Network Centrality in Strategic Alliances and Firm Innovation:
The Moderating Effect of Multilateral Alliances
Abstract
Strategic alliances have been acknowledged to be important contributors to firms’ innovativeness. While large parts of prior research have focused on alliances as isolated agreements, recent literature suggests that alliances should be regarded as connecting links in interorganizational collaboration networks. Thus, drawing on the social network theory, this study investigates the effect of a firm’s network position on its innovation performance. Furthermore, this research investigates the impact of partnering scale by testing whether a larger number of partners per alliance has a weakening effect on the aforementioned relationship. To test the hypotheses I use a panel dataset of 518 North American firms stretching over eight industrial sectors from 1990 to 2005. The results suggest that firms who uphold a central network position benefit from a significantly higher innovation performance. Meanwhile, no empirical evidence could be found for the moderation effect, implying that a higher number of partners per alliance does not mitigate the focal relationship.
Master Thesis
University of Groningen - Faculty of Economics and Business MSc BA Strategic Innovation Management
January 27th, 2020
Samuel Kaul – s3502945 s.kaul@student.rug.nl
Supervisor: Dr. Philip J. Steinberg Co-assessor: Dr. Pedro de Faria
1. Introduction
In the face of shortened product life cycles and highly competitive markets it becomes increasingly important for firms to innovate in order to sustain a competitive advantage. While this is likely to be more relevant in today’s society than ever before, Tushman & Nadler already acknowledged in 1986 that “organizations can gain competitive advantage by managing effectively for today while simultaneously creating innovation for tomorrow”. This, as well as the sheer amount of literature surrounding this topic, is a testament to the fact that managing sustained innovation is probably one of the most challenging tasks that managers are confronted with. But what tools do practitioners have in order to spur innovation? One of these tools is the use of strategic alliances. Through collaboration with other organizations, firms can access different knowledge bases, making strategic alliances an effective vehicle for organizational learning (Grant & Baden-Fuller, 2004). The newly gained knowledge can subsequently be recombined with a firm’s internal knowledge base. This, in turn, provides opportunities for innovative ideas to arise (Cassiman & Veugeleres, 2006; Lavie & Miller, 2008; Faems et al., 2012).
However, recent studies in the alliance literature have suggested shifting the notion of alliances being regarded as isolated agreements to being understood as building blocks of interorganizational collaboration networks (Gilsing et al., 2008; Dong et al., 2017). When firms form collaborations they contribute to the emergence of networks, which serve as conduits of knowledge, information and other resources. However, not every member has equal access to these resources. From a network perspective, organizations benefit to different degrees from these resource flows based on the network position they find themselves in (Tsai, 2001; Gilsing et al., 2008; Dong & Yang, 2016). This position is determined by a firm’s average path length to the other members in the network (Tsai, 2001) and thus influenced by the number of partners as well as their respective quality (Freeman, 1978). Firms positioned at the periphery of a network therefore have fewer opportunities to access new knowledge or other kinds of resources necessary for the development of innovative ideas (Gnyawali & Madhaven, 2001; Tsai, 2001). This study adds to the former literature on alliance networks by further investigating the effect of network centrality on firms’ innovative output. Based on this I formulate the following research question:
Furthermore, this work incorporates the aspect of partnering scale into the previously discussed construct. In the past, alliance research has predominantly focused on dyadic relationships and seemed content with suggesting that the same arguments, which applied to dyadic alliances, could unhesitatingly be applied to multi-partner alliances (Garcia-Canal et al., 2003). However, studies investigating the differences between these two organizational forms have produced results that suggest otherwise. In fact, increasing the number of partners introduces higher levels of complexity to the management of alliances in regards to coordination (Mirshra et al., 2015; Garcia-Canal et al., 2003), monitoring (Li et al., 2012) and communication (Garcia-Canal et al., 2003). These additional complexities might limit managerial attention and alter a firm’s ability to benefit from the advantages of network centrality. This leads me to introduce the following sub-research question:
RQ2: How do larger numbers of partners per alliance influence the relationship between network centrality and innovation performance?
Subsequently, these research questions are further developed, transformed into hypotheses, and empirically tested. The ground for further testing is a modified version of a panel dataset that was originally created by Schilling (2015). The dataset consists of information on 518 North American firms spanning eight industrial sectors, collected over a period of 16 years from 1990-2005. In order to test the interaction effect, additional alliance data was collected using the SDC database and integrated into the dataset. The empirical results show that compared to peripheral positioning, a central network position does have a positive and significant effect on firms’ innovative performance. Moreover, being in alliances with multiple partners does not attenuate firms’ abilities to benefit from the innovation-fostering effects of network centrality.
2. Theoretical Background & Hypothesis Development
2.1. Strategic Alliances and Innovation
In today’s fast-paced and highly competitive markets, firms are increasingly looking to innovate in order to sustain a competitive advantage and ensure long-term survivability (Schumpeter, 1939; Faems et al., 2005). This paper defines innovation as the invention and subsequent commercialization of new products, services or processes. Accordingly, innovation performance is defined a firm’s “ability to transform innovation capability and effort into market implementation” (Zizlavsky, 2016, p. 818).
While innovation has become instrumental for achieving a competitive advantage, a number of different views have emerged over time on the supposed sources of firms’ competitive advantage. Proponents of the resource-based view (RBV) argue that the main source of a firm’s competitive advantage stems from its set of internal resources (Barney, 1991; Wernerfelt, 1984). The knowledge-based view (KBV), which is widely accepted as an extension of the RBV, shares this sentiment, however with one exception. Scholars argue that ‘knowledge’ is considered to be more strategically significant and thus, more important than any other resource (Grant & Baden-Fuller, 1995; Grant, 1996). In contrast to the previous two, Dyer & Singh (1998) argue that the resources, which are critical to attaining a competitive advantage, may span firm boundaries and may be embedded in interfirm ties. As such, the relational view highlights the importance of interfirm collaboration, arguing that firms who operate in isolation are not able to realize the same supernormal returns, regardless of their resources and capabilities (Dyer & Singh, 1998). Throughout literature, interfirm collaboration has been recognized for its potential to share risk, gain access to new technologies, enable organizational learning, and pool complementary resources (Mowery et al., 1996; Powell et al., 1996).
