University of Groningen | Faculty of Economics and Business Department of
Innovation Management & Strategy
MASTER THESIS
MSc Business Administration
Specialization: Strategic & Innovation Management
Does location matter?
The Impact of Geographical Proximity on the Relationship between
Alliance Portfolio Centrality and Firm Innovation Performance
Angelika Steierová
S3477703
Supervisor: Dr. Pedro de Faria
Co-assessor: Dr. Florian Noseleit
II
Abstract
III
Acknowledgements
IV
Table of Contents
1. Introduction ... 5
2. Literature review ... 7
2.1. Importance of innovation and strategic alliances for the firm performance ... 7
2.2. Alliance portfolio and firm position in the networks ... 8
2.3. Centrality of firm in the strategic alliance portfolio ... 10
2.4. Influence of geographical location ... 12
3. Methodology ... 14 3.1. Research setting ... 14 3.2. Data collection ... 15 3.3. Measures ... 16 3.4. Analytical method ... 17 4. Results ... 18 4.1. Descriptive statistics ... 18
4.2. Results of regression analysis ... 21
5
1. Introduction
From the beginning of 1990´s, scholars from the field of strategic management have begun to pay a significant attention to a newly emerging form of collaboration, called strategic alliances (Schilling, 2015). Their importance rose significantly with the strengthening trend of globalization which brought a sharp rise of competition to the global markets (Pisano, 2006). The fast development of technologies leaded to the extended exchange of information and improved communication between firms. As a result of this development, more formation of strategic alliances started to occur (Dess & Picken, 2000). By the year 2000, the overall number of strategic alliances increased into thousands in the various different industries (Wassmer, 2010; Ireland & Hitt, 2002).
In relation to this development, researchers started focusing on finding the real reasons for strategic alliance formations. Gulati (1998) as the first discussed their importance from the social perspective, proposing that a firm may form relationships in the networks, leading to various additional information gains from other partners. Afterwards, scholars from the resource-based view emphasized that strategic alliances may provide the firm with additional resources that are often needed for the firm innovation activities (Das & Teng, 2000). Last but not least, Grant & Baden-Fuller (2004) proposed that the main purpose for formation of strategic alliances lies in the need of firms to obtain a one specific kind of resource, the new knowledge of other alliance partners.
Later on, as the relationships between firms became more complex, scholars shifted their attention from simple dyadic relationships between firms to investigating multiple alliances (Goerzen and Beamish, 2005; Reuer & Ragozzion, 2006). Researchers wanted to understand how to maximize the alliance potential in order to fully exploit the additional resources that could be thereafter used for the improvements of firm processes. This shift of interest called for considering more alliance aspects and characteristics at the same time. Thus, the research started to recognize different characteristics of the whole alliance portfolios and specifically, examined those characteristics that might be positively related to the firm innovation performance (Wuyts &Dutta, 2014).
6
innovation performance? Although the past research already tried to answer this question, we will test the effect of alliance portfolio centrality for the firm innovation performance in our first hypothesis. That will not only help us to verify the previous research assumptions but importantly, it will create a substantial base relationship on which we will be testing the second hypothesis thereafter.
The past papers discovered that being central to the alliance portfolio is beneficial for the firm as it may fasten the firm learning process (Zaheer & Bell, 2005) and raise the firm opportunities for knowledge exchange with other partners (Phelps et al., 2012). Thus, it can be assumed that the alliance portfolio centrality may improve the chances for further, additional gains of resources that are necessary for innovation (Tsai, 2001). Although the literature recognized these direct individual benefits of centrality for innovation performance, a more complex approach that would also take into consideration other conditional factors is still lacking.
One of such factors may be the geographical distance of firms which was investigated from other portfolio characteristics separately. The past research found that geographical proximity of corporate firms alone influences firm innovation performance (Howells, 2001). For example, it was proved that with the decreasing distance, firms tend to exchange specific, local knowledge more easily (Dyer & Nobeoka, 2000). Therefore, we suppose that a centrally positioned firm in the alliance portfolio may apart from centrality even more prosper if it is at the same time geographically close to its portfolio partners. In the context of alliance portfolio, we may thus draw the next question: how does the geographical proximity of firms influences the relationship between the firm network centrality and firm innovation performance? In other words, does the fact that a firm is centrally positioned in the alliance portfolio bring to the firm even more benefits when the firms are in the real world located close to each other?
As the both concepts were in the past studied separately, we consider relevant to fill the research gap by studying their implication for innovation performance in connection to each other. Thus, as the first aim of this study is to test the already investigated relationship between the firm alliance portfolio centrality and firm innovation performance, subsequently, the impact of geographical distance between corporate headquarters for this relationship will be examined.
7
right choice of new alliance formations. Additionally, it can help making the right payoff by utilizing the financial investments into the headquarters appropriately.
In order to test our hypothesis, we conduct a longitudinal study on the whole panel dataset of 535 firms by using a linear panel data regression analysis with the random effect. The operational findings come from the timeframe between the years 1990 – 2005. The dataset is based partially on the SDC Platinum Database and the Delphion database, supported by other internet sources.
The remainder of the thesis is structured as followed. First, we provide the reader with the past literature review on the main findings about the importance of innovation, strategic alliances and alliance portfolios and we conclude the chapter by developing hypotheses and the scheme of the conceptual model. Second, we describe the methodology for the conducted research and subsequently, the results from the descriptive statistics and regression analysis are examined. As the successive step, the discussion, research and managerial implications and the possibilities for the future research directions are provided. The limitations and conclusion are given thereafter.
2. Literature review
2.1.
Importance of innovation and strategic alliances for the
firm performance
Past literature devoted significant attention to the relationship between firm networks and firm innovation performance. In the context of networks, innovation can be defined as: “the development and implementation of new ideas by people who over time engage in transactions with others in an institutional context” (Swan et al., 1999). Nowadays, on the way to be successful in the highly competitive market environment, innovation is considered to be an important firm activity. As the fierce competition causes lowering profits, firms need to differentiate themselves by implying novel, innovative ideas into their products and services on the ongoing base (Eisenhardt & Schoonhove, 1996). However, resources that may help to improve the innovation process, are for each firm limited (Grant, 1999). Therefore, firms tend to cooperate with other entities around them. Such mutual interaction between firms leads to creation of links, gradually forming the whole network of connections (Swan et al., 1999). By cooperating in the networks, firms may acquire new resources and capabilities (Inkpen & Tsang, 2005). In the end, those can be applied on the internal processes, products or services, leading to the higher firm innovation performance (Gopalakrishnan & Damanpour, 1997).
