Exploration versus exploitation the influence of network density on firm s strategic choice between two types of innovation

Based on the idea that it is optimum for afirm to focus on one type of innovation when searching for resources from the alliance network, we posit that a low level, as well as a high leve

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ctas20

Technology Analysis & Strategic Management

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/ctas20

Exploration versus exploitation: the influence

of network density on firm’s strategic choice between two types of innovation

Mingyu Tian, Yiwei Su & Zhong Yang

To cite this article: Mingyu Tian, Yiwei Su & Zhong Yang (2024) Exploration versus

exploitation: the influence of network density on firm’s strategic choice between two types of innovation, Technology Analysis & Strategic Management, 36:3, 605-618, DOI: 10.1080/09537325.2022.2046265

To link to this article: https://doi.org/10.1080/09537325.2022.2046265

Published online: 01 Mar 2022

Submit your article to this journal

Article views: 350

View related articles

View Crossmark data

Citing articles: 1 View citing articles

Exploration versus exploitation: the in fluence of network density

Mingyu Tian , Yiwei Su and Zhong Yang

School of Business, Nanjing University, Nanjing, People’s Republic of China

ABSTRACT

Extant research has shown equivocal evidences of brokerage on firm

innovation We try to unpack the mixed findings by studying how

network density among afirm’s alliance partners influences exploratory

innovation Based on the idea that it is optimum for afirm to focus on

one type of innovation when searching for resources from the alliance

network, we posit that a low level, as well as a high level of network

density, offer a suitable environment for exploratory innovation, but a

medium level of network density is more conductive to exploitative

innovation A longitudinal investigation of 277 biopharmaceuticalfirms

indicates a U-shaped relationship between network density and

exploratory innovation This relationship is positively moderated by

absorptive capacity The study integrates related theories and provides

a comprehensive framework of how the connect tightness among a

firm’s partners impacts its exploration and exploitation behaviour

ARTICLE HISTORY

Received 14 October 2021 Revised 15 January 2022 Accepted 17 February 2022

KEYWORDS

Alliance network; network density; exploratory innovation; exploitative innovation

1 Introduction

Interorganizational relationship is one of the crucial factors toflourish business success, due to the access to knowledge and R&D capabilities from partners (Gulati, Nohria, and Zaheer2000) Under today’s increasingly competitive circumstances, it is far from sufficient for a firm to rely only on its internal resources to survive and gain a dominant position in the industry As a consequence, firms in many industries are prone to form alliances to seek external resources for their R&D

manage its alliances and take advantage of its position in the alliance network

Prior research has explored various attributes of the alliance network on the focalfirm’s inno-vation performance In addition to network ties (Liang and Liu2018; Kim, Narayanan, and Narasim-han2020; Yang et al.2021), network configuration (Guan, Zhang, and Yan2015; Kumar and Zaheer

2019; Aggarwal2020) is investigated as well There are two ideal types of network configuration, namely, structural holes (Burt1992), and network closure (Coleman1990) A network full of structural

resources As the opposite side of brokerage, closure typically leads to cooperation and trust The advantages of brokerage and closure motivate scholars to explore which of them could enhance firm innovation Some studies contend that broker firms produce greater innovation output than

as Ahuja (2000) and Fleming, Mingo, and Chen (2007) disagree and indicate a negative effect of

CONTACT Yiwei Su suyw@smail.nju.edu.cn School of Business, Nanjing University, Nanjing, 210093, People ’s Republic of China

2024, VOL 36, NO 3, 605 –618

https://doi.org/10.1080/09537325.2022.2046265

brokerage on innovation Hence, we need to seek for a solution to disentangle the equivocal evi-dences of brokerage on innovation

One possible solution for the perplexing evidences is to categorise innovation into exploratory innovation and exploitative innovation Exploratory innovation refers to the utilisation of new knowl-edge departure from the existing research domain, and exploitative innovation mainly involves

different attributes of exploratory innovation and exploitative innovation result in dissimilar knowl-edge search modes Firms search for novel knowlknowl-edge to achieve exploratory innovation and seek incremental knowledge to exploitative innovation Heterogeneous knowledge search modes lead

to different demands of external resources, and the impact of brokerage on exploratory innovation and exploitative innovation might be different Although existing research sheds light on this area, relevant conclusions are inconsistent For example, Guan and Liu (2016) indicate that brokerage hinders exploitative innovation but favours exploratory innovation, however, Phelps (2010) points

to the positive influence of closure on exploratory innovation

The incoherent conclusions might be due to the fact that the range of relevant theories is limited On the one hand, the positive impact of brokerage on innovation implies that closure, i.e the low level of brokerage, is not preferred On the other hand, closure accelerates innovation indicates that its opposite side, brokerage, is less beneficial These contradictory conclusions push

us to integrate relevant theories and to examine what degree of brokerage is beneficial for inno-vation If we study this issue in-depth and understand the mechanism behind it, we may disen-tangle the puzzle

