (mutual confiding), and

the reciprocal services which characterize the

tie’’ (Granovetter, 1973, p. 1361). Consistent

with this definition and previous research (e.g.,

Ma, Huang, & Shenkar, 2011; Perry-Smith,

2006), we included measures in this category

that captured at least one of the following

dimensions of strength: duration of the relation-

ship, the frequency of interaction, or the emo-

tional closeness characterizing the relationship.

Network strength is most frequently mea-

sured by asking respondents to report how long

a particular relationship has been in existence,

how often they interact with a particular con-

tact, or how close they feel they are to the other

person. The various formative items are then

averaged across all ties in the network and used

either as separate measures (e.g., Perry-Smith,

2006) or are combined to form one overall indi-

cator of network strength (e.g., Baer, 2010; Ma

et al., 2011). When relying on data from publi-

cation records, network strength is inferred

from the patterns of coauthorship (i.e., the num-

ber of times two scientists have coauthored a

manuscript together serves as a measure of tie

strength by capturing frequency of interaction;

e.g., Fleming et al., 2007; McFadyen &

Cannella, 2004).

In a number of instances, scholars have used

the number of either weak or strong ties to assess

the effects of network strength on innovation

(e.g., Audretsch, Aldridge, & Sanders, 2011;

Perry-Smith, 2006; Zhou et al., 2009). However,

such measures confound the strength of a tie

(or a set of ties) with the number of ties (see

Baer, 2010). As a result, it is not possible to

conclusively determine whether any observed

effect is due to the strength of a set of relation-

ships or simply due to their number. Conse-

quently, we did not consider number of weak

or strong ties as appropriate indicators of net-

work strength and did not incorporate these

measures into our path-analytic model. How-

ever, given their frequent use in the literature,

we report the relevant effect sizes in our meta-

analysis for the sake of completeness (see two

bottom rows in Table 2 for relevant estimates).

The property of brokerage was formed by

clustering a number of measures typically used

in the literature. The measures most frequently

found in our sample were constraint, or the

extent to which people are invested in rela-

tionships that lead back to a single contact

(Burt, 1992) (constraint was recoded such

that higher levels indicate higher levels of

brokerage); efficiency, the proportion of non-

redundant contacts in a network (Burt, 1992);

and betweenness centrality, the extent to which

an individual lies on paths between connections

taking into account both direct and indirect ties

(Freeman, 1977). In addition, we considered

more idiosyncratic indicators of brokerage, as

long as the variable captured the extent to which

an individual served as a bridge between

otherwise disconnected parts of the network.

For example, Totterdell, Holman, and Hukin

(2008) used a measure of gatekeeper broker-

age that reflected the number of paths that

went through a focal individual and that con-

nected employees from different work groups

with the focal individual belonging to the

same group as the destination employee.

Network closure reflects the average con-

nectedness of the actors in the network, and is

Baer et al. 205



most commonly operationalized via network

density. Network density is a proportional

measure reflecting the number of existing ties

in the network out of the total number of pos-

sible ties (Wasserman & Faust, 1994). Alter

density, alternatively, captures the proportion

of existing ties out all possible ties among an

individual’s contacts, excluding any direct ties

(e.g., Moran, 2005).

Finally, consistent with previous research

(Jackson et al., 1995; Joshi & Roh, 2009), we

distinguished between task- and relations-

oriented diversity and clustered both under the

property of network diversity. Task-oriented

diversity captures diversity with respect to

attributes such as functional background,

departmental affiliation, or tenure (e.g., Baer,

2010; Perry-Smith, 2006); relations-oriented

diversity captures diversity with respect to attri-

butes such as sex, race/ethnicity, or cultural

background (e.g., Chua, 2011; Perry-Smith,

2002). Both task- and relations-oriented diver-

sity were most commonly operationalized via

Blau’s (1977) index of heterogeneity. In

addition, some studies measured diversity by

simply counting the number of different cate-

gories present among an individual’s contacts.

For example, Paruchuri (2010) measured diver-

sity via the count of unique areas of expertise

among a focal inventor’s collaborators.

Innovation. Consistent with our definition of

innovation, we included any study that offered

a measure of idea generation, execution, or

both, excluding studies focusing on the sharing

or transfer of knowledge or the diffusion of

innovations without considering how ideas and

knowledge were developed and put into prac-

tice in the first place (e.g., Chang & Harrington,

2005; Reagans & McEvily, 2003; Weenig,

1999). Innovation was typically measured in

one of two ways—via subjective self/others’

evaluations of a person’s innovation (e.g.,

Moran, 2005; Rodan & Galunic, 2004) or via

more objective indicators, such as the number

of patents, publications, and forward citations

obtained or the prizes awarded for certain

innovative achievements (e.g., Fleming et al.,

2007; Liao, 2011).

Table 2. Meta-analytic correlations between network properties and innovationa

Network properties k N r SDρ [95% CI] [80% CV] Fail safe k

Network size 26 463,601 .29*** .17 [.21, .36] [.07, .51] 348Network strength 14 79,375 .18*** .05 [.13, .22] [.11, .24] 109Brokerage 24 362,141 .33*** .12 [.27, .39] [.17, .49] 371

Constraint 7 144,827 .17*** .08 [.06, .27] [.06, .27] 51Efficiency 6 2,870 .55*** .33 [.29, .73] [.13, .96] 158Betweenness centrality 11 17,190 .34** .30 [.12, .53] [–.05, .73] 175Other 6 199,037 .22*** .02 [.16, .28] [.19, .25] 59

Closure 12 255,428 –.10þ .10 [–.19, .00] [–.23, .03] —Overall network density 8 18,428 –.03 .16 [–.23, .18] [–.23, .18] —Alter density 5 237,210 –.06* .03 [–.12, .00] [–.13, .00] 2

Network diversity 9 223,912 .24*** .11 [.14, .33] [.10, .38] 99Functional 7 223,630 .27*** .11 [.16, .37] [.13, .41] 86Social category 3 379 .09 .11 [–.07, .24] [–.05, .23] —

Number of strong ties 8 1,273 .23** .23 [.07, .39] [–.06, .52] 85Number of weak ties 7 731 .12 .42 [–.22, .44] [–.41, .66] —

ak ¼ number of studies, N ¼ aggregated sample size, r ¼ mean estimate of corrected population correlation, SDρ¼estimated standard deviation of corrected population correlation, 95% CI ¼ 95% confidence interval, 80% CV ¼ 80%credibility interval, Fail safe k ¼ number of unpublished studies with null effect sizes that would be needed to reduce theeffect size to nonsignificance.þp < .10; *p < .05; **p < .01; ***p < .001.




To be included, subjective indicators of

innovation had to capture idea generation,

execution, or both. Examples include respon-

dents’ self-assessment of their involvement in

a set of innovations (e.g., Obstfeld, 2005),

supervisor ratings of the extent to which

employees had developed ideas for new prod-

uct or service offerings (e.g., Baer, 2010;

Perry-Smith, 2006), or judge/supervisor rat-

ings of the creativity of individuals’ ideas or

proposals (e.g., Burt, 2004; Chua, 2011). In

selected circumstances, we also included mea-

sures that combined items on innovation with

those of general performance. This was done

when there was explicit evidence that the

results for innovation were not affected by the

inclusion of items on performance or when

performance in the domain under investigation

itself required the development and implemen-

tation of new ideas (e.g., Mehra et al., 2001).

In addition, we also included studies that

used more objective indicators, such as patent

counts, to gauge innovation. Patents reflect

original contributions or novel recombination

of existing knowledge and are frequently used

to measure individual innovation (Griliches,

1990; Pavitt, 1985). In applying for a patent,

inventors must prove the novelty and utility to

an examiner, providing support for the inno-

vativeness of the claim. Similarly, we con-

sidered publication counts—an indicator often

used to assess scientific productivity (e.g.,

Azoulay, Ding, & Stuart, 2009; Ding, Levin,

Stephan, & Winkler, 2010)—as a measure of

the extent to which a person had generated and

contributed a new idea to a domain of scien-

tific inquiry. In cases in which publication

counts were differentiated by author position,

we included those counts that most directly

reflected a person’s innovation. For example, in

the biomedical sciences, last-authored papers are

more reflective of individual innovation than

first-authored papers as the last author is typi-

cally the one who determines the research

question to be investigated and is responsible

for obtaining funding (e.g., McFadyen et al.,

2009). We considered both unweighted as well

as weighted counts of patents and publications

as weighting provides additional information

about the novelty and usefulness of an inno-

vation (Frey & Rost, 2010; Stephan & Levin,

1991). Journal impact factors are typically

used to weight publication counts (e.g.,

McFadyen & Cannella, 2004); citations are

typically used to weight patent counts (e.g.,

Lee, 2010). Finally, we also included measures

of forward citations—citations subsequently

referencing a previous patent or publication—

as objective indicators of innovation. Forward

citations are frequently considered a measure

of whether an idea provides a useful and reliable

foundation for future insights (Frey & Rost,

2010; Reedijk, 1998), reflecting both its novelty

and usefulness in action (Albert, Avery, Narin,

& McAllister, 1991; Lanjouw & Schankerman,

2004; Trajtenberg, 1990).