8
technologies, or services” (Gulati, 1998). Similar as joint ventures and other long-term partnerships, alliances are a type of strategic networks that are built on interfirm ties (Gulati, Nohria & Zaheer, 2000).
In the past, research of scholars identified two main reasons for strategic alliance formations. For first, a vast number of researchers discussed their occurrence from the resource-based view perspective (Rothaermel & Hess, 2007; Grant 1999; Teng, 2007; Das & Teng, 2000). It was theorized that with the new establishment of strategic alliances, firm creates new relationships through which other potential partners in the network become available. In other words, the new alliance partners extend the access of the focal firm to the whole network of firms. Such access may help the firm to enhance its influence and strengthen its position in the alliance network. Particularly, the result of gaining access to new partners may be, that it will help the firm to acquire specific skills, technological resources and other resources critical for increasing the likelihood of successful innovation (Eisenhardt & Schoonhove, 1996). Furthermore, the firm may develop new capabilities. Those capabilities obtained from the alliance partners may in the end lead to faster learning and accumulation of new experience. Based on the newly gained experience, the firm learns from others in the alliance. Learning may help to improve the innovation processes and thus, to influence the strategy for the alliances in the future (Heimeriks & Duysters, 2007).
For second, strategic alliances may serve as the medium for obtention of one special kind of resource, the partner´s specific knowledge (Rothaermel & Hess, 2007; Teng, 2007). In the context of benefits arising from strategic alliances, knowledge was found to be an important resource which positively contributes to the firm innovation performance (Lundvall, 2016). Past research highlighted that strategic alliances are sometimes even more effective for the new knowledge acquisition than if the firm decides to develop the new knowledge internally (Cavusgil, Calantone & Zhao, 2003). If such knowledge transfer from external partners happens successfully, firms may become capable of introducing organizational changes leading to the internal process adjustments and new skills development. In the final consequence, those improvements may positively influence firm innovation performance. (Inkpen & Tsang, 2005). For securing the fluency of the knowledge transfer from other alliance partners, the firm needs to take into consideration several factors. Those factors can be for instance the position and the overall number of alliance partners in the network (Gulati, 1999).
2.2.
Alliance portfolio and firm position in the networks
9
Logically, the bigger the alliance portfolio size is, the more resources are available for the firm (Lahiri & Narayanan, 2013). A network of alliances, from which the alliance portfolio is conformed, then creates a reservoir of benefits from the whole industry (Koka & Prescott, 2008). Despite those potential advantages that arise from the formation of alliance portfolios, past research discovered that the overall failure of alliances oscillates between 40 and 70 % (Heimeriks, Duyster & Vanhaverbeke, 2007). According to those adverse numbers, it can be seen that the appropriate configuration of alliance portfolio is highly important and requires considering some other factors. For instance, it was identified that the appropriate number of partners, geographical distance from those partners (Maskell, 2001) and the partner diversity may lead to improved resource exchange (Jiang et al., 2010).
The diversity of alliance portfolio has been broadly discussed in the past literature as it reflects various different resources that may be acquired by firm from the partner´s base (Jiang et al., 2010). Some authors highlight the diversity of partner characteristics and the right partner selection as being even more important for the configuration of alliance portfolio than is the number of connections in the portfolio itself (Hagedoorn & Schakenraad, 1994). For instance, a characteristic of larger firms is that they typically have a bigger technological resource base. Thus, the firm that focuses on partnering with larger firms may prosper from access to their rich, diversified pool of resources (Jiang et al., 2010). On the other hand, by including small and medium firms into alliance portfolio, the firm may rather establish an access to key scientists and their highly specific pieces of knowledge or the tacit knowledge itself (Thorpe et al., 2006). Thus, the higher diversity in the alliance portfolio can be beneficial as it leads to the obtention of various resources emerging from the differences between the partners. That in consequence may positively influence creativity and learning of the firm (Cui & O´Connor, 2012).
Nevertheless, too diverse alliance partners whose knowledge is overly distant from the firm´s own knowledge base can create additional transaction and coordination costs and bring more complexity to the firm (Jiang et al., 2010). Also, the partner´s diversity may hinder the firms to work effectively, when the firms differ in their national origins. Then, for example, the cultural differences may cause inability to cooperate with other partners. As a consequence, the diversity may overall have a negative impact on the firm innovation performance (Goerzen & Baemish, 2005).
10
Finally, scholars identified the group of factors that influences the positioning of the firm in the alliance portfolio and which is called the network structure characteristics (Gulati, 1998). They were found to be those, that may directly influence the firm success when the alliance portfolio is being formed (Wassmer, 2010). In order to be excellent at their management, so that the firm fully exploits the advantages of access to partner knowledge base, the firm needs to be an active alliance portfolio member. Taking that approach will reflect in the need to right set the connections, specifically pursue the appropriate management of network ties (Powell, 1996).
In theory, both kinds, direct and indirect ties (Ahuja, 2000), were found to have a positive impact on the firm innovation performance. Direct ties offer an opportunity for the larger knowledge sharing when the direct partners pursue bigger projects together. The consequences for firm may be that more knowledge and complementary skills will be generated. Similarly, indirect ties that are formed through the third partners, may serve as an information conduit, so that the knowledge available from the overall network will be higher. Again, that may lead to the higher firm innovation performance (Ahuja, 2000). Firm´s number of direct and indirect ties is sometimes in the literature defined as the network density. In case that the technological knowledge of some partners is too distant for the firm and thus difficult to be understood, the higher level of alliance portfolio density, may help the firm to better absorb the partner´s knowledge. On the other hand, it was found that the density may also bring disadvantages, when it hinders the diffusion of novel knowledge. Because the more ties between the firms are, the more equally the knowledge is spread, so that overall, the less additional information is available for the firm in the portfolio (Gilsing et al., 2008).
Thus, from the above described issues and characteristics that may appear over the time of existence of the strategic alliance, we can see that management of such portfolio of alliances is highly complex activity. It requires taking into consideration many aspects from the perspective of the firm management. For instance, managers may lay questions such as whether the firm should set a specific dedicated unit aside, focusing on the management of the portfolio only (Schilke & Georzen, 2010). Or if the firm should rather focus on positioning and establishing new ties in a way that the diversity of partners is beneficial for the firm and it prospers from their social capital and knowledge flows (Koka & Prescott, 2008).
2.3.
Centrality of firm in the strategic alliance portfolio
11
Specifically, the literature examines the direct influence of centrality for the firm knowledge transfer (Dong & Yang, 2016) and organizational innovativeness (Phelps et al., 2012). According to Wasserman & Faust (1994), centrality can be described as “the extent to which the focal actor occupies a strategic position in the network by virtue of being involved in many significant ties”. In other words, it is a measure of how well the firm is connected to other partners in the overall network.