In this study, we aim to extend the boundaries of relevant theories by exploring the impact of network density among afirm’s alliance partners (network density for short) on exploratory inno-vation The main purpose of this study is to integrate related theories and solve the inconsistent findings of the relationship between brokerage and firm innovation We contend that, exploration

resources in the alliance network for exploratory innovation or exploitative innovation (Stettner

both nonredundant resources and social capital oriented from mutual trust help with novel ideas

In this sense, both a high level of network density and a low level of network density offer a favour-able environment for searching novel ideas, thereby accelerating exploratory innovation For the reason that a moderate level of network density may not promote exploratory innovation as a high level or a low level of network density does, and exploitative innovation is intrinsically preferred

byfirms (Stuart and Podolny1996),firms are more willing to focus on exploitative innovation under a moderate level of network density The foregoing implies a U-shaped effect of network density on exploratory innovation

network density and exploratory innovation A focal firm’s absorptive capacity, i.e the ability to

network Thus, absorptive capacity intensifies the U-shaped relationship In summary, we explore the following questions: How dofirms adjust recourse allocation towards exploration or exploitation

under a certain level of brokerage? In this paper, we theorise and test the consequences of

In the empirical work reported here, we test these predictions on a panel of 277 leading

relationship between network density and exploratory innovation Absorptive capacity positively moderates this relationship The results suggest that when searching for resources from alliances, managers of afirm should consider the connect tightness between its partners to find out if they should extract resources for exploration or exploitation

2 Theory and hypothesis

exploitative innovation, we build on search and recombination theoretical base From search and recombination perspective, organisations not only search for knowledge components, but are also dedicated to the way to combine them to achieve innovation (Arthur2007; Weitzman1998)

exploitative innovation, involving local search that builds on a firm’s existing knowledge base The latter one is known as exploratory innovation, requiring a distant search of new knowledge

As a means of accessing external knowledge, the alliance network could be an effective path of knowledge transfer and integration (Hamel1991) Through direct alliances with its partners, afirm could be access to knowledge and skills of its partners, adding to existing knowledge and increasing the likelihood of innovation In addition, the position that the focal firm occupies in the alliance network is considered to be crucial as well

One stream of research on network configuration stresses the importance of the brokerage pos-ition afirm occupied (Burt1992; Kumar and Zaheer 2021; Kwon et al.2020) A triadic brokerage occurs when the focalfirm is connected with two alters that are not themselves connected Accord-ing to Burt’s (1992) structural holes theory, a high level of brokerage provides the focalfirm with

research from diverse knowledge and skills, promoting successful recombination

Another stream of research on network configuration contends that network closure is beneficial (Coleman1990; Phelps2010; Zhelyazkov2018) Triadic closure refers to the situation that two part-ners of the focalfirm are themselves connected It allows the focal firm to learn about one of its part-ners from another, thereby reducing information asymmetries and increasing their trust (Gulati, Nohria, and Zaheer2000) towards repeated cooperation Trust and reciprocity mitigate opportunism

resources in a social exchange relationship that are valuable and appropriable by partners When partners of the focalfirm are densely connected, the corresponding social capital makes it easier

(Inkpen and Tsang2005) Therefore, afirm could benefit from densely connected partners as well

2.1 The influence of network density on exploratory innovation

Before entering the influences of network density on two types of innovation, we first clarify that exploratory innovation and exploitative innovation are exclusive when referring to the alliance network There is a debate over whether exploratory innovation and exploitative innovation are exclusive or orthogonal Although some research suggests that balancing exploration and exploita-tion is beneficial for innovation (Cao, Gedajlovic, and Zhang 2009; Lucena 2016; March 1991), it