First, strategic alliances enable access to complementary assets that are necessary for the successful development and commercialization of innovative products, services and technologies (Faems et al., 2005). Each firm has different asset endowments and more often than not lacks specific input. Through collaboration, organizations can obtain these valuable assets and thus, increase the likelihood of successful innovation (Mowery et al 1996; Ahuja, 2000b).
Second, strategic alliances are used as vehicles for knowledge transfer. Accessing knowledge this way is more efficient than if companies were to develop the same knowledge by themselves (Simonin, 1997; Poppo & zenger, 1998; Das & Teng, 2000). This is especially true for tacit knowledge, which is of critical importance for a firm’s innovative capability (Cavusgil et al., 2003). However, this type of knowledge is oftentimes embedded in a firms routines and processes and thus, difficult to transfer due to its nature of being not readily codifiable. Yet, strategic alliances facilitate the transfer of tacit knowledge because of the intensity and frequency of firms’ interaction, leading to higher innovation performance (Ahuja, 2000b; Cavusgil et al., 2003).
Lastly, strategic alliances make it possible for partners to split the costs and risks of R&D projects (Hagedoorn, 2002; Faems et al., 2005). This is particularly common in knowledge intensive industries and rapidly changing markets (Powell et al., 1996; Oxley & Sampson, 2004). Spreading the cost of R&D endeavors frees up resources and enables organizations to pursue other projects simultaneously.
In conclusion, it can be said that innovation rarely happens in isolation, but more so through collaboration with others (Edquist, 2005). In this context, strategic alliances embody one tool that firms can use to improve their innovative performance. In the following chapter, the focus of observation shifts from isolated inter-firm relationships to the implications of firms being embedded in a network of relationships.
2.2. Network Centrality and Innovation Performance
Faust, 1994), is a structural property that has been associated with innovation outcomes in previous literature already (Ibarra, 1993) and is also the focus of this study.
At a network level of analysis, being part of a collective confers distinct advantages. However, recent academic literature suggests that the benefits of an alliance network are not evenly distributed (Gilsing et al., 2008). As a matter of fact, firms who are more centrally positioned within a network of alliance partners have been observed to benefit more from network effects than firms who were less centrally positioned (Powell et al., 1996; Gilsing et al., 2008; Dong and Yang, 2016). Based on the definition of centrality given above, a firm is considered to occupy a central position in its network if it upholds a large number of ties to alliance partner who themselves are well connected (Freeman, 1987).
To understand how network positioning influences innovation performance it is helpful to regard a firm’s network as the locus of resources. These resources, which include but are not limited to complementary knowledge, information, and skills, are held by different actors within the network (Un & Asakawa, 2015), in which ties function as channels of transmission. Occupying a central position offers distinct advantages in regards to an organization’s ability to innovate. By virtue of being surrounded by a large number of information sources and having shorter path lengths to network partners, firms receive information much more quickly (Fleming et al., 2007; Koka & Prescott, 2002) and gain access to new technologies earlier than firms located at the periphery of the network (Gulati, 1999; Tsai, 2001). This advantage is crucial for the development of innovations as firms are able to gather information on arising market opportunities and threats and react accordingly (Dittrich & Duijsters, 2007). But the short average paths lengths have another distinct advantage towards the innovativeness of firms. By increasing the density and overall connectivity of the network, they mitigate some of the negative effects associated with technological diversity and geographical distance, which in turn facilitates the transfer of tacit knowledge (Tsai, 2001).
Furthermore, centrally positioned firms can act as connecting links between unrelated partners. This is referred to as bridging “structural holes” which, according to Ahuja (2000a), are defined as “gaps in information flows between alters linked to the same ego but not linked to each other”. In this case, ‘ego’ refers to the focal organization while ‘alters’ refer to the unrelated network partners. Those alters are oftentimes characterized by having highly different knowledge bases (Ahuja, 2000a). By positioning itself in between, the focal firm enables an indirect resource flow between its partners and can subsequently access their distinct knowledge bases. As a result of this bridging mechanism, centrally positioned firms have an additional opportunity to access diverse knowledge, which in turn promotes organizational learning and increases a firm’s recombinative potential (Ahuja, 2000a; Zaheer & Bell, 2005; Gilsing et al., 2008).
Central firms are also better informed about the network environment due to the amount of readily available information from upholding a large number of ties. This increases the focal organization’s ability to identify partners with non-redundant information more reliable and quicker (Gnyawali & Madhavan, 2001; Gilsing et al., 2008). Apart from a reduction in search costs (Chung et al., 2000), finding compatible partners also increases the chances of successful collaboration and subsequently the development of novel products and services (Gulati, 1999; Gilsing et al).
Based on the aforementioned arguments that suggest a positive relationship between a central network positioning and innovation performance I formulate the first hypothesis as follows:
H1: Network centrality positively influences a firm’s innovation performance
2.3. The Moderating Role of Multilateral Alliances
furthermore moderate the relationship between network centrality and innovation performance.