Centrality allows the firm to access and control the critically important information and resources that are likely to be relevant for the firm innovation. (Powell, 1996; Tsai, 2001). Interfirm links create channels for spreading innovative ideas, that flow significantly to the centrally positioned firm and thus, lead to a higher firm innovation performance (Tsai, 2001). The argument standing behind this claim is the relational embeddedness. It proposes that the centrally positioned firm, having more connections than other firms available, can create a higher number of strong ties. Those ties may bring to the firm timeless and diversified information resources that can improve the abilities of firm to learn from other partners in the alliance portfolio (Rowley et al. 2000).
Moreover, more centrally positioned firm will have even more information about the potential partners available, so that prospects for the formation of more successful and stable strategic alliances will be higher (Gulati 1999). Also, past literature found that the centrally positioned firm may help to bridge structural holes in the network. Partners whose knowledge base is highly different from each other, can be used by the firm to position itself between them. When the firm is spanning those structural holes, the benefits will arise from the access to highly diverse information bases of the partners. Then the firm will become the only connective link between otherwise unrelated firms. As a result of bridging the structural holes, centrally positioned firm will have a significantly higher access to more diversified sources of knowledge, extended possibilities for organizational learning and higher chances to obtain competitive advantage (Zaheer & Bell, 2005).
Furthermore, the number of established connections by a centrally positioned firm, combined with the past experience, helps the firm to sustain the speed of learning. In the end, faster learning process may lead to the higher innovation performance and firm faster development (Powell, 1996). The information about the centrally positioned firm is usually better available for others in the network which overall makes the firm perceived as more trustworthy (Phelps et al., 2012). That credibility of firm further improves its reputation and creates a stronger visibility which in consequence attracts to the firm more talented individuals who may bring more innovative ideas (Powell, 1996). Moreover, in comparison to other not centrally positioned partners, centrality raises the firm´s bargaining power. The attractivity of such powerful firm comes to be higher as it is perceived to be a dominant, strong and prestigious organization in the network. Overall, from the perspective of other firms in the network, it will lower the uncertainty about the future possible collaboration (Gilsing, 2008).
12
contrast, the direct network of partners makes the average path length the shortest (Ahuja, 2000). The shorter the path length between the firms is, the more related knowledge is exchanged and the faster the projects between them are finished. In other words, the overall effect of short connections is for the firm positive (Hansen, 2002).
As can be seen, firm centrality in the alliance portfolio may secure the firm with the additional resources that can be used for the innovation processes of the firm thereafter. Moreover, being centrally positioned means a higher possibility for bridging the structural holes, leading to higher inflow of novel information to the firm. At the same time, it can be assumed that centrality helps the firm to get higher chances for successful alliance formation, due the fact that the firm is perceived as powerful and more attractive. Finally, by building on the past experience, centrality may help the firm to achieve improved process of learning that will lead to the faster firm development and overall higher innovation performance. Therefore, we assume, that the network centrality of the firm will be positively related to the firm innovativeness and we propose the formulation for testing that relationship in the first hypothesis as follows:
Hypothesis 1: The centrality of firm in the network of alliance portfolio is positively related to the firm innovation performance.
2.4.
Influence of geographical location
Geographical location is a factor that firms may use for making strategic decisions (Alcácer, 2006). The right selection of the location may bring to the firm various advantages. Firms may decide where to make appropriately their investments and localize themselves. By doing so, they can obtain a better access to partner´s technologies, skills, or gain benefits of the possible knowledge spillovers (Alcácer & Chung, 2007). The favourable choice of location for the firm corporate headquarter may also help to better coordinate the knowledge exchange and lower the information asymmetries between firms (Maskell, 2001).
The knowledge exchange between firms is dependent on firm´s individuals. Their knowledge set is formed and shaped by the personal interaction. The characteristic of interaction between employees is that it is spatially constrained. When the individuals are geographically far from each other, the interaction between them becomes more difficult. For first, the problem arising as a consequence of large distance can be that it raises the individual scanning costs for finding the relevant knowledge. For second, with the high distance between firms, learning process between individuals becomes harder and less fluent. In order to keep the learning process within the firm fluent, individuals need to be co-located as the learning often happens while employees work together at the one place (Howells, 2001).
13
dependent on the geographical proximity of firms. The chances for tacit knowledge to be successfully transferred are higher when firm´s employees stay in the same local environment, communicate repeatedly and preferably in the meetings of low number of participants (Bathelt et al., 2004; Dyer & Nobeoka, 2000). Nevertheless, not only individual employees may take that advantage of learning codified knowledge from other firm´s employees. More importantly, the possibility of tacit knowledge transfer becomes interesting for the whole firm. Closely positioned firms may more easily pursue joint learning through the ongoing large projects leading to establishment of the whole unique base of location specific, tacit knowledge (Howells, 2001).
In the context of firm learning abilities, another research conducted by Boschma (2005) finds that if firms are geographically close, they may step through the learning process easier because their cognitive perception of reality becomes mutually similar in time. Also, learning of firms becomes for the geographically close firms more intense from the social perspective. When the social networks built on mutual relationships are localized geographically together, the interaction of firms is more intensive and the chances for appearance of knowledge spillovers become for the firm more likely (Boschma, 2005). In general, as the social relations between partners strengthen, geographical proximation facilitates the knowledge transfer between them (Bos et al., 2015). Moreover, close firms may pursue more face-to-face contacts, share more easily previously gained experience within the network and help each other when approaching the new partner selection.In those cases, firms that are located close to each other will build up trust more easily, leading to more personal partnerships and more intensive interactive learning (Boschma, 2005).
Overall, the literature so far provides evidence that geographically collocated firms may prosper from their mutual proximity in many ways. In the situation when a firm has lots of direct and indirect partners and thus, it is central to its alliance portfolio, but it is not geographically close to its partners, it may deprive itself of some potential advantages. If the firms are not geographically proximate, the local or tacit knowledge, knowledge spillovers and other learning benefits will remain out of reach. We therefore propose, that the central position in the alliance portfolio is weaker and not so beneficial for the firm if the firm is geographically distanced from other partners. The localized, hardly codifiable tacit knowledge and resources need for the transfer well established connections in the network but more importantly, close, physical presence of the firm individuals. So that the firm that is centrally positioned in the alliance portfolio but geographically far from other firms, does not prosper from its connections with other partners up to the maximal level. The high spatial distance of a centrally positioned firm in the alliance portfolio will also imply, that the firm will have relationships that will have less opportunities to become stronger over time. The cognitive distance of other firms in the alliance portfolio will thus not become closer to the centrally positioned firm and learning from those alliance partners will remain less effective and not fully exploited as for its potential.