2011) Stettner and Lavie (2014) contend that balancing exploratory and exploitative within a

types of innovation, leading to knowledge misapplication of one type to another Therefore,firms should focus on one type of innovation through the search mode of alliances Under a certain struc-ture of the alliance network, afirm should consider resource endowment to one type of innovation

to achieve superior performance

In addition, Gupta, Smith, and Shalley (2006) indicate that whenfirms need scarce resources for both types of innovation, exploration, and exploitation would be mutually exclusive As March (1991) claims, although these two types of innovation are essential for organisations, they compete for

scarce resources Resources from the alliance network are considered to be scarce, since afirm bears huge costs and risks to acquire external knowledge through the alliance network More resources devoted to exploration imply fewer towards exploitation, and vice versa (Gupta, Smith, and Shalley2006) Thus, limited resources force afirm to focus on one type of innovation when searching from the alliance network

exploitative innovation of the focalfirm based on their exclusive relationship Under a low level of network density, that is, a position with a high degree of brokerage, the focalfirm is easily accessed

to diverse knowledge and information (Burt1992) Afirm in the brokerage position is exposed to

firms (Kumar and Zaheer 2021; Kwon et al.2020) As a result, brokers may creatively synthesise diverse information and obtain a better vision of future innovation possibilities (Burt2005) Scholars contend that brokeragefirms acquire more information (Lee2007), get novel information earlier

Therefore, a low level of network density offers a favourable environment for a firm to search for

exploratory innovation over exploitative innovation under such position The foregoing analysis implies that the focalfirm should pursue exploratory innovation over exploitative innovation at a low level of network density

At a high level of network density, that is, direct partners of the focalfirm are tightly connected to form a dense network, social capital composed of trust and reciprocity increases the willingness of partners to share knowledge (Coleman1990) Density allows the focalfirm to learn about a partner through a common third partner, thereby reducing information asymmetry and increasing their

more visible in the dense network, and opportunism behaviour could jeopardise its current alliances and reduce future alliance opportunities (Gulati1995) Tightly connectedness encourages reciprocity since recipients will repay the benefits to avoid being recognised as opportunism (Coleman1988)

We believe that afirm could benefit its exploratory innovation more from a high level of network density As Phelps (2010) claims, a dense network leads to intense social interaction, joint problem solving, and experimentation, improving the ability of afirm to absorb diverse partner knowledge from partners Additionally, trust and reciprocity from network closure also increase joint

recombi-nation, thus promoting detection and transfer of diverse knowledge from partners (Dyer and Nobeoka2000) Furthermore, the rapidflow of information from tightly connected partners provides more opportunities to share and expand the understanding of technical issues and helps to establish

a shared discourse mode (Powell and Smith-Doerr1994), allowing diverse partners to share with and learn from partners more efficiently (Kogut and Zander 1996) In summary, social capital from the dense network promotes knowledge transfer among partners, broadening the knowledge scope

about its interest of the incremental of existing knowledge and tries to extract from its partners, it might be viewed as opportunistic and thus may negatively influence its current and future alliances Therefore,firms may focus more on exploratory innovation than exploitative innovation from a high level of network density

At a moderate level of network density, that is, neither the ego network is full of structural

and information or trust and reciprocity decrease Under this situation, a firm might not benefit its exploratory innovation from not-so-dispersed knowledge, and a modest level of connection may not generate trust and reciprocity to maintain a strong relationship between partners In addition, exploitative innovation is instinctively preferred byfirms, since it has a higher likelihood

of success in areas of prior experience (Stuart and Podolny1996) Therefore, when the focalfirm

does not have enough resources for exploratory innovation, it may allocate resources from the alliance network to exploitative innovation Therefore, a moderate level of network density

efforts to exploitative innovation

In summary, we contend that both a high level and a low level of network density positively

following:

Hypothesis (H1) Network density among a firm’s alliance partners has a U-shaped relationship with the firm’s subsequent exploratory innovation.

2.2 The moderating role of absorptive capacity

(2014) suggest, ‘Identifying and acquiring innovations from external sources is only half the battle’ Accordingly, to benefit from the resources of the alliance network, a firm relies on its internal learning capacity to absorb and utilise external knowledge

Absorptive capacity wasfirst defined by Cohen and Levinthal (1990) and it is viewed as afirm’s ability to acquire, assimilate, transform and exploit knowledge (Zahra and George2002) According

to Zahra and George (2002), absorptive capacity contains afirm’s ability to acquire and assimilate external knowledge and combine it with existing knowledge to further utilise it for new combi-nations It is beneficial to organisation learning and its R&D (Daghfous2004) Absorptive capacity

future information, thereby enhancing innovation (Schilling1998)