Meta-analytic procedure

Following procedures by Hunter and Schmidt

(2004), we corrected observed correlations for

measurement and sampling error. To correct

for measurement unreliability in self and other

ratings of innovation, we used Cronbach’s

alpha coefficients reported in the studies.

Whenever a study did not provide the neces-

sary reliability estimates, we used the average

of the available reliability estimates from

other studies in our sample that reported on

the same type of correlation. To avoid over-

correction, we did not adjust the objective

indicators of innovation, such as the number

of patents, publications, or forward citations

for measurement unreliability.

To maintain independence of effect sizes,

we combined correlations from studies that

provided more than one correlation per network

feature (i.e., size, strength, constraint, effi-

ciency, betweenness centrality, overall network

density, alter density, task-oriented diversity,

relations-oriented diversity). For example, Burt

(2004) reported correlations between network

Baer et al. 207



constraint and multiple indicators of innova-

tion. In such and other cases, we combined

the various correlations into one overall effect

size using the procedure outlined by Rosenthal

and Rubin (1986). In addition, because we clus-

tered certain network features into categories

(i.e., size, strength, brokerage, closure, diver-

sity), it is possible that a single study contri-

butes more than one correlation to the same

property. For instance, Liu (2011) reported cor-

relations between innovation and two indicators

of brokerage: efficiency and betweenness cen-

trality. To ensure independence, we again used

Rosenthal and Rubin’s (1986) formula to com-

bine correlations, such that a study may only

contribute one effect size per network category.

We used a random effects model correcting

for sampling error by weighting each study’s

effect size by its sample size (Hunter &

Schmidt, 2002). A random effects model allows

the population effect size to vary between

studies and protects from Type I error rate

inflation. We also computed the 95% confi-

dence interval (CI) and the 80% credibility

interval (CV) around the sample-weighted

mean correlation. Confidence intervals provide

an estimate of the variability around the esti-

mated average correlation due to sampling

error. Credibility intervals provide an estimate

of the variability in the distribution of the

average correlation and constitute a test of

heterogeneity (Hunter & Schmidt, 2004;

Whitener, 1990). Typically, researchers calcu-

late a Q-statistic to test for heterogeneity in

observed correlations across studies (Hedges

& Olkin, 1985). However, when there is large

variation in sample sizes across studies, the

Q-statistic may be inflated (Hunter & Schmidt,

2004). The sample sizes in our study ranged

from 34 to 183,204 and, therefore, the

Q-statistics may not accurately reflect whether

true heterogeneity exists. Consequently, we

inspected the credibility intervals to assess

heterogeneity in a given relation. Previous

research suggests that a random-effects model

provides more accurate estimates than fixed-

effects models when there is heterogeneity

(Cheung & Chan, 2005; Overton, 1998).

Meta-analytic path analysis

To test our overall model, we conducted a

meta-analytic path analysis, which requires

estimates of the correlations between all vari-

ables involved in the analysis (e.g., Shadish,

1996; Viswesvaran & Ones, 1995). Based on

the studies included in our sample, however,

we were able to construct only a partial

meta-analytic correlation matrix. Given that

there is no previous meta-analysis that, at least

to our knowledge, has examined the magni-

tude of relations between the various network

features considered in the present study, we

conducted a second meta-analysis to obtain

estimates of the missing correlations and to

supplement the 16 studies in our sample that

reported correlations. To ensure that these

studies represented the same population of

studies as those included in our primary

meta-analysis, we searched the top journals

of those scientific areas represented in our

sample (management, psychology, sociology,

political science) and those journals that

frequently publish relevant social network

studies. These journals included Academy of

Management Journal, Administrative Science

Quarterly, American Journal of Political

Science, American Journal of Sociology,

American Sociological Review, Journal of

Applied Psychology, Management Science,

Organization Science, Research Evaluation,

Research Policy, Scientometrics, Social Net-

works, and Strategic Management Journal. A

search using a similar set of keywords as in our

initial search yielded an additional 1,034 studies.

As before, we only included studies at the

individual level of analysis containing original

data. Second, to be included, a study had to

report at least one correlation between two of

the network properties included in the previ-

ous coding scheme. Third, we only included

studies that relied on actual network data.




Finally, included studies had to report sample

sizes and either a correlation coefficient or

enough information to calculate a correlation

coefficient. These criteria resulted in the

inclusion of an additional 38 studies.

We used the same meta-analytic procedures

(Hunter & Schmidt, 2004) to estimate the rele-

vant correlations between the five network

properties. Consistent with our previous

approach, we did not correct network measures

for unreliability. We used the created correla-

tion matrix to run a path analysis, using

maximum-likelihood estimation. Means and

standard deviations for each variable were set

to 0 and 1, respectively. The number of obser-

vations was determined by calculating the har-

monic mean of the sample sizes (N ¼ 32,541)

from the meta-analytic estimates (Viswesvaran

& Ones, 1995). To evaluate model fit, we used

established fit indices—chi-square (w2), the

comparative fit index (CFI), the root-mean-

square error of approximation (RMSEA), the

standardized root-mean square residual

(SRMR), and Akaike’s Information Criterion

(AIC). Commonly recommended cutoff values

were used to evaluate fit (CFI > .90, RMSEA

< .08, and SRMR < .10; Kline, 2005).

Results

Meta-analytic results

Table 2 summarizes the meta-analytic relations

between the five network properties (as well as

their various operationalizations) and innova-

tion. Hypothesis 1 predicted a positive relation

between network size and innovation. Consistent

with our expectation, there was a significant,

positive association between the two variables

(r ¼ .29, p < .001, k ¼ 26, 95% CI ¼ [.21,

.36]). Thus, Hypothesis 1 was supported.

Hypothesis 2 made competing predictions

regarding the nature of the link between

network strength and innovation, suggesting

that strength could have either a positive

(Hypothesis 2a) or a negative association with

innovation (Hypothesis 2b). Consistent with

Hypothesis 2a, there was a significant, positive

relation between the two variables (r¼ .18, p <

.001, k ¼ 14, 95% CI ¼ [.13, .22]). Therefore,

there was support for Hypothesis 2a; Hypoth-

esis 2b was not supported.3

Hypothesis 3 predicted that brokerage and

innovation would be positively related. This

hypothesis was supported. Correlations both for

overall brokerage (r ¼ .33, p < .001, k ¼ 24,

95% CI ¼ [.27, .39]) and for each of the

different operationalizations of brokerage—

constraint (r ¼ .17, p < .001, k ¼ 7, 95% CI ¼[.06, .27]), efficiency (r ¼ .55, p < .001, k ¼ 6,

95% CI ¼ [.29, .73]), betweenness centrality

(r ¼ .34, p < .01, k ¼ 11, 95% CI ¼ [.12, .53]),

and the ‘‘other’’ category (r¼ .22, p < .001, k¼6, 95% CI ¼ [.16, .28]) were positive and

statistically significant.

Hypothesis 4 made competing predictions

regarding the nature of the link between clo-

sure and innovation, suggesting that closure

may either have a positive (Hypothesis 4a) or a

negative relation with innovation (Hypothesis

4b). Although there was a significant, negative

relation between alter density and innovation

(r ¼ �.06, p < .05, k ¼ 5, 95% CI ¼ [�.12,

.00]), overall network density (r ¼ �.03, k ¼8, 95% CI ¼ [�.23, .18]), as well as overall

closure (r ¼ �.10, k ¼ 12, 95% CI ¼ [�.19,

.00]) exhibited nonsignificant relations with

innovation. Thus, there was only weak support

for Hypothesis 4b; Hypothesis 4a was not

supported.