In alliance literature, partnering scale refers to the number of partners that an alliance consists of (Mishra et al., 2015). The literature distinguishes between two major types of alliances: bilateral and multilateral. Both refer to collaborative partnership agreements. The former, however, involves only two parties, while the latter consists of a minimum of three. In this study I follow the definition of Lavie, Lechner & Singh (2007) who define multilateral alliances as “collective, voluntary organizational associations that interactively engage multiple members in value creating activities, such as collaborative research, development, sourcing, production, or marketing of technologies, products, or services” (p.578). Also, from heron after the terms ‘multi-partner’ and ‘multilateral’ are being used interchangeably. Multi-partner alliances have been observed throughout various sectors but are especially prominent in high-technology industries such as software development, computer hardware, and communications, as well as in the pharmaceutical industry (Doz & Hamel, 1998; Gnomes-Casseres, 2005). Among other benefits, they offer increased access to complementary resources and capabilities (Zeng & Cheng, 2003). However, it has been observed that the actual benefits from partaking in multilateral alliances have been overestimated (Sampson, 2007), while the implications of increased complexity due to the higher number of partners may have been understated.
efforts compared to dyadic alliances (Dyer & Nobeoka, 2000, Garcia-Canal et al., 2003, Li et al., 2012).
Furthermore, multi-partner collaborations bear an increased risk of conflicts (Garcia-Canal et al., 2003). With each additional partner, the number of dyadic relationships increases geometrically. Therefore, bigger alliances tend to encompass more partners whose interests might be conflicting (Park & Russo, 1996). This can lead to the failure of an alliance or, at the very least, impede alliance performance since its success is dependent on each participant’s willingness to collaboratively work towards a common goal (Heidl et al., 2014). At the same time, conflicts may not only arise between a pair of partners but also among subgroups of an alliance (Park & Russo, 1996; Heidl et al., 2014). Similarly to dyadic conflicts, these ‘fault lines’ can render an alliance dysfunctional or even lead to its premature dissolution. Along with other factors, the risk of fault lines increases with the number of alliance participants (Heidl et al., 2014). Therefore, multi partner alliances require an increased effort to achieve harmonization of interests, which is accompanied by higher coordination costs (Park & Russo, 1996; Li et al., 2012).
Moreover, a larger number of alliance participants is associated with greater communication challenges as well as coordination burden (Garcia-Canal et al., 2003; Mirshra et al., 2015). Parkhe (1993) attributes this to a potential increase in cultural diversity and the associated difficulties in regards to resource sharing and communication. However, in a more general sense, corporations are likely to operate on distinct project management processes, meaning that processes for decision-making and communication, and even management styles might differ (Li et al., 2012). As a result, collaborative efforts are less effective unless consistent project management norms are enforced among all partnering firms (Dietrich et al, 2010). In addition, multilateral R&D alliances require much higher coordination efforts due to the inherent need to continuously synchronize project tasks among all involved participants (Ahuja, 2000b). These coordinative burdens increase with partnering scale and divert attention from problem solving (Ahuja, 2000b).
elaborations, I conclude that the management of multilateral alliances requires substantial managerial resources, which however are also required to take advantage of network centrality. To make this clearer it is helpful to look at one of network centrality’s most prominent advantages: the increased access to resources, such as complementary knowledge (Gilsing et al., 2008; Kokoa & Prescott, 2008). Firms, however, do not benefit from this access alone. In order to use external knowledge for innovation purposes, there are various steps that knowledge needs to pass through – many of which are similar to the dimensions of absorptive capacity (Cohen & Levinthal, 1990; Zahra & George, 2002). First, external knowledge sources need to be screened in order to recognize potentially useful knowledge. If such is identified, it needs to be assimilated and only then can be transformed into innovative products or services. Naturally, these steps require substantial resources, among which managerial attention can be listed. However, taking into consideration that multilateral alliances require higher coordination, monitoring and communication efforts, this managerial attention is already limited. In conclusion, decision-makers need to divide their attention between managing the increased access to resources and dealing with the increased managerial complexities (Mirshra et a., 2015). This may result in decreased innovation performance on a firm level since managerial resources are diverted away from innovation promoting activities. Thus, I hypothesize:
H2: A higher average number of partners per alliance negatively influences the relationship between network centrality and innovation performance
The conceptual model, depicted in Figure 1, illustrates the aforementioned hypotheses. The main and moderating relationship are tested at firm level.
H2(-)
H1(+)
Figure 1: Conceptual model
Average Number of Partners per Alliance
Network Centrality Innovation Performance
3. Methodology
The following chapter describes the methodological choices that have been made in this study. It entails a detailed explanation of (1) how the data was collected, (2) how the variables were constructed, and (3) which analytical method has been adopted to test the hypotheses. A longitudinal approach has been chosen to capture both the effect of network centrality on innovation performance, and the moderation effect of multi-partner alliances on the aforementioned relationship.
3.1. Data Collection
In order to research the effect of network centrality on firm innovation performance I used a dataset constructed by Schilling (2015). This longitudinal panel dataset consists of information on 449 North American firms over a period of 16 years, between 1990 and 2005. Due to the fact that there are no uniform patenting norms across regions only North American firms were included in the study. To arrive at the sample size of 449 firms, a total of 13,906 organizations that participated in the global technology collaboration network were sorted using the following criteria: First, a company needed to be held publicly for at least three years to be included into the sample. In addition, it was required that a company applied for at least one subsequently granted patent during the period of observation. Reasons explaining why Schilling (2015) applied these criteria were not elaborated in the study. Later, a control group of 86 North American firms, all of which met the same criteria as the previously mentioned 449 organizations but that did not engage in any strategic alliances, was added. The sample now consisted of 535 firms operating in eight sectors: (1) Transportation equipment, air, and space; (2) construction and materials; (3) food and textiles, (4) pharmaceuticals, biotech, and medical; (5) IT, (6) machines and instruments, (7) chemicals, plastics, and oil; and (8) an aggregate of the service industry not captured in the other categories, e.g. wholesale and retail trade, lodging, and entertainment, etc.
This produced over 13.000 alliances, which were subsequently analyzed with regard to the number of partners each alliance consisted of. With this data I eventually determined the average number of partners per alliance.