We argue that absorptive capacity intensifies the U-shaped relationship between network density and exploratory innovation Since afirm could be access to external resources through the alliance network, its absorptive capacity helps with acquiring and assimilating knowledge from its partners more efficiently It accelerates the procedure of knowledge acquiring and absorbing, thereby being ready for the recombination process Thus, afirm with sufficient absorptive capacity could benefit more from partners for its innovation In contrast, if afirm lacks absorptive capacity, it would lead

to a lower ability of thefirm to detect and absorb relative new knowledge from its partners, and

it may hinder the effect of external knowledge from alliances on exploratory innovation In line with the above discussion, we hypothesise:

Hypothesis (H2) Absorptive capacity positively moderates the relationship between network density and exploratory innovation.

3 Methods

3.1 Data and sample

The research setting of this study was the global pharmaceutical industry (SICs through 2833–2836)

We chose this industry mainly for two reasons First, as a knowledge-intensive industry with knowl-edge complexity, locus of innovation is ubiquitous in this industry (Powell, Koput, and Smith-Doerr

1996) Second, firms in the pharmaceutical industry tend to patent their innovations (Paruchuri

2010), and it is convenient for us to measure innovation based on patent data

We collected alliance data from the SDC Platinum database Based on this database, we searched forfirms in the global pharmaceutical industry that announced their alliances from 1985 to 2014 There is scarce of alliances announced before 1985, so we began our search for alliances from

type of alliances (Schilling and Phelps2007), and the scope of the true activity of an alliance is often greater than reported (Powell, Koput, and Smith-Doerr1996), we included all types of alliances in our study Following the prior study, we define the network boundary as the following rules: each firm in the pharmaceutical industry must have a partner in the same industry (Østergaard, Timmermans,

Phelps2007) For the reason that alliance termination dates are rarely reported, we took a commonly usedfive-year rolling window approach to represent the duration of alliances (Guan and Liu2016) Therefore, we began in 1989, the last year of the duration of alliances that were announced in 1985 Then the alliance network was created to calculate network-based measures Finally, our data con-tains 277firms and there are 845 observations of alliances

Our patent data was collected from the frequently used Derwent Innovation Index database (DII) DII is one of the most comprehensive databases worldwide containing patent information of more than 100 countries and 40 patent offices, including USPTO, JPO, EPO and SIPO, and so on Our final data contains an unbalanced panel of 277firms and 4132 firm-year observations, which is close to reality and is preferred for avoiding survivorship bias (Baum2006)

3.2 Measures

3.2.1 Dependent variable

each patent of the focalfirm in year t to exploratory innovation or exploitative innovation Follow-ing GilsFollow-ing et al (2008), the criterion is that, if any technological class of a given patent is not in the technological profile of the focal firm in year t, it is considered as exploratory innovation If all

of the technological classes of a given patent are in the technological profile of the focal firm, it is viewed as exploitative innovation Here the technological profile of the focal firm contains all technological classes of patents published within five years before year t Existing research has

inventions (Ahuja2000) Following this criterion, we classified patents of the focal firm in year t into exploratory innovation category and exploitative innovation category Then we applied a commonly used patent citation count measure for exploratory innovation and exploitative innovation

Exploratory innovation We measured the focalfirm’s exploratory innovation performance of year

t as the forward citations of a granted exploratory patent p1 applied in year t as:

p 1

i =1

(Citationi+ 1) where i represents the ith exploratory patent of the focalfirm in year t

3.2.2 Independent variable

Network density We used the ratio of the number of existing ties among alliance partners of the focal firm to the potential maximum number of ties among them in year t-1 to measure network density of the focalfirm (Phelps2010):

TNit−1

nit −1(nit −1− 1)/2 where TNit−1is the number of connections among alliance partners of the focalfirm i in year t-1 nit−1

is the number of alliance partners of the focalfirm, and nit−1(nit−1− 1)/2 is the potential maximum number of connections It is valued at zero if there is no direct tie or only one direct tie of the focal firm.Figure 1presents examples of different levels of network density

3.2.3 Moderator

Technological base (log) We used technological base to represent the absorptive capacity of the focal

Jaffe, and Trajtenberg2005) to calculate technological base of year t-1:

t−1

a =1

(1− 15%)t − aPN

i a

where PNi ais the number of patents of the focalfirm i at yeara (a , t) Then following Funk (2014),

we logged it, with 0 replaced by a small number of 0.0001 before the transformation