14
H
1∩
+
+ H
1+ H
2effect of centrality for the firm innovation performance. Regarding these assumptions, we hypothesize the effect of geographical proximity as follows:
Hypothesis 2: The relationship between the centrality of firm in the network of alliance portfolio and innovation performance is positively moderated by the geographical proximity of firm and its alliance partners.
The conceptual model, as shown in Figure 1, outlines the baseline effect from the firm alliance portfolio centrality on its innovation performance as is predicted by the Hypothesis 1. Subsequently, it presents the moderating effect of the geographical proximity between firms, measured between their headquarters and the focal firm.
3. Methodology
The first goal of this master thesis is to investigate the impact of firm centrality in the alliance portfolio on innovation performance. Subsequently, we examine the moderating effect of geographical proximity for this relationship. To fulfill this objective, in this section, the choice of methodological design, the way the data were collected, the reason for the choice of the analytical model and the included variables, will be outlined.
3.1.
Research setting
In order to capture the effect of centrality in the alliance portfolio on the firm innovation performance, longitudinal, panel dataset approach was chosen. The advantage of this research method lies in the possibility to evaluate the data while using multiple observations of individual firms over time. The time aspect is important to take into consideration in this case as we observe and predict the influence of firm centrality on the number of patent applications with a time lag of two years. Also, this
Figure 1: Conceptual model
Firm Alliance Portfolio Centrality Firm Innovation Performance
15
approach makes it possible to control for omitted variables, that might otherwise influence the results of the analysis. Last but not least, longitudinal approach allows to observe a great sample and thereafter, from the high number of the collected data to infer the accurate predictions and generalizations for each individual firm (Hsiao, 2007). That is also relevant for the purpose of this paper, as in the final discussion, we are going to focus on implications for an individual firm only.
The paper focuses on multiple industries: Transportation equipment, air and space; Construction and materials; Food and textiles; Pharmaceutical, biotechnological and medical; IT; Machines and instruments, chemicals, plastics and oil industry. The data required for testing the first hypothesis were already available from a previous research (Schilling, 2015). For evaluating the second hypothesis, the data were collected from various internet sources.
3.2.
Data collection
The original dataset comes from the SDC Platinum Database and consists of 13 906 North American firms. The data collected for those firms come from the time frame between the years 1990 and 2005. The extracted version from the database was provided by the Department of Strategic & Innovation Management of the University of Groningen. It was reduced to 449 firms based on the next two criterions. First, there were counted only those firms that were publicly held for at least 3 years in the chosen time frame. The second condition was that the firm must have applied in the same time frame for at least one granted patent. Afterwards, another sample of 86 North American firms was added to the dataset. That one included firms with the same criterions as the sample of 449 firms except for the point of not being listed in the SDC database. The overall dataset, in the end, comprised of 535 firms. For each firm in each year, the number of newly obtained patents was collected from the Delphion database. The subsidiary patents were also included and aggregated up to the top parental firm level. The counting of patents was carried out in the year of their application (Schilling, 2015).
16
3.3.
Measures
Dependent variable
Innovation performance. There are two common ways for measuring firm performance. One stream of scholars uses for that purpose financial performance based mostly on measurements as is the return on assets or return on sales (Kostopoulous et al., 2010). The second stream examines the outcome as the firm innovation performance, often counted as the number of patents and patent citations (Hagedoorn, Cloodt, 2003). Because patents are often associated with a higher innovation performance (Griliches, 1990; Acs et al., 2002; Sampson 2007) and the data about the firm number of patents are available here, this research makes use of the second type of measuring, firm innovation performance. The moment the firm applies for a patent can be assigned as the point when the firm creates a new capability. Therefore, it would be logical to measure the innovation performance in the same year when the firm applies for it officially. Nevertheless, past research also found that there appear time lags between the firm investments into R&D and the number of patents that the firm applies for. In other words, the investments into R&D will become a pay off for the firm with a delay. (Sampson, 2007). This delayed effect is necessary to consider, as we are controlling for the effect of R&D investments in the model. Therefore, for the firm i in the year t, we observe the dependent variable, innovation performance, with a two-year time lag (Patents t+2).
Independent variables
Distance-weighted Reach. Following the prior research on the effects of firm network centrality, we overtake the measurement for capturing its impact, counted from the combined effect of the network size and reciprocal distance to every firm that is within the reach of the focal firm in the network. By using the weighted measurement, it is secured that the size and connectivity is taken into consideration when the structure of network is complex, having multiple components (Schilling, 2015). The Distance-Weighted Reach for each firm was counted according to the next formula:
∑ 1 ∕ 𝑑𝑖𝑗 𝑗
where 𝑑𝑖𝑗 expresses the lowest possible distance from a specific firm i to its partner j, where i ≠ j. The
measurement is counted such that it takes into the consideration the whole network of connections, including every other organization in the network that is reachable from the specific firm. So that for example a firm, that has a direct connection to two other, not connected firms, will have the distance-weighted reach accounting for 2. Compared to that, a firm having one direct connection to a partner and through that partner to another partner, will have the total distance-weighted reach 1.5.
Moderator
17
Control variables
In order to avoid possible distortions of the counting, it is necessary to control for some effects that might have an impact on the dependent variable, firm innovation performance.
Firm size. According to Thornhill (2006), the distinct firm size may affect the firm innovation activity differently. For instance, small firms may benefit from the access to more regional, local-specific knowledge sources (Rogers, 2004). On the other hand, big firms may have an increased innovation performance as the matter of fact, they benefit from the large-scale effects (Lahiri, Narayanan, 2013). In order to avoid the potential size effect that might influence firm innovation, it is necessary to control for this effect and include it as the control variable in the model. A logarithmical transformation was applied on the firm size in order to improve its normality. It is counted as the sales of the firm i in each whole year t.
R&D Investment. Previous research found that there is a relationship between R&D investments and innovation in terms of patenting (Sampson, 2007). The firm-level emphasis on R&D activity may differ in time. Therefore, we control for the yearly expenditures of the firm to its R&D and also in this case, the log-normalized form of the control variable, for the firm i in the year t is used (Schilling, 2015). Patent stock. Past application of patents may have an impact on the current firm growth (Coad & Rao, 2008). In other words, it represents the technological competences that the firm gained in the past and that can be used for the innovation activities also in the presence. The firm may on the patent stocks build the new research processes and thus, develop the new patents (Du et al., 2014). It can be measured either by patent citations or the overall number of patents (Bloom & Reenen, 2002). For the purpose of this research paper, the variable representing patent stocks in time t and year i, will be counted as the number of successfully applied patents for each observed firm in three years before the year of observation and ending at the year of observation. Regarding the fact that the firm innovation performance is lagged by two years, there is no overlap between the two variables in the analysis. Sectoral variables of industries. In contrast to patent stock, which is controlling for propensity to patent on the firm-level, the dummy variables are used for controlling the propensity to patent in different industries on the sectoral-level. In total, there are seven sectors, according to which the firms are aggregated in the original SDC Platinum Database and which will be controlled: Transportation equipment, air and space; Construction and materials; Food and textiles; Pharma, biotech, and medical; Machines and Instruments; It and Chemicals, plastics, and oil.