The final sample consists of 518 North American firms and incorporates 3887 observations, which amounts to 464 fewer observations than were initially included in Schilling’s (2015) dataset. Most of the missing observations are related to the two-year time lag that Schilling (2015) applied to the dependent variable. To determine the innovative performance of a company in year 2005, patent data from the year 2007 is required. However Schilling’s (2015) observation period only stretches from 1990-2005, making it impossible to determine a firm’s innovation performance in the years 2004 & 2005. Hence, there are no values for the dependent variable in these years, resulting in the negligence of said observations. The rest of the neglected observation can be attributed to the 17 companies that could not be identified. In order to maintain consistent results, only the remaining 3887 observations have been used for analysis.
3.2. Measurements
3.2.1. Dependent Variable
Innovation Performance: Throughout the literature, innovation performance has been
measured in a variety of ways. Its measures include, but are not limited to, R&D expenditure, patents and new product introduction (Hagedoorn and Cloodt, 2003). For this study, I chose a firm’s count of successful patent applications. For one, there is consensus among a large group of economic scholars that patents are among the most appropriate indicators of innovative output and knowledge creation (Freeman and Soete, 1997; Bresman et al., 1999; Ahuja, 2000a). Additionally, Basberg (1987) has demonstrated that there is a strong correlation between a firm’s overall innovation activity and its patent counts.
on patent applications did not follow a normal distribution but instead was heavily skewed to the right. In order to approximate it to a normal distribution, the dependent variable was subsequently log-transformed.
3.2.2. Independent Variable
Network Centrality: To measure the centrality of a firm’s network position I used the variable
‘distance-weighted reach’ from Schilling’s (2015) dataset. This variable describes the sum of reciprocal distances to each organization within the network that is reachable from the focal organization. Like every other network measure in her study, Schilling (2015) calculated this measure based on a 3-year window, meaning that she calculated the mean value of a firm’s distance-weighted reach over the last three years leading up to the year of observation. The following formula is a visual representation of the calculation for one year:
!"#$%&ℎ!" = 1/!!"
!
In the preceding formula, dij is defined as the minimum distance d from focal firm i to partner j. As an example, a firm has a distance-weighted reach of 3 when it is directly connected to three other organizations (who themselves are not connected to each other). However, if a firm is directly connected to two organizations, of which one is directly connected to another organization, it has a distance weighted reach of 2.5. As per this definition, the value of a firm’s distance-weighted reach increases when the firm occupies a more central position in the network, reducing the average path length to other corporations. By contrast, if a company occupies a peripheral network position, its average path length to the other network partners is much higher, thus decreasing the value of its distance-weighted reach. In order to calculate this variable, Schilling (2015) used the SDC database to collect information on the alliance activity of firms from the sample. More specifically, she only included technology collaboration agreements such as R&D alliances, cross-technology transfer agreements, and cross-licensing agreements.
3.2.3. Moderating Variable
Average Number of Partners per Alliance: In order to study the effect of multilateral
agreements. In fact, my research investigates whether the additional complexities that are associated with managing multilateral alliances have a mitigating effect on the relationship between network centrality and innovation performance. As such, it is beneficial to include additional types of alliances, as it increases the share of multi-partner alliances in my sample and allows for a higher number of instances where the potential moderating effect can be observed. Thus, I included joint ventures, marketing alliances, manufacturing alliances and supply alliances. However, to maintain comparability, I followed Schilling’s (2015) approach to calculating network measures based on 3-year windows and thus computed a firm’s average number of partners per alliance by calculating the mean value of the three years leading up to the year of observation.
3.2.4. Control Variables
In order to minimize the effect of confounding variables, I selected the variables below as controls. In addition, some of the variables were log transformed in order to prevent skewness.
Firm Size: According to Cohen & Klepper (1996) and Rogers (2004), the size of a company
can have a distinct influence on its innovation activity. Large organizations often have greater market power and possess more resources than small organizations, which they can use to develop innovations. In addition, larger firms often have dedicated units – e.g. R&D departments – that scan market conditions to identify upcoming trends (Vahlne & Jonsson, 2017). Therefore, I control for differences in firm size by incorporating yearly sales data. Since there is a high variability in firm size, the data will be log-transformed.
Patent stock: A firm’s patent stock serves as an indicator for its technological competence,
technological resources, and its absorptive capacity (Silverman, 1999). Firms with larger patent stocks might therefore be in a favorable position to develop new patents (Du et al., 2014). I control for a firm’s patent stock by counting the accumulated patents of the last three years prior to the year of observation. The figure will be log-transformed. Also, there will be no overlap with the dependent variable, since the dependent variable is always lagged by 2 years.
Investments in R&D: Historically, R&D expenditure has proven to be a significant
Sectorial dummy variables: Also, I used sectorial dummies denoting the industry affiliation
of a firm to control for industry specific effects such as the varying propensity to patent across different industries. “Transportation equipment, air, and space” was coded with “1” if the company was operational in the respective sector.
Year dummy variables: Last but not least, I used year dummies in order to control for
time-specific fixed effects, like events whose impact is limited to a certain period of time. These dummies took a value of “1” for a given year and “0” for every other year.
3.3. Analysis
In accordance with the nature of the dependent variable and the specifications of the dataset, a statistical model has been chosen to most accurately test the hypotheses. In this case we have a longitudinal dataset and a continuous dependent variable. Continuous variables have the property that they can take on an unlimited number of values between any two values. Usually, linear or non-linear models are suitable to evaluate continuous variables. To decide which model to choose, I checked the residual plot to test whether a linear model provided an adequate fit. Had this not been the case, I would have opted for non-linear model, since it gives greater flexibility to fit curves. However, the residual plot revealed that the residuals were centered nicely around the null-axis and showed no asymmetrical patterns. As a result, a linear model was adopted.
4. Results
The following section includes the depiction and analysis of the statistical results from this study. In the first part, the descriptive statistics and correlation coefficients are displayed. In the second part, the results from the regression analysis are presented to determine whether the hypotheses can be confirmed or denied. Lastly, a robustness test was performed whose results are explained and elaborated upon.