Hypothesis 5 proposed a positive association

between network diversity and innovation.

Consistent with our expectation, there was a

significant, positive correlation between the

two variables, both for overall network diver-

sity (r ¼ .24, p < .001, k ¼ 9, 95% CI ¼ [.14,

.33]) and when operationalized via task-

oriented diversity (r ¼ .27, p < .001, k ¼ 7,

95% CI ¼ [.16, .37]). However, relations-

oriented diversity did not relate significantly

to innovation (r ¼ .09, k ¼ 3, 95% CI ¼[�.07, .24]). Thus, there was good support for

Baer et al. 209



Hypothesis 5, particularly when diversity was

focused on task-oriented characteristics, such

as expertise and tenure.4,5

Path-analytic results

In Hypotheses 6 through 12 we predicted that

the effects of network size and strength on

innovation would be fully mediated by

brokerage, closure, and diversity. We tested the

proposed model (see Figure 1) via path analysis

inputting the meta-analytic correlation matrix

displayed in Table 3. As shown in Table 4, the

model fit of the hypothesized model was not

satisfactory, w2 (6) ¼ 2,356.32, CFI ¼ .93,

RMSEA ¼ .11, SRMR ¼ .05, AIC ¼2,386.32. All paths, however, were significant

and in the predicted direction.

Given that network strength has been sug-

gested to exert positive effects on innovation

but our model captured and revealed only its

negative, indirect effects via brokerage and

closure, we explored the possibility that the

positive effects on innovation ascribed to

network strength may operate directly rather

than indirectly. Hence, we included a direct

path from network strength to innovation. As

shown in Table 4, this alternative model had

excellent fit, w2 (5) ¼ 442.13, CFI ¼ .99,

RMSEA ¼ .05, SRMR ¼ .02, AIC ¼ 474.12

and fit the data significantly better than the

hypothesized model, Dw2 (1) ¼ 1,914.19, p <

.01. It is this model on which all tests of direct

and indirect effects are based on (see Figure 2).

Figure 2 displays the standardized coeffi-

cients for our final path model. Consistent with

Hypotheses 6 and 8, respectively, network size

had a significant, positive effect on brokerage

(b ¼.70, p < .01) and a significant, negative

effect on closure (b ¼ �.19, p < .01). In

addition and consistent with Hypothesis 10,

there was a significant, positive direct effect of

size on diversity (b ¼.24, p < .01). Also as

expected, network strength had a significant,

positive effect on closure (b ¼.24, p < .01)

and a significant, negative effect on brokerage

(b ¼ �.06, p < .01), supporting both Hypoth-

eses 7 and 9, respectively.

Confirming Hypothesis 11, the effect of

network size on innovation was completely

mediated by brokerage, closure, and diversity

(b ¼ .27, p < .01). However, failing to support

Hypothesis 12, the effect of strength on inno-

vation was only partially mediated by broker-

age and closure (b ¼ �.04, p < .01), as there

was a significant, positive and direct effect of

strength on innovation (b ¼.23, p < .01) over

and above the indirect effects. In sum, these

results suggest that positional (e.g., brokerage),

structural (e.g., density), and nodal (e.g., func-

tional background) network features fully or

partially mediate the effects of network size and

strength on innovation (see Table 5).

Discussion

The purpose of the present study was to system-

atically review and integrate the fragmented lit-

erature on the link between social networks and

individual innovation using meta-analytic meth-

ods. Building on previous theory and research,

we offered an approach to clustering the variety

of different ways in which network characteris-

tics have been operationalized in the past into

five general properties: size, strength, broker-

age, closure, and diversity. This approach

allowed us to succinctly summarize and com-

pare the effects on individual innovation asso-

ciated with each of the five properties. Our

analysis suggests that brokerage has the stron-

gest positive relation with innovation, followed

by network size, network diversity (particu-

larly, task-oriented diversity), and strength all

of which exhibited weak-to-moderate relations

with individual innovation. Closure was nega-

tively related to innovation, albeit only margin-

ally so.

In addition to using meta-analysis to esti-

mate the true population effect sizes associated

with each of the network properties, we also

offered a path-analytic integration of these

properties, specifying the direct and indirect




Tab

le3.

Met

a-an

alyt

icco

rrel

atio

ns

for

pat

han

alys

is.a

12

34

5

Var

iable

r[9

5%

CI]

r[9

5%

CI]

r[9

5%

CI]

r[9

5%

CI]

r[9

5%

CI]

kN

kN

kN

kN

1.N

etw

ork

size

——

——

2.N

etw

ork

stre

ngt

h.0

1[–

.05,.0

8]

14

62,9

36

3.Bro

kera

ge.7

0[.61,.7

8]

–.0

5[–

.29,.1

9]

14

312,7

75

32,9

32

4.C

losu

re–.1

9[–

.54,.2

2]

.24

[.13,.3

5]

–.1

2[–

.80,.7

0]

14

187,6

58

11

60,7

92

4186,0

68

5.N

etw

ork

div

ersi

ty.2

4[.18,.3

0]

–.0

7[–

.27,.1

3]

.21

[–.1

7,.5

3]

–.1

1[–

.39,.1

9]

18

229,2

31

13

58,4

52

4117,2

93

856,3

35

6.In

nova

tion

.29

[.21,.3

6]

.18

[.13,.2

2]

.33

[.27,.3

9]

–.1

0[–

.19,.0

0]

.24

[.14,.3

3]

18

463,6

01

14

79,3

75

24

362,1

41

12

255,4

28

9223,9

12

a r¼

mea

nes

tim

ate

ofco

rrec

ted

popula

tion

corr

elat

ion,95%

CI¼

95%

confid

ence

inte

rval

,k¼

num

ber

ofst

udie

s,N¼

aggr

egat

edsa

mple

size

.

Tab

le4.

Fit

stat

istics

for

alte

rnat

ive

model

s.a

w2df

Dw2

CFI

RM

SEA

SRM

RA

IC

Theo

retica

lm

odel

(Fig

ure

1)

2356.3

26

.93

.11

.05

2386.3

2A

lter

nat

ive

model

b(F

igure

2)

442.1

35

1914.1

9**

c.9

9.0

5.0

2474.1

2

a N¼

32,5

41;b

adds

dir

ect

pat

hfr

om

net

work

stre

ngt

hto

innova

tion;

cm

odel

fitco

mpar

edw

ith

pre

vious

model

.*p

<.0

5;**

p<

.01.

211 at DUQUESNE UNIV on August 26, 2015opr.sagepub.comDownloaded from


effects through which they impact innovation.

Specifically, we offered a comprehensive

model highlighting that the distal antecedents

of network size and strength impact innovation

through the more proximal features of broker-

age, closure, and network diversity. Relying

on meta-analytic path analysis, we tested this

model and found general support for its valid-

ity. Consistent with our logic, network size and

strength impacted innovation through a web of

relations with the more proximal positional,

structural, and nodal features of brokerage, clo-

sure, and diversity. These more proximal net-

work properties, in turn, served to mediate the

effects of network size and strength on innova-

tion either fully, as in the case of network size,

or partially, as in the case of network strength.

The direct positive effect of network

strength on innovation—which was further

validated by the finding that the number of

strong ties was positively related to innovation

in contrast to the number of weak ties, which

exhibited a nonsignificant relation with inno-

vation—highlights an interesting tension in

the way network strength shapes innovation

at work. On the one hand, strong ties are likely

to result in closure and reduced brokerage

thereby limiting access to individuals with

heterogeneous information and knowledge

sets, which should ultimately reduce the

potential to develop novel, useful ideas and

solutions. On the other hand, the trust and

solidarity associated with strong ties is not

only likely to be helpful in mobilizing the

support necessary to act upon one’s ideas

(Baer, 2012; Kanter, 1988), but strong ties are

also likely to be helpful when the knowledge to

be transmitted is more tacit in nature (Hansen,

1999), such as the knowledge involved in

formulating problems. Indeed, work by Cross

and Sproull (2004) has shown that tie strength

relates negatively to knowledge about solu-

tions but positively to knowledge about prob-

lem reformulation and validation of one’s

ideas. In fact, the results of our path analysis

suggest that the negative, indirect effect of

network strength on innovation is relatively

small as compared to its positive, direct effect.