3.2.4 Control variables

Direct ties We controlled for direct ties of the focalfirm as the number of alliance partners of the focal firm in the network of year t-1, for the reason that the knowledge exchange through direct ties might

be correlated with benefits from brokerage position (Ahuja2000)

Indirect ties To capture informal knowledge path through network (Ahuja2000), we measured indirect ties as the number of indirectly connected partners of the focalfirm in the network of year t-1

Technological diversity Based on the Blau Index of diversity, we controlled for technology diversity

to account for the technological scope of the focalfirm (Kumar and Zaheer2019) as:

j

pp2j

where j is the patent class and ppjrepresents the proportion of patents that is in class j It is assigned

to 0 if the focalfirm did not patent until year t-1 Higher Blau Index indicates a higher level of tech-nological heterogeneity (Kumar and Zaheer2019)

Alliance experience We controlled for alliance experience as the number of alliances formed by the focal firm in the seven years before year t-1 to represent the collaborative capability of the focalfirm (Phelps2010)

Repeated ties Prior collaboration between twofirms could enhance their trust (Gulati1995) Fol-lowing Phelps (2010), we calculated repeat ties as the number of repeated alliances among its part-ners divided by the number of partpart-ners in year t-1

Industry difference We measured industry difference as the proportion of alliance partners that are not in the same four-digit SIC codes as the focalfirm at year t-1 to capture the effect of market resource overlap (Kumar and Zaheer2019) High industry difference implies a low level of industry similarity

International alliance We controlled for heterogonous of alliance partners (Phelps2010) as the proportion of alliance partners that are not in the same country as the focalfirm in year t-1

Figure 1 Di fferent levels of network density.

Technological distance (cosine) To measure technological similarity between the focalfirm and its partners (Kumar and Zaheer2019), we constructed a vector I containing the cumulative distribution

of patents of afirm in different patent classes until year t-1 Thus, each element of the K-dimensional vector measured the fraction of patents in class k, k = 1, 2,… , K Then we calculated the cosine angular distance (Jaffe1986) between the focalfirm i and one of its partners j using:

cosineijt−1= Iit−1Ijt −1/ (Iit−1Iit−1)(Ijt−1Ijt−1)



The range of this measure is from 0 to 1, in which 1 represents complete similarity Then we averaged the distance of the focalfirm and all alliance partners to get the aggregate measure

Technological diversity (alliance partners) Similar to technological diversity of the focalfirm, we controlled for technological diversity of alliance partners in year t-1 as the average technological diversity of partners based on the Blau Index to capture the technological scope of partners of the focalfirm (Kumar and Zaheer2019)

Average network distance We controlled for average network distance as the average distance between connected nodes in the network in year t-1 to capture its effect on information diffusion and novel recombination (Schilling and Phelps2007)

control for the region of afirm ‘U.S.-Canada’ was coded 1 if a firm was in the U.S or Canada and

‘Europe’ was coded 1 if a firm was in Europe Asia was the omitted category

3.3 Model estimation

hypothesis This is because our dependent variable, Exploratory Innovation, is nonnegative and takes only integer values Some values of our dependent variables are zero, therefore we cannot

our dependent variables results in E[ln (D + y|x)], and it cannot be transformed back to our interest E[y|x] However, unconditional fixed-effects Poisson quasi-maximum likelihood estimator takes the form yit=aiex′itb+ 1it, and our interest could be captured as

E[yit|xit,ai]=aiex′it b= e(ln a i +x ′

it b ), t= 1, 2, , T, i = 1, 2, , N, here yit is the exploratory innovation offirm i in time t, xit includes independent variable, control variables, and time effects,b is the coefficient estimator, ai is afirm-specific unobserved time-con-stant effect, and 1it is the random error term

4 Results

Table 1presents the descriptive statistics and correlations between variables We checked for multi-collinearity through Variance Inflation Factor (VIF) and the highest VIF of the variables is 4.14, far below 10, indicating that there is no issue of multicollinearity

Table 2presents the unconditionalfixed-effects Poisson regression results Model 1 is the baseline model only including control variables Model 2 includes the independent variable, Network Density The Moderating variable of Technological Base (Log) is added in Model 3 Model 4 is our full model with all variables and interactions included