4.1. Descriptive Statistics
Table 1 presents the summary statistics and correlation matrix. While the summary statistics provide information on the mean value and standard deviation of the included variables, the correlation matrix reports the Pearson correlation coefficients between the variables. Some of the values would have been difficult to interpret after being log-transformed, and for this reason I included a pre-log-transformation version (Appendix A).
The first part of the descriptive statistics reveals that the average firm successfully applied for 92 patents within 3-year windows during the period of observation. Meanwhile, the corresponding standard deviation (350.344) reveals that there is a high variability among firms in terms of their innovative output. Furthermore, the corporations in this study vary considerably in size. The average number of yearly sales, reported in thousands, amounts to 4255 USD with a standard deviation of roughly 15000 USD. Similar phenomena can be observed with regard to ‘patent stock’ and ‘R&D expenditure’, whose standard deviations are three to four times higher than the mean. Other than that, the average firm holds close to 238 patents and cooperates with roughly 1.3 additional alliance partners in its alliance endeavors. The corresponding standard deviation is 0.846.
Note: Number of Observations: 3887. Significance levels at **** p<0.01, ** p<0.05, * p<0.1 Table 2: Variance Inflation Factor Results
Variable VIF 1/VIF
Firm Size 5.10 0.196
Patent Stock 2.28 0.439
R&D Expenditure 5.25 0.191
Distance-Weighted Reach 6.77 0.148
Average Number of Partners per Alliance 2.34 0.427
Industry Dummies YES YES
Year Dummies YES YES
Mean VIF 3.40
Note: Number of Observation: 3887. Log-transformed
(1) Innovation Performance 2.230 2.002 1.000
(2) Firm Size 5.308 2.808 0.609*** 1.000
(3) Patent Stock 2.955 2.187 0.901*** 0.628*** 1.000
(4) R&D Expenditure 3.397 1.954 0.700*** 0.804*** 0.723*** 1.000
(5) Distance-Weighted Reach 0.821 0.911 0.108*** -0.035** 0.030* 0.110*** 1.000
between ‘R&D expenditure’ and ‘patent stock’ (0.723). As firms invest larger amounts of money into R&D it is reasonable to expect a higher innovative output –Innovative output is measured in patents and after one year, these patents are then included in a firm’s patent stock. In order to rule out any concerns of multicollinearity, I conducted the Variance Inflation Factor (VIF) test. The results in Table 2 show that each value is below the suggested threshold of 10 (Robinson & Schumacker, 2009; Miles, 2014), thus refuting the previous concerns of multicollinearity between variables in the dataset.
4.2. Regression Results & Hypothesis Testing
Table 4 displays the results of the random effects linear regression analysis that was carried out in order to test the hypotheses. Overall, there are four different models whose specifications and corresponding results are explained below:
Model 1, the baseline model, only includes the control variables and shows their impact on innovation performance. ‘Firm size’ displays a positive and significant effect (β = 0.061, p < 0.010), implying that an increase in firm size improves innovation performance. More specifically, an increase in firm size by one unit is accompanied by a 0.061 increase in the dependent variable. The same effect can be observed for patent stock (β = 0.571, p < 0.010) and R&D expenditure (β = 0.151, p < 0.010), with both having a positive and significant effect on innovation performance. Since all control variables remain positive and significant at the same level in the other models, they will not be mentioned further.
Model 2 adds the independent variable ‘distance-weighted reach’ to the analysis in order to test the first hypothesis. The independent variable has coefficient of 0.094 showing a positive relationship with innovation performance at a significance level of p < 0.010. Hence, the results confirm the first hypothesis.
Model 3 introduces the moderator ‘average number of partners per alliance to the model. The results show a negative relationship (β = -0.140) at a significance level of p < 0.100. Although a direct effect on the dependent variable was not hypothesized, it becomes apparent that the moderator variable positively impacts innovation performance. Moreover, the effect of the independent variable slightly increases to β = 0.103 at a significance level of p < 0.0100
Note: Significance levels at *** p<0.01, ** p<0.05, * p<0.1. Standard deviation in parentheses Independent Variable
Distance-Weighted Reach 0.094*** 0.103*** 0.084**
Moderator Variable (0.020) (0.020) (0.037)
Average Number of Partners per Alliance -0.140* -0.166*
Interaction Effect (0.081) (0.031)
Distance-Weighted Reach x
Average Number of Partners per Alliance (0.069) 0.041
Control Variables Firm Size 0.061*** 0.064*** 0.063*** 0.062*** (0.015) (0.015) (0.015) (0.015) Patent Stock 0.571*** 0.570*** 0.572*** 0.572*** (0.013) (0.013) (0.013) (0.013) R&D Expenditure 0.151*** (0.020) 0.142*** (0.020) 0.147*** (0.021) 0.148*** (0.021)
Year Dummies YES YES YES YES
Sectorial Dummies YES YES YES YES
between the independent and the dependent variable remains positive and significant (β = 0.084, p < 0.050). Moreover, the moderator exhibits a negative linear relationship with the dependent variable at a moderately significant level (β = -0.166, p < 0.100). The second hypothesis states that a higher average number of partners per alliance negatively influences the relationship between network centrality and innovation performance. However, I observe non-significant results in regards to the interaction effect, meaning that an increase in the average number of partners per alliance does not significantly mitigate the baseline relationship. Therefore, the second hypothesis is rejected.
4.3. Robustness Check
In order to test the structural validity of my regression model, I chose to experiment with alternative measures for my dependent variable ‘innovation performance’. Due to results from some studies there are different options in regard to the lag structure used for the dependent variable. While Schilling (2015) calculated with a patent lag of two years, Hall et al. (1984) argued that a significant effect of R&D on patenting can already be observed within the first year. Griffin (2002), however, suggests an even longer patent lag. For that reason, I tested the model using patent lags of 1 and 3 years.