Thus, when the outcome organizations care

Network size

ClosureR2 = .10

Network diversityR2 = .06

Innovation R2 = .19

.70**

Network strength

BrokerageR2 = .50

.24**

−.06**

.19**

.29**

−.10**

.24**

.23**

.01*

−.19**

Figure 2. Final model of network effects on innovation.a

aStandardized coefficients in the final model (alterative Model 2) are presented; N ¼ 32,541.

*p < .05; **p < .01.

Table 5. Direct, indirect, and total effects ofnetwork size and strength on innovationa

VariableDirecteffects

Indirecteffects

Totaleffects

Network size — .27** .27**Network strength .23** –.04** .19**

aStandardized coefficients in the final model (alterativemodel; Figure 2) are presented; harmonic N ¼ 32,541.*p < .05; **p < .01.




about requires the development and implemen-

tation of fresh ideas, the trust and solidarity

benefits associated with tie strength may out-

weigh the costs associated with reduced access

to heterogeneous information and knowledge

sets. This conclusion is echoed by a number of

recent studies, all of which point towards the

conclusion that strong ties seem to be more

beneficial to innovation than weak ties (e.g.,

Kijkuit & van den Ende, 2010), particularly

when embedded in a network structure that

allows individuals to serve as brokers between

disconnected others combining the best of

both worlds—access to heterogeneous knowl-

edge sets and trust to key communication and

exchange partners (McFadyen et al., 2009;

Rost, 2011).

The argument often advanced in the extant

literature that the benefits associated with

network closure enhance innovation was not

supported—closure exhibited a significantly

negative effect on innovation. It appears that

the benefits of dense network structures for

coordinating action and for garnering support

are outweighed by the downsides of such

structures—redundant access to information,

reduced willingness to take risks, and the

adherence to existing norms and expectations.

Alternatively, it is possible that the positive

direct effect of strength that we observed

captured the effect of trust and solidarity on

innovation. This interpretation is consistent

with work by Moran (2005) who observed that

network strength (the closeness and trust

associated with interpersonal relationships)

was a significant predictor of managerial

innovation whereas the density of the network

was not. Thus, rather than closing one’s net-

work, building deep and meaningful dyadic

relationships seems to be the key conduit for

obtaining the trust and support necessary for

engaging the innovative process.

One major contribution of the present study is

that we partially reconcile the two perspectives

on the effects of networks on innovation that

often have been contrasted as polar opposites—

brokerage versus closure. Although it is true that

brokerage had uniformly positive effects and

closure had uniformly negative effects on inno-

vation, network strength—a direct antecedent to

closure—did exhibit both positive and negative

effects as reflected in our competing Hypotheses

2a and 2b. These findings hint at the potential

value not only of integrating the two perspec-

tives—the effects of closure were significant

while controlling for the effects of broker-

age—but also of acknowledging that given the

dual nature of innovation, certain network

properties, such as strength can have positive

and negative effects on innovation simultane-

ously. Although network strength had a net

positive effect on innovation, it did produce

closure, which then undermined innovation.

This finding challenges previously held

assumptions and reminds scholars to embrace

the complexities of the relations between

social network properties and innovation in

their theorizing and measurement.

Practical implications

Our results offer some guidance as to how

individuals should craft their networks.

Expanding the size of one’s network is impor-

tant as adding people to one’s social spheres is

likely to increase the opportunities for broker-

age and/or to reduce closure and, at the same

time, makes it more likely that one gets exposed

to and has access to a greater variety of people

with different information and knowledge sets.

Naturally, given the demands that an ever-

increasing web of contacts may place on the

focal individual—both in terms of cognitive and

time resources—there will be diminishing return

to this strategy, as has been well documented

(Baer, 2010; McFadyen & Cannella, 2004). The

key then is to nurture a manageable number of

direct relationships to what Uzzi and Dunlap

(2005) have termed superconnectors—powerful

brokers who are likely to provide us with access

to a wide array of different social circles. Qual-

ity, however, should not be sacrificed for

Baer et al. 213



quantity. Network strength is a powerful force

in boosting innovation and thus forging ahead

with building a vast network while not invest-

ing in each relationship is likely to provide rel-

atively limited returns. Weak ties may be

useful for learning about new ideas (e.g.,

Perry-Smith, 2006), but strong ties are instru-

mental in identifying and formulating prob-

lems in the first place (e.g., Cross & Sproull,

2004), transmitting more sensitive and tacit

knowledge often required for innovative ideas

(e.g., Hansen, 1999), and for building support

to realize the idea (e.g., Baer, 2012). Although

we acknowledge that strong ties may be most

potent when embedded in a network rich in

structural holes—a notion that we could not

test in the present study (see Limitations sec-

tion)—our findings corroborate the conclusion

by Rost (2011) that strong ties benefit innova-

tion independent of the structure in which they

are embedded.

Limitations and avenues forfuture research

Our study has a number of limitations that are

worth noting. First, in our theoretical and

empirical analysis we assumed that the rela-

tions between the various network features and

innovation are linear. However, ample research

suggests that a number of network characteris-

tics, including network size (e.g., Baer, 2010),

strength (e.g., McFadyen & Cannella, 2004;

Zhou et al., 2009), brokerage (e.g., Rost, 2011),

the extent to which individuals occupy posi-

tions between the core and periphery of the

network (e.g., Cattani & Ferriani, 2008), and

the extent to which networks exhibit properties

of a small world (e.g., Uzzi & Spiro, 2005)

exhibit inverted U-shaped relations to out-

comes, such as creativity and innovation.

Unfortunately, given the relatively few studies

that have examined each feature (and their non-

linear relations), we were not able to meta-

analytically review the potentially quadratic

nature of these relations and thus focused on the

linear effects exclusively. Nevertheless, we

recognize that many of the features examined

here may very well exhibit inverted U-shaped

relations with innovation. For example, we

highlighted the potential benefits of network

strength for innovation. However, as average tie

strength increases to the extent that networks

are populated mostly by friends who have

known each other for years and who communi-

cate on a daily basis, the costs associated with

increasing network strength may ultimately

outweigh its trust and solidarity benefits

(McFadyen & Cannella, 2004). Consequently,

as more studies examining these nonlinear rela-

tions become available, we encourage future

meta-analytic efforts to explicitly consider the

nonlinear nature of most of the associations

considered in the present paper.

Second, although we were not able to

examine the extent to which various network

features differentially impact the generative

aspect of innovation versus the refinement and

execution components—most studies in our

sample did not consider this distinction—there

is evidence to suggest that certain network

constellations may be more critical during

some phases of the innovation process than

others (e.g., Lavie & Drori, 2012). For

instance, work by Kijkuit and van den Ende

(2010) suggests that whereas larger networks

may be conducive to the generative phase of

the innovation process, networks of smaller

size may be more helpful during the latter

stages. In addition, whereas lack of closure

seems to be critical during idea generation,

during the latter stages of the innovation pro-

cess networks of greater density seem to be

relatively more helpful. Our ability to detect

such nuances was limited by the lack of

available studies and data. Thus, we encourage

future research to examine more system-

atically the coevolution of innovative ideas

and their associated networks to uncover the

differential effects that certain network fea-

tures may have over time and how individuals

manage these dynamics.




Next, many studies in our sample examined

ties that deliver informational benefits pri-

marily, or constructed their network measures

by aggregating across various tie types.

Naturally, focusing on one type of relationship

or aggregating across various types of ties may

obfuscate the fact that different types of rela-

tionships may exert different effects. A study

by Podolny and Baron (1997) nicely illustrates

this point. These authors showed that the effect

of brokerage on promotions was positive for

ties that served as conduits for information

but was negative for buy-in ties—ties that

conveyed expectations and identity. Given

these findings, Podolny and Baron (1997) con-

cluded, ‘‘the standard practice in network

research of aggregating disparate kinds of ties

when relating network structure to mobility

outcomes seems ill-conceived’’ (p. 689). We

echo this sentiment, as the different effects of

disparate kinds of ties are unlikely to be con-

fined to issues of mobility but are likely to

extend to other domains, such as innovation.

We thus would encourage future research to

more systematically consider the role that tie

type and the content conveyed through ties

may play in determining the effects of certain

network features on innovation.