3.4.
Analytical method
18
from 0 to 4390. We are overtaking the analytical method from the paper of Schilling (2015), who proposes the usage of linear data regression model with the random effect. An alternative for conducting the regression analysis might be also the usage of Poisson model. It allows to count the dependent variable that is equal to 0 or higher while working with the integer values. (Coxe et al, 2009). Also, it is a model which is appropriate for events happening randomly and independently on each other (Hausman, 1984). These conditions would be applicable for our dependent variable, the number of patents, thus creating a reason for selection of Poisson model. Nevertheless, for using the Poisson model in the regression analysis, there is also a presumption that the variance and the mean of the dependent variable is equal (Coxe et al, 2009). That is not the case of our dataset. The average not log-transformed number of patents that a firm applied for, with the two-year time lag from the year t, is 90. The variance then reaches 118 593 (Table 1). As the variance of the data is much larger than the mean, it denotes the overdispersion of the data (Rodríguez, 2013). In order to avoid the overdispersion, the data were log-transformed, so that they obtained normal distribution. The normal distribution then allowed us to conduct the linear data regression which was done by using the statistical program STATA and by running the analysis with a random effect.
When deciding about whether to use the random or fixed effect for the regression, it is necessary to look at the nature of variables. The fixed effect does not allow to use the time invariant variables which is in our case the moderating variable, geographical distance between the firm headquarters. We assume that the distance between headquarters remains over the years without change. Therefore, using the panel linear regression with the random effect is more appropriate for this case.
4. Results
In this section, we will first examine the results of descriptive statistics and then we will continue with evaluating the results of the regressions analysis on which we tested the two hypotheses.
4.1.
Descriptive statistics
19
of sales, it even reaches 15 810. Because of the high standard deviations, also the variations will be very high, which means the data are dispersed. In such a case, their values do not have to be concentrated around their mean (Williams, 1984) which signals that the data are not normally distributed. Therefore, we run the test for skewness of each variable. According to Bai & Ng (2001), we can interpret the data as being skewed if the tested variable reaches the critical value of 1.96. We included the results of test for skewness in Table 1. As we may see, except for the variable Distance weighted reach (1.037), all variables are higher than the critical value 1.96. In order to use the panel linear regression for the data analysis with the normal distribution, the data needs to be normalized with the natural logarithm.
The second table serves us for evaluating the multicollinearity between the variables because in this form they come into the regression model. According to Grewal (2004), multicollinearity may cause substantial errors if the correlation coefficients are higher than 0.6. In the Table 2, we can observe that the single industries are in this limit for both positive and negative correlations between -0.6 and 0.6. Where a problem of multicollinearity might appear is between other controlling variables. For instance, the correlation coefficient between R&D (log) and Sales accounts for 0.8019, thus crossing the critical threshold. Also, it can be found between Sales (log), R&D (log), and our dependent variable Patents t+2 (log). Regardless that other correlation coefficients are rather low, we decided to check the
multicollinearity by running the Variance Inflation Factor test (VIF). It is a tool that is broadly used by researchers for measuring the degree of collinearity between variables and which tells us how much of regressor´s variability is explained by other regressors caused by the correlation between them. The general criteria for this test is, that the regression model requires a remedy, if the VIF ≥ 10. In such a case, a usual step would be to leave highly correlated variable out of the model and check for whether the results have changed (Craney, Surles, 2002).
Table 1: Descriptive statistics of not log-transformed variables
Variable Obs. Mean S.D. Skewness
Patents t+2 4 032 89.72545 344.3739 7.606
Distance weighted reach 4 351 233.0455 310.0179 1.037
Mean distance 4 351 2 317.316 950.1568 8.757
Sales 4 351 4 534.867 15 810.63 7.311
R & D 4 351 260.8698 853.4189 6.097
Patent stock 4 351 249.9687 942.0562 7.763
20
Table 2 Descriptive statistics of log-transformed variables
Table 3 Variance Inflation Factor analysis
Variable VIF 1/VIF
Distance weighted reach (log) 1.13 0.937436 Mean distance (log) 1.07 0.888051
Sales (log) 4.94 0.202240
R & D (log) 4,08 0.245285 Patent stock 1.17 0.856148 Transportation equipment, air and space (log) 2.27 0.440935 Construction and materials (log) 2.25 0.444860 Food and textiles (log) 1.35 0.743024 Pharma, biotech, and medical (log) 8.67 0.115342
It (log) 8.65 0.115575
Machines and Instruments (log) 5.56 0.179961 Chemicals, plastics, and oil (log) 2,98 0.335582
Mean VIF 3.68
Variable Obs. Mean S.D. 1 2 3 4 5 6 7 8 9 10 11 12 13
Patents t+2 (log) 4 032 2.215 1.991 1
Distance weighted reach (log) 4 351 0.771 0.903 0.1047 1
Mean distance (log) 4 351 7.716 0.208 0.0246 0.1049 1
Sales (log) 4 351 5.392 2.825 0.6083 -0.0388 -0.0784 1
R & D (log) 4 351 3.474 1.979 0.6998 0.1059 0.0199 0.8019 1 Patent stock 4 351 249.967 942.056 0.5302 -0.0028 0.0575 0.3065 0.3620 1 Transportation equipment, air
and space (log) 4 351 0.037 0.189 0.157 -0.0079 -0.1173 0.2573 0.2217 0.0180 1 Construction and materials
(log) 4 351 0.04 0.196 -0.0065 -0.1336 -0.0659 0.1302 -0.0172 -0.0166 -0.0389 1 Food and textiles (log) 4 351 0.010 0.101 -0.0449 -0.0738 -0.0709 0.0658 -0.0411 -0.0245 -0.0196 -0.0203 1 Pharma, biotech, and medical
(log) 4 351 0.303 0.460 -0.1228 0.0046 0.0469 -0.4494 -0.1480 -0.0623 -0.128 -0.1325 -0.0669 1 It (log) 4 351 0.348 0.476 0.0029 0.2167 0.0535 0.1081 0.09970 0.089 -0.1434 -0.1484 -0.0750 -0.4884 1 Machines and Instruments
(log) 4 351 0.168 0.374 0.0366 -0.1237 0.0755 0.0788 -0.0412 -0.0028 -0.0862 -0.0892 -0.0451 -0.2937 -0.3290 1 Chemicals, plastics, and oil
21
4.2.