Model 5, applying a 1-year patent lag shows unchanged support for the first hypothesis. The relationship between network centrality and innovation performance is still positive (β = 0.068) and remains a the same significance level of p < 0.050. There is still no effect for hypothesis two, suggesting no significant moderation effect of average partners per alliance (β = 0.008; p > 0.100).
Variable Model 4 Model 5 Model 6 Independent Variable
Distance-Weighted Reach 0.084** 0.068** 0.102*
(0.037) (0.033) (0.040) Moderator Variable
Average Number of Partners per Alliance -0.166* -0.132* -0.140 (0.031) (0.076) (0.092) Interaction Effect
Distance-Weighted Reach x 0.041 0.008 0.015
Average Number of Partners per Alliance (0.069) (0.060) (0.074)
Control Variables Firm Size 0.062*** 0.030* 0.102*** (0.015) (0.012) (0.018) Patent Stock 0.572*** 0.697*** 0.443*** (0.013) (0.010) (0.016) R&D Expenditure 0.148*** 0.115*** 0.139*** (0.021) (0.017) (0.024)
Year Dummies YES YES YES
Sectorial Dummies YES YES YES
Constant -0.387** 0.462*** 0.154
(0.152) (0.111) (0.192)
Number of Observations 3,887 3,887 3720
Wald Chi-Square 4759.74*** 10809.04*** 2306.03***
R-Squared 0.819 0.875 0.754
Note: Significance levels at *** p<0.01, ** p<0.05, * p<0.1. Standard deviation in parentheses
5. Discussion
In this section, the empirical results and their interpretation are discussed in more depth. In the first part I explain the theoretical implications of my findings. This is followed up by the managerial implications with practical advice for alliance managers. Finally, I will elaborate on the limitations of this study and provide recommendations for future research.
5.1. Theoretical Implications
In conclusion there is no certain way to know whether the results from this study are sufficient proof for the absence of a moderation effect altogether. There is a possibility that the data was not robust enough or that other positive effects of being in multilateral alliances were unaccounted for.
5.2. Managerial Implications
Apart from the theoretical implications, this study provides a number of practical implications. First, managers should carefully consider their choice of alliance partners. The empirical results indicate that central network positioning is important. This measure is partly determined by the quality of a firm’s respective alliance partners. Partners who maintain a large number of ties to other well-connected firm’s are considered to be of higher quality. Consequently, managers should include these considerations in the decision-making process. Second, although an increase in the average number of partners per alliance did not show a significant influence on the relationship between network centrality and innovation performance, managers should still carefully consider how many firms to form an alliance with. Higher numbers of alliance partners have been shown to add additional complexity and negatively impact alliance effectiveness, when measured by the extent to which alliance goals were fulfilled (Garcia-Canal et al., 2003). Additional coordination burdens, monitoring efforts, and communication challenges limit managerial availability and divert attention away from R&D efforts.
5.3. Limitations and Future Research
Like other studies, this study is subject to a number of limitations that provide the opportunity for future research.
Bresman et al., 1999; Ahuja, 2000a), the sole number merely serves as a rather imperfect expression of a firm’s innovativeness. Looking at the sole number of patents disregards a patent’s distinct innovative value and thus, treats every patent the same. This could be overcome by using weighted measurements of patents, that are based on the number of forward citations they received (Lahiri & Narayanan, 2013). In literature, forward citations have been used to indicate a patent’s value (Hall et al., 2007). By implementing this change to the dependent variable, the validity of the results could be improved. However, doing so would require a substantial time-investment, which, unfortunately, goes beyond the scope of this master thesis and is material for future research.
Furthermore, as presented in the methodology section, there are arguments that support the choice of incorporating a larger number of alliance types than Schilling (2015) included in her research. Most importantly, it increases the share of multi-partner alliances in this study and enables this study to better research the effect of ‘larger numbers of partners per alliance’. Despite this reasoning it would be desirable to perform a robustness check in which the results from incorporating additional alliance types are compared with results that build on the same alliance categorization that Schilling (2015) used in her study. However, due to the limited amount of time and the fact that the process of data collection has already been completed, this step was not feasible.
Studies in recent academic literature have observed a positive impact of network centrality on innovation performance (Powell et al., 1996; Tsai, 2001; Gilsing et al., 2008). Using a panel dataset of 518 North American organizations, this study offers an empirical analysis of this phenomenon and contributes to a better understanding of the underlying effects. In accordance with prior research, the results from this study confirm a significant positive relationship. This can be explained by a number of innovation-fostering effects that central firms benefit from by virtue of upholding a large number of ties to partners who themselves are well connected in the network. These results are relevant four our understanding of alliance network dynamics and indicate that a firm’s positioning holds implications for its ability to innovate.
Ahuja, G. (2000a). Collaboration networks, structural holes, and innovation: A longitudinal study. Administrative Science Quarterly, 45(3), 425-455.
Ahuja, G. (2000b). The duality of collaboration: Inducements and opportunities in the formation of interfirm linkages. Strategic management journal, 21(3), 317-343.
Allen, M. P. (2004). Understanding regression analysis. Springer Science & Business Media. Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of
Management, 17(1), 99-120.
Basberg, B. L. (1987). Patents and the measurement of technological change: a survey of the literature. Research policy, 16(2-4), 131-141.
Bresman, H., Birkenshaw, J., Nobel, R. (1999). Knowledge transfer in international acquisitions. Journal of International Business Studies 30, 439–462.
Cassiman, B., & Veugelers, R. (2006). In search of complementarity in innovation strategy: Internal R&D and external knowledge acquisition. Management Science, 52(1), 68-82.
Cavusgil, S. T., Calantone, R. J., & Zhao, Y. (2003). Tacit knowledge transfer and firm innovation capability. Journal of business & industrial marketing.
Chung, S., Singh, H. & Lee, K., (2000). Complementarity, Status Similarity and Social Capital as Drivers of Alliance Formation. Strategic Management Journal, 21(1), 1-22.