Fourth, the focus of our path-analytic inte-

gration was on specifying and estimating the

indirect pathways through which basic net-

work features, such as size and strength may

affect innovation. In doing so, we have ignored

the role of potential moderators of these

relations. However, previous research has

highlighted that some network features may

serve to enhance the effects of certain other

network characteristics. For example, we

know that network strength provides more

substantial returns for knowledge creation and

innovation when strong ties are embedded in

networks rich in structural holes (McFadyen

et al., 2009; Rost, 2011). In addition, previous

work has suggested that brokerage is a more

powerful driver of individual innovation when

networks are composed of individuals with

heterogeneous information and knowledge

sets (Rodan & Galunic, 2004). In addition,

there is a growing body of evidence suggesting

that individual differences capturing people’s

proclivity to utilize the various resources

accessible through one’s social relationships

(Anderson, 2008; Baer, 2010; Zhou et al.,

2009), such as openness to experience or need

for cognition may serve to enhance the effects

of, for example, network size and strength on

innovation and related outcomes. However,

due to the relatively few studies that have con-

sidered network characteristics or personality

differences as potential moderators, we were

not able to test our mediating pathways for dif-

ferent levels of these moderating variables.

Future research may want to test our model

in different populations or environments that

systemically vary in personality or network

characteristics.

Fifth, there was a fair amount of hetero-

geneity in the observed relations as evidenced

by the sometimes relatively wide 80% cred-

ibility intervals. As a rule of thumb, if a cred-

ibility interval is either wide or contains zero,

the existing heterogeneity may be indicative of

the fact that moderators are operative (White-

ner, 1990). To ameliorate concerns regarding

the operation of moderators but given the fact

that only very few potential moderating vari-

ables were consistently reported across studies,

we examined the role of three methodological

artifacts as potential moderators. Specifically,

in addition to distinguishing between the type

of network data (egocentric network vs. com-

plete network) as reported in Endnote 3, we also

distinguished between subjective criterion

measures (self- and other-ratings) and objective

measures (patent/publication counts, citations;

Hulsheger et al., 2009), and coded for publi-

cation status (published vs. unpublished). None

of these methodological moderators, however,

accounted for significant amounts of hetero-

geneity in the reported relations. Given the

lack of results for the moderating variables

considered here, we would encourage scholars

Baer et al. 215



to identify and examine both methodological

and substantive moderating variables as they

continue to study the relations between network

properties on the one hand and innovation on

the other.

Finally, although our study provides an

important step forward in understanding the

various ways in which social network charac-

teristics impact innovation, our theoretical and

empirical analysis still relies on a number of

unobserved mechanisms. For example, we

argued that brokerage should boost innovation

due to informational and control benefits, yet

such benefits remained unobserved. Similarly,

network strength was found to directly impact

innovation presumably due to benefits asso-

ciated with trust and solidarity allowing for the

exchange of more sensitive information and

access to buy-in but again, we were not able to

verify the validity of these mechanisms. Most

studies do not directly examine the linkages

between network features and the types of

resources being accessed or transmitted via

the respective ties (for an exception, see

Anderson, 2008) and so our arguments about

the precise mechanisms remain somewhat

speculative. We believe that it is worth exam-

ining these mechanisms in more detail and we

encourage future research to do so.

Conclusion

The results of our meta-analysis show that

most network features examined over the past

two decades exhibit robust and mostly positive

correlations (with the exception of closure)

with individual innovation at work. Our path-

analytic integration suggests that two basic

features of social networks—the number and

average strength of ties—operate through the

positional and structural features of brokerage

and closure as well as through diversity to

impact innovation. While size (via brokerage,

closure, and diversity) exerts positive effects

on innovation only, network strength serves

as a double-edged sword—it fosters network

closure and inhibits brokerage, which constricts

access to diverse information and knowledge

sets but it also promotes trust and solidarity,

which boosts innovation with the net effect of

strength being positive.

Notes

1. Only four studies included measures with both

direct and indirect ties. To examine whether there

may be systematic differences in effect sizes asso-

ciated with network size based on direct versus

direct and indirect ties, we conducted a subgroup

analysis. No significant differences were found

(Qb ¼ 2.00, ns).

2. We did not set any inclusion criterion for the

reach of the indirect network as researchers typi-

cally do not specify or report cutoffs for the size

of the indirect network. Thus, we included all

studies that reported effect sizes for the relation

between network size and innovation. However,

we do recognize that there may be heterogeneity

among studies in terms of how far removed indi-

rect ties are from ego.

3. Inspection of Table 2 also revealed that the

number of strong ties had a positive relation to

innovation (r ¼ .23, p < .01, k ¼ 8, 95%

CI ¼ [.07, .39]), whereas the number of weak

ties was unrelated to innovation (r ¼ .12, k ¼7, 95% CI ¼ [�.22, .44]).

4. For the three properties (i.e., brokerage, closure,

diversity) for which we explicitly distinguished

between the different ways in which each prop-

erty had been measured, we conducted subgroup

analysis to see if the effect sizes varied within

each property. Qb indicates whether effect sizes

vary between subgroups (Hunter & Schmidt,

2004). There were no differences between the

different ways of measuring closure (Qb ¼ .99,

ns) and diversity (Qb ¼ 1.34, ns). However, there

were significant differences in the brokerage

category (Qb¼ 12.82, p < .05), driven by the sig-

nificant difference between those studies that

used efficiency to measure brokerage (r ¼ .55,

p < .001) and those that used either constraint

(r ¼ .17, p < .001) or more idiosyncratic mea-

sures of brokerage (r ¼ .22, p < .001).




5. We also tested whether there were any differences

in effect sizes between those studies that relied on

egocentric network data and those that were based

upon complete network data. Generally, there

were no differences across the various network

features, with the exception of network size

(Qb ¼ 6.30, p < .05)—the correlation between

network size and innovation in egocentric net-

works was significantly smaller (r ¼ .14, p <

.05) than the correlation in complete networks

(r ¼ .41, p < .001).

References

Adler, P. S., & Kwon, S. W. (2002). Social capital:

Prospects for a new concept. Academy of Man-

agement Review, 27, 17–40.

Ahuja, G. (2000). The duality of collaboration:

Inducements and opportunities in the formation

of interfirm linkages. Strategic Management

Journal, 21, 317–343.

Albert, M. B., Avery, D., Narin, F., & McAllister, P.

(1991). Direct validation of citation counts as

indicators of industrially important patents.

Research Policy, 2, 251–259.

Amabile, T. M. (1996). Creativity in context: Update

to the social psychology of creativity. Boulder,

CO: Westview Press.

Anderson, M. H. (2008). Social networks and the

cognitive motivation to realize network oppor-

tunities: A study of managers’ information

gathering behaviors. Journal of Organizational

Behavior, 29, 51–78.

*Anderson, T. (2006). In the middle of things: How

ego networks and context perceptions influence

individual creativity in work groups (Unpub-

lished doctoral dissertation). McGill University,

Montreal.

Aral, S., & van Alstyne, M. (2011). The diversity-

bandwidth trade-off. American Journal of Sociol-

ogy, 117, 90–171.

*Audretsch, D. B., Aldridge, T. T., & Sanders, M.

(2011). Social capital building and new business

formation: A case study in Silicon Valley. Inter-

national Small Business Journal, 29, 152–169.

Axtell, C. M., Holman, D. J., Unsworth, K. L., Wall,

T. D., Waterson, P. E., & Harrington, E. (2000).

Shopfloor innovation: Facilitating the sugges-

tion and implementation of ideas. Journal of

Occupational and Organizational Psychology,

73, 265–285.

*Azoulay, P., Ding, W., & Stuart, T. (2009). The

impact of academic patenting on the rate, quality

and direction of (public) research output. The

Journal of Industrial Economics, 57, 637–676.

*Baer, M. (2010). The strength-of-weak-ties per-

spective on creativity: A comprehensive exami-

nation and extension. Journal of Applied

Psychology, 95, 592–601.

*Baer, M. (2012). Putting creativity to work: The

implementation of creative ideas in organizations.

Academy of Management Journal, 55, 1102–1119.

Baer, M., & Frese, M. (2003). Innovation is not

enough: Climates for initiative and psychological

safety, process innovations, and firm perfor-

mance. Journal of Organizational Behavior, 24,

45–68.