Results of regression analysis
Table 4 shows results of the panel data random-effects regression analysis. We constructed 4 models, into which variables are added gradually in order to test the hypotheses 1 (Model 2) and hypothesis 2 (Model 4). The chi-squared test result (chi2 = 0.000) confirmed that all 4 model situations are statistically significant. Also, the R-squared value says that the model fits the data, indicating that it explains approximately 57 % of the variance around the mean of the data, that were used in the model.
Model 1 only includes control variables. The coefficients of control variables were tested for their individual significance and as we may see the 3 variables Sales (β = 0.16), R & D (β = 0.268) and Patent stock (β = 0) are all highly significant (p < 0.01). Moreover, they stay without change throughout all 4 models. The same is also valid for the sectoral control variables of the industries. Independently on adding the other independent variables (Distance weighted reach and Mean distance), their values and significance remain in all 4 models without a higher change.
Model 2 contains all control variables and adds the first independent variable Distance weighted reach in order to test if there exists a positive relationship between the firm centrality in the alliance portfolio and innovation performance (Hypothesis 1). The coefficient of Distance weighted reach is positive and statistically highly significant (β = 0.1; p < 0.01). It positively influences firm innovation performance. Thus, these results are in support of Hypothesis 1. Model 3 introduces into the regression model the variable Mean distance as the independent variable. Although not hypothesized, we look at its direct impact on the dependent variable which shows that the Mean distance has a positive impact on firm innovation performance (β = 0.318) but the result is not significant (p ≥ 0.1). The Distance weighted reach remains positive and highly significant (β = 0.099; p < 0.01).
22
Table 4 Panel data random-effects regression analysis
Variables Model 1 (only controls) Model 2 (Hypothesis 1) Model 3 (Direct effect of reach and distance Model 4
(Moderating effect of mean distance)
Distance weighted reach (log) 0.100*** (0.020) 0.099*** (0.020) 1.266** (0.554)
Mean distance (log) 0.318
(0.228)
0.491**
(0.243)
Distance Weighted reach x Mean distance (log) -0.151**
(0.072) Sales (log) 0.160*** 0.159*** 0.159*** 0.160*** (0.020) (0.020) (0.020) (0.020) R & D (log) 0.268*** 0.267*** 0.265*** 0.264*** (0.023) (0.023) (0.023) (0.023) Patent stock 0.000*** 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) (0.000)
Transportation equipment, air and space (log) 0.921** 0.918** 0.950** 0.954**
(0.382) (0.381) (0.382) (0.382)
Construction and materials (log) 0.313 0.347 0.357 0.362
(0.364) (0.363) (0.363) (0.363)
Food and textiles (log) -0.232 -0.167 -0.136 -0.118
(0.547) (0.545) (0.546) (0.546)
Pharma, biotech, and medical (log) 0.623** 0.600** 0.583** 0.581**
(0.283) (0.282) (0.282) (0.283)
It (log) 0.308 0.266 0.247 0.246
(0.279) (0.278) (0.279) (0.279)
Machines and Instruments (log) 0.692** 0.697** 0.675** 0.673**
(0.294) (0.293) (0.294) (0.294)
Chemicals, plastics, and oil (log) 1.111*** 1.125*** 1.136*** 1.142***
23
5. Discussion
The trend of firms to participate in strategic alliances with other entities for the purpose of obtaining strategic resources has been constantly growing over the past years. Formation of not only dyadic relationships, but complex, multi-partner alliances that create the whole alliance portfolios, has become a common practice of the firm strategy. In order to understand more deeply what influences a successful formation and continuation of strategic alliances, researchers from the field of strategic management started investigating portfolio different characteristics. This master thesis took the focal focus on structural characteristics, specifically, the position of a firm in the strategic alliance portfolio. First, we answered the question regarding the influence of firm centrality in the alliance portfolio on the firm innovation performance. Secondly, we filled the newly identified research gap by finding the link between the geographical proximity of firms, firm alliance portfolio centrality and firm innovation performance. So that we answered the main question: how does the geographical proximity of firms influences the relationship between the firm network centrality and firm innovation performance? To test our two hypotheses, we conducted a research on 535 firms. The research implications, managerial implications and limitations will be hereby discussed in the following sub chapters.
5.1.
Research implications
Overall, our results of the first hypothesis confirmed what past research has proposed. In line with past studies, our regression model found that a centrally positioned firm in the alliance portfolio has a higher innovation output than other firms that are located on the alliance portfolio peripheries. Thus, the aspect of centrality can be regarded as being important because it positively influences firm innovation performance. Understanding that being centrally positioned in the whole alliance portfolio is beneficial for a firm, extends our understanding behind the simple dyadic partnerships and helps us to decide how to possibly set and manage even more complex multi-partner relationships between firms. As the past research found, strategic alliances can generate an opportunity for firms to accumulate new experience. From the newly gained experience, firms develop mutual understanding and organizational routines, which overall improve the learning process (Heimeriks & Duysters, 2007). Moreover, strategic alliances also extend the firm knowledge base by acquiring knowledge of other partners. In additional effect, the internal R&D departments prosper from the partner´s unintended knowledge spillovers (Rothaermel & Hess, 2007; Teng, 2007). Hence, the centrally positioned firm in the alliance portfolio between a larger number of alliances, can deliberately profit from having those advantages at hand. It can thus freely select between a higher number of partners and choose those that seem to be most appropriately helpful for the innovation process.
24
network is the highly discussed concept of partner diversity. The too diverse partners may either have too different knowledge base, which then creates the additional transaction and coordination costs (Jiang et al., 2010) or too internationalized origin which may cause the cultural diversity, leading to difficulties to further cooperate (Goerzen & Baemish, 2005). Despite those two examples which bring evidence of a possible negative impact of centrality for the firm innovation, we believe our results of the model are strong enough to support the claim that centrality in the alliance portfolio is for the firm still beneficial.