Hypothesis (H1) predicts that network density among the focalfirm’s alliance partners has a U-shaped effect on exploratory innovation The coefficient of Network Density is negative and signifi-cant (p, 0.01) and that of Network Density Squared is positive and significant (p , 0.01) in Model 2,

3, and 4, supporting Hypothesis (H1) The focalfirm achieves better exploratory innovation under both the low level and the high level of network density, but the medium level of network density is not beneficial for exploratory innovation The results imply that the focal firm should

focus on exploratory innovation under the low level as well as the high level of network density, while pursuing exploitative innovation under the medium level of network density

Hypothesis (H2) proposes a positive moderating effect of absorptive capacity on the relationship between Network Density and Exploratory Innovation In line with Hypothesis (H2), the interaction between Network Density and Technological Base (Log) is positive (p, 0.01), and the interaction

4 Figure 2 illustrates this moderation effect, showing that the U-shape is flattened when the value of Technological Base (Log) is higher These results provide supports for Hypothesis 2 (H2)

5 Discussion

This research is motivated by the seemingly perplexing theory and results regarding the influence of brokerage on innovation Specifically, this study pays attention to the stream of research of alliance network and innovation, which neglects the benefits of specialising resources from the alliance network on exploration versus exploitation and the correspondingly optimal choice between them at different levels of brokerage

Wefill in the gap by examining the influence of network density on exploratory innovation The theoretical framework suggests that exploratory innovation and exploitative innovation are incom-patible when searching for resources from the alliance network and afirm needs to specialise in one type of innovation Since both spanning structural holes among direct partners and tightly con-nected partners are beneficial for exploratory innovation, we posit that a low level, as well as a

Table 1 Descriptive statistics and correlations.

Variable Mean SD Min Max 1 2 3 4 5 1.Exploratory Innovation 3.73 5.23 0.00 58.00 1

2.Network Density 0.03 0.15 0.00 1.00 −0.04 1

3.Technological Base (log) 3.68 2.45 −9.21 8.01 0.36 −0.01 1

4.Direct Ties 1.25 2.65 0.00 30.00 0.24 0.14 0.19 1

5.Indirect Ties 54.80 82.82 0.00 219.00 0.01 0.17 −0.05 0.53 1 6.Technological Diversity 0.85 0.14 0.00 1.00 0.14 0.01 0.05 0.01 0.01 7.Alliance Experience 0.68 1.68 0.00 28.00 0.18 −0.03 0.22 0.26 0.05 8.Repeated Ties 0.10 0.27 0.00 1.00 0.06 0.08 0.02 0.38 0.43 9.Industry Di fference 0.16 0.33 0.00 1.00 0.02 0.14 −0.04 0.31 0.42 10.International Alliance 0.26 0.41 0.00 1.00 0.10 0.13 0.09 0.34 0.46 11.Technological Distance (cosine) 0.20 0.34 0.00 1.00 0.09 0.17 0.05 0.48 0.65 12.Technological Diversity (alliance

partners)

0.29 1.68 0.00 1.00 0.09 0.17 0.05 0.55 0.70 13.Average Network Distance 3.21 1.63 0.76 6.17 0.01 0.14 −0.15 0.39 0.74 14.U.S.-Canada 0.61 0.49 0.00 1.00 0.03 0.09 −0.22 0.24 0.36 15.Europe 0.22 0.41 0.00 1.00 −0.21 −0.02 −0.21 0.03 0.11 Variable 6 7 8 9 10 11 12 13 14 15 1.Exploratory Innovation

2.Network Density

3.Technological Base (log)

4.Direct Ties

5.Indirect Ties

6.Technological Diversity 1

7.Alliance Experience 0.02 1

8.Repeated Ties −0.01 0.11 1

9.Industry Di fference 0.07 0.12 0.25 1

10.International Alliance 0.02 0.08 0.28 0.33 1

11.Technological Distance (cosine) −0.00 0.06 0.37 0.31 0.48 1

12.Technological Diversity (alliance

partners)

0.00 0.05 0.42 0.45 0.56 0.80 1 13.Average Network Distance 0.01 0.094 0.32 0.28 0.60 0.45 0.46 1

14.U.S.-Canada 0.01 0.038 0.19 0.21 0.27 0.33 0.36 0.55 1 15.Europe −0.05 0.031 0.05 0.12 −0.20 0.00 0.06 0.07 0.03 1

1

All correlations whose absolute values are greater than |0.03| are signi ficant at p < 0.05.