Balkundi, P., Kilduff, M., Barsness, Z. I., & Michael,

J. H. (2007). Demographic antecedents and per-

formance consequences of structural holes in

work teams. Journal of Organizational Behavior,

28, 241–260.

*Beaudry, C., & Allaoui, S. (2012). Impact of

public and private research funding on scien-

tific production: The case of nanotechnology.

Research Policy, 41(9), 1589–1606.

*Beaudry, C., & Kananian, R. (2013). Follow the

(industry) money – The impact of science networks

and industry-to-university contracts on academic

patenting in nanotechnology and biotechnology.

Industry and Innovation, 20(3), 241–260.

Bian, Y. (1997). Bringing strong ties back in:

Indirect ties, network bridges, and job searches

in China. American Sociological Review, 62,

366–385.

*Bjork, J., Di Vincenzo, F., Magnusson, M., & Mas-

cia, D. (2011). The impact of social capital on

ideation. Industry and Innovation, 18(6), 631–647.

Blau, P. M. (1977). Inequality and heterogeneity: A

primitive theory of social structure. New York,

NY: Free Press.

Bouty, I. (2000). Interpersonal and interaction influ-

ences on informal resource exchanges between

Baer et al. 217



R&D researchers across organizational boundaries.


Brass, D. J. (1995). A social network perspective on

human resources management. In G. Ferris (Ed.),

Research in personnel and human resources

management (Vol. 13, pp. 39–79). Greenwich,

CT: JAI Press.

Burt, R. S. (1992). Structural holes: The social

structure of competition. Cambridge, MA: Har-

vard University Press.

Burt, R. S. (2000). The network structure of social

capital. In B. Staw & R. Sutton (Eds.), Research

in organizational behavior (Vol. 22, pp.

345–423). Greenwich, CT: JAI Press.

*Burt, R. S. (2004). Structural holes and good ideas.

American Journal of Sociology, 110, 349–399.

Burt, R. S. (2005). Brokerage and closure: An intro-

duction to social capital. New York, NY: Oxford

University Press.

Buskens, V., & van de Rijt, A. (2008). Dynamics of

networks if everyone strives for structural holes.


Campbell, D. T. (1960). Blind variation and selec-

tive retentions in creative thought as in other

knowledge processes. Psychological Review,

67, 380–400.

*Cannella, A. A., & McFadyen, M. A. (2011).

Changing the exchange: Lost, added, and

reunited. Working paper, W. P. Carey School of

Business, Arizona State University, Tempe, AZ.

Carter, W. C., & Feld, S. L. (2004). Principles relat-

ing social regard to size and density of personal

networks, with applications to stigma. Social

Networks, 26, 323–329.

Casanueva, C., & Gallego, A. (2010). Social capital

and individual innovativeness in university

research networks. Innovation: Management,

Policy & Practice, 12, 105–117.

*Cattani, G., & Ferriani, S. (2008). A core/periphery

perspective on individual creative performance:

Social networks and cinematic achievements in

the Hollywood film industry. Organization

Science, 19, 824–844.

Cerne, M., Nerstad, C. G. L., Dysvik, A., &

Skerlavaj, M. (2014). What goes around comes

around: Knowledge hiding, perceived

motivational climate, and creativity. Academy of

Management Journal, 57, 172–192.

Chang, M.-H., & Harrington, J. E. (2005). Discovery

and diffusion of knowledge in an endogenous

social network. American Journal of Sociology,

110, 937–976.

*Chen, C.-S., & Liu, C.-H. (2012). Impact of net-

work position and knowledge diversity on knowl-

edge creation: The empirical setting of research

communities. Canadian Journal of Administra-

tive Sciences/Revue Canadienne Des Sciences

de l’Administration, 29(4), 297–309.

*Chen, L., & Gable, G. G. (2013). Larger or broader:

Performance implications of size and diversity of

the knowledge worker’s egocentric network.

Management and Organization Review, 9(1),

139–165.

Cheung, M. W.-L., & Chan, W. (2005). Meta-analytic

structural equation modeling: A two-stage

approach. Psychological Methods, 10, 40–64.

*Chiu, S.-C. (2013). Position, experience, and

knowledge creation: A longitudinal case study

of Chinese academic networks. Social Behavior

& Personality, 41(6), 1009–1018.

*Chua, R. Y. (2011). Innovating at the world’s cross-

roads: How multicultural networks promote crea-

tivity. Working Paper, Harvard Business School.

Clegg, C., Unsworth, K., Epitropaki, O., & Parker, G.

(2002). Implicating trust in the innovation pro-

cess. Journal of Occupational & Organizational

Psychology, 75, 409–422.

Coleman, J. S. (1988). Social capital in the creation

of human capital. American Journal of Sociology,

94, 95–120.

Coleman, J. S. (1990). Foundations of social theory.

Cambridge, MA: Harvard University Press.

Cross, R., & Parker, A. (2004). The hidden power of

social networks: Understanding how work really

gets done in organizations. Cambridge, MA:

Harvard Business Press.

Cross, R., & Sproull, L. (2004). More than an

answer: Information relationships for action-

able knowledge. Organization Science, 15,

446–462.

Damanpour, F. (1991). Organizational innovation: A

meta-analysis of effects of determinants and




moderators. Academy of Management Journal,

34, 555–590.

Damanpour, F., & Schneider, M. (2006). Phases of

the adoption of innovation: Effects of environ-

ment, organization and top managers. British

Journal of Management, 17, 215–236.

*Ding, W. W., Levin, S. G., Stephan, P. E., &

Winkler, A. E. (2010). The impact of informa-

tion technology on academic scientists’ produc-

tivity and collaboration patterns. Management

Science, 56, 1439–1461.

*Fleming, L., Mingo, S., & Chen, D. (2007). Colla-

borative brokerage, generative creativity, and

creative success. Administrative Science Quar-

terly, 52, 443–475.

*Fleming, L., & Waguespack, D. M. (2007). Broker-

age, boundary spanning, and leadership in open

innovation communities. Organization Science,

18, 165–180.

Freeman, L. C. (1977). A set of measures of cen-

trality based on betweenness. Sociometry, 40,

35–41.

Freeman, L. C. (1979). Centrality in social networks

conceptual clarification. Social Networks, 1,

215–239.

Frese, M., Teng, E., & Wijnen, C. J. (1999). Helping

to improve suggestion systems: Predictors of

making suggestions in companies. Journal of

Organizational Behavior, 20, 1139–1155.

Frey, B. S., & Rost, K. (2010). Do rankings reflect

research quality? Journal of Applied Economics,

13, 1–38.

Gabbay, S. M., & Leenders, R. T. A. (2001). Social

capital of organizations. New York, NY: JAI Press.

*Gonzalez-Brambila, C. N., Veloso, F. M., & Krac-

khardt, D. (2013). The impact of network

embeddedness on research output. Research

Policy, 42(9), 1555–1567.

Granovetter, M. (1973). The strength of weak ties.


Granovetter, M. (1983). The strength of weak ties: A

network theory revisited. Sociological Theory, 1,

201–233.

Granovetter, M. (1985). Economic action and

social structure: The problem of embeddedness.


Griliches, Z. (1990). Patent statistics as economic

indicators: A survey. Journal of Economic Liter-

ature, 28, 1661–1707.

Hallinan, M. T., & Hutchins, E. E. (1980). Structural

effects on dyadic change. SocialForces, 59, 225–245.

Hammond, M. M., Neff, N. L., Farr, J. L., Schwall,

A. R., & Zhao, X. (2011). Predictors of

individual-level innovation at work: A meta-anal-

ysis. Psychology of Aesthetics, Creativity, and the

Arts, 5, 90–105.

Hansen, M. T. (1999). The search-transfer problem:

The role of weak ties in sharing knowledge across

organization subunits. Administrative Science

Quarterly, 44, 82–111.

Hargadon, A., & Sutton, R. I. (1997). Technology

brokering and innovation in a product develop-

ment firm. Administrative Science Quarterly,

42, 716–749.

Harrison, D. A., & Klein, K. J. (2007). What’s the

difference? Diversity constructs as separation,

variety, or disparity in organizations. Academy

of Management Review, 32, 1199–1228.

Hedges, L. V., & Olkin, I. (1985). Statistical methods

for meta-analysis. New York, NY: Academic

Press.