A centrally positioned firm has the possibility of selection between the partners compared to the firm that has a lower number of alliances around itself. Perhaps a peripheral position does not bring any risks of diversity but more importantly, due the lower number of connections there even does not exist an opportunity for a noticeable improvement of the learning process, additional information gains or other resource accumulations as the centrally positioned firm has. Last but not least, it can be assumed that a centrally positioned firm in the alliance portfolio may also more prosper from the social capital it has available (Lavie, 2006). When the firm is centrally positioned in the alliance portfolio, it will have more opportunities to imply its internal bargaining power on more partners as it has more direct and indirect ties linked to them. The central firm in the alliance portfolio may also select with whom to build stronger relationships so that even acquisition of resources will be less dependent on a few specific partners. There will arise more opportunities to select between them. If the effort is put into improving the relationships with a higher number of alliance partners, it can be expected that also more knowledge inflows and information gains will become for the centrally positioned firm available. Maintaining the relationships in the alliance portfolio may thus secure the firm with a long-term supply of resources that are obvious for the innovation processes.
In the second part of the master thesis, we proved there exists a significant moderating effect of geographical proximity of firms for the relationship between the firm alliance portfolio centrality and firm innovation performance. We hypothesized that this relationship is more pronounced the closer the firms are located to each other. Although the geographical proximity and its implications for the innovation performance of firms were discussed in the literature before (Howells, 2001; Boschma, 2005), until now, the link to the alliance portfolio centrality has been lacking. Foremost, we found it relevant to investigate this link as we supposed that comprehending of how the geographical distance between firms influences firm portfolio centrality and firm innovation would uncover a possible, positive synergy effect for the firm innovation performance. In addition to that, we considered that bringing the light to the link between these concepts would not only help the firms with further formation of alliances from the network perspective but also with the locational choices for their corporate headquarters.
25
of our regression analysis turned out to be significant even after controlling for possible effects of the Firm size, R&D intensity, Patent stock and propensities to innovate in different industrial sectors, in which the innovation performance was measured.
Overall, this research paper extends the theory about strategic alliances, specifically, by adding new findings to the literature about formation of strategic alliance portfolio from the network perspective (Gulati, 1998). It recognizes that besides the importance of network direct and indirect ties (Ahuja, 2000), ties density (Schilke & Georzen, 2010) and structural holes in the network (Zaheer & Bell, 2005) the physical, real world proximity of firms also influences possible gains from the alliance networks.
For instance, as the process of personal interaction between individuals is spatially constrained (Howells, 2001) and often requires face-to-face contact (Boschma, 2005), it will be easier for the geographically proximate firms to faster interact through their employees. The result of personalized interaction will be that the scanning and searching costs for the relevant knowledge will diminish (Howells, 2001). Thus, if the firm is located in the center of alliance portfolio having the geographically proximate firms around itself, more personal interactions may occur with a high number of partners, so that more relevant knowledge can be to the firm transferred. Overall, knowledge transfer will become for the focal, centrally positioned firm less costly and more fluent. Thus, the coordination costs of the knowledge transfer for the centrally positioned firm will become lower than for the firm that is central to the alliance portfolio but geographically far from others. Especially then, if the partner´s headquarters will be located closer, the centrally positioned firm in the alliance portfolio will also have the local-specific tacit knowledge of partners easier accessible.
Geographical proximity for the centrally positioned firm in the alliance portfolio may bring further benefits in terms of improved learning process. Because innovation is sometimes a complex and difficult activity, it requires learning of specialized steps and procedures (Maskell, 2001). The effectivity and keeping up the speed of learning is dependent on accumulation of experience from other alliance partners (Powell, 1996). A firm, that is centrally positioned in the alliance portfolio has more connections with others available and thus, overall even more experience from them may be gained, leading to a more effective learning process. Nevertheless, if the other partners are geographically located far in distance, learning of their specialized and complex procedures, will be for the firm more uneasy and will require more time. If the firm will be central to its portfolio and moreover, it will be even geographically proximate to its partners, the specialized and complex learning skills of other partners will be more accessible, easily and readily transferable, than if the firm only takes the advantage of being central.
26
effect, the result of the regression model also uncovered the direct effect of geographical proximity for the innovation performance, independently of alliance portfolio centrality. The result proposes that in consequence of the raising geographical distance between firms, also the firm innovation performance will be higher. This partially contradicts our second hypothesis, when it claims that the low distance between geographically close firms will have a negative impact on the firm innovation performance. Regarding the high number of past research papers that provide evidence that geographical proximity is beneficial (for example Li et al. 2013; Bell, 2005), we find those results mostly surprising. As it has been already discussed by other authors, the geographically close partners may exchange tacit knowledge more easily (Bathelt et al., 2004; Dyer & Nobeoka, 2000) and have the learning process faster and more fluent (Boschma, 2005; Howells, 2001).
Thus, in contrast to the findings of the past scholars, this dubious result creates on the one hand a limitation of our research, on the other, forms a possibility for the future researchers to deeply investigate if such effect may truly appear. Our possible logical explanation could be related to the firm intention to isolate itself from other partners in the portfolio. Such isolation might secure a higher inimitability of the firm innovation processes, hinder the potential knowledge leakages or lower the unwanted knowledge spillovers. The firm might then more easily focus on the innovation process only, without a need to invest any additional resources to the knowledge protection mechanisms.
Sustaining this opportunity for the future investigation, we also consider relevant to propose other directions and steps that may be relevant for the future research. For instance, our paper focused predominantly on testing the impact of alliance portfolio centrality and the influence of geographical proximity between the large corporate firms that were with their headquarters mostly situated in the United States. Nevertheless, it might be also interesting to observe whether similar or different results would be achieved in different continents and their technological clusters, such as in Europe or Asia.
27
Overall, this study contributes to the field of strategy and firm innovation literature by empirically testing the past research and more importantly, by providing a support for a yet not explored connection between firm alliance portfolio centrality, innovation performance and geographical proximity of firms. Regarding the confirmation of existence of these findings by our analytical model, we also propose managerial implications in the next sub chapter.
5.2.
Managerial implications
Eventually, we consider necessary to highlight the possible usage of our research findings for managers when they pursue the alliance portfolio decision-making and strategic steps. First, it is worth strengthening that our findings are not only applicable for the beginning of alliance portfolio when it is being created but for the whole time when the firms collaborate. We propose that managers should take specifically the geographical location of their firms as one of the decision-making criteria for the right formation of strategic alliances. Geographical distance from other firms is a strategic aspect that should not be especially omitted when there exists a possibility to locate or relocate the corporate headquarter in time. It may bring benefits in sense of higher innovation performance both earlier in the beginning when the alliance portfolio is formed or later when it already exists for some time.
Despite the flawless and fast communication technologies that firms may use for the cooperation and communication, geographical location remains a remarkable aspect that may elevate the efficiency of the alliance portfolio network on different levels. We propose that these steps are also applicable for subordinate firm entities within the whole corporate hierarchy and not only for headquarters for which the research was conducted. The knowledge, information and other resources are necessary to be fluently transferred without boundaries into every part of the firm. Therefore, beyond our results, this research findings may be for instance also used for the corporate branches.