Clark, T. S., & Linzer, D. A. (2015). Should I use fixed or random effects?. Political Science
Research and Methods, 3(2), 399-408.
Cohen, W. M., & Klepper, S. (1996). A reprise of size and R & D. The Economic Journal, 925- 951.
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative science quarterly, 35(1), 128-152. Das, T., & Teng, B. (2000). A resource-based theory of strategic alliances. Journal of
Management, 26(1), 31-61
Das, T. K., & Teng, B. S. (2002). Alliance constellations: A social exchange perspective. Academy of management review, 27(3), 445-456.
Dietrich, P., Eskerod, P., Dalcher, D., & Sandhawalia, B. (2010). The dynamics of collaboration in multipartner projects. Project Management Journal, 41(4), 59-78.
Dittrich, K., & Duysters, G. (2007). Networking as a means to strategic change : The case of open innovation in mobile telephony. Journal of Product Innovation
Journal of Product Innovation Management, 34: 526-542.
Dong Qi, J., Yang, C., (2016). Being central is a double-edged sword: Knowledge network centrality and new product development in U.S. pharmaceutical industry.
Technological Forecasting and Social Change, Volume 113, Pages 379-385.
Doz, Y., Hamel, G., & Prahalad, C. K. (1989). Collaborate with your competitors and win. Harvard business review, 67(1), 133-139.
Doz, Y. L., & Hamel, G. (1998). Alliance advantage: The art of creating value through
partnering. Harvard Business Press
Draulans, J., Deman, A. P., & Volberda, H. W. (2003). Building alliance capability: Management techniques for superior alliance performance. Long range
planning, 36(2), 151-166.
Du, J., Leten, B., Vanhaverbeke, W., (2014). Managing open innovation projects with science based and market-based partners, Research Policy Volume 43, Issue 5, Pages 828-840. Dyer, J. H. and Nobeoka, K. (2000). ‘Creating and managing a high performance knowledge
sharing network: the Toyota case’. Strategic Management Journal, 21, 345–67. Dyer, J., & Singh, H. (1998). The relational view: Cooperative strategy and sources of
interorganizational competitive advantage. The Academy of Management Review, 23(4), 660-679.
Faems, D., Van Looy, B., & Debackere, K. (2005). Interorganizational collaboration and innovation: Toward a portfolio approach. Journal of product innovation management, 22(3), 238-250.
Faems, D., Janssens, M., & Neyens, I. (2012). Alliance portfolios and innovation performance: connecting structural and managerial perspectives. Group &
Organization Management, 37(2), 241-268.
Fleming, L., King III, C., & Juda, A. I. (2007). Small worlds and regional innovation.
Organization Science, 18(6), 938-954.
Freeman, L. C., (1978). Centrality in social networks conceptual clarification. Social Networks, 1(3), pp. 215-239.
Freeman, C., & Soete, L. (2009). Developing science, technology and innovation indicators: What we can learn from the past. Research policy, 38(4), 583-589.
betweenness centrality and density. Research Policy, Volume 37, Issue 10, Pages 1717-1731.
Gnomes-Casseres, B. (2005). Group versus Group: How Alliance Networks Compete. Havard Business Review, 72, 62-74.
Gnyawali, D., and R. Madhavan. (2001). Cooperative networks and competitive dynamics: A structural embeddedness perspective. Academyof Management Review 26 (3): 431–45.
Goerzen, A., & Beamish, P. W. (2005). The effect of alliance network diversity on
multinational enterprise performance. Strategic management journal, 26(4), 333-354.
Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness. American journal of sociology, 91(3), 481-510.
Grant, R. M., & Baden-Fuller, C. (1995, August). A knowledge-based theory of inter-firm collaboration. In Academy of management proceedings (Vol. 1995, No. 1, pp. 17-21). Briarcliff Manor, NY 10510: Academy of Management.
Grant, R. M. (1996). Toward a knowledge-based theory of the firm. Strategic management journal, 17(S2), 109-122.
Grant, R., & Baden-Fuller, C. (2004). A knowledge accessing theory of strategic alliances. Journal of Management Studies, 41(1), 61-84.
Gulati, R. (1998). Alliances and networks. Strategic management journal, 19(4), 293-317.
Gulati, R. (1999). Network location and learning: The influence of network resources and firm capabilities on alliance formation. Strategic Management Journal, 20(5), 397- 420.
Gulati, R., & Higgins, M. (2003). Which ties matter when? the contingent effects of
interorganizational partnerships on ipo success. Strategic Management Journal, 24(2), 127-144.
Hagedoorn, J. (2002). Inter-firm R&D partnerships: an overview of major trends and patterns since 1960. Research policy, 31(4), 477-492.
Hagedoorn, J., Cloodt, M. (2003). Measuring innovative performance: is there an advantage in using multiple indicators? Research Policy 32, 1365-1379.
Hall, B. H., Griliches, Z., & Hausman, J. A. (1984). Patents and R&D: Is there a lag? (No. w1454). National Bureau of Economic Research.
Heidl, R., Steensma, H., & Phelps, C. (2014). Divisive faultlines and the unplanned dissolutions of multipartner alliances. Organization Science, 25(5), 1351-1371. Ibarra, H. (1993). Network centrality, power, and innovation involvement: Determinants of
technical and administrative roles. The Academy of Management Journal, 36(3), 471- 501.
Kale, P., Dyer, J. H., & Singh, H. (2002). Alliance capability, stock market response, and longterm alliance success: the role of the alliance function. Strategic management
journal, 23(8),747-767.
Koka, B. R., & Prescott, J. E. (2002). Strategic alliances as social capital: a multidimensional view. Strategic Management Journal, 23(9), 795-816.
Lahiri, N., & Narayanan, S. (2013). Vertical integration, innovation, and alliance portfolio size: Implications for firm performance. Strategic Management Journal, 34(9), 1042- 1064.