Holland, P. W., & Leinhardt, S. (1970). A method for

detecting structure in sociometric data. American

Journal of Sociology, 76, 492–513.

Hulsheger, U. R., Anderson, N., & Salgado, J. F.

(2009). Team-level predictors of innovation at

work: A comprehensive meta-analysis spanning

three decades of research. Journal of Applied

Psychology, 94, 1128–1145.

Hunter, J. E., & Schmidt, F. L. (2002). Fixed

effects vs. random effects meta-analysis mod-

els: Implications for cumulative research knowl-

edge. International Journal of Selection and

Assessment, 8, 275–292.

Hunter, J. E., & Schmidt, F. L. (2004). Methods of

meta-analysis: Correcting error and bias in

research findings (2nd ed.). Thousand Oaks,

CA: Sage.

Ibarra, H. (1993). Network centrality, power, and

innovation involvement: Determinants of tech-

nical and administrative roles. Academy of

Management Journal, 36, 471–501.

Baer et al. 219



Ibarra, H. (1995). Race, opportunity, and diversity

of social circles and managerial networks.


Jackson, S. E., May, K. E., & Whitney, K. (1995).

Understanding the dynamics of diversity in

decision-making teams. In R. Guzzo & E. Salas

(Eds.), Team effectiveness and decision making

in organizations (pp. 204–261). San Francisco,

CA: Jossey-Bass.

Joshi, A., & Roh, H. (2009). The role of context in work

team diversity research: A meta-analytic review.


Kanter, R. (1988). When a thousand flowers bloom:

Structural, collective and social conditions for

innovation in organization. In B. Staw & L.

Cummings (Eds.), Research in organizational

behavior (Vol. 10, pp. 169–211). Greenwich,

CT: JAI Press.

*Keri, S. (2011). Solitary minds and social capital:

Latent inhibition, general intellectual functions

and social network size predict creative achieve-

ments. Psychology of Aesthetics, Creativity, and

the Arts, 5, 215–221.

Kijkuit, B., & van den Ende, J. (2010). With a little

help from our colleagues: A longitudinal study

of social networks for innovation. Organization

Studies, 31, 451–479.

Kilduff, M., & Tsai, W. (2003). Social networks and

organizations. Thousand Oaks, CA: Sage.

Kline, R. B. (2005). Principles and practice of

structural equation modeling. New York, NY:

The Guilford Press.

Krackhardt, D. (1992). The strength of strong ties: The

importance of philos in organizations. In N. Nohria

& R. Eccles (Eds.), Networks and organizations:

Structure, form, and action (pp. 216–239). Boston,

MA: Harvard Business School Press.

Krackhardt, D. (1999). The ties that torture: Simme-

lian tie analysis in organizations. In S. Andrews &

D. Knoke (Eds.), Research in the sociology of

organizations (Vol. 16, pp. 183–210). Greenwich,

CT: JAI Press.

*Kratzer, J., & Lettl, C. (2008). A social network

perspective of lead users and creativity: An

empirical study among children. Creativity and

Innovation Management, 17, 26–36.

Lanjouw, J. O., & Schankerman, M. (2004). Patent

quality and research productivity: Measuring

innovation with multiple indicators. The Eco-

nomic Journal, 114, 441–465.

Lavie, D., & Drori, I. (2012). Collaborating for

knowledge creation and application: The case of

nanotechnology research programs. Organization

Science, 23, 704–724.

*Lee, B. (2009, August). Effects of network posi-

tion on patent pending in pharmaceutical

industry. Paper presented at the annual meeting

of the American Sociological Association, San

Francisco, CA.

*Lee, J. (2010). Heterogeneity, brokerage, and inno-

vative performance: Endogenous formation of

collaborative inventor networks. Organization

Science, 21, 804–822.

Lee, J., & Kim, S. (2011). Exploring the role of

social networks in affective organizational com-

mitment: Network centrality, strength of ties,

and structural holes. The American Review of

Public Administration, 41, 205–223.

Leenders, R. T. A., van Engelen, J. M., & Kratzer, J.

(2003). Virtuality, communication, and new

product team creativity: A social network per-

spective. Journal of Engineering and Technology

Management, 20, 69–92.

Levin, D. Z., & Cross, R. (2004). The strength of

weak ties you can trust: The mediating role of

trust in effective knowledge transfer. Manage-

ment Science, 50, 1477–1490.

*Li, E. Y., Liao, C. H., & Yen, H. R. (2013). Co-

authorship networks and research impact: A

social capital perspective. Research Policy,

42(9), 1515–1530.

*Liao, C. H. (2011). How to improve research qual-

ity? Examining the impacts of collaboration

intensity and member diversity in collaboration

networks. Scientometrics, 86, 747–761.

*Liu, C.-H., Chiu, S., & Chiu, C. (2010). Intranet-

work relationships, creativity, knowledge diver-

sification, and network position. Social

Behavior and Personality, 38, 1173–1190.

*Liu, C.-H. (2011). The effects of innovation alli-

ance on network structure and density of cluster.

Expert Systems with Applications, 38, 299–305.




*Liu, C.-H., & Lin, J.-Y. (2012). Social relationships

and knowledge creation: The mediate of critical

network position. The Service Industries Journal,

32(9), 1469–1488.

Louch, H. (2000). Personal network integration:

Transitivity and homophily in strong-tie relations.

Social Networks, 22, 45–64.

*Ma, R., Huang, Y., & Shenkar, O. (2011). Social

networks and opportunity recognition: A cultural

comparison between Taiwan and the United

States. Strategic Management Journal, 32,

1183–1205.

Madjar, N. (2008). Emotional and informational

support from different sources and employee

creativity. Journal of Occupational and Organi-

zational Psychology, 81, 83–100.

*Maritz, A. (2010). Networking, entrepreneurship

and productivity in universities. Innovation:

Management, Policy & Practice, 12, 18–25.

*McFadyen, M., & Cannella, A. (2004). Social

capital and knowledge creation: Diminishing

returns of the number and strength of exchange

relationships. Academy of Management Jour-

nal, 47, 735–746.

*McFadyen, M., Semadeni, M., & Cannella, A.

(2009). Value of strong ties to disconnected

others: Examining knowledge creation in biome-

dicine. Organization Science, 20, 552–564.

McPherson, M., Smith-Lovin, L., & Cook, J. M.

(2001). Birds of a feather: Homophily in social

networks. Annual Review of Sociology, 27,

415–444.

*Mehra, A., Kilduff, M., & Brass, D. J. (2001). The

social networks of high and low self-monitors:

Implications for workplace performance. Admin-

istrative Science Quarterly, 46, 121–146.

Milliken, F. J., Bartel, C. A., & Kurtzberg, T. R.

(2003). Diversity and creativity in work groups:

A dynamic perspective on the affective and cog-

nitive processes that link diversity and perfor-

mance. In P. Paulus & B. Nijstad (Eds.), Group

creativity: Innovation through collaboration

(pp. 32–62). New York, NY: Oxford University

Press.

Mizruchi, M. S., & Stearns, L. B. (2001). Getting

deals done: The use of social networks in bank

decision-making. American Sociological Review,

66, 647–671.

*Moran, P. (2005). Structural vs. relational embedded-

ness: Social capital and managerial performance.

Strategic Management Journal, 26, 1129–1151.

*Mors, M. (2010). Innovation in a global consulting

firm: When the problem is too much diversity.


Mumford, M. D., & Gustafson, S. B. (1988). Creativ-

ity syndrome: Integration, application, and inno-

vation. Psychological Bulletin, 103, 27–43.

*Obstfeld, D. (2005). Social networks, the tertius

iungens and orientation involvement in innovation.

Administrative Science Quarterly, 50, 100–130.

Oldham, G. R., & Cummings, A. (1996). Employee

creativity: Personal and contextual factors at

work. Academy of Management Journal, 39,

607–634.

Overton, R. C. (1998). A comparison of fixed-effects

and mixed (random-effects) models for meta-

analysis tests of moderator variable effects. Psy-

chological Methods, 3, 354–379.

Padgett, J. F., & Ansell, C. K. (1993). Robust action

and the rise of the Medici, 1400–1434. American

Journal of Sociology, 98, 1259–1319.

Parks, M. R., Stan, C. M., & Eggert, L. L. (1983).