5.3.
Limitations
Similar to other studies, there can be also found limitations in our research. First of all, there is a significant limitation related to our whole dataset. We collected the data for measuring the geographical proximity from the internet sources that were not in all cases official. The corporate firms, that our dataset consists of, are dated back in the past, between the years 1990 – 2005 which means, there is already a more than 17 years delay at the moment we collected the data. Some of the firms may have changed their corporate headquarter addresses, others were acquired or merged with other firms. Thus, some of the firms do not exist nowadays, which brought a difficulty for the data collection. Therefore, the results and interpretation of the second hypothesis should be further tested, if possible on a different dataset, with firms where the geographical location would be searchable with more accuracy.
28
of time window differ, so that checking also the other years for which the innovation performance might be predicted, would support our final results.
6. Conclusion
29
7. References:
[1]
Acs, J. Z., Anselin, L., Varga, A., (2002). Patents and innovation counts as measures of regional production of new knowledge. Research Policy 31, 1069–1085.[2]
Ahuja, G., (2000). Collaboration Networks, Structural Holes, and Innovation: A Longitudinal Study. Administrative Science Quarterly, Vol. 45, No. 3, pp. 425-455.[3]
Alcácer, J., (2006). Location Choices across the Value Chain: How Activity and Capability Influence Collocation. Management Science, Vol. 52, No. 10, pp. 1457-1471.[4]
Alcácer, J., Chung, W., (2007). Location Strategies and Knowledge Spillovers. Management Science, Vol. 53, No. 5, pp. 760-776.[5]
Bai, J., Ng, S., (2001). Tests for Skewness, Kurtosis, and Normality for Time Series Data. Journal of Business & Economic Statistics, 23:1, 49-60[6]
Baum, C. J. A., Calabrese, T., Silverman, S. B., (2000). Don't Go It Alone: Alliance Network Composition and Startups' Performance in Canadian Biotechnology. Strategic Management Journal Vol. 21, No. 3, pp. 267-294.[7]
Bathelt, H., Malmberg, A., Maskell, P., (2004). Clusters and knowledge: local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography Vol 28, Issue 1, pp. 31 – 56.[8]
Bell, G. G., (2005). Clusters, networks and firm innovativeness, Strategic Management Journal, 26:287-295.
[9]
Bloom, N., Reenen, V. J., (2002). Patents, Real Options and Firm Performance. The Economic Journal, 112: C97-C116.[10]
Bos, B., Faems, D., Noseleit, F., (2015). The Impact of Headquarters-Subsidiary Relations on Alliance Formation Behavior of MNC Subsidiaries, Academy of Management Journal.[11]
Boschma, R., (2005). Proximity and Innovation: A Critical Assessment. Regional Studies, 39:1, 61-74.[12]
Calantone, J. R., Cavusgil, T. S., Zhao, Y., (2002). Learning orientation, firm innovation capability, and firm performance. Industrial Marketing Management 31, 515– 524.[13]
Coad, A., Rao, R., (2008). Innovation and firm growth in high-tech sectors: A quantile regression approach. Research Policy Volume 37, Issue 4, Pages 633-648.[14]
Coxe, S., West, G. S., Aiken, S. L., (2009). The Analysis of Count Data: A Gentle Introduction to Poisson Regression and Its Alternatives. Journal of Personality Assessment, 91:2, 121-136.[15]
Craney, A. T., Surles, G. J., (2002). Model-Dependent Variance Inflation Factor Cutoff Values.Quality Engineering, 14:3, 391-403.
30
[17]
Das, K. T., Teng, B., (2000). A Resource-Based Theory of Strategic Alliances. Journal of Management, Vol. 26, No. 1, 31–61.[18]
Dess, G. G., Picken, C. J., (2000). Changing roles: Leadership in the 21st century. Organizational Dynamics, Volume 28, Issue 3, 18-34.[19]
Dong Qi, J., McCarthy, K. J., Schoenmakers, W. W., (2017). How Central Is Too Central? Organizing Interorganizational Collaboration Networks for Breakthrough Innovation. Journal of Product Innovation Management, 34: 526-542.[20]
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.[21]
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.[22]
Dyer, H. J., Nobeoka, K., (2000). Creating and Managing a High-Performance Knowledge-Sharing Network: The Toyota Case. Strategic Management Journal, Vol. 21, No. 3, pp. 345-367.[23]
Eisenhardt, M. K., Schoonhoven, B. C. (1996). Resource-Based View of Strategic Alliance Formation: Strategic and Social Effects in Entrepreneurial Firms. Organization Science, Vol. 7, No. 2, pp. 136-150.[24]
Faems, D., Janssens, M., Neyens, I., (2012). Alliance Portfolios and Innovation Performance: Connecting Structural and Managerial Perspectives. Group & Organization Management, Volume: 37 issue: 2, page(s): 241-268.[25]
Gilsing, V., Nootemboom, B., Vanhaverbeke, W., Duysters, G., Oord, A., (2008). Network embeddedness and the exploration of novel technologies: Technological distance, betweenness centrality and density. Research Policy, Volume 37, Issue 10, Pages 1717-1731.[26]
Goerzen, A., Beamish, W. P., (2005). The effect of alliance network diversity on multinational enterprise performance. Strategic Management Journal, 26: 333-354.[27]
Gopalakrishnan, S., Damanpour, F., (1997). A Review Economics, Innovation Research in Sociology and Technology Management. Omega, Volume 25, Issue 1, Pages 15-28.[28]
Grant, M. R., (1999). The resource-based theory of competitive advantage: implications for strategy formulation. Knowledge and Strategy, Pages 3–23.[29]
Grant, M. R., Fuller, B., C., (2004). A Knowledge Accessing Theory of Strategic Alliances. Journal of Management Studies, Volume 41, Issue 1.[30]
Grewal, R., Cote, J. A., & Baumgartner, H. (2004). Multicollinearity and measurement error in structural equation models: Implications for theory testing. Marketing Science, 23(4), 519-529.[31]
Griliches, Z., (1990). Patent statistics as economic indicators: A survey. Journal of EconomicLiterature, Vol. 28, pp. 1661-1707.
31
[33]
Gulati, R., (1999). Network Location and Learning: The Influence of Network Resources and Firm Capabilities on Alliance Formation. Strategic Management Journal, Vol. 20, No. 5, pp. 397-420.[34]
Gulati, R., Nohria, N., Zaheer, A., (2000). Strategic networks. Strategic Management Journal, Vol.21, No. 3, pp. 203-215.