Lavie, D., Lechner, C., & Singh, H. (2007). The performance implications of timing of entry and involvement in multipartner alliances. Academy of Management Journal , 50 (3), 578-604.
Lavie, D., & Miller, S. R. (2008). Alliance portfolio internationalization and firm performance. Organization science, 19(4), 623-646.
Li, D., Eden, L., Hitt, M. A., Ireland, R. D., & Garrett, R. P. (2012). Governance in multilateral R&D alliances. Organization Science, 23(4), 1191-1210.
Miles, J. (2014). Tolerance and variance inflation factor. Wiley StatsRef: Statistics Reference Online.
Mishra, A., Chandrasekaran, A., & MacCormack, A. (2015). Collaboration in multi-partner R&D projects: The impact of partnering scale and scope. Journal of Operations
Management, 33, 1-14.
Moran, P. (2005). Structural vs. relational embeddedness: Social capital and managerial performance. Strategic management journal, 26(12), 1129-1151.
Mowery, D. C., Oxley, J. E., & Silverman, B. S. (1996). Strategic alliances and interfirm knowledge transfer. Strategic management journal, 17(S2), 77-91.
Ocasio, W. (1997). Towards an attention-based view of the firm. Strategic management
journal, 18(S1), 187-206.
Parkhe, A. 1993. Strategic Alliance Structuring: A Game Theoretic and Transaction Cost Examination of Interfirm Cooperation. The Academy of Management Journal , 36(4): 794-829.
Phelps, C. (2010). A longitudinal study of the influence of alliance network structure and composition on firm exploratory innovation. The Academy of Management Journal, 53(4), 890-913.
Phelps, C., Heidl, R. & Wadhwa, A. (2012). Knowledge, Networks, and Knowledge
Networks: A Review and Research Agenda. Journal of Management, 38(4), pp. 1115- 1166.
Poppo, L., & Zenger, T. (1998). Testing alternative theories of the firm: transaction cost, knowledge-based, and measurement explanations for make-or-buy decisions in information services. Strategic management journal, 19(9), 853-877.
Powell, W. W., Koput, K. W., & Smith-Doerr, L. (1996). Interorganizational collaboration and the locus of innovation: Networks of learning in biotechnology. Administrative science quarterly, 116-145.
Robinson, C., & Schumacker, R. E. (2009). Interaction effects: centering, variance inflation factor, and interpretation issues. Multiple Linear Regression Viewpoints, 35(1), 6-11. Rogers, M. (2004). Networks, firm size and innovation. Small Business Economics, 2(2),
pp. 141-153.
Rosenkopf, L., & Schilling, M. A. (2007). Comparing alliance network structure across industries: observations and explanations. Strategic Entrepreneurship Journal, 1(3-4), 191-209.
Salman, N., & Saives, A. L. (2005). Indirect networks: an intangible resource for biotechnology innovation. r&d Management, 35(2), 203-215.
Sampson, R. C. (2007). R&D Alliances and Firm Performance: The Impact of Technological Diversity and Alliance Organization on Innovation. Academy of Management Journal , 50(2), pp. 364-386.
Schilling, M. (2015). Technology shocks, technological collaboration, and innovation outcomes. Organization Science, 26(3), 668-686.
Schumpeter, J. A. (1939). Business cycles (Vol. 1, pp. 161-174). New York: McGraw-Hill. Silverman, B. S. (1999). Technological resources and the direction of corporate
diversification: Toward an integration of the resource-based view and transaction cost economics. Management science, 45(8), 1109-1124.
Academy of Management Journal, Vol. 44, No. 5, 996-1004.
Tushman, M., & Nadler, D. (1986). Organizing for innovation. California management
review, 28(3), 74-92.
Un, C. A., & Asakawa, K. (2015). Types of R&D collaborations and process innovation: The benefit of collaborating upstream in the knowledge chain. Journal of Product
Innovation Management, 32(1), 138-153.
Vahlne, J. E., & Jonsson, A. (2017). Ambidexterity as a dynamic capability in the globalization of the multinational business enterprise (MBE): Case studies of AB Volvo and IKEA. International Business Review, 26(1), 57-70.
Vatcheva, K. P., Lee, M., McCormick, J. B., & Rahbar, M. H. (2016). Multicollinearity in regression analyses conducted in epidemiologic studies. Epidemiology (Sunnyvale, Calif.), 6(2).
Wasserman, S., & Faust, K., (1994) Social network analysis: Methods and applications – Structuralanalysis in the social sciences 8. Cambridge: Cambridge University Press. Wernerfelt, B. (1984). A resource-based view of the firm. Strategic management
journal, 5(2), 171-180.
Zahra, S. A., & George, G. (2002). Absorptive capacity: A review, reconceptualization, and extension. Academy of management review, 27(2), 185-203.
Zaheer, A., Bell, G. G. (2005). Benefiting from network position: firm capabilities, structural holes, and performance. Strategic Management Journal, 26: 809–825.
Zeng, M., & Chen, X. P. (2003). Achieving cooperation in multiparty alliances: A social dilemma approach to partnership management. Academy of Management
Review, 28(4), 587-605.
Appendix A: Summary Statistics & Correlation Coefficients (Non-log-transformed)
Note: Number of Observations: 3887. Significance levels at *** p<0.01, ** p<0.05, * p<0.1
Variable Mean S.D. (1) (2) (3) (4) (5) (6) (1) Innovation Performance 92.173 350.344 1.000 (2) Firm Size 4255.386 15083.212 0.338*** 1.000 (3) Patent Stock 238.367 921.056 0.908*** 0.382*** 1.000 (4) R&D Expenditure 239.812 800.144 0.445*** 0.853*** 0.482*** 1.000 (5) Distance-Weighted Reach 248.573 314.901 0.059*** 0.048** 0.011 0.059*** 1.000