Romantic involvement and social network invol-

vement. Social Psychology Quarterly, 46,

116–131.

*Paruchuri, S. (2010). Intraorganizational networks,

interorganizational networks, and the impact of

central inventors: A longitudinal study of phar-

maceutical firms. Organization Science, 21,

63–80.

Pavitt, K. (1985). Patent statistics as indicators of

innovative activities: Possibilities and problems.

Scientometrics, 7, 77–99.

Perry-Smith, J. (2002). The social side of creativity:

An examination of a social network perspective

(Unpublished doctoral dissertation). Georgia

Institute of Technology, Atlanta, GA.

*Perry-Smith, J. (2006). Social yet creative: The role of

social relationships in facilitating individual creativ-

ity. Academy of Management Journal, 49, 85–101.

Perry-Smith, J., & Shalley, C. (2003). The social side

of creativity: A static and dynamic social network

Baer et al. 221



perspective. Academy of Management Review,

28, 89–106.

Phelps, C., Heidl, R., & Wadhwa, A. (2012). Knowl-

edge, networks, and knowledge networks: A

review and research agenda. Journal of Manage-

ment, 38, 1115–1166.

Podolny, J. M., & Baron, J. N. (1997). Resources and

relationships: Social networks and mobility in the

workplace. American Sociological Review, 62,

673–693.

Powell, W. W., & Smith-Doerr, L. (1994). Networks

and economic life. In N. Smelser & R. Swedberg

(Eds.), The handbook of economic sociology (pp.

368–380). Princeton, NJ: Princeton University

Press.

Reagans, R., & McEvily, B. (2003). Network struc-

ture and knowledge transfer: The effects of cohe-

sion and range. Administrative Science Quarterly,

48, 240–267.

Reagans, R., & Zuckerman, E. W. (2001). Networks,

diversity, and productivity: The social capital of

corporate R&D teams. Organization Science,

12, 502–517.

Reedijk, J. (1998). Sense and nonsense of science

citation analyses: Comments on the monopoly

position of ISI and citation inaccuracies. Risks

of possible misuse and biased citation and impact

data. New Journal of Chemistry, 22, 767–770.

*Rhee, J. H., & Ji, H. (2011). How to further exploit

social capital: An empirical investigation of three

dimensions of social capital and their contribu-

tions to individual performance. Asian Business

& Management, 10(4), 485–507.

Rodan, S. (2002). Innovation and heterogeneous

knowledge in managerial contact networks. Jour-

nal of Knowledge Management, 6, 152–163.

Rodan, S. (2010). Structural holes and managerial

performance: Identifying the underlying mechan-

isms. Social Networks, 32, 168–179.

*Rodan, S., & Galunic, C. (2004). More than network

structure: How knowledge heterogeneity influ-

ences managerial performance and innovativeness.


Rosenthal, R., & Rubin, D. B. (1986). Meta-analytic

procedures for combining studies with multiple

effect sizes. Psychological Bulletin, 99, 400–406.

*Rost, K. (2011). The strength of strong ties in the cre-

ation of innovation. Research Policy, 40, 588–604.

*Rotolo, D., & Messeni Petruzzelli, A. (2013). When

does centrality matter? Scientific productivity

and the moderating role of research specialization

and cross-community ties. Journal of Organiza-

tional Behavior, 34(5), 648–670.

*Schultz, C., & Schreyogg, J. (2013). The impact of

network ties and resource input on research perfor-

mance: An empirical investigation among surgeons

in academic medical centers. IEEE Transactions on

Engineering Management, 60(3), 457–467.

Shadish, W. R. (1996). Meta-analysis and the explo-

ration of causal mediating processes: A primer of

examples, methods, and issues. Psychological

Methods, 1, 47–65.

Shane, S. (2000). Prior knowledge and the discovery

of entrepreneurial opportunities. Organization

Science, 11, 448–469.

Sorenson, O., Rivkin, J., & Fleming, L. (2006). Com-

plexity, networks and knowledge flow. Research

Policy, 35, 994–1017.

Stephan, P. E., & Levin, S. G. (1991). Inequality in

scientific performance: Adjustment for attribu-

tion and journal impact. Social Studies of Science,

21, 351–368.

Teigland, R., & Wasko, M. M. (2003). Integrating

knowledge through information trading: Examin-

ing the relationship between boundary spanning

communication and individual performance.

Decision Sciences, 34, 261–286.

Tellis, G. J., Prabhu, J. C., & Chandy, R. K. (2009).

Radical innovation across nations: The preemi-

nence of corporate culture. Journal of Marketing,

73, 3–23.

*Tortoriello, M., & Krackhardt, D. (2010). Activat-

ing cross-boundary knowledge: The role of Sim-

melian ties in the generation of innovations.


*Totterdell, P., Holman, D., & Hukin, A. (2008).

Social networkers: Measuring and examining

individual differences in propensity to connect

with others. Social Networks, 30, 283–296.

Trajtenberg, M. (1990). A penny for your quotes:

Patent citations and the value of innovations. The

RAND Journal of Economics, 21, 172–187.




Tsai, W., & Ghoshal, S. (1998). Social capital and

value creation: The role of intrafirm networks.


Uzzi, B. (1997). Social structure and competition in

interfirm networks: The paradox of embeddedness.

Administrative Science Quarterly, 42, 35–67.

Uzzi, B., & Dunlap, S. (2005). How to build your

network. Harvard Business Review, 83(12),

53–60.

Uzzi, B., & Spiro, J. (2005). Collaboration and crea-

tivity: The small world problem. American Jour-

nal of Sociology, 111, 447–504.

Van de Ven, A. H. (1986). Central problems in the

management of innovation. Management Sci-

ence, 32, 590–607.

Viswesvaran, C., & Ones, D. S. (1995). Theory test-

ing: Combining psychometric meta-analysis and

structural equations modeling. Personnel Psy-

chology, 48, 865–885.

Wassermann, S., & Faust, K. (1994). Social network

analysis: Methods and applications. New York,

NY: Cambridge University Press.

Weenig, M. W. (1999). Communication networks in

the diffusion of an innovation in an organization.

Journal of Applied Social Psychology, 29,

1072–1092.

West, M. A. (2002). Sparkling fountains or stagnant

ponds: An integrative model of creativity and

innovation implementation in work groups.

Applied Psychology: An International Review,

51, 355–387.

Whitener, E. M. (1990). Confusion of confidence

intervals and credibility intervals in meta-anal-

ysis. Journal of Applied Psychology, 75,

315–321.

Wuchty, S., Jones, B. F., & Uzzi, B. (2007). The

increasing dominance of teams in production of

knowledge. Science, 316, 1036–1039.

*Zhou, J, Shin, S., Brass, D., Choi, J., & Zhang, Z.

(2009). Social networks, personal values, and

creativity: Evidence for curvilinear and

interaction effects. Journal of Applied Psychol-

ogy, 94, 1544–1552.

*Zou, X., & Ingram, P. (2013). Bonds and bound-

aries: Network structure, organizational bound-

aries, and job performance. Organizational

Behavior and Human Decision Processes, 120,

98–109.

Author biographies

Markus Baer is an Associate Professor of

Organizational Behavior at the Olin Business

School, Washington University in St. Louis.

He earned his PhD from the University of

Illinois at Urbana-Champaign. His research

examines the various activities (i.e., problem

formulation, idea generation, solution imple-

mentation) underlying creativity and innova-

tion in organizations.

Karoline Evans is a PhD candidate in Organi-

zational Behavior at the Olin Business School,

Washington University in St. Louis. Karoline’s

research interests include team innovation and

intragroup dynamics.

Greg R. Oldham is the J. F., Jr. and Jesse Lee

Seinsheimer Chair in Business at the A. B.

Freeman School of Business, Tulane Univer-

sity. He received his PhD from Yale University.

His research focuses on the contextual and per-

sonal conditions that prompt the creativity of

individuals and teams in organizations.

Alyssa Boasso is a staff scientist at Thync Inc.

She earned her PhD from Tulane University. Her

research examines acute and chronic stress and

factors that lead to adaptive outcomes.

Baer et al. 223



DOI: 10.1177/2041386614564105 A meta-analysis and path ...

Documents

Transcript of DOI: 10.1177/2041386614564105 A meta-analysis and path ...